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} -Wabi-tag -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 -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-sections=@var{s1,s2,...} @gol
305 -fsanitize-undefined-trap-on-error @gol
306 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
307 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
308 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
309 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
310 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
311 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
312 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
313 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
314 -fchkp-use-wrappers @gol
315 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
316 -fdisable-ipa-@var{pass_name} @gol
317 -fdisable-rtl-@var{pass_name} @gol
318 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
319 -fdisable-tree-@var{pass_name} @gol
320 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
321 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
322 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
323 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
324 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
326 -fdump-statistics @gol
328 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
329 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
330 -fdump-tree-cfg -fdump-tree-alias @gol
332 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
333 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
334 -fdump-tree-gimple@r{[}-raw@r{]} @gol
335 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
337 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
338 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
339 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
340 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
341 -fdump-tree-nrv -fdump-tree-vect @gol
342 -fdump-tree-sink @gol
343 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
344 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-vtable-verify @gol
347 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
348 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
349 -fdump-final-insns=@var{file} @gol
350 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
351 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
352 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
353 -fenable-@var{kind}-@var{pass} @gol
354 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
355 -fdebug-types-section -fmem-report-wpa @gol
356 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
358 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
359 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
360 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
361 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
362 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
363 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
364 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
365 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
366 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
367 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
368 -fdebug-prefix-map=@var{old}=@var{new} @gol
369 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
370 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
371 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
372 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
373 -print-prog-name=@var{program} -print-search-dirs -Q @gol
374 -print-sysroot -print-sysroot-headers-suffix @gol
375 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
377 @item Optimization Options
378 @xref{Optimize Options,,Options that Control Optimization}.
379 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
380 -falign-jumps[=@var{n}] @gol
381 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
382 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
383 -fauto-inc-dec -fbranch-probabilities @gol
384 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
385 -fbtr-bb-exclusive -fcaller-saves @gol
386 -fcheck-data-deps -fcombine-stack-adjustments -fconserve-stack @gol
387 -fcompare-elim -fcprop-registers -fcrossjumping @gol
388 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
389 -fcx-limited-range @gol
390 -fdata-sections -fdce -fdelayed-branch @gol
391 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
392 -fdevirtualize-at-ltrans -fdse @gol
393 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
394 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
395 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
396 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
397 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
398 -fif-conversion2 -findirect-inlining @gol
399 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
400 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
401 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
402 -fira-algorithm=@var{algorithm} @gol
403 -fira-region=@var{region} -fira-hoist-pressure @gol
404 -fira-loop-pressure -fno-ira-share-save-slots @gol
405 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
406 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
407 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
408 -flive-range-shrinkage @gol
409 -floop-block -floop-interchange -floop-strip-mine @gol
410 -floop-unroll-and-jam -floop-nest-optimize @gol
411 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
412 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
413 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
414 -fmove-loop-invariants -fno-branch-count-reg @gol
415 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
416 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
417 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
418 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
419 -fomit-frame-pointer -foptimize-sibling-calls @gol
420 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
421 -fprefetch-loop-arrays -fprofile-report @gol
422 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
423 -fprofile-generate=@var{path} @gol
424 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
425 -fprofile-reorder-functions @gol
426 -freciprocal-math -free -frename-registers -freorder-blocks @gol
427 -freorder-blocks-and-partition -freorder-functions @gol
428 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
429 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
430 -fsched-spec-load -fsched-spec-load-dangerous @gol
431 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
432 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
433 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
434 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
435 -fschedule-fusion @gol
436 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
437 -fselective-scheduling -fselective-scheduling2 @gol
438 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
439 -fsemantic-interposition @gol
440 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
441 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt @gol
442 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
443 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
444 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
445 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
446 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
447 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
448 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
449 -ftree-loop-if-convert-stores -ftree-loop-im @gol
450 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
451 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
452 -ftree-loop-vectorize @gol
453 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
454 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
455 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
456 -ftree-vectorize -ftree-vrp @gol
457 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
458 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
459 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
460 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
461 --param @var{name}=@var{value}
462 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
464 @item Preprocessor Options
465 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
466 @gccoptlist{-A@var{question}=@var{answer} @gol
467 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
468 -C -dD -dI -dM -dN @gol
469 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
470 -idirafter @var{dir} @gol
471 -include @var{file} -imacros @var{file} @gol
472 -iprefix @var{file} -iwithprefix @var{dir} @gol
473 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
474 -imultilib @var{dir} -isysroot @var{dir} @gol
475 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
476 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
477 -remap -trigraphs -undef -U@var{macro} @gol
478 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
480 @item Assembler Option
481 @xref{Assembler Options,,Passing Options to the Assembler}.
482 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
485 @xref{Link Options,,Options for Linking}.
486 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
487 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
488 -s -static -static-libgcc -static-libstdc++ @gol
489 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
490 -static-libmpx -static-libmpxwrappers @gol
491 -shared -shared-libgcc -symbolic @gol
492 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
493 -u @var{symbol} -z @var{keyword}}
495 @item Directory Options
496 @xref{Directory Options,,Options for Directory Search}.
497 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
498 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
499 --sysroot=@var{dir} --no-sysroot-suffix}
501 @item Machine Dependent Options
502 @xref{Submodel Options,,Hardware Models and Configurations}.
503 @c This list is ordered alphanumerically by subsection name.
504 @c Try and put the significant identifier (CPU or system) first,
505 @c so users have a clue at guessing where the ones they want will be.
507 @emph{AArch64 Options}
508 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
509 -mgeneral-regs-only @gol
510 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
512 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
513 -mtls-dialect=desc -mtls-dialect=traditional @gol
514 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
515 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
517 @emph{Adapteva Epiphany Options}
518 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
519 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
520 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
521 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
522 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
523 -msplit-vecmove-early -m1reg-@var{reg}}
526 @gccoptlist{-mbarrel-shifter @gol
527 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
528 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
529 -mea -mno-mpy -mmul32x16 -mmul64 @gol
530 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
531 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
532 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
533 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
534 -mucb-mcount -mvolatile-cache @gol
535 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
536 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
537 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
538 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
539 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
540 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
543 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
544 -mabi=@var{name} @gol
545 -mapcs-stack-check -mno-apcs-stack-check @gol
546 -mapcs-float -mno-apcs-float @gol
547 -mapcs-reentrant -mno-apcs-reentrant @gol
548 -msched-prolog -mno-sched-prolog @gol
549 -mlittle-endian -mbig-endian @gol
550 -mfloat-abi=@var{name} @gol
551 -mfp16-format=@var{name}
552 -mthumb-interwork -mno-thumb-interwork @gol
553 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
554 -mtune=@var{name} -mprint-tune-info @gol
555 -mstructure-size-boundary=@var{n} @gol
556 -mabort-on-noreturn @gol
557 -mlong-calls -mno-long-calls @gol
558 -msingle-pic-base -mno-single-pic-base @gol
559 -mpic-register=@var{reg} @gol
560 -mnop-fun-dllimport @gol
561 -mpoke-function-name @gol
563 -mtpcs-frame -mtpcs-leaf-frame @gol
564 -mcaller-super-interworking -mcallee-super-interworking @gol
565 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
566 -mword-relocations @gol
567 -mfix-cortex-m3-ldrd @gol
568 -munaligned-access @gol
569 -mneon-for-64bits @gol
570 -mslow-flash-data @gol
571 -masm-syntax-unified @gol
575 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
576 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
577 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
579 @emph{Blackfin Options}
580 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
581 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
582 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
583 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
584 -mno-id-shared-library -mshared-library-id=@var{n} @gol
585 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
586 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
587 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
591 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
592 -msim -msdata=@var{sdata-type}}
595 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
596 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
597 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
598 -mstack-align -mdata-align -mconst-align @gol
599 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
600 -melf -maout -melinux -mlinux -sim -sim2 @gol
601 -mmul-bug-workaround -mno-mul-bug-workaround}
604 @gccoptlist{-mmac @gol
605 -mcr16cplus -mcr16c @gol
606 -msim -mint32 -mbit-ops
607 -mdata-model=@var{model}}
609 @emph{Darwin Options}
610 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
611 -arch_only -bind_at_load -bundle -bundle_loader @gol
612 -client_name -compatibility_version -current_version @gol
614 -dependency-file -dylib_file -dylinker_install_name @gol
615 -dynamic -dynamiclib -exported_symbols_list @gol
616 -filelist -flat_namespace -force_cpusubtype_ALL @gol
617 -force_flat_namespace -headerpad_max_install_names @gol
619 -image_base -init -install_name -keep_private_externs @gol
620 -multi_module -multiply_defined -multiply_defined_unused @gol
621 -noall_load -no_dead_strip_inits_and_terms @gol
622 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
623 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
624 -private_bundle -read_only_relocs -sectalign @gol
625 -sectobjectsymbols -whyload -seg1addr @gol
626 -sectcreate -sectobjectsymbols -sectorder @gol
627 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
628 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
629 -segprot -segs_read_only_addr -segs_read_write_addr @gol
630 -single_module -static -sub_library -sub_umbrella @gol
631 -twolevel_namespace -umbrella -undefined @gol
632 -unexported_symbols_list -weak_reference_mismatches @gol
633 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
634 -mkernel -mone-byte-bool}
636 @emph{DEC Alpha Options}
637 @gccoptlist{-mno-fp-regs -msoft-float @gol
638 -mieee -mieee-with-inexact -mieee-conformant @gol
639 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
640 -mtrap-precision=@var{mode} -mbuild-constants @gol
641 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
642 -mbwx -mmax -mfix -mcix @gol
643 -mfloat-vax -mfloat-ieee @gol
644 -mexplicit-relocs -msmall-data -mlarge-data @gol
645 -msmall-text -mlarge-text @gol
646 -mmemory-latency=@var{time}}
649 @gccoptlist{-msmall-model -mno-lsim}
652 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
653 -mhard-float -msoft-float @gol
654 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
655 -mdouble -mno-double @gol
656 -mmedia -mno-media -mmuladd -mno-muladd @gol
657 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
658 -mlinked-fp -mlong-calls -malign-labels @gol
659 -mlibrary-pic -macc-4 -macc-8 @gol
660 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
661 -moptimize-membar -mno-optimize-membar @gol
662 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
663 -mvliw-branch -mno-vliw-branch @gol
664 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
665 -mno-nested-cond-exec -mtomcat-stats @gol
669 @emph{GNU/Linux Options}
670 @gccoptlist{-mglibc -muclibc -mbionic -mandroid @gol
671 -tno-android-cc -tno-android-ld}
673 @emph{H8/300 Options}
674 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
677 @gccoptlist{-march=@var{architecture-type} @gol
678 -mdisable-fpregs -mdisable-indexing @gol
679 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
680 -mfixed-range=@var{register-range} @gol
681 -mjump-in-delay -mlinker-opt -mlong-calls @gol
682 -mlong-load-store -mno-disable-fpregs @gol
683 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
684 -mno-jump-in-delay -mno-long-load-store @gol
685 -mno-portable-runtime -mno-soft-float @gol
686 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
687 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
688 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
689 -munix=@var{unix-std} -nolibdld -static -threads}
692 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
693 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
694 -mconstant-gp -mauto-pic -mfused-madd @gol
695 -minline-float-divide-min-latency @gol
696 -minline-float-divide-max-throughput @gol
697 -mno-inline-float-divide @gol
698 -minline-int-divide-min-latency @gol
699 -minline-int-divide-max-throughput @gol
700 -mno-inline-int-divide @gol
701 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
702 -mno-inline-sqrt @gol
703 -mdwarf2-asm -mearly-stop-bits @gol
704 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
705 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
706 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
707 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
708 -msched-spec-ldc -msched-spec-control-ldc @gol
709 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
710 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
711 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
712 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
715 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
716 -msign-extend-enabled -muser-enabled}
718 @emph{M32R/D Options}
719 @gccoptlist{-m32r2 -m32rx -m32r @gol
721 -malign-loops -mno-align-loops @gol
722 -missue-rate=@var{number} @gol
723 -mbranch-cost=@var{number} @gol
724 -mmodel=@var{code-size-model-type} @gol
725 -msdata=@var{sdata-type} @gol
726 -mno-flush-func -mflush-func=@var{name} @gol
727 -mno-flush-trap -mflush-trap=@var{number} @gol
731 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
733 @emph{M680x0 Options}
734 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
735 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
736 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
737 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
738 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
739 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
740 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
741 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
745 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
746 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
747 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
748 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
749 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
752 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
753 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
754 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
755 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
758 @emph{MicroBlaze Options}
759 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
760 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
761 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
762 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
763 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
766 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
767 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
768 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
769 -mips16 -mno-mips16 -mflip-mips16 @gol
770 -minterlink-compressed -mno-interlink-compressed @gol
771 -minterlink-mips16 -mno-interlink-mips16 @gol
772 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
773 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
774 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
775 -mno-float -msingle-float -mdouble-float @gol
776 -modd-spreg -mno-odd-spreg @gol
777 -mabs=@var{mode} -mnan=@var{encoding} @gol
778 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
781 -mvirt -mno-virt @gol
783 -mmicromips -mno-micromips @gol
784 -mfpu=@var{fpu-type} @gol
785 -msmartmips -mno-smartmips @gol
786 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
787 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
788 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
789 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
790 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
791 -membedded-data -mno-embedded-data @gol
792 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
793 -mcode-readable=@var{setting} @gol
794 -msplit-addresses -mno-split-addresses @gol
795 -mexplicit-relocs -mno-explicit-relocs @gol
796 -mcheck-zero-division -mno-check-zero-division @gol
797 -mdivide-traps -mdivide-breaks @gol
798 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
799 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
800 -mfix-24k -mno-fix-24k @gol
801 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
802 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
803 -mfix-vr4120 -mno-fix-vr4120 @gol
804 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
805 -mflush-func=@var{func} -mno-flush-func @gol
806 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
807 -mfp-exceptions -mno-fp-exceptions @gol
808 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
809 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
812 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
813 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
814 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
815 -mno-base-addresses -msingle-exit -mno-single-exit}
817 @emph{MN10300 Options}
818 @gccoptlist{-mmult-bug -mno-mult-bug @gol
819 -mno-am33 -mam33 -mam33-2 -mam34 @gol
820 -mtune=@var{cpu-type} @gol
821 -mreturn-pointer-on-d0 @gol
822 -mno-crt0 -mrelax -mliw -msetlb}
825 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
827 @emph{MSP430 Options}
828 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
832 @gccoptlist{-mbig-endian -mlittle-endian @gol
833 -mreduced-regs -mfull-regs @gol
834 -mcmov -mno-cmov @gol
835 -mperf-ext -mno-perf-ext @gol
836 -mv3push -mno-v3push @gol
837 -m16bit -mno-16bit @gol
838 -misr-vector-size=@var{num} @gol
839 -mcache-block-size=@var{num} @gol
840 -march=@var{arch} @gol
841 -mcmodel=@var{code-model} @gol
844 @emph{Nios II Options}
845 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
847 -mno-bypass-cache -mbypass-cache @gol
848 -mno-cache-volatile -mcache-volatile @gol
849 -mno-fast-sw-div -mfast-sw-div @gol
850 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
851 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
852 -mcustom-fpu-cfg=@var{name} @gol
853 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name}}
855 @emph{Nvidia PTX Options}
856 @gccoptlist{-m32 -m64 -mmainkernel}
858 @emph{PDP-11 Options}
859 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
860 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
861 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
862 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
863 -mbranch-expensive -mbranch-cheap @gol
864 -munix-asm -mdec-asm}
866 @emph{picoChip Options}
867 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
868 -msymbol-as-address -mno-inefficient-warnings}
870 @emph{PowerPC Options}
871 See RS/6000 and PowerPC Options.
874 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
875 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
876 -m64bit-doubles -m32bit-doubles}
878 @emph{RS/6000 and PowerPC Options}
879 @gccoptlist{-mcpu=@var{cpu-type} @gol
880 -mtune=@var{cpu-type} @gol
881 -mcmodel=@var{code-model} @gol
883 -maltivec -mno-altivec @gol
884 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
885 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
886 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
887 -mfprnd -mno-fprnd @gol
888 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
889 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
890 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
891 -malign-power -malign-natural @gol
892 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
893 -msingle-float -mdouble-float -msimple-fpu @gol
894 -mstring -mno-string -mupdate -mno-update @gol
895 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
896 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
897 -mstrict-align -mno-strict-align -mrelocatable @gol
898 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
899 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
900 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
901 -mprioritize-restricted-insns=@var{priority} @gol
902 -msched-costly-dep=@var{dependence_type} @gol
903 -minsert-sched-nops=@var{scheme} @gol
904 -mcall-sysv -mcall-netbsd @gol
905 -maix-struct-return -msvr4-struct-return @gol
906 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
907 -mblock-move-inline-limit=@var{num} @gol
908 -misel -mno-isel @gol
909 -misel=yes -misel=no @gol
911 -mspe=yes -mspe=no @gol
913 -mgen-cell-microcode -mwarn-cell-microcode @gol
914 -mvrsave -mno-vrsave @gol
915 -mmulhw -mno-mulhw @gol
916 -mdlmzb -mno-dlmzb @gol
917 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
918 -mprototype -mno-prototype @gol
919 -msim -mmvme -mads -myellowknife -memb -msdata @gol
920 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
921 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
922 -mno-recip-precision @gol
923 -mveclibabi=@var{type} -mfriz -mno-friz @gol
924 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
925 -msave-toc-indirect -mno-save-toc-indirect @gol
926 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
927 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
928 -mquad-memory -mno-quad-memory @gol
929 -mquad-memory-atomic -mno-quad-memory-atomic @gol
930 -mcompat-align-parm -mno-compat-align-parm @gol
931 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
932 -mupper-regs -mno-upper-regs}
935 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
937 -mbig-endian-data -mlittle-endian-data @gol
940 -mas100-syntax -mno-as100-syntax@gol
942 -mmax-constant-size=@gol
945 -mallow-string-insns -mno-allow-string-insns@gol
946 -mno-warn-multiple-fast-interrupts@gol
947 -msave-acc-in-interrupts}
949 @emph{S/390 and zSeries Options}
950 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
951 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
952 -mlong-double-64 -mlong-double-128 @gol
953 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
954 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
955 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
956 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
957 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
958 -mhotpatch=@var{halfwords},@var{halfwords}}
961 @gccoptlist{-meb -mel @gol
965 -mscore5 -mscore5u -mscore7 -mscore7d}
968 @gccoptlist{-m1 -m2 -m2e @gol
969 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
971 -m4-nofpu -m4-single-only -m4-single -m4 @gol
972 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
973 -m5-64media -m5-64media-nofpu @gol
974 -m5-32media -m5-32media-nofpu @gol
975 -m5-compact -m5-compact-nofpu @gol
976 -mb -ml -mdalign -mrelax @gol
977 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
978 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
979 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
980 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
981 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
982 -maccumulate-outgoing-args -minvalid-symbols @gol
983 -matomic-model=@var{atomic-model} @gol
984 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
985 -mcbranch-force-delay-slot @gol
986 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
987 -mpretend-cmove -mtas}
989 @emph{Solaris 2 Options}
990 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
994 @gccoptlist{-mcpu=@var{cpu-type} @gol
995 -mtune=@var{cpu-type} @gol
996 -mcmodel=@var{code-model} @gol
997 -mmemory-model=@var{mem-model} @gol
998 -m32 -m64 -mapp-regs -mno-app-regs @gol
999 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1000 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1001 -mhard-quad-float -msoft-quad-float @gol
1002 -mstack-bias -mno-stack-bias @gol
1003 -munaligned-doubles -mno-unaligned-doubles @gol
1004 -muser-mode -mno-user-mode @gol
1005 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1006 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1007 -mcbcond -mno-cbcond @gol
1008 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1009 -mfix-at697f -mfix-ut699}
1012 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1013 -msafe-dma -munsafe-dma @gol
1015 -msmall-mem -mlarge-mem -mstdmain @gol
1016 -mfixed-range=@var{register-range} @gol
1018 -maddress-space-conversion -mno-address-space-conversion @gol
1019 -mcache-size=@var{cache-size} @gol
1020 -matomic-updates -mno-atomic-updates}
1022 @emph{System V Options}
1023 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1025 @emph{TILE-Gx Options}
1026 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1027 -mcmodel=@var{code-model}}
1029 @emph{TILEPro Options}
1030 @gccoptlist{-mcpu=@var{cpu} -m32}
1033 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1034 -mprolog-function -mno-prolog-function -mspace @gol
1035 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1036 -mapp-regs -mno-app-regs @gol
1037 -mdisable-callt -mno-disable-callt @gol
1038 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1039 -mv850e -mv850 -mv850e3v5 @gol
1050 @gccoptlist{-mg -mgnu -munix}
1052 @emph{Visium Options}
1053 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1054 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1057 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1058 -mpointer-size=@var{size}}
1060 @emph{VxWorks Options}
1061 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1062 -Xbind-lazy -Xbind-now}
1065 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1066 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1067 -mfpmath=@var{unit} @gol
1068 -masm=@var{dialect} -mno-fancy-math-387 @gol
1069 -mno-fp-ret-in-387 -msoft-float @gol
1070 -mno-wide-multiply -mrtd -malign-double @gol
1071 -mpreferred-stack-boundary=@var{num} @gol
1072 -mincoming-stack-boundary=@var{num} @gol
1073 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1074 -mrecip -mrecip=@var{opt} @gol
1075 -mvzeroupper -mprefer-avx128 @gol
1076 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1077 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1078 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1079 -mclflushopt -mxsavec -mxsaves @gol
1080 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1081 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mthreads @gol
1082 -mno-align-stringops -minline-all-stringops @gol
1083 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1084 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1085 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1086 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1087 -mregparm=@var{num} -msseregparm @gol
1088 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1089 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1090 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1091 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1092 -m32 -m64 -mx32 -m16 -mlarge-data-threshold=@var{num} @gol
1093 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1094 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1095 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1097 @emph{x86 Windows Options}
1098 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1099 -mnop-fun-dllimport -mthread @gol
1100 -municode -mwin32 -mwindows -fno-set-stack-executable}
1102 @emph{Xstormy16 Options}
1105 @emph{Xtensa Options}
1106 @gccoptlist{-mconst16 -mno-const16 @gol
1107 -mfused-madd -mno-fused-madd @gol
1109 -mserialize-volatile -mno-serialize-volatile @gol
1110 -mtext-section-literals -mno-text-section-literals @gol
1111 -mtarget-align -mno-target-align @gol
1112 -mlongcalls -mno-longcalls}
1114 @emph{zSeries Options}
1115 See S/390 and zSeries Options.
1117 @item Code Generation Options
1118 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1119 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1120 -ffixed-@var{reg} -fexceptions @gol
1121 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1122 -fasynchronous-unwind-tables @gol
1123 -fno-gnu-unique @gol
1124 -finhibit-size-directive -finstrument-functions @gol
1125 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1126 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1127 -fno-common -fno-ident @gol
1128 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
1129 -fno-jump-tables @gol
1130 -frecord-gcc-switches @gol
1131 -freg-struct-return -fshort-enums @gol
1132 -fshort-double -fshort-wchar @gol
1133 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1134 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1135 -fno-stack-limit -fsplit-stack @gol
1136 -fleading-underscore -ftls-model=@var{model} @gol
1137 -fstack-reuse=@var{reuse_level} @gol
1138 -ftrapv -fwrapv -fbounds-check @gol
1139 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1140 -fstrict-volatile-bitfields -fsync-libcalls}
1144 @node Overall Options
1145 @section Options Controlling the Kind of Output
1147 Compilation can involve up to four stages: preprocessing, compilation
1148 proper, assembly and linking, always in that order. GCC is capable of
1149 preprocessing and compiling several files either into several
1150 assembler input files, or into one assembler input file; then each
1151 assembler input file produces an object file, and linking combines all
1152 the object files (those newly compiled, and those specified as input)
1153 into an executable file.
1155 @cindex file name suffix
1156 For any given input file, the file name suffix determines what kind of
1157 compilation is done:
1161 C source code that must be preprocessed.
1164 C source code that should not be preprocessed.
1167 C++ source code that should not be preprocessed.
1170 Objective-C source code. Note that you must link with the @file{libobjc}
1171 library to make an Objective-C program work.
1174 Objective-C source code that should not be preprocessed.
1178 Objective-C++ source code. Note that you must link with the @file{libobjc}
1179 library to make an Objective-C++ program work. Note that @samp{.M} refers
1180 to a literal capital M@.
1182 @item @var{file}.mii
1183 Objective-C++ source code that should not be preprocessed.
1186 C, C++, Objective-C or Objective-C++ header file to be turned into a
1187 precompiled header (default), or C, C++ header file to be turned into an
1188 Ada spec (via the @option{-fdump-ada-spec} switch).
1191 @itemx @var{file}.cp
1192 @itemx @var{file}.cxx
1193 @itemx @var{file}.cpp
1194 @itemx @var{file}.CPP
1195 @itemx @var{file}.c++
1197 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1198 the last two letters must both be literally @samp{x}. Likewise,
1199 @samp{.C} refers to a literal capital C@.
1203 Objective-C++ source code that must be preprocessed.
1205 @item @var{file}.mii
1206 Objective-C++ source code that should not be preprocessed.
1210 @itemx @var{file}.hp
1211 @itemx @var{file}.hxx
1212 @itemx @var{file}.hpp
1213 @itemx @var{file}.HPP
1214 @itemx @var{file}.h++
1215 @itemx @var{file}.tcc
1216 C++ header file to be turned into a precompiled header or Ada spec.
1219 @itemx @var{file}.for
1220 @itemx @var{file}.ftn
1221 Fixed form Fortran source code that should not be preprocessed.
1224 @itemx @var{file}.FOR
1225 @itemx @var{file}.fpp
1226 @itemx @var{file}.FPP
1227 @itemx @var{file}.FTN
1228 Fixed form Fortran source code that must be preprocessed (with the traditional
1231 @item @var{file}.f90
1232 @itemx @var{file}.f95
1233 @itemx @var{file}.f03
1234 @itemx @var{file}.f08
1235 Free form Fortran source code that should not be preprocessed.
1237 @item @var{file}.F90
1238 @itemx @var{file}.F95
1239 @itemx @var{file}.F03
1240 @itemx @var{file}.F08
1241 Free form Fortran source code that must be preprocessed (with the
1242 traditional preprocessor).
1247 @c FIXME: Descriptions of Java file types.
1253 @item @var{file}.ads
1254 Ada source code file that contains a library unit declaration (a
1255 declaration of a package, subprogram, or generic, or a generic
1256 instantiation), or a library unit renaming declaration (a package,
1257 generic, or subprogram renaming declaration). Such files are also
1260 @item @var{file}.adb
1261 Ada source code file containing a library unit body (a subprogram or
1262 package body). Such files are also called @dfn{bodies}.
1264 @c GCC also knows about some suffixes for languages not yet included:
1275 @itemx @var{file}.sx
1276 Assembler code that must be preprocessed.
1279 An object file to be fed straight into linking.
1280 Any file name with no recognized suffix is treated this way.
1284 You can specify the input language explicitly with the @option{-x} option:
1287 @item -x @var{language}
1288 Specify explicitly the @var{language} for the following input files
1289 (rather than letting the compiler choose a default based on the file
1290 name suffix). This option applies to all following input files until
1291 the next @option{-x} option. Possible values for @var{language} are:
1293 c c-header cpp-output
1294 c++ c++-header c++-cpp-output
1295 objective-c objective-c-header objective-c-cpp-output
1296 objective-c++ objective-c++-header objective-c++-cpp-output
1297 assembler assembler-with-cpp
1299 f77 f77-cpp-input f95 f95-cpp-input
1305 Turn off any specification of a language, so that subsequent files are
1306 handled according to their file name suffixes (as they are if @option{-x}
1307 has not been used at all).
1309 @item -pass-exit-codes
1310 @opindex pass-exit-codes
1311 Normally the @command{gcc} program exits with the code of 1 if any
1312 phase of the compiler returns a non-success return code. If you specify
1313 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1314 the numerically highest error produced by any phase returning an error
1315 indication. The C, C++, and Fortran front ends return 4 if an internal
1316 compiler error is encountered.
1319 If you only want some of the stages of compilation, you can use
1320 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1321 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1322 @command{gcc} is to stop. Note that some combinations (for example,
1323 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1328 Compile or assemble the source files, but do not link. The linking
1329 stage simply is not done. The ultimate output is in the form of an
1330 object file for each source file.
1332 By default, the object file name for a source file is made by replacing
1333 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1335 Unrecognized input files, not requiring compilation or assembly, are
1340 Stop after the stage of compilation proper; do not assemble. The output
1341 is in the form of an assembler code file for each non-assembler input
1344 By default, the assembler file name for a source file is made by
1345 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1347 Input files that don't require compilation are ignored.
1351 Stop after the preprocessing stage; do not run the compiler proper. The
1352 output is in the form of preprocessed source code, which is sent to the
1355 Input files that don't require preprocessing are ignored.
1357 @cindex output file option
1360 Place output in file @var{file}. This applies to whatever
1361 sort of output is being produced, whether it be an executable file,
1362 an object file, an assembler file or preprocessed C code.
1364 If @option{-o} is not specified, the default is to put an executable
1365 file in @file{a.out}, the object file for
1366 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1367 assembler file in @file{@var{source}.s}, a precompiled header file in
1368 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1373 Print (on standard error output) the commands executed to run the stages
1374 of compilation. Also print the version number of the compiler driver
1375 program and of the preprocessor and the compiler proper.
1379 Like @option{-v} except the commands are not executed and arguments
1380 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1381 This is useful for shell scripts to capture the driver-generated command lines.
1385 Use pipes rather than temporary files for communication between the
1386 various stages of compilation. This fails to work on some systems where
1387 the assembler is unable to read from a pipe; but the GNU assembler has
1392 Print (on the standard output) a description of the command-line options
1393 understood by @command{gcc}. If the @option{-v} option is also specified
1394 then @option{--help} is also passed on to the various processes
1395 invoked by @command{gcc}, so that they can display the command-line options
1396 they accept. If the @option{-Wextra} option has also been specified
1397 (prior to the @option{--help} option), then command-line options that
1398 have no documentation associated with them are also displayed.
1401 @opindex target-help
1402 Print (on the standard output) a description of target-specific command-line
1403 options for each tool. For some targets extra target-specific
1404 information may also be printed.
1406 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1407 Print (on the standard output) a description of the command-line
1408 options understood by the compiler that fit into all specified classes
1409 and qualifiers. These are the supported classes:
1412 @item @samp{optimizers}
1413 Display all of the optimization options supported by the
1416 @item @samp{warnings}
1417 Display all of the options controlling warning messages
1418 produced by the compiler.
1421 Display target-specific options. Unlike the
1422 @option{--target-help} option however, target-specific options of the
1423 linker and assembler are not displayed. This is because those
1424 tools do not currently support the extended @option{--help=} syntax.
1427 Display the values recognized by the @option{--param}
1430 @item @var{language}
1431 Display the options supported for @var{language}, where
1432 @var{language} is the name of one of the languages supported in this
1436 Display the options that are common to all languages.
1439 These are the supported qualifiers:
1442 @item @samp{undocumented}
1443 Display only those options that are undocumented.
1446 Display options taking an argument that appears after an equal
1447 sign in the same continuous piece of text, such as:
1448 @samp{--help=target}.
1450 @item @samp{separate}
1451 Display options taking an argument that appears as a separate word
1452 following the original option, such as: @samp{-o output-file}.
1455 Thus for example to display all the undocumented target-specific
1456 switches supported by the compiler, use:
1459 --help=target,undocumented
1462 The sense of a qualifier can be inverted by prefixing it with the
1463 @samp{^} character, so for example to display all binary warning
1464 options (i.e., ones that are either on or off and that do not take an
1465 argument) that have a description, use:
1468 --help=warnings,^joined,^undocumented
1471 The argument to @option{--help=} should not consist solely of inverted
1474 Combining several classes is possible, although this usually
1475 restricts the output so much that there is nothing to display. One
1476 case where it does work, however, is when one of the classes is
1477 @var{target}. For example, to display all the target-specific
1478 optimization options, use:
1481 --help=target,optimizers
1484 The @option{--help=} option can be repeated on the command line. Each
1485 successive use displays its requested class of options, skipping
1486 those that have already been displayed.
1488 If the @option{-Q} option appears on the command line before the
1489 @option{--help=} option, then the descriptive text displayed by
1490 @option{--help=} is changed. Instead of describing the displayed
1491 options, an indication is given as to whether the option is enabled,
1492 disabled or set to a specific value (assuming that the compiler
1493 knows this at the point where the @option{--help=} option is used).
1495 Here is a truncated example from the ARM port of @command{gcc}:
1498 % gcc -Q -mabi=2 --help=target -c
1499 The following options are target specific:
1501 -mabort-on-noreturn [disabled]
1505 The output is sensitive to the effects of previous command-line
1506 options, so for example it is possible to find out which optimizations
1507 are enabled at @option{-O2} by using:
1510 -Q -O2 --help=optimizers
1513 Alternatively you can discover which binary optimizations are enabled
1514 by @option{-O3} by using:
1517 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1518 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1519 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1522 @item -no-canonical-prefixes
1523 @opindex no-canonical-prefixes
1524 Do not expand any symbolic links, resolve references to @samp{/../}
1525 or @samp{/./}, or make the path absolute when generating a relative
1530 Display the version number and copyrights of the invoked GCC@.
1534 Invoke all subcommands under a wrapper program. The name of the
1535 wrapper program and its parameters are passed as a comma separated
1539 gcc -c t.c -wrapper gdb,--args
1543 This invokes all subprograms of @command{gcc} under
1544 @samp{gdb --args}, thus the invocation of @command{cc1} is
1545 @samp{gdb --args cc1 @dots{}}.
1547 @item -fplugin=@var{name}.so
1549 Load the plugin code in file @var{name}.so, assumed to be a
1550 shared object to be dlopen'd by the compiler. The base name of
1551 the shared object file is used to identify the plugin for the
1552 purposes of argument parsing (See
1553 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1554 Each plugin should define the callback functions specified in the
1557 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1558 @opindex fplugin-arg
1559 Define an argument called @var{key} with a value of @var{value}
1560 for the plugin called @var{name}.
1562 @item -fdump-ada-spec@r{[}-slim@r{]}
1563 @opindex fdump-ada-spec
1564 For C and C++ source and include files, generate corresponding Ada specs.
1565 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1566 GNAT User's Guide}, which provides detailed documentation on this feature.
1568 @item -fada-spec-parent=@var{unit}
1569 @opindex fada-spec-parent
1570 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1571 Ada specs as child units of parent @var{unit}.
1573 @item -fdump-go-spec=@var{file}
1574 @opindex fdump-go-spec
1575 For input files in any language, generate corresponding Go
1576 declarations in @var{file}. This generates Go @code{const},
1577 @code{type}, @code{var}, and @code{func} declarations which may be a
1578 useful way to start writing a Go interface to code written in some
1581 @include @value{srcdir}/../libiberty/at-file.texi
1585 @section Compiling C++ Programs
1587 @cindex suffixes for C++ source
1588 @cindex C++ source file suffixes
1589 C++ source files conventionally use one of the suffixes @samp{.C},
1590 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1591 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1592 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1593 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1594 files with these names and compiles them as C++ programs even if you
1595 call the compiler the same way as for compiling C programs (usually
1596 with the name @command{gcc}).
1600 However, the use of @command{gcc} does not add the C++ library.
1601 @command{g++} is a program that calls GCC and automatically specifies linking
1602 against the C++ library. It treats @samp{.c},
1603 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1604 files unless @option{-x} is used. This program is also useful when
1605 precompiling a C header file with a @samp{.h} extension for use in C++
1606 compilations. On many systems, @command{g++} is also installed with
1607 the name @command{c++}.
1609 @cindex invoking @command{g++}
1610 When you compile C++ programs, you may specify many of the same
1611 command-line options that you use for compiling programs in any
1612 language; or command-line options meaningful for C and related
1613 languages; or options that are meaningful only for C++ programs.
1614 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1615 explanations of options for languages related to C@.
1616 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1617 explanations of options that are meaningful only for C++ programs.
1619 @node C Dialect Options
1620 @section Options Controlling C Dialect
1621 @cindex dialect options
1622 @cindex language dialect options
1623 @cindex options, dialect
1625 The following options control the dialect of C (or languages derived
1626 from C, such as C++, Objective-C and Objective-C++) that the compiler
1630 @cindex ANSI support
1634 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1635 equivalent to @option{-std=c++98}.
1637 This turns off certain features of GCC that are incompatible with ISO
1638 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1639 such as the @code{asm} and @code{typeof} keywords, and
1640 predefined macros such as @code{unix} and @code{vax} that identify the
1641 type of system you are using. It also enables the undesirable and
1642 rarely used ISO trigraph feature. For the C compiler,
1643 it disables recognition of C++ style @samp{//} comments as well as
1644 the @code{inline} keyword.
1646 The alternate keywords @code{__asm__}, @code{__extension__},
1647 @code{__inline__} and @code{__typeof__} continue to work despite
1648 @option{-ansi}. You would not want to use them in an ISO C program, of
1649 course, but it is useful to put them in header files that might be included
1650 in compilations done with @option{-ansi}. Alternate predefined macros
1651 such as @code{__unix__} and @code{__vax__} are also available, with or
1652 without @option{-ansi}.
1654 The @option{-ansi} option does not cause non-ISO programs to be
1655 rejected gratuitously. For that, @option{-Wpedantic} is required in
1656 addition to @option{-ansi}. @xref{Warning Options}.
1658 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1659 option is used. Some header files may notice this macro and refrain
1660 from declaring certain functions or defining certain macros that the
1661 ISO standard doesn't call for; this is to avoid interfering with any
1662 programs that might use these names for other things.
1664 Functions that are normally built in but do not have semantics
1665 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1666 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1667 built-in functions provided by GCC}, for details of the functions
1672 Determine the language standard. @xref{Standards,,Language Standards
1673 Supported by GCC}, for details of these standard versions. This option
1674 is currently only supported when compiling C or C++.
1676 The compiler can accept several base standards, such as @samp{c90} or
1677 @samp{c++98}, and GNU dialects of those standards, such as
1678 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1679 compiler accepts all programs following that standard plus those
1680 using GNU extensions that do not contradict it. For example,
1681 @option{-std=c90} turns off certain features of GCC that are
1682 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1683 keywords, but not other GNU extensions that do not have a meaning in
1684 ISO C90, such as omitting the middle term of a @code{?:}
1685 expression. On the other hand, when a GNU dialect of a standard is
1686 specified, all features supported by the compiler are enabled, even when
1687 those features change the meaning of the base standard. As a result, some
1688 strict-conforming programs may be rejected. The particular standard
1689 is used by @option{-Wpedantic} to identify which features are GNU
1690 extensions given that version of the standard. For example
1691 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1692 comments, while @option{-std=gnu99 -Wpedantic} does not.
1694 A value for this option must be provided; possible values are
1700 Support all ISO C90 programs (certain GNU extensions that conflict
1701 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1703 @item iso9899:199409
1704 ISO C90 as modified in amendment 1.
1710 ISO C99. This standard is substantially completely supported, modulo
1711 bugs and floating-point issues
1712 (mainly but not entirely relating to optional C99 features from
1713 Annexes F and G). See
1714 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1715 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1720 ISO C11, the 2011 revision of the ISO C standard. This standard is
1721 substantially completely supported, modulo bugs, floating-point issues
1722 (mainly but not entirely relating to optional C11 features from
1723 Annexes F and G) and the optional Annexes K (Bounds-checking
1724 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1728 GNU dialect of ISO C90 (including some C99 features).
1732 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1736 GNU dialect of ISO C11. This is the default for C code.
1737 The name @samp{gnu1x} is deprecated.
1741 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1742 additional defect reports. Same as @option{-ansi} for C++ code.
1746 GNU dialect of @option{-std=c++98}. This is the default for
1751 The 2011 ISO C++ standard plus amendments.
1752 The name @samp{c++0x} is deprecated.
1756 GNU dialect of @option{-std=c++11}.
1757 The name @samp{gnu++0x} is deprecated.
1761 The 2014 ISO C++ standard plus amendments.
1762 The name @samp{c++1y} is deprecated.
1766 GNU dialect of @option{-std=c++14}.
1767 The name @samp{gnu++1y} is deprecated.
1770 The next revision of the ISO C++ standard, tentatively planned for
1771 2017. Support is highly experimental, and will almost certainly
1772 change in incompatible ways in future releases.
1775 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1776 and will almost certainly change in incompatible ways in future
1780 @item -fgnu89-inline
1781 @opindex fgnu89-inline
1782 The option @option{-fgnu89-inline} tells GCC to use the traditional
1783 GNU semantics for @code{inline} functions when in C99 mode.
1784 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1785 Using this option is roughly equivalent to adding the
1786 @code{gnu_inline} function attribute to all inline functions
1787 (@pxref{Function Attributes}).
1789 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1790 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1791 specifies the default behavior).
1792 This option is not supported in @option{-std=c90} or
1793 @option{-std=gnu90} mode.
1795 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1796 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1797 in effect for @code{inline} functions. @xref{Common Predefined
1798 Macros,,,cpp,The C Preprocessor}.
1800 @item -aux-info @var{filename}
1802 Output to the given filename prototyped declarations for all functions
1803 declared and/or defined in a translation unit, including those in header
1804 files. This option is silently ignored in any language other than C@.
1806 Besides declarations, the file indicates, in comments, the origin of
1807 each declaration (source file and line), whether the declaration was
1808 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1809 @samp{O} for old, respectively, in the first character after the line
1810 number and the colon), and whether it came from a declaration or a
1811 definition (@samp{C} or @samp{F}, respectively, in the following
1812 character). In the case of function definitions, a K&R-style list of
1813 arguments followed by their declarations is also provided, inside
1814 comments, after the declaration.
1816 @item -fallow-parameterless-variadic-functions
1817 @opindex fallow-parameterless-variadic-functions
1818 Accept variadic functions without named parameters.
1820 Although it is possible to define such a function, this is not very
1821 useful as it is not possible to read the arguments. This is only
1822 supported for C as this construct is allowed by C++.
1826 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1827 keyword, so that code can use these words as identifiers. You can use
1828 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1829 instead. @option{-ansi} implies @option{-fno-asm}.
1831 In C++, this switch only affects the @code{typeof} keyword, since
1832 @code{asm} and @code{inline} are standard keywords. You may want to
1833 use the @option{-fno-gnu-keywords} flag instead, which has the same
1834 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1835 switch only affects the @code{asm} and @code{typeof} keywords, since
1836 @code{inline} is a standard keyword in ISO C99.
1839 @itemx -fno-builtin-@var{function}
1840 @opindex fno-builtin
1841 @cindex built-in functions
1842 Don't recognize built-in functions that do not begin with
1843 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1844 functions provided by GCC}, for details of the functions affected,
1845 including those which are not built-in functions when @option{-ansi} or
1846 @option{-std} options for strict ISO C conformance are used because they
1847 do not have an ISO standard meaning.
1849 GCC normally generates special code to handle certain built-in functions
1850 more efficiently; for instance, calls to @code{alloca} may become single
1851 instructions which adjust the stack directly, and calls to @code{memcpy}
1852 may become inline copy loops. The resulting code is often both smaller
1853 and faster, but since the function calls no longer appear as such, you
1854 cannot set a breakpoint on those calls, nor can you change the behavior
1855 of the functions by linking with a different library. In addition,
1856 when a function is recognized as a built-in function, GCC may use
1857 information about that function to warn about problems with calls to
1858 that function, or to generate more efficient code, even if the
1859 resulting code still contains calls to that function. For example,
1860 warnings are given with @option{-Wformat} for bad calls to
1861 @code{printf} when @code{printf} is built in and @code{strlen} is
1862 known not to modify global memory.
1864 With the @option{-fno-builtin-@var{function}} option
1865 only the built-in function @var{function} is
1866 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1867 function is named that is not built-in in this version of GCC, this
1868 option is ignored. There is no corresponding
1869 @option{-fbuiltin-@var{function}} option; if you wish to enable
1870 built-in functions selectively when using @option{-fno-builtin} or
1871 @option{-ffreestanding}, you may define macros such as:
1874 #define abs(n) __builtin_abs ((n))
1875 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1880 @cindex hosted environment
1882 Assert that compilation targets a hosted environment. This implies
1883 @option{-fbuiltin}. A hosted environment is one in which the
1884 entire standard library is available, and in which @code{main} has a return
1885 type of @code{int}. Examples are nearly everything except a kernel.
1886 This is equivalent to @option{-fno-freestanding}.
1888 @item -ffreestanding
1889 @opindex ffreestanding
1890 @cindex hosted environment
1892 Assert that compilation targets a freestanding environment. This
1893 implies @option{-fno-builtin}. A freestanding environment
1894 is one in which the standard library may not exist, and program startup may
1895 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1896 This is equivalent to @option{-fno-hosted}.
1898 @xref{Standards,,Language Standards Supported by GCC}, for details of
1899 freestanding and hosted environments.
1903 @cindex OpenACC accelerator programming
1904 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1905 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1906 compiler generates accelerated code according to the OpenACC Application
1907 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1908 implies @option{-pthread}, and thus is only supported on targets that
1909 have support for @option{-pthread}.
1911 Note that this is an experimental feature, incomplete, and subject to
1912 change in future versions of GCC. See
1913 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1917 @cindex OpenMP parallel
1918 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1919 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1920 compiler generates parallel code according to the OpenMP Application
1921 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1922 implies @option{-pthread}, and thus is only supported on targets that
1923 have support for @option{-pthread}. @option{-fopenmp} implies
1924 @option{-fopenmp-simd}.
1927 @opindex fopenmp-simd
1930 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1931 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1936 @cindex Enable Cilk Plus
1937 Enable the usage of Cilk Plus language extension features for C/C++.
1938 When the option @option{-fcilkplus} is specified, enable the usage of
1939 the Cilk Plus Language extension features for C/C++. The present
1940 implementation follows ABI version 1.2. This is an experimental
1941 feature that is only partially complete, and whose interface may
1942 change in future versions of GCC as the official specification
1943 changes. Currently, all features but @code{_Cilk_for} have been
1948 When the option @option{-fgnu-tm} is specified, the compiler
1949 generates code for the Linux variant of Intel's current Transactional
1950 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1951 an experimental feature whose interface may change in future versions
1952 of GCC, as the official specification changes. Please note that not
1953 all architectures are supported for this feature.
1955 For more information on GCC's support for transactional memory,
1956 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1957 Transactional Memory Library}.
1959 Note that the transactional memory feature is not supported with
1960 non-call exceptions (@option{-fnon-call-exceptions}).
1962 @item -fms-extensions
1963 @opindex fms-extensions
1964 Accept some non-standard constructs used in Microsoft header files.
1966 In C++ code, this allows member names in structures to be similar
1967 to previous types declarations.
1976 Some cases of unnamed fields in structures and unions are only
1977 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1978 fields within structs/unions}, for details.
1980 Note that this option is off for all targets but x86
1981 targets using ms-abi.
1983 @item -fplan9-extensions
1984 @opindex fplan9-extensions
1985 Accept some non-standard constructs used in Plan 9 code.
1987 This enables @option{-fms-extensions}, permits passing pointers to
1988 structures with anonymous fields to functions that expect pointers to
1989 elements of the type of the field, and permits referring to anonymous
1990 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
1991 struct/union fields within structs/unions}, for details. This is only
1992 supported for C, not C++.
1996 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1997 options for strict ISO C conformance) implies @option{-trigraphs}.
1999 @cindex traditional C language
2000 @cindex C language, traditional
2002 @itemx -traditional-cpp
2003 @opindex traditional-cpp
2004 @opindex traditional
2005 Formerly, these options caused GCC to attempt to emulate a pre-standard
2006 C compiler. They are now only supported with the @option{-E} switch.
2007 The preprocessor continues to support a pre-standard mode. See the GNU
2008 CPP manual for details.
2010 @item -fcond-mismatch
2011 @opindex fcond-mismatch
2012 Allow conditional expressions with mismatched types in the second and
2013 third arguments. The value of such an expression is void. This option
2014 is not supported for C++.
2016 @item -flax-vector-conversions
2017 @opindex flax-vector-conversions
2018 Allow implicit conversions between vectors with differing numbers of
2019 elements and/or incompatible element types. This option should not be
2022 @item -funsigned-char
2023 @opindex funsigned-char
2024 Let the type @code{char} be unsigned, like @code{unsigned char}.
2026 Each kind of machine has a default for what @code{char} should
2027 be. It is either like @code{unsigned char} by default or like
2028 @code{signed char} by default.
2030 Ideally, a portable program should always use @code{signed char} or
2031 @code{unsigned char} when it depends on the signedness of an object.
2032 But many programs have been written to use plain @code{char} and
2033 expect it to be signed, or expect it to be unsigned, depending on the
2034 machines they were written for. This option, and its inverse, let you
2035 make such a program work with the opposite default.
2037 The type @code{char} is always a distinct type from each of
2038 @code{signed char} or @code{unsigned char}, even though its behavior
2039 is always just like one of those two.
2042 @opindex fsigned-char
2043 Let the type @code{char} be signed, like @code{signed char}.
2045 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2046 the negative form of @option{-funsigned-char}. Likewise, the option
2047 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2049 @item -fsigned-bitfields
2050 @itemx -funsigned-bitfields
2051 @itemx -fno-signed-bitfields
2052 @itemx -fno-unsigned-bitfields
2053 @opindex fsigned-bitfields
2054 @opindex funsigned-bitfields
2055 @opindex fno-signed-bitfields
2056 @opindex fno-unsigned-bitfields
2057 These options control whether a bit-field is signed or unsigned, when the
2058 declaration does not use either @code{signed} or @code{unsigned}. By
2059 default, such a bit-field is signed, because this is consistent: the
2060 basic integer types such as @code{int} are signed types.
2063 @node C++ Dialect Options
2064 @section Options Controlling C++ Dialect
2066 @cindex compiler options, C++
2067 @cindex C++ options, command-line
2068 @cindex options, C++
2069 This section describes the command-line options that are only meaningful
2070 for C++ programs. You can also use most of the GNU compiler options
2071 regardless of what language your program is in. For example, you
2072 might compile a file @file{firstClass.C} like this:
2075 g++ -g -frepo -O -c firstClass.C
2079 In this example, only @option{-frepo} is an option meant
2080 only for C++ programs; you can use the other options with any
2081 language supported by GCC@.
2083 Here is a list of options that are @emph{only} for compiling C++ programs:
2087 @item -fabi-version=@var{n}
2088 @opindex fabi-version
2089 Use version @var{n} of the C++ ABI@. The default is version 0.
2091 Version 0 refers to the version conforming most closely to
2092 the C++ ABI specification. Therefore, the ABI obtained using version 0
2093 will change in different versions of G++ as ABI bugs are fixed.
2095 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2097 Version 2 is the version of the C++ ABI that first appeared in G++
2098 3.4, and was the default through G++ 4.9.
2100 Version 3 corrects an error in mangling a constant address as a
2103 Version 4, which first appeared in G++ 4.5, implements a standard
2104 mangling for vector types.
2106 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2107 attribute const/volatile on function pointer types, decltype of a
2108 plain decl, and use of a function parameter in the declaration of
2111 Version 6, which first appeared in G++ 4.7, corrects the promotion
2112 behavior of C++11 scoped enums and the mangling of template argument
2113 packs, const/static_cast, prefix ++ and --, and a class scope function
2114 used as a template argument.
2116 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2117 builtin type and corrects the mangling of lambdas in default argument
2120 Version 8, which first appeared in G++ 4.9, corrects the substitution
2121 behavior of function types with function-cv-qualifiers.
2123 See also @option{-Wabi}.
2125 @item -fabi-compat-version=@var{n}
2126 @opindex fabi-compat-version
2127 On targets that support strong aliases, G++
2128 works around mangling changes by creating an alias with the correct
2129 mangled name when defining a symbol with an incorrect mangled name.
2130 This switch specifies which ABI version to use for the alias.
2132 With @option{-fabi-version=0} (the default), this defaults to 2. If
2133 another ABI version is explicitly selected, this defaults to 0.
2135 The compatibility version is also set by @option{-Wabi=@var{n}}.
2137 @item -fno-access-control
2138 @opindex fno-access-control
2139 Turn off all access checking. This switch is mainly useful for working
2140 around bugs in the access control code.
2144 Check that the pointer returned by @code{operator new} is non-null
2145 before attempting to modify the storage allocated. This check is
2146 normally unnecessary because the C++ standard specifies that
2147 @code{operator new} only returns @code{0} if it is declared
2148 @code{throw()}, in which case the compiler always checks the
2149 return value even without this option. In all other cases, when
2150 @code{operator new} has a non-empty exception specification, memory
2151 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2152 @samp{new (nothrow)}.
2154 @item -fconstexpr-depth=@var{n}
2155 @opindex fconstexpr-depth
2156 Set the maximum nested evaluation depth for C++11 constexpr functions
2157 to @var{n}. A limit is needed to detect endless recursion during
2158 constant expression evaluation. The minimum specified by the standard
2161 @item -fdeduce-init-list
2162 @opindex fdeduce-init-list
2163 Enable deduction of a template type parameter as
2164 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2167 template <class T> auto forward(T t) -> decltype (realfn (t))
2174 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2178 This deduction was implemented as a possible extension to the
2179 originally proposed semantics for the C++11 standard, but was not part
2180 of the final standard, so it is disabled by default. This option is
2181 deprecated, and may be removed in a future version of G++.
2183 @item -ffriend-injection
2184 @opindex ffriend-injection
2185 Inject friend functions into the enclosing namespace, so that they are
2186 visible outside the scope of the class in which they are declared.
2187 Friend functions were documented to work this way in the old Annotated
2188 C++ Reference Manual.
2189 However, in ISO C++ a friend function that is not declared
2190 in an enclosing scope can only be found using argument dependent
2191 lookup. GCC defaults to the standard behavior.
2193 This option is for compatibility, and may be removed in a future
2196 @item -fno-elide-constructors
2197 @opindex fno-elide-constructors
2198 The C++ standard allows an implementation to omit creating a temporary
2199 that is only used to initialize another object of the same type.
2200 Specifying this option disables that optimization, and forces G++ to
2201 call the copy constructor in all cases.
2203 @item -fno-enforce-eh-specs
2204 @opindex fno-enforce-eh-specs
2205 Don't generate code to check for violation of exception specifications
2206 at run time. This option violates the C++ standard, but may be useful
2207 for reducing code size in production builds, much like defining
2208 @code{NDEBUG}. This does not give user code permission to throw
2209 exceptions in violation of the exception specifications; the compiler
2210 still optimizes based on the specifications, so throwing an
2211 unexpected exception results in undefined behavior at run time.
2213 @item -fextern-tls-init
2214 @itemx -fno-extern-tls-init
2215 @opindex fextern-tls-init
2216 @opindex fno-extern-tls-init
2217 The C++11 and OpenMP standards allow @code{thread_local} and
2218 @code{threadprivate} variables to have dynamic (runtime)
2219 initialization. To support this, any use of such a variable goes
2220 through a wrapper function that performs any necessary initialization.
2221 When the use and definition of the variable are in the same
2222 translation unit, this overhead can be optimized away, but when the
2223 use is in a different translation unit there is significant overhead
2224 even if the variable doesn't actually need dynamic initialization. If
2225 the programmer can be sure that no use of the variable in a
2226 non-defining TU needs to trigger dynamic initialization (either
2227 because the variable is statically initialized, or a use of the
2228 variable in the defining TU will be executed before any uses in
2229 another TU), they can avoid this overhead with the
2230 @option{-fno-extern-tls-init} option.
2232 On targets that support symbol aliases, the default is
2233 @option{-fextern-tls-init}. On targets that do not support symbol
2234 aliases, the default is @option{-fno-extern-tls-init}.
2237 @itemx -fno-for-scope
2239 @opindex fno-for-scope
2240 If @option{-ffor-scope} is specified, the scope of variables declared in
2241 a @i{for-init-statement} is limited to the @code{for} loop itself,
2242 as specified by the C++ standard.
2243 If @option{-fno-for-scope} is specified, the scope of variables declared in
2244 a @i{for-init-statement} extends to the end of the enclosing scope,
2245 as was the case in old versions of G++, and other (traditional)
2246 implementations of C++.
2248 If neither flag is given, the default is to follow the standard,
2249 but to allow and give a warning for old-style code that would
2250 otherwise be invalid, or have different behavior.
2252 @item -fno-gnu-keywords
2253 @opindex fno-gnu-keywords
2254 Do not recognize @code{typeof} as a keyword, so that code can use this
2255 word as an identifier. You can use the keyword @code{__typeof__} instead.
2256 @option{-ansi} implies @option{-fno-gnu-keywords}.
2258 @item -fno-implicit-templates
2259 @opindex fno-implicit-templates
2260 Never emit code for non-inline templates that are instantiated
2261 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2262 @xref{Template Instantiation}, for more information.
2264 @item -fno-implicit-inline-templates
2265 @opindex fno-implicit-inline-templates
2266 Don't emit code for implicit instantiations of inline templates, either.
2267 The default is to handle inlines differently so that compiles with and
2268 without optimization need the same set of explicit instantiations.
2270 @item -fno-implement-inlines
2271 @opindex fno-implement-inlines
2272 To save space, do not emit out-of-line copies of inline functions
2273 controlled by @code{#pragma implementation}. This causes linker
2274 errors if these functions are not inlined everywhere they are called.
2276 @item -fms-extensions
2277 @opindex fms-extensions
2278 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2279 int and getting a pointer to member function via non-standard syntax.
2281 @item -fno-nonansi-builtins
2282 @opindex fno-nonansi-builtins
2283 Disable built-in declarations of functions that are not mandated by
2284 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2285 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2288 @opindex fnothrow-opt
2289 Treat a @code{throw()} exception specification as if it were a
2290 @code{noexcept} specification to reduce or eliminate the text size
2291 overhead relative to a function with no exception specification. If
2292 the function has local variables of types with non-trivial
2293 destructors, the exception specification actually makes the
2294 function smaller because the EH cleanups for those variables can be
2295 optimized away. The semantic effect is that an exception thrown out of
2296 a function with such an exception specification results in a call
2297 to @code{terminate} rather than @code{unexpected}.
2299 @item -fno-operator-names
2300 @opindex fno-operator-names
2301 Do not treat the operator name keywords @code{and}, @code{bitand},
2302 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2303 synonyms as keywords.
2305 @item -fno-optional-diags
2306 @opindex fno-optional-diags
2307 Disable diagnostics that the standard says a compiler does not need to
2308 issue. Currently, the only such diagnostic issued by G++ is the one for
2309 a name having multiple meanings within a class.
2312 @opindex fpermissive
2313 Downgrade some diagnostics about nonconformant code from errors to
2314 warnings. Thus, using @option{-fpermissive} allows some
2315 nonconforming code to compile.
2317 @item -fno-pretty-templates
2318 @opindex fno-pretty-templates
2319 When an error message refers to a specialization of a function
2320 template, the compiler normally prints the signature of the
2321 template followed by the template arguments and any typedefs or
2322 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2323 rather than @code{void f(int)}) so that it's clear which template is
2324 involved. When an error message refers to a specialization of a class
2325 template, the compiler omits any template arguments that match
2326 the default template arguments for that template. If either of these
2327 behaviors make it harder to understand the error message rather than
2328 easier, you can use @option{-fno-pretty-templates} to disable them.
2332 Enable automatic template instantiation at link time. This option also
2333 implies @option{-fno-implicit-templates}. @xref{Template
2334 Instantiation}, for more information.
2338 Disable generation of information about every class with virtual
2339 functions for use by the C++ run-time type identification features
2340 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2341 of the language, you can save some space by using this flag. Note that
2342 exception handling uses the same information, but G++ generates it as
2343 needed. The @code{dynamic_cast} operator can still be used for casts that
2344 do not require run-time type information, i.e.@: casts to @code{void *} or to
2345 unambiguous base classes.
2347 @item -fsized-deallocation
2348 @opindex fsized-deallocation
2349 Enable the built-in global declarations
2351 void operator delete (void *, std::size_t) noexcept;
2352 void operator delete[] (void *, std::size_t) noexcept;
2354 as introduced in C++14. This is useful for user-defined replacement
2355 deallocation functions that, for example, use the size of the object
2356 to make deallocation faster. Enabled by default under
2357 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2358 warns about places that might want to add a definition.
2362 Emit statistics about front-end processing at the end of the compilation.
2363 This information is generally only useful to the G++ development team.
2365 @item -fstrict-enums
2366 @opindex fstrict-enums
2367 Allow the compiler to optimize using the assumption that a value of
2368 enumerated type can only be one of the values of the enumeration (as
2369 defined in the C++ standard; basically, a value that can be
2370 represented in the minimum number of bits needed to represent all the
2371 enumerators). This assumption may not be valid if the program uses a
2372 cast to convert an arbitrary integer value to the enumerated type.
2374 @item -ftemplate-backtrace-limit=@var{n}
2375 @opindex ftemplate-backtrace-limit
2376 Set the maximum number of template instantiation notes for a single
2377 warning or error to @var{n}. The default value is 10.
2379 @item -ftemplate-depth=@var{n}
2380 @opindex ftemplate-depth
2381 Set the maximum instantiation depth for template classes to @var{n}.
2382 A limit on the template instantiation depth is needed to detect
2383 endless recursions during template class instantiation. ANSI/ISO C++
2384 conforming programs must not rely on a maximum depth greater than 17
2385 (changed to 1024 in C++11). The default value is 900, as the compiler
2386 can run out of stack space before hitting 1024 in some situations.
2388 @item -fno-threadsafe-statics
2389 @opindex fno-threadsafe-statics
2390 Do not emit the extra code to use the routines specified in the C++
2391 ABI for thread-safe initialization of local statics. You can use this
2392 option to reduce code size slightly in code that doesn't need to be
2395 @item -fuse-cxa-atexit
2396 @opindex fuse-cxa-atexit
2397 Register destructors for objects with static storage duration with the
2398 @code{__cxa_atexit} function rather than the @code{atexit} function.
2399 This option is required for fully standards-compliant handling of static
2400 destructors, but only works if your C library supports
2401 @code{__cxa_atexit}.
2403 @item -fno-use-cxa-get-exception-ptr
2404 @opindex fno-use-cxa-get-exception-ptr
2405 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2406 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2407 if the runtime routine is not available.
2409 @item -fvisibility-inlines-hidden
2410 @opindex fvisibility-inlines-hidden
2411 This switch declares that the user does not attempt to compare
2412 pointers to inline functions or methods where the addresses of the two functions
2413 are taken in different shared objects.
2415 The effect of this is that GCC may, effectively, mark inline methods with
2416 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2417 appear in the export table of a DSO and do not require a PLT indirection
2418 when used within the DSO@. Enabling this option can have a dramatic effect
2419 on load and link times of a DSO as it massively reduces the size of the
2420 dynamic export table when the library makes heavy use of templates.
2422 The behavior of this switch is not quite the same as marking the
2423 methods as hidden directly, because it does not affect static variables
2424 local to the function or cause the compiler to deduce that
2425 the function is defined in only one shared object.
2427 You may mark a method as having a visibility explicitly to negate the
2428 effect of the switch for that method. For example, if you do want to
2429 compare pointers to a particular inline method, you might mark it as
2430 having default visibility. Marking the enclosing class with explicit
2431 visibility has no effect.
2433 Explicitly instantiated inline methods are unaffected by this option
2434 as their linkage might otherwise cross a shared library boundary.
2435 @xref{Template Instantiation}.
2437 @item -fvisibility-ms-compat
2438 @opindex fvisibility-ms-compat
2439 This flag attempts to use visibility settings to make GCC's C++
2440 linkage model compatible with that of Microsoft Visual Studio.
2442 The flag makes these changes to GCC's linkage model:
2446 It sets the default visibility to @code{hidden}, like
2447 @option{-fvisibility=hidden}.
2450 Types, but not their members, are not hidden by default.
2453 The One Definition Rule is relaxed for types without explicit
2454 visibility specifications that are defined in more than one
2455 shared object: those declarations are permitted if they are
2456 permitted when this option is not used.
2459 In new code it is better to use @option{-fvisibility=hidden} and
2460 export those classes that are intended to be externally visible.
2461 Unfortunately it is possible for code to rely, perhaps accidentally,
2462 on the Visual Studio behavior.
2464 Among the consequences of these changes are that static data members
2465 of the same type with the same name but defined in different shared
2466 objects are different, so changing one does not change the other;
2467 and that pointers to function members defined in different shared
2468 objects may not compare equal. When this flag is given, it is a
2469 violation of the ODR to define types with the same name differently.
2471 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2472 @opindex fvtable-verify
2473 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2474 feature that verifies at run time, for every virtual call, that
2475 the vtable pointer through which the call is made is valid for the type of
2476 the object, and has not been corrupted or overwritten. If an invalid vtable
2477 pointer is detected at run time, an error is reported and execution of the
2478 program is immediately halted.
2480 This option causes run-time data structures to be built at program startup,
2481 which are used for verifying the vtable pointers.
2482 The options @samp{std} and @samp{preinit}
2483 control the timing of when these data structures are built. In both cases the
2484 data structures are built before execution reaches @code{main}. Using
2485 @option{-fvtable-verify=std} causes the data structures to be built after
2486 shared libraries have been loaded and initialized.
2487 @option{-fvtable-verify=preinit} causes them to be built before shared
2488 libraries have been loaded and initialized.
2490 If this option appears multiple times in the command line with different
2491 values specified, @samp{none} takes highest priority over both @samp{std} and
2492 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2496 When used in conjunction with @option{-fvtable-verify=std} or
2497 @option{-fvtable-verify=preinit}, causes debug versions of the
2498 runtime functions for the vtable verification feature to be called.
2499 This flag also causes the compiler to log information about which
2500 vtable pointers it finds for each class.
2501 This information is written to a file named @file{vtv_set_ptr_data.log}
2502 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2503 if that is defined or the current working directory otherwise.
2505 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2506 file, be sure to delete any existing one.
2509 @opindex fvtv-counts
2510 This is a debugging flag. When used in conjunction with
2511 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2512 causes the compiler to keep track of the total number of virtual calls
2513 it encounters and the number of verifications it inserts. It also
2514 counts the number of calls to certain run-time library functions
2515 that it inserts and logs this information for each compilation unit.
2516 The compiler writes this information to a file named
2517 @file{vtv_count_data.log} in the directory named by the environment
2518 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2519 directory otherwise. It also counts the size of the vtable pointer sets
2520 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2521 in the same directory.
2523 Note: This feature @emph{appends} data to the log files. To get fresh log
2524 files, be sure to delete any existing ones.
2528 Do not use weak symbol support, even if it is provided by the linker.
2529 By default, G++ uses weak symbols if they are available. This
2530 option exists only for testing, and should not be used by end-users;
2531 it results in inferior code and has no benefits. This option may
2532 be removed in a future release of G++.
2536 Do not search for header files in the standard directories specific to
2537 C++, but do still search the other standard directories. (This option
2538 is used when building the C++ library.)
2541 In addition, these optimization, warning, and code generation options
2542 have meanings only for C++ programs:
2545 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2548 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2549 G++ to warn when it generates code that is probably not compatible with the
2550 vendor-neutral C++ ABI@. Since G++ now defaults to
2551 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2552 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2553 or an older compatibility version is selected (with
2554 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2556 Although an effort has been made to warn about
2557 all such cases, there are probably some cases that are not warned about,
2558 even though G++ is generating incompatible code. There may also be
2559 cases where warnings are emitted even though the code that is generated
2562 You should rewrite your code to avoid these warnings if you are
2563 concerned about the fact that code generated by G++ may not be binary
2564 compatible with code generated by other compilers.
2566 @option{-Wabi} can also be used with an explicit version number to
2567 warn about compatibility with a particular @option{-fabi-version}
2568 level, e.g. @option{-Wabi=2} to warn about changes relative to
2569 @option{-fabi-version=2}. Specifying a version number also sets
2570 @option{-fabi-compat-version=@var{n}}.
2572 The known incompatibilities in @option{-fabi-version=2} (which was the
2573 default from GCC 3.4 to 4.9) include:
2578 A template with a non-type template parameter of reference type was
2579 mangled incorrectly:
2582 template <int &> struct S @{@};
2586 This was fixed in @option{-fabi-version=3}.
2589 SIMD vector types declared using @code{__attribute ((vector_size))} were
2590 mangled in a non-standard way that does not allow for overloading of
2591 functions taking vectors of different sizes.
2593 The mangling was changed in @option{-fabi-version=4}.
2596 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2597 qualifiers, and @code{decltype} of a plain declaration was folded away.
2599 These mangling issues were fixed in @option{-fabi-version=5}.
2602 Scoped enumerators passed as arguments to a variadic function are
2603 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2604 On most targets this does not actually affect the parameter passing
2605 ABI, as there is no way to pass an argument smaller than @code{int}.
2607 Also, the ABI changed the mangling of template argument packs,
2608 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2609 a class scope function used as a template argument.
2611 These issues were corrected in @option{-fabi-version=6}.
2614 Lambdas in default argument scope were mangled incorrectly, and the
2615 ABI changed the mangling of @code{nullptr_t}.
2617 These issues were corrected in @option{-fabi-version=7}.
2620 When mangling a function type with function-cv-qualifiers, the
2621 un-qualified function type was incorrectly treated as a substitution
2624 This was fixed in @option{-fabi-version=8}.
2627 It also warns about psABI-related changes. The known psABI changes at this
2633 For SysV/x86-64, unions with @code{long double} members are
2634 passed in memory as specified in psABI. For example:
2644 @code{union U} is always passed in memory.
2648 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2651 Warn when a type with an ABI tag is used in a context that does not
2652 have that ABI tag. See @ref{C++ Attributes} for more information
2655 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2656 @opindex Wctor-dtor-privacy
2657 @opindex Wno-ctor-dtor-privacy
2658 Warn when a class seems unusable because all the constructors or
2659 destructors in that class are private, and it has neither friends nor
2660 public static member functions. Also warn if there are no non-private
2661 methods, and there's at least one private member function that isn't
2662 a constructor or destructor.
2664 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2665 @opindex Wdelete-non-virtual-dtor
2666 @opindex Wno-delete-non-virtual-dtor
2667 Warn when @code{delete} is used to destroy an instance of a class that
2668 has virtual functions and non-virtual destructor. It is unsafe to delete
2669 an instance of a derived class through a pointer to a base class if the
2670 base class does not have a virtual destructor. This warning is enabled
2673 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2674 @opindex Wliteral-suffix
2675 @opindex Wno-literal-suffix
2676 Warn when a string or character literal is followed by a ud-suffix which does
2677 not begin with an underscore. As a conforming extension, GCC treats such
2678 suffixes as separate preprocessing tokens in order to maintain backwards
2679 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2683 #define __STDC_FORMAT_MACROS
2684 #include <inttypes.h>
2689 printf("My int64: %"PRId64"\n", i64);
2693 In this case, @code{PRId64} is treated as a separate preprocessing token.
2695 This warning is enabled by default.
2697 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2699 @opindex Wno-narrowing
2700 Warn when a narrowing conversion prohibited by C++11 occurs within
2704 int i = @{ 2.2 @}; // error: narrowing from double to int
2707 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2709 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2710 required by the standard. Note that this does not affect the meaning
2711 of well-formed code; narrowing conversions are still considered
2712 ill-formed in SFINAE context.
2714 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2716 @opindex Wno-noexcept
2717 Warn when a noexcept-expression evaluates to false because of a call
2718 to a function that does not have a non-throwing exception
2719 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2720 the compiler to never throw an exception.
2722 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2723 @opindex Wnon-virtual-dtor
2724 @opindex Wno-non-virtual-dtor
2725 Warn when a class has virtual functions and an accessible non-virtual
2726 destructor itself or in an accessible polymorphic base class, in which
2727 case it is possible but unsafe to delete an instance of a derived
2728 class through a pointer to the class itself or base class. This
2729 warning is automatically enabled if @option{-Weffc++} is specified.
2731 @item -Wreorder @r{(C++ and Objective-C++ only)}
2733 @opindex Wno-reorder
2734 @cindex reordering, warning
2735 @cindex warning for reordering of member initializers
2736 Warn when the order of member initializers given in the code does not
2737 match the order in which they must be executed. For instance:
2743 A(): j (0), i (1) @{ @}
2748 The compiler rearranges the member initializers for @code{i}
2749 and @code{j} to match the declaration order of the members, emitting
2750 a warning to that effect. This warning is enabled by @option{-Wall}.
2752 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2753 @opindex fext-numeric-literals
2754 @opindex fno-ext-numeric-literals
2755 Accept imaginary, fixed-point, or machine-defined
2756 literal number suffixes as GNU extensions.
2757 When this option is turned off these suffixes are treated
2758 as C++11 user-defined literal numeric suffixes.
2759 This is on by default for all pre-C++11 dialects and all GNU dialects:
2760 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2761 @option{-std=gnu++14}.
2762 This option is off by default
2763 for ISO C++11 onwards (@option{-std=c++11}, ...).
2766 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2769 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2772 Warn about violations of the following style guidelines from Scott Meyers'
2773 @cite{Effective C++} series of books:
2777 Define a copy constructor and an assignment operator for classes
2778 with dynamically-allocated memory.
2781 Prefer initialization to assignment in constructors.
2784 Have @code{operator=} return a reference to @code{*this}.
2787 Don't try to return a reference when you must return an object.
2790 Distinguish between prefix and postfix forms of increment and
2791 decrement operators.
2794 Never overload @code{&&}, @code{||}, or @code{,}.
2798 This option also enables @option{-Wnon-virtual-dtor}, which is also
2799 one of the effective C++ recommendations. However, the check is
2800 extended to warn about the lack of virtual destructor in accessible
2801 non-polymorphic bases classes too.
2803 When selecting this option, be aware that the standard library
2804 headers do not obey all of these guidelines; use @samp{grep -v}
2805 to filter out those warnings.
2807 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2808 @opindex Wstrict-null-sentinel
2809 @opindex Wno-strict-null-sentinel
2810 Warn about the use of an uncasted @code{NULL} as sentinel. When
2811 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2812 to @code{__null}. Although it is a null pointer constant rather than a
2813 null pointer, it is guaranteed to be of the same size as a pointer.
2814 But this use is not portable across different compilers.
2816 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2817 @opindex Wno-non-template-friend
2818 @opindex Wnon-template-friend
2819 Disable warnings when non-templatized friend functions are declared
2820 within a template. Since the advent of explicit template specification
2821 support in G++, if the name of the friend is an unqualified-id (i.e.,
2822 @samp{friend foo(int)}), the C++ language specification demands that the
2823 friend declare or define an ordinary, nontemplate function. (Section
2824 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2825 could be interpreted as a particular specialization of a templatized
2826 function. Because this non-conforming behavior is no longer the default
2827 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2828 check existing code for potential trouble spots and is on by default.
2829 This new compiler behavior can be turned off with
2830 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2831 but disables the helpful warning.
2833 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2834 @opindex Wold-style-cast
2835 @opindex Wno-old-style-cast
2836 Warn if an old-style (C-style) cast to a non-void type is used within
2837 a C++ program. The new-style casts (@code{dynamic_cast},
2838 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2839 less vulnerable to unintended effects and much easier to search for.
2841 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2842 @opindex Woverloaded-virtual
2843 @opindex Wno-overloaded-virtual
2844 @cindex overloaded virtual function, warning
2845 @cindex warning for overloaded virtual function
2846 Warn when a function declaration hides virtual functions from a
2847 base class. For example, in:
2854 struct B: public A @{
2859 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2870 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2871 @opindex Wno-pmf-conversions
2872 @opindex Wpmf-conversions
2873 Disable the diagnostic for converting a bound pointer to member function
2876 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2877 @opindex Wsign-promo
2878 @opindex Wno-sign-promo
2879 Warn when overload resolution chooses a promotion from unsigned or
2880 enumerated type to a signed type, over a conversion to an unsigned type of
2881 the same size. Previous versions of G++ tried to preserve
2882 unsignedness, but the standard mandates the current behavior.
2885 @node Objective-C and Objective-C++ Dialect Options
2886 @section Options Controlling Objective-C and Objective-C++ Dialects
2888 @cindex compiler options, Objective-C and Objective-C++
2889 @cindex Objective-C and Objective-C++ options, command-line
2890 @cindex options, Objective-C and Objective-C++
2891 (NOTE: This manual does not describe the Objective-C and Objective-C++
2892 languages themselves. @xref{Standards,,Language Standards
2893 Supported by GCC}, for references.)
2895 This section describes the command-line options that are only meaningful
2896 for Objective-C and Objective-C++ programs. You can also use most of
2897 the language-independent GNU compiler options.
2898 For example, you might compile a file @file{some_class.m} like this:
2901 gcc -g -fgnu-runtime -O -c some_class.m
2905 In this example, @option{-fgnu-runtime} is an option meant only for
2906 Objective-C and Objective-C++ programs; you can use the other options with
2907 any language supported by GCC@.
2909 Note that since Objective-C is an extension of the C language, Objective-C
2910 compilations may also use options specific to the C front-end (e.g.,
2911 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2912 C++-specific options (e.g., @option{-Wabi}).
2914 Here is a list of options that are @emph{only} for compiling Objective-C
2915 and Objective-C++ programs:
2918 @item -fconstant-string-class=@var{class-name}
2919 @opindex fconstant-string-class
2920 Use @var{class-name} as the name of the class to instantiate for each
2921 literal string specified with the syntax @code{@@"@dots{}"}. The default
2922 class name is @code{NXConstantString} if the GNU runtime is being used, and
2923 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2924 @option{-fconstant-cfstrings} option, if also present, overrides the
2925 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2926 to be laid out as constant CoreFoundation strings.
2929 @opindex fgnu-runtime
2930 Generate object code compatible with the standard GNU Objective-C
2931 runtime. This is the default for most types of systems.
2933 @item -fnext-runtime
2934 @opindex fnext-runtime
2935 Generate output compatible with the NeXT runtime. This is the default
2936 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2937 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2940 @item -fno-nil-receivers
2941 @opindex fno-nil-receivers
2942 Assume that all Objective-C message dispatches (@code{[receiver
2943 message:arg]}) in this translation unit ensure that the receiver is
2944 not @code{nil}. This allows for more efficient entry points in the
2945 runtime to be used. This option is only available in conjunction with
2946 the NeXT runtime and ABI version 0 or 1.
2948 @item -fobjc-abi-version=@var{n}
2949 @opindex fobjc-abi-version
2950 Use version @var{n} of the Objective-C ABI for the selected runtime.
2951 This option is currently supported only for the NeXT runtime. In that
2952 case, Version 0 is the traditional (32-bit) ABI without support for
2953 properties and other Objective-C 2.0 additions. Version 1 is the
2954 traditional (32-bit) ABI with support for properties and other
2955 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2956 nothing is specified, the default is Version 0 on 32-bit target
2957 machines, and Version 2 on 64-bit target machines.
2959 @item -fobjc-call-cxx-cdtors
2960 @opindex fobjc-call-cxx-cdtors
2961 For each Objective-C class, check if any of its instance variables is a
2962 C++ object with a non-trivial default constructor. If so, synthesize a
2963 special @code{- (id) .cxx_construct} instance method which runs
2964 non-trivial default constructors on any such instance variables, in order,
2965 and then return @code{self}. Similarly, check if any instance variable
2966 is a C++ object with a non-trivial destructor, and if so, synthesize a
2967 special @code{- (void) .cxx_destruct} method which runs
2968 all such default destructors, in reverse order.
2970 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2971 methods thusly generated only operate on instance variables
2972 declared in the current Objective-C class, and not those inherited
2973 from superclasses. It is the responsibility of the Objective-C
2974 runtime to invoke all such methods in an object's inheritance
2975 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
2976 by the runtime immediately after a new object instance is allocated;
2977 the @code{- (void) .cxx_destruct} methods are invoked immediately
2978 before the runtime deallocates an object instance.
2980 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2981 support for invoking the @code{- (id) .cxx_construct} and
2982 @code{- (void) .cxx_destruct} methods.
2984 @item -fobjc-direct-dispatch
2985 @opindex fobjc-direct-dispatch
2986 Allow fast jumps to the message dispatcher. On Darwin this is
2987 accomplished via the comm page.
2989 @item -fobjc-exceptions
2990 @opindex fobjc-exceptions
2991 Enable syntactic support for structured exception handling in
2992 Objective-C, similar to what is offered by C++ and Java. This option
2993 is required to use the Objective-C keywords @code{@@try},
2994 @code{@@throw}, @code{@@catch}, @code{@@finally} and
2995 @code{@@synchronized}. This option is available with both the GNU
2996 runtime and the NeXT runtime (but not available in conjunction with
2997 the NeXT runtime on Mac OS X 10.2 and earlier).
3001 Enable garbage collection (GC) in Objective-C and Objective-C++
3002 programs. This option is only available with the NeXT runtime; the
3003 GNU runtime has a different garbage collection implementation that
3004 does not require special compiler flags.
3006 @item -fobjc-nilcheck
3007 @opindex fobjc-nilcheck
3008 For the NeXT runtime with version 2 of the ABI, check for a nil
3009 receiver in method invocations before doing the actual method call.
3010 This is the default and can be disabled using
3011 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3012 checked for nil in this way no matter what this flag is set to.
3013 Currently this flag does nothing when the GNU runtime, or an older
3014 version of the NeXT runtime ABI, is used.
3016 @item -fobjc-std=objc1
3018 Conform to the language syntax of Objective-C 1.0, the language
3019 recognized by GCC 4.0. This only affects the Objective-C additions to
3020 the C/C++ language; it does not affect conformance to C/C++ standards,
3021 which is controlled by the separate C/C++ dialect option flags. When
3022 this option is used with the Objective-C or Objective-C++ compiler,
3023 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3024 This is useful if you need to make sure that your Objective-C code can
3025 be compiled with older versions of GCC@.
3027 @item -freplace-objc-classes
3028 @opindex freplace-objc-classes
3029 Emit a special marker instructing @command{ld(1)} not to statically link in
3030 the resulting object file, and allow @command{dyld(1)} to load it in at
3031 run time instead. This is used in conjunction with the Fix-and-Continue
3032 debugging mode, where the object file in question may be recompiled and
3033 dynamically reloaded in the course of program execution, without the need
3034 to restart the program itself. Currently, Fix-and-Continue functionality
3035 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3040 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3041 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3042 compile time) with static class references that get initialized at load time,
3043 which improves run-time performance. Specifying the @option{-fzero-link} flag
3044 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3045 to be retained. This is useful in Zero-Link debugging mode, since it allows
3046 for individual class implementations to be modified during program execution.
3047 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3048 regardless of command-line options.
3050 @item -fno-local-ivars
3051 @opindex fno-local-ivars
3052 @opindex flocal-ivars
3053 By default instance variables in Objective-C can be accessed as if
3054 they were local variables from within the methods of the class they're
3055 declared in. This can lead to shadowing between instance variables
3056 and other variables declared either locally inside a class method or
3057 globally with the same name. Specifying the @option{-fno-local-ivars}
3058 flag disables this behavior thus avoiding variable shadowing issues.
3060 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3061 @opindex fivar-visibility
3062 Set the default instance variable visibility to the specified option
3063 so that instance variables declared outside the scope of any access
3064 modifier directives default to the specified visibility.
3068 Dump interface declarations for all classes seen in the source file to a
3069 file named @file{@var{sourcename}.decl}.
3071 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3072 @opindex Wassign-intercept
3073 @opindex Wno-assign-intercept
3074 Warn whenever an Objective-C assignment is being intercepted by the
3077 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3078 @opindex Wno-protocol
3080 If a class is declared to implement a protocol, a warning is issued for
3081 every method in the protocol that is not implemented by the class. The
3082 default behavior is to issue a warning for every method not explicitly
3083 implemented in the class, even if a method implementation is inherited
3084 from the superclass. If you use the @option{-Wno-protocol} option, then
3085 methods inherited from the superclass are considered to be implemented,
3086 and no warning is issued for them.
3088 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3090 @opindex Wno-selector
3091 Warn if multiple methods of different types for the same selector are
3092 found during compilation. The check is performed on the list of methods
3093 in the final stage of compilation. Additionally, a check is performed
3094 for each selector appearing in a @code{@@selector(@dots{})}
3095 expression, and a corresponding method for that selector has been found
3096 during compilation. Because these checks scan the method table only at
3097 the end of compilation, these warnings are not produced if the final
3098 stage of compilation is not reached, for example because an error is
3099 found during compilation, or because the @option{-fsyntax-only} option is
3102 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3103 @opindex Wstrict-selector-match
3104 @opindex Wno-strict-selector-match
3105 Warn if multiple methods with differing argument and/or return types are
3106 found for a given selector when attempting to send a message using this
3107 selector to a receiver of type @code{id} or @code{Class}. When this flag
3108 is off (which is the default behavior), the compiler omits such warnings
3109 if any differences found are confined to types that share the same size
3112 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3113 @opindex Wundeclared-selector
3114 @opindex Wno-undeclared-selector
3115 Warn if a @code{@@selector(@dots{})} expression referring to an
3116 undeclared selector is found. A selector is considered undeclared if no
3117 method with that name has been declared before the
3118 @code{@@selector(@dots{})} expression, either explicitly in an
3119 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3120 an @code{@@implementation} section. This option always performs its
3121 checks as soon as a @code{@@selector(@dots{})} expression is found,
3122 while @option{-Wselector} only performs its checks in the final stage of
3123 compilation. This also enforces the coding style convention
3124 that methods and selectors must be declared before being used.
3126 @item -print-objc-runtime-info
3127 @opindex print-objc-runtime-info
3128 Generate C header describing the largest structure that is passed by
3133 @node Language Independent Options
3134 @section Options to Control Diagnostic Messages Formatting
3135 @cindex options to control diagnostics formatting
3136 @cindex diagnostic messages
3137 @cindex message formatting
3139 Traditionally, diagnostic messages have been formatted irrespective of
3140 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3141 options described below
3142 to control the formatting algorithm for diagnostic messages,
3143 e.g.@: how many characters per line, how often source location
3144 information should be reported. Note that some language front ends may not
3145 honor these options.
3148 @item -fmessage-length=@var{n}
3149 @opindex fmessage-length
3150 Try to format error messages so that they fit on lines of about
3151 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3152 done; each error message appears on a single line. This is the
3153 default for all front ends.
3155 @item -fdiagnostics-show-location=once
3156 @opindex fdiagnostics-show-location
3157 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3158 reporter to emit source location information @emph{once}; that is, in
3159 case the message is too long to fit on a single physical line and has to
3160 be wrapped, the source location won't be emitted (as prefix) again,
3161 over and over, in subsequent continuation lines. This is the default
3164 @item -fdiagnostics-show-location=every-line
3165 Only meaningful in line-wrapping mode. Instructs the diagnostic
3166 messages reporter to emit the same source location information (as
3167 prefix) for physical lines that result from the process of breaking
3168 a message which is too long to fit on a single line.
3170 @item -fdiagnostics-color[=@var{WHEN}]
3171 @itemx -fno-diagnostics-color
3172 @opindex fdiagnostics-color
3173 @cindex highlight, color, colour
3174 @vindex GCC_COLORS @r{environment variable}
3175 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3176 or @samp{auto}. The default depends on how the compiler has been configured,
3177 it can be any of the above @var{WHEN} options or also @samp{never}
3178 if @env{GCC_COLORS} environment variable isn't present in the environment,
3179 and @samp{auto} otherwise.
3180 @samp{auto} means to use color only when the standard error is a terminal.
3181 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3182 aliases for @option{-fdiagnostics-color=always} and
3183 @option{-fdiagnostics-color=never}, respectively.
3185 The colors are defined by the environment variable @env{GCC_COLORS}.
3186 Its value is a colon-separated list of capabilities and Select Graphic
3187 Rendition (SGR) substrings. SGR commands are interpreted by the
3188 terminal or terminal emulator. (See the section in the documentation
3189 of your text terminal for permitted values and their meanings as
3190 character attributes.) These substring values are integers in decimal
3191 representation and can be concatenated with semicolons.
3192 Common values to concatenate include
3194 @samp{4} for underline,
3196 @samp{7} for inverse,
3197 @samp{39} for default foreground color,
3198 @samp{30} to @samp{37} for foreground colors,
3199 @samp{90} to @samp{97} for 16-color mode foreground colors,
3200 @samp{38;5;0} to @samp{38;5;255}
3201 for 88-color and 256-color modes foreground colors,
3202 @samp{49} for default background color,
3203 @samp{40} to @samp{47} for background colors,
3204 @samp{100} to @samp{107} for 16-color mode background colors,
3205 and @samp{48;5;0} to @samp{48;5;255}
3206 for 88-color and 256-color modes background colors.
3208 The default @env{GCC_COLORS} is
3210 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3213 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3214 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3215 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3216 string disables colors.
3217 Supported capabilities are as follows.
3221 @vindex error GCC_COLORS @r{capability}
3222 SGR substring for error: markers.
3225 @vindex warning GCC_COLORS @r{capability}
3226 SGR substring for warning: markers.
3229 @vindex note GCC_COLORS @r{capability}
3230 SGR substring for note: markers.
3233 @vindex caret GCC_COLORS @r{capability}
3234 SGR substring for caret line.
3237 @vindex locus GCC_COLORS @r{capability}
3238 SGR substring for location information, @samp{file:line} or
3239 @samp{file:line:column} etc.
3242 @vindex quote GCC_COLORS @r{capability}
3243 SGR substring for information printed within quotes.
3246 @item -fno-diagnostics-show-option
3247 @opindex fno-diagnostics-show-option
3248 @opindex fdiagnostics-show-option
3249 By default, each diagnostic emitted includes text indicating the
3250 command-line option that directly controls the diagnostic (if such an
3251 option is known to the diagnostic machinery). Specifying the
3252 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3254 @item -fno-diagnostics-show-caret
3255 @opindex fno-diagnostics-show-caret
3256 @opindex fdiagnostics-show-caret
3257 By default, each diagnostic emitted includes the original source line
3258 and a caret '^' indicating the column. This option suppresses this
3259 information. The source line is truncated to @var{n} characters, if
3260 the @option{-fmessage-length=n} option is given. When the output is done
3261 to the terminal, the width is limited to the width given by the
3262 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3266 @node Warning Options
3267 @section Options to Request or Suppress Warnings
3268 @cindex options to control warnings
3269 @cindex warning messages
3270 @cindex messages, warning
3271 @cindex suppressing warnings
3273 Warnings are diagnostic messages that report constructions that
3274 are not inherently erroneous but that are risky or suggest there
3275 may have been an error.
3277 The following language-independent options do not enable specific
3278 warnings but control the kinds of diagnostics produced by GCC@.
3281 @cindex syntax checking
3283 @opindex fsyntax-only
3284 Check the code for syntax errors, but don't do anything beyond that.
3286 @item -fmax-errors=@var{n}
3287 @opindex fmax-errors
3288 Limits the maximum number of error messages to @var{n}, at which point
3289 GCC bails out rather than attempting to continue processing the source
3290 code. If @var{n} is 0 (the default), there is no limit on the number
3291 of error messages produced. If @option{-Wfatal-errors} is also
3292 specified, then @option{-Wfatal-errors} takes precedence over this
3297 Inhibit all warning messages.
3302 Make all warnings into errors.
3307 Make the specified warning into an error. The specifier for a warning
3308 is appended; for example @option{-Werror=switch} turns the warnings
3309 controlled by @option{-Wswitch} into errors. This switch takes a
3310 negative form, to be used to negate @option{-Werror} for specific
3311 warnings; for example @option{-Wno-error=switch} makes
3312 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3315 The warning message for each controllable warning includes the
3316 option that controls the warning. That option can then be used with
3317 @option{-Werror=} and @option{-Wno-error=} as described above.
3318 (Printing of the option in the warning message can be disabled using the
3319 @option{-fno-diagnostics-show-option} flag.)
3321 Note that specifying @option{-Werror=}@var{foo} automatically implies
3322 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3325 @item -Wfatal-errors
3326 @opindex Wfatal-errors
3327 @opindex Wno-fatal-errors
3328 This option causes the compiler to abort compilation on the first error
3329 occurred rather than trying to keep going and printing further error
3334 You can request many specific warnings with options beginning with
3335 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3336 implicit declarations. Each of these specific warning options also
3337 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3338 example, @option{-Wno-implicit}. This manual lists only one of the
3339 two forms, whichever is not the default. For further
3340 language-specific options also refer to @ref{C++ Dialect Options} and
3341 @ref{Objective-C and Objective-C++ Dialect Options}.
3343 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3344 options, such as @option{-Wunused}, which may turn on further options,
3345 such as @option{-Wunused-value}. The combined effect of positive and
3346 negative forms is that more specific options have priority over less
3347 specific ones, independently of their position in the command-line. For
3348 options of the same specificity, the last one takes effect. Options
3349 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3350 as if they appeared at the end of the command-line.
3352 When an unrecognized warning option is requested (e.g.,
3353 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3354 that the option is not recognized. However, if the @option{-Wno-} form
3355 is used, the behavior is slightly different: no diagnostic is
3356 produced for @option{-Wno-unknown-warning} unless other diagnostics
3357 are being produced. This allows the use of new @option{-Wno-} options
3358 with old compilers, but if something goes wrong, the compiler
3359 warns that an unrecognized option is present.
3366 Issue all the warnings demanded by strict ISO C and ISO C++;
3367 reject all programs that use forbidden extensions, and some other
3368 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3369 version of the ISO C standard specified by any @option{-std} option used.
3371 Valid ISO C and ISO C++ programs should compile properly with or without
3372 this option (though a rare few require @option{-ansi} or a
3373 @option{-std} option specifying the required version of ISO C)@. However,
3374 without this option, certain GNU extensions and traditional C and C++
3375 features are supported as well. With this option, they are rejected.
3377 @option{-Wpedantic} does not cause warning messages for use of the
3378 alternate keywords whose names begin and end with @samp{__}. Pedantic
3379 warnings are also disabled in the expression that follows
3380 @code{__extension__}. However, only system header files should use
3381 these escape routes; application programs should avoid them.
3382 @xref{Alternate Keywords}.
3384 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3385 C conformance. They soon find that it does not do quite what they want:
3386 it finds some non-ISO practices, but not all---only those for which
3387 ISO C @emph{requires} a diagnostic, and some others for which
3388 diagnostics have been added.
3390 A feature to report any failure to conform to ISO C might be useful in
3391 some instances, but would require considerable additional work and would
3392 be quite different from @option{-Wpedantic}. We don't have plans to
3393 support such a feature in the near future.
3395 Where the standard specified with @option{-std} represents a GNU
3396 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3397 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3398 extended dialect is based. Warnings from @option{-Wpedantic} are given
3399 where they are required by the base standard. (It does not make sense
3400 for such warnings to be given only for features not in the specified GNU
3401 C dialect, since by definition the GNU dialects of C include all
3402 features the compiler supports with the given option, and there would be
3403 nothing to warn about.)
3405 @item -pedantic-errors
3406 @opindex pedantic-errors
3407 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3408 requires a diagnostic, in some cases where there is undefined behavior
3409 at compile-time and in some other cases that do not prevent compilation
3410 of programs that are valid according to the standard. This is not
3411 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3412 by this option and not enabled by the latter and vice versa.
3417 This enables all the warnings about constructions that some users
3418 consider questionable, and that are easy to avoid (or modify to
3419 prevent the warning), even in conjunction with macros. This also
3420 enables some language-specific warnings described in @ref{C++ Dialect
3421 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3423 @option{-Wall} turns on the following warning flags:
3425 @gccoptlist{-Waddress @gol
3426 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3427 -Wc++11-compat -Wc++14-compat@gol
3428 -Wchar-subscripts @gol
3429 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3430 -Wimplicit-int @r{(C and Objective-C only)} @gol
3431 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3434 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3435 -Wmaybe-uninitialized @gol
3436 -Wmissing-braces @r{(only for C/ObjC)} @gol
3443 -Wsequence-point @gol
3444 -Wsign-compare @r{(only in C++)} @gol
3445 -Wstrict-aliasing @gol
3446 -Wstrict-overflow=1 @gol
3449 -Wuninitialized @gol
3450 -Wunknown-pragmas @gol
3451 -Wunused-function @gol
3454 -Wunused-variable @gol
3455 -Wvolatile-register-var @gol
3458 Note that some warning flags are not implied by @option{-Wall}. Some of
3459 them warn about constructions that users generally do not consider
3460 questionable, but which occasionally you might wish to check for;
3461 others warn about constructions that are necessary or hard to avoid in
3462 some cases, and there is no simple way to modify the code to suppress
3463 the warning. Some of them are enabled by @option{-Wextra} but many of
3464 them must be enabled individually.
3470 This enables some extra warning flags that are not enabled by
3471 @option{-Wall}. (This option used to be called @option{-W}. The older
3472 name is still supported, but the newer name is more descriptive.)
3474 @gccoptlist{-Wclobbered @gol
3476 -Wignored-qualifiers @gol
3477 -Wmissing-field-initializers @gol
3478 -Wmissing-parameter-type @r{(C only)} @gol
3479 -Wold-style-declaration @r{(C only)} @gol
3480 -Woverride-init @gol
3483 -Wuninitialized @gol
3484 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3485 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3488 The option @option{-Wextra} also prints warning messages for the
3494 A pointer is compared against integer zero with @code{<}, @code{<=},
3495 @code{>}, or @code{>=}.
3498 (C++ only) An enumerator and a non-enumerator both appear in a
3499 conditional expression.
3502 (C++ only) Ambiguous virtual bases.
3505 (C++ only) Subscripting an array that has been declared @code{register}.
3508 (C++ only) Taking the address of a variable that has been declared
3512 (C++ only) A base class is not initialized in a derived class's copy
3517 @item -Wchar-subscripts
3518 @opindex Wchar-subscripts
3519 @opindex Wno-char-subscripts
3520 Warn if an array subscript has type @code{char}. This is a common cause
3521 of error, as programmers often forget that this type is signed on some
3523 This warning is enabled by @option{-Wall}.
3527 @opindex Wno-comment
3528 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3529 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3530 This warning is enabled by @option{-Wall}.
3532 @item -Wno-coverage-mismatch
3533 @opindex Wno-coverage-mismatch
3534 Warn if feedback profiles do not match when using the
3535 @option{-fprofile-use} option.
3536 If a source file is changed between compiling with @option{-fprofile-gen} and
3537 with @option{-fprofile-use}, the files with the profile feedback can fail
3538 to match the source file and GCC cannot use the profile feedback
3539 information. By default, this warning is enabled and is treated as an
3540 error. @option{-Wno-coverage-mismatch} can be used to disable the
3541 warning or @option{-Wno-error=coverage-mismatch} can be used to
3542 disable the error. Disabling the error for this warning can result in
3543 poorly optimized code and is useful only in the
3544 case of very minor changes such as bug fixes to an existing code-base.
3545 Completely disabling the warning is not recommended.
3548 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3550 Suppress warning messages emitted by @code{#warning} directives.
3552 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3553 @opindex Wdouble-promotion
3554 @opindex Wno-double-promotion
3555 Give a warning when a value of type @code{float} is implicitly
3556 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3557 floating-point unit implement @code{float} in hardware, but emulate
3558 @code{double} in software. On such a machine, doing computations
3559 using @code{double} values is much more expensive because of the
3560 overhead required for software emulation.
3562 It is easy to accidentally do computations with @code{double} because
3563 floating-point literals are implicitly of type @code{double}. For
3567 float area(float radius)
3569 return 3.14159 * radius * radius;
3573 the compiler performs the entire computation with @code{double}
3574 because the floating-point literal is a @code{double}.
3577 @itemx -Wformat=@var{n}
3580 @opindex ffreestanding
3581 @opindex fno-builtin
3583 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3584 the arguments supplied have types appropriate to the format string
3585 specified, and that the conversions specified in the format string make
3586 sense. This includes standard functions, and others specified by format
3587 attributes (@pxref{Function Attributes}), in the @code{printf},
3588 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3589 not in the C standard) families (or other target-specific families).
3590 Which functions are checked without format attributes having been
3591 specified depends on the standard version selected, and such checks of
3592 functions without the attribute specified are disabled by
3593 @option{-ffreestanding} or @option{-fno-builtin}.
3595 The formats are checked against the format features supported by GNU
3596 libc version 2.2. These include all ISO C90 and C99 features, as well
3597 as features from the Single Unix Specification and some BSD and GNU
3598 extensions. Other library implementations may not support all these
3599 features; GCC does not support warning about features that go beyond a
3600 particular library's limitations. However, if @option{-Wpedantic} is used
3601 with @option{-Wformat}, warnings are given about format features not
3602 in the selected standard version (but not for @code{strfmon} formats,
3603 since those are not in any version of the C standard). @xref{C Dialect
3604 Options,,Options Controlling C Dialect}.
3611 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3612 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3613 @option{-Wformat} also checks for null format arguments for several
3614 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3615 aspects of this level of format checking can be disabled by the
3616 options: @option{-Wno-format-contains-nul},
3617 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3618 @option{-Wformat} is enabled by @option{-Wall}.
3620 @item -Wno-format-contains-nul
3621 @opindex Wno-format-contains-nul
3622 @opindex Wformat-contains-nul
3623 If @option{-Wformat} is specified, do not warn about format strings that
3626 @item -Wno-format-extra-args
3627 @opindex Wno-format-extra-args
3628 @opindex Wformat-extra-args
3629 If @option{-Wformat} is specified, do not warn about excess arguments to a
3630 @code{printf} or @code{scanf} format function. The C standard specifies
3631 that such arguments are ignored.
3633 Where the unused arguments lie between used arguments that are
3634 specified with @samp{$} operand number specifications, normally
3635 warnings are still given, since the implementation could not know what
3636 type to pass to @code{va_arg} to skip the unused arguments. However,
3637 in the case of @code{scanf} formats, this option suppresses the
3638 warning if the unused arguments are all pointers, since the Single
3639 Unix Specification says that such unused arguments are allowed.
3641 @item -Wno-format-zero-length
3642 @opindex Wno-format-zero-length
3643 @opindex Wformat-zero-length
3644 If @option{-Wformat} is specified, do not warn about zero-length formats.
3645 The C standard specifies that zero-length formats are allowed.
3650 Enable @option{-Wformat} plus additional format checks. Currently
3651 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3654 @item -Wformat-nonliteral
3655 @opindex Wformat-nonliteral
3656 @opindex Wno-format-nonliteral
3657 If @option{-Wformat} is specified, also warn if the format string is not a
3658 string literal and so cannot be checked, unless the format function
3659 takes its format arguments as a @code{va_list}.
3661 @item -Wformat-security
3662 @opindex Wformat-security
3663 @opindex Wno-format-security
3664 If @option{-Wformat} is specified, also warn about uses of format
3665 functions that represent possible security problems. At present, this
3666 warns about calls to @code{printf} and @code{scanf} functions where the
3667 format string is not a string literal and there are no format arguments,
3668 as in @code{printf (foo);}. This may be a security hole if the format
3669 string came from untrusted input and contains @samp{%n}. (This is
3670 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3671 in future warnings may be added to @option{-Wformat-security} that are not
3672 included in @option{-Wformat-nonliteral}.)
3674 @item -Wformat-signedness
3675 @opindex Wformat-signedness
3676 @opindex Wno-format-signedness
3677 If @option{-Wformat} is specified, also warn if the format string
3678 requires an unsigned argument and the argument is signed and vice versa.
3681 @opindex Wformat-y2k
3682 @opindex Wno-format-y2k
3683 If @option{-Wformat} is specified, also warn about @code{strftime}
3684 formats that may yield only a two-digit year.
3689 @opindex Wno-nonnull
3690 Warn about passing a null pointer for arguments marked as
3691 requiring a non-null value by the @code{nonnull} function attribute.
3693 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3694 can be disabled with the @option{-Wno-nonnull} option.
3696 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3698 @opindex Wno-init-self
3699 Warn about uninitialized variables that are initialized with themselves.
3700 Note this option can only be used with the @option{-Wuninitialized} option.
3702 For example, GCC warns about @code{i} being uninitialized in the
3703 following snippet only when @option{-Winit-self} has been specified:
3714 This warning is enabled by @option{-Wall} in C++.
3716 @item -Wimplicit-int @r{(C and Objective-C only)}
3717 @opindex Wimplicit-int
3718 @opindex Wno-implicit-int
3719 Warn when a declaration does not specify a type.
3720 This warning is enabled by @option{-Wall}.
3722 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3723 @opindex Wimplicit-function-declaration
3724 @opindex Wno-implicit-function-declaration
3725 Give a warning whenever a function is used before being declared. In
3726 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3727 enabled by default and it is made into an error by
3728 @option{-pedantic-errors}. This warning is also enabled by
3731 @item -Wimplicit @r{(C and Objective-C only)}
3733 @opindex Wno-implicit
3734 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3735 This warning is enabled by @option{-Wall}.
3737 @item -Wignored-qualifiers @r{(C and C++ only)}
3738 @opindex Wignored-qualifiers
3739 @opindex Wno-ignored-qualifiers
3740 Warn if the return type of a function has a type qualifier
3741 such as @code{const}. For ISO C such a type qualifier has no effect,
3742 since the value returned by a function is not an lvalue.
3743 For C++, the warning is only emitted for scalar types or @code{void}.
3744 ISO C prohibits qualified @code{void} return types on function
3745 definitions, so such return types always receive a warning
3746 even without this option.
3748 This warning is also enabled by @option{-Wextra}.
3753 Warn if the type of @code{main} is suspicious. @code{main} should be
3754 a function with external linkage, returning int, taking either zero
3755 arguments, two, or three arguments of appropriate types. This warning
3756 is enabled by default in C++ and is enabled by either @option{-Wall}
3757 or @option{-Wpedantic}.
3759 @item -Wmissing-braces
3760 @opindex Wmissing-braces
3761 @opindex Wno-missing-braces
3762 Warn if an aggregate or union initializer is not fully bracketed. In
3763 the following example, the initializer for @code{a} is not fully
3764 bracketed, but that for @code{b} is fully bracketed. This warning is
3765 enabled by @option{-Wall} in C.
3768 int a[2][2] = @{ 0, 1, 2, 3 @};
3769 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3772 This warning is enabled by @option{-Wall}.
3774 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3775 @opindex Wmissing-include-dirs
3776 @opindex Wno-missing-include-dirs
3777 Warn if a user-supplied include directory does not exist.
3780 @opindex Wparentheses
3781 @opindex Wno-parentheses
3782 Warn if parentheses are omitted in certain contexts, such
3783 as when there is an assignment in a context where a truth value
3784 is expected, or when operators are nested whose precedence people
3785 often get confused about.
3787 Also warn if a comparison like @code{x<=y<=z} appears; this is
3788 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3789 interpretation from that of ordinary mathematical notation.
3791 Also warn about constructions where there may be confusion to which
3792 @code{if} statement an @code{else} branch belongs. Here is an example of
3807 In C/C++, every @code{else} branch belongs to the innermost possible
3808 @code{if} statement, which in this example is @code{if (b)}. This is
3809 often not what the programmer expected, as illustrated in the above
3810 example by indentation the programmer chose. When there is the
3811 potential for this confusion, GCC issues a warning when this flag
3812 is specified. To eliminate the warning, add explicit braces around
3813 the innermost @code{if} statement so there is no way the @code{else}
3814 can belong to the enclosing @code{if}. The resulting code
3831 Also warn for dangerous uses of the GNU extension to
3832 @code{?:} with omitted middle operand. When the condition
3833 in the @code{?}: operator is a boolean expression, the omitted value is
3834 always 1. Often programmers expect it to be a value computed
3835 inside the conditional expression instead.
3837 This warning is enabled by @option{-Wall}.
3839 @item -Wsequence-point
3840 @opindex Wsequence-point
3841 @opindex Wno-sequence-point
3842 Warn about code that may have undefined semantics because of violations
3843 of sequence point rules in the C and C++ standards.
3845 The C and C++ standards define the order in which expressions in a C/C++
3846 program are evaluated in terms of @dfn{sequence points}, which represent
3847 a partial ordering between the execution of parts of the program: those
3848 executed before the sequence point, and those executed after it. These
3849 occur after the evaluation of a full expression (one which is not part
3850 of a larger expression), after the evaluation of the first operand of a
3851 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3852 function is called (but after the evaluation of its arguments and the
3853 expression denoting the called function), and in certain other places.
3854 Other than as expressed by the sequence point rules, the order of
3855 evaluation of subexpressions of an expression is not specified. All
3856 these rules describe only a partial order rather than a total order,
3857 since, for example, if two functions are called within one expression
3858 with no sequence point between them, the order in which the functions
3859 are called is not specified. However, the standards committee have
3860 ruled that function calls do not overlap.
3862 It is not specified when between sequence points modifications to the
3863 values of objects take effect. Programs whose behavior depends on this
3864 have undefined behavior; the C and C++ standards specify that ``Between
3865 the previous and next sequence point an object shall have its stored
3866 value modified at most once by the evaluation of an expression.
3867 Furthermore, the prior value shall be read only to determine the value
3868 to be stored.''. If a program breaks these rules, the results on any
3869 particular implementation are entirely unpredictable.
3871 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3872 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3873 diagnosed by this option, and it may give an occasional false positive
3874 result, but in general it has been found fairly effective at detecting
3875 this sort of problem in programs.
3877 The standard is worded confusingly, therefore there is some debate
3878 over the precise meaning of the sequence point rules in subtle cases.
3879 Links to discussions of the problem, including proposed formal
3880 definitions, may be found on the GCC readings page, at
3881 @uref{http://gcc.gnu.org/@/readings.html}.
3883 This warning is enabled by @option{-Wall} for C and C++.
3885 @item -Wno-return-local-addr
3886 @opindex Wno-return-local-addr
3887 @opindex Wreturn-local-addr
3888 Do not warn about returning a pointer (or in C++, a reference) to a
3889 variable that goes out of scope after the function returns.
3892 @opindex Wreturn-type
3893 @opindex Wno-return-type
3894 Warn whenever a function is defined with a return type that defaults
3895 to @code{int}. Also warn about any @code{return} statement with no
3896 return value in a function whose return type is not @code{void}
3897 (falling off the end of the function body is considered returning
3898 without a value), and about a @code{return} statement with an
3899 expression in a function whose return type is @code{void}.
3901 For C++, a function without return type always produces a diagnostic
3902 message, even when @option{-Wno-return-type} is specified. The only
3903 exceptions are @code{main} and functions defined in system headers.
3905 This warning is enabled by @option{-Wall}.
3907 @item -Wshift-count-negative
3908 @opindex Wshift-count-negative
3909 @opindex Wno-shift-count-negative
3910 Warn if shift count is negative. This warning is enabled by default.
3912 @item -Wshift-count-overflow
3913 @opindex Wshift-count-overflow
3914 @opindex Wno-shift-count-overflow
3915 Warn if shift count >= width of type. This warning is enabled by default.
3920 Warn whenever a @code{switch} statement has an index of enumerated type
3921 and lacks a @code{case} for one or more of the named codes of that
3922 enumeration. (The presence of a @code{default} label prevents this
3923 warning.) @code{case} labels outside the enumeration range also
3924 provoke warnings when this option is used (even if there is a
3925 @code{default} label).
3926 This warning is enabled by @option{-Wall}.
3928 @item -Wswitch-default
3929 @opindex Wswitch-default
3930 @opindex Wno-switch-default
3931 Warn whenever a @code{switch} statement does not have a @code{default}
3935 @opindex Wswitch-enum
3936 @opindex Wno-switch-enum
3937 Warn whenever a @code{switch} statement has an index of enumerated type
3938 and lacks a @code{case} for one or more of the named codes of that
3939 enumeration. @code{case} labels outside the enumeration range also
3940 provoke warnings when this option is used. The only difference
3941 between @option{-Wswitch} and this option is that this option gives a
3942 warning about an omitted enumeration code even if there is a
3943 @code{default} label.
3946 @opindex Wswitch-bool
3947 @opindex Wno-switch-bool
3948 Warn whenever a @code{switch} statement has an index of boolean type.
3949 It is possible to suppress this warning by casting the controlling
3950 expression to a type other than @code{bool}. For example:
3953 switch ((int) (a == 4))
3959 This warning is enabled by default for C and C++ programs.
3961 @item -Wsync-nand @r{(C and C++ only)}
3963 @opindex Wno-sync-nand
3964 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
3965 built-in functions are used. These functions changed semantics in GCC 4.4.
3969 @opindex Wno-trigraphs
3970 Warn if any trigraphs are encountered that might change the meaning of
3971 the program (trigraphs within comments are not warned about).
3972 This warning is enabled by @option{-Wall}.
3974 @item -Wunused-but-set-parameter
3975 @opindex Wunused-but-set-parameter
3976 @opindex Wno-unused-but-set-parameter
3977 Warn whenever a function parameter is assigned to, but otherwise unused
3978 (aside from its declaration).
3980 To suppress this warning use the @code{unused} attribute
3981 (@pxref{Variable Attributes}).
3983 This warning is also enabled by @option{-Wunused} together with
3986 @item -Wunused-but-set-variable
3987 @opindex Wunused-but-set-variable
3988 @opindex Wno-unused-but-set-variable
3989 Warn whenever a local variable is assigned to, but otherwise unused
3990 (aside from its declaration).
3991 This warning is enabled by @option{-Wall}.
3993 To suppress this warning use the @code{unused} attribute
3994 (@pxref{Variable Attributes}).
3996 This warning is also enabled by @option{-Wunused}, which is enabled
3999 @item -Wunused-function
4000 @opindex Wunused-function
4001 @opindex Wno-unused-function
4002 Warn whenever a static function is declared but not defined or a
4003 non-inline static function is unused.
4004 This warning is enabled by @option{-Wall}.
4006 @item -Wunused-label
4007 @opindex Wunused-label
4008 @opindex Wno-unused-label
4009 Warn whenever a label is declared but not used.
4010 This warning is enabled by @option{-Wall}.
4012 To suppress this warning use the @code{unused} attribute
4013 (@pxref{Variable Attributes}).
4015 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4016 @opindex Wunused-local-typedefs
4017 Warn when a typedef locally defined in a function is not used.
4018 This warning is enabled by @option{-Wall}.
4020 @item -Wunused-parameter
4021 @opindex Wunused-parameter
4022 @opindex Wno-unused-parameter
4023 Warn whenever a function parameter is unused aside from its declaration.
4025 To suppress this warning use the @code{unused} attribute
4026 (@pxref{Variable Attributes}).
4028 @item -Wno-unused-result
4029 @opindex Wunused-result
4030 @opindex Wno-unused-result
4031 Do not warn if a caller of a function marked with attribute
4032 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4033 its return value. The default is @option{-Wunused-result}.
4035 @item -Wunused-variable
4036 @opindex Wunused-variable
4037 @opindex Wno-unused-variable
4038 Warn whenever a local variable or non-constant static variable is unused
4039 aside from its declaration.
4040 This warning is enabled by @option{-Wall}.
4042 To suppress this warning use the @code{unused} attribute
4043 (@pxref{Variable Attributes}).
4045 @item -Wunused-value
4046 @opindex Wunused-value
4047 @opindex Wno-unused-value
4048 Warn whenever a statement computes a result that is explicitly not
4049 used. To suppress this warning cast the unused expression to
4050 @code{void}. This includes an expression-statement or the left-hand
4051 side of a comma expression that contains no side effects. For example,
4052 an expression such as @code{x[i,j]} causes a warning, while
4053 @code{x[(void)i,j]} does not.
4055 This warning is enabled by @option{-Wall}.
4060 All the above @option{-Wunused} options combined.
4062 In order to get a warning about an unused function parameter, you must
4063 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4064 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4066 @item -Wuninitialized
4067 @opindex Wuninitialized
4068 @opindex Wno-uninitialized
4069 Warn if an automatic variable is used without first being initialized
4070 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4071 warn if a non-static reference or non-static @code{const} member
4072 appears in a class without constructors.
4074 If you want to warn about code that uses the uninitialized value of the
4075 variable in its own initializer, use the @option{-Winit-self} option.
4077 These warnings occur for individual uninitialized or clobbered
4078 elements of structure, union or array variables as well as for
4079 variables that are uninitialized or clobbered as a whole. They do
4080 not occur for variables or elements declared @code{volatile}. Because
4081 these warnings depend on optimization, the exact variables or elements
4082 for which there are warnings depends on the precise optimization
4083 options and version of GCC used.
4085 Note that there may be no warning about a variable that is used only
4086 to compute a value that itself is never used, because such
4087 computations may be deleted by data flow analysis before the warnings
4090 @item -Wmaybe-uninitialized
4091 @opindex Wmaybe-uninitialized
4092 @opindex Wno-maybe-uninitialized
4093 For an automatic variable, if there exists a path from the function
4094 entry to a use of the variable that is initialized, but there exist
4095 some other paths for which the variable is not initialized, the compiler
4096 emits a warning if it cannot prove the uninitialized paths are not
4097 executed at run time. These warnings are made optional because GCC is
4098 not smart enough to see all the reasons why the code might be correct
4099 in spite of appearing to have an error. Here is one example of how
4120 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4121 always initialized, but GCC doesn't know this. To suppress the
4122 warning, you need to provide a default case with assert(0) or
4125 @cindex @code{longjmp} warnings
4126 This option also warns when a non-volatile automatic variable might be
4127 changed by a call to @code{longjmp}. These warnings as well are possible
4128 only in optimizing compilation.
4130 The compiler sees only the calls to @code{setjmp}. It cannot know
4131 where @code{longjmp} will be called; in fact, a signal handler could
4132 call it at any point in the code. As a result, you may get a warning
4133 even when there is in fact no problem because @code{longjmp} cannot
4134 in fact be called at the place that would cause a problem.
4136 Some spurious warnings can be avoided if you declare all the functions
4137 you use that never return as @code{noreturn}. @xref{Function
4140 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4142 @item -Wunknown-pragmas
4143 @opindex Wunknown-pragmas
4144 @opindex Wno-unknown-pragmas
4145 @cindex warning for unknown pragmas
4146 @cindex unknown pragmas, warning
4147 @cindex pragmas, warning of unknown
4148 Warn when a @code{#pragma} directive is encountered that is not understood by
4149 GCC@. If this command-line option is used, warnings are even issued
4150 for unknown pragmas in system header files. This is not the case if
4151 the warnings are only enabled by the @option{-Wall} command-line option.
4154 @opindex Wno-pragmas
4156 Do not warn about misuses of pragmas, such as incorrect parameters,
4157 invalid syntax, or conflicts between pragmas. See also
4158 @option{-Wunknown-pragmas}.
4160 @item -Wstrict-aliasing
4161 @opindex Wstrict-aliasing
4162 @opindex Wno-strict-aliasing
4163 This option is only active when @option{-fstrict-aliasing} is active.
4164 It warns about code that might break the strict aliasing rules that the
4165 compiler is using for optimization. The warning does not catch all
4166 cases, but does attempt to catch the more common pitfalls. It is
4167 included in @option{-Wall}.
4168 It is equivalent to @option{-Wstrict-aliasing=3}
4170 @item -Wstrict-aliasing=n
4171 @opindex Wstrict-aliasing=n
4172 This option is only active when @option{-fstrict-aliasing} is active.
4173 It warns about code that might break the strict aliasing rules that the
4174 compiler is using for optimization.
4175 Higher levels correspond to higher accuracy (fewer false positives).
4176 Higher levels also correspond to more effort, similar to the way @option{-O}
4178 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4180 Level 1: Most aggressive, quick, least accurate.
4181 Possibly useful when higher levels
4182 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4183 false negatives. However, it has many false positives.
4184 Warns for all pointer conversions between possibly incompatible types,
4185 even if never dereferenced. Runs in the front end only.
4187 Level 2: Aggressive, quick, not too precise.
4188 May still have many false positives (not as many as level 1 though),
4189 and few false negatives (but possibly more than level 1).
4190 Unlike level 1, it only warns when an address is taken. Warns about
4191 incomplete types. Runs in the front end only.
4193 Level 3 (default for @option{-Wstrict-aliasing}):
4194 Should have very few false positives and few false
4195 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4196 Takes care of the common pun+dereference pattern in the front end:
4197 @code{*(int*)&some_float}.
4198 If optimization is enabled, it also runs in the back end, where it deals
4199 with multiple statement cases using flow-sensitive points-to information.
4200 Only warns when the converted pointer is dereferenced.
4201 Does not warn about incomplete types.
4203 @item -Wstrict-overflow
4204 @itemx -Wstrict-overflow=@var{n}
4205 @opindex Wstrict-overflow
4206 @opindex Wno-strict-overflow
4207 This option is only active when @option{-fstrict-overflow} is active.
4208 It warns about cases where the compiler optimizes based on the
4209 assumption that signed overflow does not occur. Note that it does not
4210 warn about all cases where the code might overflow: it only warns
4211 about cases where the compiler implements some optimization. Thus
4212 this warning depends on the optimization level.
4214 An optimization that assumes that signed overflow does not occur is
4215 perfectly safe if the values of the variables involved are such that
4216 overflow never does, in fact, occur. Therefore this warning can
4217 easily give a false positive: a warning about code that is not
4218 actually a problem. To help focus on important issues, several
4219 warning levels are defined. No warnings are issued for the use of
4220 undefined signed overflow when estimating how many iterations a loop
4221 requires, in particular when determining whether a loop will be
4225 @item -Wstrict-overflow=1
4226 Warn about cases that are both questionable and easy to avoid. For
4227 example, with @option{-fstrict-overflow}, the compiler simplifies
4228 @code{x + 1 > x} to @code{1}. This level of
4229 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4230 are not, and must be explicitly requested.
4232 @item -Wstrict-overflow=2
4233 Also warn about other cases where a comparison is simplified to a
4234 constant. For example: @code{abs (x) >= 0}. This can only be
4235 simplified when @option{-fstrict-overflow} is in effect, because
4236 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4237 zero. @option{-Wstrict-overflow} (with no level) is the same as
4238 @option{-Wstrict-overflow=2}.
4240 @item -Wstrict-overflow=3
4241 Also warn about other cases where a comparison is simplified. For
4242 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4244 @item -Wstrict-overflow=4
4245 Also warn about other simplifications not covered by the above cases.
4246 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4248 @item -Wstrict-overflow=5
4249 Also warn about cases where the compiler reduces the magnitude of a
4250 constant involved in a comparison. For example: @code{x + 2 > y} is
4251 simplified to @code{x + 1 >= y}. This is reported only at the
4252 highest warning level because this simplification applies to many
4253 comparisons, so this warning level gives a very large number of
4257 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4258 @opindex Wsuggest-attribute=
4259 @opindex Wno-suggest-attribute=
4260 Warn for cases where adding an attribute may be beneficial. The
4261 attributes currently supported are listed below.
4264 @item -Wsuggest-attribute=pure
4265 @itemx -Wsuggest-attribute=const
4266 @itemx -Wsuggest-attribute=noreturn
4267 @opindex Wsuggest-attribute=pure
4268 @opindex Wno-suggest-attribute=pure
4269 @opindex Wsuggest-attribute=const
4270 @opindex Wno-suggest-attribute=const
4271 @opindex Wsuggest-attribute=noreturn
4272 @opindex Wno-suggest-attribute=noreturn
4274 Warn about functions that might be candidates for attributes
4275 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4276 functions visible in other compilation units or (in the case of @code{pure} and
4277 @code{const}) if it cannot prove that the function returns normally. A function
4278 returns normally if it doesn't contain an infinite loop or return abnormally
4279 by throwing, calling @code{abort} or trapping. This analysis requires option
4280 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4281 higher. Higher optimization levels improve the accuracy of the analysis.
4283 @item -Wsuggest-attribute=format
4284 @itemx -Wmissing-format-attribute
4285 @opindex Wsuggest-attribute=format
4286 @opindex Wmissing-format-attribute
4287 @opindex Wno-suggest-attribute=format
4288 @opindex Wno-missing-format-attribute
4292 Warn about function pointers that might be candidates for @code{format}
4293 attributes. Note these are only possible candidates, not absolute ones.
4294 GCC guesses that function pointers with @code{format} attributes that
4295 are used in assignment, initialization, parameter passing or return
4296 statements should have a corresponding @code{format} attribute in the
4297 resulting type. I.e.@: the left-hand side of the assignment or
4298 initialization, the type of the parameter variable, or the return type
4299 of the containing function respectively should also have a @code{format}
4300 attribute to avoid the warning.
4302 GCC also warns about function definitions that might be
4303 candidates for @code{format} attributes. Again, these are only
4304 possible candidates. GCC guesses that @code{format} attributes
4305 might be appropriate for any function that calls a function like
4306 @code{vprintf} or @code{vscanf}, but this might not always be the
4307 case, and some functions for which @code{format} attributes are
4308 appropriate may not be detected.
4311 @item -Wsuggest-final-types
4312 @opindex Wno-suggest-final-types
4313 @opindex Wsuggest-final-types
4314 Warn about types with virtual methods where code quality would be improved
4315 if the type were declared with the C++11 @code{final} specifier,
4317 declared in an anonymous namespace. This allows GCC to more aggressively
4318 devirtualize the polymorphic calls. This warning is more effective with link
4319 time optimization, where the information about the class hierarchy graph is
4322 @item -Wsuggest-final-methods
4323 @opindex Wno-suggest-final-methods
4324 @opindex Wsuggest-final-methods
4325 Warn about virtual methods where code quality would be improved if the method
4326 were declared with the C++11 @code{final} specifier,
4327 or, if possible, its type were
4328 declared in an anonymous namespace or with the @code{final} specifier.
4330 more effective with link time optimization, where the information about the
4331 class hierarchy graph is more complete. It is recommended to first consider
4332 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4335 @item -Wsuggest-override
4336 Warn about overriding virtual functions that are not marked with the override
4339 @item -Warray-bounds
4340 @itemx -Warray-bounds=@var{n}
4341 @opindex Wno-array-bounds
4342 @opindex Warray-bounds
4343 This option is only active when @option{-ftree-vrp} is active
4344 (default for @option{-O2} and above). It warns about subscripts to arrays
4345 that are always out of bounds. This warning is enabled by @option{-Wall}.
4348 @item -Warray-bounds=1
4349 This is the warning level of @option{-Warray-bounds} and is enabled
4350 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4352 @item -Warray-bounds=2
4353 This warning level also warns about out of bounds access for
4354 arrays at the end of a struct and for arrays accessed through
4355 pointers. This warning level may give a larger number of
4356 false positives and is deactivated by default.
4359 @item -Wbool-compare
4360 @opindex Wno-bool-compare
4361 @opindex Wbool-compare
4362 Warn about boolean expression compared with an integer value different from
4363 @code{true}/@code{false}. For instance, the following comparison is
4368 if ((n > 1) == 2) @{ @dots{} @}
4370 This warning is enabled by @option{-Wall}.
4372 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4373 @opindex Wno-discarded-qualifiers
4374 @opindex Wdiscarded-qualifiers
4375 Do not warn if type qualifiers on pointers are being discarded.
4376 Typically, the compiler warns if a @code{const char *} variable is
4377 passed to a function that takes a @code{char *} parameter. This option
4378 can be used to suppress such a warning.
4380 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4381 @opindex Wno-discarded-array-qualifiers
4382 @opindex Wdiscarded-array-qualifiers
4383 Do not warn if type qualifiers on arrays which are pointer targets
4384 are being discarded. Typically, the compiler warns if a
4385 @code{const int (*)[]} variable is passed to a function that
4386 takes a @code{int (*)[]} parameter. This option can be used to
4387 suppress such a warning.
4389 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4390 @opindex Wno-incompatible-pointer-types
4391 @opindex Wincompatible-pointer-types
4392 Do not warn when there is a conversion between pointers that have incompatible
4393 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4394 which warns for pointer argument passing or assignment with different
4397 @item -Wno-int-conversion @r{(C and Objective-C only)}
4398 @opindex Wno-int-conversion
4399 @opindex Wint-conversion
4400 Do not warn about incompatible integer to pointer and pointer to integer
4401 conversions. This warning is about implicit conversions; for explicit
4402 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4403 @option{-Wno-pointer-to-int-cast} may be used.
4405 @item -Wno-div-by-zero
4406 @opindex Wno-div-by-zero
4407 @opindex Wdiv-by-zero
4408 Do not warn about compile-time integer division by zero. Floating-point
4409 division by zero is not warned about, as it can be a legitimate way of
4410 obtaining infinities and NaNs.
4412 @item -Wsystem-headers
4413 @opindex Wsystem-headers
4414 @opindex Wno-system-headers
4415 @cindex warnings from system headers
4416 @cindex system headers, warnings from
4417 Print warning messages for constructs found in system header files.
4418 Warnings from system headers are normally suppressed, on the assumption
4419 that they usually do not indicate real problems and would only make the
4420 compiler output harder to read. Using this command-line option tells
4421 GCC to emit warnings from system headers as if they occurred in user
4422 code. However, note that using @option{-Wall} in conjunction with this
4423 option does @emph{not} warn about unknown pragmas in system
4424 headers---for that, @option{-Wunknown-pragmas} must also be used.
4427 @opindex Wtrampolines
4428 @opindex Wno-trampolines
4429 Warn about trampolines generated for pointers to nested functions.
4430 A trampoline is a small piece of data or code that is created at run
4431 time on the stack when the address of a nested function is taken, and is
4432 used to call the nested function indirectly. For some targets, it is
4433 made up of data only and thus requires no special treatment. But, for
4434 most targets, it is made up of code and thus requires the stack to be
4435 made executable in order for the program to work properly.
4438 @opindex Wfloat-equal
4439 @opindex Wno-float-equal
4440 Warn if floating-point values are used in equality comparisons.
4442 The idea behind this is that sometimes it is convenient (for the
4443 programmer) to consider floating-point values as approximations to
4444 infinitely precise real numbers. If you are doing this, then you need
4445 to compute (by analyzing the code, or in some other way) the maximum or
4446 likely maximum error that the computation introduces, and allow for it
4447 when performing comparisons (and when producing output, but that's a
4448 different problem). In particular, instead of testing for equality, you
4449 should check to see whether the two values have ranges that overlap; and
4450 this is done with the relational operators, so equality comparisons are
4453 @item -Wtraditional @r{(C and Objective-C only)}
4454 @opindex Wtraditional
4455 @opindex Wno-traditional
4456 Warn about certain constructs that behave differently in traditional and
4457 ISO C@. Also warn about ISO C constructs that have no traditional C
4458 equivalent, and/or problematic constructs that should be avoided.
4462 Macro parameters that appear within string literals in the macro body.
4463 In traditional C macro replacement takes place within string literals,
4464 but in ISO C it does not.
4467 In traditional C, some preprocessor directives did not exist.
4468 Traditional preprocessors only considered a line to be a directive
4469 if the @samp{#} appeared in column 1 on the line. Therefore
4470 @option{-Wtraditional} warns about directives that traditional C
4471 understands but ignores because the @samp{#} does not appear as the
4472 first character on the line. It also suggests you hide directives like
4473 @code{#pragma} not understood by traditional C by indenting them. Some
4474 traditional implementations do not recognize @code{#elif}, so this option
4475 suggests avoiding it altogether.
4478 A function-like macro that appears without arguments.
4481 The unary plus operator.
4484 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4485 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4486 constants.) Note, these suffixes appear in macros defined in the system
4487 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4488 Use of these macros in user code might normally lead to spurious
4489 warnings, however GCC's integrated preprocessor has enough context to
4490 avoid warning in these cases.
4493 A function declared external in one block and then used after the end of
4497 A @code{switch} statement has an operand of type @code{long}.
4500 A non-@code{static} function declaration follows a @code{static} one.
4501 This construct is not accepted by some traditional C compilers.
4504 The ISO type of an integer constant has a different width or
4505 signedness from its traditional type. This warning is only issued if
4506 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4507 typically represent bit patterns, are not warned about.
4510 Usage of ISO string concatenation is detected.
4513 Initialization of automatic aggregates.
4516 Identifier conflicts with labels. Traditional C lacks a separate
4517 namespace for labels.
4520 Initialization of unions. If the initializer is zero, the warning is
4521 omitted. This is done under the assumption that the zero initializer in
4522 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4523 initializer warnings and relies on default initialization to zero in the
4527 Conversions by prototypes between fixed/floating-point values and vice
4528 versa. The absence of these prototypes when compiling with traditional
4529 C causes serious problems. This is a subset of the possible
4530 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4533 Use of ISO C style function definitions. This warning intentionally is
4534 @emph{not} issued for prototype declarations or variadic functions
4535 because these ISO C features appear in your code when using
4536 libiberty's traditional C compatibility macros, @code{PARAMS} and
4537 @code{VPARAMS}. This warning is also bypassed for nested functions
4538 because that feature is already a GCC extension and thus not relevant to
4539 traditional C compatibility.
4542 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4543 @opindex Wtraditional-conversion
4544 @opindex Wno-traditional-conversion
4545 Warn if a prototype causes a type conversion that is different from what
4546 would happen to the same argument in the absence of a prototype. This
4547 includes conversions of fixed point to floating and vice versa, and
4548 conversions changing the width or signedness of a fixed-point argument
4549 except when the same as the default promotion.
4551 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4552 @opindex Wdeclaration-after-statement
4553 @opindex Wno-declaration-after-statement
4554 Warn when a declaration is found after a statement in a block. This
4555 construct, known from C++, was introduced with ISO C99 and is by default
4556 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4561 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4563 @item -Wno-endif-labels
4564 @opindex Wno-endif-labels
4565 @opindex Wendif-labels
4566 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4571 Warn whenever a local variable or type declaration shadows another
4572 variable, parameter, type, class member (in C++), or instance variable
4573 (in Objective-C) or whenever a built-in function is shadowed. Note
4574 that in C++, the compiler warns if a local variable shadows an
4575 explicit typedef, but not if it shadows a struct/class/enum.
4577 @item -Wno-shadow-ivar @r{(Objective-C only)}
4578 @opindex Wno-shadow-ivar
4579 @opindex Wshadow-ivar
4580 Do not warn whenever a local variable shadows an instance variable in an
4583 @item -Wlarger-than=@var{len}
4584 @opindex Wlarger-than=@var{len}
4585 @opindex Wlarger-than-@var{len}
4586 Warn whenever an object of larger than @var{len} bytes is defined.
4588 @item -Wframe-larger-than=@var{len}
4589 @opindex Wframe-larger-than
4590 Warn if the size of a function frame is larger than @var{len} bytes.
4591 The computation done to determine the stack frame size is approximate
4592 and not conservative.
4593 The actual requirements may be somewhat greater than @var{len}
4594 even if you do not get a warning. In addition, any space allocated
4595 via @code{alloca}, variable-length arrays, or related constructs
4596 is not included by the compiler when determining
4597 whether or not to issue a warning.
4599 @item -Wno-free-nonheap-object
4600 @opindex Wno-free-nonheap-object
4601 @opindex Wfree-nonheap-object
4602 Do not warn when attempting to free an object that was not allocated
4605 @item -Wstack-usage=@var{len}
4606 @opindex Wstack-usage
4607 Warn if the stack usage of a function might be larger than @var{len} bytes.
4608 The computation done to determine the stack usage is conservative.
4609 Any space allocated via @code{alloca}, variable-length arrays, or related
4610 constructs is included by the compiler when determining whether or not to
4613 The message is in keeping with the output of @option{-fstack-usage}.
4617 If the stack usage is fully static but exceeds the specified amount, it's:
4620 warning: stack usage is 1120 bytes
4623 If the stack usage is (partly) dynamic but bounded, it's:
4626 warning: stack usage might be 1648 bytes
4629 If the stack usage is (partly) dynamic and not bounded, it's:
4632 warning: stack usage might be unbounded
4636 @item -Wunsafe-loop-optimizations
4637 @opindex Wunsafe-loop-optimizations
4638 @opindex Wno-unsafe-loop-optimizations
4639 Warn if the loop cannot be optimized because the compiler cannot
4640 assume anything on the bounds of the loop indices. With
4641 @option{-funsafe-loop-optimizations} warn if the compiler makes
4644 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4645 @opindex Wno-pedantic-ms-format
4646 @opindex Wpedantic-ms-format
4647 When used in combination with @option{-Wformat}
4648 and @option{-pedantic} without GNU extensions, this option
4649 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4650 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4651 which depend on the MS runtime.
4653 @item -Wpointer-arith
4654 @opindex Wpointer-arith
4655 @opindex Wno-pointer-arith
4656 Warn about anything that depends on the ``size of'' a function type or
4657 of @code{void}. GNU C assigns these types a size of 1, for
4658 convenience in calculations with @code{void *} pointers and pointers
4659 to functions. In C++, warn also when an arithmetic operation involves
4660 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4663 @opindex Wtype-limits
4664 @opindex Wno-type-limits
4665 Warn if a comparison is always true or always false due to the limited
4666 range of the data type, but do not warn for constant expressions. For
4667 example, warn if an unsigned variable is compared against zero with
4668 @code{<} or @code{>=}. This warning is also enabled by
4671 @item -Wbad-function-cast @r{(C and Objective-C only)}
4672 @opindex Wbad-function-cast
4673 @opindex Wno-bad-function-cast
4674 Warn when a function call is cast to a non-matching type.
4675 For example, warn if a call to a function returning an integer type
4676 is cast to a pointer type.
4678 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4679 @opindex Wc90-c99-compat
4680 @opindex Wno-c90-c99-compat
4681 Warn about features not present in ISO C90, but present in ISO C99.
4682 For instance, warn about use of variable length arrays, @code{long long}
4683 type, @code{bool} type, compound literals, designated initializers, and so
4684 on. This option is independent of the standards mode. Warnings are disabled
4685 in the expression that follows @code{__extension__}.
4687 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4688 @opindex Wc99-c11-compat
4689 @opindex Wno-c99-c11-compat
4690 Warn about features not present in ISO C99, but present in ISO C11.
4691 For instance, warn about use of anonymous structures and unions,
4692 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4693 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4694 and so on. This option is independent of the standards mode. Warnings are
4695 disabled in the expression that follows @code{__extension__}.
4697 @item -Wc++-compat @r{(C and Objective-C only)}
4698 @opindex Wc++-compat
4699 Warn about ISO C constructs that are outside of the common subset of
4700 ISO C and ISO C++, e.g.@: request for implicit conversion from
4701 @code{void *} to a pointer to non-@code{void} type.
4703 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4704 @opindex Wc++11-compat
4705 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4706 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4707 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4708 enabled by @option{-Wall}.
4710 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4711 @opindex Wc++14-compat
4712 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4713 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4717 @opindex Wno-cast-qual
4718 Warn whenever a pointer is cast so as to remove a type qualifier from
4719 the target type. For example, warn if a @code{const char *} is cast
4720 to an ordinary @code{char *}.
4722 Also warn when making a cast that introduces a type qualifier in an
4723 unsafe way. For example, casting @code{char **} to @code{const char **}
4724 is unsafe, as in this example:
4727 /* p is char ** value. */
4728 const char **q = (const char **) p;
4729 /* Assignment of readonly string to const char * is OK. */
4731 /* Now char** pointer points to read-only memory. */
4736 @opindex Wcast-align
4737 @opindex Wno-cast-align
4738 Warn whenever a pointer is cast such that the required alignment of the
4739 target is increased. For example, warn if a @code{char *} is cast to
4740 an @code{int *} on machines where integers can only be accessed at
4741 two- or four-byte boundaries.
4743 @item -Wwrite-strings
4744 @opindex Wwrite-strings
4745 @opindex Wno-write-strings
4746 When compiling C, give string constants the type @code{const
4747 char[@var{length}]} so that copying the address of one into a
4748 non-@code{const} @code{char *} pointer produces a warning. These
4749 warnings help you find at compile time code that can try to write
4750 into a string constant, but only if you have been very careful about
4751 using @code{const} in declarations and prototypes. Otherwise, it is
4752 just a nuisance. This is why we did not make @option{-Wall} request
4755 When compiling C++, warn about the deprecated conversion from string
4756 literals to @code{char *}. This warning is enabled by default for C++
4761 @opindex Wno-clobbered
4762 Warn for variables that might be changed by @code{longjmp} or
4763 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4765 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4766 @opindex Wconditionally-supported
4767 @opindex Wno-conditionally-supported
4768 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4771 @opindex Wconversion
4772 @opindex Wno-conversion
4773 Warn for implicit conversions that may alter a value. This includes
4774 conversions between real and integer, like @code{abs (x)} when
4775 @code{x} is @code{double}; conversions between signed and unsigned,
4776 like @code{unsigned ui = -1}; and conversions to smaller types, like
4777 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4778 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4779 changed by the conversion like in @code{abs (2.0)}. Warnings about
4780 conversions between signed and unsigned integers can be disabled by
4781 using @option{-Wno-sign-conversion}.
4783 For C++, also warn for confusing overload resolution for user-defined
4784 conversions; and conversions that never use a type conversion
4785 operator: conversions to @code{void}, the same type, a base class or a
4786 reference to them. Warnings about conversions between signed and
4787 unsigned integers are disabled by default in C++ unless
4788 @option{-Wsign-conversion} is explicitly enabled.
4790 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4791 @opindex Wconversion-null
4792 @opindex Wno-conversion-null
4793 Do not warn for conversions between @code{NULL} and non-pointer
4794 types. @option{-Wconversion-null} is enabled by default.
4796 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4797 @opindex Wzero-as-null-pointer-constant
4798 @opindex Wno-zero-as-null-pointer-constant
4799 Warn when a literal '0' is used as null pointer constant. This can
4800 be useful to facilitate the conversion to @code{nullptr} in C++11.
4804 @opindex Wno-date-time
4805 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4806 are encountered as they might prevent bit-wise-identical reproducible
4809 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4810 @opindex Wdelete-incomplete
4811 @opindex Wno-delete-incomplete
4812 Warn when deleting a pointer to incomplete type, which may cause
4813 undefined behavior at runtime. This warning is enabled by default.
4815 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4816 @opindex Wuseless-cast
4817 @opindex Wno-useless-cast
4818 Warn when an expression is casted to its own type.
4821 @opindex Wempty-body
4822 @opindex Wno-empty-body
4823 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4824 while} statement. This warning is also enabled by @option{-Wextra}.
4826 @item -Wenum-compare
4827 @opindex Wenum-compare
4828 @opindex Wno-enum-compare
4829 Warn about a comparison between values of different enumerated types.
4830 In C++ enumeral mismatches in conditional expressions are also
4831 diagnosed and the warning is enabled by default. In C this warning is
4832 enabled by @option{-Wall}.
4834 @item -Wjump-misses-init @r{(C, Objective-C only)}
4835 @opindex Wjump-misses-init
4836 @opindex Wno-jump-misses-init
4837 Warn if a @code{goto} statement or a @code{switch} statement jumps
4838 forward across the initialization of a variable, or jumps backward to a
4839 label after the variable has been initialized. This only warns about
4840 variables that are initialized when they are declared. This warning is
4841 only supported for C and Objective-C; in C++ this sort of branch is an
4844 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4845 can be disabled with the @option{-Wno-jump-misses-init} option.
4847 @item -Wsign-compare
4848 @opindex Wsign-compare
4849 @opindex Wno-sign-compare
4850 @cindex warning for comparison of signed and unsigned values
4851 @cindex comparison of signed and unsigned values, warning
4852 @cindex signed and unsigned values, comparison warning
4853 Warn when a comparison between signed and unsigned values could produce
4854 an incorrect result when the signed value is converted to unsigned.
4855 This warning is also enabled by @option{-Wextra}; to get the other warnings
4856 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4858 @item -Wsign-conversion
4859 @opindex Wsign-conversion
4860 @opindex Wno-sign-conversion
4861 Warn for implicit conversions that may change the sign of an integer
4862 value, like assigning a signed integer expression to an unsigned
4863 integer variable. An explicit cast silences the warning. In C, this
4864 option is enabled also by @option{-Wconversion}.
4866 @item -Wfloat-conversion
4867 @opindex Wfloat-conversion
4868 @opindex Wno-float-conversion
4869 Warn for implicit conversions that reduce the precision of a real value.
4870 This includes conversions from real to integer, and from higher precision
4871 real to lower precision real values. This option is also enabled by
4872 @option{-Wconversion}.
4874 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4875 @opindex Wsized-deallocation
4876 @opindex Wno-sized-deallocation
4877 Warn about a definition of an unsized deallocation function
4879 void operator delete (void *) noexcept;
4880 void operator delete[] (void *) noexcept;
4882 without a definition of the corresponding sized deallocation function
4884 void operator delete (void *, std::size_t) noexcept;
4885 void operator delete[] (void *, std::size_t) noexcept;
4887 or vice versa. Enabled by @option{-Wextra} along with
4888 @option{-fsized-deallocation}.
4890 @item -Wsizeof-pointer-memaccess
4891 @opindex Wsizeof-pointer-memaccess
4892 @opindex Wno-sizeof-pointer-memaccess
4893 Warn for suspicious length parameters to certain string and memory built-in
4894 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
4895 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
4896 but a pointer, and suggests a possible fix, or about
4897 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
4900 @item -Wsizeof-array-argument
4901 @opindex Wsizeof-array-argument
4902 @opindex Wno-sizeof-array-argument
4903 Warn when the @code{sizeof} operator is applied to a parameter that is
4904 declared as an array in a function definition. This warning is enabled by
4905 default for C and C++ programs.
4907 @item -Wmemset-transposed-args
4908 @opindex Wmemset-transposed-args
4909 @opindex Wno-memset-transposed-args
4910 Warn for suspicious calls to the @code{memset} built-in function, if the
4911 second argument is not zero and the third argument is zero. This warns e.g.@
4912 about @code{memset (buf, sizeof buf, 0)} where most probably
4913 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
4914 is only emitted if the third argument is literal zero. If it is some
4915 expression that is folded to zero, a cast of zero to some type, etc.,
4916 it is far less likely that the user has mistakenly exchanged the arguments
4917 and no warning is emitted. This warning is enabled by @option{-Wall}.
4921 @opindex Wno-address
4922 Warn about suspicious uses of memory addresses. These include using
4923 the address of a function in a conditional expression, such as
4924 @code{void func(void); if (func)}, and comparisons against the memory
4925 address of a string literal, such as @code{if (x == "abc")}. Such
4926 uses typically indicate a programmer error: the address of a function
4927 always evaluates to true, so their use in a conditional usually
4928 indicate that the programmer forgot the parentheses in a function
4929 call; and comparisons against string literals result in unspecified
4930 behavior and are not portable in C, so they usually indicate that the
4931 programmer intended to use @code{strcmp}. This warning is enabled by
4935 @opindex Wlogical-op
4936 @opindex Wno-logical-op
4937 Warn about suspicious uses of logical operators in expressions.
4938 This includes using logical operators in contexts where a
4939 bit-wise operator is likely to be expected. Also warns when
4940 the operands of a logical operator are the same:
4943 if (a < 0 && a < 0) @{ @dots{} @}
4946 @item -Wlogical-not-parentheses
4947 @opindex Wlogical-not-parentheses
4948 @opindex Wno-logical-not-parentheses
4949 Warn about logical not used on the left hand side operand of a comparison.
4950 This option does not warn if the RHS operand is of a boolean type. Its
4951 purpose is to detect suspicious code like the following:
4955 if (!a > 1) @{ @dots{} @}
4958 It is possible to suppress the warning by wrapping the LHS into
4961 if ((!a) > 1) @{ @dots{} @}
4964 This warning is enabled by @option{-Wall}.
4966 @item -Waggregate-return
4967 @opindex Waggregate-return
4968 @opindex Wno-aggregate-return
4969 Warn if any functions that return structures or unions are defined or
4970 called. (In languages where you can return an array, this also elicits
4973 @item -Wno-aggressive-loop-optimizations
4974 @opindex Wno-aggressive-loop-optimizations
4975 @opindex Waggressive-loop-optimizations
4976 Warn if in a loop with constant number of iterations the compiler detects
4977 undefined behavior in some statement during one or more of the iterations.
4979 @item -Wno-attributes
4980 @opindex Wno-attributes
4981 @opindex Wattributes
4982 Do not warn if an unexpected @code{__attribute__} is used, such as
4983 unrecognized attributes, function attributes applied to variables,
4984 etc. This does not stop errors for incorrect use of supported
4987 @item -Wno-builtin-macro-redefined
4988 @opindex Wno-builtin-macro-redefined
4989 @opindex Wbuiltin-macro-redefined
4990 Do not warn if certain built-in macros are redefined. This suppresses
4991 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
4992 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
4994 @item -Wstrict-prototypes @r{(C and Objective-C only)}
4995 @opindex Wstrict-prototypes
4996 @opindex Wno-strict-prototypes
4997 Warn if a function is declared or defined without specifying the
4998 argument types. (An old-style function definition is permitted without
4999 a warning if preceded by a declaration that specifies the argument
5002 @item -Wold-style-declaration @r{(C and Objective-C only)}
5003 @opindex Wold-style-declaration
5004 @opindex Wno-old-style-declaration
5005 Warn for obsolescent usages, according to the C Standard, in a
5006 declaration. For example, warn if storage-class specifiers like
5007 @code{static} are not the first things in a declaration. This warning
5008 is also enabled by @option{-Wextra}.
5010 @item -Wold-style-definition @r{(C and Objective-C only)}
5011 @opindex Wold-style-definition
5012 @opindex Wno-old-style-definition
5013 Warn if an old-style function definition is used. A warning is given
5014 even if there is a previous prototype.
5016 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5017 @opindex Wmissing-parameter-type
5018 @opindex Wno-missing-parameter-type
5019 A function parameter is declared without a type specifier in K&R-style
5026 This warning is also enabled by @option{-Wextra}.
5028 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5029 @opindex Wmissing-prototypes
5030 @opindex Wno-missing-prototypes
5031 Warn if a global function is defined without a previous prototype
5032 declaration. This warning is issued even if the definition itself
5033 provides a prototype. Use this option to detect global functions
5034 that do not have a matching prototype declaration in a header file.
5035 This option is not valid for C++ because all function declarations
5036 provide prototypes and a non-matching declaration declares an
5037 overload rather than conflict with an earlier declaration.
5038 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5040 @item -Wmissing-declarations
5041 @opindex Wmissing-declarations
5042 @opindex Wno-missing-declarations
5043 Warn if a global function is defined without a previous declaration.
5044 Do so even if the definition itself provides a prototype.
5045 Use this option to detect global functions that are not declared in
5046 header files. In C, no warnings are issued for functions with previous
5047 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5048 missing prototypes. In C++, no warnings are issued for function templates,
5049 or for inline functions, or for functions in anonymous namespaces.
5051 @item -Wmissing-field-initializers
5052 @opindex Wmissing-field-initializers
5053 @opindex Wno-missing-field-initializers
5057 Warn if a structure's initializer has some fields missing. For
5058 example, the following code causes such a warning, because
5059 @code{x.h} is implicitly zero:
5062 struct s @{ int f, g, h; @};
5063 struct s x = @{ 3, 4 @};
5066 This option does not warn about designated initializers, so the following
5067 modification does not trigger a warning:
5070 struct s @{ int f, g, h; @};
5071 struct s x = @{ .f = 3, .g = 4 @};
5074 In C++ this option does not warn either about the empty @{ @}
5075 initializer, for example:
5078 struct s @{ int f, g, h; @};
5082 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5083 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5085 @item -Wno-multichar
5086 @opindex Wno-multichar
5088 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5089 Usually they indicate a typo in the user's code, as they have
5090 implementation-defined values, and should not be used in portable code.
5092 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5093 @opindex Wnormalized=
5094 @opindex Wnormalized
5095 @opindex Wno-normalized
5098 @cindex character set, input normalization
5099 In ISO C and ISO C++, two identifiers are different if they are
5100 different sequences of characters. However, sometimes when characters
5101 outside the basic ASCII character set are used, you can have two
5102 different character sequences that look the same. To avoid confusion,
5103 the ISO 10646 standard sets out some @dfn{normalization rules} which
5104 when applied ensure that two sequences that look the same are turned into
5105 the same sequence. GCC can warn you if you are using identifiers that
5106 have not been normalized; this option controls that warning.
5108 There are four levels of warning supported by GCC@. The default is
5109 @option{-Wnormalized=nfc}, which warns about any identifier that is
5110 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5111 recommended form for most uses. It is equivalent to
5112 @option{-Wnormalized}.
5114 Unfortunately, there are some characters allowed in identifiers by
5115 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5116 identifiers. That is, there's no way to use these symbols in portable
5117 ISO C or C++ and have all your identifiers in NFC@.
5118 @option{-Wnormalized=id} suppresses the warning for these characters.
5119 It is hoped that future versions of the standards involved will correct
5120 this, which is why this option is not the default.
5122 You can switch the warning off for all characters by writing
5123 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5124 only do this if you are using some other normalization scheme (like
5125 ``D''), because otherwise you can easily create bugs that are
5126 literally impossible to see.
5128 Some characters in ISO 10646 have distinct meanings but look identical
5129 in some fonts or display methodologies, especially once formatting has
5130 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5131 LETTER N'', displays just like a regular @code{n} that has been
5132 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5133 normalization scheme to convert all these into a standard form as
5134 well, and GCC warns if your code is not in NFKC if you use
5135 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5136 about every identifier that contains the letter O because it might be
5137 confused with the digit 0, and so is not the default, but may be
5138 useful as a local coding convention if the programming environment
5139 cannot be fixed to display these characters distinctly.
5141 @item -Wno-deprecated
5142 @opindex Wno-deprecated
5143 @opindex Wdeprecated
5144 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5146 @item -Wno-deprecated-declarations
5147 @opindex Wno-deprecated-declarations
5148 @opindex Wdeprecated-declarations
5149 Do not warn about uses of functions (@pxref{Function Attributes}),
5150 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5151 Attributes}) marked as deprecated by using the @code{deprecated}
5155 @opindex Wno-overflow
5157 Do not warn about compile-time overflow in constant expressions.
5162 Warn about One Definition Rule violations during link-time optimization.
5163 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5166 @opindex Wopenm-simd
5167 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5168 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5169 option can be used to relax the cost model.
5171 @item -Woverride-init @r{(C and Objective-C only)}
5172 @opindex Woverride-init
5173 @opindex Wno-override-init
5177 Warn if an initialized field without side effects is overridden when
5178 using designated initializers (@pxref{Designated Inits, , Designated
5181 This warning is included in @option{-Wextra}. To get other
5182 @option{-Wextra} warnings without this one, use @option{-Wextra
5183 -Wno-override-init}.
5188 Warn if a structure is given the packed attribute, but the packed
5189 attribute has no effect on the layout or size of the structure.
5190 Such structures may be mis-aligned for little benefit. For
5191 instance, in this code, the variable @code{f.x} in @code{struct bar}
5192 is misaligned even though @code{struct bar} does not itself
5193 have the packed attribute:
5200 @} __attribute__((packed));
5208 @item -Wpacked-bitfield-compat
5209 @opindex Wpacked-bitfield-compat
5210 @opindex Wno-packed-bitfield-compat
5211 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5212 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5213 the change can lead to differences in the structure layout. GCC
5214 informs you when the offset of such a field has changed in GCC 4.4.
5215 For example there is no longer a 4-bit padding between field @code{a}
5216 and @code{b} in this structure:
5223 @} __attribute__ ((packed));
5226 This warning is enabled by default. Use
5227 @option{-Wno-packed-bitfield-compat} to disable this warning.
5232 Warn if padding is included in a structure, either to align an element
5233 of the structure or to align the whole structure. Sometimes when this
5234 happens it is possible to rearrange the fields of the structure to
5235 reduce the padding and so make the structure smaller.
5237 @item -Wredundant-decls
5238 @opindex Wredundant-decls
5239 @opindex Wno-redundant-decls
5240 Warn if anything is declared more than once in the same scope, even in
5241 cases where multiple declaration is valid and changes nothing.
5243 @item -Wnested-externs @r{(C and Objective-C only)}
5244 @opindex Wnested-externs
5245 @opindex Wno-nested-externs
5246 Warn if an @code{extern} declaration is encountered within a function.
5248 @item -Wno-inherited-variadic-ctor
5249 @opindex Winherited-variadic-ctor
5250 @opindex Wno-inherited-variadic-ctor
5251 Suppress warnings about use of C++11 inheriting constructors when the
5252 base class inherited from has a C variadic constructor; the warning is
5253 on by default because the ellipsis is not inherited.
5258 Warn if a function that is declared as inline cannot be inlined.
5259 Even with this option, the compiler does not warn about failures to
5260 inline functions declared in system headers.
5262 The compiler uses a variety of heuristics to determine whether or not
5263 to inline a function. For example, the compiler takes into account
5264 the size of the function being inlined and the amount of inlining
5265 that has already been done in the current function. Therefore,
5266 seemingly insignificant changes in the source program can cause the
5267 warnings produced by @option{-Winline} to appear or disappear.
5269 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5270 @opindex Wno-invalid-offsetof
5271 @opindex Winvalid-offsetof
5272 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5273 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5274 to a non-standard-layout type is undefined. In existing C++ implementations,
5275 however, @code{offsetof} typically gives meaningful results.
5276 This flag is for users who are aware that they are
5277 writing nonportable code and who have deliberately chosen to ignore the
5280 The restrictions on @code{offsetof} may be relaxed in a future version
5281 of the C++ standard.
5283 @item -Wno-int-to-pointer-cast
5284 @opindex Wno-int-to-pointer-cast
5285 @opindex Wint-to-pointer-cast
5286 Suppress warnings from casts to pointer type of an integer of a
5287 different size. In C++, casting to a pointer type of smaller size is
5288 an error. @option{Wint-to-pointer-cast} is enabled by default.
5291 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5292 @opindex Wno-pointer-to-int-cast
5293 @opindex Wpointer-to-int-cast
5294 Suppress warnings from casts from a pointer to an integer type of a
5298 @opindex Winvalid-pch
5299 @opindex Wno-invalid-pch
5300 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5301 the search path but can't be used.
5305 @opindex Wno-long-long
5306 Warn if @code{long long} type is used. This is enabled by either
5307 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5308 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5310 @item -Wvariadic-macros
5311 @opindex Wvariadic-macros
5312 @opindex Wno-variadic-macros
5313 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5314 alternate syntax is used in ISO C99 mode. This is enabled by either
5315 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5316 messages, use @option{-Wno-variadic-macros}.
5320 @opindex Wno-varargs
5321 Warn upon questionable usage of the macros used to handle variable
5322 arguments like @code{va_start}. This is default. To inhibit the
5323 warning messages, use @option{-Wno-varargs}.
5325 @item -Wvector-operation-performance
5326 @opindex Wvector-operation-performance
5327 @opindex Wno-vector-operation-performance
5328 Warn if vector operation is not implemented via SIMD capabilities of the
5329 architecture. Mainly useful for the performance tuning.
5330 Vector operation can be implemented @code{piecewise}, which means that the
5331 scalar operation is performed on every vector element;
5332 @code{in parallel}, which means that the vector operation is implemented
5333 using scalars of wider type, which normally is more performance efficient;
5334 and @code{as a single scalar}, which means that vector fits into a
5337 @item -Wno-virtual-move-assign
5338 @opindex Wvirtual-move-assign
5339 @opindex Wno-virtual-move-assign
5340 Suppress warnings about inheriting from a virtual base with a
5341 non-trivial C++11 move assignment operator. This is dangerous because
5342 if the virtual base is reachable along more than one path, it is
5343 moved multiple times, which can mean both objects end up in the
5344 moved-from state. If the move assignment operator is written to avoid
5345 moving from a moved-from object, this warning can be disabled.
5350 Warn if variable length array is used in the code.
5351 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5352 the variable length array.
5354 @item -Wvolatile-register-var
5355 @opindex Wvolatile-register-var
5356 @opindex Wno-volatile-register-var
5357 Warn if a register variable is declared volatile. The volatile
5358 modifier does not inhibit all optimizations that may eliminate reads
5359 and/or writes to register variables. This warning is enabled by
5362 @item -Wdisabled-optimization
5363 @opindex Wdisabled-optimization
5364 @opindex Wno-disabled-optimization
5365 Warn if a requested optimization pass is disabled. This warning does
5366 not generally indicate that there is anything wrong with your code; it
5367 merely indicates that GCC's optimizers are unable to handle the code
5368 effectively. Often, the problem is that your code is too big or too
5369 complex; GCC refuses to optimize programs when the optimization
5370 itself is likely to take inordinate amounts of time.
5372 @item -Wpointer-sign @r{(C and Objective-C only)}
5373 @opindex Wpointer-sign
5374 @opindex Wno-pointer-sign
5375 Warn for pointer argument passing or assignment with different signedness.
5376 This option is only supported for C and Objective-C@. It is implied by
5377 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5378 @option{-Wno-pointer-sign}.
5380 @item -Wstack-protector
5381 @opindex Wstack-protector
5382 @opindex Wno-stack-protector
5383 This option is only active when @option{-fstack-protector} is active. It
5384 warns about functions that are not protected against stack smashing.
5386 @item -Woverlength-strings
5387 @opindex Woverlength-strings
5388 @opindex Wno-overlength-strings
5389 Warn about string constants that are longer than the ``minimum
5390 maximum'' length specified in the C standard. Modern compilers
5391 generally allow string constants that are much longer than the
5392 standard's minimum limit, but very portable programs should avoid
5393 using longer strings.
5395 The limit applies @emph{after} string constant concatenation, and does
5396 not count the trailing NUL@. In C90, the limit was 509 characters; in
5397 C99, it was raised to 4095. C++98 does not specify a normative
5398 minimum maximum, so we do not diagnose overlength strings in C++@.
5400 This option is implied by @option{-Wpedantic}, and can be disabled with
5401 @option{-Wno-overlength-strings}.
5403 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5404 @opindex Wunsuffixed-float-constants
5406 Issue a warning for any floating constant that does not have
5407 a suffix. When used together with @option{-Wsystem-headers} it
5408 warns about such constants in system header files. This can be useful
5409 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5410 from the decimal floating-point extension to C99.
5412 @item -Wno-designated-init @r{(C and Objective-C only)}
5413 Suppress warnings when a positional initializer is used to initialize
5414 a structure that has been marked with the @code{designated_init}
5419 @node Debugging Options
5420 @section Options for Debugging Your Program or GCC
5421 @cindex options, debugging
5422 @cindex debugging information options
5424 GCC has various special options that are used for debugging
5425 either your program or GCC:
5430 Produce debugging information in the operating system's native format
5431 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5434 On most systems that use stabs format, @option{-g} enables use of extra
5435 debugging information that only GDB can use; this extra information
5436 makes debugging work better in GDB but probably makes other debuggers
5438 refuse to read the program. If you want to control for certain whether
5439 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5440 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5442 GCC allows you to use @option{-g} with
5443 @option{-O}. The shortcuts taken by optimized code may occasionally
5444 produce surprising results: some variables you declared may not exist
5445 at all; flow of control may briefly move where you did not expect it;
5446 some statements may not be executed because they compute constant
5447 results or their values are already at hand; some statements may
5448 execute in different places because they have been moved out of loops.
5450 Nevertheless it proves possible to debug optimized output. This makes
5451 it reasonable to use the optimizer for programs that might have bugs.
5453 The following options are useful when GCC is generated with the
5454 capability for more than one debugging format.
5457 @opindex gsplit-dwarf
5458 Separate as much dwarf debugging information as possible into a
5459 separate output file with the extension .dwo. This option allows
5460 the build system to avoid linking files with debug information. To
5461 be useful, this option requires a debugger capable of reading .dwo
5466 Produce debugging information for use by GDB@. This means to use the
5467 most expressive format available (DWARF 2, stabs, or the native format
5468 if neither of those are supported), including GDB extensions if at all
5473 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5475 @item -ggnu-pubnames
5476 @opindex ggnu-pubnames
5477 Generate .debug_pubnames and .debug_pubtypes sections in a format
5478 suitable for conversion into a GDB@ index. This option is only useful
5479 with a linker that can produce GDB@ index version 7.
5483 Produce debugging information in stabs format (if that is supported),
5484 without GDB extensions. This is the format used by DBX on most BSD
5485 systems. On MIPS, Alpha and System V Release 4 systems this option
5486 produces stabs debugging output that is not understood by DBX or SDB@.
5487 On System V Release 4 systems this option requires the GNU assembler.
5489 @item -feliminate-unused-debug-symbols
5490 @opindex feliminate-unused-debug-symbols
5491 Produce debugging information in stabs format (if that is supported),
5492 for only symbols that are actually used.
5494 @item -femit-class-debug-always
5495 @opindex femit-class-debug-always
5496 Instead of emitting debugging information for a C++ class in only one
5497 object file, emit it in all object files using the class. This option
5498 should be used only with debuggers that are unable to handle the way GCC
5499 normally emits debugging information for classes because using this
5500 option increases the size of debugging information by as much as a
5503 @item -fdebug-types-section
5504 @opindex fdebug-types-section
5505 @opindex fno-debug-types-section
5506 When using DWARF Version 4 or higher, type DIEs can be put into
5507 their own @code{.debug_types} section instead of making them part of the
5508 @code{.debug_info} section. It is more efficient to put them in a separate
5509 comdat sections since the linker can then remove duplicates.
5510 But not all DWARF consumers support @code{.debug_types} sections yet
5511 and on some objects @code{.debug_types} produces larger instead of smaller
5512 debugging information.
5516 Produce debugging information in stabs format (if that is supported),
5517 using GNU extensions understood only by the GNU debugger (GDB)@. The
5518 use of these extensions is likely to make other debuggers crash or
5519 refuse to read the program.
5523 Produce debugging information in COFF format (if that is supported).
5524 This is the format used by SDB on most System V systems prior to
5529 Produce debugging information in XCOFF format (if that is supported).
5530 This is the format used by the DBX debugger on IBM RS/6000 systems.
5534 Produce debugging information in XCOFF format (if that is supported),
5535 using GNU extensions understood only by the GNU debugger (GDB)@. The
5536 use of these extensions is likely to make other debuggers crash or
5537 refuse to read the program, and may cause assemblers other than the GNU
5538 assembler (GAS) to fail with an error.
5540 @item -gdwarf-@var{version}
5541 @opindex gdwarf-@var{version}
5542 Produce debugging information in DWARF format (if that is supported).
5543 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5544 for most targets is 4. DWARF Version 5 is only experimental.
5546 Note that with DWARF Version 2, some ports require and always
5547 use some non-conflicting DWARF 3 extensions in the unwind tables.
5549 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5550 for maximum benefit.
5552 @item -grecord-gcc-switches
5553 @opindex grecord-gcc-switches
5554 This switch causes the command-line options used to invoke the
5555 compiler that may affect code generation to be appended to the
5556 DW_AT_producer attribute in DWARF debugging information. The options
5557 are concatenated with spaces separating them from each other and from
5558 the compiler version. See also @option{-frecord-gcc-switches} for another
5559 way of storing compiler options into the object file. This is the default.
5561 @item -gno-record-gcc-switches
5562 @opindex gno-record-gcc-switches
5563 Disallow appending command-line options to the DW_AT_producer attribute
5564 in DWARF debugging information.
5566 @item -gstrict-dwarf
5567 @opindex gstrict-dwarf
5568 Disallow using extensions of later DWARF standard version than selected
5569 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5570 DWARF extensions from later standard versions is allowed.
5572 @item -gno-strict-dwarf
5573 @opindex gno-strict-dwarf
5574 Allow using extensions of later DWARF standard version than selected with
5575 @option{-gdwarf-@var{version}}.
5577 @item -gz@r{[}=@var{type}@r{]}
5579 Produce compressed debug sections in DWARF format, if that is supported.
5580 If @var{type} is not given, the default type depends on the capabilities
5581 of the assembler and linker used. @var{type} may be one of
5582 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5583 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5584 compression in traditional GNU format). If the linker doesn't support
5585 writing compressed debug sections, the option is rejected. Otherwise,
5586 if the assembler does not support them, @option{-gz} is silently ignored
5587 when producing object files.
5591 Produce debugging information in Alpha/VMS debug format (if that is
5592 supported). This is the format used by DEBUG on Alpha/VMS systems.
5595 @itemx -ggdb@var{level}
5596 @itemx -gstabs@var{level}
5597 @itemx -gcoff@var{level}
5598 @itemx -gxcoff@var{level}
5599 @itemx -gvms@var{level}
5600 Request debugging information and also use @var{level} to specify how
5601 much information. The default level is 2.
5603 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5606 Level 1 produces minimal information, enough for making backtraces in
5607 parts of the program that you don't plan to debug. This includes
5608 descriptions of functions and external variables, and line number
5609 tables, but no information about local variables.
5611 Level 3 includes extra information, such as all the macro definitions
5612 present in the program. Some debuggers support macro expansion when
5613 you use @option{-g3}.
5615 @option{-gdwarf-2} does not accept a concatenated debug level, because
5616 GCC used to support an option @option{-gdwarf} that meant to generate
5617 debug information in version 1 of the DWARF format (which is very
5618 different from version 2), and it would have been too confusing. That
5619 debug format is long obsolete, but the option cannot be changed now.
5620 Instead use an additional @option{-g@var{level}} option to change the
5621 debug level for DWARF.
5625 Turn off generation of debug info, if leaving out this option
5626 generates it, or turn it on at level 2 otherwise. The position of this
5627 argument in the command line does not matter; it takes effect after all
5628 other options are processed, and it does so only once, no matter how
5629 many times it is given. This is mainly intended to be used with
5630 @option{-fcompare-debug}.
5632 @item -fsanitize=address
5633 @opindex fsanitize=address
5634 Enable AddressSanitizer, a fast memory error detector.
5635 Memory access instructions are instrumented to detect
5636 out-of-bounds and use-after-free bugs.
5637 See @uref{http://code.google.com/p/address-sanitizer/} for
5638 more details. The run-time behavior can be influenced using the
5639 @env{ASAN_OPTIONS} environment variable; see
5640 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5641 a list of supported options.
5643 @item -fsanitize=kernel-address
5644 @opindex fsanitize=kernel-address
5645 Enable AddressSanitizer for Linux kernel.
5646 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5648 @item -fsanitize=thread
5649 @opindex fsanitize=thread
5650 Enable ThreadSanitizer, a fast data race detector.
5651 Memory access instructions are instrumented to detect
5652 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5653 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5654 environment variable; see
5655 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5658 @item -fsanitize=leak
5659 @opindex fsanitize=leak
5660 Enable LeakSanitizer, a memory leak detector.
5661 This option only matters for linking of executables and if neither
5662 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5663 case the executable is linked against a library that overrides @code{malloc}
5664 and other allocator functions. See
5665 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5666 details. The run-time behavior can be influenced using the
5667 @env{LSAN_OPTIONS} environment variable.
5669 @item -fsanitize=undefined
5670 @opindex fsanitize=undefined
5671 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5672 Various computations are instrumented to detect undefined behavior
5673 at runtime. Current suboptions are:
5677 @item -fsanitize=shift
5678 @opindex fsanitize=shift
5679 This option enables checking that the result of a shift operation is
5680 not undefined. Note that what exactly is considered undefined differs
5681 slightly between C and C++, as well as between ISO C90 and C99, etc.
5683 @item -fsanitize=integer-divide-by-zero
5684 @opindex fsanitize=integer-divide-by-zero
5685 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5687 @item -fsanitize=unreachable
5688 @opindex fsanitize=unreachable
5689 With this option, the compiler turns the @code{__builtin_unreachable}
5690 call into a diagnostics message call instead. When reaching the
5691 @code{__builtin_unreachable} call, the behavior is undefined.
5693 @item -fsanitize=vla-bound
5694 @opindex fsanitize=vla-bound
5695 This option instructs the compiler to check that the size of a variable
5696 length array is positive.
5698 @item -fsanitize=null
5699 @opindex fsanitize=null
5700 This option enables pointer checking. Particularly, the application
5701 built with this option turned on will issue an error message when it
5702 tries to dereference a NULL pointer, or if a reference (possibly an
5703 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5704 on an object pointed by a NULL pointer.
5706 @item -fsanitize=return
5707 @opindex fsanitize=return
5708 This option enables return statement checking. Programs
5709 built with this option turned on will issue an error message
5710 when the end of a non-void function is reached without actually
5711 returning a value. This option works in C++ only.
5713 @item -fsanitize=signed-integer-overflow
5714 @opindex fsanitize=signed-integer-overflow
5715 This option enables signed integer overflow checking. We check that
5716 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5717 does not overflow in the signed arithmetics. Note, integer promotion
5718 rules must be taken into account. That is, the following is not an
5721 signed char a = SCHAR_MAX;
5725 @item -fsanitize=bounds
5726 @opindex fsanitize=bounds
5727 This option enables instrumentation of array bounds. Various out of bounds
5728 accesses are detected. Flexible array members, flexible array member-like
5729 arrays, and initializers of variables with static storage are not instrumented.
5731 @item -fsanitize=alignment
5732 @opindex fsanitize=alignment
5734 This option enables checking of alignment of pointers when they are
5735 dereferenced, or when a reference is bound to insufficiently aligned target,
5736 or when a method or constructor is invoked on insufficiently aligned object.
5738 @item -fsanitize=object-size
5739 @opindex fsanitize=object-size
5740 This option enables instrumentation of memory references using the
5741 @code{__builtin_object_size} function. Various out of bounds pointer
5742 accesses are detected.
5744 @item -fsanitize=float-divide-by-zero
5745 @opindex fsanitize=float-divide-by-zero
5746 Detect floating-point division by zero. Unlike other similar options,
5747 @option{-fsanitize=float-divide-by-zero} is not enabled by
5748 @option{-fsanitize=undefined}, since floating-point division by zero can
5749 be a legitimate way of obtaining infinities and NaNs.
5751 @item -fsanitize=float-cast-overflow
5752 @opindex fsanitize=float-cast-overflow
5753 This option enables floating-point type to integer conversion checking.
5754 We check that the result of the conversion does not overflow.
5755 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5756 not enabled by @option{-fsanitize=undefined}.
5757 This option does not work well with @code{FE_INVALID} exceptions enabled.
5759 @item -fsanitize=nonnull-attribute
5760 @opindex fsanitize=nonnull-attribute
5762 This option enables instrumentation of calls, checking whether null values
5763 are not passed to arguments marked as requiring a non-null value by the
5764 @code{nonnull} function attribute.
5766 @item -fsanitize=returns-nonnull-attribute
5767 @opindex fsanitize=returns-nonnull-attribute
5769 This option enables instrumentation of return statements in functions
5770 marked with @code{returns_nonnull} function attribute, to detect returning
5771 of null values from such functions.
5773 @item -fsanitize=bool
5774 @opindex fsanitize=bool
5776 This option enables instrumentation of loads from bool. If a value other
5777 than 0/1 is loaded, a run-time error is issued.
5779 @item -fsanitize=enum
5780 @opindex fsanitize=enum
5782 This option enables instrumentation of loads from an enum type. If
5783 a value outside the range of values for the enum type is loaded,
5784 a run-time error is issued.
5786 @item -fsanitize=vptr
5787 @opindex fsanitize=vptr
5789 This option enables instrumentation of C++ member function calls, member
5790 accesses and some conversions between pointers to base and derived classes,
5791 to verify the referenced object has the correct dynamic type.
5795 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5796 @option{-fsanitize=undefined} gives a diagnostic message.
5797 This currently works only for the C family of languages.
5799 @item -fno-sanitize=all
5800 @opindex fno-sanitize=all
5802 This option disables all previously enabled sanitizers.
5803 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5806 @item -fasan-shadow-offset=@var{number}
5807 @opindex fasan-shadow-offset
5808 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5809 It is useful for experimenting with different shadow memory layouts in
5810 Kernel AddressSanitizer.
5812 @item -fsanitize-sections=@var{s1,s2,...}
5813 @opindex fsanitize-sections
5814 Sanitize global variables in selected user-defined sections. @var{si} may
5817 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5818 @opindex fsanitize-recover
5819 @opindex fno-sanitize-recover
5820 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5821 mentioned in comma-separated list of @var{opts}. Enabling this option
5822 for a sanitizer component causes it to attempt to continue
5823 running the program as if no error happened. This means multiple
5824 runtime errors can be reported in a single program run, and the exit
5825 code of the program may indicate success even when errors
5826 have been reported. The @option{-fno-sanitize-recover=} option
5827 can be used to alter
5828 this behavior: only the first detected error is reported
5829 and program then exits with a non-zero exit code.
5831 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5832 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5833 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5834 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5835 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5836 accepted, the former enables recovery for all sanitizers that support it,
5837 the latter disables recovery for all sanitizers that support it.
5839 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5841 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5844 Similarly @option{-fno-sanitize-recover} is equivalent to
5846 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5849 @item -fsanitize-undefined-trap-on-error
5850 @opindex fsanitize-undefined-trap-on-error
5851 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
5852 report undefined behavior using @code{__builtin_trap} rather than
5853 a @code{libubsan} library routine. The advantage of this is that the
5854 @code{libubsan} library is not needed and is not linked in, so this
5855 is usable even in freestanding environments.
5857 @item -fcheck-pointer-bounds
5858 @opindex fcheck-pointer-bounds
5859 @opindex fno-check-pointer-bounds
5860 @cindex Pointer Bounds Checker options
5861 Enable Pointer Bounds Checker instrumentation. Each memory reference
5862 is instrumented with checks of the pointer used for memory access against
5863 bounds associated with that pointer.
5866 is only an implementation for Intel MPX available, thus x86 target
5867 and @option{-mmpx} are required to enable this feature.
5868 MPX-based instrumentation requires
5869 a runtime library to enable MPX in hardware and handle bounds
5870 violation signals. By default when @option{-fcheck-pointer-bounds}
5871 and @option{-mmpx} options are used to link a program, the GCC driver
5872 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
5873 library. It also passes '-z bndplt' to a linker in case it supports this
5874 option (which is checked on libmpx configuration). Note that old versions
5875 of linker may ignore option. Gold linker doesn't support '-z bndplt'
5876 option. With no '-z bndplt' support in linker all calls to dynamic libraries
5877 lose passed bounds reducing overall protection level. It's highly
5878 recommended to use linker with '-z bndplt' support. In case such linker
5879 is not available it is adviced to always use @option{-static-libmpxwrappers}
5880 for better protection level or use @option{-static} to completely avoid
5881 external calls to dynamic libraries. MPX-based instrumentation
5882 may be used for debugging and also may be included in production code
5883 to increase program security. Depending on usage, you may
5884 have different requirements for the runtime library. The current version
5885 of the MPX runtime library is more oriented for use as a debugging
5886 tool. MPX runtime library usage implies @option{-lpthread}. See
5887 also @option{-static-libmpx}. The runtime library behavior can be
5888 influenced using various @env{CHKP_RT_*} environment variables. See
5889 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
5892 Generated instrumentation may be controlled by various
5893 @option{-fchkp-*} options and by the @code{bnd_variable_size}
5894 structure field attribute (@pxref{Type Attributes}) and
5895 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
5896 (@pxref{Function Attributes}). GCC also provides a number of built-in
5897 functions for controlling the Pointer Bounds Checker. @xref{Pointer
5898 Bounds Checker builtins}, for more information.
5900 @item -fchkp-check-incomplete-type
5901 @opindex fchkp-check-incomplete-type
5902 @opindex fno-chkp-check-incomplete-type
5903 Generate pointer bounds checks for variables with incomplete type.
5906 @item -fchkp-narrow-bounds
5907 @opindex fchkp-narrow-bounds
5908 @opindex fno-chkp-narrow-bounds
5909 Controls bounds used by Pointer Bounds Checker for pointers to object
5910 fields. If narrowing is enabled then field bounds are used. Otherwise
5911 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
5912 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
5914 @item -fchkp-first-field-has-own-bounds
5915 @opindex fchkp-first-field-has-own-bounds
5916 @opindex fno-chkp-first-field-has-own-bounds
5917 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
5918 first field in the structure. By default a pointer to the first field has
5919 the same bounds as a pointer to the whole structure.
5921 @item -fchkp-narrow-to-innermost-array
5922 @opindex fchkp-narrow-to-innermost-array
5923 @opindex fno-chkp-narrow-to-innermost-array
5924 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
5925 case of nested static array access. By default this option is disabled and
5926 bounds of the outermost array are used.
5928 @item -fchkp-optimize
5929 @opindex fchkp-optimize
5930 @opindex fno-chkp-optimize
5931 Enables Pointer Bounds Checker optimizations. Enabled by default at
5932 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
5934 @item -fchkp-use-fast-string-functions
5935 @opindex fchkp-use-fast-string-functions
5936 @opindex fno-chkp-use-fast-string-functions
5937 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
5938 by Pointer Bounds Checker. Disabled by default.
5940 @item -fchkp-use-nochk-string-functions
5941 @opindex fchkp-use-nochk-string-functions
5942 @opindex fno-chkp-use-nochk-string-functions
5943 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
5944 by Pointer Bounds Checker. Disabled by default.
5946 @item -fchkp-use-static-bounds
5947 @opindex fchkp-use-static-bounds
5948 @opindex fno-chkp-use-static-bounds
5949 Allow Pointer Bounds Checker to generate static bounds holding
5950 bounds of static variables. Enabled by default.
5952 @item -fchkp-use-static-const-bounds
5953 @opindex fchkp-use-static-const-bounds
5954 @opindex fno-chkp-use-static-const-bounds
5955 Use statically-initialized bounds for constant bounds instead of
5956 generating them each time they are required. By default enabled when
5957 @option{-fchkp-use-static-bounds} is enabled.
5959 @item -fchkp-treat-zero-dynamic-size-as-infinite
5960 @opindex fchkp-treat-zero-dynamic-size-as-infinite
5961 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
5962 With this option, objects with incomplete type whose
5963 dynamically-obtained size is zero are treated as having infinite size
5964 instead by Pointer Bounds
5965 Checker. This option may be helpful if a program is linked with a library
5966 missing size information for some symbols. Disabled by default.
5968 @item -fchkp-check-read
5969 @opindex fchkp-check-read
5970 @opindex fno-chkp-check-read
5971 Instructs Pointer Bounds Checker to generate checks for all read
5972 accesses to memory. Enabled by default.
5974 @item -fchkp-check-write
5975 @opindex fchkp-check-write
5976 @opindex fno-chkp-check-write
5977 Instructs Pointer Bounds Checker to generate checks for all write
5978 accesses to memory. Enabled by default.
5980 @item -fchkp-store-bounds
5981 @opindex fchkp-store-bounds
5982 @opindex fno-chkp-store-bounds
5983 Instructs Pointer Bounds Checker to generate bounds stores for
5984 pointer writes. Enabled by default.
5986 @item -fchkp-instrument-calls
5987 @opindex fchkp-instrument-calls
5988 @opindex fno-chkp-instrument-calls
5989 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
5992 @item -fchkp-instrument-marked-only
5993 @opindex fchkp-instrument-marked-only
5994 @opindex fno-chkp-instrument-marked-only
5995 Instructs Pointer Bounds Checker to instrument only functions
5996 marked with the @code{bnd_instrument} attribute
5997 (@pxref{Function Attributes}). Disabled by default.
5999 @item -fchkp-use-wrappers
6000 @opindex fchkp-use-wrappers
6001 @opindex fno-chkp-use-wrappers
6002 Allows Pointer Bounds Checker to replace calls to built-in functions
6003 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6004 is used to link a program, the GCC driver automatically links
6005 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6008 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6009 @opindex fdump-final-insns
6010 Dump the final internal representation (RTL) to @var{file}. If the
6011 optional argument is omitted (or if @var{file} is @code{.}), the name
6012 of the dump file is determined by appending @code{.gkd} to the
6013 compilation output file name.
6015 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6016 @opindex fcompare-debug
6017 @opindex fno-compare-debug
6018 If no error occurs during compilation, run the compiler a second time,
6019 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6020 passed to the second compilation. Dump the final internal
6021 representation in both compilations, and print an error if they differ.
6023 If the equal sign is omitted, the default @option{-gtoggle} is used.
6025 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6026 and nonzero, implicitly enables @option{-fcompare-debug}. If
6027 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6028 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6031 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6032 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6033 of the final representation and the second compilation, preventing even
6034 @env{GCC_COMPARE_DEBUG} from taking effect.
6036 To verify full coverage during @option{-fcompare-debug} testing, set
6037 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6038 which GCC rejects as an invalid option in any actual compilation
6039 (rather than preprocessing, assembly or linking). To get just a
6040 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6041 not overridden} will do.
6043 @item -fcompare-debug-second
6044 @opindex fcompare-debug-second
6045 This option is implicitly passed to the compiler for the second
6046 compilation requested by @option{-fcompare-debug}, along with options to
6047 silence warnings, and omitting other options that would cause
6048 side-effect compiler outputs to files or to the standard output. Dump
6049 files and preserved temporary files are renamed so as to contain the
6050 @code{.gk} additional extension during the second compilation, to avoid
6051 overwriting those generated by the first.
6053 When this option is passed to the compiler driver, it causes the
6054 @emph{first} compilation to be skipped, which makes it useful for little
6055 other than debugging the compiler proper.
6057 @item -feliminate-dwarf2-dups
6058 @opindex feliminate-dwarf2-dups
6059 Compress DWARF 2 debugging information by eliminating duplicated
6060 information about each symbol. This option only makes sense when
6061 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6063 @item -femit-struct-debug-baseonly
6064 @opindex femit-struct-debug-baseonly
6065 Emit debug information for struct-like types
6066 only when the base name of the compilation source file
6067 matches the base name of file in which the struct is defined.
6069 This option substantially reduces the size of debugging information,
6070 but at significant potential loss in type information to the debugger.
6071 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6072 See @option{-femit-struct-debug-detailed} for more detailed control.
6074 This option works only with DWARF 2.
6076 @item -femit-struct-debug-reduced
6077 @opindex femit-struct-debug-reduced
6078 Emit debug information for struct-like types
6079 only when the base name of the compilation source file
6080 matches the base name of file in which the type is defined,
6081 unless the struct is a template or defined in a system header.
6083 This option significantly reduces the size of debugging information,
6084 with some potential loss in type information to the debugger.
6085 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6086 See @option{-femit-struct-debug-detailed} for more detailed control.
6088 This option works only with DWARF 2.
6090 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6091 @opindex femit-struct-debug-detailed
6092 Specify the struct-like types
6093 for which the compiler generates debug information.
6094 The intent is to reduce duplicate struct debug information
6095 between different object files within the same program.
6097 This option is a detailed version of
6098 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6099 which serves for most needs.
6101 A specification has the syntax@*
6102 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6104 The optional first word limits the specification to
6105 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6106 A struct type is used directly when it is the type of a variable, member.
6107 Indirect uses arise through pointers to structs.
6108 That is, when use of an incomplete struct is valid, the use is indirect.
6110 @samp{struct one direct; struct two * indirect;}.
6112 The optional second word limits the specification to
6113 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6114 Generic structs are a bit complicated to explain.
6115 For C++, these are non-explicit specializations of template classes,
6116 or non-template classes within the above.
6117 Other programming languages have generics,
6118 but @option{-femit-struct-debug-detailed} does not yet implement them.
6120 The third word specifies the source files for those
6121 structs for which the compiler should emit debug information.
6122 The values @samp{none} and @samp{any} have the normal meaning.
6123 The value @samp{base} means that
6124 the base of name of the file in which the type declaration appears
6125 must match the base of the name of the main compilation file.
6126 In practice, this means that when compiling @file{foo.c}, debug information
6127 is generated for types declared in that file and @file{foo.h},
6128 but not other header files.
6129 The value @samp{sys} means those types satisfying @samp{base}
6130 or declared in system or compiler headers.
6132 You may need to experiment to determine the best settings for your application.
6134 The default is @option{-femit-struct-debug-detailed=all}.
6136 This option works only with DWARF 2.
6138 @item -fno-merge-debug-strings
6139 @opindex fmerge-debug-strings
6140 @opindex fno-merge-debug-strings
6141 Direct the linker to not merge together strings in the debugging
6142 information that are identical in different object files. Merging is
6143 not supported by all assemblers or linkers. Merging decreases the size
6144 of the debug information in the output file at the cost of increasing
6145 link processing time. Merging is enabled by default.
6147 @item -fdebug-prefix-map=@var{old}=@var{new}
6148 @opindex fdebug-prefix-map
6149 When compiling files in directory @file{@var{old}}, record debugging
6150 information describing them as in @file{@var{new}} instead.
6152 @item -fno-dwarf2-cfi-asm
6153 @opindex fdwarf2-cfi-asm
6154 @opindex fno-dwarf2-cfi-asm
6155 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6156 instead of using GAS @code{.cfi_*} directives.
6158 @cindex @command{prof}
6161 Generate extra code to write profile information suitable for the
6162 analysis program @command{prof}. You must use this option when compiling
6163 the source files you want data about, and you must also use it when
6166 @cindex @command{gprof}
6169 Generate extra code to write profile information suitable for the
6170 analysis program @command{gprof}. You must use this option when compiling
6171 the source files you want data about, and you must also use it when
6176 Makes the compiler print out each function name as it is compiled, and
6177 print some statistics about each pass when it finishes.
6180 @opindex ftime-report
6181 Makes the compiler print some statistics about the time consumed by each
6182 pass when it finishes.
6185 @opindex fmem-report
6186 Makes the compiler print some statistics about permanent memory
6187 allocation when it finishes.
6189 @item -fmem-report-wpa
6190 @opindex fmem-report-wpa
6191 Makes the compiler print some statistics about permanent memory
6192 allocation for the WPA phase only.
6194 @item -fpre-ipa-mem-report
6195 @opindex fpre-ipa-mem-report
6196 @item -fpost-ipa-mem-report
6197 @opindex fpost-ipa-mem-report
6198 Makes the compiler print some statistics about permanent memory
6199 allocation before or after interprocedural optimization.
6201 @item -fprofile-report
6202 @opindex fprofile-report
6203 Makes the compiler print some statistics about consistency of the
6204 (estimated) profile and effect of individual passes.
6207 @opindex fstack-usage
6208 Makes the compiler output stack usage information for the program, on a
6209 per-function basis. The filename for the dump is made by appending
6210 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6211 the output file, if explicitly specified and it is not an executable,
6212 otherwise it is the basename of the source file. An entry is made up
6217 The name of the function.
6221 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6224 The qualifier @code{static} means that the function manipulates the stack
6225 statically: a fixed number of bytes are allocated for the frame on function
6226 entry and released on function exit; no stack adjustments are otherwise made
6227 in the function. The second field is this fixed number of bytes.
6229 The qualifier @code{dynamic} means that the function manipulates the stack
6230 dynamically: in addition to the static allocation described above, stack
6231 adjustments are made in the body of the function, for example to push/pop
6232 arguments around function calls. If the qualifier @code{bounded} is also
6233 present, the amount of these adjustments is bounded at compile time and
6234 the second field is an upper bound of the total amount of stack used by
6235 the function. If it is not present, the amount of these adjustments is
6236 not bounded at compile time and the second field only represents the
6239 @item -fprofile-arcs
6240 @opindex fprofile-arcs
6241 Add code so that program flow @dfn{arcs} are instrumented. During
6242 execution the program records how many times each branch and call is
6243 executed and how many times it is taken or returns. When the compiled
6244 program exits it saves this data to a file called
6245 @file{@var{auxname}.gcda} for each source file. The data may be used for
6246 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6247 test coverage analysis (@option{-ftest-coverage}). Each object file's
6248 @var{auxname} is generated from the name of the output file, if
6249 explicitly specified and it is not the final executable, otherwise it is
6250 the basename of the source file. In both cases any suffix is removed
6251 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6252 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6253 @xref{Cross-profiling}.
6255 @cindex @command{gcov}
6259 This option is used to compile and link code instrumented for coverage
6260 analysis. The option is a synonym for @option{-fprofile-arcs}
6261 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6262 linking). See the documentation for those options for more details.
6267 Compile the source files with @option{-fprofile-arcs} plus optimization
6268 and code generation options. For test coverage analysis, use the
6269 additional @option{-ftest-coverage} option. You do not need to profile
6270 every source file in a program.
6273 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6274 (the latter implies the former).
6277 Run the program on a representative workload to generate the arc profile
6278 information. This may be repeated any number of times. You can run
6279 concurrent instances of your program, and provided that the file system
6280 supports locking, the data files will be correctly updated. Also
6281 @code{fork} calls are detected and correctly handled (double counting
6285 For profile-directed optimizations, compile the source files again with
6286 the same optimization and code generation options plus
6287 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6288 Control Optimization}).
6291 For test coverage analysis, use @command{gcov} to produce human readable
6292 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6293 @command{gcov} documentation for further information.
6297 With @option{-fprofile-arcs}, for each function of your program GCC
6298 creates a program flow graph, then finds a spanning tree for the graph.
6299 Only arcs that are not on the spanning tree have to be instrumented: the
6300 compiler adds code to count the number of times that these arcs are
6301 executed. When an arc is the only exit or only entrance to a block, the
6302 instrumentation code can be added to the block; otherwise, a new basic
6303 block must be created to hold the instrumentation code.
6306 @item -ftest-coverage
6307 @opindex ftest-coverage
6308 Produce a notes file that the @command{gcov} code-coverage utility
6309 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6310 show program coverage. Each source file's note file is called
6311 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6312 above for a description of @var{auxname} and instructions on how to
6313 generate test coverage data. Coverage data matches the source files
6314 more closely if you do not optimize.
6316 @item -fdbg-cnt-list
6317 @opindex fdbg-cnt-list
6318 Print the name and the counter upper bound for all debug counters.
6321 @item -fdbg-cnt=@var{counter-value-list}
6323 Set the internal debug counter upper bound. @var{counter-value-list}
6324 is a comma-separated list of @var{name}:@var{value} pairs
6325 which sets the upper bound of each debug counter @var{name} to @var{value}.
6326 All debug counters have the initial upper bound of @code{UINT_MAX};
6327 thus @code{dbg_cnt} returns true always unless the upper bound
6328 is set by this option.
6329 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6330 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6332 @item -fenable-@var{kind}-@var{pass}
6333 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6337 This is a set of options that are used to explicitly disable/enable
6338 optimization passes. These options are intended for use for debugging GCC.
6339 Compiler users should use regular options for enabling/disabling
6344 @item -fdisable-ipa-@var{pass}
6345 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6346 statically invoked in the compiler multiple times, the pass name should be
6347 appended with a sequential number starting from 1.
6349 @item -fdisable-rtl-@var{pass}
6350 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6351 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6352 statically invoked in the compiler multiple times, the pass name should be
6353 appended with a sequential number starting from 1. @var{range-list} is a
6354 comma-separated list of function ranges or assembler names. Each range is a number
6355 pair separated by a colon. The range is inclusive in both ends. If the range
6356 is trivial, the number pair can be simplified as a single number. If the
6357 function's call graph node's @var{uid} falls within one of the specified ranges,
6358 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6359 function header of a dump file, and the pass names can be dumped by using
6360 option @option{-fdump-passes}.
6362 @item -fdisable-tree-@var{pass}
6363 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6364 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6367 @item -fenable-ipa-@var{pass}
6368 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6369 statically invoked in the compiler multiple times, the pass name should be
6370 appended with a sequential number starting from 1.
6372 @item -fenable-rtl-@var{pass}
6373 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6374 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6375 description and examples.
6377 @item -fenable-tree-@var{pass}
6378 @itemx -fenable-tree-@var{pass}=@var{range-list}
6379 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6380 of option arguments.
6384 Here are some examples showing uses of these options.
6388 # disable ccp1 for all functions
6390 # disable complete unroll for function whose cgraph node uid is 1
6391 -fenable-tree-cunroll=1
6392 # disable gcse2 for functions at the following ranges [1,1],
6393 # [300,400], and [400,1000]
6394 # disable gcse2 for functions foo and foo2
6395 -fdisable-rtl-gcse2=foo,foo2
6396 # disable early inlining
6397 -fdisable-tree-einline
6398 # disable ipa inlining
6399 -fdisable-ipa-inline
6400 # enable tree full unroll
6401 -fenable-tree-unroll
6405 @item -d@var{letters}
6406 @itemx -fdump-rtl-@var{pass}
6407 @itemx -fdump-rtl-@var{pass}=@var{filename}
6409 @opindex fdump-rtl-@var{pass}
6410 Says to make debugging dumps during compilation at times specified by
6411 @var{letters}. This is used for debugging the RTL-based passes of the
6412 compiler. The file names for most of the dumps are made by appending
6413 a pass number and a word to the @var{dumpname}, and the files are
6414 created in the directory of the output file. In case of
6415 @option{=@var{filename}} option, the dump is output on the given file
6416 instead of the pass numbered dump files. Note that the pass number is
6417 computed statically as passes get registered into the pass manager.
6418 Thus the numbering is not related to the dynamic order of execution of
6419 passes. In particular, a pass installed by a plugin could have a
6420 number over 200 even if it executed quite early. @var{dumpname} is
6421 generated from the name of the output file, if explicitly specified
6422 and it is not an executable, otherwise it is the basename of the
6423 source file. These switches may have different effects when
6424 @option{-E} is used for preprocessing.
6426 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6427 @option{-d} option @var{letters}. Here are the possible
6428 letters for use in @var{pass} and @var{letters}, and their meanings:
6432 @item -fdump-rtl-alignments
6433 @opindex fdump-rtl-alignments
6434 Dump after branch alignments have been computed.
6436 @item -fdump-rtl-asmcons
6437 @opindex fdump-rtl-asmcons
6438 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6440 @item -fdump-rtl-auto_inc_dec
6441 @opindex fdump-rtl-auto_inc_dec
6442 Dump after auto-inc-dec discovery. This pass is only run on
6443 architectures that have auto inc or auto dec instructions.
6445 @item -fdump-rtl-barriers
6446 @opindex fdump-rtl-barriers
6447 Dump after cleaning up the barrier instructions.
6449 @item -fdump-rtl-bbpart
6450 @opindex fdump-rtl-bbpart
6451 Dump after partitioning hot and cold basic blocks.
6453 @item -fdump-rtl-bbro
6454 @opindex fdump-rtl-bbro
6455 Dump after block reordering.
6457 @item -fdump-rtl-btl1
6458 @itemx -fdump-rtl-btl2
6459 @opindex fdump-rtl-btl2
6460 @opindex fdump-rtl-btl2
6461 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6462 after the two branch
6463 target load optimization passes.
6465 @item -fdump-rtl-bypass
6466 @opindex fdump-rtl-bypass
6467 Dump after jump bypassing and control flow optimizations.
6469 @item -fdump-rtl-combine
6470 @opindex fdump-rtl-combine
6471 Dump after the RTL instruction combination pass.
6473 @item -fdump-rtl-compgotos
6474 @opindex fdump-rtl-compgotos
6475 Dump after duplicating the computed gotos.
6477 @item -fdump-rtl-ce1
6478 @itemx -fdump-rtl-ce2
6479 @itemx -fdump-rtl-ce3
6480 @opindex fdump-rtl-ce1
6481 @opindex fdump-rtl-ce2
6482 @opindex fdump-rtl-ce3
6483 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6484 @option{-fdump-rtl-ce3} enable dumping after the three
6485 if conversion passes.
6487 @item -fdump-rtl-cprop_hardreg
6488 @opindex fdump-rtl-cprop_hardreg
6489 Dump after hard register copy propagation.
6491 @item -fdump-rtl-csa
6492 @opindex fdump-rtl-csa
6493 Dump after combining stack adjustments.
6495 @item -fdump-rtl-cse1
6496 @itemx -fdump-rtl-cse2
6497 @opindex fdump-rtl-cse1
6498 @opindex fdump-rtl-cse2
6499 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6500 the two common subexpression elimination passes.
6502 @item -fdump-rtl-dce
6503 @opindex fdump-rtl-dce
6504 Dump after the standalone dead code elimination passes.
6506 @item -fdump-rtl-dbr
6507 @opindex fdump-rtl-dbr
6508 Dump after delayed branch scheduling.
6510 @item -fdump-rtl-dce1
6511 @itemx -fdump-rtl-dce2
6512 @opindex fdump-rtl-dce1
6513 @opindex fdump-rtl-dce2
6514 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6515 the two dead store elimination passes.
6518 @opindex fdump-rtl-eh
6519 Dump after finalization of EH handling code.
6521 @item -fdump-rtl-eh_ranges
6522 @opindex fdump-rtl-eh_ranges
6523 Dump after conversion of EH handling range regions.
6525 @item -fdump-rtl-expand
6526 @opindex fdump-rtl-expand
6527 Dump after RTL generation.
6529 @item -fdump-rtl-fwprop1
6530 @itemx -fdump-rtl-fwprop2
6531 @opindex fdump-rtl-fwprop1
6532 @opindex fdump-rtl-fwprop2
6533 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6534 dumping after the two forward propagation passes.
6536 @item -fdump-rtl-gcse1
6537 @itemx -fdump-rtl-gcse2
6538 @opindex fdump-rtl-gcse1
6539 @opindex fdump-rtl-gcse2
6540 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6541 after global common subexpression elimination.
6543 @item -fdump-rtl-init-regs
6544 @opindex fdump-rtl-init-regs
6545 Dump after the initialization of the registers.
6547 @item -fdump-rtl-initvals
6548 @opindex fdump-rtl-initvals
6549 Dump after the computation of the initial value sets.
6551 @item -fdump-rtl-into_cfglayout
6552 @opindex fdump-rtl-into_cfglayout
6553 Dump after converting to cfglayout mode.
6555 @item -fdump-rtl-ira
6556 @opindex fdump-rtl-ira
6557 Dump after iterated register allocation.
6559 @item -fdump-rtl-jump
6560 @opindex fdump-rtl-jump
6561 Dump after the second jump optimization.
6563 @item -fdump-rtl-loop2
6564 @opindex fdump-rtl-loop2
6565 @option{-fdump-rtl-loop2} enables dumping after the rtl
6566 loop optimization passes.
6568 @item -fdump-rtl-mach
6569 @opindex fdump-rtl-mach
6570 Dump after performing the machine dependent reorganization pass, if that
6573 @item -fdump-rtl-mode_sw
6574 @opindex fdump-rtl-mode_sw
6575 Dump after removing redundant mode switches.
6577 @item -fdump-rtl-rnreg
6578 @opindex fdump-rtl-rnreg
6579 Dump after register renumbering.
6581 @item -fdump-rtl-outof_cfglayout
6582 @opindex fdump-rtl-outof_cfglayout
6583 Dump after converting from cfglayout mode.
6585 @item -fdump-rtl-peephole2
6586 @opindex fdump-rtl-peephole2
6587 Dump after the peephole pass.
6589 @item -fdump-rtl-postreload
6590 @opindex fdump-rtl-postreload
6591 Dump after post-reload optimizations.
6593 @item -fdump-rtl-pro_and_epilogue
6594 @opindex fdump-rtl-pro_and_epilogue
6595 Dump after generating the function prologues and epilogues.
6597 @item -fdump-rtl-sched1
6598 @itemx -fdump-rtl-sched2
6599 @opindex fdump-rtl-sched1
6600 @opindex fdump-rtl-sched2
6601 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6602 after the basic block scheduling passes.
6604 @item -fdump-rtl-ree
6605 @opindex fdump-rtl-ree
6606 Dump after sign/zero extension elimination.
6608 @item -fdump-rtl-seqabstr
6609 @opindex fdump-rtl-seqabstr
6610 Dump after common sequence discovery.
6612 @item -fdump-rtl-shorten
6613 @opindex fdump-rtl-shorten
6614 Dump after shortening branches.
6616 @item -fdump-rtl-sibling
6617 @opindex fdump-rtl-sibling
6618 Dump after sibling call optimizations.
6620 @item -fdump-rtl-split1
6621 @itemx -fdump-rtl-split2
6622 @itemx -fdump-rtl-split3
6623 @itemx -fdump-rtl-split4
6624 @itemx -fdump-rtl-split5
6625 @opindex fdump-rtl-split1
6626 @opindex fdump-rtl-split2
6627 @opindex fdump-rtl-split3
6628 @opindex fdump-rtl-split4
6629 @opindex fdump-rtl-split5
6630 These options enable dumping after five rounds of
6631 instruction splitting.
6633 @item -fdump-rtl-sms
6634 @opindex fdump-rtl-sms
6635 Dump after modulo scheduling. This pass is only run on some
6638 @item -fdump-rtl-stack
6639 @opindex fdump-rtl-stack
6640 Dump after conversion from GCC's ``flat register file'' registers to the
6641 x87's stack-like registers. This pass is only run on x86 variants.
6643 @item -fdump-rtl-subreg1
6644 @itemx -fdump-rtl-subreg2
6645 @opindex fdump-rtl-subreg1
6646 @opindex fdump-rtl-subreg2
6647 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6648 the two subreg expansion passes.
6650 @item -fdump-rtl-unshare
6651 @opindex fdump-rtl-unshare
6652 Dump after all rtl has been unshared.
6654 @item -fdump-rtl-vartrack
6655 @opindex fdump-rtl-vartrack
6656 Dump after variable tracking.
6658 @item -fdump-rtl-vregs
6659 @opindex fdump-rtl-vregs
6660 Dump after converting virtual registers to hard registers.
6662 @item -fdump-rtl-web
6663 @opindex fdump-rtl-web
6664 Dump after live range splitting.
6666 @item -fdump-rtl-regclass
6667 @itemx -fdump-rtl-subregs_of_mode_init
6668 @itemx -fdump-rtl-subregs_of_mode_finish
6669 @itemx -fdump-rtl-dfinit
6670 @itemx -fdump-rtl-dfinish
6671 @opindex fdump-rtl-regclass
6672 @opindex fdump-rtl-subregs_of_mode_init
6673 @opindex fdump-rtl-subregs_of_mode_finish
6674 @opindex fdump-rtl-dfinit
6675 @opindex fdump-rtl-dfinish
6676 These dumps are defined but always produce empty files.
6679 @itemx -fdump-rtl-all
6681 @opindex fdump-rtl-all
6682 Produce all the dumps listed above.
6686 Annotate the assembler output with miscellaneous debugging information.
6690 Dump all macro definitions, at the end of preprocessing, in addition to
6695 Produce a core dump whenever an error occurs.
6699 Annotate the assembler output with a comment indicating which
6700 pattern and alternative is used. The length of each instruction is
6705 Dump the RTL in the assembler output as a comment before each instruction.
6706 Also turns on @option{-dp} annotation.
6710 Just generate RTL for a function instead of compiling it. Usually used
6711 with @option{-fdump-rtl-expand}.
6715 @opindex fdump-noaddr
6716 When doing debugging dumps, suppress address output. This makes it more
6717 feasible to use diff on debugging dumps for compiler invocations with
6718 different compiler binaries and/or different
6719 text / bss / data / heap / stack / dso start locations.
6722 @opindex freport-bug
6723 Collect and dump debug information into temporary file if ICE in C/C++
6726 @item -fdump-unnumbered
6727 @opindex fdump-unnumbered
6728 When doing debugging dumps, suppress instruction numbers and address output.
6729 This makes it more feasible to use diff on debugging dumps for compiler
6730 invocations with different options, in particular with and without
6733 @item -fdump-unnumbered-links
6734 @opindex fdump-unnumbered-links
6735 When doing debugging dumps (see @option{-d} option above), suppress
6736 instruction numbers for the links to the previous and next instructions
6739 @item -fdump-translation-unit @r{(C++ only)}
6740 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6741 @opindex fdump-translation-unit
6742 Dump a representation of the tree structure for the entire translation
6743 unit to a file. The file name is made by appending @file{.tu} to the
6744 source file name, and the file is created in the same directory as the
6745 output file. If the @samp{-@var{options}} form is used, @var{options}
6746 controls the details of the dump as described for the
6747 @option{-fdump-tree} options.
6749 @item -fdump-class-hierarchy @r{(C++ only)}
6750 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6751 @opindex fdump-class-hierarchy
6752 Dump a representation of each class's hierarchy and virtual function
6753 table layout to a file. The file name is made by appending
6754 @file{.class} to the source file name, and the file is created in the
6755 same directory as the output file. If the @samp{-@var{options}} form
6756 is used, @var{options} controls the details of the dump as described
6757 for the @option{-fdump-tree} options.
6759 @item -fdump-ipa-@var{switch}
6761 Control the dumping at various stages of inter-procedural analysis
6762 language tree to a file. The file name is generated by appending a
6763 switch specific suffix to the source file name, and the file is created
6764 in the same directory as the output file. The following dumps are
6769 Enables all inter-procedural analysis dumps.
6772 Dumps information about call-graph optimization, unused function removal,
6773 and inlining decisions.
6776 Dump after function inlining.
6781 @opindex fdump-passes
6782 Dump the list of optimization passes that are turned on and off by
6783 the current command-line options.
6785 @item -fdump-statistics-@var{option}
6786 @opindex fdump-statistics
6787 Enable and control dumping of pass statistics in a separate file. The
6788 file name is generated by appending a suffix ending in
6789 @samp{.statistics} to the source file name, and the file is created in
6790 the same directory as the output file. If the @samp{-@var{option}}
6791 form is used, @samp{-stats} causes counters to be summed over the
6792 whole compilation unit while @samp{-details} dumps every event as
6793 the passes generate them. The default with no option is to sum
6794 counters for each function compiled.
6796 @item -fdump-tree-@var{switch}
6797 @itemx -fdump-tree-@var{switch}-@var{options}
6798 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6800 Control the dumping at various stages of processing the intermediate
6801 language tree to a file. The file name is generated by appending a
6802 switch-specific suffix to the source file name, and the file is
6803 created in the same directory as the output file. In case of
6804 @option{=@var{filename}} option, the dump is output on the given file
6805 instead of the auto named dump files. If the @samp{-@var{options}}
6806 form is used, @var{options} is a list of @samp{-} separated options
6807 which control the details of the dump. Not all options are applicable
6808 to all dumps; those that are not meaningful are ignored. The
6809 following options are available
6813 Print the address of each node. Usually this is not meaningful as it
6814 changes according to the environment and source file. Its primary use
6815 is for tying up a dump file with a debug environment.
6817 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6818 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6819 use working backward from mangled names in the assembly file.
6821 When dumping front-end intermediate representations, inhibit dumping
6822 of members of a scope or body of a function merely because that scope
6823 has been reached. Only dump such items when they are directly reachable
6826 When dumping pretty-printed trees, this option inhibits dumping the
6827 bodies of control structures.
6829 When dumping RTL, print the RTL in slim (condensed) form instead of
6830 the default LISP-like representation.
6832 Print a raw representation of the tree. By default, trees are
6833 pretty-printed into a C-like representation.
6835 Enable more detailed dumps (not honored by every dump option). Also
6836 include information from the optimization passes.
6838 Enable dumping various statistics about the pass (not honored by every dump
6841 Enable showing basic block boundaries (disabled in raw dumps).
6843 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6844 dump a representation of the control flow graph suitable for viewing with
6845 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6846 the file is pretty-printed as a subgraph, so that GraphViz can render them
6847 all in a single plot.
6849 This option currently only works for RTL dumps, and the RTL is always
6850 dumped in slim form.
6852 Enable showing virtual operands for every statement.
6854 Enable showing line numbers for statements.
6856 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6858 Enable showing the tree dump for each statement.
6860 Enable showing the EH region number holding each statement.
6862 Enable showing scalar evolution analysis details.
6864 Enable showing optimization information (only available in certain
6867 Enable showing missed optimization information (only available in certain
6870 Enable other detailed optimization information (only available in
6872 @item =@var{filename}
6873 Instead of an auto named dump file, output into the given file
6874 name. The file names @file{stdout} and @file{stderr} are treated
6875 specially and are considered already open standard streams. For
6879 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
6880 -fdump-tree-pre=stderr file.c
6883 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
6884 output on to @file{stderr}. If two conflicting dump filenames are
6885 given for the same pass, then the latter option overrides the earlier
6889 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
6890 and @option{lineno}.
6893 Turn on all optimization options, i.e., @option{optimized},
6894 @option{missed}, and @option{note}.
6897 The following tree dumps are possible:
6901 @opindex fdump-tree-original
6902 Dump before any tree based optimization, to @file{@var{file}.original}.
6905 @opindex fdump-tree-optimized
6906 Dump after all tree based optimization, to @file{@var{file}.optimized}.
6909 @opindex fdump-tree-gimple
6910 Dump each function before and after the gimplification pass to a file. The
6911 file name is made by appending @file{.gimple} to the source file name.
6914 @opindex fdump-tree-cfg
6915 Dump the control flow graph of each function to a file. The file name is
6916 made by appending @file{.cfg} to the source file name.
6919 @opindex fdump-tree-ch
6920 Dump each function after copying loop headers. The file name is made by
6921 appending @file{.ch} to the source file name.
6924 @opindex fdump-tree-ssa
6925 Dump SSA related information to a file. The file name is made by appending
6926 @file{.ssa} to the source file name.
6929 @opindex fdump-tree-alias
6930 Dump aliasing information for each function. The file name is made by
6931 appending @file{.alias} to the source file name.
6934 @opindex fdump-tree-ccp
6935 Dump each function after CCP@. The file name is made by appending
6936 @file{.ccp} to the source file name.
6939 @opindex fdump-tree-storeccp
6940 Dump each function after STORE-CCP@. The file name is made by appending
6941 @file{.storeccp} to the source file name.
6944 @opindex fdump-tree-pre
6945 Dump trees after partial redundancy elimination. The file name is made
6946 by appending @file{.pre} to the source file name.
6949 @opindex fdump-tree-fre
6950 Dump trees after full redundancy elimination. The file name is made
6951 by appending @file{.fre} to the source file name.
6954 @opindex fdump-tree-copyprop
6955 Dump trees after copy propagation. The file name is made
6956 by appending @file{.copyprop} to the source file name.
6958 @item store_copyprop
6959 @opindex fdump-tree-store_copyprop
6960 Dump trees after store copy-propagation. The file name is made
6961 by appending @file{.store_copyprop} to the source file name.
6964 @opindex fdump-tree-dce
6965 Dump each function after dead code elimination. The file name is made by
6966 appending @file{.dce} to the source file name.
6969 @opindex fdump-tree-sra
6970 Dump each function after performing scalar replacement of aggregates. The
6971 file name is made by appending @file{.sra} to the source file name.
6974 @opindex fdump-tree-sink
6975 Dump each function after performing code sinking. The file name is made
6976 by appending @file{.sink} to the source file name.
6979 @opindex fdump-tree-dom
6980 Dump each function after applying dominator tree optimizations. The file
6981 name is made by appending @file{.dom} to the source file name.
6984 @opindex fdump-tree-dse
6985 Dump each function after applying dead store elimination. The file
6986 name is made by appending @file{.dse} to the source file name.
6989 @opindex fdump-tree-phiopt
6990 Dump each function after optimizing PHI nodes into straightline code. The file
6991 name is made by appending @file{.phiopt} to the source file name.
6994 @opindex fdump-tree-forwprop
6995 Dump each function after forward propagating single use variables. The file
6996 name is made by appending @file{.forwprop} to the source file name.
6999 @opindex fdump-tree-copyrename
7000 Dump each function after applying the copy rename optimization. The file
7001 name is made by appending @file{.copyrename} to the source file name.
7004 @opindex fdump-tree-nrv
7005 Dump each function after applying the named return value optimization on
7006 generic trees. The file name is made by appending @file{.nrv} to the source
7010 @opindex fdump-tree-vect
7011 Dump each function after applying vectorization of loops. The file name is
7012 made by appending @file{.vect} to the source file name.
7015 @opindex fdump-tree-slp
7016 Dump each function after applying vectorization of basic blocks. The file name
7017 is made by appending @file{.slp} to the source file name.
7020 @opindex fdump-tree-vrp
7021 Dump each function after Value Range Propagation (VRP). The file name
7022 is made by appending @file{.vrp} to the source file name.
7025 @opindex fdump-tree-all
7026 Enable all the available tree dumps with the flags provided in this option.
7030 @itemx -fopt-info-@var{options}
7031 @itemx -fopt-info-@var{options}=@var{filename}
7033 Controls optimization dumps from various optimization passes. If the
7034 @samp{-@var{options}} form is used, @var{options} is a list of
7035 @samp{-} separated option keywords to select the dump details and
7038 The @var{options} can be divided into two groups: options describing the
7039 verbosity of the dump, and options describing which optimizations
7040 should be included. The options from both the groups can be freely
7041 mixed as they are non-overlapping. However, in case of any conflicts,
7042 the later options override the earlier options on the command
7045 The following options control the dump verbosity:
7049 Print information when an optimization is successfully applied. It is
7050 up to a pass to decide which information is relevant. For example, the
7051 vectorizer passes print the source location of loops which are
7052 successfully vectorized.
7054 Print information about missed optimizations. Individual passes
7055 control which information to include in the output.
7057 Print verbose information about optimizations, such as certain
7058 transformations, more detailed messages about decisions etc.
7060 Print detailed optimization information. This includes
7061 @samp{optimized}, @samp{missed}, and @samp{note}.
7064 One or more of the following option keywords can be used to describe a
7065 group of optimizations:
7069 Enable dumps from all interprocedural optimizations.
7071 Enable dumps from all loop optimizations.
7073 Enable dumps from all inlining optimizations.
7075 Enable dumps from all vectorization optimizations.
7077 Enable dumps from all optimizations. This is a superset of
7078 the optimization groups listed above.
7082 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7083 info about successful optimizations from all the passes.
7085 If the @var{filename} is provided, then the dumps from all the
7086 applicable optimizations are concatenated into the @var{filename}.
7087 Otherwise the dump is output onto @file{stderr}. Though multiple
7088 @option{-fopt-info} options are accepted, only one of them can include
7089 a @var{filename}. If other filenames are provided then all but the
7090 first such option are ignored.
7092 Note that the output @var{filename} is overwritten
7093 in case of multiple translation units. If a combined output from
7094 multiple translation units is desired, @file{stderr} should be used
7097 In the following example, the optimization info is output to
7106 gcc -O3 -fopt-info-missed=missed.all
7110 outputs missed optimization report from all the passes into
7111 @file{missed.all}, and this one:
7114 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7118 prints information about missed optimization opportunities from
7119 vectorization passes on @file{stderr}.
7120 Note that @option{-fopt-info-vec-missed} is equivalent to
7121 @option{-fopt-info-missed-vec}.
7125 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7129 outputs information about missed optimizations as well as
7130 optimized locations from all the inlining passes into
7136 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7140 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7141 in conflict since only one output file is allowed. In this case, only
7142 the first option takes effect and the subsequent options are
7143 ignored. Thus only @file{vec.miss} is produced which contains
7144 dumps from the vectorizer about missed opportunities.
7146 @item -frandom-seed=@var{number}
7147 @opindex frandom-seed
7148 This option provides a seed that GCC uses in place of
7149 random numbers in generating certain symbol names
7150 that have to be different in every compiled file. It is also used to
7151 place unique stamps in coverage data files and the object files that
7152 produce them. You can use the @option{-frandom-seed} option to produce
7153 reproducibly identical object files.
7155 The @var{number} should be different for every file you compile.
7157 @item -fsched-verbose=@var{n}
7158 @opindex fsched-verbose
7159 On targets that use instruction scheduling, this option controls the
7160 amount of debugging output the scheduler prints. This information is
7161 written to standard error, unless @option{-fdump-rtl-sched1} or
7162 @option{-fdump-rtl-sched2} is specified, in which case it is output
7163 to the usual dump listing file, @file{.sched1} or @file{.sched2}
7164 respectively. However for @var{n} greater than nine, the output is
7165 always printed to standard error.
7167 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7168 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7169 For @var{n} greater than one, it also output basic block probabilities,
7170 detailed ready list information and unit/insn info. For @var{n} greater
7171 than two, it includes RTL at abort point, control-flow and regions info.
7172 And for @var{n} over four, @option{-fsched-verbose} also includes
7176 @itemx -save-temps=cwd
7178 Store the usual ``temporary'' intermediate files permanently; place them
7179 in the current directory and name them based on the source file. Thus,
7180 compiling @file{foo.c} with @option{-c -save-temps} produces files
7181 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7182 preprocessed @file{foo.i} output file even though the compiler now
7183 normally uses an integrated preprocessor.
7185 When used in combination with the @option{-x} command-line option,
7186 @option{-save-temps} is sensible enough to avoid over writing an
7187 input source file with the same extension as an intermediate file.
7188 The corresponding intermediate file may be obtained by renaming the
7189 source file before using @option{-save-temps}.
7191 If you invoke GCC in parallel, compiling several different source
7192 files that share a common base name in different subdirectories or the
7193 same source file compiled for multiple output destinations, it is
7194 likely that the different parallel compilers will interfere with each
7195 other, and overwrite the temporary files. For instance:
7198 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7199 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7202 may result in @file{foo.i} and @file{foo.o} being written to
7203 simultaneously by both compilers.
7205 @item -save-temps=obj
7206 @opindex save-temps=obj
7207 Store the usual ``temporary'' intermediate files permanently. If the
7208 @option{-o} option is used, the temporary files are based on the
7209 object file. If the @option{-o} option is not used, the
7210 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7215 gcc -save-temps=obj -c foo.c
7216 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7217 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7221 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7222 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7223 @file{dir2/yfoobar.o}.
7225 @item -time@r{[}=@var{file}@r{]}
7227 Report the CPU time taken by each subprocess in the compilation
7228 sequence. For C source files, this is the compiler proper and assembler
7229 (plus the linker if linking is done).
7231 Without the specification of an output file, the output looks like this:
7238 The first number on each line is the ``user time'', that is time spent
7239 executing the program itself. The second number is ``system time'',
7240 time spent executing operating system routines on behalf of the program.
7241 Both numbers are in seconds.
7243 With the specification of an output file, the output is appended to the
7244 named file, and it looks like this:
7247 0.12 0.01 cc1 @var{options}
7248 0.00 0.01 as @var{options}
7251 The ``user time'' and the ``system time'' are moved before the program
7252 name, and the options passed to the program are displayed, so that one
7253 can later tell what file was being compiled, and with which options.
7255 @item -fvar-tracking
7256 @opindex fvar-tracking
7257 Run variable tracking pass. It computes where variables are stored at each
7258 position in code. Better debugging information is then generated
7259 (if the debugging information format supports this information).
7261 It is enabled by default when compiling with optimization (@option{-Os},
7262 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7263 the debug info format supports it.
7265 @item -fvar-tracking-assignments
7266 @opindex fvar-tracking-assignments
7267 @opindex fno-var-tracking-assignments
7268 Annotate assignments to user variables early in the compilation and
7269 attempt to carry the annotations over throughout the compilation all the
7270 way to the end, in an attempt to improve debug information while
7271 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7273 It can be enabled even if var-tracking is disabled, in which case
7274 annotations are created and maintained, but discarded at the end.
7275 By default, this flag is enabled together with @option{-fvar-tracking},
7276 except when selective scheduling is enabled.
7278 @item -fvar-tracking-assignments-toggle
7279 @opindex fvar-tracking-assignments-toggle
7280 @opindex fno-var-tracking-assignments-toggle
7281 Toggle @option{-fvar-tracking-assignments}, in the same way that
7282 @option{-gtoggle} toggles @option{-g}.
7284 @item -print-file-name=@var{library}
7285 @opindex print-file-name
7286 Print the full absolute name of the library file @var{library} that
7287 would be used when linking---and don't do anything else. With this
7288 option, GCC does not compile or link anything; it just prints the
7291 @item -print-multi-directory
7292 @opindex print-multi-directory
7293 Print the directory name corresponding to the multilib selected by any
7294 other switches present in the command line. This directory is supposed
7295 to exist in @env{GCC_EXEC_PREFIX}.
7297 @item -print-multi-lib
7298 @opindex print-multi-lib
7299 Print the mapping from multilib directory names to compiler switches
7300 that enable them. The directory name is separated from the switches by
7301 @samp{;}, and each switch starts with an @samp{@@} instead of the
7302 @samp{-}, without spaces between multiple switches. This is supposed to
7303 ease shell processing.
7305 @item -print-multi-os-directory
7306 @opindex print-multi-os-directory
7307 Print the path to OS libraries for the selected
7308 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7309 present in the @file{lib} subdirectory and no multilibs are used, this is
7310 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7311 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7312 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7313 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7315 @item -print-multiarch
7316 @opindex print-multiarch
7317 Print the path to OS libraries for the selected multiarch,
7318 relative to some @file{lib} subdirectory.
7320 @item -print-prog-name=@var{program}
7321 @opindex print-prog-name
7322 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7324 @item -print-libgcc-file-name
7325 @opindex print-libgcc-file-name
7326 Same as @option{-print-file-name=libgcc.a}.
7328 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7329 but you do want to link with @file{libgcc.a}. You can do:
7332 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7335 @item -print-search-dirs
7336 @opindex print-search-dirs
7337 Print the name of the configured installation directory and a list of
7338 program and library directories @command{gcc} searches---and don't do anything else.
7340 This is useful when @command{gcc} prints the error message
7341 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7342 To resolve this you either need to put @file{cpp0} and the other compiler
7343 components where @command{gcc} expects to find them, or you can set the environment
7344 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7345 Don't forget the trailing @samp{/}.
7346 @xref{Environment Variables}.
7348 @item -print-sysroot
7349 @opindex print-sysroot
7350 Print the target sysroot directory that is used during
7351 compilation. This is the target sysroot specified either at configure
7352 time or using the @option{--sysroot} option, possibly with an extra
7353 suffix that depends on compilation options. If no target sysroot is
7354 specified, the option prints nothing.
7356 @item -print-sysroot-headers-suffix
7357 @opindex print-sysroot-headers-suffix
7358 Print the suffix added to the target sysroot when searching for
7359 headers, or give an error if the compiler is not configured with such
7360 a suffix---and don't do anything else.
7363 @opindex dumpmachine
7364 Print the compiler's target machine (for example,
7365 @samp{i686-pc-linux-gnu})---and don't do anything else.
7368 @opindex dumpversion
7369 Print the compiler version (for example, @code{3.0})---and don't do
7374 Print the compiler's built-in specs---and don't do anything else. (This
7375 is used when GCC itself is being built.) @xref{Spec Files}.
7377 @item -fno-eliminate-unused-debug-types
7378 @opindex feliminate-unused-debug-types
7379 @opindex fno-eliminate-unused-debug-types
7380 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7381 output for types that are nowhere used in the source file being compiled.
7382 Sometimes it is useful to have GCC emit debugging
7383 information for all types declared in a compilation
7384 unit, regardless of whether or not they are actually used
7385 in that compilation unit, for example
7386 if, in the debugger, you want to cast a value to a type that is
7387 not actually used in your program (but is declared). More often,
7388 however, this results in a significant amount of wasted space.
7391 @node Optimize Options
7392 @section Options That Control Optimization
7393 @cindex optimize options
7394 @cindex options, optimization
7396 These options control various sorts of optimizations.
7398 Without any optimization option, the compiler's goal is to reduce the
7399 cost of compilation and to make debugging produce the expected
7400 results. Statements are independent: if you stop the program with a
7401 breakpoint between statements, you can then assign a new value to any
7402 variable or change the program counter to any other statement in the
7403 function and get exactly the results you expect from the source
7406 Turning on optimization flags makes the compiler attempt to improve
7407 the performance and/or code size at the expense of compilation time
7408 and possibly the ability to debug the program.
7410 The compiler performs optimization based on the knowledge it has of the
7411 program. Compiling multiple files at once to a single output file mode allows
7412 the compiler to use information gained from all of the files when compiling
7415 Not all optimizations are controlled directly by a flag. Only
7416 optimizations that have a flag are listed in this section.
7418 Most optimizations are only enabled if an @option{-O} level is set on
7419 the command line. Otherwise they are disabled, even if individual
7420 optimization flags are specified.
7422 Depending on the target and how GCC was configured, a slightly different
7423 set of optimizations may be enabled at each @option{-O} level than
7424 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7425 to find out the exact set of optimizations that are enabled at each level.
7426 @xref{Overall Options}, for examples.
7433 Optimize. Optimizing compilation takes somewhat more time, and a lot
7434 more memory for a large function.
7436 With @option{-O}, the compiler tries to reduce code size and execution
7437 time, without performing any optimizations that take a great deal of
7440 @option{-O} turns on the following optimization flags:
7443 -fbranch-count-reg @gol
7444 -fcombine-stack-adjustments @gol
7446 -fcprop-registers @gol
7449 -fdelayed-branch @gol
7451 -fforward-propagate @gol
7452 -fguess-branch-probability @gol
7453 -fif-conversion2 @gol
7454 -fif-conversion @gol
7455 -finline-functions-called-once @gol
7456 -fipa-pure-const @gol
7458 -fipa-reference @gol
7459 -fmerge-constants @gol
7460 -fmove-loop-invariants @gol
7462 -fsplit-wide-types @gol
7467 -ftree-copy-prop @gol
7468 -ftree-copyrename @gol
7470 -ftree-dominator-opts @gol
7472 -ftree-forwprop @gol
7482 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7483 where doing so does not interfere with debugging.
7487 Optimize even more. GCC performs nearly all supported optimizations
7488 that do not involve a space-speed tradeoff.
7489 As compared to @option{-O}, this option increases both compilation time
7490 and the performance of the generated code.
7492 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7493 also turns on the following optimization flags:
7494 @gccoptlist{-fthread-jumps @gol
7495 -falign-functions -falign-jumps @gol
7496 -falign-loops -falign-labels @gol
7499 -fcse-follow-jumps -fcse-skip-blocks @gol
7500 -fdelete-null-pointer-checks @gol
7501 -fdevirtualize -fdevirtualize-speculatively @gol
7502 -fexpensive-optimizations @gol
7503 -fgcse -fgcse-lm @gol
7504 -fhoist-adjacent-loads @gol
7505 -finline-small-functions @gol
7506 -findirect-inlining @gol
7508 -fipa-cp-alignment @gol
7511 -fisolate-erroneous-paths-dereference @gol
7513 -foptimize-sibling-calls @gol
7514 -foptimize-strlen @gol
7515 -fpartial-inlining @gol
7517 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7518 -frerun-cse-after-loop @gol
7519 -fsched-interblock -fsched-spec @gol
7520 -fschedule-insns -fschedule-insns2 @gol
7521 -fstrict-aliasing -fstrict-overflow @gol
7522 -ftree-builtin-call-dce @gol
7523 -ftree-switch-conversion -ftree-tail-merge @gol
7528 Please note the warning under @option{-fgcse} about
7529 invoking @option{-O2} on programs that use computed gotos.
7533 Optimize yet more. @option{-O3} turns on all optimizations specified
7534 by @option{-O2} and also turns on the @option{-finline-functions},
7535 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7536 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7537 @option{-ftree-loop-distribute-patterns},
7538 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7539 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7543 Reduce compilation time and make debugging produce the expected
7544 results. This is the default.
7548 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7549 do not typically increase code size. It also performs further
7550 optimizations designed to reduce code size.
7552 @option{-Os} disables the following optimization flags:
7553 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7554 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7555 -fprefetch-loop-arrays}
7559 Disregard strict standards compliance. @option{-Ofast} enables all
7560 @option{-O3} optimizations. It also enables optimizations that are not
7561 valid for all standard-compliant programs.
7562 It turns on @option{-ffast-math} and the Fortran-specific
7563 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7567 Optimize debugging experience. @option{-Og} enables optimizations
7568 that do not interfere with debugging. It should be the optimization
7569 level of choice for the standard edit-compile-debug cycle, offering
7570 a reasonable level of optimization while maintaining fast compilation
7571 and a good debugging experience.
7573 If you use multiple @option{-O} options, with or without level numbers,
7574 the last such option is the one that is effective.
7577 Options of the form @option{-f@var{flag}} specify machine-independent
7578 flags. Most flags have both positive and negative forms; the negative
7579 form of @option{-ffoo} is @option{-fno-foo}. In the table
7580 below, only one of the forms is listed---the one you typically
7581 use. You can figure out the other form by either removing @samp{no-}
7584 The following options control specific optimizations. They are either
7585 activated by @option{-O} options or are related to ones that are. You
7586 can use the following flags in the rare cases when ``fine-tuning'' of
7587 optimizations to be performed is desired.
7590 @item -fno-defer-pop
7591 @opindex fno-defer-pop
7592 Always pop the arguments to each function call as soon as that function
7593 returns. For machines that must pop arguments after a function call,
7594 the compiler normally lets arguments accumulate on the stack for several
7595 function calls and pops them all at once.
7597 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7599 @item -fforward-propagate
7600 @opindex fforward-propagate
7601 Perform a forward propagation pass on RTL@. The pass tries to combine two
7602 instructions and checks if the result can be simplified. If loop unrolling
7603 is active, two passes are performed and the second is scheduled after
7606 This option is enabled by default at optimization levels @option{-O},
7607 @option{-O2}, @option{-O3}, @option{-Os}.
7609 @item -ffp-contract=@var{style}
7610 @opindex ffp-contract
7611 @option{-ffp-contract=off} disables floating-point expression contraction.
7612 @option{-ffp-contract=fast} enables floating-point expression contraction
7613 such as forming of fused multiply-add operations if the target has
7614 native support for them.
7615 @option{-ffp-contract=on} enables floating-point expression contraction
7616 if allowed by the language standard. This is currently not implemented
7617 and treated equal to @option{-ffp-contract=off}.
7619 The default is @option{-ffp-contract=fast}.
7621 @item -fomit-frame-pointer
7622 @opindex fomit-frame-pointer
7623 Don't keep the frame pointer in a register for functions that
7624 don't need one. This avoids the instructions to save, set up and
7625 restore frame pointers; it also makes an extra register available
7626 in many functions. @strong{It also makes debugging impossible on
7629 On some machines, such as the VAX, this flag has no effect, because
7630 the standard calling sequence automatically handles the frame pointer
7631 and nothing is saved by pretending it doesn't exist. The
7632 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7633 whether a target machine supports this flag. @xref{Registers,,Register
7634 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7636 The default setting (when not optimizing for
7637 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7638 @option{-fomit-frame-pointer}. You can configure GCC with the
7639 @option{--enable-frame-pointer} configure option to change the default.
7641 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7643 @item -foptimize-sibling-calls
7644 @opindex foptimize-sibling-calls
7645 Optimize sibling and tail recursive calls.
7647 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7649 @item -foptimize-strlen
7650 @opindex foptimize-strlen
7651 Optimize various standard C string functions (e.g. @code{strlen},
7652 @code{strchr} or @code{strcpy}) and
7653 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7655 Enabled at levels @option{-O2}, @option{-O3}.
7659 Do not expand any functions inline apart from those marked with
7660 the @code{always_inline} attribute. This is the default when not
7663 Single functions can be exempted from inlining by marking them
7664 with the @code{noinline} attribute.
7666 @item -finline-small-functions
7667 @opindex finline-small-functions
7668 Integrate functions into their callers when their body is smaller than expected
7669 function call code (so overall size of program gets smaller). The compiler
7670 heuristically decides which functions are simple enough to be worth integrating
7671 in this way. This inlining applies to all functions, even those not declared
7674 Enabled at level @option{-O2}.
7676 @item -findirect-inlining
7677 @opindex findirect-inlining
7678 Inline also indirect calls that are discovered to be known at compile
7679 time thanks to previous inlining. This option has any effect only
7680 when inlining itself is turned on by the @option{-finline-functions}
7681 or @option{-finline-small-functions} options.
7683 Enabled at level @option{-O2}.
7685 @item -finline-functions
7686 @opindex finline-functions
7687 Consider all functions for inlining, even if they are not declared inline.
7688 The compiler heuristically decides which functions are worth integrating
7691 If all calls to a given function are integrated, and the function is
7692 declared @code{static}, then the function is normally not output as
7693 assembler code in its own right.
7695 Enabled at level @option{-O3}.
7697 @item -finline-functions-called-once
7698 @opindex finline-functions-called-once
7699 Consider all @code{static} functions called once for inlining into their
7700 caller even if they are not marked @code{inline}. If a call to a given
7701 function is integrated, then the function is not output as assembler code
7704 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7706 @item -fearly-inlining
7707 @opindex fearly-inlining
7708 Inline functions marked by @code{always_inline} and functions whose body seems
7709 smaller than the function call overhead early before doing
7710 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7711 makes profiling significantly cheaper and usually inlining faster on programs
7712 having large chains of nested wrapper functions.
7718 Perform interprocedural scalar replacement of aggregates, removal of
7719 unused parameters and replacement of parameters passed by reference
7720 by parameters passed by value.
7722 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7724 @item -finline-limit=@var{n}
7725 @opindex finline-limit
7726 By default, GCC limits the size of functions that can be inlined. This flag
7727 allows coarse control of this limit. @var{n} is the size of functions that
7728 can be inlined in number of pseudo instructions.
7730 Inlining is actually controlled by a number of parameters, which may be
7731 specified individually by using @option{--param @var{name}=@var{value}}.
7732 The @option{-finline-limit=@var{n}} option sets some of these parameters
7736 @item max-inline-insns-single
7737 is set to @var{n}/2.
7738 @item max-inline-insns-auto
7739 is set to @var{n}/2.
7742 See below for a documentation of the individual
7743 parameters controlling inlining and for the defaults of these parameters.
7745 @emph{Note:} there may be no value to @option{-finline-limit} that results
7746 in default behavior.
7748 @emph{Note:} pseudo instruction represents, in this particular context, an
7749 abstract measurement of function's size. In no way does it represent a count
7750 of assembly instructions and as such its exact meaning might change from one
7751 release to an another.
7753 @item -fno-keep-inline-dllexport
7754 @opindex fno-keep-inline-dllexport
7755 This is a more fine-grained version of @option{-fkeep-inline-functions},
7756 which applies only to functions that are declared using the @code{dllexport}
7757 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7760 @item -fkeep-inline-functions
7761 @opindex fkeep-inline-functions
7762 In C, emit @code{static} functions that are declared @code{inline}
7763 into the object file, even if the function has been inlined into all
7764 of its callers. This switch does not affect functions using the
7765 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7766 inline functions into the object file.
7768 @item -fkeep-static-consts
7769 @opindex fkeep-static-consts
7770 Emit variables declared @code{static const} when optimization isn't turned
7771 on, even if the variables aren't referenced.
7773 GCC enables this option by default. If you want to force the compiler to
7774 check if a variable is referenced, regardless of whether or not
7775 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7777 @item -fmerge-constants
7778 @opindex fmerge-constants
7779 Attempt to merge identical constants (string constants and floating-point
7780 constants) across compilation units.
7782 This option is the default for optimized compilation if the assembler and
7783 linker support it. Use @option{-fno-merge-constants} to inhibit this
7786 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7788 @item -fmerge-all-constants
7789 @opindex fmerge-all-constants
7790 Attempt to merge identical constants and identical variables.
7792 This option implies @option{-fmerge-constants}. In addition to
7793 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7794 arrays or initialized constant variables with integral or floating-point
7795 types. Languages like C or C++ require each variable, including multiple
7796 instances of the same variable in recursive calls, to have distinct locations,
7797 so using this option results in non-conforming
7800 @item -fmodulo-sched
7801 @opindex fmodulo-sched
7802 Perform swing modulo scheduling immediately before the first scheduling
7803 pass. This pass looks at innermost loops and reorders their
7804 instructions by overlapping different iterations.
7806 @item -fmodulo-sched-allow-regmoves
7807 @opindex fmodulo-sched-allow-regmoves
7808 Perform more aggressive SMS-based modulo scheduling with register moves
7809 allowed. By setting this flag certain anti-dependences edges are
7810 deleted, which triggers the generation of reg-moves based on the
7811 life-range analysis. This option is effective only with
7812 @option{-fmodulo-sched} enabled.
7814 @item -fno-branch-count-reg
7815 @opindex fno-branch-count-reg
7816 Do not use ``decrement and branch'' instructions on a count register,
7817 but instead generate a sequence of instructions that decrement a
7818 register, compare it against zero, then branch based upon the result.
7819 This option is only meaningful on architectures that support such
7820 instructions, which include x86, PowerPC, IA-64 and S/390.
7822 Enabled by default at @option{-O1} and higher.
7824 The default is @option{-fbranch-count-reg}.
7826 @item -fno-function-cse
7827 @opindex fno-function-cse
7828 Do not put function addresses in registers; make each instruction that
7829 calls a constant function contain the function's address explicitly.
7831 This option results in less efficient code, but some strange hacks
7832 that alter the assembler output may be confused by the optimizations
7833 performed when this option is not used.
7835 The default is @option{-ffunction-cse}
7837 @item -fno-zero-initialized-in-bss
7838 @opindex fno-zero-initialized-in-bss
7839 If the target supports a BSS section, GCC by default puts variables that
7840 are initialized to zero into BSS@. This can save space in the resulting
7843 This option turns off this behavior because some programs explicitly
7844 rely on variables going to the data section---e.g., so that the
7845 resulting executable can find the beginning of that section and/or make
7846 assumptions based on that.
7848 The default is @option{-fzero-initialized-in-bss}.
7850 @item -fthread-jumps
7851 @opindex fthread-jumps
7852 Perform optimizations that check to see if a jump branches to a
7853 location where another comparison subsumed by the first is found. If
7854 so, the first branch is redirected to either the destination of the
7855 second branch or a point immediately following it, depending on whether
7856 the condition is known to be true or false.
7858 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7860 @item -fsplit-wide-types
7861 @opindex fsplit-wide-types
7862 When using a type that occupies multiple registers, such as @code{long
7863 long} on a 32-bit system, split the registers apart and allocate them
7864 independently. This normally generates better code for those types,
7865 but may make debugging more difficult.
7867 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
7870 @item -fcse-follow-jumps
7871 @opindex fcse-follow-jumps
7872 In common subexpression elimination (CSE), scan through jump instructions
7873 when the target of the jump is not reached by any other path. For
7874 example, when CSE encounters an @code{if} statement with an
7875 @code{else} clause, CSE follows the jump when the condition
7878 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7880 @item -fcse-skip-blocks
7881 @opindex fcse-skip-blocks
7882 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
7883 follow jumps that conditionally skip over blocks. When CSE
7884 encounters a simple @code{if} statement with no else clause,
7885 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
7886 body of the @code{if}.
7888 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7890 @item -frerun-cse-after-loop
7891 @opindex frerun-cse-after-loop
7892 Re-run common subexpression elimination after loop optimizations are
7895 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7899 Perform a global common subexpression elimination pass.
7900 This pass also performs global constant and copy propagation.
7902 @emph{Note:} When compiling a program using computed gotos, a GCC
7903 extension, you may get better run-time performance if you disable
7904 the global common subexpression elimination pass by adding
7905 @option{-fno-gcse} to the command line.
7907 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7911 When @option{-fgcse-lm} is enabled, global common subexpression elimination
7912 attempts to move loads that are only killed by stores into themselves. This
7913 allows a loop containing a load/store sequence to be changed to a load outside
7914 the loop, and a copy/store within the loop.
7916 Enabled by default when @option{-fgcse} is enabled.
7920 When @option{-fgcse-sm} is enabled, a store motion pass is run after
7921 global common subexpression elimination. This pass attempts to move
7922 stores out of loops. When used in conjunction with @option{-fgcse-lm},
7923 loops containing a load/store sequence can be changed to a load before
7924 the loop and a store after the loop.
7926 Not enabled at any optimization level.
7930 When @option{-fgcse-las} is enabled, the global common subexpression
7931 elimination pass eliminates redundant loads that come after stores to the
7932 same memory location (both partial and full redundancies).
7934 Not enabled at any optimization level.
7936 @item -fgcse-after-reload
7937 @opindex fgcse-after-reload
7938 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
7939 pass is performed after reload. The purpose of this pass is to clean up
7942 @item -faggressive-loop-optimizations
7943 @opindex faggressive-loop-optimizations
7944 This option tells the loop optimizer to use language constraints to
7945 derive bounds for the number of iterations of a loop. This assumes that
7946 loop code does not invoke undefined behavior by for example causing signed
7947 integer overflows or out-of-bound array accesses. The bounds for the
7948 number of iterations of a loop are used to guide loop unrolling and peeling
7949 and loop exit test optimizations.
7950 This option is enabled by default.
7952 @item -funsafe-loop-optimizations
7953 @opindex funsafe-loop-optimizations
7954 This option tells the loop optimizer to assume that loop indices do not
7955 overflow, and that loops with nontrivial exit condition are not
7956 infinite. This enables a wider range of loop optimizations even if
7957 the loop optimizer itself cannot prove that these assumptions are valid.
7958 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
7959 if it finds this kind of loop.
7961 @item -fcrossjumping
7962 @opindex fcrossjumping
7963 Perform cross-jumping transformation.
7964 This transformation unifies equivalent code and saves code size. The
7965 resulting code may or may not perform better than without cross-jumping.
7967 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7969 @item -fauto-inc-dec
7970 @opindex fauto-inc-dec
7971 Combine increments or decrements of addresses with memory accesses.
7972 This pass is always skipped on architectures that do not have
7973 instructions to support this. Enabled by default at @option{-O} and
7974 higher on architectures that support this.
7978 Perform dead code elimination (DCE) on RTL@.
7979 Enabled by default at @option{-O} and higher.
7983 Perform dead store elimination (DSE) on RTL@.
7984 Enabled by default at @option{-O} and higher.
7986 @item -fif-conversion
7987 @opindex fif-conversion
7988 Attempt to transform conditional jumps into branch-less equivalents. This
7989 includes use of conditional moves, min, max, set flags and abs instructions, and
7990 some tricks doable by standard arithmetics. The use of conditional execution
7991 on chips where it is available is controlled by @option{-fif-conversion2}.
7993 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7995 @item -fif-conversion2
7996 @opindex fif-conversion2
7997 Use conditional execution (where available) to transform conditional jumps into
7998 branch-less equivalents.
8000 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8002 @item -fdeclone-ctor-dtor
8003 @opindex fdeclone-ctor-dtor
8004 The C++ ABI requires multiple entry points for constructors and
8005 destructors: one for a base subobject, one for a complete object, and
8006 one for a virtual destructor that calls operator delete afterwards.
8007 For a hierarchy with virtual bases, the base and complete variants are
8008 clones, which means two copies of the function. With this option, the
8009 base and complete variants are changed to be thunks that call a common
8012 Enabled by @option{-Os}.
8014 @item -fdelete-null-pointer-checks
8015 @opindex fdelete-null-pointer-checks
8016 Assume that programs cannot safely dereference null pointers, and that
8017 no code or data element resides at address zero.
8018 This option enables simple constant
8019 folding optimizations at all optimization levels. In addition, other
8020 optimization passes in GCC use this flag to control global dataflow
8021 analyses that eliminate useless checks for null pointers; these assume
8022 that a memory access to address zero always results in a trap, so
8023 that if a pointer is checked after it has already been dereferenced,
8026 Note however that in some environments this assumption is not true.
8027 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8028 for programs that depend on that behavior.
8030 This option is enabled by default on most targets. On Nios II ELF, it
8031 defaults to off. On AVR and CR16, this option is completely disabled.
8033 Passes that use the dataflow information
8034 are enabled independently at different optimization levels.
8036 @item -fdevirtualize
8037 @opindex fdevirtualize
8038 Attempt to convert calls to virtual functions to direct calls. This
8039 is done both within a procedure and interprocedurally as part of
8040 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8041 propagation (@option{-fipa-cp}).
8042 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8044 @item -fdevirtualize-speculatively
8045 @opindex fdevirtualize-speculatively
8046 Attempt to convert calls to virtual functions to speculative direct calls.
8047 Based on the analysis of the type inheritance graph, determine for a given call
8048 the set of likely targets. If the set is small, preferably of size 1, change
8049 the call into a conditional deciding between direct and indirect calls. The
8050 speculative calls enable more optimizations, such as inlining. When they seem
8051 useless after further optimization, they are converted back into original form.
8053 @item -fdevirtualize-at-ltrans
8054 @opindex fdevirtualize-at-ltrans
8055 Stream extra information needed for aggressive devirtualization when running
8056 the link-time optimizer in local transformation mode.
8057 This option enables more devirtualization but
8058 significantly increases the size of streamed data. For this reason it is
8059 disabled by default.
8061 @item -fexpensive-optimizations
8062 @opindex fexpensive-optimizations
8063 Perform a number of minor optimizations that are relatively expensive.
8065 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8069 Attempt to remove redundant extension instructions. This is especially
8070 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8071 registers after writing to their lower 32-bit half.
8073 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8074 @option{-O3}, @option{-Os}.
8076 @item -fno-lifetime-dse
8077 @opindex fno-lifetime-dse
8078 In C++ the value of an object is only affected by changes within its
8079 lifetime: when the constructor begins, the object has an indeterminate
8080 value, and any changes during the lifetime of the object are dead when
8081 the object is destroyed. Normally dead store elimination will take
8082 advantage of this; if your code relies on the value of the object
8083 storage persisting beyond the lifetime of the object, you can use this
8084 flag to disable this optimization.
8086 @item -flive-range-shrinkage
8087 @opindex flive-range-shrinkage
8088 Attempt to decrease register pressure through register live range
8089 shrinkage. This is helpful for fast processors with small or moderate
8092 @item -fira-algorithm=@var{algorithm}
8093 @opindex fira-algorithm
8094 Use the specified coloring algorithm for the integrated register
8095 allocator. The @var{algorithm} argument can be @samp{priority}, which
8096 specifies Chow's priority coloring, or @samp{CB}, which specifies
8097 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8098 for all architectures, but for those targets that do support it, it is
8099 the default because it generates better code.
8101 @item -fira-region=@var{region}
8102 @opindex fira-region
8103 Use specified regions for the integrated register allocator. The
8104 @var{region} argument should be one of the following:
8109 Use all loops as register allocation regions.
8110 This can give the best results for machines with a small and/or
8111 irregular register set.
8114 Use all loops except for loops with small register pressure
8115 as the regions. This value usually gives
8116 the best results in most cases and for most architectures,
8117 and is enabled by default when compiling with optimization for speed
8118 (@option{-O}, @option{-O2}, @dots{}).
8121 Use all functions as a single region.
8122 This typically results in the smallest code size, and is enabled by default for
8123 @option{-Os} or @option{-O0}.
8127 @item -fira-hoist-pressure
8128 @opindex fira-hoist-pressure
8129 Use IRA to evaluate register pressure in the code hoisting pass for
8130 decisions to hoist expressions. This option usually results in smaller
8131 code, but it can slow the compiler down.
8133 This option is enabled at level @option{-Os} for all targets.
8135 @item -fira-loop-pressure
8136 @opindex fira-loop-pressure
8137 Use IRA to evaluate register pressure in loops for decisions to move
8138 loop invariants. This option usually results in generation
8139 of faster and smaller code on machines with large register files (>= 32
8140 registers), but it can slow the compiler down.
8142 This option is enabled at level @option{-O3} for some targets.
8144 @item -fno-ira-share-save-slots
8145 @opindex fno-ira-share-save-slots
8146 Disable sharing of stack slots used for saving call-used hard
8147 registers living through a call. Each hard register gets a
8148 separate stack slot, and as a result function stack frames are
8151 @item -fno-ira-share-spill-slots
8152 @opindex fno-ira-share-spill-slots
8153 Disable sharing of stack slots allocated for pseudo-registers. Each
8154 pseudo-register that does not get a hard register gets a separate
8155 stack slot, and as a result function stack frames are larger.
8157 @item -fira-verbose=@var{n}
8158 @opindex fira-verbose
8159 Control the verbosity of the dump file for the integrated register allocator.
8160 The default value is 5. If the value @var{n} is greater or equal to 10,
8161 the dump output is sent to stderr using the same format as @var{n} minus 10.
8165 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8166 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8167 values if it is profitable.
8169 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8171 @item -fdelayed-branch
8172 @opindex fdelayed-branch
8173 If supported for the target machine, attempt to reorder instructions
8174 to exploit instruction slots available after delayed branch
8177 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8179 @item -fschedule-insns
8180 @opindex fschedule-insns
8181 If supported for the target machine, attempt to reorder instructions to
8182 eliminate execution stalls due to required data being unavailable. This
8183 helps machines that have slow floating point or memory load instructions
8184 by allowing other instructions to be issued until the result of the load
8185 or floating-point instruction is required.
8187 Enabled at levels @option{-O2}, @option{-O3}.
8189 @item -fschedule-insns2
8190 @opindex fschedule-insns2
8191 Similar to @option{-fschedule-insns}, but requests an additional pass of
8192 instruction scheduling after register allocation has been done. This is
8193 especially useful on machines with a relatively small number of
8194 registers and where memory load instructions take more than one cycle.
8196 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8198 @item -fno-sched-interblock
8199 @opindex fno-sched-interblock
8200 Don't schedule instructions across basic blocks. This is normally
8201 enabled by default when scheduling before register allocation, i.e.@:
8202 with @option{-fschedule-insns} or at @option{-O2} or higher.
8204 @item -fno-sched-spec
8205 @opindex fno-sched-spec
8206 Don't allow speculative motion of non-load instructions. This is normally
8207 enabled by default when scheduling before register allocation, i.e.@:
8208 with @option{-fschedule-insns} or at @option{-O2} or higher.
8210 @item -fsched-pressure
8211 @opindex fsched-pressure
8212 Enable register pressure sensitive insn scheduling before register
8213 allocation. This only makes sense when scheduling before register
8214 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8215 @option{-O2} or higher. Usage of this option can improve the
8216 generated code and decrease its size by preventing register pressure
8217 increase above the number of available hard registers and subsequent
8218 spills in register allocation.
8220 @item -fsched-spec-load
8221 @opindex fsched-spec-load
8222 Allow speculative motion of some load instructions. This only makes
8223 sense when scheduling before register allocation, i.e.@: with
8224 @option{-fschedule-insns} or at @option{-O2} or higher.
8226 @item -fsched-spec-load-dangerous
8227 @opindex fsched-spec-load-dangerous
8228 Allow speculative motion of more load instructions. This only makes
8229 sense when scheduling before register allocation, i.e.@: with
8230 @option{-fschedule-insns} or at @option{-O2} or higher.
8232 @item -fsched-stalled-insns
8233 @itemx -fsched-stalled-insns=@var{n}
8234 @opindex fsched-stalled-insns
8235 Define how many insns (if any) can be moved prematurely from the queue
8236 of stalled insns into the ready list during the second scheduling pass.
8237 @option{-fno-sched-stalled-insns} means that no insns are moved
8238 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8239 on how many queued insns can be moved prematurely.
8240 @option{-fsched-stalled-insns} without a value is equivalent to
8241 @option{-fsched-stalled-insns=1}.
8243 @item -fsched-stalled-insns-dep
8244 @itemx -fsched-stalled-insns-dep=@var{n}
8245 @opindex fsched-stalled-insns-dep
8246 Define how many insn groups (cycles) are examined for a dependency
8247 on a stalled insn that is a candidate for premature removal from the queue
8248 of stalled insns. This has an effect only during the second scheduling pass,
8249 and only if @option{-fsched-stalled-insns} is used.
8250 @option{-fno-sched-stalled-insns-dep} is equivalent to
8251 @option{-fsched-stalled-insns-dep=0}.
8252 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8253 @option{-fsched-stalled-insns-dep=1}.
8255 @item -fsched2-use-superblocks
8256 @opindex fsched2-use-superblocks
8257 When scheduling after register allocation, use superblock scheduling.
8258 This allows motion across basic block boundaries,
8259 resulting in faster schedules. This option is experimental, as not all machine
8260 descriptions used by GCC model the CPU closely enough to avoid unreliable
8261 results from the algorithm.
8263 This only makes sense when scheduling after register allocation, i.e.@: with
8264 @option{-fschedule-insns2} or at @option{-O2} or higher.
8266 @item -fsched-group-heuristic
8267 @opindex fsched-group-heuristic
8268 Enable the group heuristic in the scheduler. This heuristic favors
8269 the instruction that belongs to a schedule group. This is enabled
8270 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8271 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8273 @item -fsched-critical-path-heuristic
8274 @opindex fsched-critical-path-heuristic
8275 Enable the critical-path heuristic in the scheduler. This heuristic favors
8276 instructions on the critical path. This is enabled by default when
8277 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8278 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8280 @item -fsched-spec-insn-heuristic
8281 @opindex fsched-spec-insn-heuristic
8282 Enable the speculative instruction heuristic in the scheduler. This
8283 heuristic favors speculative instructions with greater dependency weakness.
8284 This is enabled by default when scheduling is enabled, i.e.@:
8285 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8286 or at @option{-O2} or higher.
8288 @item -fsched-rank-heuristic
8289 @opindex fsched-rank-heuristic
8290 Enable the rank heuristic in the scheduler. This heuristic favors
8291 the instruction belonging to a basic block with greater size or frequency.
8292 This is enabled by default when scheduling is enabled, i.e.@:
8293 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8294 at @option{-O2} or higher.
8296 @item -fsched-last-insn-heuristic
8297 @opindex fsched-last-insn-heuristic
8298 Enable the last-instruction heuristic in the scheduler. This heuristic
8299 favors the instruction that is less dependent on the last instruction
8300 scheduled. This is enabled by default when scheduling is enabled,
8301 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8302 at @option{-O2} or higher.
8304 @item -fsched-dep-count-heuristic
8305 @opindex fsched-dep-count-heuristic
8306 Enable the dependent-count heuristic in the scheduler. This heuristic
8307 favors the instruction that has more instructions depending on it.
8308 This is enabled by default when scheduling is enabled, i.e.@:
8309 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8310 at @option{-O2} or higher.
8312 @item -freschedule-modulo-scheduled-loops
8313 @opindex freschedule-modulo-scheduled-loops
8314 Modulo scheduling is performed before traditional scheduling. If a loop
8315 is modulo scheduled, later scheduling passes may change its schedule.
8316 Use this option to control that behavior.
8318 @item -fselective-scheduling
8319 @opindex fselective-scheduling
8320 Schedule instructions using selective scheduling algorithm. Selective
8321 scheduling runs instead of the first scheduler pass.
8323 @item -fselective-scheduling2
8324 @opindex fselective-scheduling2
8325 Schedule instructions using selective scheduling algorithm. Selective
8326 scheduling runs instead of the second scheduler pass.
8328 @item -fsel-sched-pipelining
8329 @opindex fsel-sched-pipelining
8330 Enable software pipelining of innermost loops during selective scheduling.
8331 This option has no effect unless one of @option{-fselective-scheduling} or
8332 @option{-fselective-scheduling2} is turned on.
8334 @item -fsel-sched-pipelining-outer-loops
8335 @opindex fsel-sched-pipelining-outer-loops
8336 When pipelining loops during selective scheduling, also pipeline outer loops.
8337 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8339 @item -fsemantic-interposition
8340 @opindex fsemantic-interposition
8341 Some object formats, like ELF, allow interposing of symbols by the
8343 This means that for symbols exported from the DSO, the compiler cannot perform
8344 interprocedural propagation, inlining and other optimizations in anticipation
8345 that the function or variable in question may change. While this feature is
8346 useful, for example, to rewrite memory allocation functions by a debugging
8347 implementation, it is expensive in the terms of code quality.
8348 With @option{-fno-semantic-interposition} the compiler assumes that
8349 if interposition happens for functions the overwriting function will have
8350 precisely the same semantics (and side effects).
8351 Similarly if interposition happens
8352 for variables, the constructor of the variable will be the same. The flag
8353 has no effect for functions explicitly declared inline
8354 (where it is never allowed for interposition to change semantics)
8355 and for symbols explicitly declared weak.
8358 @opindex fshrink-wrap
8359 Emit function prologues only before parts of the function that need it,
8360 rather than at the top of the function. This flag is enabled by default at
8361 @option{-O} and higher.
8363 @item -fcaller-saves
8364 @opindex fcaller-saves
8365 Enable allocation of values to registers that are clobbered by
8366 function calls, by emitting extra instructions to save and restore the
8367 registers around such calls. Such allocation is done only when it
8368 seems to result in better code.
8370 This option is always enabled by default on certain machines, usually
8371 those which have no call-preserved registers to use instead.
8373 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8375 @item -fcombine-stack-adjustments
8376 @opindex fcombine-stack-adjustments
8377 Tracks stack adjustments (pushes and pops) and stack memory references
8378 and then tries to find ways to combine them.
8380 Enabled by default at @option{-O1} and higher.
8384 Use caller save registers for allocation if those registers are not used by
8385 any called function. In that case it is not necessary to save and restore
8386 them around calls. This is only possible if called functions are part of
8387 same compilation unit as current function and they are compiled before it.
8389 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8391 @item -fconserve-stack
8392 @opindex fconserve-stack
8393 Attempt to minimize stack usage. The compiler attempts to use less
8394 stack space, even if that makes the program slower. This option
8395 implies setting the @option{large-stack-frame} parameter to 100
8396 and the @option{large-stack-frame-growth} parameter to 400.
8398 @item -ftree-reassoc
8399 @opindex ftree-reassoc
8400 Perform reassociation on trees. This flag is enabled by default
8401 at @option{-O} and higher.
8405 Perform partial redundancy elimination (PRE) on trees. This flag is
8406 enabled by default at @option{-O2} and @option{-O3}.
8408 @item -ftree-partial-pre
8409 @opindex ftree-partial-pre
8410 Make partial redundancy elimination (PRE) more aggressive. This flag is
8411 enabled by default at @option{-O3}.
8413 @item -ftree-forwprop
8414 @opindex ftree-forwprop
8415 Perform forward propagation on trees. This flag is enabled by default
8416 at @option{-O} and higher.
8420 Perform full redundancy elimination (FRE) on trees. The difference
8421 between FRE and PRE is that FRE only considers expressions
8422 that are computed on all paths leading to the redundant computation.
8423 This analysis is faster than PRE, though it exposes fewer redundancies.
8424 This flag is enabled by default at @option{-O} and higher.
8426 @item -ftree-phiprop
8427 @opindex ftree-phiprop
8428 Perform hoisting of loads from conditional pointers on trees. This
8429 pass is enabled by default at @option{-O} and higher.
8431 @item -fhoist-adjacent-loads
8432 @opindex fhoist-adjacent-loads
8433 Speculatively hoist loads from both branches of an if-then-else if the
8434 loads are from adjacent locations in the same structure and the target
8435 architecture has a conditional move instruction. This flag is enabled
8436 by default at @option{-O2} and higher.
8438 @item -ftree-copy-prop
8439 @opindex ftree-copy-prop
8440 Perform copy propagation on trees. This pass eliminates unnecessary
8441 copy operations. This flag is enabled by default at @option{-O} and
8444 @item -fipa-pure-const
8445 @opindex fipa-pure-const
8446 Discover which functions are pure or constant.
8447 Enabled by default at @option{-O} and higher.
8449 @item -fipa-reference
8450 @opindex fipa-reference
8451 Discover which static variables do not escape the
8453 Enabled by default at @option{-O} and higher.
8457 Perform interprocedural pointer analysis and interprocedural modification
8458 and reference analysis. This option can cause excessive memory and
8459 compile-time usage on large compilation units. It is not enabled by
8460 default at any optimization level.
8463 @opindex fipa-profile
8464 Perform interprocedural profile propagation. The functions called only from
8465 cold functions are marked as cold. Also functions executed once (such as
8466 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8467 functions and loop less parts of functions executed once are then optimized for
8469 Enabled by default at @option{-O} and higher.
8473 Perform interprocedural constant propagation.
8474 This optimization analyzes the program to determine when values passed
8475 to functions are constants and then optimizes accordingly.
8476 This optimization can substantially increase performance
8477 if the application has constants passed to functions.
8478 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8480 @item -fipa-cp-clone
8481 @opindex fipa-cp-clone
8482 Perform function cloning to make interprocedural constant propagation stronger.
8483 When enabled, interprocedural constant propagation performs function cloning
8484 when externally visible function can be called with constant arguments.
8485 Because this optimization can create multiple copies of functions,
8486 it may significantly increase code size
8487 (see @option{--param ipcp-unit-growth=@var{value}}).
8488 This flag is enabled by default at @option{-O3}.
8490 @item -fipa-cp-alignment
8491 @opindex -fipa-cp-alignment
8492 When enabled, this optimization propagates alignment of function
8493 parameters to support better vectorization and string operations.
8495 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8496 requires that @option{-fipa-cp} is enabled.
8500 Perform Identical Code Folding for functions and read-only variables.
8501 The optimization reduces code size and may disturb unwind stacks by replacing
8502 a function by equivalent one with a different name. The optimization works
8503 more effectively with link time optimization enabled.
8505 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8506 works on different levels and thus the optimizations are not same - there are
8507 equivalences that are found only by GCC and equivalences found only by Gold.
8509 This flag is enabled by default at @option{-O2} and @option{-Os}.
8511 @item -fisolate-erroneous-paths-dereference
8512 @opindex fisolate-erroneous-paths-dereference
8513 Detect paths that trigger erroneous or undefined behavior due to
8514 dereferencing a null pointer. Isolate those paths from the main control
8515 flow and turn the statement with erroneous or undefined behavior into a trap.
8516 This flag is enabled by default at @option{-O2} and higher and depends on
8517 @option{-fdelete-null-pointer-checks} also being enabled.
8519 @item -fisolate-erroneous-paths-attribute
8520 @opindex fisolate-erroneous-paths-attribute
8521 Detect paths that trigger erroneous or undefined behavior due a null value
8522 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8523 attribute. Isolate those paths from the main control flow and turn the
8524 statement with erroneous or undefined behavior into a trap. This is not
8525 currently enabled, but may be enabled by @option{-O2} in the future.
8529 Perform forward store motion on trees. This flag is
8530 enabled by default at @option{-O} and higher.
8532 @item -ftree-bit-ccp
8533 @opindex ftree-bit-ccp
8534 Perform sparse conditional bit constant propagation on trees and propagate
8535 pointer alignment information.
8536 This pass only operates on local scalar variables and is enabled by default
8537 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8541 Perform sparse conditional constant propagation (CCP) on trees. This
8542 pass only operates on local scalar variables and is enabled by default
8543 at @option{-O} and higher.
8546 @opindex fssa-phiopt
8547 Perform pattern matching on SSA PHI nodes to optimize conditional
8548 code. This pass is enabled by default at @option{-O} and higher.
8550 @item -ftree-switch-conversion
8551 @opindex ftree-switch-conversion
8552 Perform conversion of simple initializations in a switch to
8553 initializations from a scalar array. This flag is enabled by default
8554 at @option{-O2} and higher.
8556 @item -ftree-tail-merge
8557 @opindex ftree-tail-merge
8558 Look for identical code sequences. When found, replace one with a jump to the
8559 other. This optimization is known as tail merging or cross jumping. This flag
8560 is enabled by default at @option{-O2} and higher. The compilation time
8562 be limited using @option{max-tail-merge-comparisons} parameter and
8563 @option{max-tail-merge-iterations} parameter.
8567 Perform dead code elimination (DCE) on trees. This flag is enabled by
8568 default at @option{-O} and higher.
8570 @item -ftree-builtin-call-dce
8571 @opindex ftree-builtin-call-dce
8572 Perform conditional dead code elimination (DCE) for calls to built-in functions
8573 that may set @code{errno} but are otherwise side-effect free. This flag is
8574 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8577 @item -ftree-dominator-opts
8578 @opindex ftree-dominator-opts
8579 Perform a variety of simple scalar cleanups (constant/copy
8580 propagation, redundancy elimination, range propagation and expression
8581 simplification) based on a dominator tree traversal. This also
8582 performs jump threading (to reduce jumps to jumps). This flag is
8583 enabled by default at @option{-O} and higher.
8587 Perform dead store elimination (DSE) on trees. A dead store is a store into
8588 a memory location that is later overwritten by another store without
8589 any intervening loads. In this case the earlier store can be deleted. This
8590 flag is enabled by default at @option{-O} and higher.
8594 Perform loop header copying on trees. This is beneficial since it increases
8595 effectiveness of code motion optimizations. It also saves one jump. This flag
8596 is enabled by default at @option{-O} and higher. It is not enabled
8597 for @option{-Os}, since it usually increases code size.
8599 @item -ftree-loop-optimize
8600 @opindex ftree-loop-optimize
8601 Perform loop optimizations on trees. This flag is enabled by default
8602 at @option{-O} and higher.
8604 @item -ftree-loop-linear
8605 @opindex ftree-loop-linear
8606 Perform loop interchange transformations on tree. Same as
8607 @option{-floop-interchange}. To use this code transformation, GCC has
8608 to be configured with @option{--with-isl} to enable the Graphite loop
8609 transformation infrastructure.
8611 @item -floop-interchange
8612 @opindex floop-interchange
8613 Perform loop interchange transformations on loops. Interchanging two
8614 nested loops switches the inner and outer loops. For example, given a
8619 A(J, I) = A(J, I) * C
8624 loop interchange transforms the loop as if it were written:
8628 A(J, I) = A(J, I) * C
8632 which can be beneficial when @code{N} is larger than the caches,
8633 because in Fortran, the elements of an array are stored in memory
8634 contiguously by column, and the original loop iterates over rows,
8635 potentially creating at each access a cache miss. This optimization
8636 applies to all the languages supported by GCC and is not limited to
8637 Fortran. To use this code transformation, GCC has to be configured
8638 with @option{--with-isl} to enable the Graphite loop transformation
8641 @item -floop-strip-mine
8642 @opindex floop-strip-mine
8643 Perform loop strip mining transformations on loops. Strip mining
8644 splits a loop into two nested loops. The outer loop has strides
8645 equal to the strip size and the inner loop has strides of the
8646 original loop within a strip. The strip length can be changed
8647 using the @option{loop-block-tile-size} parameter. For example,
8655 loop strip mining transforms the loop as if it were written:
8658 DO I = II, min (II + 50, N)
8663 This optimization applies to all the languages supported by GCC and is
8664 not limited to Fortran. To use this code transformation, GCC has to
8665 be configured with @option{--with-isl} to enable the Graphite loop
8666 transformation infrastructure.
8669 @opindex floop-block
8670 Perform loop blocking transformations on loops. Blocking strip mines
8671 each loop in the loop nest such that the memory accesses of the
8672 element loops fit inside caches. The strip length can be changed
8673 using the @option{loop-block-tile-size} parameter. For example, given
8678 A(J, I) = B(I) + C(J)
8683 loop blocking transforms the loop as if it were written:
8687 DO I = II, min (II + 50, N)
8688 DO J = JJ, min (JJ + 50, M)
8689 A(J, I) = B(I) + C(J)
8695 which can be beneficial when @code{M} is larger than the caches,
8696 because the innermost loop iterates over a smaller amount of data
8697 which can be kept in the caches. This optimization applies to all the
8698 languages supported by GCC and is not limited to Fortran. To use this
8699 code transformation, GCC has to be configured with @option{--with-isl}
8700 to enable the Graphite loop transformation infrastructure.
8702 @item -fgraphite-identity
8703 @opindex fgraphite-identity
8704 Enable the identity transformation for graphite. For every SCoP we generate
8705 the polyhedral representation and transform it back to gimple. Using
8706 @option{-fgraphite-identity} we can check the costs or benefits of the
8707 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8708 are also performed by the code generator ISL, like index splitting and
8709 dead code elimination in loops.
8711 @item -floop-nest-optimize
8712 @opindex floop-nest-optimize
8713 Enable the ISL based loop nest optimizer. This is a generic loop nest
8714 optimizer based on the Pluto optimization algorithms. It calculates a loop
8715 structure optimized for data-locality and parallelism. This option
8718 @item -floop-unroll-and-jam
8719 @opindex floop-unroll-and-jam
8720 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8721 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8722 The unrolled dimension (counting from the most inner one) can be changed
8723 using the @option{loop-unroll-jam-depth} parameter. .
8725 @item -floop-parallelize-all
8726 @opindex floop-parallelize-all
8727 Use the Graphite data dependence analysis to identify loops that can
8728 be parallelized. Parallelize all the loops that can be analyzed to
8729 not contain loop carried dependences without checking that it is
8730 profitable to parallelize the loops.
8732 @item -fcheck-data-deps
8733 @opindex fcheck-data-deps
8734 Compare the results of several data dependence analyzers. This option
8735 is used for debugging the data dependence analyzers.
8737 @item -ftree-loop-if-convert
8738 @opindex ftree-loop-if-convert
8739 Attempt to transform conditional jumps in the innermost loops to
8740 branch-less equivalents. The intent is to remove control-flow from
8741 the innermost loops in order to improve the ability of the
8742 vectorization pass to handle these loops. This is enabled by default
8743 if vectorization is enabled.
8745 @item -ftree-loop-if-convert-stores
8746 @opindex ftree-loop-if-convert-stores
8747 Attempt to also if-convert conditional jumps containing memory writes.
8748 This transformation can be unsafe for multi-threaded programs as it
8749 transforms conditional memory writes into unconditional memory writes.
8752 for (i = 0; i < N; i++)
8758 for (i = 0; i < N; i++)
8759 A[i] = cond ? expr : A[i];
8761 potentially producing data races.
8763 @item -ftree-loop-distribution
8764 @opindex ftree-loop-distribution
8765 Perform loop distribution. This flag can improve cache performance on
8766 big loop bodies and allow further loop optimizations, like
8767 parallelization or vectorization, to take place. For example, the loop
8784 @item -ftree-loop-distribute-patterns
8785 @opindex ftree-loop-distribute-patterns
8786 Perform loop distribution of patterns that can be code generated with
8787 calls to a library. This flag is enabled by default at @option{-O3}.
8789 This pass distributes the initialization loops and generates a call to
8790 memset zero. For example, the loop
8806 and the initialization loop is transformed into a call to memset zero.
8808 @item -ftree-loop-im
8809 @opindex ftree-loop-im
8810 Perform loop invariant motion on trees. This pass moves only invariants that
8811 are hard to handle at RTL level (function calls, operations that expand to
8812 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8813 operands of conditions that are invariant out of the loop, so that we can use
8814 just trivial invariantness analysis in loop unswitching. The pass also includes
8817 @item -ftree-loop-ivcanon
8818 @opindex ftree-loop-ivcanon
8819 Create a canonical counter for number of iterations in loops for which
8820 determining number of iterations requires complicated analysis. Later
8821 optimizations then may determine the number easily. Useful especially
8822 in connection with unrolling.
8826 Perform induction variable optimizations (strength reduction, induction
8827 variable merging and induction variable elimination) on trees.
8829 @item -ftree-parallelize-loops=n
8830 @opindex ftree-parallelize-loops
8831 Parallelize loops, i.e., split their iteration space to run in n threads.
8832 This is only possible for loops whose iterations are independent
8833 and can be arbitrarily reordered. The optimization is only
8834 profitable on multiprocessor machines, for loops that are CPU-intensive,
8835 rather than constrained e.g.@: by memory bandwidth. This option
8836 implies @option{-pthread}, and thus is only supported on targets
8837 that have support for @option{-pthread}.
8841 Perform function-local points-to analysis on trees. This flag is
8842 enabled by default at @option{-O} and higher.
8846 Perform scalar replacement of aggregates. This pass replaces structure
8847 references with scalars to prevent committing structures to memory too
8848 early. This flag is enabled by default at @option{-O} and higher.
8850 @item -ftree-copyrename
8851 @opindex ftree-copyrename
8852 Perform copy renaming on trees. This pass attempts to rename compiler
8853 temporaries to other variables at copy locations, usually resulting in
8854 variable names which more closely resemble the original variables. This flag
8855 is enabled by default at @option{-O} and higher.
8857 @item -ftree-coalesce-inlined-vars
8858 @opindex ftree-coalesce-inlined-vars
8859 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8860 combine small user-defined variables too, but only if they are inlined
8861 from other functions. It is a more limited form of
8862 @option{-ftree-coalesce-vars}. This may harm debug information of such
8863 inlined variables, but it keeps variables of the inlined-into
8864 function apart from each other, such that they are more likely to
8865 contain the expected values in a debugging session.
8867 @item -ftree-coalesce-vars
8868 @opindex ftree-coalesce-vars
8869 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8870 combine small user-defined variables too, instead of just compiler
8871 temporaries. This may severely limit the ability to debug an optimized
8872 program compiled with @option{-fno-var-tracking-assignments}. In the
8873 negated form, this flag prevents SSA coalescing of user variables,
8874 including inlined ones. This option is enabled by default.
8878 Perform temporary expression replacement during the SSA->normal phase. Single
8879 use/single def temporaries are replaced at their use location with their
8880 defining expression. This results in non-GIMPLE code, but gives the expanders
8881 much more complex trees to work on resulting in better RTL generation. This is
8882 enabled by default at @option{-O} and higher.
8886 Perform straight-line strength reduction on trees. This recognizes related
8887 expressions involving multiplications and replaces them by less expensive
8888 calculations when possible. This is enabled by default at @option{-O} and
8891 @item -ftree-vectorize
8892 @opindex ftree-vectorize
8893 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
8894 and @option{-ftree-slp-vectorize} if not explicitly specified.
8896 @item -ftree-loop-vectorize
8897 @opindex ftree-loop-vectorize
8898 Perform loop vectorization on trees. This flag is enabled by default at
8899 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8901 @item -ftree-slp-vectorize
8902 @opindex ftree-slp-vectorize
8903 Perform basic block vectorization on trees. This flag is enabled by default at
8904 @option{-O3} and when @option{-ftree-vectorize} is enabled.
8906 @item -fvect-cost-model=@var{model}
8907 @opindex fvect-cost-model
8908 Alter the cost model used for vectorization. The @var{model} argument
8909 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
8910 With the @samp{unlimited} model the vectorized code-path is assumed
8911 to be profitable while with the @samp{dynamic} model a runtime check
8912 guards the vectorized code-path to enable it only for iteration
8913 counts that will likely execute faster than when executing the original
8914 scalar loop. The @samp{cheap} model disables vectorization of
8915 loops where doing so would be cost prohibitive for example due to
8916 required runtime checks for data dependence or alignment but otherwise
8917 is equal to the @samp{dynamic} model.
8918 The default cost model depends on other optimization flags and is
8919 either @samp{dynamic} or @samp{cheap}.
8921 @item -fsimd-cost-model=@var{model}
8922 @opindex fsimd-cost-model
8923 Alter the cost model used for vectorization of loops marked with the OpenMP
8924 or Cilk Plus simd directive. The @var{model} argument should be one of
8925 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
8926 have the same meaning as described in @option{-fvect-cost-model} and by
8927 default a cost model defined with @option{-fvect-cost-model} is used.
8931 Perform Value Range Propagation on trees. This is similar to the
8932 constant propagation pass, but instead of values, ranges of values are
8933 propagated. This allows the optimizers to remove unnecessary range
8934 checks like array bound checks and null pointer checks. This is
8935 enabled by default at @option{-O2} and higher. Null pointer check
8936 elimination is only done if @option{-fdelete-null-pointer-checks} is
8939 @item -fsplit-ivs-in-unroller
8940 @opindex fsplit-ivs-in-unroller
8941 Enables expression of values of induction variables in later iterations
8942 of the unrolled loop using the value in the first iteration. This breaks
8943 long dependency chains, thus improving efficiency of the scheduling passes.
8945 A combination of @option{-fweb} and CSE is often sufficient to obtain the
8946 same effect. However, that is not reliable in cases where the loop body
8947 is more complicated than a single basic block. It also does not work at all
8948 on some architectures due to restrictions in the CSE pass.
8950 This optimization is enabled by default.
8952 @item -fvariable-expansion-in-unroller
8953 @opindex fvariable-expansion-in-unroller
8954 With this option, the compiler creates multiple copies of some
8955 local variables when unrolling a loop, which can result in superior code.
8957 @item -fpartial-inlining
8958 @opindex fpartial-inlining
8959 Inline parts of functions. This option has any effect only
8960 when inlining itself is turned on by the @option{-finline-functions}
8961 or @option{-finline-small-functions} options.
8963 Enabled at level @option{-O2}.
8965 @item -fpredictive-commoning
8966 @opindex fpredictive-commoning
8967 Perform predictive commoning optimization, i.e., reusing computations
8968 (especially memory loads and stores) performed in previous
8969 iterations of loops.
8971 This option is enabled at level @option{-O3}.
8973 @item -fprefetch-loop-arrays
8974 @opindex fprefetch-loop-arrays
8975 If supported by the target machine, generate instructions to prefetch
8976 memory to improve the performance of loops that access large arrays.
8978 This option may generate better or worse code; results are highly
8979 dependent on the structure of loops within the source code.
8981 Disabled at level @option{-Os}.
8984 @itemx -fno-peephole2
8985 @opindex fno-peephole
8986 @opindex fno-peephole2
8987 Disable any machine-specific peephole optimizations. The difference
8988 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
8989 are implemented in the compiler; some targets use one, some use the
8990 other, a few use both.
8992 @option{-fpeephole} is enabled by default.
8993 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8995 @item -fno-guess-branch-probability
8996 @opindex fno-guess-branch-probability
8997 Do not guess branch probabilities using heuristics.
8999 GCC uses heuristics to guess branch probabilities if they are
9000 not provided by profiling feedback (@option{-fprofile-arcs}). These
9001 heuristics are based on the control flow graph. If some branch probabilities
9002 are specified by @code{__builtin_expect}, then the heuristics are
9003 used to guess branch probabilities for the rest of the control flow graph,
9004 taking the @code{__builtin_expect} info into account. The interactions
9005 between the heuristics and @code{__builtin_expect} can be complex, and in
9006 some cases, it may be useful to disable the heuristics so that the effects
9007 of @code{__builtin_expect} are easier to understand.
9009 The default is @option{-fguess-branch-probability} at levels
9010 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9012 @item -freorder-blocks
9013 @opindex freorder-blocks
9014 Reorder basic blocks in the compiled function in order to reduce number of
9015 taken branches and improve code locality.
9017 Enabled at levels @option{-O2}, @option{-O3}.
9019 @item -freorder-blocks-and-partition
9020 @opindex freorder-blocks-and-partition
9021 In addition to reordering basic blocks in the compiled function, in order
9022 to reduce number of taken branches, partitions hot and cold basic blocks
9023 into separate sections of the assembly and .o files, to improve
9024 paging and cache locality performance.
9026 This optimization is automatically turned off in the presence of
9027 exception handling, for linkonce sections, for functions with a user-defined
9028 section attribute and on any architecture that does not support named
9031 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9033 @item -freorder-functions
9034 @opindex freorder-functions
9035 Reorder functions in the object file in order to
9036 improve code locality. This is implemented by using special
9037 subsections @code{.text.hot} for most frequently executed functions and
9038 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9039 the linker so object file format must support named sections and linker must
9040 place them in a reasonable way.
9042 Also profile feedback must be available to make this option effective. See
9043 @option{-fprofile-arcs} for details.
9045 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9047 @item -fstrict-aliasing
9048 @opindex fstrict-aliasing
9049 Allow the compiler to assume the strictest aliasing rules applicable to
9050 the language being compiled. For C (and C++), this activates
9051 optimizations based on the type of expressions. In particular, an
9052 object of one type is assumed never to reside at the same address as an
9053 object of a different type, unless the types are almost the same. For
9054 example, an @code{unsigned int} can alias an @code{int}, but not a
9055 @code{void*} or a @code{double}. A character type may alias any other
9058 @anchor{Type-punning}Pay special attention to code like this:
9071 The practice of reading from a different union member than the one most
9072 recently written to (called ``type-punning'') is common. Even with
9073 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9074 is accessed through the union type. So, the code above works as
9075 expected. @xref{Structures unions enumerations and bit-fields
9076 implementation}. However, this code might not:
9087 Similarly, access by taking the address, casting the resulting pointer
9088 and dereferencing the result has undefined behavior, even if the cast
9089 uses a union type, e.g.:
9093 return ((union a_union *) &d)->i;
9097 The @option{-fstrict-aliasing} option is enabled at levels
9098 @option{-O2}, @option{-O3}, @option{-Os}.
9100 @item -fstrict-overflow
9101 @opindex fstrict-overflow
9102 Allow the compiler to assume strict signed overflow rules, depending
9103 on the language being compiled. For C (and C++) this means that
9104 overflow when doing arithmetic with signed numbers is undefined, which
9105 means that the compiler may assume that it does not happen. This
9106 permits various optimizations. For example, the compiler assumes
9107 that an expression like @code{i + 10 > i} is always true for
9108 signed @code{i}. This assumption is only valid if signed overflow is
9109 undefined, as the expression is false if @code{i + 10} overflows when
9110 using twos complement arithmetic. When this option is in effect any
9111 attempt to determine whether an operation on signed numbers
9112 overflows must be written carefully to not actually involve overflow.
9114 This option also allows the compiler to assume strict pointer
9115 semantics: given a pointer to an object, if adding an offset to that
9116 pointer does not produce a pointer to the same object, the addition is
9117 undefined. This permits the compiler to conclude that @code{p + u >
9118 p} is always true for a pointer @code{p} and unsigned integer
9119 @code{u}. This assumption is only valid because pointer wraparound is
9120 undefined, as the expression is false if @code{p + u} overflows using
9121 twos complement arithmetic.
9123 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9124 that integer signed overflow is fully defined: it wraps. When
9125 @option{-fwrapv} is used, there is no difference between
9126 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9127 integers. With @option{-fwrapv} certain types of overflow are
9128 permitted. For example, if the compiler gets an overflow when doing
9129 arithmetic on constants, the overflowed value can still be used with
9130 @option{-fwrapv}, but not otherwise.
9132 The @option{-fstrict-overflow} option is enabled at levels
9133 @option{-O2}, @option{-O3}, @option{-Os}.
9135 @item -falign-functions
9136 @itemx -falign-functions=@var{n}
9137 @opindex falign-functions
9138 Align the start of functions to the next power-of-two greater than
9139 @var{n}, skipping up to @var{n} bytes. For instance,
9140 @option{-falign-functions=32} aligns functions to the next 32-byte
9141 boundary, but @option{-falign-functions=24} aligns to the next
9142 32-byte boundary only if this can be done by skipping 23 bytes or less.
9144 @option{-fno-align-functions} and @option{-falign-functions=1} are
9145 equivalent and mean that functions are not aligned.
9147 Some assemblers only support this flag when @var{n} is a power of two;
9148 in that case, it is rounded up.
9150 If @var{n} is not specified or is zero, use a machine-dependent default.
9152 Enabled at levels @option{-O2}, @option{-O3}.
9154 @item -falign-labels
9155 @itemx -falign-labels=@var{n}
9156 @opindex falign-labels
9157 Align all branch targets to a power-of-two boundary, skipping up to
9158 @var{n} bytes like @option{-falign-functions}. This option can easily
9159 make code slower, because it must insert dummy operations for when the
9160 branch target is reached in the usual flow of the code.
9162 @option{-fno-align-labels} and @option{-falign-labels=1} are
9163 equivalent and mean that labels are not aligned.
9165 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9166 are greater than this value, then their values are used instead.
9168 If @var{n} is not specified or is zero, use a machine-dependent default
9169 which is very likely to be @samp{1}, meaning no alignment.
9171 Enabled at levels @option{-O2}, @option{-O3}.
9174 @itemx -falign-loops=@var{n}
9175 @opindex falign-loops
9176 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9177 like @option{-falign-functions}. If the loops are
9178 executed many times, this makes up for any execution of the dummy
9181 @option{-fno-align-loops} and @option{-falign-loops=1} are
9182 equivalent and mean that loops are not aligned.
9184 If @var{n} is not specified or is zero, use a machine-dependent default.
9186 Enabled at levels @option{-O2}, @option{-O3}.
9189 @itemx -falign-jumps=@var{n}
9190 @opindex falign-jumps
9191 Align branch targets to a power-of-two boundary, for branch targets
9192 where the targets can only be reached by jumping, skipping up to @var{n}
9193 bytes like @option{-falign-functions}. In this case, no dummy operations
9196 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9197 equivalent and mean that loops are not aligned.
9199 If @var{n} is not specified or is zero, use a machine-dependent default.
9201 Enabled at levels @option{-O2}, @option{-O3}.
9203 @item -funit-at-a-time
9204 @opindex funit-at-a-time
9205 This option is left for compatibility reasons. @option{-funit-at-a-time}
9206 has no effect, while @option{-fno-unit-at-a-time} implies
9207 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9211 @item -fno-toplevel-reorder
9212 @opindex fno-toplevel-reorder
9213 Do not reorder top-level functions, variables, and @code{asm}
9214 statements. Output them in the same order that they appear in the
9215 input file. When this option is used, unreferenced static variables
9216 are not removed. This option is intended to support existing code
9217 that relies on a particular ordering. For new code, it is better to
9218 use attributes when possible.
9220 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9221 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9226 Constructs webs as commonly used for register allocation purposes and assign
9227 each web individual pseudo register. This allows the register allocation pass
9228 to operate on pseudos directly, but also strengthens several other optimization
9229 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9230 however, make debugging impossible, since variables no longer stay in a
9233 Enabled by default with @option{-funroll-loops}.
9235 @item -fwhole-program
9236 @opindex fwhole-program
9237 Assume that the current compilation unit represents the whole program being
9238 compiled. All public functions and variables with the exception of @code{main}
9239 and those merged by attribute @code{externally_visible} become static functions
9240 and in effect are optimized more aggressively by interprocedural optimizers.
9242 This option should not be used in combination with @option{-flto}.
9243 Instead relying on a linker plugin should provide safer and more precise
9246 @item -flto[=@var{n}]
9248 This option runs the standard link-time optimizer. When invoked
9249 with source code, it generates GIMPLE (one of GCC's internal
9250 representations) and writes it to special ELF sections in the object
9251 file. When the object files are linked together, all the function
9252 bodies are read from these ELF sections and instantiated as if they
9253 had been part of the same translation unit.
9255 To use the link-time optimizer, @option{-flto} and optimization
9256 options should be specified at compile time and during the final link.
9260 gcc -c -O2 -flto foo.c
9261 gcc -c -O2 -flto bar.c
9262 gcc -o myprog -flto -O2 foo.o bar.o
9265 The first two invocations to GCC save a bytecode representation
9266 of GIMPLE into special ELF sections inside @file{foo.o} and
9267 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9268 @file{foo.o} and @file{bar.o}, merges the two files into a single
9269 internal image, and compiles the result as usual. Since both
9270 @file{foo.o} and @file{bar.o} are merged into a single image, this
9271 causes all the interprocedural analyses and optimizations in GCC to
9272 work across the two files as if they were a single one. This means,
9273 for example, that the inliner is able to inline functions in
9274 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9276 Another (simpler) way to enable link-time optimization is:
9279 gcc -o myprog -flto -O2 foo.c bar.c
9282 The above generates bytecode for @file{foo.c} and @file{bar.c},
9283 merges them together into a single GIMPLE representation and optimizes
9284 them as usual to produce @file{myprog}.
9286 The only important thing to keep in mind is that to enable link-time
9287 optimizations you need to use the GCC driver to perform the link-step.
9288 GCC then automatically performs link-time optimization if any of the
9289 objects involved were compiled with the @option{-flto} command-line option.
9291 should specify the optimization options to be used for link-time
9292 optimization though GCC tries to be clever at guessing an
9293 optimization level to use from the options used at compile-time
9294 if you fail to specify one at link-time. You can always override
9295 the automatic decision to do link-time optimization at link-time
9296 by passing @option{-fno-lto} to the link command.
9298 To make whole program optimization effective, it is necessary to make
9299 certain whole program assumptions. The compiler needs to know
9300 what functions and variables can be accessed by libraries and runtime
9301 outside of the link-time optimized unit. When supported by the linker,
9302 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9303 to the compiler about used and externally visible symbols. When
9304 the linker plugin is not available, @option{-fwhole-program} should be
9305 used to allow the compiler to make these assumptions, which leads
9306 to more aggressive optimization decisions.
9308 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9309 compiled with @option{-flto}, the generated object file is larger than
9310 a regular object file because it contains GIMPLE bytecodes and the usual
9311 final code (see @option{-ffat-lto-objects}. This means that
9312 object files with LTO information can be linked as normal object
9313 files; if @option{-fno-lto} is passed to the linker, no
9314 interprocedural optimizations are applied. Note that when
9315 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9316 but you cannot perform a regular, non-LTO link on them.
9318 Additionally, the optimization flags used to compile individual files
9319 are not necessarily related to those used at link time. For instance,
9322 gcc -c -O0 -ffat-lto-objects -flto foo.c
9323 gcc -c -O0 -ffat-lto-objects -flto bar.c
9324 gcc -o myprog -O3 foo.o bar.o
9327 This produces individual object files with unoptimized assembler
9328 code, but the resulting binary @file{myprog} is optimized at
9329 @option{-O3}. If, instead, the final binary is generated with
9330 @option{-fno-lto}, then @file{myprog} is not optimized.
9332 When producing the final binary, GCC only
9333 applies link-time optimizations to those files that contain bytecode.
9334 Therefore, you can mix and match object files and libraries with
9335 GIMPLE bytecodes and final object code. GCC automatically selects
9336 which files to optimize in LTO mode and which files to link without
9339 There are some code generation flags preserved by GCC when
9340 generating bytecodes, as they need to be used during the final link
9341 stage. Generally options specified at link-time override those
9342 specified at compile-time.
9344 If you do not specify an optimization level option @option{-O} at
9345 link-time then GCC computes one based on the optimization levels
9346 used when compiling the object files. The highest optimization
9349 Currently, the following options and their setting are take from
9350 the first object file that explicitely specified it:
9351 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9352 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9353 and all the @option{-m} target flags.
9355 Certain ABI changing flags are required to match in all compilation-units
9356 and trying to override this at link-time with a conflicting value
9357 is ignored. This includes options such as @option{-freg-struct-return}
9358 and @option{-fpcc-struct-return}.
9360 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9361 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9362 are passed through to the link stage and merged conservatively for
9363 conflicting translation units. Specifically
9364 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9365 precedence and for example @option{-ffp-contract=off} takes precedence
9366 over @option{-ffp-contract=fast}. You can override them at linke-time.
9368 It is recommended that you compile all the files participating in the
9369 same link with the same options and also specify those options at
9372 If LTO encounters objects with C linkage declared with incompatible
9373 types in separate translation units to be linked together (undefined
9374 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9375 issued. The behavior is still undefined at run time. Similar
9376 diagnostics may be raised for other languages.
9378 Another feature of LTO is that it is possible to apply interprocedural
9379 optimizations on files written in different languages:
9384 gfortran -c -flto baz.f90
9385 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9388 Notice that the final link is done with @command{g++} to get the C++
9389 runtime libraries and @option{-lgfortran} is added to get the Fortran
9390 runtime libraries. In general, when mixing languages in LTO mode, you
9391 should use the same link command options as when mixing languages in a
9392 regular (non-LTO) compilation.
9394 If object files containing GIMPLE bytecode are stored in a library archive, say
9395 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9396 are using a linker with plugin support. To create static libraries suitable
9397 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9398 and @command{ranlib};
9399 to show the symbols of object files with GIMPLE bytecode, use
9400 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9401 and @command{nm} have been compiled with plugin support. At link time, use the the
9402 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9403 the LTO optimization process:
9406 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9409 With the linker plugin enabled, the linker extracts the needed
9410 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9411 to make them part of the aggregated GIMPLE image to be optimized.
9413 If you are not using a linker with plugin support and/or do not
9414 enable the linker plugin, then the objects inside @file{libfoo.a}
9415 are extracted and linked as usual, but they do not participate
9416 in the LTO optimization process. In order to make a static library suitable
9417 for both LTO optimization and usual linkage, compile its object files with
9418 @option{-flto} @option{-ffat-lto-objects}.
9420 Link-time optimizations do not require the presence of the whole program to
9421 operate. If the program does not require any symbols to be exported, it is
9422 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9423 the interprocedural optimizers to use more aggressive assumptions which may
9424 lead to improved optimization opportunities.
9425 Use of @option{-fwhole-program} is not needed when linker plugin is
9426 active (see @option{-fuse-linker-plugin}).
9428 The current implementation of LTO makes no
9429 attempt to generate bytecode that is portable between different
9430 types of hosts. The bytecode files are versioned and there is a
9431 strict version check, so bytecode files generated in one version of
9432 GCC do not work with an older or newer version of GCC.
9434 Link-time optimization does not work well with generation of debugging
9435 information. Combining @option{-flto} with
9436 @option{-g} is currently experimental and expected to produce unexpected
9439 If you specify the optional @var{n}, the optimization and code
9440 generation done at link time is executed in parallel using @var{n}
9441 parallel jobs by utilizing an installed @command{make} program. The
9442 environment variable @env{MAKE} may be used to override the program
9443 used. The default value for @var{n} is 1.
9445 You can also specify @option{-flto=jobserver} to use GNU make's
9446 job server mode to determine the number of parallel jobs. This
9447 is useful when the Makefile calling GCC is already executing in parallel.
9448 You must prepend a @samp{+} to the command recipe in the parent Makefile
9449 for this to work. This option likely only works if @env{MAKE} is
9452 @item -flto-partition=@var{alg}
9453 @opindex flto-partition
9454 Specify the partitioning algorithm used by the link-time optimizer.
9455 The value is either @samp{1to1} to specify a partitioning mirroring
9456 the original source files or @samp{balanced} to specify partitioning
9457 into equally sized chunks (whenever possible) or @samp{max} to create
9458 new partition for every symbol where possible. Specifying @samp{none}
9459 as an algorithm disables partitioning and streaming completely.
9460 The default value is @samp{balanced}. While @samp{1to1} can be used
9461 as an workaround for various code ordering issues, the @samp{max}
9462 partitioning is intended for internal testing only.
9463 The value @samp{one} specifies that exactly one partition should be
9464 used while the value @samp{none} bypasses partitioning and executes
9465 the link-time optimization step directly from the WPA phase.
9467 @item -flto-odr-type-merging
9468 @opindex flto-odr-type-merging
9469 Enable streaming of mangled types names of C++ types and their unification
9470 at linktime. This increases size of LTO object files, but enable
9471 diagnostics about One Definition Rule violations.
9473 @item -flto-compression-level=@var{n}
9474 @opindex flto-compression-level
9475 This option specifies the level of compression used for intermediate
9476 language written to LTO object files, and is only meaningful in
9477 conjunction with LTO mode (@option{-flto}). Valid
9478 values are 0 (no compression) to 9 (maximum compression). Values
9479 outside this range are clamped to either 0 or 9. If the option is not
9480 given, a default balanced compression setting is used.
9483 @opindex flto-report
9484 Prints a report with internal details on the workings of the link-time
9485 optimizer. The contents of this report vary from version to version.
9486 It is meant to be useful to GCC developers when processing object
9487 files in LTO mode (via @option{-flto}).
9489 Disabled by default.
9491 @item -flto-report-wpa
9492 @opindex flto-report-wpa
9493 Like @option{-flto-report}, but only print for the WPA phase of Link
9496 @item -fuse-linker-plugin
9497 @opindex fuse-linker-plugin
9498 Enables the use of a linker plugin during link-time optimization. This
9499 option relies on plugin support in the linker, which is available in gold
9500 or in GNU ld 2.21 or newer.
9502 This option enables the extraction of object files with GIMPLE bytecode out
9503 of library archives. This improves the quality of optimization by exposing
9504 more code to the link-time optimizer. This information specifies what
9505 symbols can be accessed externally (by non-LTO object or during dynamic
9506 linking). Resulting code quality improvements on binaries (and shared
9507 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9508 See @option{-flto} for a description of the effect of this flag and how to
9511 This option is enabled by default when LTO support in GCC is enabled
9512 and GCC was configured for use with
9513 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9515 @item -ffat-lto-objects
9516 @opindex ffat-lto-objects
9517 Fat LTO objects are object files that contain both the intermediate language
9518 and the object code. This makes them usable for both LTO linking and normal
9519 linking. This option is effective only when compiling with @option{-flto}
9520 and is ignored at link time.
9522 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9523 requires the complete toolchain to be aware of LTO. It requires a linker with
9524 linker plugin support for basic functionality. Additionally,
9525 @command{nm}, @command{ar} and @command{ranlib}
9526 need to support linker plugins to allow a full-featured build environment
9527 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9528 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9529 to these tools. With non fat LTO makefiles need to be modified to use them.
9531 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9534 @item -fcompare-elim
9535 @opindex fcompare-elim
9536 After register allocation and post-register allocation instruction splitting,
9537 identify arithmetic instructions that compute processor flags similar to a
9538 comparison operation based on that arithmetic. If possible, eliminate the
9539 explicit comparison operation.
9541 This pass only applies to certain targets that cannot explicitly represent
9542 the comparison operation before register allocation is complete.
9544 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9546 @item -fcprop-registers
9547 @opindex fcprop-registers
9548 After register allocation and post-register allocation instruction splitting,
9549 perform a copy-propagation pass to try to reduce scheduling dependencies
9550 and occasionally eliminate the copy.
9552 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9554 @item -fprofile-correction
9555 @opindex fprofile-correction
9556 Profiles collected using an instrumented binary for multi-threaded programs may
9557 be inconsistent due to missed counter updates. When this option is specified,
9558 GCC uses heuristics to correct or smooth out such inconsistencies. By
9559 default, GCC emits an error message when an inconsistent profile is detected.
9561 @item -fprofile-dir=@var{path}
9562 @opindex fprofile-dir
9564 Set the directory to search for the profile data files in to @var{path}.
9565 This option affects only the profile data generated by
9566 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9567 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9568 and its related options. Both absolute and relative paths can be used.
9569 By default, GCC uses the current directory as @var{path}, thus the
9570 profile data file appears in the same directory as the object file.
9572 @item -fprofile-generate
9573 @itemx -fprofile-generate=@var{path}
9574 @opindex fprofile-generate
9576 Enable options usually used for instrumenting application to produce
9577 profile useful for later recompilation with profile feedback based
9578 optimization. You must use @option{-fprofile-generate} both when
9579 compiling and when linking your program.
9581 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9583 If @var{path} is specified, GCC looks at the @var{path} to find
9584 the profile feedback data files. See @option{-fprofile-dir}.
9587 @itemx -fprofile-use=@var{path}
9588 @opindex fprofile-use
9589 Enable profile feedback-directed optimizations,
9590 and the following optimizations
9591 which are generally profitable only with profile feedback available:
9592 @option{-fbranch-probabilities}, @option{-fvpt},
9593 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9594 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9596 By default, GCC emits an error message if the feedback profiles do not
9597 match the source code. This error can be turned into a warning by using
9598 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9601 If @var{path} is specified, GCC looks at the @var{path} to find
9602 the profile feedback data files. See @option{-fprofile-dir}.
9604 @item -fauto-profile
9605 @itemx -fauto-profile=@var{path}
9606 @opindex fauto-profile
9607 Enable sampling-based feedback-directed optimizations,
9608 and the following optimizations
9609 which are generally profitable only with profile feedback available:
9610 @option{-fbranch-probabilities}, @option{-fvpt},
9611 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9612 @option{-ftree-vectorize},
9613 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9614 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9615 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9617 @var{path} is the name of a file containing AutoFDO profile information.
9618 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9620 Producing an AutoFDO profile data file requires running your program
9621 with the @command{perf} utility on a supported GNU/Linux target system.
9622 For more information, see @uref{https://perf.wiki.kernel.org/}.
9626 perf record -e br_inst_retired:near_taken -b -o perf.data \
9630 Then use the @command{create_gcov} tool to convert the raw profile data
9631 to a format that can be used by GCC.@ You must also supply the
9632 unstripped binary for your program to this tool.
9633 See @uref{https://github.com/google/autofdo}.
9637 create_gcov --binary=your_program.unstripped --profile=perf.data \
9642 The following options control compiler behavior regarding floating-point
9643 arithmetic. These options trade off between speed and
9644 correctness. All must be specifically enabled.
9648 @opindex ffloat-store
9649 Do not store floating-point variables in registers, and inhibit other
9650 options that might change whether a floating-point value is taken from a
9653 @cindex floating-point precision
9654 This option prevents undesirable excess precision on machines such as
9655 the 68000 where the floating registers (of the 68881) keep more
9656 precision than a @code{double} is supposed to have. Similarly for the
9657 x86 architecture. For most programs, the excess precision does only
9658 good, but a few programs rely on the precise definition of IEEE floating
9659 point. Use @option{-ffloat-store} for such programs, after modifying
9660 them to store all pertinent intermediate computations into variables.
9662 @item -fexcess-precision=@var{style}
9663 @opindex fexcess-precision
9664 This option allows further control over excess precision on machines
9665 where floating-point registers have more precision than the IEEE
9666 @code{float} and @code{double} types and the processor does not
9667 support operations rounding to those types. By default,
9668 @option{-fexcess-precision=fast} is in effect; this means that
9669 operations are carried out in the precision of the registers and that
9670 it is unpredictable when rounding to the types specified in the source
9671 code takes place. When compiling C, if
9672 @option{-fexcess-precision=standard} is specified then excess
9673 precision follows the rules specified in ISO C99; in particular,
9674 both casts and assignments cause values to be rounded to their
9675 semantic types (whereas @option{-ffloat-store} only affects
9676 assignments). This option is enabled by default for C if a strict
9677 conformance option such as @option{-std=c99} is used.
9680 @option{-fexcess-precision=standard} is not implemented for languages
9681 other than C, and has no effect if
9682 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9683 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9684 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9685 semantics apply without excess precision, and in the latter, rounding
9690 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9691 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9692 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9694 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9696 This option is not turned on by any @option{-O} option besides
9697 @option{-Ofast} since it can result in incorrect output for programs
9698 that depend on an exact implementation of IEEE or ISO rules/specifications
9699 for math functions. It may, however, yield faster code for programs
9700 that do not require the guarantees of these specifications.
9702 @item -fno-math-errno
9703 @opindex fno-math-errno
9704 Do not set @code{errno} after calling math functions that are executed
9705 with a single instruction, e.g., @code{sqrt}. A program that relies on
9706 IEEE exceptions for math error handling may want to use this flag
9707 for speed while maintaining IEEE arithmetic compatibility.
9709 This option is not turned on by any @option{-O} option since
9710 it can result in incorrect output for programs that depend on
9711 an exact implementation of IEEE or ISO rules/specifications for
9712 math functions. It may, however, yield faster code for programs
9713 that do not require the guarantees of these specifications.
9715 The default is @option{-fmath-errno}.
9717 On Darwin systems, the math library never sets @code{errno}. There is
9718 therefore no reason for the compiler to consider the possibility that
9719 it might, and @option{-fno-math-errno} is the default.
9721 @item -funsafe-math-optimizations
9722 @opindex funsafe-math-optimizations
9724 Allow optimizations for floating-point arithmetic that (a) assume
9725 that arguments and results are valid and (b) may violate IEEE or
9726 ANSI standards. When used at link-time, it may include libraries
9727 or startup files that change the default FPU control word or other
9728 similar optimizations.
9730 This option is not turned on by any @option{-O} option since
9731 it can result in incorrect output for programs that depend on
9732 an exact implementation of IEEE or ISO rules/specifications for
9733 math functions. It may, however, yield faster code for programs
9734 that do not require the guarantees of these specifications.
9735 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9736 @option{-fassociative-math} and @option{-freciprocal-math}.
9738 The default is @option{-fno-unsafe-math-optimizations}.
9740 @item -fassociative-math
9741 @opindex fassociative-math
9743 Allow re-association of operands in series of floating-point operations.
9744 This violates the ISO C and C++ language standard by possibly changing
9745 computation result. NOTE: re-ordering may change the sign of zero as
9746 well as ignore NaNs and inhibit or create underflow or overflow (and
9747 thus cannot be used on code that relies on rounding behavior like
9748 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9749 and thus may not be used when ordered comparisons are required.
9750 This option requires that both @option{-fno-signed-zeros} and
9751 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9752 much sense with @option{-frounding-math}. For Fortran the option
9753 is automatically enabled when both @option{-fno-signed-zeros} and
9754 @option{-fno-trapping-math} are in effect.
9756 The default is @option{-fno-associative-math}.
9758 @item -freciprocal-math
9759 @opindex freciprocal-math
9761 Allow the reciprocal of a value to be used instead of dividing by
9762 the value if this enables optimizations. For example @code{x / y}
9763 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9764 is subject to common subexpression elimination. Note that this loses
9765 precision and increases the number of flops operating on the value.
9767 The default is @option{-fno-reciprocal-math}.
9769 @item -ffinite-math-only
9770 @opindex ffinite-math-only
9771 Allow optimizations for floating-point arithmetic that assume
9772 that arguments and results are not NaNs or +-Infs.
9774 This option is not turned on by any @option{-O} option since
9775 it can result in incorrect output for programs that depend on
9776 an exact implementation of IEEE or ISO rules/specifications for
9777 math functions. It may, however, yield faster code for programs
9778 that do not require the guarantees of these specifications.
9780 The default is @option{-fno-finite-math-only}.
9782 @item -fno-signed-zeros
9783 @opindex fno-signed-zeros
9784 Allow optimizations for floating-point arithmetic that ignore the
9785 signedness of zero. IEEE arithmetic specifies the behavior of
9786 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9787 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9788 This option implies that the sign of a zero result isn't significant.
9790 The default is @option{-fsigned-zeros}.
9792 @item -fno-trapping-math
9793 @opindex fno-trapping-math
9794 Compile code assuming that floating-point operations cannot generate
9795 user-visible traps. These traps include division by zero, overflow,
9796 underflow, inexact result and invalid operation. This option requires
9797 that @option{-fno-signaling-nans} be in effect. Setting this option may
9798 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9800 This option should never be turned on by any @option{-O} option since
9801 it can result in incorrect output for programs that depend on
9802 an exact implementation of IEEE or ISO rules/specifications for
9805 The default is @option{-ftrapping-math}.
9807 @item -frounding-math
9808 @opindex frounding-math
9809 Disable transformations and optimizations that assume default floating-point
9810 rounding behavior. This is round-to-zero for all floating point
9811 to integer conversions, and round-to-nearest for all other arithmetic
9812 truncations. This option should be specified for programs that change
9813 the FP rounding mode dynamically, or that may be executed with a
9814 non-default rounding mode. This option disables constant folding of
9815 floating-point expressions at compile time (which may be affected by
9816 rounding mode) and arithmetic transformations that are unsafe in the
9817 presence of sign-dependent rounding modes.
9819 The default is @option{-fno-rounding-math}.
9821 This option is experimental and does not currently guarantee to
9822 disable all GCC optimizations that are affected by rounding mode.
9823 Future versions of GCC may provide finer control of this setting
9824 using C99's @code{FENV_ACCESS} pragma. This command-line option
9825 will be used to specify the default state for @code{FENV_ACCESS}.
9827 @item -fsignaling-nans
9828 @opindex fsignaling-nans
9829 Compile code assuming that IEEE signaling NaNs may generate user-visible
9830 traps during floating-point operations. Setting this option disables
9831 optimizations that may change the number of exceptions visible with
9832 signaling NaNs. This option implies @option{-ftrapping-math}.
9834 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9837 The default is @option{-fno-signaling-nans}.
9839 This option is experimental and does not currently guarantee to
9840 disable all GCC optimizations that affect signaling NaN behavior.
9842 @item -fsingle-precision-constant
9843 @opindex fsingle-precision-constant
9844 Treat floating-point constants as single precision instead of
9845 implicitly converting them to double-precision constants.
9847 @item -fcx-limited-range
9848 @opindex fcx-limited-range
9849 When enabled, this option states that a range reduction step is not
9850 needed when performing complex division. Also, there is no checking
9851 whether the result of a complex multiplication or division is @code{NaN
9852 + I*NaN}, with an attempt to rescue the situation in that case. The
9853 default is @option{-fno-cx-limited-range}, but is enabled by
9854 @option{-ffast-math}.
9856 This option controls the default setting of the ISO C99
9857 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9860 @item -fcx-fortran-rules
9861 @opindex fcx-fortran-rules
9862 Complex multiplication and division follow Fortran rules. Range
9863 reduction is done as part of complex division, but there is no checking
9864 whether the result of a complex multiplication or division is @code{NaN
9865 + I*NaN}, with an attempt to rescue the situation in that case.
9867 The default is @option{-fno-cx-fortran-rules}.
9871 The following options control optimizations that may improve
9872 performance, but are not enabled by any @option{-O} options. This
9873 section includes experimental options that may produce broken code.
9876 @item -fbranch-probabilities
9877 @opindex fbranch-probabilities
9878 After running a program compiled with @option{-fprofile-arcs}
9879 (@pxref{Debugging Options,, Options for Debugging Your Program or
9880 @command{gcc}}), you can compile it a second time using
9881 @option{-fbranch-probabilities}, to improve optimizations based on
9882 the number of times each branch was taken. When a program
9883 compiled with @option{-fprofile-arcs} exits, it saves arc execution
9884 counts to a file called @file{@var{sourcename}.gcda} for each source
9885 file. The information in this data file is very dependent on the
9886 structure of the generated code, so you must use the same source code
9887 and the same optimization options for both compilations.
9889 With @option{-fbranch-probabilities}, GCC puts a
9890 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
9891 These can be used to improve optimization. Currently, they are only
9892 used in one place: in @file{reorg.c}, instead of guessing which path a
9893 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
9894 exactly determine which path is taken more often.
9896 @item -fprofile-values
9897 @opindex fprofile-values
9898 If combined with @option{-fprofile-arcs}, it adds code so that some
9899 data about values of expressions in the program is gathered.
9901 With @option{-fbranch-probabilities}, it reads back the data gathered
9902 from profiling values of expressions for usage in optimizations.
9904 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
9906 @item -fprofile-reorder-functions
9907 @opindex fprofile-reorder-functions
9908 Function reordering based on profile instrumentation collects
9909 first time of execution of a function and orders these functions
9912 Enabled with @option{-fprofile-use}.
9916 If combined with @option{-fprofile-arcs}, this option instructs the compiler
9917 to add code to gather information about values of expressions.
9919 With @option{-fbranch-probabilities}, it reads back the data gathered
9920 and actually performs the optimizations based on them.
9921 Currently the optimizations include specialization of division operations
9922 using the knowledge about the value of the denominator.
9924 @item -frename-registers
9925 @opindex frename-registers
9926 Attempt to avoid false dependencies in scheduled code by making use
9927 of registers left over after register allocation. This optimization
9928 most benefits processors with lots of registers. Depending on the
9929 debug information format adopted by the target, however, it can
9930 make debugging impossible, since variables no longer stay in
9931 a ``home register''.
9933 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
9935 @item -fschedule-fusion
9936 @opindex fschedule-fusion
9937 Performs a target dependent pass over the instruction stream to schedule
9938 instructions of same type together because target machine can execute them
9939 more efficiently if they are adjacent to each other in the instruction flow.
9941 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9945 Perform tail duplication to enlarge superblock size. This transformation
9946 simplifies the control flow of the function allowing other optimizations to do
9949 Enabled with @option{-fprofile-use}.
9951 @item -funroll-loops
9952 @opindex funroll-loops
9953 Unroll loops whose number of iterations can be determined at compile time or
9954 upon entry to the loop. @option{-funroll-loops} implies
9955 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
9956 It also turns on complete loop peeling (i.e.@: complete removal of loops with
9957 a small constant number of iterations). This option makes code larger, and may
9958 or may not make it run faster.
9960 Enabled with @option{-fprofile-use}.
9962 @item -funroll-all-loops
9963 @opindex funroll-all-loops
9964 Unroll all loops, even if their number of iterations is uncertain when
9965 the loop is entered. This usually makes programs run more slowly.
9966 @option{-funroll-all-loops} implies the same options as
9967 @option{-funroll-loops}.
9970 @opindex fpeel-loops
9971 Peels loops for which there is enough information that they do not
9972 roll much (from profile feedback). It also turns on complete loop peeling
9973 (i.e.@: complete removal of loops with small constant number of iterations).
9975 Enabled with @option{-fprofile-use}.
9977 @item -fmove-loop-invariants
9978 @opindex fmove-loop-invariants
9979 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
9980 at level @option{-O1}
9982 @item -funswitch-loops
9983 @opindex funswitch-loops
9984 Move branches with loop invariant conditions out of the loop, with duplicates
9985 of the loop on both branches (modified according to result of the condition).
9987 @item -ffunction-sections
9988 @itemx -fdata-sections
9989 @opindex ffunction-sections
9990 @opindex fdata-sections
9991 Place each function or data item into its own section in the output
9992 file if the target supports arbitrary sections. The name of the
9993 function or the name of the data item determines the section's name
9996 Use these options on systems where the linker can perform optimizations
9997 to improve locality of reference in the instruction space. Most systems
9998 using the ELF object format and SPARC processors running Solaris 2 have
9999 linkers with such optimizations. AIX may have these optimizations in
10002 Only use these options when there are significant benefits from doing
10003 so. When you specify these options, the assembler and linker
10004 create larger object and executable files and are also slower.
10005 You cannot use @command{gprof} on all systems if you
10006 specify this option, and you may have problems with debugging if
10007 you specify both this option and @option{-g}.
10009 @item -fbranch-target-load-optimize
10010 @opindex fbranch-target-load-optimize
10011 Perform branch target register load optimization before prologue / epilogue
10013 The use of target registers can typically be exposed only during reload,
10014 thus hoisting loads out of loops and doing inter-block scheduling needs
10015 a separate optimization pass.
10017 @item -fbranch-target-load-optimize2
10018 @opindex fbranch-target-load-optimize2
10019 Perform branch target register load optimization after prologue / epilogue
10022 @item -fbtr-bb-exclusive
10023 @opindex fbtr-bb-exclusive
10024 When performing branch target register load optimization, don't reuse
10025 branch target registers within any basic block.
10027 @item -fstack-protector
10028 @opindex fstack-protector
10029 Emit extra code to check for buffer overflows, such as stack smashing
10030 attacks. This is done by adding a guard variable to functions with
10031 vulnerable objects. This includes functions that call @code{alloca}, and
10032 functions with buffers larger than 8 bytes. The guards are initialized
10033 when a function is entered and then checked when the function exits.
10034 If a guard check fails, an error message is printed and the program exits.
10036 @item -fstack-protector-all
10037 @opindex fstack-protector-all
10038 Like @option{-fstack-protector} except that all functions are protected.
10040 @item -fstack-protector-strong
10041 @opindex fstack-protector-strong
10042 Like @option{-fstack-protector} but includes additional functions to
10043 be protected --- those that have local array definitions, or have
10044 references to local frame addresses.
10046 @item -fstack-protector-explicit
10047 @opindex fstack-protector-explicit
10048 Like @option{-fstack-protector} but only protects those functions which
10049 have the @code{stack_protect} attribute
10052 @opindex fstdarg-opt
10053 Optimize the prologue of variadic argument functions with respect to usage of
10056 @item -fsection-anchors
10057 @opindex fsection-anchors
10058 Try to reduce the number of symbolic address calculations by using
10059 shared ``anchor'' symbols to address nearby objects. This transformation
10060 can help to reduce the number of GOT entries and GOT accesses on some
10063 For example, the implementation of the following function @code{foo}:
10066 static int a, b, c;
10067 int foo (void) @{ return a + b + c; @}
10071 usually calculates the addresses of all three variables, but if you
10072 compile it with @option{-fsection-anchors}, it accesses the variables
10073 from a common anchor point instead. The effect is similar to the
10074 following pseudocode (which isn't valid C):
10079 register int *xr = &x;
10080 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10084 Not all targets support this option.
10086 @item --param @var{name}=@var{value}
10088 In some places, GCC uses various constants to control the amount of
10089 optimization that is done. For example, GCC does not inline functions
10090 that contain more than a certain number of instructions. You can
10091 control some of these constants on the command line using the
10092 @option{--param} option.
10094 The names of specific parameters, and the meaning of the values, are
10095 tied to the internals of the compiler, and are subject to change
10096 without notice in future releases.
10098 In each case, the @var{value} is an integer. The allowable choices for
10102 @item predictable-branch-outcome
10103 When branch is predicted to be taken with probability lower than this threshold
10104 (in percent), then it is considered well predictable. The default is 10.
10106 @item max-crossjump-edges
10107 The maximum number of incoming edges to consider for cross-jumping.
10108 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10109 the number of edges incoming to each block. Increasing values mean
10110 more aggressive optimization, making the compilation time increase with
10111 probably small improvement in executable size.
10113 @item min-crossjump-insns
10114 The minimum number of instructions that must be matched at the end
10115 of two blocks before cross-jumping is performed on them. This
10116 value is ignored in the case where all instructions in the block being
10117 cross-jumped from are matched. The default value is 5.
10119 @item max-grow-copy-bb-insns
10120 The maximum code size expansion factor when copying basic blocks
10121 instead of jumping. The expansion is relative to a jump instruction.
10122 The default value is 8.
10124 @item max-goto-duplication-insns
10125 The maximum number of instructions to duplicate to a block that jumps
10126 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10127 passes, GCC factors computed gotos early in the compilation process,
10128 and unfactors them as late as possible. Only computed jumps at the
10129 end of a basic blocks with no more than max-goto-duplication-insns are
10130 unfactored. The default value is 8.
10132 @item max-delay-slot-insn-search
10133 The maximum number of instructions to consider when looking for an
10134 instruction to fill a delay slot. If more than this arbitrary number of
10135 instructions are searched, the time savings from filling the delay slot
10136 are minimal, so stop searching. Increasing values mean more
10137 aggressive optimization, making the compilation time increase with probably
10138 small improvement in execution time.
10140 @item max-delay-slot-live-search
10141 When trying to fill delay slots, the maximum number of instructions to
10142 consider when searching for a block with valid live register
10143 information. Increasing this arbitrarily chosen value means more
10144 aggressive optimization, increasing the compilation time. This parameter
10145 should be removed when the delay slot code is rewritten to maintain the
10146 control-flow graph.
10148 @item max-gcse-memory
10149 The approximate maximum amount of memory that can be allocated in
10150 order to perform the global common subexpression elimination
10151 optimization. If more memory than specified is required, the
10152 optimization is not done.
10154 @item max-gcse-insertion-ratio
10155 If the ratio of expression insertions to deletions is larger than this value
10156 for any expression, then RTL PRE inserts or removes the expression and thus
10157 leaves partially redundant computations in the instruction stream. The default value is 20.
10159 @item max-pending-list-length
10160 The maximum number of pending dependencies scheduling allows
10161 before flushing the current state and starting over. Large functions
10162 with few branches or calls can create excessively large lists which
10163 needlessly consume memory and resources.
10165 @item max-modulo-backtrack-attempts
10166 The maximum number of backtrack attempts the scheduler should make
10167 when modulo scheduling a loop. Larger values can exponentially increase
10170 @item max-inline-insns-single
10171 Several parameters control the tree inliner used in GCC@.
10172 This number sets the maximum number of instructions (counted in GCC's
10173 internal representation) in a single function that the tree inliner
10174 considers for inlining. This only affects functions declared
10175 inline and methods implemented in a class declaration (C++).
10176 The default value is 400.
10178 @item max-inline-insns-auto
10179 When you use @option{-finline-functions} (included in @option{-O3}),
10180 a lot of functions that would otherwise not be considered for inlining
10181 by the compiler are investigated. To those functions, a different
10182 (more restrictive) limit compared to functions declared inline can
10184 The default value is 40.
10186 @item inline-min-speedup
10187 When estimated performance improvement of caller + callee runtime exceeds this
10188 threshold (in precent), the function can be inlined regardless the limit on
10189 @option{--param max-inline-insns-single} and @option{--param
10190 max-inline-insns-auto}.
10192 @item large-function-insns
10193 The limit specifying really large functions. For functions larger than this
10194 limit after inlining, inlining is constrained by
10195 @option{--param large-function-growth}. This parameter is useful primarily
10196 to avoid extreme compilation time caused by non-linear algorithms used by the
10198 The default value is 2700.
10200 @item large-function-growth
10201 Specifies maximal growth of large function caused by inlining in percents.
10202 The default value is 100 which limits large function growth to 2.0 times
10205 @item large-unit-insns
10206 The limit specifying large translation unit. Growth caused by inlining of
10207 units larger than this limit is limited by @option{--param inline-unit-growth}.
10208 For small units this might be too tight.
10209 For example, consider a unit consisting of function A
10210 that is inline and B that just calls A three times. If B is small relative to
10211 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10212 large units consisting of small inlineable functions, however, the overall unit
10213 growth limit is needed to avoid exponential explosion of code size. Thus for
10214 smaller units, the size is increased to @option{--param large-unit-insns}
10215 before applying @option{--param inline-unit-growth}. The default is 10000.
10217 @item inline-unit-growth
10218 Specifies maximal overall growth of the compilation unit caused by inlining.
10219 The default value is 20 which limits unit growth to 1.2 times the original
10220 size. Cold functions (either marked cold via an attribute or by profile
10221 feedback) are not accounted into the unit size.
10223 @item ipcp-unit-growth
10224 Specifies maximal overall growth of the compilation unit caused by
10225 interprocedural constant propagation. The default value is 10 which limits
10226 unit growth to 1.1 times the original size.
10228 @item large-stack-frame
10229 The limit specifying large stack frames. While inlining the algorithm is trying
10230 to not grow past this limit too much. The default value is 256 bytes.
10232 @item large-stack-frame-growth
10233 Specifies maximal growth of large stack frames caused by inlining in percents.
10234 The default value is 1000 which limits large stack frame growth to 11 times
10237 @item max-inline-insns-recursive
10238 @itemx max-inline-insns-recursive-auto
10239 Specifies the maximum number of instructions an out-of-line copy of a
10240 self-recursive inline
10241 function can grow into by performing recursive inlining.
10243 @option{--param max-inline-insns-recursive} applies to functions
10245 For functions not declared inline, recursive inlining
10246 happens only when @option{-finline-functions} (included in @option{-O3}) is
10247 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10248 default value is 450.
10250 @item max-inline-recursive-depth
10251 @itemx max-inline-recursive-depth-auto
10252 Specifies the maximum recursion depth used for recursive inlining.
10254 @option{--param max-inline-recursive-depth} applies to functions
10255 declared inline. For functions not declared inline, recursive inlining
10256 happens only when @option{-finline-functions} (included in @option{-O3}) is
10257 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10258 default value is 8.
10260 @item min-inline-recursive-probability
10261 Recursive inlining is profitable only for function having deep recursion
10262 in average and can hurt for function having little recursion depth by
10263 increasing the prologue size or complexity of function body to other
10266 When profile feedback is available (see @option{-fprofile-generate}) the actual
10267 recursion depth can be guessed from probability that function recurses via a
10268 given call expression. This parameter limits inlining only to call expressions
10269 whose probability exceeds the given threshold (in percents).
10270 The default value is 10.
10272 @item early-inlining-insns
10273 Specify growth that the early inliner can make. In effect it increases
10274 the amount of inlining for code having a large abstraction penalty.
10275 The default value is 14.
10277 @item max-early-inliner-iterations
10278 Limit of iterations of the early inliner. This basically bounds
10279 the number of nested indirect calls the early inliner can resolve.
10280 Deeper chains are still handled by late inlining.
10282 @item comdat-sharing-probability
10283 Probability (in percent) that C++ inline function with comdat visibility
10284 are shared across multiple compilation units. The default value is 20.
10286 @item profile-func-internal-id
10287 A parameter to control whether to use function internal id in profile
10288 database lookup. If the value is 0, the compiler uses an id that
10289 is based on function assembler name and filename, which makes old profile
10290 data more tolerant to source changes such as function reordering etc.
10291 The default value is 0.
10293 @item min-vect-loop-bound
10294 The minimum number of iterations under which loops are not vectorized
10295 when @option{-ftree-vectorize} is used. The number of iterations after
10296 vectorization needs to be greater than the value specified by this option
10297 to allow vectorization. The default value is 0.
10299 @item gcse-cost-distance-ratio
10300 Scaling factor in calculation of maximum distance an expression
10301 can be moved by GCSE optimizations. This is currently supported only in the
10302 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10303 is with simple expressions, i.e., the expressions that have cost
10304 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10305 hoisting of simple expressions. The default value is 10.
10307 @item gcse-unrestricted-cost
10308 Cost, roughly measured as the cost of a single typical machine
10309 instruction, at which GCSE optimizations do not constrain
10310 the distance an expression can travel. This is currently
10311 supported only in the code hoisting pass. The lesser the cost,
10312 the more aggressive code hoisting is. Specifying 0
10313 allows all expressions to travel unrestricted distances.
10314 The default value is 3.
10316 @item max-hoist-depth
10317 The depth of search in the dominator tree for expressions to hoist.
10318 This is used to avoid quadratic behavior in hoisting algorithm.
10319 The value of 0 does not limit on the search, but may slow down compilation
10320 of huge functions. The default value is 30.
10322 @item max-tail-merge-comparisons
10323 The maximum amount of similar bbs to compare a bb with. This is used to
10324 avoid quadratic behavior in tree tail merging. The default value is 10.
10326 @item max-tail-merge-iterations
10327 The maximum amount of iterations of the pass over the function. This is used to
10328 limit compilation time in tree tail merging. The default value is 2.
10330 @item max-unrolled-insns
10331 The maximum number of instructions that a loop may have to be unrolled.
10332 If a loop is unrolled, this parameter also determines how many times
10333 the loop code is unrolled.
10335 @item max-average-unrolled-insns
10336 The maximum number of instructions biased by probabilities of their execution
10337 that a loop may have to be unrolled. If a loop is unrolled,
10338 this parameter also determines how many times the loop code is unrolled.
10340 @item max-unroll-times
10341 The maximum number of unrollings of a single loop.
10343 @item max-peeled-insns
10344 The maximum number of instructions that a loop may have to be peeled.
10345 If a loop is peeled, this parameter also determines how many times
10346 the loop code is peeled.
10348 @item max-peel-times
10349 The maximum number of peelings of a single loop.
10351 @item max-peel-branches
10352 The maximum number of branches on the hot path through the peeled sequence.
10354 @item max-completely-peeled-insns
10355 The maximum number of insns of a completely peeled loop.
10357 @item max-completely-peel-times
10358 The maximum number of iterations of a loop to be suitable for complete peeling.
10360 @item max-completely-peel-loop-nest-depth
10361 The maximum depth of a loop nest suitable for complete peeling.
10363 @item max-unswitch-insns
10364 The maximum number of insns of an unswitched loop.
10366 @item max-unswitch-level
10367 The maximum number of branches unswitched in a single loop.
10369 @item lim-expensive
10370 The minimum cost of an expensive expression in the loop invariant motion.
10372 @item iv-consider-all-candidates-bound
10373 Bound on number of candidates for induction variables, below which
10374 all candidates are considered for each use in induction variable
10375 optimizations. If there are more candidates than this,
10376 only the most relevant ones are considered to avoid quadratic time complexity.
10378 @item iv-max-considered-uses
10379 The induction variable optimizations give up on loops that contain more
10380 induction variable uses.
10382 @item iv-always-prune-cand-set-bound
10383 If the number of candidates in the set is smaller than this value,
10384 always try to remove unnecessary ivs from the set
10385 when adding a new one.
10387 @item scev-max-expr-size
10388 Bound on size of expressions used in the scalar evolutions analyzer.
10389 Large expressions slow the analyzer.
10391 @item scev-max-expr-complexity
10392 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10393 Complex expressions slow the analyzer.
10395 @item omega-max-vars
10396 The maximum number of variables in an Omega constraint system.
10397 The default value is 128.
10399 @item omega-max-geqs
10400 The maximum number of inequalities in an Omega constraint system.
10401 The default value is 256.
10403 @item omega-max-eqs
10404 The maximum number of equalities in an Omega constraint system.
10405 The default value is 128.
10407 @item omega-max-wild-cards
10408 The maximum number of wildcard variables that the Omega solver is
10409 able to insert. The default value is 18.
10411 @item omega-hash-table-size
10412 The size of the hash table in the Omega solver. The default value is
10415 @item omega-max-keys
10416 The maximal number of keys used by the Omega solver. The default
10419 @item omega-eliminate-redundant-constraints
10420 When set to 1, use expensive methods to eliminate all redundant
10421 constraints. The default value is 0.
10423 @item vect-max-version-for-alignment-checks
10424 The maximum number of run-time checks that can be performed when
10425 doing loop versioning for alignment in the vectorizer.
10427 @item vect-max-version-for-alias-checks
10428 The maximum number of run-time checks that can be performed when
10429 doing loop versioning for alias in the vectorizer.
10431 @item vect-max-peeling-for-alignment
10432 The maximum number of loop peels to enhance access alignment
10433 for vectorizer. Value -1 means 'no limit'.
10435 @item max-iterations-to-track
10436 The maximum number of iterations of a loop the brute-force algorithm
10437 for analysis of the number of iterations of the loop tries to evaluate.
10439 @item hot-bb-count-ws-permille
10440 A basic block profile count is considered hot if it contributes to
10441 the given permillage (i.e. 0...1000) of the entire profiled execution.
10443 @item hot-bb-frequency-fraction
10444 Select fraction of the entry block frequency of executions of basic block in
10445 function given basic block needs to have to be considered hot.
10447 @item max-predicted-iterations
10448 The maximum number of loop iterations we predict statically. This is useful
10449 in cases where a function contains a single loop with known bound and
10450 another loop with unknown bound.
10451 The known number of iterations is predicted correctly, while
10452 the unknown number of iterations average to roughly 10. This means that the
10453 loop without bounds appears artificially cold relative to the other one.
10455 @item builtin-expect-probability
10456 Control the probability of the expression having the specified value. This
10457 parameter takes a percentage (i.e. 0 ... 100) as input.
10458 The default probability of 90 is obtained empirically.
10460 @item align-threshold
10462 Select fraction of the maximal frequency of executions of a basic block in
10463 a function to align the basic block.
10465 @item align-loop-iterations
10467 A loop expected to iterate at least the selected number of iterations is
10470 @item tracer-dynamic-coverage
10471 @itemx tracer-dynamic-coverage-feedback
10473 This value is used to limit superblock formation once the given percentage of
10474 executed instructions is covered. This limits unnecessary code size
10477 The @option{tracer-dynamic-coverage-feedback} parameter
10478 is used only when profile
10479 feedback is available. The real profiles (as opposed to statically estimated
10480 ones) are much less balanced allowing the threshold to be larger value.
10482 @item tracer-max-code-growth
10483 Stop tail duplication once code growth has reached given percentage. This is
10484 a rather artificial limit, as most of the duplicates are eliminated later in
10485 cross jumping, so it may be set to much higher values than is the desired code
10488 @item tracer-min-branch-ratio
10490 Stop reverse growth when the reverse probability of best edge is less than this
10491 threshold (in percent).
10493 @item tracer-min-branch-ratio
10494 @itemx tracer-min-branch-ratio-feedback
10496 Stop forward growth if the best edge has probability lower than this
10499 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10500 compilation for profile feedback and one for compilation without. The value
10501 for compilation with profile feedback needs to be more conservative (higher) in
10502 order to make tracer effective.
10504 @item max-cse-path-length
10506 The maximum number of basic blocks on path that CSE considers.
10509 @item max-cse-insns
10510 The maximum number of instructions CSE processes before flushing.
10511 The default is 1000.
10513 @item ggc-min-expand
10515 GCC uses a garbage collector to manage its own memory allocation. This
10516 parameter specifies the minimum percentage by which the garbage
10517 collector's heap should be allowed to expand between collections.
10518 Tuning this may improve compilation speed; it has no effect on code
10521 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10522 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10523 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10524 GCC is not able to calculate RAM on a particular platform, the lower
10525 bound of 30% is used. Setting this parameter and
10526 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10527 every opportunity. This is extremely slow, but can be useful for
10530 @item ggc-min-heapsize
10532 Minimum size of the garbage collector's heap before it begins bothering
10533 to collect garbage. The first collection occurs after the heap expands
10534 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10535 tuning this may improve compilation speed, and has no effect on code
10538 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10539 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10540 with a lower bound of 4096 (four megabytes) and an upper bound of
10541 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10542 particular platform, the lower bound is used. Setting this parameter
10543 very large effectively disables garbage collection. Setting this
10544 parameter and @option{ggc-min-expand} to zero causes a full collection
10545 to occur at every opportunity.
10547 @item max-reload-search-insns
10548 The maximum number of instruction reload should look backward for equivalent
10549 register. Increasing values mean more aggressive optimization, making the
10550 compilation time increase with probably slightly better performance.
10551 The default value is 100.
10553 @item max-cselib-memory-locations
10554 The maximum number of memory locations cselib should take into account.
10555 Increasing values mean more aggressive optimization, making the compilation time
10556 increase with probably slightly better performance. The default value is 500.
10558 @item reorder-blocks-duplicate
10559 @itemx reorder-blocks-duplicate-feedback
10561 Used by the basic block reordering pass to decide whether to use unconditional
10562 branch or duplicate the code on its destination. Code is duplicated when its
10563 estimated size is smaller than this value multiplied by the estimated size of
10564 unconditional jump in the hot spots of the program.
10566 The @option{reorder-block-duplicate-feedback} parameter
10567 is used only when profile
10568 feedback is available. It may be set to higher values than
10569 @option{reorder-block-duplicate} since information about the hot spots is more
10572 @item max-sched-ready-insns
10573 The maximum number of instructions ready to be issued the scheduler should
10574 consider at any given time during the first scheduling pass. Increasing
10575 values mean more thorough searches, making the compilation time increase
10576 with probably little benefit. The default value is 100.
10578 @item max-sched-region-blocks
10579 The maximum number of blocks in a region to be considered for
10580 interblock scheduling. The default value is 10.
10582 @item max-pipeline-region-blocks
10583 The maximum number of blocks in a region to be considered for
10584 pipelining in the selective scheduler. The default value is 15.
10586 @item max-sched-region-insns
10587 The maximum number of insns in a region to be considered for
10588 interblock scheduling. The default value is 100.
10590 @item max-pipeline-region-insns
10591 The maximum number of insns in a region to be considered for
10592 pipelining in the selective scheduler. The default value is 200.
10594 @item min-spec-prob
10595 The minimum probability (in percents) of reaching a source block
10596 for interblock speculative scheduling. The default value is 40.
10598 @item max-sched-extend-regions-iters
10599 The maximum number of iterations through CFG to extend regions.
10600 A value of 0 (the default) disables region extensions.
10602 @item max-sched-insn-conflict-delay
10603 The maximum conflict delay for an insn to be considered for speculative motion.
10604 The default value is 3.
10606 @item sched-spec-prob-cutoff
10607 The minimal probability of speculation success (in percents), so that
10608 speculative insns are scheduled.
10609 The default value is 40.
10611 @item sched-spec-state-edge-prob-cutoff
10612 The minimum probability an edge must have for the scheduler to save its
10614 The default value is 10.
10616 @item sched-mem-true-dep-cost
10617 Minimal distance (in CPU cycles) between store and load targeting same
10618 memory locations. The default value is 1.
10620 @item selsched-max-lookahead
10621 The maximum size of the lookahead window of selective scheduling. It is a
10622 depth of search for available instructions.
10623 The default value is 50.
10625 @item selsched-max-sched-times
10626 The maximum number of times that an instruction is scheduled during
10627 selective scheduling. This is the limit on the number of iterations
10628 through which the instruction may be pipelined. The default value is 2.
10630 @item selsched-max-insns-to-rename
10631 The maximum number of best instructions in the ready list that are considered
10632 for renaming in the selective scheduler. The default value is 2.
10635 The minimum value of stage count that swing modulo scheduler
10636 generates. The default value is 2.
10638 @item max-last-value-rtl
10639 The maximum size measured as number of RTLs that can be recorded in an expression
10640 in combiner for a pseudo register as last known value of that register. The default
10643 @item max-combine-insns
10644 The maximum number of instructions the RTL combiner tries to combine.
10645 The default value is 2 at @option{-Og} and 4 otherwise.
10647 @item integer-share-limit
10648 Small integer constants can use a shared data structure, reducing the
10649 compiler's memory usage and increasing its speed. This sets the maximum
10650 value of a shared integer constant. The default value is 256.
10652 @item ssp-buffer-size
10653 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10654 protection when @option{-fstack-protection} is used.
10656 @item min-size-for-stack-sharing
10657 The minimum size of variables taking part in stack slot sharing when not
10658 optimizing. The default value is 32.
10660 @item max-jump-thread-duplication-stmts
10661 Maximum number of statements allowed in a block that needs to be
10662 duplicated when threading jumps.
10664 @item max-fields-for-field-sensitive
10665 Maximum number of fields in a structure treated in
10666 a field sensitive manner during pointer analysis. The default is zero
10667 for @option{-O0} and @option{-O1},
10668 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10670 @item prefetch-latency
10671 Estimate on average number of instructions that are executed before
10672 prefetch finishes. The distance prefetched ahead is proportional
10673 to this constant. Increasing this number may also lead to less
10674 streams being prefetched (see @option{simultaneous-prefetches}).
10676 @item simultaneous-prefetches
10677 Maximum number of prefetches that can run at the same time.
10679 @item l1-cache-line-size
10680 The size of cache line in L1 cache, in bytes.
10682 @item l1-cache-size
10683 The size of L1 cache, in kilobytes.
10685 @item l2-cache-size
10686 The size of L2 cache, in kilobytes.
10688 @item min-insn-to-prefetch-ratio
10689 The minimum ratio between the number of instructions and the
10690 number of prefetches to enable prefetching in a loop.
10692 @item prefetch-min-insn-to-mem-ratio
10693 The minimum ratio between the number of instructions and the
10694 number of memory references to enable prefetching in a loop.
10696 @item use-canonical-types
10697 Whether the compiler should use the ``canonical'' type system. By
10698 default, this should always be 1, which uses a more efficient internal
10699 mechanism for comparing types in C++ and Objective-C++. However, if
10700 bugs in the canonical type system are causing compilation failures,
10701 set this value to 0 to disable canonical types.
10703 @item switch-conversion-max-branch-ratio
10704 Switch initialization conversion refuses to create arrays that are
10705 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10706 branches in the switch.
10708 @item max-partial-antic-length
10709 Maximum length of the partial antic set computed during the tree
10710 partial redundancy elimination optimization (@option{-ftree-pre}) when
10711 optimizing at @option{-O3} and above. For some sorts of source code
10712 the enhanced partial redundancy elimination optimization can run away,
10713 consuming all of the memory available on the host machine. This
10714 parameter sets a limit on the length of the sets that are computed,
10715 which prevents the runaway behavior. Setting a value of 0 for
10716 this parameter allows an unlimited set length.
10718 @item sccvn-max-scc-size
10719 Maximum size of a strongly connected component (SCC) during SCCVN
10720 processing. If this limit is hit, SCCVN processing for the whole
10721 function is not done and optimizations depending on it are
10722 disabled. The default maximum SCC size is 10000.
10724 @item sccvn-max-alias-queries-per-access
10725 Maximum number of alias-oracle queries we perform when looking for
10726 redundancies for loads and stores. If this limit is hit the search
10727 is aborted and the load or store is not considered redundant. The
10728 number of queries is algorithmically limited to the number of
10729 stores on all paths from the load to the function entry.
10730 The default maxmimum number of queries is 1000.
10732 @item ira-max-loops-num
10733 IRA uses regional register allocation by default. If a function
10734 contains more loops than the number given by this parameter, only at most
10735 the given number of the most frequently-executed loops form regions
10736 for regional register allocation. The default value of the
10739 @item ira-max-conflict-table-size
10740 Although IRA uses a sophisticated algorithm to compress the conflict
10741 table, the table can still require excessive amounts of memory for
10742 huge functions. If the conflict table for a function could be more
10743 than the size in MB given by this parameter, the register allocator
10744 instead uses a faster, simpler, and lower-quality
10745 algorithm that does not require building a pseudo-register conflict table.
10746 The default value of the parameter is 2000.
10748 @item ira-loop-reserved-regs
10749 IRA can be used to evaluate more accurate register pressure in loops
10750 for decisions to move loop invariants (see @option{-O3}). The number
10751 of available registers reserved for some other purposes is given
10752 by this parameter. The default value of the parameter is 2, which is
10753 the minimal number of registers needed by typical instructions.
10754 This value is the best found from numerous experiments.
10756 @item lra-inheritance-ebb-probability-cutoff
10757 LRA tries to reuse values reloaded in registers in subsequent insns.
10758 This optimization is called inheritance. EBB is used as a region to
10759 do this optimization. The parameter defines a minimal fall-through
10760 edge probability in percentage used to add BB to inheritance EBB in
10761 LRA. The default value of the parameter is 40. The value was chosen
10762 from numerous runs of SPEC2000 on x86-64.
10764 @item loop-invariant-max-bbs-in-loop
10765 Loop invariant motion can be very expensive, both in compilation time and
10766 in amount of needed compile-time memory, with very large loops. Loops
10767 with more basic blocks than this parameter won't have loop invariant
10768 motion optimization performed on them. The default value of the
10769 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10771 @item loop-max-datarefs-for-datadeps
10772 Building data dapendencies is expensive for very large loops. This
10773 parameter limits the number of data references in loops that are
10774 considered for data dependence analysis. These large loops are no
10775 handled by the optimizations using loop data dependencies.
10776 The default value is 1000.
10778 @item max-vartrack-size
10779 Sets a maximum number of hash table slots to use during variable
10780 tracking dataflow analysis of any function. If this limit is exceeded
10781 with variable tracking at assignments enabled, analysis for that
10782 function is retried without it, after removing all debug insns from
10783 the function. If the limit is exceeded even without debug insns, var
10784 tracking analysis is completely disabled for the function. Setting
10785 the parameter to zero makes it unlimited.
10787 @item max-vartrack-expr-depth
10788 Sets a maximum number of recursion levels when attempting to map
10789 variable names or debug temporaries to value expressions. This trades
10790 compilation time for more complete debug information. If this is set too
10791 low, value expressions that are available and could be represented in
10792 debug information may end up not being used; setting this higher may
10793 enable the compiler to find more complex debug expressions, but compile
10794 time and memory use may grow. The default is 12.
10796 @item min-nondebug-insn-uid
10797 Use uids starting at this parameter for nondebug insns. The range below
10798 the parameter is reserved exclusively for debug insns created by
10799 @option{-fvar-tracking-assignments}, but debug insns may get
10800 (non-overlapping) uids above it if the reserved range is exhausted.
10802 @item ipa-sra-ptr-growth-factor
10803 IPA-SRA replaces a pointer to an aggregate with one or more new
10804 parameters only when their cumulative size is less or equal to
10805 @option{ipa-sra-ptr-growth-factor} times the size of the original
10808 @item sra-max-scalarization-size-Ospeed
10809 @item sra-max-scalarization-size-Osize
10810 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10811 replace scalar parts of aggregates with uses of independent scalar
10812 variables. These parameters control the maximum size, in storage units,
10813 of aggregate which is considered for replacement when compiling for
10815 (@option{sra-max-scalarization-size-Ospeed}) or size
10816 (@option{sra-max-scalarization-size-Osize}) respectively.
10818 @item tm-max-aggregate-size
10819 When making copies of thread-local variables in a transaction, this
10820 parameter specifies the size in bytes after which variables are
10821 saved with the logging functions as opposed to save/restore code
10822 sequence pairs. This option only applies when using
10825 @item graphite-max-nb-scop-params
10826 To avoid exponential effects in the Graphite loop transforms, the
10827 number of parameters in a Static Control Part (SCoP) is bounded. The
10828 default value is 10 parameters. A variable whose value is unknown at
10829 compilation time and defined outside a SCoP is a parameter of the SCoP.
10831 @item graphite-max-bbs-per-function
10832 To avoid exponential effects in the detection of SCoPs, the size of
10833 the functions analyzed by Graphite is bounded. The default value is
10836 @item loop-block-tile-size
10837 Loop blocking or strip mining transforms, enabled with
10838 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10839 loop in the loop nest by a given number of iterations. The strip
10840 length can be changed using the @option{loop-block-tile-size}
10841 parameter. The default value is 51 iterations.
10843 @item loop-unroll-jam-size
10844 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10845 default value is 4.
10847 @item loop-unroll-jam-depth
10848 Specify the dimension to be unrolled (counting from the most inner loop)
10849 for the @option{-floop-unroll-and-jam}. The default value is 2.
10851 @item ipa-cp-value-list-size
10852 IPA-CP attempts to track all possible values and types passed to a function's
10853 parameter in order to propagate them and perform devirtualization.
10854 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10855 stores per one formal parameter of a function.
10857 @item ipa-cp-eval-threshold
10858 IPA-CP calculates its own score of cloning profitability heuristics
10859 and performs those cloning opportunities with scores that exceed
10860 @option{ipa-cp-eval-threshold}.
10862 @item ipa-cp-recursion-penalty
10863 Percentage penalty the recursive functions will receive when they
10864 are evaluated for cloning.
10866 @item ipa-cp-single-call-penalty
10867 Percentage penalty functions containg a single call to another
10868 function will receive when they are evaluated for cloning.
10871 @item ipa-max-agg-items
10872 IPA-CP is also capable to propagate a number of scalar values passed
10873 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10874 number of such values per one parameter.
10876 @item ipa-cp-loop-hint-bonus
10877 When IPA-CP determines that a cloning candidate would make the number
10878 of iterations of a loop known, it adds a bonus of
10879 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10882 @item ipa-cp-array-index-hint-bonus
10883 When IPA-CP determines that a cloning candidate would make the index of
10884 an array access known, it adds a bonus of
10885 @option{ipa-cp-array-index-hint-bonus} to the profitability
10886 score of the candidate.
10888 @item ipa-max-aa-steps
10889 During its analysis of function bodies, IPA-CP employs alias analysis
10890 in order to track values pointed to by function parameters. In order
10891 not spend too much time analyzing huge functions, it gives up and
10892 consider all memory clobbered after examining
10893 @option{ipa-max-aa-steps} statements modifying memory.
10895 @item lto-partitions
10896 Specify desired number of partitions produced during WHOPR compilation.
10897 The number of partitions should exceed the number of CPUs used for compilation.
10898 The default value is 32.
10900 @item lto-minpartition
10901 Size of minimal partition for WHOPR (in estimated instructions).
10902 This prevents expenses of splitting very small programs into too many
10905 @item cxx-max-namespaces-for-diagnostic-help
10906 The maximum number of namespaces to consult for suggestions when C++
10907 name lookup fails for an identifier. The default is 1000.
10909 @item sink-frequency-threshold
10910 The maximum relative execution frequency (in percents) of the target block
10911 relative to a statement's original block to allow statement sinking of a
10912 statement. Larger numbers result in more aggressive statement sinking.
10913 The default value is 75. A small positive adjustment is applied for
10914 statements with memory operands as those are even more profitable so sink.
10916 @item max-stores-to-sink
10917 The maximum number of conditional stores paires that can be sunk. Set to 0
10918 if either vectorization (@option{-ftree-vectorize}) or if-conversion
10919 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
10921 @item allow-store-data-races
10922 Allow optimizers to introduce new data races on stores.
10923 Set to 1 to allow, otherwise to 0. This option is enabled by default
10924 at optimization level @option{-Ofast}.
10926 @item case-values-threshold
10927 The smallest number of different values for which it is best to use a
10928 jump-table instead of a tree of conditional branches. If the value is
10929 0, use the default for the machine. The default is 0.
10931 @item tree-reassoc-width
10932 Set the maximum number of instructions executed in parallel in
10933 reassociated tree. This parameter overrides target dependent
10934 heuristics used by default if has non zero value.
10936 @item sched-pressure-algorithm
10937 Choose between the two available implementations of
10938 @option{-fsched-pressure}. Algorithm 1 is the original implementation
10939 and is the more likely to prevent instructions from being reordered.
10940 Algorithm 2 was designed to be a compromise between the relatively
10941 conservative approach taken by algorithm 1 and the rather aggressive
10942 approach taken by the default scheduler. It relies more heavily on
10943 having a regular register file and accurate register pressure classes.
10944 See @file{haifa-sched.c} in the GCC sources for more details.
10946 The default choice depends on the target.
10948 @item max-slsr-cand-scan
10949 Set the maximum number of existing candidates that are considered when
10950 seeking a basis for a new straight-line strength reduction candidate.
10953 Enable buffer overflow detection for global objects. This kind
10954 of protection is enabled by default if you are using
10955 @option{-fsanitize=address} option.
10956 To disable global objects protection use @option{--param asan-globals=0}.
10959 Enable buffer overflow detection for stack objects. This kind of
10960 protection is enabled by default when using@option{-fsanitize=address}.
10961 To disable stack protection use @option{--param asan-stack=0} option.
10963 @item asan-instrument-reads
10964 Enable buffer overflow detection for memory reads. This kind of
10965 protection is enabled by default when using @option{-fsanitize=address}.
10966 To disable memory reads protection use
10967 @option{--param asan-instrument-reads=0}.
10969 @item asan-instrument-writes
10970 Enable buffer overflow detection for memory writes. This kind of
10971 protection is enabled by default when using @option{-fsanitize=address}.
10972 To disable memory writes protection use
10973 @option{--param asan-instrument-writes=0} option.
10975 @item asan-memintrin
10976 Enable detection for built-in functions. This kind of protection
10977 is enabled by default when using @option{-fsanitize=address}.
10978 To disable built-in functions protection use
10979 @option{--param asan-memintrin=0}.
10981 @item asan-use-after-return
10982 Enable detection of use-after-return. This kind of protection
10983 is enabled by default when using @option{-fsanitize=address} option.
10984 To disable use-after-return detection use
10985 @option{--param asan-use-after-return=0}.
10987 @item asan-instrumentation-with-call-threshold
10988 If number of memory accesses in function being instrumented
10989 is greater or equal to this number, use callbacks instead of inline checks.
10990 E.g. to disable inline code use
10991 @option{--param asan-instrumentation-with-call-threshold=0}.
10993 @item chkp-max-ctor-size
10994 Static constructors generated by Pointer Bounds Checker may become very
10995 large and significantly increase compile time at optimization level
10996 @option{-O1} and higher. This parameter is a maximum nubmer of statements
10997 in a single generated constructor. Default value is 5000.
10999 @item max-fsm-thread-path-insns
11000 Maximum number of instructions to copy when duplicating blocks on a
11001 finite state automaton jump thread path. The default is 100.
11003 @item max-fsm-thread-length
11004 Maximum number of basic blocks on a finite state automaton jump thread
11005 path. The default is 10.
11007 @item max-fsm-thread-paths
11008 Maximum number of new jump thread paths to create for a finite state
11009 automaton. The default is 50.
11014 @node Preprocessor Options
11015 @section Options Controlling the Preprocessor
11016 @cindex preprocessor options
11017 @cindex options, preprocessor
11019 These options control the C preprocessor, which is run on each C source
11020 file before actual compilation.
11022 If you use the @option{-E} option, nothing is done except preprocessing.
11023 Some of these options make sense only together with @option{-E} because
11024 they cause the preprocessor output to be unsuitable for actual
11028 @item -Wp,@var{option}
11030 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11031 and pass @var{option} directly through to the preprocessor. If
11032 @var{option} contains commas, it is split into multiple options at the
11033 commas. However, many options are modified, translated or interpreted
11034 by the compiler driver before being passed to the preprocessor, and
11035 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11036 interface is undocumented and subject to change, so whenever possible
11037 you should avoid using @option{-Wp} and let the driver handle the
11040 @item -Xpreprocessor @var{option}
11041 @opindex Xpreprocessor
11042 Pass @var{option} as an option to the preprocessor. You can use this to
11043 supply system-specific preprocessor options that GCC does not
11046 If you want to pass an option that takes an argument, you must use
11047 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11049 @item -no-integrated-cpp
11050 @opindex no-integrated-cpp
11051 Perform preprocessing as a separate pass before compilation.
11052 By default, GCC performs preprocessing as an integrated part of
11053 input tokenization and parsing.
11054 If this option is provided, the appropriate language front end
11055 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11056 and Objective-C, respectively) is instead invoked twice,
11057 once for preprocessing only and once for actual compilation
11058 of the preprocessed input.
11059 This option may be useful in conjunction with the @option{-B} or
11060 @option{-wrapper} options to specify an alternate preprocessor or
11061 perform additional processing of the program source between
11062 normal preprocessing and compilation.
11065 @include cppopts.texi
11067 @node Assembler Options
11068 @section Passing Options to the Assembler
11070 @c prevent bad page break with this line
11071 You can pass options to the assembler.
11074 @item -Wa,@var{option}
11076 Pass @var{option} as an option to the assembler. If @var{option}
11077 contains commas, it is split into multiple options at the commas.
11079 @item -Xassembler @var{option}
11080 @opindex Xassembler
11081 Pass @var{option} as an option to the assembler. You can use this to
11082 supply system-specific assembler options that GCC does not
11085 If you want to pass an option that takes an argument, you must use
11086 @option{-Xassembler} twice, once for the option and once for the argument.
11091 @section Options for Linking
11092 @cindex link options
11093 @cindex options, linking
11095 These options come into play when the compiler links object files into
11096 an executable output file. They are meaningless if the compiler is
11097 not doing a link step.
11101 @item @var{object-file-name}
11102 A file name that does not end in a special recognized suffix is
11103 considered to name an object file or library. (Object files are
11104 distinguished from libraries by the linker according to the file
11105 contents.) If linking is done, these object files are used as input
11114 If any of these options is used, then the linker is not run, and
11115 object file names should not be used as arguments. @xref{Overall
11119 @opindex fuse-ld=bfd
11120 Use the @command{bfd} linker instead of the default linker.
11122 @item -fuse-ld=gold
11123 @opindex fuse-ld=gold
11124 Use the @command{gold} linker instead of the default linker.
11127 @item -l@var{library}
11128 @itemx -l @var{library}
11130 Search the library named @var{library} when linking. (The second
11131 alternative with the library as a separate argument is only for
11132 POSIX compliance and is not recommended.)
11134 It makes a difference where in the command you write this option; the
11135 linker searches and processes libraries and object files in the order they
11136 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11137 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11138 to functions in @samp{z}, those functions may not be loaded.
11140 The linker searches a standard list of directories for the library,
11141 which is actually a file named @file{lib@var{library}.a}. The linker
11142 then uses this file as if it had been specified precisely by name.
11144 The directories searched include several standard system directories
11145 plus any that you specify with @option{-L}.
11147 Normally the files found this way are library files---archive files
11148 whose members are object files. The linker handles an archive file by
11149 scanning through it for members which define symbols that have so far
11150 been referenced but not defined. But if the file that is found is an
11151 ordinary object file, it is linked in the usual fashion. The only
11152 difference between using an @option{-l} option and specifying a file name
11153 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11154 and searches several directories.
11158 You need this special case of the @option{-l} option in order to
11159 link an Objective-C or Objective-C++ program.
11161 @item -nostartfiles
11162 @opindex nostartfiles
11163 Do not use the standard system startup files when linking.
11164 The standard system libraries are used normally, unless @option{-nostdlib}
11165 or @option{-nodefaultlibs} is used.
11167 @item -nodefaultlibs
11168 @opindex nodefaultlibs
11169 Do not use the standard system libraries when linking.
11170 Only the libraries you specify are passed to the linker, and options
11171 specifying linkage of the system libraries, such as @option{-static-libgcc}
11172 or @option{-shared-libgcc}, are ignored.
11173 The standard startup files are used normally, unless @option{-nostartfiles}
11176 The compiler may generate calls to @code{memcmp},
11177 @code{memset}, @code{memcpy} and @code{memmove}.
11178 These entries are usually resolved by entries in
11179 libc. These entry points should be supplied through some other
11180 mechanism when this option is specified.
11184 Do not use the standard system startup files or libraries when linking.
11185 No startup files and only the libraries you specify are passed to
11186 the linker, and options specifying linkage of the system libraries, such as
11187 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11189 The compiler may generate calls to @code{memcmp}, @code{memset},
11190 @code{memcpy} and @code{memmove}.
11191 These entries are usually resolved by entries in
11192 libc. These entry points should be supplied through some other
11193 mechanism when this option is specified.
11195 @cindex @option{-lgcc}, use with @option{-nostdlib}
11196 @cindex @option{-nostdlib} and unresolved references
11197 @cindex unresolved references and @option{-nostdlib}
11198 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11199 @cindex @option{-nodefaultlibs} and unresolved references
11200 @cindex unresolved references and @option{-nodefaultlibs}
11201 One of the standard libraries bypassed by @option{-nostdlib} and
11202 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11203 which GCC uses to overcome shortcomings of particular machines, or special
11204 needs for some languages.
11205 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11206 Collection (GCC) Internals},
11207 for more discussion of @file{libgcc.a}.)
11208 In most cases, you need @file{libgcc.a} even when you want to avoid
11209 other standard libraries. In other words, when you specify @option{-nostdlib}
11210 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11211 This ensures that you have no unresolved references to internal GCC
11212 library subroutines.
11213 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11214 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11215 GNU Compiler Collection (GCC) Internals}.)
11219 Produce a position independent executable on targets that support it.
11220 For predictable results, you must also specify the same set of options
11221 used for compilation (@option{-fpie}, @option{-fPIE},
11222 or model suboptions) when you specify this linker option.
11226 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11227 that support it. This instructs the linker to add all symbols, not
11228 only used ones, to the dynamic symbol table. This option is needed
11229 for some uses of @code{dlopen} or to allow obtaining backtraces
11230 from within a program.
11234 Remove all symbol table and relocation information from the executable.
11238 On systems that support dynamic linking, this prevents linking with the shared
11239 libraries. On other systems, this option has no effect.
11243 Produce a shared object which can then be linked with other objects to
11244 form an executable. Not all systems support this option. For predictable
11245 results, you must also specify the same set of options used for compilation
11246 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11247 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11248 needs to build supplementary stub code for constructors to work. On
11249 multi-libbed systems, @samp{gcc -shared} must select the correct support
11250 libraries to link against. Failing to supply the correct flags may lead
11251 to subtle defects. Supplying them in cases where they are not necessary
11254 @item -shared-libgcc
11255 @itemx -static-libgcc
11256 @opindex shared-libgcc
11257 @opindex static-libgcc
11258 On systems that provide @file{libgcc} as a shared library, these options
11259 force the use of either the shared or static version, respectively.
11260 If no shared version of @file{libgcc} was built when the compiler was
11261 configured, these options have no effect.
11263 There are several situations in which an application should use the
11264 shared @file{libgcc} instead of the static version. The most common
11265 of these is when the application wishes to throw and catch exceptions
11266 across different shared libraries. In that case, each of the libraries
11267 as well as the application itself should use the shared @file{libgcc}.
11269 Therefore, the G++ and GCJ drivers automatically add
11270 @option{-shared-libgcc} whenever you build a shared library or a main
11271 executable, because C++ and Java programs typically use exceptions, so
11272 this is the right thing to do.
11274 If, instead, you use the GCC driver to create shared libraries, you may
11275 find that they are not always linked with the shared @file{libgcc}.
11276 If GCC finds, at its configuration time, that you have a non-GNU linker
11277 or a GNU linker that does not support option @option{--eh-frame-hdr},
11278 it links the shared version of @file{libgcc} into shared libraries
11279 by default. Otherwise, it takes advantage of the linker and optimizes
11280 away the linking with the shared version of @file{libgcc}, linking with
11281 the static version of libgcc by default. This allows exceptions to
11282 propagate through such shared libraries, without incurring relocation
11283 costs at library load time.
11285 However, if a library or main executable is supposed to throw or catch
11286 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11287 for the languages used in the program, or using the option
11288 @option{-shared-libgcc}, such that it is linked with the shared
11291 @item -static-libasan
11292 @opindex static-libasan
11293 When the @option{-fsanitize=address} option is used to link a program,
11294 the GCC driver automatically links against @option{libasan}. If
11295 @file{libasan} is available as a shared library, and the @option{-static}
11296 option is not used, then this links against the shared version of
11297 @file{libasan}. The @option{-static-libasan} option directs the GCC
11298 driver to link @file{libasan} statically, without necessarily linking
11299 other libraries statically.
11301 @item -static-libtsan
11302 @opindex static-libtsan
11303 When the @option{-fsanitize=thread} option is used to link a program,
11304 the GCC driver automatically links against @option{libtsan}. If
11305 @file{libtsan} is available as a shared library, and the @option{-static}
11306 option is not used, then this links against the shared version of
11307 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11308 driver to link @file{libtsan} statically, without necessarily linking
11309 other libraries statically.
11311 @item -static-liblsan
11312 @opindex static-liblsan
11313 When the @option{-fsanitize=leak} option is used to link a program,
11314 the GCC driver automatically links against @option{liblsan}. If
11315 @file{liblsan} is available as a shared library, and the @option{-static}
11316 option is not used, then this links against the shared version of
11317 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11318 driver to link @file{liblsan} statically, without necessarily linking
11319 other libraries statically.
11321 @item -static-libubsan
11322 @opindex static-libubsan
11323 When the @option{-fsanitize=undefined} option is used to link a program,
11324 the GCC driver automatically links against @option{libubsan}. If
11325 @file{libubsan} is available as a shared library, and the @option{-static}
11326 option is not used, then this links against the shared version of
11327 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11328 driver to link @file{libubsan} statically, without necessarily linking
11329 other libraries statically.
11331 @item -static-libmpx
11332 @opindex static-libmpx
11333 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11334 used to link a program, the GCC driver automatically links against
11335 @file{libmpx}. If @file{libmpx} is available as a shared library,
11336 and the @option{-static} option is not used, then this links against
11337 the shared version of @file{libmpx}. The @option{-static-libmpx}
11338 option directs the GCC driver to link @file{libmpx} statically,
11339 without necessarily linking other libraries statically.
11341 @item -static-libmpxwrappers
11342 @opindex static-libmpxwrappers
11343 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11344 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11345 GCC driver automatically links against @file{libmpxwrappers}. If
11346 @file{libmpxwrappers} is available as a shared library, and the
11347 @option{-static} option is not used, then this links against the shared
11348 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11349 option directs the GCC driver to link @file{libmpxwrappers} statically,
11350 without necessarily linking other libraries statically.
11352 @item -static-libstdc++
11353 @opindex static-libstdc++
11354 When the @command{g++} program is used to link a C++ program, it
11355 normally automatically links against @option{libstdc++}. If
11356 @file{libstdc++} is available as a shared library, and the
11357 @option{-static} option is not used, then this links against the
11358 shared version of @file{libstdc++}. That is normally fine. However, it
11359 is sometimes useful to freeze the version of @file{libstdc++} used by
11360 the program without going all the way to a fully static link. The
11361 @option{-static-libstdc++} option directs the @command{g++} driver to
11362 link @file{libstdc++} statically, without necessarily linking other
11363 libraries statically.
11367 Bind references to global symbols when building a shared object. Warn
11368 about any unresolved references (unless overridden by the link editor
11369 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11372 @item -T @var{script}
11374 @cindex linker script
11375 Use @var{script} as the linker script. This option is supported by most
11376 systems using the GNU linker. On some targets, such as bare-board
11377 targets without an operating system, the @option{-T} option may be required
11378 when linking to avoid references to undefined symbols.
11380 @item -Xlinker @var{option}
11382 Pass @var{option} as an option to the linker. You can use this to
11383 supply system-specific linker options that GCC does not recognize.
11385 If you want to pass an option that takes a separate argument, you must use
11386 @option{-Xlinker} twice, once for the option and once for the argument.
11387 For example, to pass @option{-assert definitions}, you must write
11388 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11389 @option{-Xlinker "-assert definitions"}, because this passes the entire
11390 string as a single argument, which is not what the linker expects.
11392 When using the GNU linker, it is usually more convenient to pass
11393 arguments to linker options using the @option{@var{option}=@var{value}}
11394 syntax than as separate arguments. For example, you can specify
11395 @option{-Xlinker -Map=output.map} rather than
11396 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11397 this syntax for command-line options.
11399 @item -Wl,@var{option}
11401 Pass @var{option} as an option to the linker. If @var{option} contains
11402 commas, it is split into multiple options at the commas. You can use this
11403 syntax to pass an argument to the option.
11404 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11405 linker. When using the GNU linker, you can also get the same effect with
11406 @option{-Wl,-Map=output.map}.
11408 @item -u @var{symbol}
11410 Pretend the symbol @var{symbol} is undefined, to force linking of
11411 library modules to define it. You can use @option{-u} multiple times with
11412 different symbols to force loading of additional library modules.
11414 @item -z @var{keyword}
11416 @option{-z} is passed directly on to the linker along with the keyword
11417 @var{keyword}. See the section in the documentation of your linker for
11418 permitted values and their meanings.
11421 @node Directory Options
11422 @section Options for Directory Search
11423 @cindex directory options
11424 @cindex options, directory search
11425 @cindex search path
11427 These options specify directories to search for header files, for
11428 libraries and for parts of the compiler:
11433 Add the directory @var{dir} to the head of the list of directories to be
11434 searched for header files. This can be used to override a system header
11435 file, substituting your own version, since these directories are
11436 searched before the system header file directories. However, you should
11437 not use this option to add directories that contain vendor-supplied
11438 system header files (use @option{-isystem} for that). If you use more than
11439 one @option{-I} option, the directories are scanned in left-to-right
11440 order; the standard system directories come after.
11442 If a standard system include directory, or a directory specified with
11443 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11444 option is ignored. The directory is still searched but as a
11445 system directory at its normal position in the system include chain.
11446 This is to ensure that GCC's procedure to fix buggy system headers and
11447 the ordering for the @code{include_next} directive are not inadvertently changed.
11448 If you really need to change the search order for system directories,
11449 use the @option{-nostdinc} and/or @option{-isystem} options.
11451 @item -iplugindir=@var{dir}
11452 @opindex iplugindir=
11453 Set the directory to search for plugins that are passed
11454 by @option{-fplugin=@var{name}} instead of
11455 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11456 to be used by the user, but only passed by the driver.
11458 @item -iquote@var{dir}
11460 Add the directory @var{dir} to the head of the list of directories to
11461 be searched for header files only for the case of @code{#include
11462 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11463 otherwise just like @option{-I}.
11467 Add directory @var{dir} to the list of directories to be searched
11470 @item -B@var{prefix}
11472 This option specifies where to find the executables, libraries,
11473 include files, and data files of the compiler itself.
11475 The compiler driver program runs one or more of the subprograms
11476 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11477 @var{prefix} as a prefix for each program it tries to run, both with and
11478 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11480 For each subprogram to be run, the compiler driver first tries the
11481 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11482 is not specified, the driver tries two standard prefixes,
11483 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11484 those results in a file name that is found, the unmodified program
11485 name is searched for using the directories specified in your
11486 @env{PATH} environment variable.
11488 The compiler checks to see if the path provided by @option{-B}
11489 refers to a directory, and if necessary it adds a directory
11490 separator character at the end of the path.
11492 @option{-B} prefixes that effectively specify directory names also apply
11493 to libraries in the linker, because the compiler translates these
11494 options into @option{-L} options for the linker. They also apply to
11495 include files in the preprocessor, because the compiler translates these
11496 options into @option{-isystem} options for the preprocessor. In this case,
11497 the compiler appends @samp{include} to the prefix.
11499 The runtime support file @file{libgcc.a} can also be searched for using
11500 the @option{-B} prefix, if needed. If it is not found there, the two
11501 standard prefixes above are tried, and that is all. The file is left
11502 out of the link if it is not found by those means.
11504 Another way to specify a prefix much like the @option{-B} prefix is to use
11505 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11508 As a special kludge, if the path provided by @option{-B} is
11509 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11510 9, then it is replaced by @file{[dir/]include}. This is to help
11511 with boot-strapping the compiler.
11513 @item -specs=@var{file}
11515 Process @var{file} after the compiler reads in the standard @file{specs}
11516 file, in order to override the defaults which the @command{gcc} driver
11517 program uses when determining what switches to pass to @command{cc1},
11518 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11519 @option{-specs=@var{file}} can be specified on the command line, and they
11520 are processed in order, from left to right.
11522 @item --sysroot=@var{dir}
11524 Use @var{dir} as the logical root directory for headers and libraries.
11525 For example, if the compiler normally searches for headers in
11526 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11527 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11529 If you use both this option and the @option{-isysroot} option, then
11530 the @option{--sysroot} option applies to libraries, but the
11531 @option{-isysroot} option applies to header files.
11533 The GNU linker (beginning with version 2.16) has the necessary support
11534 for this option. If your linker does not support this option, the
11535 header file aspect of @option{--sysroot} still works, but the
11536 library aspect does not.
11538 @item --no-sysroot-suffix
11539 @opindex no-sysroot-suffix
11540 For some targets, a suffix is added to the root directory specified
11541 with @option{--sysroot}, depending on the other options used, so that
11542 headers may for example be found in
11543 @file{@var{dir}/@var{suffix}/usr/include} instead of
11544 @file{@var{dir}/usr/include}. This option disables the addition of
11549 This option has been deprecated. Please use @option{-iquote} instead for
11550 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11552 Any directories you specify with @option{-I} options before the @option{-I-}
11553 option are searched only for the case of @code{#include "@var{file}"};
11554 they are not searched for @code{#include <@var{file}>}.
11556 If additional directories are specified with @option{-I} options after
11557 the @option{-I-} option, these directories are searched for all @code{#include}
11558 directives. (Ordinarily @emph{all} @option{-I} directories are used
11561 In addition, the @option{-I-} option inhibits the use of the current
11562 directory (where the current input file came from) as the first search
11563 directory for @code{#include "@var{file}"}. There is no way to
11564 override this effect of @option{-I-}. With @option{-I.} you can specify
11565 searching the directory that is current when the compiler is
11566 invoked. That is not exactly the same as what the preprocessor does
11567 by default, but it is often satisfactory.
11569 @option{-I-} does not inhibit the use of the standard system directories
11570 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11577 @section Specifying Subprocesses and the Switches to Pass to Them
11580 @command{gcc} is a driver program. It performs its job by invoking a
11581 sequence of other programs to do the work of compiling, assembling and
11582 linking. GCC interprets its command-line parameters and uses these to
11583 deduce which programs it should invoke, and which command-line options
11584 it ought to place on their command lines. This behavior is controlled
11585 by @dfn{spec strings}. In most cases there is one spec string for each
11586 program that GCC can invoke, but a few programs have multiple spec
11587 strings to control their behavior. The spec strings built into GCC can
11588 be overridden by using the @option{-specs=} command-line switch to specify
11591 @dfn{Spec files} are plaintext files that are used to construct spec
11592 strings. They consist of a sequence of directives separated by blank
11593 lines. The type of directive is determined by the first non-whitespace
11594 character on the line, which can be one of the following:
11597 @item %@var{command}
11598 Issues a @var{command} to the spec file processor. The commands that can
11602 @item %include <@var{file}>
11603 @cindex @code{%include}
11604 Search for @var{file} and insert its text at the current point in the
11607 @item %include_noerr <@var{file}>
11608 @cindex @code{%include_noerr}
11609 Just like @samp{%include}, but do not generate an error message if the include
11610 file cannot be found.
11612 @item %rename @var{old_name} @var{new_name}
11613 @cindex @code{%rename}
11614 Rename the spec string @var{old_name} to @var{new_name}.
11618 @item *[@var{spec_name}]:
11619 This tells the compiler to create, override or delete the named spec
11620 string. All lines after this directive up to the next directive or
11621 blank line are considered to be the text for the spec string. If this
11622 results in an empty string then the spec is deleted. (Or, if the
11623 spec did not exist, then nothing happens.) Otherwise, if the spec
11624 does not currently exist a new spec is created. If the spec does
11625 exist then its contents are overridden by the text of this
11626 directive, unless the first character of that text is the @samp{+}
11627 character, in which case the text is appended to the spec.
11629 @item [@var{suffix}]:
11630 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11631 and up to the next directive or blank line are considered to make up the
11632 spec string for the indicated suffix. When the compiler encounters an
11633 input file with the named suffix, it processes the spec string in
11634 order to work out how to compile that file. For example:
11638 z-compile -input %i
11641 This says that any input file whose name ends in @samp{.ZZ} should be
11642 passed to the program @samp{z-compile}, which should be invoked with the
11643 command-line switch @option{-input} and with the result of performing the
11644 @samp{%i} substitution. (See below.)
11646 As an alternative to providing a spec string, the text following a
11647 suffix directive can be one of the following:
11650 @item @@@var{language}
11651 This says that the suffix is an alias for a known @var{language}. This is
11652 similar to using the @option{-x} command-line switch to GCC to specify a
11653 language explicitly. For example:
11660 Says that .ZZ files are, in fact, C++ source files.
11663 This causes an error messages saying:
11666 @var{name} compiler not installed on this system.
11670 GCC already has an extensive list of suffixes built into it.
11671 This directive adds an entry to the end of the list of suffixes, but
11672 since the list is searched from the end backwards, it is effectively
11673 possible to override earlier entries using this technique.
11677 GCC has the following spec strings built into it. Spec files can
11678 override these strings or create their own. Note that individual
11679 targets can also add their own spec strings to this list.
11682 asm Options to pass to the assembler
11683 asm_final Options to pass to the assembler post-processor
11684 cpp Options to pass to the C preprocessor
11685 cc1 Options to pass to the C compiler
11686 cc1plus Options to pass to the C++ compiler
11687 endfile Object files to include at the end of the link
11688 link Options to pass to the linker
11689 lib Libraries to include on the command line to the linker
11690 libgcc Decides which GCC support library to pass to the linker
11691 linker Sets the name of the linker
11692 predefines Defines to be passed to the C preprocessor
11693 signed_char Defines to pass to CPP to say whether @code{char} is signed
11695 startfile Object files to include at the start of the link
11698 Here is a small example of a spec file:
11701 %rename lib old_lib
11704 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11707 This example renames the spec called @samp{lib} to @samp{old_lib} and
11708 then overrides the previous definition of @samp{lib} with a new one.
11709 The new definition adds in some extra command-line options before
11710 including the text of the old definition.
11712 @dfn{Spec strings} are a list of command-line options to be passed to their
11713 corresponding program. In addition, the spec strings can contain
11714 @samp{%}-prefixed sequences to substitute variable text or to
11715 conditionally insert text into the command line. Using these constructs
11716 it is possible to generate quite complex command lines.
11718 Here is a table of all defined @samp{%}-sequences for spec
11719 strings. Note that spaces are not generated automatically around the
11720 results of expanding these sequences. Therefore you can concatenate them
11721 together or combine them with constant text in a single argument.
11725 Substitute one @samp{%} into the program name or argument.
11728 Substitute the name of the input file being processed.
11731 Substitute the basename of the input file being processed.
11732 This is the substring up to (and not including) the last period
11733 and not including the directory.
11736 This is the same as @samp{%b}, but include the file suffix (text after
11740 Marks the argument containing or following the @samp{%d} as a
11741 temporary file name, so that that file is deleted if GCC exits
11742 successfully. Unlike @samp{%g}, this contributes no text to the
11745 @item %g@var{suffix}
11746 Substitute a file name that has suffix @var{suffix} and is chosen
11747 once per compilation, and mark the argument in the same way as
11748 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11749 name is now chosen in a way that is hard to predict even when previously
11750 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11751 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11752 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11753 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11754 was simply substituted with a file name chosen once per compilation,
11755 without regard to any appended suffix (which was therefore treated
11756 just like ordinary text), making such attacks more likely to succeed.
11758 @item %u@var{suffix}
11759 Like @samp{%g}, but generates a new temporary file name
11760 each time it appears instead of once per compilation.
11762 @item %U@var{suffix}
11763 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11764 new one if there is no such last file name. In the absence of any
11765 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11766 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11767 involves the generation of two distinct file names, one
11768 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11769 simply substituted with a file name chosen for the previous @samp{%u},
11770 without regard to any appended suffix.
11772 @item %j@var{suffix}
11773 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11774 writable, and if @option{-save-temps} is not used;
11775 otherwise, substitute the name
11776 of a temporary file, just like @samp{%u}. This temporary file is not
11777 meant for communication between processes, but rather as a junk
11778 disposal mechanism.
11780 @item %|@var{suffix}
11781 @itemx %m@var{suffix}
11782 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11783 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11784 all. These are the two most common ways to instruct a program that it
11785 should read from standard input or write to standard output. If you
11786 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11787 construct: see for example @file{f/lang-specs.h}.
11789 @item %.@var{SUFFIX}
11790 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11791 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11792 terminated by the next space or %.
11795 Marks the argument containing or following the @samp{%w} as the
11796 designated output file of this compilation. This puts the argument
11797 into the sequence of arguments that @samp{%o} substitutes.
11800 Substitutes the names of all the output files, with spaces
11801 automatically placed around them. You should write spaces
11802 around the @samp{%o} as well or the results are undefined.
11803 @samp{%o} is for use in the specs for running the linker.
11804 Input files whose names have no recognized suffix are not compiled
11805 at all, but they are included among the output files, so they are
11809 Substitutes the suffix for object files. Note that this is
11810 handled specially when it immediately follows @samp{%g, %u, or %U},
11811 because of the need for those to form complete file names. The
11812 handling is such that @samp{%O} is treated exactly as if it had already
11813 been substituted, except that @samp{%g, %u, and %U} do not currently
11814 support additional @var{suffix} characters following @samp{%O} as they do
11815 following, for example, @samp{.o}.
11818 Substitutes the standard macro predefinitions for the
11819 current target machine. Use this when running @command{cpp}.
11822 Like @samp{%p}, but puts @samp{__} before and after the name of each
11823 predefined macro, except for macros that start with @samp{__} or with
11824 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11828 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11829 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11830 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11831 and @option{-imultilib} as necessary.
11834 Current argument is the name of a library or startup file of some sort.
11835 Search for that file in a standard list of directories and substitute
11836 the full name found. The current working directory is included in the
11837 list of directories scanned.
11840 Current argument is the name of a linker script. Search for that file
11841 in the current list of directories to scan for libraries. If the file
11842 is located insert a @option{--script} option into the command line
11843 followed by the full path name found. If the file is not found then
11844 generate an error message. Note: the current working directory is not
11848 Print @var{str} as an error message. @var{str} is terminated by a newline.
11849 Use this when inconsistent options are detected.
11851 @item %(@var{name})
11852 Substitute the contents of spec string @var{name} at this point.
11854 @item %x@{@var{option}@}
11855 Accumulate an option for @samp{%X}.
11858 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11862 Output the accumulated assembler options specified by @option{-Wa}.
11865 Output the accumulated preprocessor options specified by @option{-Wp}.
11868 Process the @code{asm} spec. This is used to compute the
11869 switches to be passed to the assembler.
11872 Process the @code{asm_final} spec. This is a spec string for
11873 passing switches to an assembler post-processor, if such a program is
11877 Process the @code{link} spec. This is the spec for computing the
11878 command line passed to the linker. Typically it makes use of the
11879 @samp{%L %G %S %D and %E} sequences.
11882 Dump out a @option{-L} option for each directory that GCC believes might
11883 contain startup files. If the target supports multilibs then the
11884 current multilib directory is prepended to each of these paths.
11887 Process the @code{lib} spec. This is a spec string for deciding which
11888 libraries are included on the command line to the linker.
11891 Process the @code{libgcc} spec. This is a spec string for deciding
11892 which GCC support library is included on the command line to the linker.
11895 Process the @code{startfile} spec. This is a spec for deciding which
11896 object files are the first ones passed to the linker. Typically
11897 this might be a file named @file{crt0.o}.
11900 Process the @code{endfile} spec. This is a spec string that specifies
11901 the last object files that are passed to the linker.
11904 Process the @code{cpp} spec. This is used to construct the arguments
11905 to be passed to the C preprocessor.
11908 Process the @code{cc1} spec. This is used to construct the options to be
11909 passed to the actual C compiler (@command{cc1}).
11912 Process the @code{cc1plus} spec. This is used to construct the options to be
11913 passed to the actual C++ compiler (@command{cc1plus}).
11916 Substitute the variable part of a matched option. See below.
11917 Note that each comma in the substituted string is replaced by
11921 Remove all occurrences of @code{-S} from the command line. Note---this
11922 command is position dependent. @samp{%} commands in the spec string
11923 before this one see @code{-S}, @samp{%} commands in the spec string
11924 after this one do not.
11926 @item %:@var{function}(@var{args})
11927 Call the named function @var{function}, passing it @var{args}.
11928 @var{args} is first processed as a nested spec string, then split
11929 into an argument vector in the usual fashion. The function returns
11930 a string which is processed as if it had appeared literally as part
11931 of the current spec.
11933 The following built-in spec functions are provided:
11936 @item @code{getenv}
11937 The @code{getenv} spec function takes two arguments: an environment
11938 variable name and a string. If the environment variable is not
11939 defined, a fatal error is issued. Otherwise, the return value is the
11940 value of the environment variable concatenated with the string. For
11941 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
11944 %:getenv(TOPDIR /include)
11947 expands to @file{/path/to/top/include}.
11949 @item @code{if-exists}
11950 The @code{if-exists} spec function takes one argument, an absolute
11951 pathname to a file. If the file exists, @code{if-exists} returns the
11952 pathname. Here is a small example of its usage:
11956 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
11959 @item @code{if-exists-else}
11960 The @code{if-exists-else} spec function is similar to the @code{if-exists}
11961 spec function, except that it takes two arguments. The first argument is
11962 an absolute pathname to a file. If the file exists, @code{if-exists-else}
11963 returns the pathname. If it does not exist, it returns the second argument.
11964 This way, @code{if-exists-else} can be used to select one file or another,
11965 based on the existence of the first. Here is a small example of its usage:
11969 crt0%O%s %:if-exists(crti%O%s) \
11970 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
11973 @item @code{replace-outfile}
11974 The @code{replace-outfile} spec function takes two arguments. It looks for the
11975 first argument in the outfiles array and replaces it with the second argument. Here
11976 is a small example of its usage:
11979 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
11982 @item @code{remove-outfile}
11983 The @code{remove-outfile} spec function takes one argument. It looks for the
11984 first argument in the outfiles array and removes it. Here is a small example
11988 %:remove-outfile(-lm)
11991 @item @code{pass-through-libs}
11992 The @code{pass-through-libs} spec function takes any number of arguments. It
11993 finds any @option{-l} options and any non-options ending in @file{.a} (which it
11994 assumes are the names of linker input library archive files) and returns a
11995 result containing all the found arguments each prepended by
11996 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
11997 intended to be passed to the LTO linker plugin.
12000 %:pass-through-libs(%G %L %G)
12003 @item @code{print-asm-header}
12004 The @code{print-asm-header} function takes no arguments and simply
12005 prints a banner like:
12011 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12014 It is used to separate compiler options from assembler options
12015 in the @option{--target-help} output.
12018 @item %@{@code{S}@}
12019 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12020 If that switch is not specified, this substitutes nothing. Note that
12021 the leading dash is omitted when specifying this option, and it is
12022 automatically inserted if the substitution is performed. Thus the spec
12023 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12024 and outputs the command-line option @option{-foo}.
12026 @item %W@{@code{S}@}
12027 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12028 deleted on failure.
12030 @item %@{@code{S}*@}
12031 Substitutes all the switches specified to GCC whose names start
12032 with @code{-S}, but which also take an argument. This is used for
12033 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12034 GCC considers @option{-o foo} as being
12035 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12036 text, including the space. Thus two arguments are generated.
12038 @item %@{@code{S}*&@code{T}*@}
12039 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12040 (the order of @code{S} and @code{T} in the spec is not significant).
12041 There can be any number of ampersand-separated variables; for each the
12042 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12044 @item %@{@code{S}:@code{X}@}
12045 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12047 @item %@{!@code{S}:@code{X}@}
12048 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12050 @item %@{@code{S}*:@code{X}@}
12051 Substitutes @code{X} if one or more switches whose names start with
12052 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12053 once, no matter how many such switches appeared. However, if @code{%*}
12054 appears somewhere in @code{X}, then @code{X} is substituted once
12055 for each matching switch, with the @code{%*} replaced by the part of
12056 that switch matching the @code{*}.
12058 If @code{%*} appears as the last part of a spec sequence then a space
12059 is added after the end of the last substitution. If there is more
12060 text in the sequence, however, then a space is not generated. This
12061 allows the @code{%*} substitution to be used as part of a larger
12062 string. For example, a spec string like this:
12065 %@{mcu=*:--script=%*/memory.ld@}
12069 when matching an option like @option{-mcu=newchip} produces:
12072 --script=newchip/memory.ld
12075 @item %@{.@code{S}:@code{X}@}
12076 Substitutes @code{X}, if processing a file with suffix @code{S}.
12078 @item %@{!.@code{S}:@code{X}@}
12079 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12081 @item %@{,@code{S}:@code{X}@}
12082 Substitutes @code{X}, if processing a file for language @code{S}.
12084 @item %@{!,@code{S}:@code{X}@}
12085 Substitutes @code{X}, if not processing a file for language @code{S}.
12087 @item %@{@code{S}|@code{P}:@code{X}@}
12088 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12089 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12090 @code{*} sequences as well, although they have a stronger binding than
12091 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12092 alternatives must be starred, and only the first matching alternative
12095 For example, a spec string like this:
12098 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12102 outputs the following command-line options from the following input
12103 command-line options:
12108 -d fred.c -foo -baz -boggle
12109 -d jim.d -bar -baz -boggle
12112 @item %@{S:X; T:Y; :D@}
12114 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12115 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12116 be as many clauses as you need. This may be combined with @code{.},
12117 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12122 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12123 construct may contain other nested @samp{%} constructs or spaces, or
12124 even newlines. They are processed as usual, as described above.
12125 Trailing white space in @code{X} is ignored. White space may also
12126 appear anywhere on the left side of the colon in these constructs,
12127 except between @code{.} or @code{*} and the corresponding word.
12129 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12130 handled specifically in these constructs. If another value of
12131 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12132 @option{-W} switch is found later in the command line, the earlier
12133 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12134 just one letter, which passes all matching options.
12136 The character @samp{|} at the beginning of the predicate text is used to
12137 indicate that a command should be piped to the following command, but
12138 only if @option{-pipe} is specified.
12140 It is built into GCC which switches take arguments and which do not.
12141 (You might think it would be useful to generalize this to allow each
12142 compiler's spec to say which switches take arguments. But this cannot
12143 be done in a consistent fashion. GCC cannot even decide which input
12144 files have been specified without knowing which switches take arguments,
12145 and it must know which input files to compile in order to tell which
12148 GCC also knows implicitly that arguments starting in @option{-l} are to be
12149 treated as compiler output files, and passed to the linker in their
12150 proper position among the other output files.
12152 @c man begin OPTIONS
12154 @node Target Options
12155 @section Specifying Target Machine and Compiler Version
12156 @cindex target options
12157 @cindex cross compiling
12158 @cindex specifying machine version
12159 @cindex specifying compiler version and target machine
12160 @cindex compiler version, specifying
12161 @cindex target machine, specifying
12163 The usual way to run GCC is to run the executable called @command{gcc}, or
12164 @command{@var{machine}-gcc} when cross-compiling, or
12165 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12166 one that was installed last.
12168 @node Submodel Options
12169 @section Hardware Models and Configurations
12170 @cindex submodel options
12171 @cindex specifying hardware config
12172 @cindex hardware models and configurations, specifying
12173 @cindex machine dependent options
12175 Each target machine types can have its own
12176 special options, starting with @samp{-m}, to choose among various
12177 hardware models or configurations---for example, 68010 vs 68020,
12178 floating coprocessor or none. A single installed version of the
12179 compiler can compile for any model or configuration, according to the
12182 Some configurations of the compiler also support additional special
12183 options, usually for compatibility with other compilers on the same
12186 @c This list is ordered alphanumerically by subsection name.
12187 @c It should be the same order and spelling as these options are listed
12188 @c in Machine Dependent Options
12191 * AArch64 Options::
12192 * Adapteva Epiphany Options::
12196 * Blackfin Options::
12201 * DEC Alpha Options::
12204 * GNU/Linux Options::
12214 * MicroBlaze Options::
12217 * MN10300 Options::
12221 * Nios II Options::
12222 * Nvidia PTX Options::
12224 * picoChip Options::
12225 * PowerPC Options::
12227 * RS/6000 and PowerPC Options::
12229 * S/390 and zSeries Options::
12232 * Solaris 2 Options::
12235 * System V Options::
12236 * TILE-Gx Options::
12237 * TILEPro Options::
12242 * VxWorks Options::
12244 * x86 Windows Options::
12245 * Xstormy16 Options::
12247 * zSeries Options::
12250 @node AArch64 Options
12251 @subsection AArch64 Options
12252 @cindex AArch64 Options
12254 These options are defined for AArch64 implementations:
12258 @item -mabi=@var{name}
12260 Generate code for the specified data model. Permissible values
12261 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12262 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12263 but long int and pointer are 64-bit.
12265 The default depends on the specific target configuration. Note that
12266 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12267 entire program with the same ABI, and link with a compatible set of libraries.
12270 @opindex mbig-endian
12271 Generate big-endian code. This is the default when GCC is configured for an
12272 @samp{aarch64_be-*-*} target.
12274 @item -mgeneral-regs-only
12275 @opindex mgeneral-regs-only
12276 Generate code which uses only the general registers.
12278 @item -mlittle-endian
12279 @opindex mlittle-endian
12280 Generate little-endian code. This is the default when GCC is configured for an
12281 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12283 @item -mcmodel=tiny
12284 @opindex mcmodel=tiny
12285 Generate code for the tiny code model. The program and its statically defined
12286 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12287 be statically or dynamically linked. This model is not fully implemented and
12288 mostly treated as @samp{small}.
12290 @item -mcmodel=small
12291 @opindex mcmodel=small
12292 Generate code for the small code model. The program and its statically defined
12293 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12294 be statically or dynamically linked. This is the default code model.
12296 @item -mcmodel=large
12297 @opindex mcmodel=large
12298 Generate code for the large code model. This makes no assumptions about
12299 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12300 statically linked only.
12302 @item -mstrict-align
12303 @opindex mstrict-align
12304 Do not assume that unaligned memory references are handled by the system.
12306 @item -momit-leaf-frame-pointer
12307 @itemx -mno-omit-leaf-frame-pointer
12308 @opindex momit-leaf-frame-pointer
12309 @opindex mno-omit-leaf-frame-pointer
12310 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12313 @item -mtls-dialect=desc
12314 @opindex mtls-dialect=desc
12315 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12316 of TLS variables. This is the default.
12318 @item -mtls-dialect=traditional
12319 @opindex mtls-dialect=traditional
12320 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12323 @item -mfix-cortex-a53-835769
12324 @itemx -mno-fix-cortex-a53-835769
12325 @opindex mfix-cortex-a53-835769
12326 @opindex mno-fix-cortex-a53-835769
12327 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12328 This involves inserting a NOP instruction between memory instructions and
12329 64-bit integer multiply-accumulate instructions.
12331 @item -march=@var{name}
12333 Specify the name of the target architecture, optionally suffixed by one or
12334 more feature modifiers. This option has the form
12335 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12336 only permissible value for @var{arch} is @samp{armv8-a}.
12337 The permissible values for @var{feature} are documented in the sub-section
12338 below. Additionally on native AArch64 GNU/Linux systems the value
12339 @samp{native} is available. This option causes the compiler to pick the
12340 architecture of the host system. If the compiler is unable to recognize the
12341 architecture of the host system this option has no effect.
12343 Where conflicting feature modifiers are specified, the right-most feature is
12346 GCC uses this name to determine what kind of instructions it can emit when
12347 generating assembly code.
12349 Where @option{-march} is specified without either of @option{-mtune}
12350 or @option{-mcpu} also being specified, the code is tuned to perform
12351 well across a range of target processors implementing the target
12354 @item -mtune=@var{name}
12356 Specify the name of the target processor for which GCC should tune the
12357 performance of the code. Permissible values for this option are:
12358 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
12359 @samp{exynos-m1}, @samp{thunderx}, @samp{xgene1}.
12361 Additionally, this option can specify that GCC should tune the performance
12362 of the code for a big.LITTLE system. Permissible values for this
12363 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12365 Additionally on native AArch64 GNU/Linux systems the value @samp{native}
12367 This option causes the compiler to pick the architecture of and tune the
12368 performance of the code for the processor of the host system.
12369 If the compiler is unable to recognize the processor of the host system
12370 this option has no effect.
12372 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12373 are specified, the code is tuned to perform well across a range
12374 of target processors.
12376 This option cannot be suffixed by feature modifiers.
12378 @item -mcpu=@var{name}
12380 Specify the name of the target processor, optionally suffixed by one or more
12381 feature modifiers. This option has the form
12382 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where the
12383 permissible values for @var{cpu} are the same as those available for
12384 @option{-mtune}. Additionally on native AArch64 GNU/Linux systems the
12385 value @samp{native} is available.
12386 This option causes the compiler to tune the performance of the code for the
12387 processor of the host system. If the compiler is unable to recognize the
12388 processor of the host system this option has no effect.
12390 The permissible values for @var{feature} are documented in the sub-section
12393 Where conflicting feature modifiers are specified, the right-most feature is
12396 GCC uses this name to determine what kind of instructions it can emit when
12397 generating assembly code (as if by @option{-march}) and to determine
12398 the target processor for which to tune for performance (as if
12399 by @option{-mtune}). Where this option is used in conjunction
12400 with @option{-march} or @option{-mtune}, those options take precedence
12401 over the appropriate part of this option.
12404 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12405 @cindex @option{-march} feature modifiers
12406 @cindex @option{-mcpu} feature modifiers
12407 Feature modifiers used with @option{-march} and @option{-mcpu} can be one
12412 Enable CRC extension.
12414 Enable Crypto extension. This implies Advanced SIMD is enabled.
12416 Enable floating-point instructions.
12418 Enable Advanced SIMD instructions. This implies floating-point instructions
12419 are enabled. This is the default for all current possible values for options
12420 @option{-march} and @option{-mcpu=}.
12423 @node Adapteva Epiphany Options
12424 @subsection Adapteva Epiphany Options
12426 These @samp{-m} options are defined for Adapteva Epiphany:
12429 @item -mhalf-reg-file
12430 @opindex mhalf-reg-file
12431 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12432 That allows code to run on hardware variants that lack these registers.
12434 @item -mprefer-short-insn-regs
12435 @opindex mprefer-short-insn-regs
12436 Preferrentially allocate registers that allow short instruction generation.
12437 This can result in increased instruction count, so this may either reduce or
12438 increase overall code size.
12440 @item -mbranch-cost=@var{num}
12441 @opindex mbranch-cost
12442 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12443 This cost is only a heuristic and is not guaranteed to produce
12444 consistent results across releases.
12448 Enable the generation of conditional moves.
12450 @item -mnops=@var{num}
12452 Emit @var{num} NOPs before every other generated instruction.
12454 @item -mno-soft-cmpsf
12455 @opindex mno-soft-cmpsf
12456 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12457 and test the flags. This is faster than a software comparison, but can
12458 get incorrect results in the presence of NaNs, or when two different small
12459 numbers are compared such that their difference is calculated as zero.
12460 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12461 software comparisons.
12463 @item -mstack-offset=@var{num}
12464 @opindex mstack-offset
12465 Set the offset between the top of the stack and the stack pointer.
12466 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12467 can be used by leaf functions without stack allocation.
12468 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12469 Note also that this option changes the ABI; compiling a program with a
12470 different stack offset than the libraries have been compiled with
12471 generally does not work.
12472 This option can be useful if you want to evaluate if a different stack
12473 offset would give you better code, but to actually use a different stack
12474 offset to build working programs, it is recommended to configure the
12475 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12477 @item -mno-round-nearest
12478 @opindex mno-round-nearest
12479 Make the scheduler assume that the rounding mode has been set to
12480 truncating. The default is @option{-mround-nearest}.
12483 @opindex mlong-calls
12484 If not otherwise specified by an attribute, assume all calls might be beyond
12485 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12486 function address into a register before performing a (otherwise direct) call.
12487 This is the default.
12489 @item -mshort-calls
12490 @opindex short-calls
12491 If not otherwise specified by an attribute, assume all direct calls are
12492 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12493 for direct calls. The default is @option{-mlong-calls}.
12497 Assume addresses can be loaded as 16-bit unsigned values. This does not
12498 apply to function addresses for which @option{-mlong-calls} semantics
12501 @item -mfp-mode=@var{mode}
12503 Set the prevailing mode of the floating-point unit.
12504 This determines the floating-point mode that is provided and expected
12505 at function call and return time. Making this mode match the mode you
12506 predominantly need at function start can make your programs smaller and
12507 faster by avoiding unnecessary mode switches.
12509 @var{mode} can be set to one the following values:
12513 Any mode at function entry is valid, and retained or restored when
12514 the function returns, and when it calls other functions.
12515 This mode is useful for compiling libraries or other compilation units
12516 you might want to incorporate into different programs with different
12517 prevailing FPU modes, and the convenience of being able to use a single
12518 object file outweighs the size and speed overhead for any extra
12519 mode switching that might be needed, compared with what would be needed
12520 with a more specific choice of prevailing FPU mode.
12523 This is the mode used for floating-point calculations with
12524 truncating (i.e.@: round towards zero) rounding mode. That includes
12525 conversion from floating point to integer.
12527 @item round-nearest
12528 This is the mode used for floating-point calculations with
12529 round-to-nearest-or-even rounding mode.
12532 This is the mode used to perform integer calculations in the FPU, e.g.@:
12533 integer multiply, or integer multiply-and-accumulate.
12536 The default is @option{-mfp-mode=caller}
12538 @item -mnosplit-lohi
12539 @itemx -mno-postinc
12540 @itemx -mno-postmodify
12541 @opindex mnosplit-lohi
12542 @opindex mno-postinc
12543 @opindex mno-postmodify
12544 Code generation tweaks that disable, respectively, splitting of 32-bit
12545 loads, generation of post-increment addresses, and generation of
12546 post-modify addresses. The defaults are @option{msplit-lohi},
12547 @option{-mpost-inc}, and @option{-mpost-modify}.
12549 @item -mnovect-double
12550 @opindex mno-vect-double
12551 Change the preferred SIMD mode to SImode. The default is
12552 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12554 @item -max-vect-align=@var{num}
12555 @opindex max-vect-align
12556 The maximum alignment for SIMD vector mode types.
12557 @var{num} may be 4 or 8. The default is 8.
12558 Note that this is an ABI change, even though many library function
12559 interfaces are unaffected if they don't use SIMD vector modes
12560 in places that affect size and/or alignment of relevant types.
12562 @item -msplit-vecmove-early
12563 @opindex msplit-vecmove-early
12564 Split vector moves into single word moves before reload. In theory this
12565 can give better register allocation, but so far the reverse seems to be
12566 generally the case.
12568 @item -m1reg-@var{reg}
12570 Specify a register to hold the constant @minus{}1, which makes loading small negative
12571 constants and certain bitmasks faster.
12572 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12573 which specify use of that register as a fixed register,
12574 and @samp{none}, which means that no register is used for this
12575 purpose. The default is @option{-m1reg-none}.
12580 @subsection ARC Options
12581 @cindex ARC options
12583 The following options control the architecture variant for which code
12586 @c architecture variants
12589 @item -mbarrel-shifter
12590 @opindex mbarrel-shifter
12591 Generate instructions supported by barrel shifter. This is the default
12592 unless @option{-mcpu=ARC601} is in effect.
12594 @item -mcpu=@var{cpu}
12596 Set architecture type, register usage, and instruction scheduling
12597 parameters for @var{cpu}. There are also shortcut alias options
12598 available for backward compatibility and convenience. Supported
12599 values for @var{cpu} are
12605 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12609 Compile for ARC601. Alias: @option{-mARC601}.
12614 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12615 This is the default when configured with @option{--with-cpu=arc700}@.
12620 @itemx -mdpfp-compact
12621 @opindex mdpfp-compact
12622 FPX: Generate Double Precision FPX instructions, tuned for the compact
12626 @opindex mdpfp-fast
12627 FPX: Generate Double Precision FPX instructions, tuned for the fast
12630 @item -mno-dpfp-lrsr
12631 @opindex mno-dpfp-lrsr
12632 Disable LR and SR instructions from using FPX extension aux registers.
12636 Generate Extended arithmetic instructions. Currently only
12637 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12638 supported. This is always enabled for @option{-mcpu=ARC700}.
12642 Do not generate mpy instructions for ARC700.
12646 Generate 32x16 bit multiply and mac instructions.
12650 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12654 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12659 @itemx -mspfp-compact
12660 @opindex mspfp-compact
12661 FPX: Generate Single Precision FPX instructions, tuned for the compact
12665 @opindex mspfp-fast
12666 FPX: Generate Single Precision FPX instructions, tuned for the fast
12671 Enable generation of ARC SIMD instructions via target-specific
12672 builtins. Only valid for @option{-mcpu=ARC700}.
12675 @opindex msoft-float
12676 This option ignored; it is provided for compatibility purposes only.
12677 Software floating point code is emitted by default, and this default
12678 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12679 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12680 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12684 Generate swap instructions.
12688 The following options are passed through to the assembler, and also
12689 define preprocessor macro symbols.
12691 @c Flags used by the assembler, but for which we define preprocessor
12692 @c macro symbols as well.
12695 @opindex mdsp-packa
12696 Passed down to the assembler to enable the DSP Pack A extensions.
12697 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12701 Passed down to the assembler to enable the dual viterbi butterfly
12702 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12704 @c ARC700 4.10 extension instruction
12707 Passed down to the assembler to enable the Locked Load/Store
12708 Conditional extension. Also sets the preprocessor symbol
12713 Passed down to the assembler. Also sets the preprocessor symbol
12714 @code{__Xxmac_d16}.
12718 Passed down to the assembler. Also sets the preprocessor symbol
12721 @c ARC700 4.10 extension instruction
12724 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12725 extension instruction. Also sets the preprocessor symbol
12728 @c ARC700 4.10 extension instruction
12731 Passed down to the assembler to enable the swap byte ordering
12732 extension instruction. Also sets the preprocessor symbol
12736 @opindex mtelephony
12737 Passed down to the assembler to enable dual and single operand
12738 instructions for telephony. Also sets the preprocessor symbol
12739 @code{__Xtelephony}.
12743 Passed down to the assembler to enable the XY Memory extension. Also
12744 sets the preprocessor symbol @code{__Xxy}.
12748 The following options control how the assembly code is annotated:
12750 @c Assembly annotation options
12754 Annotate assembler instructions with estimated addresses.
12756 @item -mannotate-align
12757 @opindex mannotate-align
12758 Explain what alignment considerations lead to the decision to make an
12759 instruction short or long.
12763 The following options are passed through to the linker:
12765 @c options passed through to the linker
12769 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12770 This option is enabled by default in tool chains built for
12771 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12772 when profiling is not requested.
12774 @item -marclinux_prof
12775 @opindex marclinux_prof
12776 Passed through to the linker, to specify use of the
12777 @code{arclinux_prof} emulation. This option is enabled by default in
12778 tool chains built for @w{@code{arc-linux-uclibc}} and
12779 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12783 The following options control the semantics of generated code:
12785 @c semantically relevant code generation options
12787 @item -mepilogue-cfi
12788 @opindex mepilogue-cfi
12789 Enable generation of call frame information for epilogues.
12791 @item -mno-epilogue-cfi
12792 @opindex mno-epilogue-cfi
12793 Disable generation of call frame information for epilogues.
12796 @opindex mlong-calls
12797 Generate call insns as register indirect calls, thus providing access
12798 to the full 32-bit address range.
12800 @item -mmedium-calls
12801 @opindex mmedium-calls
12802 Don't use less than 25 bit addressing range for calls, which is the
12803 offset available for an unconditional branch-and-link
12804 instruction. Conditional execution of function calls is suppressed, to
12805 allow use of the 25-bit range, rather than the 21-bit range with
12806 conditional branch-and-link. This is the default for tool chains built
12807 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12811 Do not generate sdata references. This is the default for tool chains
12812 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12816 @opindex mucb-mcount
12817 Instrument with mcount calls as used in UCB code. I.e. do the
12818 counting in the callee, not the caller. By default ARC instrumentation
12819 counts in the caller.
12821 @item -mvolatile-cache
12822 @opindex mvolatile-cache
12823 Use ordinarily cached memory accesses for volatile references. This is the
12826 @item -mno-volatile-cache
12827 @opindex mno-volatile-cache
12828 Enable cache bypass for volatile references.
12832 The following options fine tune code generation:
12833 @c code generation tuning options
12836 @opindex malign-call
12837 Do alignment optimizations for call instructions.
12839 @item -mauto-modify-reg
12840 @opindex mauto-modify-reg
12841 Enable the use of pre/post modify with register displacement.
12843 @item -mbbit-peephole
12844 @opindex mbbit-peephole
12845 Enable bbit peephole2.
12849 This option disables a target-specific pass in @file{arc_reorg} to
12850 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12851 generation driven by the combiner pass.
12853 @item -mcase-vector-pcrel
12854 @opindex mcase-vector-pcrel
12855 Use pc-relative switch case tables - this enables case table shortening.
12856 This is the default for @option{-Os}.
12858 @item -mcompact-casesi
12859 @opindex mcompact-casesi
12860 Enable compact casesi pattern.
12861 This is the default for @option{-Os}.
12863 @item -mno-cond-exec
12864 @opindex mno-cond-exec
12865 Disable ARCompact specific pass to generate conditional execution instructions.
12866 Due to delay slot scheduling and interactions between operand numbers,
12867 literal sizes, instruction lengths, and the support for conditional execution,
12868 the target-independent pass to generate conditional execution is often lacking,
12869 so the ARC port has kept a special pass around that tries to find more
12870 conditional execution generating opportunities after register allocation,
12871 branch shortening, and delay slot scheduling have been done. This pass
12872 generally, but not always, improves performance and code size, at the cost of
12873 extra compilation time, which is why there is an option to switch it off.
12874 If you have a problem with call instructions exceeding their allowable
12875 offset range because they are conditionalized, you should consider using
12876 @option{-mmedium-calls} instead.
12878 @item -mearly-cbranchsi
12879 @opindex mearly-cbranchsi
12880 Enable pre-reload use of the cbranchsi pattern.
12882 @item -mexpand-adddi
12883 @opindex mexpand-adddi
12884 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
12885 @code{add.f}, @code{adc} etc.
12887 @item -mindexed-loads
12888 @opindex mindexed-loads
12889 Enable the use of indexed loads. This can be problematic because some
12890 optimizers then assume that indexed stores exist, which is not
12895 Enable Local Register Allocation. This is still experimental for ARC,
12896 so by default the compiler uses standard reload
12897 (i.e. @option{-mno-lra}).
12899 @item -mlra-priority-none
12900 @opindex mlra-priority-none
12901 Don't indicate any priority for target registers.
12903 @item -mlra-priority-compact
12904 @opindex mlra-priority-compact
12905 Indicate target register priority for r0..r3 / r12..r15.
12907 @item -mlra-priority-noncompact
12908 @opindex mlra-priority-noncompact
12909 Reduce target regsiter priority for r0..r3 / r12..r15.
12911 @item -mno-millicode
12912 @opindex mno-millicode
12913 When optimizing for size (using @option{-Os}), prologues and epilogues
12914 that have to save or restore a large number of registers are often
12915 shortened by using call to a special function in libgcc; this is
12916 referred to as a @emph{millicode} call. As these calls can pose
12917 performance issues, and/or cause linking issues when linking in a
12918 nonstandard way, this option is provided to turn off millicode call
12922 @opindex mmixed-code
12923 Tweak register allocation to help 16-bit instruction generation.
12924 This generally has the effect of decreasing the average instruction size
12925 while increasing the instruction count.
12929 Enable 'q' instruction alternatives.
12930 This is the default for @option{-Os}.
12934 Enable Rcq constraint handling - most short code generation depends on this.
12935 This is the default.
12939 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
12940 This is the default.
12942 @item -msize-level=@var{level}
12943 @opindex msize-level
12944 Fine-tune size optimization with regards to instruction lengths and alignment.
12945 The recognized values for @var{level} are:
12948 No size optimization. This level is deprecated and treated like @samp{1}.
12951 Short instructions are used opportunistically.
12954 In addition, alignment of loops and of code after barriers are dropped.
12957 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
12961 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
12962 the behavior when this is not set is equivalent to level @samp{1}.
12964 @item -mtune=@var{cpu}
12966 Set instruction scheduling parameters for @var{cpu}, overriding any implied
12967 by @option{-mcpu=}.
12969 Supported values for @var{cpu} are
12973 Tune for ARC600 cpu.
12976 Tune for ARC601 cpu.
12979 Tune for ARC700 cpu with standard multiplier block.
12982 Tune for ARC700 cpu with XMAC block.
12985 Tune for ARC725D cpu.
12988 Tune for ARC750D cpu.
12992 @item -mmultcost=@var{num}
12994 Cost to assume for a multiply instruction, with @samp{4} being equal to a
12995 normal instruction.
12997 @item -munalign-prob-threshold=@var{probability}
12998 @opindex munalign-prob-threshold
12999 Set probability threshold for unaligning branches.
13000 When tuning for @samp{ARC700} and optimizing for speed, branches without
13001 filled delay slot are preferably emitted unaligned and long, unless
13002 profiling indicates that the probability for the branch to be taken
13003 is below @var{probability}. @xref{Cross-profiling}.
13004 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13008 The following options are maintained for backward compatibility, but
13009 are now deprecated and will be removed in a future release:
13011 @c Deprecated options
13019 @opindex mbig-endian
13022 Compile code for big endian targets. Use of these options is now
13023 deprecated. Users wanting big-endian code, should use the
13024 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13025 building the tool chain, for which big-endian is the default.
13027 @item -mlittle-endian
13028 @opindex mlittle-endian
13031 Compile code for little endian targets. Use of these options is now
13032 deprecated. Users wanting little-endian code should use the
13033 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13034 building the tool chain, for which little-endian is the default.
13036 @item -mbarrel_shifter
13037 @opindex mbarrel_shifter
13038 Replaced by @option{-mbarrel-shifter}.
13040 @item -mdpfp_compact
13041 @opindex mdpfp_compact
13042 Replaced by @option{-mdpfp-compact}.
13045 @opindex mdpfp_fast
13046 Replaced by @option{-mdpfp-fast}.
13049 @opindex mdsp_packa
13050 Replaced by @option{-mdsp-packa}.
13054 Replaced by @option{-mea}.
13058 Replaced by @option{-mmac-24}.
13062 Replaced by @option{-mmac-d16}.
13064 @item -mspfp_compact
13065 @opindex mspfp_compact
13066 Replaced by @option{-mspfp-compact}.
13069 @opindex mspfp_fast
13070 Replaced by @option{-mspfp-fast}.
13072 @item -mtune=@var{cpu}
13074 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13075 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13076 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13078 @item -multcost=@var{num}
13080 Replaced by @option{-mmultcost}.
13085 @subsection ARM Options
13086 @cindex ARM options
13088 These @samp{-m} options are defined for the ARM port:
13091 @item -mabi=@var{name}
13093 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13094 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13097 @opindex mapcs-frame
13098 Generate a stack frame that is compliant with the ARM Procedure Call
13099 Standard for all functions, even if this is not strictly necessary for
13100 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13101 with this option causes the stack frames not to be generated for
13102 leaf functions. The default is @option{-mno-apcs-frame}.
13103 This option is deprecated.
13107 This is a synonym for @option{-mapcs-frame} and is deprecated.
13110 @c not currently implemented
13111 @item -mapcs-stack-check
13112 @opindex mapcs-stack-check
13113 Generate code to check the amount of stack space available upon entry to
13114 every function (that actually uses some stack space). If there is
13115 insufficient space available then either the function
13116 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13117 called, depending upon the amount of stack space required. The runtime
13118 system is required to provide these functions. The default is
13119 @option{-mno-apcs-stack-check}, since this produces smaller code.
13121 @c not currently implemented
13123 @opindex mapcs-float
13124 Pass floating-point arguments using the floating-point registers. This is
13125 one of the variants of the APCS@. This option is recommended if the
13126 target hardware has a floating-point unit or if a lot of floating-point
13127 arithmetic is going to be performed by the code. The default is
13128 @option{-mno-apcs-float}, since the size of integer-only code is
13129 slightly increased if @option{-mapcs-float} is used.
13131 @c not currently implemented
13132 @item -mapcs-reentrant
13133 @opindex mapcs-reentrant
13134 Generate reentrant, position-independent code. The default is
13135 @option{-mno-apcs-reentrant}.
13138 @item -mthumb-interwork
13139 @opindex mthumb-interwork
13140 Generate code that supports calling between the ARM and Thumb
13141 instruction sets. Without this option, on pre-v5 architectures, the
13142 two instruction sets cannot be reliably used inside one program. The
13143 default is @option{-mno-thumb-interwork}, since slightly larger code
13144 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13145 configurations this option is meaningless.
13147 @item -mno-sched-prolog
13148 @opindex mno-sched-prolog
13149 Prevent the reordering of instructions in the function prologue, or the
13150 merging of those instruction with the instructions in the function's
13151 body. This means that all functions start with a recognizable set
13152 of instructions (or in fact one of a choice from a small set of
13153 different function prologues), and this information can be used to
13154 locate the start of functions inside an executable piece of code. The
13155 default is @option{-msched-prolog}.
13157 @item -mfloat-abi=@var{name}
13158 @opindex mfloat-abi
13159 Specifies which floating-point ABI to use. Permissible values
13160 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13162 Specifying @samp{soft} causes GCC to generate output containing
13163 library calls for floating-point operations.
13164 @samp{softfp} allows the generation of code using hardware floating-point
13165 instructions, but still uses the soft-float calling conventions.
13166 @samp{hard} allows generation of floating-point instructions
13167 and uses FPU-specific calling conventions.
13169 The default depends on the specific target configuration. Note that
13170 the hard-float and soft-float ABIs are not link-compatible; you must
13171 compile your entire program with the same ABI, and link with a
13172 compatible set of libraries.
13174 @item -mlittle-endian
13175 @opindex mlittle-endian
13176 Generate code for a processor running in little-endian mode. This is
13177 the default for all standard configurations.
13180 @opindex mbig-endian
13181 Generate code for a processor running in big-endian mode; the default is
13182 to compile code for a little-endian processor.
13184 @item -march=@var{name}
13186 This specifies the name of the target ARM architecture. GCC uses this
13187 name to determine what kind of instructions it can emit when generating
13188 assembly code. This option can be used in conjunction with or instead
13189 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13190 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13191 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13192 @samp{armv6}, @samp{armv6j},
13193 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv6-m},
13194 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13195 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
13196 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13198 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13201 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13202 architecture together with the optional CRC32 extensions.
13204 @option{-march=native} causes the compiler to auto-detect the architecture
13205 of the build computer. At present, this feature is only supported on
13206 GNU/Linux, and not all architectures are recognized. If the auto-detect
13207 is unsuccessful the option has no effect.
13209 @item -mtune=@var{name}
13211 This option specifies the name of the target ARM processor for
13212 which GCC should tune the performance of the code.
13213 For some ARM implementations better performance can be obtained by using
13215 Permissible names are: @samp{arm2}, @samp{arm250},
13216 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13217 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13218 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13219 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13221 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13222 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13223 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13224 @samp{strongarm1110},
13225 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13226 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13227 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13228 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13229 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13230 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13231 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13232 @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8}, @samp{cortex-a9},
13233 @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a53},
13234 @samp{cortex-a57}, @samp{cortex-a72},
13236 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13241 @samp{cortex-m0plus},
13242 @samp{cortex-m1.small-multiply},
13243 @samp{cortex-m0.small-multiply},
13244 @samp{cortex-m0plus.small-multiply},
13246 @samp{marvell-pj4},
13247 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13248 @samp{fa526}, @samp{fa626},
13249 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13252 Additionally, this option can specify that GCC should tune the performance
13253 of the code for a big.LITTLE system. Permissible names are:
13254 @samp{cortex-a15.cortex-a7}, @samp{cortex-a57.cortex-a53},
13255 @samp{cortex-a72.cortex-a53}.
13257 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13258 performance for a blend of processors within architecture @var{arch}.
13259 The aim is to generate code that run well on the current most popular
13260 processors, balancing between optimizations that benefit some CPUs in the
13261 range, and avoiding performance pitfalls of other CPUs. The effects of
13262 this option may change in future GCC versions as CPU models come and go.
13264 @option{-mtune=native} causes the compiler to auto-detect the CPU
13265 of the build computer. At present, this feature is only supported on
13266 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13267 unsuccessful the option has no effect.
13269 @item -mcpu=@var{name}
13271 This specifies the name of the target ARM processor. GCC uses this name
13272 to derive the name of the target ARM architecture (as if specified
13273 by @option{-march}) and the ARM processor type for which to tune for
13274 performance (as if specified by @option{-mtune}). Where this option
13275 is used in conjunction with @option{-march} or @option{-mtune},
13276 those options take precedence over the appropriate part of this option.
13278 Permissible names for this option are the same as those for
13281 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13282 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13283 See @option{-mtune} for more information.
13285 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13286 of the build computer. At present, this feature is only supported on
13287 GNU/Linux, and not all architectures are recognized. If the auto-detect
13288 is unsuccessful the option has no effect.
13290 @item -mfpu=@var{name}
13292 This specifies what floating-point hardware (or hardware emulation) is
13293 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13294 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13295 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13296 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13297 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13298 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13300 If @option{-msoft-float} is specified this specifies the format of
13301 floating-point values.
13303 If the selected floating-point hardware includes the NEON extension
13304 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13305 operations are not generated by GCC's auto-vectorization pass unless
13306 @option{-funsafe-math-optimizations} is also specified. This is
13307 because NEON hardware does not fully implement the IEEE 754 standard for
13308 floating-point arithmetic (in particular denormal values are treated as
13309 zero), so the use of NEON instructions may lead to a loss of precision.
13311 @item -mfp16-format=@var{name}
13312 @opindex mfp16-format
13313 Specify the format of the @code{__fp16} half-precision floating-point type.
13314 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13315 the default is @samp{none}, in which case the @code{__fp16} type is not
13316 defined. @xref{Half-Precision}, for more information.
13318 @item -mstructure-size-boundary=@var{n}
13319 @opindex mstructure-size-boundary
13320 The sizes of all structures and unions are rounded up to a multiple
13321 of the number of bits set by this option. Permissible values are 8, 32
13322 and 64. The default value varies for different toolchains. For the COFF
13323 targeted toolchain the default value is 8. A value of 64 is only allowed
13324 if the underlying ABI supports it.
13326 Specifying a larger number can produce faster, more efficient code, but
13327 can also increase the size of the program. Different values are potentially
13328 incompatible. Code compiled with one value cannot necessarily expect to
13329 work with code or libraries compiled with another value, if they exchange
13330 information using structures or unions.
13332 @item -mabort-on-noreturn
13333 @opindex mabort-on-noreturn
13334 Generate a call to the function @code{abort} at the end of a
13335 @code{noreturn} function. It is executed if the function tries to
13339 @itemx -mno-long-calls
13340 @opindex mlong-calls
13341 @opindex mno-long-calls
13342 Tells the compiler to perform function calls by first loading the
13343 address of the function into a register and then performing a subroutine
13344 call on this register. This switch is needed if the target function
13345 lies outside of the 64-megabyte addressing range of the offset-based
13346 version of subroutine call instruction.
13348 Even if this switch is enabled, not all function calls are turned
13349 into long calls. The heuristic is that static functions, functions
13350 that have the @code{short_call} attribute, functions that are inside
13351 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13352 definitions have already been compiled within the current compilation
13353 unit are not turned into long calls. The exceptions to this rule are
13354 that weak function definitions, functions with the @code{long_call}
13355 attribute or the @code{section} attribute, and functions that are within
13356 the scope of a @code{#pragma long_calls} directive are always
13357 turned into long calls.
13359 This feature is not enabled by default. Specifying
13360 @option{-mno-long-calls} restores the default behavior, as does
13361 placing the function calls within the scope of a @code{#pragma
13362 long_calls_off} directive. Note these switches have no effect on how
13363 the compiler generates code to handle function calls via function
13366 @item -msingle-pic-base
13367 @opindex msingle-pic-base
13368 Treat the register used for PIC addressing as read-only, rather than
13369 loading it in the prologue for each function. The runtime system is
13370 responsible for initializing this register with an appropriate value
13371 before execution begins.
13373 @item -mpic-register=@var{reg}
13374 @opindex mpic-register
13375 Specify the register to be used for PIC addressing.
13376 For standard PIC base case, the default is any suitable register
13377 determined by compiler. For single PIC base case, the default is
13378 @samp{R9} if target is EABI based or stack-checking is enabled,
13379 otherwise the default is @samp{R10}.
13381 @item -mpic-data-is-text-relative
13382 @opindex mpic-data-is-text-relative
13383 Assume that each data segments are relative to text segment at load time.
13384 Therefore, it permits addressing data using PC-relative operations.
13385 This option is on by default for targets other than VxWorks RTP.
13387 @item -mpoke-function-name
13388 @opindex mpoke-function-name
13389 Write the name of each function into the text section, directly
13390 preceding the function prologue. The generated code is similar to this:
13394 .ascii "arm_poke_function_name", 0
13397 .word 0xff000000 + (t1 - t0)
13398 arm_poke_function_name
13400 stmfd sp!, @{fp, ip, lr, pc@}
13404 When performing a stack backtrace, code can inspect the value of
13405 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13406 location @code{pc - 12} and the top 8 bits are set, then we know that
13407 there is a function name embedded immediately preceding this location
13408 and has length @code{((pc[-3]) & 0xff000000)}.
13415 Select between generating code that executes in ARM and Thumb
13416 states. The default for most configurations is to generate code
13417 that executes in ARM state, but the default can be changed by
13418 configuring GCC with the @option{--with-mode=}@var{state}
13422 @opindex mtpcs-frame
13423 Generate a stack frame that is compliant with the Thumb Procedure Call
13424 Standard for all non-leaf functions. (A leaf function is one that does
13425 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13427 @item -mtpcs-leaf-frame
13428 @opindex mtpcs-leaf-frame
13429 Generate a stack frame that is compliant with the Thumb Procedure Call
13430 Standard for all leaf functions. (A leaf function is one that does
13431 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13433 @item -mcallee-super-interworking
13434 @opindex mcallee-super-interworking
13435 Gives all externally visible functions in the file being compiled an ARM
13436 instruction set header which switches to Thumb mode before executing the
13437 rest of the function. This allows these functions to be called from
13438 non-interworking code. This option is not valid in AAPCS configurations
13439 because interworking is enabled by default.
13441 @item -mcaller-super-interworking
13442 @opindex mcaller-super-interworking
13443 Allows calls via function pointers (including virtual functions) to
13444 execute correctly regardless of whether the target code has been
13445 compiled for interworking or not. There is a small overhead in the cost
13446 of executing a function pointer if this option is enabled. This option
13447 is not valid in AAPCS configurations because interworking is enabled
13450 @item -mtp=@var{name}
13452 Specify the access model for the thread local storage pointer. The valid
13453 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13454 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13455 (supported in the arm6k architecture), and @samp{auto}, which uses the
13456 best available method for the selected processor. The default setting is
13459 @item -mtls-dialect=@var{dialect}
13460 @opindex mtls-dialect
13461 Specify the dialect to use for accessing thread local storage. Two
13462 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13463 @samp{gnu} dialect selects the original GNU scheme for supporting
13464 local and global dynamic TLS models. The @samp{gnu2} dialect
13465 selects the GNU descriptor scheme, which provides better performance
13466 for shared libraries. The GNU descriptor scheme is compatible with
13467 the original scheme, but does require new assembler, linker and
13468 library support. Initial and local exec TLS models are unaffected by
13469 this option and always use the original scheme.
13471 @item -mword-relocations
13472 @opindex mword-relocations
13473 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13474 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13475 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13478 @item -mfix-cortex-m3-ldrd
13479 @opindex mfix-cortex-m3-ldrd
13480 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13481 with overlapping destination and base registers are used. This option avoids
13482 generating these instructions. This option is enabled by default when
13483 @option{-mcpu=cortex-m3} is specified.
13485 @item -munaligned-access
13486 @itemx -mno-unaligned-access
13487 @opindex munaligned-access
13488 @opindex mno-unaligned-access
13489 Enables (or disables) reading and writing of 16- and 32- bit values
13490 from addresses that are not 16- or 32- bit aligned. By default
13491 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13492 architectures, and enabled for all other architectures. If unaligned
13493 access is not enabled then words in packed data structures are
13494 accessed a byte at a time.
13496 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13497 generated object file to either true or false, depending upon the
13498 setting of this option. If unaligned access is enabled then the
13499 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13502 @item -mneon-for-64bits
13503 @opindex mneon-for-64bits
13504 Enables using Neon to handle scalar 64-bits operations. This is
13505 disabled by default since the cost of moving data from core registers
13508 @item -mslow-flash-data
13509 @opindex mslow-flash-data
13510 Assume loading data from flash is slower than fetching instruction.
13511 Therefore literal load is minimized for better performance.
13512 This option is only supported when compiling for ARMv7 M-profile and
13515 @item -masm-syntax-unified
13516 @opindex masm-syntax-unified
13517 Assume inline assembler is using unified asm syntax. The default is
13518 currently off which implies divided syntax. Currently this option is
13519 available only for Thumb1 and has no effect on ARM state and Thumb2.
13520 However, this may change in future releases of GCC. Divided syntax
13521 should be considered deprecated.
13523 @item -mrestrict-it
13524 @opindex mrestrict-it
13525 Restricts generation of IT blocks to conform to the rules of ARMv8.
13526 IT blocks can only contain a single 16-bit instruction from a select
13527 set of instructions. This option is on by default for ARMv8 Thumb mode.
13529 @item -mprint-tune-info
13530 @opindex mprint-tune-info
13531 Print CPU tuning information as comment in assembler file. This is
13532 an option used only for regression testing of the compiler and not
13533 intended for ordinary use in compiling code. This option is disabled
13538 @subsection AVR Options
13539 @cindex AVR Options
13541 These options are defined for AVR implementations:
13544 @item -mmcu=@var{mcu}
13546 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13548 The default for this option is@tie{}@samp{avr2}.
13550 GCC supports the following AVR devices and ISAs:
13552 @include avr-mmcu.texi
13554 @item -maccumulate-args
13555 @opindex maccumulate-args
13556 Accumulate outgoing function arguments and acquire/release the needed
13557 stack space for outgoing function arguments once in function
13558 prologue/epilogue. Without this option, outgoing arguments are pushed
13559 before calling a function and popped afterwards.
13561 Popping the arguments after the function call can be expensive on
13562 AVR so that accumulating the stack space might lead to smaller
13563 executables because arguments need not to be removed from the
13564 stack after such a function call.
13566 This option can lead to reduced code size for functions that perform
13567 several calls to functions that get their arguments on the stack like
13568 calls to printf-like functions.
13570 @item -mbranch-cost=@var{cost}
13571 @opindex mbranch-cost
13572 Set the branch costs for conditional branch instructions to
13573 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13574 integers. The default branch cost is 0.
13576 @item -mcall-prologues
13577 @opindex mcall-prologues
13578 Functions prologues/epilogues are expanded as calls to appropriate
13579 subroutines. Code size is smaller.
13583 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13584 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13585 and @code{long long} is 4 bytes. Please note that this option does not
13586 conform to the C standards, but it results in smaller code
13589 @item -mn-flash=@var{num}
13591 Assume that the flash memory has a size of
13592 @var{num} times 64@tie{}KiB.
13594 @item -mno-interrupts
13595 @opindex mno-interrupts
13596 Generated code is not compatible with hardware interrupts.
13597 Code size is smaller.
13601 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13602 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13603 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13604 the assembler's command line and the @option{--relax} option to the
13605 linker's command line.
13607 Jump relaxing is performed by the linker because jump offsets are not
13608 known before code is located. Therefore, the assembler code generated by the
13609 compiler is the same, but the instructions in the executable may
13610 differ from instructions in the assembler code.
13612 Relaxing must be turned on if linker stubs are needed, see the
13613 section on @code{EIND} and linker stubs below.
13617 Assume that the device supports the Read-Modify-Write
13618 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13622 Treat the stack pointer register as an 8-bit register,
13623 i.e.@: assume the high byte of the stack pointer is zero.
13624 In general, you don't need to set this option by hand.
13626 This option is used internally by the compiler to select and
13627 build multilibs for architectures @code{avr2} and @code{avr25}.
13628 These architectures mix devices with and without @code{SPH}.
13629 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13630 the compiler driver adds or removes this option from the compiler
13631 proper's command line, because the compiler then knows if the device
13632 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13637 Use address register @code{X} in a way proposed by the hardware. This means
13638 that @code{X} is only used in indirect, post-increment or
13639 pre-decrement addressing.
13641 Without this option, the @code{X} register may be used in the same way
13642 as @code{Y} or @code{Z} which then is emulated by additional
13644 For example, loading a value with @code{X+const} addressing with a
13645 small non-negative @code{const < 64} to a register @var{Rn} is
13649 adiw r26, const ; X += const
13650 ld @var{Rn}, X ; @var{Rn} = *X
13651 sbiw r26, const ; X -= const
13655 @opindex mtiny-stack
13656 Only change the lower 8@tie{}bits of the stack pointer.
13659 @opindex nodevicelib
13660 Don't link against AVR-LibC's device specific library @code{libdev.a}.
13662 @item -Waddr-space-convert
13663 @opindex Waddr-space-convert
13664 Warn about conversions between address spaces in the case where the
13665 resulting address space is not contained in the incoming address space.
13668 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13669 @cindex @code{EIND}
13670 Pointers in the implementation are 16@tie{}bits wide.
13671 The address of a function or label is represented as word address so
13672 that indirect jumps and calls can target any code address in the
13673 range of 64@tie{}Ki words.
13675 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13676 bytes of program memory space, there is a special function register called
13677 @code{EIND} that serves as most significant part of the target address
13678 when @code{EICALL} or @code{EIJMP} instructions are used.
13680 Indirect jumps and calls on these devices are handled as follows by
13681 the compiler and are subject to some limitations:
13686 The compiler never sets @code{EIND}.
13689 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13690 instructions or might read @code{EIND} directly in order to emulate an
13691 indirect call/jump by means of a @code{RET} instruction.
13694 The compiler assumes that @code{EIND} never changes during the startup
13695 code or during the application. In particular, @code{EIND} is not
13696 saved/restored in function or interrupt service routine
13700 For indirect calls to functions and computed goto, the linker
13701 generates @emph{stubs}. Stubs are jump pads sometimes also called
13702 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13703 The stub contains a direct jump to the desired address.
13706 Linker relaxation must be turned on so that the linker generates
13707 the stubs correctly in all situations. See the compiler option
13708 @option{-mrelax} and the linker option @option{--relax}.
13709 There are corner cases where the linker is supposed to generate stubs
13710 but aborts without relaxation and without a helpful error message.
13713 The default linker script is arranged for code with @code{EIND = 0}.
13714 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13715 linker script has to be used in order to place the sections whose
13716 name start with @code{.trampolines} into the segment where @code{EIND}
13720 The startup code from libgcc never sets @code{EIND}.
13721 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13722 For the impact of AVR-LibC on @code{EIND}, see the
13723 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13726 It is legitimate for user-specific startup code to set up @code{EIND}
13727 early, for example by means of initialization code located in
13728 section @code{.init3}. Such code runs prior to general startup code
13729 that initializes RAM and calls constructors, but after the bit
13730 of startup code from AVR-LibC that sets @code{EIND} to the segment
13731 where the vector table is located.
13733 #include <avr/io.h>
13736 __attribute__((section(".init3"),naked,used,no_instrument_function))
13737 init3_set_eind (void)
13739 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13740 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13745 The @code{__trampolines_start} symbol is defined in the linker script.
13748 Stubs are generated automatically by the linker if
13749 the following two conditions are met:
13752 @item The address of a label is taken by means of the @code{gs} modifier
13753 (short for @emph{generate stubs}) like so:
13755 LDI r24, lo8(gs(@var{func}))
13756 LDI r25, hi8(gs(@var{func}))
13758 @item The final location of that label is in a code segment
13759 @emph{outside} the segment where the stubs are located.
13763 The compiler emits such @code{gs} modifiers for code labels in the
13764 following situations:
13766 @item Taking address of a function or code label.
13767 @item Computed goto.
13768 @item If prologue-save function is used, see @option{-mcall-prologues}
13769 command-line option.
13770 @item Switch/case dispatch tables. If you do not want such dispatch
13771 tables you can specify the @option{-fno-jump-tables} command-line option.
13772 @item C and C++ constructors/destructors called during startup/shutdown.
13773 @item If the tools hit a @code{gs()} modifier explained above.
13777 Jumping to non-symbolic addresses like so is @emph{not} supported:
13782 /* Call function at word address 0x2 */
13783 return ((int(*)(void)) 0x2)();
13787 Instead, a stub has to be set up, i.e.@: the function has to be called
13788 through a symbol (@code{func_4} in the example):
13793 extern int func_4 (void);
13795 /* Call function at byte address 0x4 */
13800 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13801 Alternatively, @code{func_4} can be defined in the linker script.
13804 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13805 @cindex @code{RAMPD}
13806 @cindex @code{RAMPX}
13807 @cindex @code{RAMPY}
13808 @cindex @code{RAMPZ}
13809 Some AVR devices support memories larger than the 64@tie{}KiB range
13810 that can be accessed with 16-bit pointers. To access memory locations
13811 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13812 register is used as high part of the address:
13813 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13814 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13815 register, respectively, to get a wide address. Similarly,
13816 @code{RAMPD} is used together with direct addressing.
13820 The startup code initializes the @code{RAMP} special function
13821 registers with zero.
13824 If a @ref{AVR Named Address Spaces,named address space} other than
13825 generic or @code{__flash} is used, then @code{RAMPZ} is set
13826 as needed before the operation.
13829 If the device supports RAM larger than 64@tie{}KiB and the compiler
13830 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13831 is reset to zero after the operation.
13834 If the device comes with a specific @code{RAMP} register, the ISR
13835 prologue/epilogue saves/restores that SFR and initializes it with
13836 zero in case the ISR code might (implicitly) use it.
13839 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13840 If you use inline assembler to read from locations outside the
13841 16-bit address range and change one of the @code{RAMP} registers,
13842 you must reset it to zero after the access.
13846 @subsubsection AVR Built-in Macros
13848 GCC defines several built-in macros so that the user code can test
13849 for the presence or absence of features. Almost any of the following
13850 built-in macros are deduced from device capabilities and thus
13851 triggered by the @option{-mmcu=} command-line option.
13853 For even more AVR-specific built-in macros see
13854 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
13859 Build-in macro that resolves to a decimal number that identifies the
13860 architecture and depends on the @option{-mmcu=@var{mcu}} option.
13861 Possible values are:
13863 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
13864 @code{4}, @code{5}, @code{51}, @code{6}
13866 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
13867 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
13871 @code{100}, @code{102}, @code{104},
13872 @code{105}, @code{106}, @code{107}
13874 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
13875 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
13876 If @var{mcu} specifies a device, this built-in macro is set
13877 accordingly. For example, with @option{-mmcu=atmega8} the macro is
13878 defined to @code{4}.
13880 @item __AVR_@var{Device}__
13881 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
13882 the device's name. For example, @option{-mmcu=atmega8} defines the
13883 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
13884 @code{__AVR_ATtiny261A__}, etc.
13886 The built-in macros' names follow
13887 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
13888 the device name as from the AVR user manual. The difference between
13889 @var{Device} in the built-in macro and @var{device} in
13890 @option{-mmcu=@var{device}} is that the latter is always lowercase.
13892 If @var{device} is not a device but only a core architecture like
13893 @samp{avr51}, this macro is not defined.
13895 @item __AVR_DEVICE_NAME__
13896 Setting @option{-mmcu=@var{device}} defines this built-in macro to
13897 the device's name. For example, with @option{-mmcu=atmega8} the macro
13898 is defined to @code{atmega8}.
13900 If @var{device} is not a device but only a core architecture like
13901 @samp{avr51}, this macro is not defined.
13903 @item __AVR_XMEGA__
13904 The device / architecture belongs to the XMEGA family of devices.
13906 @item __AVR_HAVE_ELPM__
13907 The device has the the @code{ELPM} instruction.
13909 @item __AVR_HAVE_ELPMX__
13910 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
13911 R@var{n},Z+} instructions.
13913 @item __AVR_HAVE_MOVW__
13914 The device has the @code{MOVW} instruction to perform 16-bit
13915 register-register moves.
13917 @item __AVR_HAVE_LPMX__
13918 The device has the @code{LPM R@var{n},Z} and
13919 @code{LPM R@var{n},Z+} instructions.
13921 @item __AVR_HAVE_MUL__
13922 The device has a hardware multiplier.
13924 @item __AVR_HAVE_JMP_CALL__
13925 The device has the @code{JMP} and @code{CALL} instructions.
13926 This is the case for devices with at least 16@tie{}KiB of program
13929 @item __AVR_HAVE_EIJMP_EICALL__
13930 @itemx __AVR_3_BYTE_PC__
13931 The device has the @code{EIJMP} and @code{EICALL} instructions.
13932 This is the case for devices with more than 128@tie{}KiB of program memory.
13933 This also means that the program counter
13934 (PC) is 3@tie{}bytes wide.
13936 @item __AVR_2_BYTE_PC__
13937 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
13938 with up to 128@tie{}KiB of program memory.
13940 @item __AVR_HAVE_8BIT_SP__
13941 @itemx __AVR_HAVE_16BIT_SP__
13942 The stack pointer (SP) register is treated as 8-bit respectively
13943 16-bit register by the compiler.
13944 The definition of these macros is affected by @option{-mtiny-stack}.
13946 @item __AVR_HAVE_SPH__
13948 The device has the SPH (high part of stack pointer) special function
13949 register or has an 8-bit stack pointer, respectively.
13950 The definition of these macros is affected by @option{-mmcu=} and
13951 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
13954 @item __AVR_HAVE_RAMPD__
13955 @itemx __AVR_HAVE_RAMPX__
13956 @itemx __AVR_HAVE_RAMPY__
13957 @itemx __AVR_HAVE_RAMPZ__
13958 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
13959 @code{RAMPZ} special function register, respectively.
13961 @item __NO_INTERRUPTS__
13962 This macro reflects the @option{-mno-interrupts} command-line option.
13964 @item __AVR_ERRATA_SKIP__
13965 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
13966 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
13967 instructions because of a hardware erratum. Skip instructions are
13968 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
13969 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
13972 @item __AVR_ISA_RMW__
13973 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
13975 @item __AVR_SFR_OFFSET__=@var{offset}
13976 Instructions that can address I/O special function registers directly
13977 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
13978 address as if addressed by an instruction to access RAM like @code{LD}
13979 or @code{STS}. This offset depends on the device architecture and has
13980 to be subtracted from the RAM address in order to get the
13981 respective I/O@tie{}address.
13983 @item __WITH_AVRLIBC__
13984 The compiler is configured to be used together with AVR-Libc.
13985 See the @option{--with-avrlibc} configure option.
13989 @node Blackfin Options
13990 @subsection Blackfin Options
13991 @cindex Blackfin Options
13994 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
13996 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
13997 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
13998 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
13999 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14000 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14001 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14002 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14003 @samp{bf561}, @samp{bf592}.
14005 The optional @var{sirevision} specifies the silicon revision of the target
14006 Blackfin processor. Any workarounds available for the targeted silicon revision
14007 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14008 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14009 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14010 hexadecimal digits representing the major and minor numbers in the silicon
14011 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14012 is not defined. If @var{sirevision} is @samp{any}, the
14013 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14014 If this optional @var{sirevision} is not used, GCC assumes the latest known
14015 silicon revision of the targeted Blackfin processor.
14017 GCC defines a preprocessor macro for the specified @var{cpu}.
14018 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14019 provided by libgloss to be linked in if @option{-msim} is not given.
14021 Without this option, @samp{bf532} is used as the processor by default.
14023 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14024 only the preprocessor macro is defined.
14028 Specifies that the program will be run on the simulator. This causes
14029 the simulator BSP provided by libgloss to be linked in. This option
14030 has effect only for @samp{bfin-elf} toolchain.
14031 Certain other options, such as @option{-mid-shared-library} and
14032 @option{-mfdpic}, imply @option{-msim}.
14034 @item -momit-leaf-frame-pointer
14035 @opindex momit-leaf-frame-pointer
14036 Don't keep the frame pointer in a register for leaf functions. This
14037 avoids the instructions to save, set up and restore frame pointers and
14038 makes an extra register available in leaf functions. The option
14039 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14040 which might make debugging harder.
14042 @item -mspecld-anomaly
14043 @opindex mspecld-anomaly
14044 When enabled, the compiler ensures that the generated code does not
14045 contain speculative loads after jump instructions. If this option is used,
14046 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14048 @item -mno-specld-anomaly
14049 @opindex mno-specld-anomaly
14050 Don't generate extra code to prevent speculative loads from occurring.
14052 @item -mcsync-anomaly
14053 @opindex mcsync-anomaly
14054 When enabled, the compiler ensures that the generated code does not
14055 contain CSYNC or SSYNC instructions too soon after conditional branches.
14056 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14058 @item -mno-csync-anomaly
14059 @opindex mno-csync-anomaly
14060 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14061 occurring too soon after a conditional branch.
14065 When enabled, the compiler is free to take advantage of the knowledge that
14066 the entire program fits into the low 64k of memory.
14069 @opindex mno-low-64k
14070 Assume that the program is arbitrarily large. This is the default.
14072 @item -mstack-check-l1
14073 @opindex mstack-check-l1
14074 Do stack checking using information placed into L1 scratchpad memory by the
14077 @item -mid-shared-library
14078 @opindex mid-shared-library
14079 Generate code that supports shared libraries via the library ID method.
14080 This allows for execute in place and shared libraries in an environment
14081 without virtual memory management. This option implies @option{-fPIC}.
14082 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14084 @item -mno-id-shared-library
14085 @opindex mno-id-shared-library
14086 Generate code that doesn't assume ID-based shared libraries are being used.
14087 This is the default.
14089 @item -mleaf-id-shared-library
14090 @opindex mleaf-id-shared-library
14091 Generate code that supports shared libraries via the library ID method,
14092 but assumes that this library or executable won't link against any other
14093 ID shared libraries. That allows the compiler to use faster code for jumps
14096 @item -mno-leaf-id-shared-library
14097 @opindex mno-leaf-id-shared-library
14098 Do not assume that the code being compiled won't link against any ID shared
14099 libraries. Slower code is generated for jump and call insns.
14101 @item -mshared-library-id=n
14102 @opindex mshared-library-id
14103 Specifies the identification number of the ID-based shared library being
14104 compiled. Specifying a value of 0 generates more compact code; specifying
14105 other values forces the allocation of that number to the current
14106 library but is no more space- or time-efficient than omitting this option.
14110 Generate code that allows the data segment to be located in a different
14111 area of memory from the text segment. This allows for execute in place in
14112 an environment without virtual memory management by eliminating relocations
14113 against the text section.
14115 @item -mno-sep-data
14116 @opindex mno-sep-data
14117 Generate code that assumes that the data segment follows the text segment.
14118 This is the default.
14121 @itemx -mno-long-calls
14122 @opindex mlong-calls
14123 @opindex mno-long-calls
14124 Tells the compiler to perform function calls by first loading the
14125 address of the function into a register and then performing a subroutine
14126 call on this register. This switch is needed if the target function
14127 lies outside of the 24-bit addressing range of the offset-based
14128 version of subroutine call instruction.
14130 This feature is not enabled by default. Specifying
14131 @option{-mno-long-calls} restores the default behavior. Note these
14132 switches have no effect on how the compiler generates code to handle
14133 function calls via function pointers.
14137 Link with the fast floating-point library. This library relaxes some of
14138 the IEEE floating-point standard's rules for checking inputs against
14139 Not-a-Number (NAN), in the interest of performance.
14142 @opindex minline-plt
14143 Enable inlining of PLT entries in function calls to functions that are
14144 not known to bind locally. It has no effect without @option{-mfdpic}.
14147 @opindex mmulticore
14148 Build a standalone application for multicore Blackfin processors.
14149 This option causes proper start files and link scripts supporting
14150 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14151 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14153 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14154 selects the one-application-per-core programming model. Without
14155 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14156 programming model is used. In this model, the main function of Core B
14157 should be named as @code{coreb_main}.
14159 If this option is not used, the single-core application programming
14164 Build a standalone application for Core A of BF561 when using
14165 the one-application-per-core programming model. Proper start files
14166 and link scripts are used to support Core A, and the macro
14167 @code{__BFIN_COREA} is defined.
14168 This option can only be used in conjunction with @option{-mmulticore}.
14172 Build a standalone application for Core B of BF561 when using
14173 the one-application-per-core programming model. Proper start files
14174 and link scripts are used to support Core B, and the macro
14175 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14176 should be used instead of @code{main}.
14177 This option can only be used in conjunction with @option{-mmulticore}.
14181 Build a standalone application for SDRAM. Proper start files and
14182 link scripts are used to put the application into SDRAM, and the macro
14183 @code{__BFIN_SDRAM} is defined.
14184 The loader should initialize SDRAM before loading the application.
14188 Assume that ICPLBs are enabled at run time. This has an effect on certain
14189 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14190 are enabled; for standalone applications the default is off.
14194 @subsection C6X Options
14195 @cindex C6X Options
14198 @item -march=@var{name}
14200 This specifies the name of the target architecture. GCC uses this
14201 name to determine what kind of instructions it can emit when generating
14202 assembly code. Permissible names are: @samp{c62x},
14203 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14206 @opindex mbig-endian
14207 Generate code for a big-endian target.
14209 @item -mlittle-endian
14210 @opindex mlittle-endian
14211 Generate code for a little-endian target. This is the default.
14215 Choose startup files and linker script suitable for the simulator.
14217 @item -msdata=default
14218 @opindex msdata=default
14219 Put small global and static data in the @code{.neardata} section,
14220 which is pointed to by register @code{B14}. Put small uninitialized
14221 global and static data in the @code{.bss} section, which is adjacent
14222 to the @code{.neardata} section. Put small read-only data into the
14223 @code{.rodata} section. The corresponding sections used for large
14224 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14227 @opindex msdata=all
14228 Put all data, not just small objects, into the sections reserved for
14229 small data, and use addressing relative to the @code{B14} register to
14233 @opindex msdata=none
14234 Make no use of the sections reserved for small data, and use absolute
14235 addresses to access all data. Put all initialized global and static
14236 data in the @code{.fardata} section, and all uninitialized data in the
14237 @code{.far} section. Put all constant data into the @code{.const}
14242 @subsection CRIS Options
14243 @cindex CRIS Options
14245 These options are defined specifically for the CRIS ports.
14248 @item -march=@var{architecture-type}
14249 @itemx -mcpu=@var{architecture-type}
14252 Generate code for the specified architecture. The choices for
14253 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14254 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14255 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14258 @item -mtune=@var{architecture-type}
14260 Tune to @var{architecture-type} everything applicable about the generated
14261 code, except for the ABI and the set of available instructions. The
14262 choices for @var{architecture-type} are the same as for
14263 @option{-march=@var{architecture-type}}.
14265 @item -mmax-stack-frame=@var{n}
14266 @opindex mmax-stack-frame
14267 Warn when the stack frame of a function exceeds @var{n} bytes.
14273 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14274 @option{-march=v3} and @option{-march=v8} respectively.
14276 @item -mmul-bug-workaround
14277 @itemx -mno-mul-bug-workaround
14278 @opindex mmul-bug-workaround
14279 @opindex mno-mul-bug-workaround
14280 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14281 models where it applies. This option is active by default.
14285 Enable CRIS-specific verbose debug-related information in the assembly
14286 code. This option also has the effect of turning off the @samp{#NO_APP}
14287 formatted-code indicator to the assembler at the beginning of the
14292 Do not use condition-code results from previous instruction; always emit
14293 compare and test instructions before use of condition codes.
14295 @item -mno-side-effects
14296 @opindex mno-side-effects
14297 Do not emit instructions with side effects in addressing modes other than
14300 @item -mstack-align
14301 @itemx -mno-stack-align
14302 @itemx -mdata-align
14303 @itemx -mno-data-align
14304 @itemx -mconst-align
14305 @itemx -mno-const-align
14306 @opindex mstack-align
14307 @opindex mno-stack-align
14308 @opindex mdata-align
14309 @opindex mno-data-align
14310 @opindex mconst-align
14311 @opindex mno-const-align
14312 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14313 stack frame, individual data and constants to be aligned for the maximum
14314 single data access size for the chosen CPU model. The default is to
14315 arrange for 32-bit alignment. ABI details such as structure layout are
14316 not affected by these options.
14324 Similar to the stack- data- and const-align options above, these options
14325 arrange for stack frame, writable data and constants to all be 32-bit,
14326 16-bit or 8-bit aligned. The default is 32-bit alignment.
14328 @item -mno-prologue-epilogue
14329 @itemx -mprologue-epilogue
14330 @opindex mno-prologue-epilogue
14331 @opindex mprologue-epilogue
14332 With @option{-mno-prologue-epilogue}, the normal function prologue and
14333 epilogue which set up the stack frame are omitted and no return
14334 instructions or return sequences are generated in the code. Use this
14335 option only together with visual inspection of the compiled code: no
14336 warnings or errors are generated when call-saved registers must be saved,
14337 or storage for local variables needs to be allocated.
14341 @opindex mno-gotplt
14343 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14344 instruction sequences that load addresses for functions from the PLT part
14345 of the GOT rather than (traditional on other architectures) calls to the
14346 PLT@. The default is @option{-mgotplt}.
14350 Legacy no-op option only recognized with the cris-axis-elf and
14351 cris-axis-linux-gnu targets.
14355 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14359 This option, recognized for the cris-axis-elf, arranges
14360 to link with input-output functions from a simulator library. Code,
14361 initialized data and zero-initialized data are allocated consecutively.
14365 Like @option{-sim}, but pass linker options to locate initialized data at
14366 0x40000000 and zero-initialized data at 0x80000000.
14370 @subsection CR16 Options
14371 @cindex CR16 Options
14373 These options are defined specifically for the CR16 ports.
14379 Enable the use of multiply-accumulate instructions. Disabled by default.
14383 @opindex mcr16cplus
14385 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14390 Links the library libsim.a which is in compatible with simulator. Applicable
14391 to ELF compiler only.
14395 Choose integer type as 32-bit wide.
14399 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14401 @item -mdata-model=@var{model}
14402 @opindex mdata-model
14403 Choose a data model. The choices for @var{model} are @samp{near},
14404 @samp{far} or @samp{medium}. @samp{medium} is default.
14405 However, @samp{far} is not valid with @option{-mcr16c}, as the
14406 CR16C architecture does not support the far data model.
14409 @node Darwin Options
14410 @subsection Darwin Options
14411 @cindex Darwin options
14413 These options are defined for all architectures running the Darwin operating
14416 FSF GCC on Darwin does not create ``fat'' object files; it creates
14417 an object file for the single architecture that GCC was built to
14418 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14419 @option{-arch} options are used; it does so by running the compiler or
14420 linker multiple times and joining the results together with
14423 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14424 @samp{i686}) is determined by the flags that specify the ISA
14425 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14426 @option{-force_cpusubtype_ALL} option can be used to override this.
14428 The Darwin tools vary in their behavior when presented with an ISA
14429 mismatch. The assembler, @file{as}, only permits instructions to
14430 be used that are valid for the subtype of the file it is generating,
14431 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14432 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14433 and prints an error if asked to create a shared library with a less
14434 restrictive subtype than its input files (for instance, trying to put
14435 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14436 for executables, @command{ld}, quietly gives the executable the most
14437 restrictive subtype of any of its input files.
14442 Add the framework directory @var{dir} to the head of the list of
14443 directories to be searched for header files. These directories are
14444 interleaved with those specified by @option{-I} options and are
14445 scanned in a left-to-right order.
14447 A framework directory is a directory with frameworks in it. A
14448 framework is a directory with a @file{Headers} and/or
14449 @file{PrivateHeaders} directory contained directly in it that ends
14450 in @file{.framework}. The name of a framework is the name of this
14451 directory excluding the @file{.framework}. Headers associated with
14452 the framework are found in one of those two directories, with
14453 @file{Headers} being searched first. A subframework is a framework
14454 directory that is in a framework's @file{Frameworks} directory.
14455 Includes of subframework headers can only appear in a header of a
14456 framework that contains the subframework, or in a sibling subframework
14457 header. Two subframeworks are siblings if they occur in the same
14458 framework. A subframework should not have the same name as a
14459 framework; a warning is issued if this is violated. Currently a
14460 subframework cannot have subframeworks; in the future, the mechanism
14461 may be extended to support this. The standard frameworks can be found
14462 in @file{/System/Library/Frameworks} and
14463 @file{/Library/Frameworks}. An example include looks like
14464 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14465 the name of the framework and @file{header.h} is found in the
14466 @file{PrivateHeaders} or @file{Headers} directory.
14468 @item -iframework@var{dir}
14469 @opindex iframework
14470 Like @option{-F} except the directory is a treated as a system
14471 directory. The main difference between this @option{-iframework} and
14472 @option{-F} is that with @option{-iframework} the compiler does not
14473 warn about constructs contained within header files found via
14474 @var{dir}. This option is valid only for the C family of languages.
14478 Emit debugging information for symbols that are used. For stabs
14479 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14480 This is by default ON@.
14484 Emit debugging information for all symbols and types.
14486 @item -mmacosx-version-min=@var{version}
14487 The earliest version of MacOS X that this executable will run on
14488 is @var{version}. Typical values of @var{version} include @code{10.1},
14489 @code{10.2}, and @code{10.3.9}.
14491 If the compiler was built to use the system's headers by default,
14492 then the default for this option is the system version on which the
14493 compiler is running, otherwise the default is to make choices that
14494 are compatible with as many systems and code bases as possible.
14498 Enable kernel development mode. The @option{-mkernel} option sets
14499 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14500 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14501 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14502 applicable. This mode also sets @option{-mno-altivec},
14503 @option{-msoft-float}, @option{-fno-builtin} and
14504 @option{-mlong-branch} for PowerPC targets.
14506 @item -mone-byte-bool
14507 @opindex mone-byte-bool
14508 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14509 By default @code{sizeof(bool)} is @code{4} when compiling for
14510 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14511 option has no effect on x86.
14513 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14514 to generate code that is not binary compatible with code generated
14515 without that switch. Using this switch may require recompiling all
14516 other modules in a program, including system libraries. Use this
14517 switch to conform to a non-default data model.
14519 @item -mfix-and-continue
14520 @itemx -ffix-and-continue
14521 @itemx -findirect-data
14522 @opindex mfix-and-continue
14523 @opindex ffix-and-continue
14524 @opindex findirect-data
14525 Generate code suitable for fast turnaround development, such as to
14526 allow GDB to dynamically load @file{.o} files into already-running
14527 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14528 are provided for backwards compatibility.
14532 Loads all members of static archive libraries.
14533 See man ld(1) for more information.
14535 @item -arch_errors_fatal
14536 @opindex arch_errors_fatal
14537 Cause the errors having to do with files that have the wrong architecture
14540 @item -bind_at_load
14541 @opindex bind_at_load
14542 Causes the output file to be marked such that the dynamic linker will
14543 bind all undefined references when the file is loaded or launched.
14547 Produce a Mach-o bundle format file.
14548 See man ld(1) for more information.
14550 @item -bundle_loader @var{executable}
14551 @opindex bundle_loader
14552 This option specifies the @var{executable} that will load the build
14553 output file being linked. See man ld(1) for more information.
14556 @opindex dynamiclib
14557 When passed this option, GCC produces a dynamic library instead of
14558 an executable when linking, using the Darwin @file{libtool} command.
14560 @item -force_cpusubtype_ALL
14561 @opindex force_cpusubtype_ALL
14562 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14563 one controlled by the @option{-mcpu} or @option{-march} option.
14565 @item -allowable_client @var{client_name}
14566 @itemx -client_name
14567 @itemx -compatibility_version
14568 @itemx -current_version
14570 @itemx -dependency-file
14572 @itemx -dylinker_install_name
14574 @itemx -exported_symbols_list
14577 @itemx -flat_namespace
14578 @itemx -force_flat_namespace
14579 @itemx -headerpad_max_install_names
14582 @itemx -install_name
14583 @itemx -keep_private_externs
14584 @itemx -multi_module
14585 @itemx -multiply_defined
14586 @itemx -multiply_defined_unused
14589 @itemx -no_dead_strip_inits_and_terms
14590 @itemx -nofixprebinding
14591 @itemx -nomultidefs
14593 @itemx -noseglinkedit
14594 @itemx -pagezero_size
14596 @itemx -prebind_all_twolevel_modules
14597 @itemx -private_bundle
14599 @itemx -read_only_relocs
14601 @itemx -sectobjectsymbols
14605 @itemx -sectobjectsymbols
14608 @itemx -segs_read_only_addr
14610 @itemx -segs_read_write_addr
14611 @itemx -seg_addr_table
14612 @itemx -seg_addr_table_filename
14613 @itemx -seglinkedit
14615 @itemx -segs_read_only_addr
14616 @itemx -segs_read_write_addr
14617 @itemx -single_module
14619 @itemx -sub_library
14621 @itemx -sub_umbrella
14622 @itemx -twolevel_namespace
14625 @itemx -unexported_symbols_list
14626 @itemx -weak_reference_mismatches
14627 @itemx -whatsloaded
14628 @opindex allowable_client
14629 @opindex client_name
14630 @opindex compatibility_version
14631 @opindex current_version
14632 @opindex dead_strip
14633 @opindex dependency-file
14634 @opindex dylib_file
14635 @opindex dylinker_install_name
14637 @opindex exported_symbols_list
14639 @opindex flat_namespace
14640 @opindex force_flat_namespace
14641 @opindex headerpad_max_install_names
14642 @opindex image_base
14644 @opindex install_name
14645 @opindex keep_private_externs
14646 @opindex multi_module
14647 @opindex multiply_defined
14648 @opindex multiply_defined_unused
14649 @opindex noall_load
14650 @opindex no_dead_strip_inits_and_terms
14651 @opindex nofixprebinding
14652 @opindex nomultidefs
14654 @opindex noseglinkedit
14655 @opindex pagezero_size
14657 @opindex prebind_all_twolevel_modules
14658 @opindex private_bundle
14659 @opindex read_only_relocs
14661 @opindex sectobjectsymbols
14664 @opindex sectcreate
14665 @opindex sectobjectsymbols
14668 @opindex segs_read_only_addr
14669 @opindex segs_read_write_addr
14670 @opindex seg_addr_table
14671 @opindex seg_addr_table_filename
14672 @opindex seglinkedit
14674 @opindex segs_read_only_addr
14675 @opindex segs_read_write_addr
14676 @opindex single_module
14678 @opindex sub_library
14679 @opindex sub_umbrella
14680 @opindex twolevel_namespace
14683 @opindex unexported_symbols_list
14684 @opindex weak_reference_mismatches
14685 @opindex whatsloaded
14686 These options are passed to the Darwin linker. The Darwin linker man page
14687 describes them in detail.
14690 @node DEC Alpha Options
14691 @subsection DEC Alpha Options
14693 These @samp{-m} options are defined for the DEC Alpha implementations:
14696 @item -mno-soft-float
14697 @itemx -msoft-float
14698 @opindex mno-soft-float
14699 @opindex msoft-float
14700 Use (do not use) the hardware floating-point instructions for
14701 floating-point operations. When @option{-msoft-float} is specified,
14702 functions in @file{libgcc.a} are used to perform floating-point
14703 operations. Unless they are replaced by routines that emulate the
14704 floating-point operations, or compiled in such a way as to call such
14705 emulations routines, these routines issue floating-point
14706 operations. If you are compiling for an Alpha without floating-point
14707 operations, you must ensure that the library is built so as not to call
14710 Note that Alpha implementations without floating-point operations are
14711 required to have floating-point registers.
14714 @itemx -mno-fp-regs
14716 @opindex mno-fp-regs
14717 Generate code that uses (does not use) the floating-point register set.
14718 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14719 register set is not used, floating-point operands are passed in integer
14720 registers as if they were integers and floating-point results are passed
14721 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14722 so any function with a floating-point argument or return value called by code
14723 compiled with @option{-mno-fp-regs} must also be compiled with that
14726 A typical use of this option is building a kernel that does not use,
14727 and hence need not save and restore, any floating-point registers.
14731 The Alpha architecture implements floating-point hardware optimized for
14732 maximum performance. It is mostly compliant with the IEEE floating-point
14733 standard. However, for full compliance, software assistance is
14734 required. This option generates code fully IEEE-compliant code
14735 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14736 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14737 defined during compilation. The resulting code is less efficient but is
14738 able to correctly support denormalized numbers and exceptional IEEE
14739 values such as not-a-number and plus/minus infinity. Other Alpha
14740 compilers call this option @option{-ieee_with_no_inexact}.
14742 @item -mieee-with-inexact
14743 @opindex mieee-with-inexact
14744 This is like @option{-mieee} except the generated code also maintains
14745 the IEEE @var{inexact-flag}. Turning on this option causes the
14746 generated code to implement fully-compliant IEEE math. In addition to
14747 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14748 macro. On some Alpha implementations the resulting code may execute
14749 significantly slower than the code generated by default. Since there is
14750 very little code that depends on the @var{inexact-flag}, you should
14751 normally not specify this option. Other Alpha compilers call this
14752 option @option{-ieee_with_inexact}.
14754 @item -mfp-trap-mode=@var{trap-mode}
14755 @opindex mfp-trap-mode
14756 This option controls what floating-point related traps are enabled.
14757 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14758 The trap mode can be set to one of four values:
14762 This is the default (normal) setting. The only traps that are enabled
14763 are the ones that cannot be disabled in software (e.g., division by zero
14767 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14771 Like @samp{u}, but the instructions are marked to be safe for software
14772 completion (see Alpha architecture manual for details).
14775 Like @samp{su}, but inexact traps are enabled as well.
14778 @item -mfp-rounding-mode=@var{rounding-mode}
14779 @opindex mfp-rounding-mode
14780 Selects the IEEE rounding mode. Other Alpha compilers call this option
14781 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14786 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14787 the nearest machine number or towards the even machine number in case
14791 Round towards minus infinity.
14794 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14797 Dynamic rounding mode. A field in the floating-point control register
14798 (@var{fpcr}, see Alpha architecture reference manual) controls the
14799 rounding mode in effect. The C library initializes this register for
14800 rounding towards plus infinity. Thus, unless your program modifies the
14801 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14804 @item -mtrap-precision=@var{trap-precision}
14805 @opindex mtrap-precision
14806 In the Alpha architecture, floating-point traps are imprecise. This
14807 means without software assistance it is impossible to recover from a
14808 floating trap and program execution normally needs to be terminated.
14809 GCC can generate code that can assist operating system trap handlers
14810 in determining the exact location that caused a floating-point trap.
14811 Depending on the requirements of an application, different levels of
14812 precisions can be selected:
14816 Program precision. This option is the default and means a trap handler
14817 can only identify which program caused a floating-point exception.
14820 Function precision. The trap handler can determine the function that
14821 caused a floating-point exception.
14824 Instruction precision. The trap handler can determine the exact
14825 instruction that caused a floating-point exception.
14828 Other Alpha compilers provide the equivalent options called
14829 @option{-scope_safe} and @option{-resumption_safe}.
14831 @item -mieee-conformant
14832 @opindex mieee-conformant
14833 This option marks the generated code as IEEE conformant. You must not
14834 use this option unless you also specify @option{-mtrap-precision=i} and either
14835 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14836 is to emit the line @samp{.eflag 48} in the function prologue of the
14837 generated assembly file.
14839 @item -mbuild-constants
14840 @opindex mbuild-constants
14841 Normally GCC examines a 32- or 64-bit integer constant to
14842 see if it can construct it from smaller constants in two or three
14843 instructions. If it cannot, it outputs the constant as a literal and
14844 generates code to load it from the data segment at run time.
14846 Use this option to require GCC to construct @emph{all} integer constants
14847 using code, even if it takes more instructions (the maximum is six).
14849 You typically use this option to build a shared library dynamic
14850 loader. Itself a shared library, it must relocate itself in memory
14851 before it can find the variables and constants in its own data segment.
14869 Indicate whether GCC should generate code to use the optional BWX,
14870 CIX, FIX and MAX instruction sets. The default is to use the instruction
14871 sets supported by the CPU type specified via @option{-mcpu=} option or that
14872 of the CPU on which GCC was built if none is specified.
14875 @itemx -mfloat-ieee
14876 @opindex mfloat-vax
14877 @opindex mfloat-ieee
14878 Generate code that uses (does not use) VAX F and G floating-point
14879 arithmetic instead of IEEE single and double precision.
14881 @item -mexplicit-relocs
14882 @itemx -mno-explicit-relocs
14883 @opindex mexplicit-relocs
14884 @opindex mno-explicit-relocs
14885 Older Alpha assemblers provided no way to generate symbol relocations
14886 except via assembler macros. Use of these macros does not allow
14887 optimal instruction scheduling. GNU binutils as of version 2.12
14888 supports a new syntax that allows the compiler to explicitly mark
14889 which relocations should apply to which instructions. This option
14890 is mostly useful for debugging, as GCC detects the capabilities of
14891 the assembler when it is built and sets the default accordingly.
14894 @itemx -mlarge-data
14895 @opindex msmall-data
14896 @opindex mlarge-data
14897 When @option{-mexplicit-relocs} is in effect, static data is
14898 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
14899 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
14900 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
14901 16-bit relocations off of the @code{$gp} register. This limits the
14902 size of the small data area to 64KB, but allows the variables to be
14903 directly accessed via a single instruction.
14905 The default is @option{-mlarge-data}. With this option the data area
14906 is limited to just below 2GB@. Programs that require more than 2GB of
14907 data must use @code{malloc} or @code{mmap} to allocate the data in the
14908 heap instead of in the program's data segment.
14910 When generating code for shared libraries, @option{-fpic} implies
14911 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
14914 @itemx -mlarge-text
14915 @opindex msmall-text
14916 @opindex mlarge-text
14917 When @option{-msmall-text} is used, the compiler assumes that the
14918 code of the entire program (or shared library) fits in 4MB, and is
14919 thus reachable with a branch instruction. When @option{-msmall-data}
14920 is used, the compiler can assume that all local symbols share the
14921 same @code{$gp} value, and thus reduce the number of instructions
14922 required for a function call from 4 to 1.
14924 The default is @option{-mlarge-text}.
14926 @item -mcpu=@var{cpu_type}
14928 Set the instruction set and instruction scheduling parameters for
14929 machine type @var{cpu_type}. You can specify either the @samp{EV}
14930 style name or the corresponding chip number. GCC supports scheduling
14931 parameters for the EV4, EV5 and EV6 family of processors and
14932 chooses the default values for the instruction set from the processor
14933 you specify. If you do not specify a processor type, GCC defaults
14934 to the processor on which the compiler was built.
14936 Supported values for @var{cpu_type} are
14942 Schedules as an EV4 and has no instruction set extensions.
14946 Schedules as an EV5 and has no instruction set extensions.
14950 Schedules as an EV5 and supports the BWX extension.
14955 Schedules as an EV5 and supports the BWX and MAX extensions.
14959 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
14963 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
14966 Native toolchains also support the value @samp{native},
14967 which selects the best architecture option for the host processor.
14968 @option{-mcpu=native} has no effect if GCC does not recognize
14971 @item -mtune=@var{cpu_type}
14973 Set only the instruction scheduling parameters for machine type
14974 @var{cpu_type}. The instruction set is not changed.
14976 Native toolchains also support the value @samp{native},
14977 which selects the best architecture option for the host processor.
14978 @option{-mtune=native} has no effect if GCC does not recognize
14981 @item -mmemory-latency=@var{time}
14982 @opindex mmemory-latency
14983 Sets the latency the scheduler should assume for typical memory
14984 references as seen by the application. This number is highly
14985 dependent on the memory access patterns used by the application
14986 and the size of the external cache on the machine.
14988 Valid options for @var{time} are
14992 A decimal number representing clock cycles.
14998 The compiler contains estimates of the number of clock cycles for
14999 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15000 (also called Dcache, Scache, and Bcache), as well as to main memory.
15001 Note that L3 is only valid for EV5.
15007 @subsection FR30 Options
15008 @cindex FR30 Options
15010 These options are defined specifically for the FR30 port.
15014 @item -msmall-model
15015 @opindex msmall-model
15016 Use the small address space model. This can produce smaller code, but
15017 it does assume that all symbolic values and addresses fit into a
15022 Assume that runtime support has been provided and so there is no need
15023 to include the simulator library (@file{libsim.a}) on the linker
15029 @subsection FRV Options
15030 @cindex FRV Options
15036 Only use the first 32 general-purpose registers.
15041 Use all 64 general-purpose registers.
15046 Use only the first 32 floating-point registers.
15051 Use all 64 floating-point registers.
15054 @opindex mhard-float
15056 Use hardware instructions for floating-point operations.
15059 @opindex msoft-float
15061 Use library routines for floating-point operations.
15066 Dynamically allocate condition code registers.
15071 Do not try to dynamically allocate condition code registers, only
15072 use @code{icc0} and @code{fcc0}.
15077 Change ABI to use double word insns.
15082 Do not use double word instructions.
15087 Use floating-point double instructions.
15090 @opindex mno-double
15092 Do not use floating-point double instructions.
15097 Use media instructions.
15102 Do not use media instructions.
15107 Use multiply and add/subtract instructions.
15110 @opindex mno-muladd
15112 Do not use multiply and add/subtract instructions.
15117 Select the FDPIC ABI, which uses function descriptors to represent
15118 pointers to functions. Without any PIC/PIE-related options, it
15119 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15120 assumes GOT entries and small data are within a 12-bit range from the
15121 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15122 are computed with 32 bits.
15123 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15126 @opindex minline-plt
15128 Enable inlining of PLT entries in function calls to functions that are
15129 not known to bind locally. It has no effect without @option{-mfdpic}.
15130 It's enabled by default if optimizing for speed and compiling for
15131 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15132 optimization option such as @option{-O3} or above is present in the
15138 Assume a large TLS segment when generating thread-local code.
15143 Do not assume a large TLS segment when generating thread-local code.
15148 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15149 that is known to be in read-only sections. It's enabled by default,
15150 except for @option{-fpic} or @option{-fpie}: even though it may help
15151 make the global offset table smaller, it trades 1 instruction for 4.
15152 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15153 one of which may be shared by multiple symbols, and it avoids the need
15154 for a GOT entry for the referenced symbol, so it's more likely to be a
15155 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15157 @item -multilib-library-pic
15158 @opindex multilib-library-pic
15160 Link with the (library, not FD) pic libraries. It's implied by
15161 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15162 @option{-fpic} without @option{-mfdpic}. You should never have to use
15166 @opindex mlinked-fp
15168 Follow the EABI requirement of always creating a frame pointer whenever
15169 a stack frame is allocated. This option is enabled by default and can
15170 be disabled with @option{-mno-linked-fp}.
15173 @opindex mlong-calls
15175 Use indirect addressing to call functions outside the current
15176 compilation unit. This allows the functions to be placed anywhere
15177 within the 32-bit address space.
15179 @item -malign-labels
15180 @opindex malign-labels
15182 Try to align labels to an 8-byte boundary by inserting NOPs into the
15183 previous packet. This option only has an effect when VLIW packing
15184 is enabled. It doesn't create new packets; it merely adds NOPs to
15187 @item -mlibrary-pic
15188 @opindex mlibrary-pic
15190 Generate position-independent EABI code.
15195 Use only the first four media accumulator registers.
15200 Use all eight media accumulator registers.
15205 Pack VLIW instructions.
15210 Do not pack VLIW instructions.
15213 @opindex mno-eflags
15215 Do not mark ABI switches in e_flags.
15218 @opindex mcond-move
15220 Enable the use of conditional-move instructions (default).
15222 This switch is mainly for debugging the compiler and will likely be removed
15223 in a future version.
15225 @item -mno-cond-move
15226 @opindex mno-cond-move
15228 Disable the use of conditional-move instructions.
15230 This switch is mainly for debugging the compiler and will likely be removed
15231 in a future version.
15236 Enable the use of conditional set instructions (default).
15238 This switch is mainly for debugging the compiler and will likely be removed
15239 in a future version.
15244 Disable the use of conditional set instructions.
15246 This switch is mainly for debugging the compiler and will likely be removed
15247 in a future version.
15250 @opindex mcond-exec
15252 Enable the use of conditional execution (default).
15254 This switch is mainly for debugging the compiler and will likely be removed
15255 in a future version.
15257 @item -mno-cond-exec
15258 @opindex mno-cond-exec
15260 Disable the use of conditional execution.
15262 This switch is mainly for debugging the compiler and will likely be removed
15263 in a future version.
15265 @item -mvliw-branch
15266 @opindex mvliw-branch
15268 Run a pass to pack branches into VLIW instructions (default).
15270 This switch is mainly for debugging the compiler and will likely be removed
15271 in a future version.
15273 @item -mno-vliw-branch
15274 @opindex mno-vliw-branch
15276 Do not run a pass to pack branches into VLIW instructions.
15278 This switch is mainly for debugging the compiler and will likely be removed
15279 in a future version.
15281 @item -mmulti-cond-exec
15282 @opindex mmulti-cond-exec
15284 Enable optimization of @code{&&} and @code{||} in conditional execution
15287 This switch is mainly for debugging the compiler and will likely be removed
15288 in a future version.
15290 @item -mno-multi-cond-exec
15291 @opindex mno-multi-cond-exec
15293 Disable optimization of @code{&&} and @code{||} in conditional execution.
15295 This switch is mainly for debugging the compiler and will likely be removed
15296 in a future version.
15298 @item -mnested-cond-exec
15299 @opindex mnested-cond-exec
15301 Enable nested conditional execution optimizations (default).
15303 This switch is mainly for debugging the compiler and will likely be removed
15304 in a future version.
15306 @item -mno-nested-cond-exec
15307 @opindex mno-nested-cond-exec
15309 Disable nested conditional execution optimizations.
15311 This switch is mainly for debugging the compiler and will likely be removed
15312 in a future version.
15314 @item -moptimize-membar
15315 @opindex moptimize-membar
15317 This switch removes redundant @code{membar} instructions from the
15318 compiler-generated code. It is enabled by default.
15320 @item -mno-optimize-membar
15321 @opindex mno-optimize-membar
15323 This switch disables the automatic removal of redundant @code{membar}
15324 instructions from the generated code.
15326 @item -mtomcat-stats
15327 @opindex mtomcat-stats
15329 Cause gas to print out tomcat statistics.
15331 @item -mcpu=@var{cpu}
15334 Select the processor type for which to generate code. Possible values are
15335 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15336 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15340 @node GNU/Linux Options
15341 @subsection GNU/Linux Options
15343 These @samp{-m} options are defined for GNU/Linux targets:
15348 Use the GNU C library. This is the default except
15349 on @samp{*-*-linux-*uclibc*} and @samp{*-*-linux-*android*} targets.
15353 Use uClibc C library. This is the default on
15354 @samp{*-*-linux-*uclibc*} targets.
15358 Use Bionic C library. This is the default on
15359 @samp{*-*-linux-*android*} targets.
15363 Compile code compatible with Android platform. This is the default on
15364 @samp{*-*-linux-*android*} targets.
15366 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15367 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15368 this option makes the GCC driver pass Android-specific options to the linker.
15369 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15372 @item -tno-android-cc
15373 @opindex tno-android-cc
15374 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15375 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15376 @option{-fno-rtti} by default.
15378 @item -tno-android-ld
15379 @opindex tno-android-ld
15380 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15381 linking options to the linker.
15385 @node H8/300 Options
15386 @subsection H8/300 Options
15388 These @samp{-m} options are defined for the H8/300 implementations:
15393 Shorten some address references at link time, when possible; uses the
15394 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15395 ld, Using ld}, for a fuller description.
15399 Generate code for the H8/300H@.
15403 Generate code for the H8S@.
15407 Generate code for the H8S and H8/300H in the normal mode. This switch
15408 must be used either with @option{-mh} or @option{-ms}.
15412 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15416 Extended registers are stored on stack before execution of function
15417 with monitor attribute. Default option is @option{-mexr}.
15418 This option is valid only for H8S targets.
15422 Extended registers are not stored on stack before execution of function
15423 with monitor attribute. Default option is @option{-mno-exr}.
15424 This option is valid only for H8S targets.
15428 Make @code{int} data 32 bits by default.
15431 @opindex malign-300
15432 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15433 The default for the H8/300H and H8S is to align longs and floats on
15435 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15436 This option has no effect on the H8/300.
15440 @subsection HPPA Options
15441 @cindex HPPA Options
15443 These @samp{-m} options are defined for the HPPA family of computers:
15446 @item -march=@var{architecture-type}
15448 Generate code for the specified architecture. The choices for
15449 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15450 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15451 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15452 architecture option for your machine. Code compiled for lower numbered
15453 architectures runs on higher numbered architectures, but not the
15456 @item -mpa-risc-1-0
15457 @itemx -mpa-risc-1-1
15458 @itemx -mpa-risc-2-0
15459 @opindex mpa-risc-1-0
15460 @opindex mpa-risc-1-1
15461 @opindex mpa-risc-2-0
15462 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15464 @item -mjump-in-delay
15465 @opindex mjump-in-delay
15466 This option is ignored and provided for compatibility purposes only.
15468 @item -mdisable-fpregs
15469 @opindex mdisable-fpregs
15470 Prevent floating-point registers from being used in any manner. This is
15471 necessary for compiling kernels that perform lazy context switching of
15472 floating-point registers. If you use this option and attempt to perform
15473 floating-point operations, the compiler aborts.
15475 @item -mdisable-indexing
15476 @opindex mdisable-indexing
15477 Prevent the compiler from using indexing address modes. This avoids some
15478 rather obscure problems when compiling MIG generated code under MACH@.
15480 @item -mno-space-regs
15481 @opindex mno-space-regs
15482 Generate code that assumes the target has no space registers. This allows
15483 GCC to generate faster indirect calls and use unscaled index address modes.
15485 Such code is suitable for level 0 PA systems and kernels.
15487 @item -mfast-indirect-calls
15488 @opindex mfast-indirect-calls
15489 Generate code that assumes calls never cross space boundaries. This
15490 allows GCC to emit code that performs faster indirect calls.
15492 This option does not work in the presence of shared libraries or nested
15495 @item -mfixed-range=@var{register-range}
15496 @opindex mfixed-range
15497 Generate code treating the given register range as fixed registers.
15498 A fixed register is one that the register allocator cannot use. This is
15499 useful when compiling kernel code. A register range is specified as
15500 two registers separated by a dash. Multiple register ranges can be
15501 specified separated by a comma.
15503 @item -mlong-load-store
15504 @opindex mlong-load-store
15505 Generate 3-instruction load and store sequences as sometimes required by
15506 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15509 @item -mportable-runtime
15510 @opindex mportable-runtime
15511 Use the portable calling conventions proposed by HP for ELF systems.
15515 Enable the use of assembler directives only GAS understands.
15517 @item -mschedule=@var{cpu-type}
15519 Schedule code according to the constraints for the machine type
15520 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15521 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15522 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15523 proper scheduling option for your machine. The default scheduling is
15527 @opindex mlinker-opt
15528 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15529 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15530 linkers in which they give bogus error messages when linking some programs.
15533 @opindex msoft-float
15534 Generate output containing library calls for floating point.
15535 @strong{Warning:} the requisite libraries are not available for all HPPA
15536 targets. Normally the facilities of the machine's usual C compiler are
15537 used, but this cannot be done directly in cross-compilation. You must make
15538 your own arrangements to provide suitable library functions for
15541 @option{-msoft-float} changes the calling convention in the output file;
15542 therefore, it is only useful if you compile @emph{all} of a program with
15543 this option. In particular, you need to compile @file{libgcc.a}, the
15544 library that comes with GCC, with @option{-msoft-float} in order for
15549 Generate the predefine, @code{_SIO}, for server IO@. The default is
15550 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15551 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15552 options are available under HP-UX and HI-UX@.
15556 Use options specific to GNU @command{ld}.
15557 This passes @option{-shared} to @command{ld} when
15558 building a shared library. It is the default when GCC is configured,
15559 explicitly or implicitly, with the GNU linker. This option does not
15560 affect which @command{ld} is called; it only changes what parameters
15561 are passed to that @command{ld}.
15562 The @command{ld} that is called is determined by the
15563 @option{--with-ld} configure option, GCC's program search path, and
15564 finally by the user's @env{PATH}. The linker used by GCC can be printed
15565 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15566 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15570 Use options specific to HP @command{ld}.
15571 This passes @option{-b} to @command{ld} when building
15572 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15573 links. It is the default when GCC is configured, explicitly or
15574 implicitly, with the HP linker. This option does not affect
15575 which @command{ld} is called; it only changes what parameters are passed to that
15577 The @command{ld} that is called is determined by the @option{--with-ld}
15578 configure option, GCC's program search path, and finally by the user's
15579 @env{PATH}. The linker used by GCC can be printed using @samp{which
15580 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15581 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15584 @opindex mno-long-calls
15585 Generate code that uses long call sequences. This ensures that a call
15586 is always able to reach linker generated stubs. The default is to generate
15587 long calls only when the distance from the call site to the beginning
15588 of the function or translation unit, as the case may be, exceeds a
15589 predefined limit set by the branch type being used. The limits for
15590 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15591 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15594 Distances are measured from the beginning of functions when using the
15595 @option{-ffunction-sections} option, or when using the @option{-mgas}
15596 and @option{-mno-portable-runtime} options together under HP-UX with
15599 It is normally not desirable to use this option as it degrades
15600 performance. However, it may be useful in large applications,
15601 particularly when partial linking is used to build the application.
15603 The types of long calls used depends on the capabilities of the
15604 assembler and linker, and the type of code being generated. The
15605 impact on systems that support long absolute calls, and long pic
15606 symbol-difference or pc-relative calls should be relatively small.
15607 However, an indirect call is used on 32-bit ELF systems in pic code
15608 and it is quite long.
15610 @item -munix=@var{unix-std}
15612 Generate compiler predefines and select a startfile for the specified
15613 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15614 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15615 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15616 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15617 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15620 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15621 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15622 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15623 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15624 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15625 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15627 It is @emph{important} to note that this option changes the interfaces
15628 for various library routines. It also affects the operational behavior
15629 of the C library. Thus, @emph{extreme} care is needed in using this
15632 Library code that is intended to operate with more than one UNIX
15633 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15634 as appropriate. Most GNU software doesn't provide this capability.
15638 Suppress the generation of link options to search libdld.sl when the
15639 @option{-static} option is specified on HP-UX 10 and later.
15643 The HP-UX implementation of setlocale in libc has a dependency on
15644 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15645 when the @option{-static} option is specified, special link options
15646 are needed to resolve this dependency.
15648 On HP-UX 10 and later, the GCC driver adds the necessary options to
15649 link with libdld.sl when the @option{-static} option is specified.
15650 This causes the resulting binary to be dynamic. On the 64-bit port,
15651 the linkers generate dynamic binaries by default in any case. The
15652 @option{-nolibdld} option can be used to prevent the GCC driver from
15653 adding these link options.
15657 Add support for multithreading with the @dfn{dce thread} library
15658 under HP-UX@. This option sets flags for both the preprocessor and
15662 @node IA-64 Options
15663 @subsection IA-64 Options
15664 @cindex IA-64 Options
15666 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15670 @opindex mbig-endian
15671 Generate code for a big-endian target. This is the default for HP-UX@.
15673 @item -mlittle-endian
15674 @opindex mlittle-endian
15675 Generate code for a little-endian target. This is the default for AIX5
15681 @opindex mno-gnu-as
15682 Generate (or don't) code for the GNU assembler. This is the default.
15683 @c Also, this is the default if the configure option @option{--with-gnu-as}
15689 @opindex mno-gnu-ld
15690 Generate (or don't) code for the GNU linker. This is the default.
15691 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15696 Generate code that does not use a global pointer register. The result
15697 is not position independent code, and violates the IA-64 ABI@.
15699 @item -mvolatile-asm-stop
15700 @itemx -mno-volatile-asm-stop
15701 @opindex mvolatile-asm-stop
15702 @opindex mno-volatile-asm-stop
15703 Generate (or don't) a stop bit immediately before and after volatile asm
15706 @item -mregister-names
15707 @itemx -mno-register-names
15708 @opindex mregister-names
15709 @opindex mno-register-names
15710 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15711 the stacked registers. This may make assembler output more readable.
15717 Disable (or enable) optimizations that use the small data section. This may
15718 be useful for working around optimizer bugs.
15720 @item -mconstant-gp
15721 @opindex mconstant-gp
15722 Generate code that uses a single constant global pointer value. This is
15723 useful when compiling kernel code.
15727 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15728 This is useful when compiling firmware code.
15730 @item -minline-float-divide-min-latency
15731 @opindex minline-float-divide-min-latency
15732 Generate code for inline divides of floating-point values
15733 using the minimum latency algorithm.
15735 @item -minline-float-divide-max-throughput
15736 @opindex minline-float-divide-max-throughput
15737 Generate code for inline divides of floating-point values
15738 using the maximum throughput algorithm.
15740 @item -mno-inline-float-divide
15741 @opindex mno-inline-float-divide
15742 Do not generate inline code for divides of floating-point values.
15744 @item -minline-int-divide-min-latency
15745 @opindex minline-int-divide-min-latency
15746 Generate code for inline divides of integer values
15747 using the minimum latency algorithm.
15749 @item -minline-int-divide-max-throughput
15750 @opindex minline-int-divide-max-throughput
15751 Generate code for inline divides of integer values
15752 using the maximum throughput algorithm.
15754 @item -mno-inline-int-divide
15755 @opindex mno-inline-int-divide
15756 Do not generate inline code for divides of integer values.
15758 @item -minline-sqrt-min-latency
15759 @opindex minline-sqrt-min-latency
15760 Generate code for inline square roots
15761 using the minimum latency algorithm.
15763 @item -minline-sqrt-max-throughput
15764 @opindex minline-sqrt-max-throughput
15765 Generate code for inline square roots
15766 using the maximum throughput algorithm.
15768 @item -mno-inline-sqrt
15769 @opindex mno-inline-sqrt
15770 Do not generate inline code for @code{sqrt}.
15773 @itemx -mno-fused-madd
15774 @opindex mfused-madd
15775 @opindex mno-fused-madd
15776 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15777 instructions. The default is to use these instructions.
15779 @item -mno-dwarf2-asm
15780 @itemx -mdwarf2-asm
15781 @opindex mno-dwarf2-asm
15782 @opindex mdwarf2-asm
15783 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15784 info. This may be useful when not using the GNU assembler.
15786 @item -mearly-stop-bits
15787 @itemx -mno-early-stop-bits
15788 @opindex mearly-stop-bits
15789 @opindex mno-early-stop-bits
15790 Allow stop bits to be placed earlier than immediately preceding the
15791 instruction that triggered the stop bit. This can improve instruction
15792 scheduling, but does not always do so.
15794 @item -mfixed-range=@var{register-range}
15795 @opindex mfixed-range
15796 Generate code treating the given register range as fixed registers.
15797 A fixed register is one that the register allocator cannot use. This is
15798 useful when compiling kernel code. A register range is specified as
15799 two registers separated by a dash. Multiple register ranges can be
15800 specified separated by a comma.
15802 @item -mtls-size=@var{tls-size}
15804 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15807 @item -mtune=@var{cpu-type}
15809 Tune the instruction scheduling for a particular CPU, Valid values are
15810 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15811 and @samp{mckinley}.
15817 Generate code for a 32-bit or 64-bit environment.
15818 The 32-bit environment sets int, long and pointer to 32 bits.
15819 The 64-bit environment sets int to 32 bits and long and pointer
15820 to 64 bits. These are HP-UX specific flags.
15822 @item -mno-sched-br-data-spec
15823 @itemx -msched-br-data-spec
15824 @opindex mno-sched-br-data-spec
15825 @opindex msched-br-data-spec
15826 (Dis/En)able data speculative scheduling before reload.
15827 This results in generation of @code{ld.a} instructions and
15828 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15829 The default is 'disable'.
15831 @item -msched-ar-data-spec
15832 @itemx -mno-sched-ar-data-spec
15833 @opindex msched-ar-data-spec
15834 @opindex mno-sched-ar-data-spec
15835 (En/Dis)able data speculative scheduling after reload.
15836 This results in generation of @code{ld.a} instructions and
15837 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
15838 The default is 'enable'.
15840 @item -mno-sched-control-spec
15841 @itemx -msched-control-spec
15842 @opindex mno-sched-control-spec
15843 @opindex msched-control-spec
15844 (Dis/En)able control speculative scheduling. This feature is
15845 available only during region scheduling (i.e.@: before reload).
15846 This results in generation of the @code{ld.s} instructions and
15847 the corresponding check instructions @code{chk.s}.
15848 The default is 'disable'.
15850 @item -msched-br-in-data-spec
15851 @itemx -mno-sched-br-in-data-spec
15852 @opindex msched-br-in-data-spec
15853 @opindex mno-sched-br-in-data-spec
15854 (En/Dis)able speculative scheduling of the instructions that
15855 are dependent on the data speculative loads before reload.
15856 This is effective only with @option{-msched-br-data-spec} enabled.
15857 The default is 'enable'.
15859 @item -msched-ar-in-data-spec
15860 @itemx -mno-sched-ar-in-data-spec
15861 @opindex msched-ar-in-data-spec
15862 @opindex mno-sched-ar-in-data-spec
15863 (En/Dis)able speculative scheduling of the instructions that
15864 are dependent on the data speculative loads after reload.
15865 This is effective only with @option{-msched-ar-data-spec} enabled.
15866 The default is 'enable'.
15868 @item -msched-in-control-spec
15869 @itemx -mno-sched-in-control-spec
15870 @opindex msched-in-control-spec
15871 @opindex mno-sched-in-control-spec
15872 (En/Dis)able speculative scheduling of the instructions that
15873 are dependent on the control speculative loads.
15874 This is effective only with @option{-msched-control-spec} enabled.
15875 The default is 'enable'.
15877 @item -mno-sched-prefer-non-data-spec-insns
15878 @itemx -msched-prefer-non-data-spec-insns
15879 @opindex mno-sched-prefer-non-data-spec-insns
15880 @opindex msched-prefer-non-data-spec-insns
15881 If enabled, data-speculative instructions are chosen for schedule
15882 only if there are no other choices at the moment. This makes
15883 the use of the data speculation much more conservative.
15884 The default is 'disable'.
15886 @item -mno-sched-prefer-non-control-spec-insns
15887 @itemx -msched-prefer-non-control-spec-insns
15888 @opindex mno-sched-prefer-non-control-spec-insns
15889 @opindex msched-prefer-non-control-spec-insns
15890 If enabled, control-speculative instructions are chosen for schedule
15891 only if there are no other choices at the moment. This makes
15892 the use of the control speculation much more conservative.
15893 The default is 'disable'.
15895 @item -mno-sched-count-spec-in-critical-path
15896 @itemx -msched-count-spec-in-critical-path
15897 @opindex mno-sched-count-spec-in-critical-path
15898 @opindex msched-count-spec-in-critical-path
15899 If enabled, speculative dependencies are considered during
15900 computation of the instructions priorities. This makes the use of the
15901 speculation a bit more conservative.
15902 The default is 'disable'.
15904 @item -msched-spec-ldc
15905 @opindex msched-spec-ldc
15906 Use a simple data speculation check. This option is on by default.
15908 @item -msched-control-spec-ldc
15909 @opindex msched-spec-ldc
15910 Use a simple check for control speculation. This option is on by default.
15912 @item -msched-stop-bits-after-every-cycle
15913 @opindex msched-stop-bits-after-every-cycle
15914 Place a stop bit after every cycle when scheduling. This option is on
15917 @item -msched-fp-mem-deps-zero-cost
15918 @opindex msched-fp-mem-deps-zero-cost
15919 Assume that floating-point stores and loads are not likely to cause a conflict
15920 when placed into the same instruction group. This option is disabled by
15923 @item -msel-sched-dont-check-control-spec
15924 @opindex msel-sched-dont-check-control-spec
15925 Generate checks for control speculation in selective scheduling.
15926 This flag is disabled by default.
15928 @item -msched-max-memory-insns=@var{max-insns}
15929 @opindex msched-max-memory-insns
15930 Limit on the number of memory insns per instruction group, giving lower
15931 priority to subsequent memory insns attempting to schedule in the same
15932 instruction group. Frequently useful to prevent cache bank conflicts.
15933 The default value is 1.
15935 @item -msched-max-memory-insns-hard-limit
15936 @opindex msched-max-memory-insns-hard-limit
15937 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
15938 disallowing more than that number in an instruction group.
15939 Otherwise, the limit is ``soft'', meaning that non-memory operations
15940 are preferred when the limit is reached, but memory operations may still
15946 @subsection LM32 Options
15947 @cindex LM32 options
15949 These @option{-m} options are defined for the LatticeMico32 architecture:
15952 @item -mbarrel-shift-enabled
15953 @opindex mbarrel-shift-enabled
15954 Enable barrel-shift instructions.
15956 @item -mdivide-enabled
15957 @opindex mdivide-enabled
15958 Enable divide and modulus instructions.
15960 @item -mmultiply-enabled
15961 @opindex multiply-enabled
15962 Enable multiply instructions.
15964 @item -msign-extend-enabled
15965 @opindex msign-extend-enabled
15966 Enable sign extend instructions.
15968 @item -muser-enabled
15969 @opindex muser-enabled
15970 Enable user-defined instructions.
15975 @subsection M32C Options
15976 @cindex M32C options
15979 @item -mcpu=@var{name}
15981 Select the CPU for which code is generated. @var{name} may be one of
15982 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
15983 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
15984 the M32C/80 series.
15988 Specifies that the program will be run on the simulator. This causes
15989 an alternate runtime library to be linked in which supports, for
15990 example, file I/O@. You must not use this option when generating
15991 programs that will run on real hardware; you must provide your own
15992 runtime library for whatever I/O functions are needed.
15994 @item -memregs=@var{number}
15996 Specifies the number of memory-based pseudo-registers GCC uses
15997 during code generation. These pseudo-registers are used like real
15998 registers, so there is a tradeoff between GCC's ability to fit the
15999 code into available registers, and the performance penalty of using
16000 memory instead of registers. Note that all modules in a program must
16001 be compiled with the same value for this option. Because of that, you
16002 must not use this option with GCC's default runtime libraries.
16006 @node M32R/D Options
16007 @subsection M32R/D Options
16008 @cindex M32R/D options
16010 These @option{-m} options are defined for Renesas M32R/D architectures:
16015 Generate code for the M32R/2@.
16019 Generate code for the M32R/X@.
16023 Generate code for the M32R@. This is the default.
16025 @item -mmodel=small
16026 @opindex mmodel=small
16027 Assume all objects live in the lower 16MB of memory (so that their addresses
16028 can be loaded with the @code{ld24} instruction), and assume all subroutines
16029 are reachable with the @code{bl} instruction.
16030 This is the default.
16032 The addressability of a particular object can be set with the
16033 @code{model} attribute.
16035 @item -mmodel=medium
16036 @opindex mmodel=medium
16037 Assume objects may be anywhere in the 32-bit address space (the compiler
16038 generates @code{seth/add3} instructions to load their addresses), and
16039 assume all subroutines are reachable with the @code{bl} instruction.
16041 @item -mmodel=large
16042 @opindex mmodel=large
16043 Assume objects may be anywhere in the 32-bit address space (the compiler
16044 generates @code{seth/add3} instructions to load their addresses), and
16045 assume subroutines may not be reachable with the @code{bl} instruction
16046 (the compiler generates the much slower @code{seth/add3/jl}
16047 instruction sequence).
16050 @opindex msdata=none
16051 Disable use of the small data area. Variables are put into
16052 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16053 @code{section} attribute has been specified).
16054 This is the default.
16056 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16057 Objects may be explicitly put in the small data area with the
16058 @code{section} attribute using one of these sections.
16060 @item -msdata=sdata
16061 @opindex msdata=sdata
16062 Put small global and static data in the small data area, but do not
16063 generate special code to reference them.
16066 @opindex msdata=use
16067 Put small global and static data in the small data area, and generate
16068 special instructions to reference them.
16072 @cindex smaller data references
16073 Put global and static objects less than or equal to @var{num} bytes
16074 into the small data or BSS sections instead of the normal data or BSS
16075 sections. The default value of @var{num} is 8.
16076 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16077 for this option to have any effect.
16079 All modules should be compiled with the same @option{-G @var{num}} value.
16080 Compiling with different values of @var{num} may or may not work; if it
16081 doesn't the linker gives an error message---incorrect code is not
16086 Makes the M32R-specific code in the compiler display some statistics
16087 that might help in debugging programs.
16089 @item -malign-loops
16090 @opindex malign-loops
16091 Align all loops to a 32-byte boundary.
16093 @item -mno-align-loops
16094 @opindex mno-align-loops
16095 Do not enforce a 32-byte alignment for loops. This is the default.
16097 @item -missue-rate=@var{number}
16098 @opindex missue-rate=@var{number}
16099 Issue @var{number} instructions per cycle. @var{number} can only be 1
16102 @item -mbranch-cost=@var{number}
16103 @opindex mbranch-cost=@var{number}
16104 @var{number} can only be 1 or 2. If it is 1 then branches are
16105 preferred over conditional code, if it is 2, then the opposite applies.
16107 @item -mflush-trap=@var{number}
16108 @opindex mflush-trap=@var{number}
16109 Specifies the trap number to use to flush the cache. The default is
16110 12. Valid numbers are between 0 and 15 inclusive.
16112 @item -mno-flush-trap
16113 @opindex mno-flush-trap
16114 Specifies that the cache cannot be flushed by using a trap.
16116 @item -mflush-func=@var{name}
16117 @opindex mflush-func=@var{name}
16118 Specifies the name of the operating system function to call to flush
16119 the cache. The default is @samp{_flush_cache}, but a function call
16120 is only used if a trap is not available.
16122 @item -mno-flush-func
16123 @opindex mno-flush-func
16124 Indicates that there is no OS function for flushing the cache.
16128 @node M680x0 Options
16129 @subsection M680x0 Options
16130 @cindex M680x0 options
16132 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16133 The default settings depend on which architecture was selected when
16134 the compiler was configured; the defaults for the most common choices
16138 @item -march=@var{arch}
16140 Generate code for a specific M680x0 or ColdFire instruction set
16141 architecture. Permissible values of @var{arch} for M680x0
16142 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16143 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16144 architectures are selected according to Freescale's ISA classification
16145 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16146 @samp{isab} and @samp{isac}.
16148 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16149 code for a ColdFire target. The @var{arch} in this macro is one of the
16150 @option{-march} arguments given above.
16152 When used together, @option{-march} and @option{-mtune} select code
16153 that runs on a family of similar processors but that is optimized
16154 for a particular microarchitecture.
16156 @item -mcpu=@var{cpu}
16158 Generate code for a specific M680x0 or ColdFire processor.
16159 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16160 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16161 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16162 below, which also classifies the CPUs into families:
16164 @multitable @columnfractions 0.20 0.80
16165 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16166 @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}
16167 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16168 @item @samp{5206e} @tab @samp{5206e}
16169 @item @samp{5208} @tab @samp{5207} @samp{5208}
16170 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16171 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16172 @item @samp{5216} @tab @samp{5214} @samp{5216}
16173 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16174 @item @samp{5225} @tab @samp{5224} @samp{5225}
16175 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16176 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16177 @item @samp{5249} @tab @samp{5249}
16178 @item @samp{5250} @tab @samp{5250}
16179 @item @samp{5271} @tab @samp{5270} @samp{5271}
16180 @item @samp{5272} @tab @samp{5272}
16181 @item @samp{5275} @tab @samp{5274} @samp{5275}
16182 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16183 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16184 @item @samp{5307} @tab @samp{5307}
16185 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16186 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16187 @item @samp{5407} @tab @samp{5407}
16188 @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}
16191 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16192 @var{arch} is compatible with @var{cpu}. Other combinations of
16193 @option{-mcpu} and @option{-march} are rejected.
16195 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16196 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16197 where the value of @var{family} is given by the table above.
16199 @item -mtune=@var{tune}
16201 Tune the code for a particular microarchitecture within the
16202 constraints set by @option{-march} and @option{-mcpu}.
16203 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16204 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16205 and @samp{cpu32}. The ColdFire microarchitectures
16206 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16208 You can also use @option{-mtune=68020-40} for code that needs
16209 to run relatively well on 68020, 68030 and 68040 targets.
16210 @option{-mtune=68020-60} is similar but includes 68060 targets
16211 as well. These two options select the same tuning decisions as
16212 @option{-m68020-40} and @option{-m68020-60} respectively.
16214 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16215 when tuning for 680x0 architecture @var{arch}. It also defines
16216 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16217 option is used. If GCC is tuning for a range of architectures,
16218 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16219 it defines the macros for every architecture in the range.
16221 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16222 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16223 of the arguments given above.
16229 Generate output for a 68000. This is the default
16230 when the compiler is configured for 68000-based systems.
16231 It is equivalent to @option{-march=68000}.
16233 Use this option for microcontrollers with a 68000 or EC000 core,
16234 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16238 Generate output for a 68010. This is the default
16239 when the compiler is configured for 68010-based systems.
16240 It is equivalent to @option{-march=68010}.
16246 Generate output for a 68020. This is the default
16247 when the compiler is configured for 68020-based systems.
16248 It is equivalent to @option{-march=68020}.
16252 Generate output for a 68030. This is the default when the compiler is
16253 configured for 68030-based systems. It is equivalent to
16254 @option{-march=68030}.
16258 Generate output for a 68040. This is the default when the compiler is
16259 configured for 68040-based systems. It is equivalent to
16260 @option{-march=68040}.
16262 This option inhibits the use of 68881/68882 instructions that have to be
16263 emulated by software on the 68040. Use this option if your 68040 does not
16264 have code to emulate those instructions.
16268 Generate output for a 68060. This is the default when the compiler is
16269 configured for 68060-based systems. It is equivalent to
16270 @option{-march=68060}.
16272 This option inhibits the use of 68020 and 68881/68882 instructions that
16273 have to be emulated by software on the 68060. Use this option if your 68060
16274 does not have code to emulate those instructions.
16278 Generate output for a CPU32. This is the default
16279 when the compiler is configured for CPU32-based systems.
16280 It is equivalent to @option{-march=cpu32}.
16282 Use this option for microcontrollers with a
16283 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16284 68336, 68340, 68341, 68349 and 68360.
16288 Generate output for a 520X ColdFire CPU@. This is the default
16289 when the compiler is configured for 520X-based systems.
16290 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16291 in favor of that option.
16293 Use this option for microcontroller with a 5200 core, including
16294 the MCF5202, MCF5203, MCF5204 and MCF5206.
16298 Generate output for a 5206e ColdFire CPU@. The option is now
16299 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16303 Generate output for a member of the ColdFire 528X family.
16304 The option is now deprecated in favor of the equivalent
16305 @option{-mcpu=528x}.
16309 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16310 in favor of the equivalent @option{-mcpu=5307}.
16314 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16315 in favor of the equivalent @option{-mcpu=5407}.
16319 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16320 This includes use of hardware floating-point instructions.
16321 The option is equivalent to @option{-mcpu=547x}, and is now
16322 deprecated in favor of that option.
16326 Generate output for a 68040, without using any of the new instructions.
16327 This results in code that can run relatively efficiently on either a
16328 68020/68881 or a 68030 or a 68040. The generated code does use the
16329 68881 instructions that are emulated on the 68040.
16331 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16335 Generate output for a 68060, without using any of the new instructions.
16336 This results in code that can run relatively efficiently on either a
16337 68020/68881 or a 68030 or a 68040. The generated code does use the
16338 68881 instructions that are emulated on the 68060.
16340 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16344 @opindex mhard-float
16346 Generate floating-point instructions. This is the default for 68020
16347 and above, and for ColdFire devices that have an FPU@. It defines the
16348 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16349 on ColdFire targets.
16352 @opindex msoft-float
16353 Do not generate floating-point instructions; use library calls instead.
16354 This is the default for 68000, 68010, and 68832 targets. It is also
16355 the default for ColdFire devices that have no FPU.
16361 Generate (do not generate) ColdFire hardware divide and remainder
16362 instructions. If @option{-march} is used without @option{-mcpu},
16363 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16364 architectures. Otherwise, the default is taken from the target CPU
16365 (either the default CPU, or the one specified by @option{-mcpu}). For
16366 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16367 @option{-mcpu=5206e}.
16369 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16373 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16374 Additionally, parameters passed on the stack are also aligned to a
16375 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16379 Do not consider type @code{int} to be 16 bits wide. This is the default.
16382 @itemx -mno-bitfield
16383 @opindex mnobitfield
16384 @opindex mno-bitfield
16385 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16386 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16390 Do use the bit-field instructions. The @option{-m68020} option implies
16391 @option{-mbitfield}. This is the default if you use a configuration
16392 designed for a 68020.
16396 Use a different function-calling convention, in which functions
16397 that take a fixed number of arguments return with the @code{rtd}
16398 instruction, which pops their arguments while returning. This
16399 saves one instruction in the caller since there is no need to pop
16400 the arguments there.
16402 This calling convention is incompatible with the one normally
16403 used on Unix, so you cannot use it if you need to call libraries
16404 compiled with the Unix compiler.
16406 Also, you must provide function prototypes for all functions that
16407 take variable numbers of arguments (including @code{printf});
16408 otherwise incorrect code is generated for calls to those
16411 In addition, seriously incorrect code results if you call a
16412 function with too many arguments. (Normally, extra arguments are
16413 harmlessly ignored.)
16415 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16416 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16420 Do not use the calling conventions selected by @option{-mrtd}.
16421 This is the default.
16424 @itemx -mno-align-int
16425 @opindex malign-int
16426 @opindex mno-align-int
16427 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16428 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16429 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16430 Aligning variables on 32-bit boundaries produces code that runs somewhat
16431 faster on processors with 32-bit busses at the expense of more memory.
16433 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16434 aligns structures containing the above types differently than
16435 most published application binary interface specifications for the m68k.
16439 Use the pc-relative addressing mode of the 68000 directly, instead of
16440 using a global offset table. At present, this option implies @option{-fpic},
16441 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16442 not presently supported with @option{-mpcrel}, though this could be supported for
16443 68020 and higher processors.
16445 @item -mno-strict-align
16446 @itemx -mstrict-align
16447 @opindex mno-strict-align
16448 @opindex mstrict-align
16449 Do not (do) assume that unaligned memory references are handled by
16453 Generate code that allows the data segment to be located in a different
16454 area of memory from the text segment. This allows for execute-in-place in
16455 an environment without virtual memory management. This option implies
16458 @item -mno-sep-data
16459 Generate code that assumes that the data segment follows the text segment.
16460 This is the default.
16462 @item -mid-shared-library
16463 Generate code that supports shared libraries via the library ID method.
16464 This allows for execute-in-place and shared libraries in an environment
16465 without virtual memory management. This option implies @option{-fPIC}.
16467 @item -mno-id-shared-library
16468 Generate code that doesn't assume ID-based shared libraries are being used.
16469 This is the default.
16471 @item -mshared-library-id=n
16472 Specifies the identification number of the ID-based shared library being
16473 compiled. Specifying a value of 0 generates more compact code; specifying
16474 other values forces the allocation of that number to the current
16475 library, but is no more space- or time-efficient than omitting this option.
16481 When generating position-independent code for ColdFire, generate code
16482 that works if the GOT has more than 8192 entries. This code is
16483 larger and slower than code generated without this option. On M680x0
16484 processors, this option is not needed; @option{-fPIC} suffices.
16486 GCC normally uses a single instruction to load values from the GOT@.
16487 While this is relatively efficient, it only works if the GOT
16488 is smaller than about 64k. Anything larger causes the linker
16489 to report an error such as:
16491 @cindex relocation truncated to fit (ColdFire)
16493 relocation truncated to fit: R_68K_GOT16O foobar
16496 If this happens, you should recompile your code with @option{-mxgot}.
16497 It should then work with very large GOTs. However, code generated with
16498 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16499 the value of a global symbol.
16501 Note that some linkers, including newer versions of the GNU linker,
16502 can create multiple GOTs and sort GOT entries. If you have such a linker,
16503 you should only need to use @option{-mxgot} when compiling a single
16504 object file that accesses more than 8192 GOT entries. Very few do.
16506 These options have no effect unless GCC is generating
16507 position-independent code.
16511 @node MCore Options
16512 @subsection MCore Options
16513 @cindex MCore options
16515 These are the @samp{-m} options defined for the Motorola M*Core
16521 @itemx -mno-hardlit
16523 @opindex mno-hardlit
16524 Inline constants into the code stream if it can be done in two
16525 instructions or less.
16531 Use the divide instruction. (Enabled by default).
16533 @item -mrelax-immediate
16534 @itemx -mno-relax-immediate
16535 @opindex mrelax-immediate
16536 @opindex mno-relax-immediate
16537 Allow arbitrary-sized immediates in bit operations.
16539 @item -mwide-bitfields
16540 @itemx -mno-wide-bitfields
16541 @opindex mwide-bitfields
16542 @opindex mno-wide-bitfields
16543 Always treat bit-fields as @code{int}-sized.
16545 @item -m4byte-functions
16546 @itemx -mno-4byte-functions
16547 @opindex m4byte-functions
16548 @opindex mno-4byte-functions
16549 Force all functions to be aligned to a 4-byte boundary.
16551 @item -mcallgraph-data
16552 @itemx -mno-callgraph-data
16553 @opindex mcallgraph-data
16554 @opindex mno-callgraph-data
16555 Emit callgraph information.
16558 @itemx -mno-slow-bytes
16559 @opindex mslow-bytes
16560 @opindex mno-slow-bytes
16561 Prefer word access when reading byte quantities.
16563 @item -mlittle-endian
16564 @itemx -mbig-endian
16565 @opindex mlittle-endian
16566 @opindex mbig-endian
16567 Generate code for a little-endian target.
16573 Generate code for the 210 processor.
16577 Assume that runtime support has been provided and so omit the
16578 simulator library (@file{libsim.a)} from the linker command line.
16580 @item -mstack-increment=@var{size}
16581 @opindex mstack-increment
16582 Set the maximum amount for a single stack increment operation. Large
16583 values can increase the speed of programs that contain functions
16584 that need a large amount of stack space, but they can also trigger a
16585 segmentation fault if the stack is extended too much. The default
16591 @subsection MeP Options
16592 @cindex MeP options
16598 Enables the @code{abs} instruction, which is the absolute difference
16599 between two registers.
16603 Enables all the optional instructions---average, multiply, divide, bit
16604 operations, leading zero, absolute difference, min/max, clip, and
16610 Enables the @code{ave} instruction, which computes the average of two
16613 @item -mbased=@var{n}
16615 Variables of size @var{n} bytes or smaller are placed in the
16616 @code{.based} section by default. Based variables use the @code{$tp}
16617 register as a base register, and there is a 128-byte limit to the
16618 @code{.based} section.
16622 Enables the bit operation instructions---bit test (@code{btstm}), set
16623 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16624 test-and-set (@code{tas}).
16626 @item -mc=@var{name}
16628 Selects which section constant data is placed in. @var{name} may
16629 be @samp{tiny}, @samp{near}, or @samp{far}.
16633 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16634 useful unless you also provide @option{-mminmax}.
16636 @item -mconfig=@var{name}
16638 Selects one of the built-in core configurations. Each MeP chip has
16639 one or more modules in it; each module has a core CPU and a variety of
16640 coprocessors, optional instructions, and peripherals. The
16641 @code{MeP-Integrator} tool, not part of GCC, provides these
16642 configurations through this option; using this option is the same as
16643 using all the corresponding command-line options. The default
16644 configuration is @samp{default}.
16648 Enables the coprocessor instructions. By default, this is a 32-bit
16649 coprocessor. Note that the coprocessor is normally enabled via the
16650 @option{-mconfig=} option.
16654 Enables the 32-bit coprocessor's instructions.
16658 Enables the 64-bit coprocessor's instructions.
16662 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16666 Causes constant variables to be placed in the @code{.near} section.
16670 Enables the @code{div} and @code{divu} instructions.
16674 Generate big-endian code.
16678 Generate little-endian code.
16680 @item -mio-volatile
16681 @opindex mio-volatile
16682 Tells the compiler that any variable marked with the @code{io}
16683 attribute is to be considered volatile.
16687 Causes variables to be assigned to the @code{.far} section by default.
16691 Enables the @code{leadz} (leading zero) instruction.
16695 Causes variables to be assigned to the @code{.near} section by default.
16699 Enables the @code{min} and @code{max} instructions.
16703 Enables the multiplication and multiply-accumulate instructions.
16707 Disables all the optional instructions enabled by @option{-mall-opts}.
16711 Enables the @code{repeat} and @code{erepeat} instructions, used for
16712 low-overhead looping.
16716 Causes all variables to default to the @code{.tiny} section. Note
16717 that there is a 65536-byte limit to this section. Accesses to these
16718 variables use the @code{%gp} base register.
16722 Enables the saturation instructions. Note that the compiler does not
16723 currently generate these itself, but this option is included for
16724 compatibility with other tools, like @code{as}.
16728 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16732 Link the simulator run-time libraries.
16736 Link the simulator runtime libraries, excluding built-in support
16737 for reset and exception vectors and tables.
16741 Causes all functions to default to the @code{.far} section. Without
16742 this option, functions default to the @code{.near} section.
16744 @item -mtiny=@var{n}
16746 Variables that are @var{n} bytes or smaller are allocated to the
16747 @code{.tiny} section. These variables use the @code{$gp} base
16748 register. The default for this option is 4, but note that there's a
16749 65536-byte limit to the @code{.tiny} section.
16753 @node MicroBlaze Options
16754 @subsection MicroBlaze Options
16755 @cindex MicroBlaze Options
16760 @opindex msoft-float
16761 Use software emulation for floating point (default).
16764 @opindex mhard-float
16765 Use hardware floating-point instructions.
16769 Do not optimize block moves, use @code{memcpy}.
16771 @item -mno-clearbss
16772 @opindex mno-clearbss
16773 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16775 @item -mcpu=@var{cpu-type}
16777 Use features of, and schedule code for, the given CPU.
16778 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16779 where @var{X} is a major version, @var{YY} is the minor version, and
16780 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16781 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16783 @item -mxl-soft-mul
16784 @opindex mxl-soft-mul
16785 Use software multiply emulation (default).
16787 @item -mxl-soft-div
16788 @opindex mxl-soft-div
16789 Use software emulation for divides (default).
16791 @item -mxl-barrel-shift
16792 @opindex mxl-barrel-shift
16793 Use the hardware barrel shifter.
16795 @item -mxl-pattern-compare
16796 @opindex mxl-pattern-compare
16797 Use pattern compare instructions.
16799 @item -msmall-divides
16800 @opindex msmall-divides
16801 Use table lookup optimization for small signed integer divisions.
16803 @item -mxl-stack-check
16804 @opindex mxl-stack-check
16805 This option is deprecated. Use @option{-fstack-check} instead.
16808 @opindex mxl-gp-opt
16809 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16811 @item -mxl-multiply-high
16812 @opindex mxl-multiply-high
16813 Use multiply high instructions for high part of 32x32 multiply.
16815 @item -mxl-float-convert
16816 @opindex mxl-float-convert
16817 Use hardware floating-point conversion instructions.
16819 @item -mxl-float-sqrt
16820 @opindex mxl-float-sqrt
16821 Use hardware floating-point square root instruction.
16824 @opindex mbig-endian
16825 Generate code for a big-endian target.
16827 @item -mlittle-endian
16828 @opindex mlittle-endian
16829 Generate code for a little-endian target.
16832 @opindex mxl-reorder
16833 Use reorder instructions (swap and byte reversed load/store).
16835 @item -mxl-mode-@var{app-model}
16836 Select application model @var{app-model}. Valid models are
16839 normal executable (default), uses startup code @file{crt0.o}.
16842 for use with Xilinx Microprocessor Debugger (XMD) based
16843 software intrusive debug agent called xmdstub. This uses startup file
16844 @file{crt1.o} and sets the start address of the program to 0x800.
16847 for applications that are loaded using a bootloader.
16848 This model uses startup file @file{crt2.o} which does not contain a processor
16849 reset vector handler. This is suitable for transferring control on a
16850 processor reset to the bootloader rather than the application.
16853 for applications that do not require any of the
16854 MicroBlaze vectors. This option may be useful for applications running
16855 within a monitoring application. This model uses @file{crt3.o} as a startup file.
16858 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
16859 @option{-mxl-mode-@var{app-model}}.
16864 @subsection MIPS Options
16865 @cindex MIPS options
16871 Generate big-endian code.
16875 Generate little-endian code. This is the default for @samp{mips*el-*-*}
16878 @item -march=@var{arch}
16880 Generate code that runs on @var{arch}, which can be the name of a
16881 generic MIPS ISA, or the name of a particular processor.
16883 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
16884 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
16885 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
16886 @samp{mips64r5} and @samp{mips64r6}.
16887 The processor names are:
16888 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
16889 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
16890 @samp{5kc}, @samp{5kf},
16892 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
16893 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
16894 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
16895 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
16896 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
16897 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
16899 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
16900 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
16903 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
16904 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
16905 @samp{rm7000}, @samp{rm9000},
16906 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
16909 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
16910 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
16911 @samp{xlr} and @samp{xlp}.
16912 The special value @samp{from-abi} selects the
16913 most compatible architecture for the selected ABI (that is,
16914 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
16916 The native Linux/GNU toolchain also supports the value @samp{native},
16917 which selects the best architecture option for the host processor.
16918 @option{-march=native} has no effect if GCC does not recognize
16921 In processor names, a final @samp{000} can be abbreviated as @samp{k}
16922 (for example, @option{-march=r2k}). Prefixes are optional, and
16923 @samp{vr} may be written @samp{r}.
16925 Names of the form @samp{@var{n}f2_1} refer to processors with
16926 FPUs clocked at half the rate of the core, names of the form
16927 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
16928 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
16929 processors with FPUs clocked a ratio of 3:2 with respect to the core.
16930 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
16931 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
16932 accepted as synonyms for @samp{@var{n}f1_1}.
16934 GCC defines two macros based on the value of this option. The first
16935 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
16936 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
16937 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
16938 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
16939 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
16941 Note that the @code{_MIPS_ARCH} macro uses the processor names given
16942 above. In other words, it has the full prefix and does not
16943 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
16944 the macro names the resolved architecture (either @code{"mips1"} or
16945 @code{"mips3"}). It names the default architecture when no
16946 @option{-march} option is given.
16948 @item -mtune=@var{arch}
16950 Optimize for @var{arch}. Among other things, this option controls
16951 the way instructions are scheduled, and the perceived cost of arithmetic
16952 operations. The list of @var{arch} values is the same as for
16955 When this option is not used, GCC optimizes for the processor
16956 specified by @option{-march}. By using @option{-march} and
16957 @option{-mtune} together, it is possible to generate code that
16958 runs on a family of processors, but optimize the code for one
16959 particular member of that family.
16961 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
16962 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
16963 @option{-march} ones described above.
16967 Equivalent to @option{-march=mips1}.
16971 Equivalent to @option{-march=mips2}.
16975 Equivalent to @option{-march=mips3}.
16979 Equivalent to @option{-march=mips4}.
16983 Equivalent to @option{-march=mips32}.
16987 Equivalent to @option{-march=mips32r3}.
16991 Equivalent to @option{-march=mips32r5}.
16995 Equivalent to @option{-march=mips32r6}.
16999 Equivalent to @option{-march=mips64}.
17003 Equivalent to @option{-march=mips64r2}.
17007 Equivalent to @option{-march=mips64r3}.
17011 Equivalent to @option{-march=mips64r5}.
17015 Equivalent to @option{-march=mips64r6}.
17020 @opindex mno-mips16
17021 Generate (do not generate) MIPS16 code. If GCC is targeting a
17022 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17024 MIPS16 code generation can also be controlled on a per-function basis
17025 by means of @code{mips16} and @code{nomips16} attributes.
17026 @xref{Function Attributes}, for more information.
17028 @item -mflip-mips16
17029 @opindex mflip-mips16
17030 Generate MIPS16 code on alternating functions. This option is provided
17031 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17032 not intended for ordinary use in compiling user code.
17034 @item -minterlink-compressed
17035 @item -mno-interlink-compressed
17036 @opindex minterlink-compressed
17037 @opindex mno-interlink-compressed
17038 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17039 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17041 For example, code using the standard ISA encoding cannot jump directly
17042 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17043 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17044 knows that the target of the jump is not compressed.
17046 @item -minterlink-mips16
17047 @itemx -mno-interlink-mips16
17048 @opindex minterlink-mips16
17049 @opindex mno-interlink-mips16
17050 Aliases of @option{-minterlink-compressed} and
17051 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17052 and are retained for backwards compatibility.
17064 Generate code for the given ABI@.
17066 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17067 generates 64-bit code when you select a 64-bit architecture, but you
17068 can use @option{-mgp32} to get 32-bit code instead.
17070 For information about the O64 ABI, see
17071 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17073 GCC supports a variant of the o32 ABI in which floating-point registers
17074 are 64 rather than 32 bits wide. You can select this combination with
17075 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17076 and @code{mfhc1} instructions and is therefore only supported for
17077 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17079 The register assignments for arguments and return values remain the
17080 same, but each scalar value is passed in a single 64-bit register
17081 rather than a pair of 32-bit registers. For example, scalar
17082 floating-point values are returned in @samp{$f0} only, not a
17083 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17084 remains the same in that the even-numbered double-precision registers
17087 Two additional variants of the o32 ABI are supported to enable
17088 a transition from 32-bit to 64-bit registers. These are FPXX
17089 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17090 The FPXX extension mandates that all code must execute correctly
17091 when run using 32-bit or 64-bit registers. The code can be interlinked
17092 with either FP32 or FP64, but not both.
17093 The FP64A extension is similar to the FP64 extension but forbids the
17094 use of odd-numbered single-precision registers. This can be used
17095 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17096 processors and allows both FP32 and FP64A code to interlink and
17097 run in the same process without changing FPU modes.
17100 @itemx -mno-abicalls
17102 @opindex mno-abicalls
17103 Generate (do not generate) code that is suitable for SVR4-style
17104 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17109 Generate (do not generate) code that is fully position-independent,
17110 and that can therefore be linked into shared libraries. This option
17111 only affects @option{-mabicalls}.
17113 All @option{-mabicalls} code has traditionally been position-independent,
17114 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17115 as an extension, the GNU toolchain allows executables to use absolute
17116 accesses for locally-binding symbols. It can also use shorter GP
17117 initialization sequences and generate direct calls to locally-defined
17118 functions. This mode is selected by @option{-mno-shared}.
17120 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17121 objects that can only be linked by the GNU linker. However, the option
17122 does not affect the ABI of the final executable; it only affects the ABI
17123 of relocatable objects. Using @option{-mno-shared} generally makes
17124 executables both smaller and quicker.
17126 @option{-mshared} is the default.
17132 Assume (do not assume) that the static and dynamic linkers
17133 support PLTs and copy relocations. This option only affects
17134 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17135 has no effect without @option{-msym32}.
17137 You can make @option{-mplt} the default by configuring
17138 GCC with @option{--with-mips-plt}. The default is
17139 @option{-mno-plt} otherwise.
17145 Lift (do not lift) the usual restrictions on the size of the global
17148 GCC normally uses a single instruction to load values from the GOT@.
17149 While this is relatively efficient, it only works if the GOT
17150 is smaller than about 64k. Anything larger causes the linker
17151 to report an error such as:
17153 @cindex relocation truncated to fit (MIPS)
17155 relocation truncated to fit: R_MIPS_GOT16 foobar
17158 If this happens, you should recompile your code with @option{-mxgot}.
17159 This works with very large GOTs, although the code is also
17160 less efficient, since it takes three instructions to fetch the
17161 value of a global symbol.
17163 Note that some linkers can create multiple GOTs. If you have such a
17164 linker, you should only need to use @option{-mxgot} when a single object
17165 file accesses more than 64k's worth of GOT entries. Very few do.
17167 These options have no effect unless GCC is generating position
17172 Assume that general-purpose registers are 32 bits wide.
17176 Assume that general-purpose registers are 64 bits wide.
17180 Assume that floating-point registers are 32 bits wide.
17184 Assume that floating-point registers are 64 bits wide.
17188 Do not assume the width of floating-point registers.
17191 @opindex mhard-float
17192 Use floating-point coprocessor instructions.
17195 @opindex msoft-float
17196 Do not use floating-point coprocessor instructions. Implement
17197 floating-point calculations using library calls instead.
17201 Equivalent to @option{-msoft-float}, but additionally asserts that the
17202 program being compiled does not perform any floating-point operations.
17203 This option is presently supported only by some bare-metal MIPS
17204 configurations, where it may select a special set of libraries
17205 that lack all floating-point support (including, for example, the
17206 floating-point @code{printf} formats).
17207 If code compiled with @option{-mno-float} accidentally contains
17208 floating-point operations, it is likely to suffer a link-time
17209 or run-time failure.
17211 @item -msingle-float
17212 @opindex msingle-float
17213 Assume that the floating-point coprocessor only supports single-precision
17216 @item -mdouble-float
17217 @opindex mdouble-float
17218 Assume that the floating-point coprocessor supports double-precision
17219 operations. This is the default.
17222 @itemx -mno-odd-spreg
17223 @opindex modd-spreg
17224 @opindex mno-odd-spreg
17225 Enable the use of odd-numbered single-precision floating-point registers
17226 for the o32 ABI. This is the default for processors that are known to
17227 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17231 @itemx -mabs=legacy
17233 @opindex mabs=legacy
17234 These options control the treatment of the special not-a-number (NaN)
17235 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17236 @code{neg.@i{fmt}} machine instructions.
17238 By default or when @option{-mabs=legacy} is used the legacy
17239 treatment is selected. In this case these instructions are considered
17240 arithmetic and avoided where correct operation is required and the
17241 input operand might be a NaN. A longer sequence of instructions that
17242 manipulate the sign bit of floating-point datum manually is used
17243 instead unless the @option{-ffinite-math-only} option has also been
17246 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17247 this case these instructions are considered non-arithmetic and therefore
17248 operating correctly in all cases, including in particular where the
17249 input operand is a NaN. These instructions are therefore always used
17250 for the respective operations.
17253 @itemx -mnan=legacy
17255 @opindex mnan=legacy
17256 These options control the encoding of the special not-a-number (NaN)
17257 IEEE 754 floating-point data.
17259 The @option{-mnan=legacy} option selects the legacy encoding. In this
17260 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17261 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17262 by the first bit of their trailing significand field being 1.
17264 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17265 this case qNaNs are denoted by the first bit of their trailing
17266 significand field being 1, whereas sNaNs are denoted by the first bit of
17267 their trailing significand field being 0.
17269 The default is @option{-mnan=legacy} unless GCC has been configured with
17270 @option{--with-nan=2008}.
17276 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17277 implement atomic memory built-in functions. When neither option is
17278 specified, GCC uses the instructions if the target architecture
17281 @option{-mllsc} is useful if the runtime environment can emulate the
17282 instructions and @option{-mno-llsc} can be useful when compiling for
17283 nonstandard ISAs. You can make either option the default by
17284 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17285 respectively. @option{--with-llsc} is the default for some
17286 configurations; see the installation documentation for details.
17292 Use (do not use) revision 1 of the MIPS DSP ASE@.
17293 @xref{MIPS DSP Built-in Functions}. This option defines the
17294 preprocessor macro @code{__mips_dsp}. It also defines
17295 @code{__mips_dsp_rev} to 1.
17301 Use (do not use) revision 2 of the MIPS DSP ASE@.
17302 @xref{MIPS DSP Built-in Functions}. This option defines the
17303 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17304 It also defines @code{__mips_dsp_rev} to 2.
17307 @itemx -mno-smartmips
17308 @opindex msmartmips
17309 @opindex mno-smartmips
17310 Use (do not use) the MIPS SmartMIPS ASE.
17312 @item -mpaired-single
17313 @itemx -mno-paired-single
17314 @opindex mpaired-single
17315 @opindex mno-paired-single
17316 Use (do not use) paired-single floating-point instructions.
17317 @xref{MIPS Paired-Single Support}. This option requires
17318 hardware floating-point support to be enabled.
17324 Use (do not use) MIPS Digital Media Extension instructions.
17325 This option can only be used when generating 64-bit code and requires
17326 hardware floating-point support to be enabled.
17331 @opindex mno-mips3d
17332 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17333 The option @option{-mips3d} implies @option{-mpaired-single}.
17336 @itemx -mno-micromips
17337 @opindex mmicromips
17338 @opindex mno-mmicromips
17339 Generate (do not generate) microMIPS code.
17341 MicroMIPS code generation can also be controlled on a per-function basis
17342 by means of @code{micromips} and @code{nomicromips} attributes.
17343 @xref{Function Attributes}, for more information.
17349 Use (do not use) MT Multithreading instructions.
17355 Use (do not use) the MIPS MCU ASE instructions.
17361 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17367 Use (do not use) the MIPS Virtualization Application Specific instructions.
17373 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17377 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17378 an explanation of the default and the way that the pointer size is
17383 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17385 The default size of @code{int}s, @code{long}s and pointers depends on
17386 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17387 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17388 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17389 or the same size as integer registers, whichever is smaller.
17395 Assume (do not assume) that all symbols have 32-bit values, regardless
17396 of the selected ABI@. This option is useful in combination with
17397 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17398 to generate shorter and faster references to symbolic addresses.
17402 Put definitions of externally-visible data in a small data section
17403 if that data is no bigger than @var{num} bytes. GCC can then generate
17404 more efficient accesses to the data; see @option{-mgpopt} for details.
17406 The default @option{-G} option depends on the configuration.
17408 @item -mlocal-sdata
17409 @itemx -mno-local-sdata
17410 @opindex mlocal-sdata
17411 @opindex mno-local-sdata
17412 Extend (do not extend) the @option{-G} behavior to local data too,
17413 such as to static variables in C@. @option{-mlocal-sdata} is the
17414 default for all configurations.
17416 If the linker complains that an application is using too much small data,
17417 you might want to try rebuilding the less performance-critical parts with
17418 @option{-mno-local-sdata}. You might also want to build large
17419 libraries with @option{-mno-local-sdata}, so that the libraries leave
17420 more room for the main program.
17422 @item -mextern-sdata
17423 @itemx -mno-extern-sdata
17424 @opindex mextern-sdata
17425 @opindex mno-extern-sdata
17426 Assume (do not assume) that externally-defined data is in
17427 a small data section if the size of that data is within the @option{-G} limit.
17428 @option{-mextern-sdata} is the default for all configurations.
17430 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17431 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17432 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17433 is placed in a small data section. If @var{Var} is defined by another
17434 module, you must either compile that module with a high-enough
17435 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17436 definition. If @var{Var} is common, you must link the application
17437 with a high-enough @option{-G} setting.
17439 The easiest way of satisfying these restrictions is to compile
17440 and link every module with the same @option{-G} option. However,
17441 you may wish to build a library that supports several different
17442 small data limits. You can do this by compiling the library with
17443 the highest supported @option{-G} setting and additionally using
17444 @option{-mno-extern-sdata} to stop the library from making assumptions
17445 about externally-defined data.
17451 Use (do not use) GP-relative accesses for symbols that are known to be
17452 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17453 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17456 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17457 might not hold the value of @code{_gp}. For example, if the code is
17458 part of a library that might be used in a boot monitor, programs that
17459 call boot monitor routines pass an unknown value in @code{$gp}.
17460 (In such situations, the boot monitor itself is usually compiled
17461 with @option{-G0}.)
17463 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17464 @option{-mno-extern-sdata}.
17466 @item -membedded-data
17467 @itemx -mno-embedded-data
17468 @opindex membedded-data
17469 @opindex mno-embedded-data
17470 Allocate variables to the read-only data section first if possible, then
17471 next in the small data section if possible, otherwise in data. This gives
17472 slightly slower code than the default, but reduces the amount of RAM required
17473 when executing, and thus may be preferred for some embedded systems.
17475 @item -muninit-const-in-rodata
17476 @itemx -mno-uninit-const-in-rodata
17477 @opindex muninit-const-in-rodata
17478 @opindex mno-uninit-const-in-rodata
17479 Put uninitialized @code{const} variables in the read-only data section.
17480 This option is only meaningful in conjunction with @option{-membedded-data}.
17482 @item -mcode-readable=@var{setting}
17483 @opindex mcode-readable
17484 Specify whether GCC may generate code that reads from executable sections.
17485 There are three possible settings:
17488 @item -mcode-readable=yes
17489 Instructions may freely access executable sections. This is the
17492 @item -mcode-readable=pcrel
17493 MIPS16 PC-relative load instructions can access executable sections,
17494 but other instructions must not do so. This option is useful on 4KSc
17495 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17496 It is also useful on processors that can be configured to have a dual
17497 instruction/data SRAM interface and that, like the M4K, automatically
17498 redirect PC-relative loads to the instruction RAM.
17500 @item -mcode-readable=no
17501 Instructions must not access executable sections. This option can be
17502 useful on targets that are configured to have a dual instruction/data
17503 SRAM interface but that (unlike the M4K) do not automatically redirect
17504 PC-relative loads to the instruction RAM.
17507 @item -msplit-addresses
17508 @itemx -mno-split-addresses
17509 @opindex msplit-addresses
17510 @opindex mno-split-addresses
17511 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17512 relocation operators. This option has been superseded by
17513 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17515 @item -mexplicit-relocs
17516 @itemx -mno-explicit-relocs
17517 @opindex mexplicit-relocs
17518 @opindex mno-explicit-relocs
17519 Use (do not use) assembler relocation operators when dealing with symbolic
17520 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17521 is to use assembler macros instead.
17523 @option{-mexplicit-relocs} is the default if GCC was configured
17524 to use an assembler that supports relocation operators.
17526 @item -mcheck-zero-division
17527 @itemx -mno-check-zero-division
17528 @opindex mcheck-zero-division
17529 @opindex mno-check-zero-division
17530 Trap (do not trap) on integer division by zero.
17532 The default is @option{-mcheck-zero-division}.
17534 @item -mdivide-traps
17535 @itemx -mdivide-breaks
17536 @opindex mdivide-traps
17537 @opindex mdivide-breaks
17538 MIPS systems check for division by zero by generating either a
17539 conditional trap or a break instruction. Using traps results in
17540 smaller code, but is only supported on MIPS II and later. Also, some
17541 versions of the Linux kernel have a bug that prevents trap from
17542 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17543 allow conditional traps on architectures that support them and
17544 @option{-mdivide-breaks} to force the use of breaks.
17546 The default is usually @option{-mdivide-traps}, but this can be
17547 overridden at configure time using @option{--with-divide=breaks}.
17548 Divide-by-zero checks can be completely disabled using
17549 @option{-mno-check-zero-division}.
17554 @opindex mno-memcpy
17555 Force (do not force) the use of @code{memcpy} for non-trivial block
17556 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17557 most constant-sized copies.
17560 @itemx -mno-long-calls
17561 @opindex mlong-calls
17562 @opindex mno-long-calls
17563 Disable (do not disable) use of the @code{jal} instruction. Calling
17564 functions using @code{jal} is more efficient but requires the caller
17565 and callee to be in the same 256 megabyte segment.
17567 This option has no effect on abicalls code. The default is
17568 @option{-mno-long-calls}.
17574 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17575 instructions, as provided by the R4650 ISA@.
17581 Enable (disable) use of the @code{madd} and @code{msub} integer
17582 instructions. The default is @option{-mimadd} on architectures
17583 that support @code{madd} and @code{msub} except for the 74k
17584 architecture where it was found to generate slower code.
17587 @itemx -mno-fused-madd
17588 @opindex mfused-madd
17589 @opindex mno-fused-madd
17590 Enable (disable) use of the floating-point multiply-accumulate
17591 instructions, when they are available. The default is
17592 @option{-mfused-madd}.
17594 On the R8000 CPU when multiply-accumulate instructions are used,
17595 the intermediate product is calculated to infinite precision
17596 and is not subject to the FCSR Flush to Zero bit. This may be
17597 undesirable in some circumstances. On other processors the result
17598 is numerically identical to the equivalent computation using
17599 separate multiply, add, subtract and negate instructions.
17603 Tell the MIPS assembler to not run its preprocessor over user
17604 assembler files (with a @samp{.s} suffix) when assembling them.
17609 @opindex mno-fix-24k
17610 Work around the 24K E48 (lost data on stores during refill) errata.
17611 The workarounds are implemented by the assembler rather than by GCC@.
17614 @itemx -mno-fix-r4000
17615 @opindex mfix-r4000
17616 @opindex mno-fix-r4000
17617 Work around certain R4000 CPU errata:
17620 A double-word or a variable shift may give an incorrect result if executed
17621 immediately after starting an integer division.
17623 A double-word or a variable shift may give an incorrect result if executed
17624 while an integer multiplication is in progress.
17626 An integer division may give an incorrect result if started in a delay slot
17627 of a taken branch or a jump.
17631 @itemx -mno-fix-r4400
17632 @opindex mfix-r4400
17633 @opindex mno-fix-r4400
17634 Work around certain R4400 CPU errata:
17637 A double-word or a variable shift may give an incorrect result if executed
17638 immediately after starting an integer division.
17642 @itemx -mno-fix-r10000
17643 @opindex mfix-r10000
17644 @opindex mno-fix-r10000
17645 Work around certain R10000 errata:
17648 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17649 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17652 This option can only be used if the target architecture supports
17653 branch-likely instructions. @option{-mfix-r10000} is the default when
17654 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17658 @itemx -mno-fix-rm7000
17659 @opindex mfix-rm7000
17660 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17661 workarounds are implemented by the assembler rather than by GCC@.
17664 @itemx -mno-fix-vr4120
17665 @opindex mfix-vr4120
17666 Work around certain VR4120 errata:
17669 @code{dmultu} does not always produce the correct result.
17671 @code{div} and @code{ddiv} do not always produce the correct result if one
17672 of the operands is negative.
17674 The workarounds for the division errata rely on special functions in
17675 @file{libgcc.a}. At present, these functions are only provided by
17676 the @code{mips64vr*-elf} configurations.
17678 Other VR4120 errata require a NOP to be inserted between certain pairs of
17679 instructions. These errata are handled by the assembler, not by GCC itself.
17682 @opindex mfix-vr4130
17683 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17684 workarounds are implemented by the assembler rather than by GCC,
17685 although GCC avoids using @code{mflo} and @code{mfhi} if the
17686 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17687 instructions are available instead.
17690 @itemx -mno-fix-sb1
17692 Work around certain SB-1 CPU core errata.
17693 (This flag currently works around the SB-1 revision 2
17694 ``F1'' and ``F2'' floating-point errata.)
17696 @item -mr10k-cache-barrier=@var{setting}
17697 @opindex mr10k-cache-barrier
17698 Specify whether GCC should insert cache barriers to avoid the
17699 side-effects of speculation on R10K processors.
17701 In common with many processors, the R10K tries to predict the outcome
17702 of a conditional branch and speculatively executes instructions from
17703 the ``taken'' branch. It later aborts these instructions if the
17704 predicted outcome is wrong. However, on the R10K, even aborted
17705 instructions can have side effects.
17707 This problem only affects kernel stores and, depending on the system,
17708 kernel loads. As an example, a speculatively-executed store may load
17709 the target memory into cache and mark the cache line as dirty, even if
17710 the store itself is later aborted. If a DMA operation writes to the
17711 same area of memory before the ``dirty'' line is flushed, the cached
17712 data overwrites the DMA-ed data. See the R10K processor manual
17713 for a full description, including other potential problems.
17715 One workaround is to insert cache barrier instructions before every memory
17716 access that might be speculatively executed and that might have side
17717 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17718 controls GCC's implementation of this workaround. It assumes that
17719 aborted accesses to any byte in the following regions does not have
17724 the memory occupied by the current function's stack frame;
17727 the memory occupied by an incoming stack argument;
17730 the memory occupied by an object with a link-time-constant address.
17733 It is the kernel's responsibility to ensure that speculative
17734 accesses to these regions are indeed safe.
17736 If the input program contains a function declaration such as:
17742 then the implementation of @code{foo} must allow @code{j foo} and
17743 @code{jal foo} to be executed speculatively. GCC honors this
17744 restriction for functions it compiles itself. It expects non-GCC
17745 functions (such as hand-written assembly code) to do the same.
17747 The option has three forms:
17750 @item -mr10k-cache-barrier=load-store
17751 Insert a cache barrier before a load or store that might be
17752 speculatively executed and that might have side effects even
17755 @item -mr10k-cache-barrier=store
17756 Insert a cache barrier before a store that might be speculatively
17757 executed and that might have side effects even if aborted.
17759 @item -mr10k-cache-barrier=none
17760 Disable the insertion of cache barriers. This is the default setting.
17763 @item -mflush-func=@var{func}
17764 @itemx -mno-flush-func
17765 @opindex mflush-func
17766 Specifies the function to call to flush the I and D caches, or to not
17767 call any such function. If called, the function must take the same
17768 arguments as the common @code{_flush_func}, that is, the address of the
17769 memory range for which the cache is being flushed, the size of the
17770 memory range, and the number 3 (to flush both caches). The default
17771 depends on the target GCC was configured for, but commonly is either
17772 @code{_flush_func} or @code{__cpu_flush}.
17774 @item mbranch-cost=@var{num}
17775 @opindex mbranch-cost
17776 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17777 This cost is only a heuristic and is not guaranteed to produce
17778 consistent results across releases. A zero cost redundantly selects
17779 the default, which is based on the @option{-mtune} setting.
17781 @item -mbranch-likely
17782 @itemx -mno-branch-likely
17783 @opindex mbranch-likely
17784 @opindex mno-branch-likely
17785 Enable or disable use of Branch Likely instructions, regardless of the
17786 default for the selected architecture. By default, Branch Likely
17787 instructions may be generated if they are supported by the selected
17788 architecture. An exception is for the MIPS32 and MIPS64 architectures
17789 and processors that implement those architectures; for those, Branch
17790 Likely instructions are not be generated by default because the MIPS32
17791 and MIPS64 architectures specifically deprecate their use.
17793 @item -mfp-exceptions
17794 @itemx -mno-fp-exceptions
17795 @opindex mfp-exceptions
17796 Specifies whether FP exceptions are enabled. This affects how
17797 FP instructions are scheduled for some processors.
17798 The default is that FP exceptions are
17801 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17802 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17805 @item -mvr4130-align
17806 @itemx -mno-vr4130-align
17807 @opindex mvr4130-align
17808 The VR4130 pipeline is two-way superscalar, but can only issue two
17809 instructions together if the first one is 8-byte aligned. When this
17810 option is enabled, GCC aligns pairs of instructions that it
17811 thinks should execute in parallel.
17813 This option only has an effect when optimizing for the VR4130.
17814 It normally makes code faster, but at the expense of making it bigger.
17815 It is enabled by default at optimization level @option{-O3}.
17820 Enable (disable) generation of @code{synci} instructions on
17821 architectures that support it. The @code{synci} instructions (if
17822 enabled) are generated when @code{__builtin___clear_cache} is
17825 This option defaults to @option{-mno-synci}, but the default can be
17826 overridden by configuring GCC with @option{--with-synci}.
17828 When compiling code for single processor systems, it is generally safe
17829 to use @code{synci}. However, on many multi-core (SMP) systems, it
17830 does not invalidate the instruction caches on all cores and may lead
17831 to undefined behavior.
17833 @item -mrelax-pic-calls
17834 @itemx -mno-relax-pic-calls
17835 @opindex mrelax-pic-calls
17836 Try to turn PIC calls that are normally dispatched via register
17837 @code{$25} into direct calls. This is only possible if the linker can
17838 resolve the destination at link-time and if the destination is within
17839 range for a direct call.
17841 @option{-mrelax-pic-calls} is the default if GCC was configured to use
17842 an assembler and a linker that support the @code{.reloc} assembly
17843 directive and @option{-mexplicit-relocs} is in effect. With
17844 @option{-mno-explicit-relocs}, this optimization can be performed by the
17845 assembler and the linker alone without help from the compiler.
17847 @item -mmcount-ra-address
17848 @itemx -mno-mcount-ra-address
17849 @opindex mmcount-ra-address
17850 @opindex mno-mcount-ra-address
17851 Emit (do not emit) code that allows @code{_mcount} to modify the
17852 calling function's return address. When enabled, this option extends
17853 the usual @code{_mcount} interface with a new @var{ra-address}
17854 parameter, which has type @code{intptr_t *} and is passed in register
17855 @code{$12}. @code{_mcount} can then modify the return address by
17856 doing both of the following:
17859 Returning the new address in register @code{$31}.
17861 Storing the new address in @code{*@var{ra-address}},
17862 if @var{ra-address} is nonnull.
17865 The default is @option{-mno-mcount-ra-address}.
17870 @subsection MMIX Options
17871 @cindex MMIX Options
17873 These options are defined for the MMIX:
17877 @itemx -mno-libfuncs
17879 @opindex mno-libfuncs
17880 Specify that intrinsic library functions are being compiled, passing all
17881 values in registers, no matter the size.
17884 @itemx -mno-epsilon
17886 @opindex mno-epsilon
17887 Generate floating-point comparison instructions that compare with respect
17888 to the @code{rE} epsilon register.
17890 @item -mabi=mmixware
17892 @opindex mabi=mmixware
17894 Generate code that passes function parameters and return values that (in
17895 the called function) are seen as registers @code{$0} and up, as opposed to
17896 the GNU ABI which uses global registers @code{$231} and up.
17898 @item -mzero-extend
17899 @itemx -mno-zero-extend
17900 @opindex mzero-extend
17901 @opindex mno-zero-extend
17902 When reading data from memory in sizes shorter than 64 bits, use (do not
17903 use) zero-extending load instructions by default, rather than
17904 sign-extending ones.
17907 @itemx -mno-knuthdiv
17909 @opindex mno-knuthdiv
17910 Make the result of a division yielding a remainder have the same sign as
17911 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
17912 remainder follows the sign of the dividend. Both methods are
17913 arithmetically valid, the latter being almost exclusively used.
17915 @item -mtoplevel-symbols
17916 @itemx -mno-toplevel-symbols
17917 @opindex mtoplevel-symbols
17918 @opindex mno-toplevel-symbols
17919 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
17920 code can be used with the @code{PREFIX} assembly directive.
17924 Generate an executable in the ELF format, rather than the default
17925 @samp{mmo} format used by the @command{mmix} simulator.
17927 @item -mbranch-predict
17928 @itemx -mno-branch-predict
17929 @opindex mbranch-predict
17930 @opindex mno-branch-predict
17931 Use (do not use) the probable-branch instructions, when static branch
17932 prediction indicates a probable branch.
17934 @item -mbase-addresses
17935 @itemx -mno-base-addresses
17936 @opindex mbase-addresses
17937 @opindex mno-base-addresses
17938 Generate (do not generate) code that uses @emph{base addresses}. Using a
17939 base address automatically generates a request (handled by the assembler
17940 and the linker) for a constant to be set up in a global register. The
17941 register is used for one or more base address requests within the range 0
17942 to 255 from the value held in the register. The generally leads to short
17943 and fast code, but the number of different data items that can be
17944 addressed is limited. This means that a program that uses lots of static
17945 data may require @option{-mno-base-addresses}.
17947 @item -msingle-exit
17948 @itemx -mno-single-exit
17949 @opindex msingle-exit
17950 @opindex mno-single-exit
17951 Force (do not force) generated code to have a single exit point in each
17955 @node MN10300 Options
17956 @subsection MN10300 Options
17957 @cindex MN10300 options
17959 These @option{-m} options are defined for Matsushita MN10300 architectures:
17964 Generate code to avoid bugs in the multiply instructions for the MN10300
17965 processors. This is the default.
17967 @item -mno-mult-bug
17968 @opindex mno-mult-bug
17969 Do not generate code to avoid bugs in the multiply instructions for the
17970 MN10300 processors.
17974 Generate code using features specific to the AM33 processor.
17978 Do not generate code using features specific to the AM33 processor. This
17983 Generate code using features specific to the AM33/2.0 processor.
17987 Generate code using features specific to the AM34 processor.
17989 @item -mtune=@var{cpu-type}
17991 Use the timing characteristics of the indicated CPU type when
17992 scheduling instructions. This does not change the targeted processor
17993 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
17994 @samp{am33-2} or @samp{am34}.
17996 @item -mreturn-pointer-on-d0
17997 @opindex mreturn-pointer-on-d0
17998 When generating a function that returns a pointer, return the pointer
17999 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18000 only in @code{a0}, and attempts to call such functions without a prototype
18001 result in errors. Note that this option is on by default; use
18002 @option{-mno-return-pointer-on-d0} to disable it.
18006 Do not link in the C run-time initialization object file.
18010 Indicate to the linker that it should perform a relaxation optimization pass
18011 to shorten branches, calls and absolute memory addresses. This option only
18012 has an effect when used on the command line for the final link step.
18014 This option makes symbolic debugging impossible.
18018 Allow the compiler to generate @emph{Long Instruction Word}
18019 instructions if the target is the @samp{AM33} or later. This is the
18020 default. This option defines the preprocessor macro @code{__LIW__}.
18024 Do not allow the compiler to generate @emph{Long Instruction Word}
18025 instructions. This option defines the preprocessor macro
18030 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18031 instructions if the target is the @samp{AM33} or later. This is the
18032 default. This option defines the preprocessor macro @code{__SETLB__}.
18036 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18037 instructions. This option defines the preprocessor macro
18038 @code{__NO_SETLB__}.
18042 @node Moxie Options
18043 @subsection Moxie Options
18044 @cindex Moxie Options
18050 Generate big-endian code. This is the default for @samp{moxie-*-*}
18055 Generate little-endian code.
18059 Generate mul.x and umul.x instructions. This is the default for
18060 @samp{moxiebox-*-*} configurations.
18064 Do not link in the C run-time initialization object file.
18068 @node MSP430 Options
18069 @subsection MSP430 Options
18070 @cindex MSP430 Options
18072 These options are defined for the MSP430:
18078 Force assembly output to always use hex constants. Normally such
18079 constants are signed decimals, but this option is available for
18080 testsuite and/or aesthetic purposes.
18084 Select the MCU to target. This is used to create a C preprocessor
18085 symbol based upon the MCU name, converted to upper case and pre- and
18086 post-fixed with @samp{__}. This in turn is used by the
18087 @file{msp430.h} header file to select an MCU-specific supplementary
18090 The option also sets the ISA to use. If the MCU name is one that is
18091 known to only support the 430 ISA then that is selected, otherwise the
18092 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18093 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18094 name selects the 430X ISA.
18096 In addition an MCU-specific linker script is added to the linker
18097 command line. The script's name is the name of the MCU with
18098 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18099 command line defines the C preprocessor symbol @code{__XXX__} and
18100 cause the linker to search for a script called @file{xxx.ld}.
18102 This option is also passed on to the assembler.
18106 Specifies the ISA to use. Accepted values are @samp{msp430},
18107 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18108 @option{-mmcu=} option should be used to select the ISA.
18112 Link to the simulator runtime libraries and linker script. Overrides
18113 any scripts that would be selected by the @option{-mmcu=} option.
18117 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18121 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18125 This option is passed to the assembler and linker, and allows the
18126 linker to perform certain optimizations that cannot be done until
18131 Describes the type of hardware multiply supported by the target.
18132 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18133 for the original 16-bit-only multiply supported by early MCUs.
18134 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18135 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18136 A value of @samp{auto} can also be given. This tells GCC to deduce
18137 the hardware multiply support based upon the MCU name provided by the
18138 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18139 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
18142 Hardware multiplies are normally performed by calling a library
18143 routine. This saves space in the generated code. When compiling at
18144 @option{-O3} or higher however the hardware multiplier is invoked
18145 inline. This makes for bigger, but faster code.
18147 The hardware multiply routines disable interrupts whilst running and
18148 restore the previous interrupt state when they finish. This makes
18149 them safe to use inside interrupt handlers as well as in normal code.
18153 Enable the use of a minimum runtime environment - no static
18154 initializers or constructors. This is intended for memory-constrained
18155 devices. The compiler includes special symbols in some objects
18156 that tell the linker and runtime which code fragments are required.
18160 @node NDS32 Options
18161 @subsection NDS32 Options
18162 @cindex NDS32 Options
18164 These options are defined for NDS32 implementations:
18169 @opindex mbig-endian
18170 Generate code in big-endian mode.
18172 @item -mlittle-endian
18173 @opindex mlittle-endian
18174 Generate code in little-endian mode.
18176 @item -mreduced-regs
18177 @opindex mreduced-regs
18178 Use reduced-set registers for register allocation.
18181 @opindex mfull-regs
18182 Use full-set registers for register allocation.
18186 Generate conditional move instructions.
18190 Do not generate conditional move instructions.
18194 Generate performance extension instructions.
18196 @item -mno-perf-ext
18197 @opindex mno-perf-ext
18198 Do not generate performance extension instructions.
18202 Generate v3 push25/pop25 instructions.
18205 @opindex mno-v3push
18206 Do not generate v3 push25/pop25 instructions.
18210 Generate 16-bit instructions.
18213 @opindex mno-16-bit
18214 Do not generate 16-bit instructions.
18216 @item -misr-vector-size=@var{num}
18217 @opindex misr-vector-size
18218 Specify the size of each interrupt vector, which must be 4 or 16.
18220 @item -mcache-block-size=@var{num}
18221 @opindex mcache-block-size
18222 Specify the size of each cache block,
18223 which must be a power of 2 between 4 and 512.
18225 @item -march=@var{arch}
18227 Specify the name of the target architecture.
18229 @item -mcmodel=@var{code-model}
18231 Set the code model to one of
18234 All the data and read-only data segments must be within 512KB addressing space.
18235 The text segment must be within 16MB addressing space.
18236 @item @samp{medium}
18237 The data segment must be within 512KB while the read-only data segment can be
18238 within 4GB addressing space. The text segment should be still within 16MB
18241 All the text and data segments can be within 4GB addressing space.
18245 @opindex mctor-dtor
18246 Enable constructor/destructor feature.
18250 Guide linker to relax instructions.
18254 @node Nios II Options
18255 @subsection Nios II Options
18256 @cindex Nios II options
18257 @cindex Altera Nios II options
18259 These are the options defined for the Altera Nios II processor.
18265 @cindex smaller data references
18266 Put global and static objects less than or equal to @var{num} bytes
18267 into the small data or BSS sections instead of the normal data or BSS
18268 sections. The default value of @var{num} is 8.
18270 @item -mgpopt=@var{option}
18275 Generate (do not generate) GP-relative accesses. The following
18276 @var{option} names are recognized:
18281 Do not generate GP-relative accesses.
18284 Generate GP-relative accesses for small data objects that are not
18285 external or weak. Also use GP-relative addressing for objects that
18286 have been explicitly placed in a small data section via a @code{section}
18290 As for @samp{local}, but also generate GP-relative accesses for
18291 small data objects that are external or weak. If you use this option,
18292 you must ensure that all parts of your program (including libraries) are
18293 compiled with the same @option{-G} setting.
18296 Generate GP-relative accesses for all data objects in the program. If you
18297 use this option, the entire data and BSS segments
18298 of your program must fit in 64K of memory and you must use an appropriate
18299 linker script to allocate them within the addressible range of the
18303 Generate GP-relative addresses for function pointers as well as data
18304 pointers. If you use this option, the entire text, data, and BSS segments
18305 of your program must fit in 64K of memory and you must use an appropriate
18306 linker script to allocate them within the addressible range of the
18311 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18312 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18314 The default is @option{-mgpopt} except when @option{-fpic} or
18315 @option{-fPIC} is specified to generate position-independent code.
18316 Note that the Nios II ABI does not permit GP-relative accesses from
18319 You may need to specify @option{-mno-gpopt} explicitly when building
18320 programs that include large amounts of small data, including large
18321 GOT data sections. In this case, the 16-bit offset for GP-relative
18322 addressing may not be large enough to allow access to the entire
18323 small data section.
18329 Generate little-endian (default) or big-endian (experimental) code,
18332 @item -mbypass-cache
18333 @itemx -mno-bypass-cache
18334 @opindex mno-bypass-cache
18335 @opindex mbypass-cache
18336 Force all load and store instructions to always bypass cache by
18337 using I/O variants of the instructions. The default is not to
18340 @item -mno-cache-volatile
18341 @itemx -mcache-volatile
18342 @opindex mcache-volatile
18343 @opindex mno-cache-volatile
18344 Volatile memory access bypass the cache using the I/O variants of
18345 the load and store instructions. The default is not to bypass the cache.
18347 @item -mno-fast-sw-div
18348 @itemx -mfast-sw-div
18349 @opindex mno-fast-sw-div
18350 @opindex mfast-sw-div
18351 Do not use table-based fast divide for small numbers. The default
18352 is to use the fast divide at @option{-O3} and above.
18356 @itemx -mno-hw-mulx
18360 @opindex mno-hw-mul
18362 @opindex mno-hw-mulx
18364 @opindex mno-hw-div
18366 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18367 instructions by the compiler. The default is to emit @code{mul}
18368 and not emit @code{div} and @code{mulx}.
18370 @item -mcustom-@var{insn}=@var{N}
18371 @itemx -mno-custom-@var{insn}
18372 @opindex mcustom-@var{insn}
18373 @opindex mno-custom-@var{insn}
18374 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18375 custom instruction with encoding @var{N} when generating code that uses
18376 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18377 instruction 253 for single-precision floating-point add operations instead
18378 of the default behavior of using a library call.
18380 The following values of @var{insn} are supported. Except as otherwise
18381 noted, floating-point operations are expected to be implemented with
18382 normal IEEE 754 semantics and correspond directly to the C operators or the
18383 equivalent GCC built-in functions (@pxref{Other Builtins}).
18385 Single-precision floating point:
18388 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18389 Binary arithmetic operations.
18395 Unary absolute value.
18397 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18398 Comparison operations.
18400 @item @samp{fmins}, @samp{fmaxs}
18401 Floating-point minimum and maximum. These instructions are only
18402 generated if @option{-ffinite-math-only} is specified.
18404 @item @samp{fsqrts}
18405 Unary square root operation.
18407 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18408 Floating-point trigonometric and exponential functions. These instructions
18409 are only generated if @option{-funsafe-math-optimizations} is also specified.
18413 Double-precision floating point:
18416 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18417 Binary arithmetic operations.
18423 Unary absolute value.
18425 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18426 Comparison operations.
18428 @item @samp{fmind}, @samp{fmaxd}
18429 Double-precision minimum and maximum. These instructions are only
18430 generated if @option{-ffinite-math-only} is specified.
18432 @item @samp{fsqrtd}
18433 Unary square root operation.
18435 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18436 Double-precision trigonometric and exponential functions. These instructions
18437 are only generated if @option{-funsafe-math-optimizations} is also specified.
18443 @item @samp{fextsd}
18444 Conversion from single precision to double precision.
18446 @item @samp{ftruncds}
18447 Conversion from double precision to single precision.
18449 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18450 Conversion from floating point to signed or unsigned integer types, with
18451 truncation towards zero.
18454 Conversion from single-precision floating point to signed integer,
18455 rounding to the nearest integer and ties away from zero.
18456 This corresponds to the @code{__builtin_lroundf} function when
18457 @option{-fno-math-errno} is used.
18459 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18460 Conversion from signed or unsigned integer types to floating-point types.
18464 In addition, all of the following transfer instructions for internal
18465 registers X and Y must be provided to use any of the double-precision
18466 floating-point instructions. Custom instructions taking two
18467 double-precision source operands expect the first operand in the
18468 64-bit register X. The other operand (or only operand of a unary
18469 operation) is given to the custom arithmetic instruction with the
18470 least significant half in source register @var{src1} and the most
18471 significant half in @var{src2}. A custom instruction that returns a
18472 double-precision result returns the most significant 32 bits in the
18473 destination register and the other half in 32-bit register Y.
18474 GCC automatically generates the necessary code sequences to write
18475 register X and/or read register Y when double-precision floating-point
18476 instructions are used.
18481 Write @var{src1} into the least significant half of X and @var{src2} into
18482 the most significant half of X.
18485 Write @var{src1} into Y.
18487 @item @samp{frdxhi}, @samp{frdxlo}
18488 Read the most or least (respectively) significant half of X and store it in
18492 Read the value of Y and store it into @var{dest}.
18495 Note that you can gain more local control over generation of Nios II custom
18496 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18497 and @code{target("no-custom-@var{insn}")} function attributes
18498 (@pxref{Function Attributes})
18499 or pragmas (@pxref{Function Specific Option Pragmas}).
18501 @item -mcustom-fpu-cfg=@var{name}
18502 @opindex mcustom-fpu-cfg
18504 This option enables a predefined, named set of custom instruction encodings
18505 (see @option{-mcustom-@var{insn}} above).
18506 Currently, the following sets are defined:
18508 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18509 @gccoptlist{-mcustom-fmuls=252 @gol
18510 -mcustom-fadds=253 @gol
18511 -mcustom-fsubs=254 @gol
18512 -fsingle-precision-constant}
18514 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18515 @gccoptlist{-mcustom-fmuls=252 @gol
18516 -mcustom-fadds=253 @gol
18517 -mcustom-fsubs=254 @gol
18518 -mcustom-fdivs=255 @gol
18519 -fsingle-precision-constant}
18521 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18522 @gccoptlist{-mcustom-floatus=243 @gol
18523 -mcustom-fixsi=244 @gol
18524 -mcustom-floatis=245 @gol
18525 -mcustom-fcmpgts=246 @gol
18526 -mcustom-fcmples=249 @gol
18527 -mcustom-fcmpeqs=250 @gol
18528 -mcustom-fcmpnes=251 @gol
18529 -mcustom-fmuls=252 @gol
18530 -mcustom-fadds=253 @gol
18531 -mcustom-fsubs=254 @gol
18532 -mcustom-fdivs=255 @gol
18533 -fsingle-precision-constant}
18535 Custom instruction assignments given by individual
18536 @option{-mcustom-@var{insn}=} options override those given by
18537 @option{-mcustom-fpu-cfg=}, regardless of the
18538 order of the options on the command line.
18540 Note that you can gain more local control over selection of a FPU
18541 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18542 function attribute (@pxref{Function Attributes})
18543 or pragma (@pxref{Function Specific Option Pragmas}).
18547 These additional @samp{-m} options are available for the Altera Nios II
18548 ELF (bare-metal) target:
18554 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18555 startup and termination code, and is typically used in conjunction with
18556 @option{-msys-crt0=} to specify the location of the alternate startup code
18557 provided by the HAL BSP.
18561 Link with a limited version of the C library, @option{-lsmallc}, rather than
18564 @item -msys-crt0=@var{startfile}
18566 @var{startfile} is the file name of the startfile (crt0) to use
18567 when linking. This option is only useful in conjunction with @option{-mhal}.
18569 @item -msys-lib=@var{systemlib}
18571 @var{systemlib} is the library name of the library that provides
18572 low-level system calls required by the C library,
18573 e.g. @code{read} and @code{write}.
18574 This option is typically used to link with a library provided by a HAL BSP.
18578 @node Nvidia PTX Options
18579 @subsection Nvidia PTX Options
18580 @cindex Nvidia PTX options
18581 @cindex nvptx options
18583 These options are defined for Nvidia PTX:
18591 Generate code for 32-bit or 64-bit ABI.
18594 @opindex mmainkernel
18595 Link in code for a __main kernel. This is for stand-alone instead of
18596 offloading execution.
18600 @node PDP-11 Options
18601 @subsection PDP-11 Options
18602 @cindex PDP-11 Options
18604 These options are defined for the PDP-11:
18609 Use hardware FPP floating point. This is the default. (FIS floating
18610 point on the PDP-11/40 is not supported.)
18613 @opindex msoft-float
18614 Do not use hardware floating point.
18618 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18622 Return floating-point results in memory. This is the default.
18626 Generate code for a PDP-11/40.
18630 Generate code for a PDP-11/45. This is the default.
18634 Generate code for a PDP-11/10.
18636 @item -mbcopy-builtin
18637 @opindex mbcopy-builtin
18638 Use inline @code{movmemhi} patterns for copying memory. This is the
18643 Do not use inline @code{movmemhi} patterns for copying memory.
18649 Use 16-bit @code{int}. This is the default.
18655 Use 32-bit @code{int}.
18658 @itemx -mno-float32
18660 @opindex mno-float32
18661 Use 64-bit @code{float}. This is the default.
18664 @itemx -mno-float64
18666 @opindex mno-float64
18667 Use 32-bit @code{float}.
18671 Use @code{abshi2} pattern. This is the default.
18675 Do not use @code{abshi2} pattern.
18677 @item -mbranch-expensive
18678 @opindex mbranch-expensive
18679 Pretend that branches are expensive. This is for experimenting with
18680 code generation only.
18682 @item -mbranch-cheap
18683 @opindex mbranch-cheap
18684 Do not pretend that branches are expensive. This is the default.
18688 Use Unix assembler syntax. This is the default when configured for
18689 @samp{pdp11-*-bsd}.
18693 Use DEC assembler syntax. This is the default when configured for any
18694 PDP-11 target other than @samp{pdp11-*-bsd}.
18697 @node picoChip Options
18698 @subsection picoChip Options
18699 @cindex picoChip options
18701 These @samp{-m} options are defined for picoChip implementations:
18705 @item -mae=@var{ae_type}
18707 Set the instruction set, register set, and instruction scheduling
18708 parameters for array element type @var{ae_type}. Supported values
18709 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18711 @option{-mae=ANY} selects a completely generic AE type. Code
18712 generated with this option runs on any of the other AE types. The
18713 code is not as efficient as it would be if compiled for a specific
18714 AE type, and some types of operation (e.g., multiplication) do not
18715 work properly on all types of AE.
18717 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18718 for compiled code, and is the default.
18720 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18721 option may suffer from poor performance of byte (char) manipulation,
18722 since the DSP AE does not provide hardware support for byte load/stores.
18724 @item -msymbol-as-address
18725 Enable the compiler to directly use a symbol name as an address in a
18726 load/store instruction, without first loading it into a
18727 register. Typically, the use of this option generates larger
18728 programs, which run faster than when the option isn't used. However, the
18729 results vary from program to program, so it is left as a user option,
18730 rather than being permanently enabled.
18732 @item -mno-inefficient-warnings
18733 Disables warnings about the generation of inefficient code. These
18734 warnings can be generated, for example, when compiling code that
18735 performs byte-level memory operations on the MAC AE type. The MAC AE has
18736 no hardware support for byte-level memory operations, so all byte
18737 load/stores must be synthesized from word load/store operations. This is
18738 inefficient and a warning is generated to indicate
18739 that you should rewrite the code to avoid byte operations, or to target
18740 an AE type that has the necessary hardware support. This option disables
18745 @node PowerPC Options
18746 @subsection PowerPC Options
18747 @cindex PowerPC options
18749 These are listed under @xref{RS/6000 and PowerPC Options}.
18752 @subsection RL78 Options
18753 @cindex RL78 Options
18759 Links in additional target libraries to support operation within a
18768 Specifies the type of hardware multiplication and division support to
18769 be used. The simplest is @code{none}, which uses software for both
18770 multiplication and division. This is the default. The @code{g13}
18771 value is for the hardware multiply/divide peripheral found on the
18772 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
18773 the multiplication and division instructions supported by the RL78/G14
18774 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
18775 the value @code{mg10} is an alias for @code{none}.
18777 In addition a C preprocessor macro is defined, based upon the setting
18778 of this option. Possible values are: @code{__RL78_MUL_NONE__},
18779 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
18786 Specifies the RL78 core to target. The default is the G14 core, also
18787 known as an S3 core or just RL78. The G13 or S2 core does not have
18788 multiply or divide instructions, instead it uses a hardware peripheral
18789 for these operations. The G10 or S1 core does not have register
18790 banks, so it uses a different calling convention.
18792 If this option is set it also selects the type of hardware multiply
18793 support to use, unless this is overridden by an explicit
18794 @option{-mmul=none} option on the command line. Thus specifying
18795 @option{-mcpu=g13} enables the use of the G13 hardware multiply
18796 peripheral and specifying @option{-mcpu=g10} disables the use of
18797 hardware multipications altogether.
18799 Note, although the RL78/G14 core is the default target, specifying
18800 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
18801 change the behaviour of the toolchain since it also enables G14
18802 hardware multiply support. If these options are not specified on the
18803 command line then software multiplication routines will be used even
18804 though the code targets the RL78 core. This is for backwards
18805 compatibility with older toolchains which did not have hardware
18806 multiply and divide support.
18808 In addition a C preprocessor macro is defined, based upon the setting
18809 of this option. Possible values are: @code{__RL78_G10__},
18810 @code{__RL78_G13__} or @code{__RL78_G14__}.
18820 These are aliases for the corresponding @option{-mcpu=} option. They
18821 are provided for backwards compatibility.
18825 Allow the compiler to use all of the available registers. By default
18826 registers @code{r24..r31} are reserved for use in interrupt handlers.
18827 With this option enabled these registers can be used in ordinary
18830 @item -m64bit-doubles
18831 @itemx -m32bit-doubles
18832 @opindex m64bit-doubles
18833 @opindex m32bit-doubles
18834 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
18835 or 32 bits (@option{-m32bit-doubles}) in size. The default is
18836 @option{-m32bit-doubles}.
18840 @node RS/6000 and PowerPC Options
18841 @subsection IBM RS/6000 and PowerPC Options
18842 @cindex RS/6000 and PowerPC Options
18843 @cindex IBM RS/6000 and PowerPC Options
18845 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
18847 @item -mpowerpc-gpopt
18848 @itemx -mno-powerpc-gpopt
18849 @itemx -mpowerpc-gfxopt
18850 @itemx -mno-powerpc-gfxopt
18853 @itemx -mno-powerpc64
18857 @itemx -mno-popcntb
18859 @itemx -mno-popcntd
18868 @itemx -mno-hard-dfp
18869 @opindex mpowerpc-gpopt
18870 @opindex mno-powerpc-gpopt
18871 @opindex mpowerpc-gfxopt
18872 @opindex mno-powerpc-gfxopt
18873 @opindex mpowerpc64
18874 @opindex mno-powerpc64
18878 @opindex mno-popcntb
18880 @opindex mno-popcntd
18886 @opindex mno-mfpgpr
18888 @opindex mno-hard-dfp
18889 You use these options to specify which instructions are available on the
18890 processor you are using. The default value of these options is
18891 determined when configuring GCC@. Specifying the
18892 @option{-mcpu=@var{cpu_type}} overrides the specification of these
18893 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
18894 rather than the options listed above.
18896 Specifying @option{-mpowerpc-gpopt} allows
18897 GCC to use the optional PowerPC architecture instructions in the
18898 General Purpose group, including floating-point square root. Specifying
18899 @option{-mpowerpc-gfxopt} allows GCC to
18900 use the optional PowerPC architecture instructions in the Graphics
18901 group, including floating-point select.
18903 The @option{-mmfcrf} option allows GCC to generate the move from
18904 condition register field instruction implemented on the POWER4
18905 processor and other processors that support the PowerPC V2.01
18907 The @option{-mpopcntb} option allows GCC to generate the popcount and
18908 double-precision FP reciprocal estimate instruction implemented on the
18909 POWER5 processor and other processors that support the PowerPC V2.02
18911 The @option{-mpopcntd} option allows GCC to generate the popcount
18912 instruction implemented on the POWER7 processor and other processors
18913 that support the PowerPC V2.06 architecture.
18914 The @option{-mfprnd} option allows GCC to generate the FP round to
18915 integer instructions implemented on the POWER5+ processor and other
18916 processors that support the PowerPC V2.03 architecture.
18917 The @option{-mcmpb} option allows GCC to generate the compare bytes
18918 instruction implemented on the POWER6 processor and other processors
18919 that support the PowerPC V2.05 architecture.
18920 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
18921 general-purpose register instructions implemented on the POWER6X
18922 processor and other processors that support the extended PowerPC V2.05
18924 The @option{-mhard-dfp} option allows GCC to generate the decimal
18925 floating-point instructions implemented on some POWER processors.
18927 The @option{-mpowerpc64} option allows GCC to generate the additional
18928 64-bit instructions that are found in the full PowerPC64 architecture
18929 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
18930 @option{-mno-powerpc64}.
18932 @item -mcpu=@var{cpu_type}
18934 Set architecture type, register usage, and
18935 instruction scheduling parameters for machine type @var{cpu_type}.
18936 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
18937 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
18938 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
18939 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
18940 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
18941 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
18942 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
18943 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
18944 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
18945 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
18946 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
18948 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
18949 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
18950 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
18951 architecture machine types, with an appropriate, generic processor
18952 model assumed for scheduling purposes.
18954 The other options specify a specific processor. Code generated under
18955 those options runs best on that processor, and may not run at all on
18958 The @option{-mcpu} options automatically enable or disable the
18961 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
18962 -mpopcntb -mpopcntd -mpowerpc64 @gol
18963 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
18964 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
18965 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
18966 -mquad-memory -mquad-memory-atomic}
18968 The particular options set for any particular CPU varies between
18969 compiler versions, depending on what setting seems to produce optimal
18970 code for that CPU; it doesn't necessarily reflect the actual hardware's
18971 capabilities. If you wish to set an individual option to a particular
18972 value, you may specify it after the @option{-mcpu} option, like
18973 @option{-mcpu=970 -mno-altivec}.
18975 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
18976 not enabled or disabled by the @option{-mcpu} option at present because
18977 AIX does not have full support for these options. You may still
18978 enable or disable them individually if you're sure it'll work in your
18981 @item -mtune=@var{cpu_type}
18983 Set the instruction scheduling parameters for machine type
18984 @var{cpu_type}, but do not set the architecture type or register usage,
18985 as @option{-mcpu=@var{cpu_type}} does. The same
18986 values for @var{cpu_type} are used for @option{-mtune} as for
18987 @option{-mcpu}. If both are specified, the code generated uses the
18988 architecture and registers set by @option{-mcpu}, but the
18989 scheduling parameters set by @option{-mtune}.
18991 @item -mcmodel=small
18992 @opindex mcmodel=small
18993 Generate PowerPC64 code for the small model: The TOC is limited to
18996 @item -mcmodel=medium
18997 @opindex mcmodel=medium
18998 Generate PowerPC64 code for the medium model: The TOC and other static
18999 data may be up to a total of 4G in size.
19001 @item -mcmodel=large
19002 @opindex mcmodel=large
19003 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19004 in size. Other data and code is only limited by the 64-bit address
19008 @itemx -mno-altivec
19010 @opindex mno-altivec
19011 Generate code that uses (does not use) AltiVec instructions, and also
19012 enable the use of built-in functions that allow more direct access to
19013 the AltiVec instruction set. You may also need to set
19014 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19017 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19018 @option{-maltivec=be}, the element order for Altivec intrinsics such
19019 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19020 match array element order corresponding to the endianness of the
19021 target. That is, element zero identifies the leftmost element in a
19022 vector register when targeting a big-endian platform, and identifies
19023 the rightmost element in a vector register when targeting a
19024 little-endian platform.
19027 @opindex maltivec=be
19028 Generate Altivec instructions using big-endian element order,
19029 regardless of whether the target is big- or little-endian. This is
19030 the default when targeting a big-endian platform.
19032 The element order is used to interpret element numbers in Altivec
19033 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19034 @code{vec_insert}. By default, these match array element order
19035 corresponding to the endianness for the target.
19038 @opindex maltivec=le
19039 Generate Altivec instructions using little-endian element order,
19040 regardless of whether the target is big- or little-endian. This is
19041 the default when targeting a little-endian platform. This option is
19042 currently ignored when targeting a big-endian platform.
19044 The element order is used to interpret element numbers in Altivec
19045 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19046 @code{vec_insert}. By default, these match array element order
19047 corresponding to the endianness for the target.
19052 @opindex mno-vrsave
19053 Generate VRSAVE instructions when generating AltiVec code.
19055 @item -mgen-cell-microcode
19056 @opindex mgen-cell-microcode
19057 Generate Cell microcode instructions.
19059 @item -mwarn-cell-microcode
19060 @opindex mwarn-cell-microcode
19061 Warn when a Cell microcode instruction is emitted. An example
19062 of a Cell microcode instruction is a variable shift.
19065 @opindex msecure-plt
19066 Generate code that allows @command{ld} and @command{ld.so}
19067 to build executables and shared
19068 libraries with non-executable @code{.plt} and @code{.got} sections.
19070 32-bit SYSV ABI option.
19074 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19076 requires @code{.plt} and @code{.got}
19077 sections that are both writable and executable.
19078 This is a PowerPC 32-bit SYSV ABI option.
19084 This switch enables or disables the generation of ISEL instructions.
19086 @item -misel=@var{yes/no}
19087 This switch has been deprecated. Use @option{-misel} and
19088 @option{-mno-isel} instead.
19094 This switch enables or disables the generation of SPE simd
19100 @opindex mno-paired
19101 This switch enables or disables the generation of PAIRED simd
19104 @item -mspe=@var{yes/no}
19105 This option has been deprecated. Use @option{-mspe} and
19106 @option{-mno-spe} instead.
19112 Generate code that uses (does not use) vector/scalar (VSX)
19113 instructions, and also enable the use of built-in functions that allow
19114 more direct access to the VSX instruction set.
19119 @opindex mno-crypto
19120 Enable the use (disable) of the built-in functions that allow direct
19121 access to the cryptographic instructions that were added in version
19122 2.07 of the PowerPC ISA.
19124 @item -mdirect-move
19125 @itemx -mno-direct-move
19126 @opindex mdirect-move
19127 @opindex mno-direct-move
19128 Generate code that uses (does not use) the instructions to move data
19129 between the general purpose registers and the vector/scalar (VSX)
19130 registers that were added in version 2.07 of the PowerPC ISA.
19132 @item -mpower8-fusion
19133 @itemx -mno-power8-fusion
19134 @opindex mpower8-fusion
19135 @opindex mno-power8-fusion
19136 Generate code that keeps (does not keeps) some integer operations
19137 adjacent so that the instructions can be fused together on power8 and
19140 @item -mpower8-vector
19141 @itemx -mno-power8-vector
19142 @opindex mpower8-vector
19143 @opindex mno-power8-vector
19144 Generate code that uses (does not use) the vector and scalar
19145 instructions that were added in version 2.07 of the PowerPC ISA. Also
19146 enable the use of built-in functions that allow more direct access to
19147 the vector instructions.
19149 @item -mquad-memory
19150 @itemx -mno-quad-memory
19151 @opindex mquad-memory
19152 @opindex mno-quad-memory
19153 Generate code that uses (does not use) the non-atomic quad word memory
19154 instructions. The @option{-mquad-memory} option requires use of
19157 @item -mquad-memory-atomic
19158 @itemx -mno-quad-memory-atomic
19159 @opindex mquad-memory-atomic
19160 @opindex mno-quad-memory-atomic
19161 Generate code that uses (does not use) the atomic quad word memory
19162 instructions. The @option{-mquad-memory-atomic} option requires use of
19165 @item -mupper-regs-df
19166 @itemx -mno-upper-regs-df
19167 @opindex mupper-regs-df
19168 @opindex mno-upper-regs-df
19169 Generate code that uses (does not use) the scalar double precision
19170 instructions that target all 64 registers in the vector/scalar
19171 floating point register set that were added in version 2.06 of the
19172 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19173 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19174 @option{-mvsx} options.
19176 @item -mupper-regs-sf
19177 @itemx -mno-upper-regs-sf
19178 @opindex mupper-regs-sf
19179 @opindex mno-upper-regs-sf
19180 Generate code that uses (does not use) the scalar single precision
19181 instructions that target all 64 registers in the vector/scalar
19182 floating point register set that were added in version 2.07 of the
19183 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19184 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19188 @itemx -mno-upper-regs
19189 @opindex mupper-regs
19190 @opindex mno-upper-regs
19191 Generate code that uses (does not use) the scalar
19192 instructions that target all 64 registers in the vector/scalar
19193 floating point register set, depending on the model of the machine.
19195 If the @option{-mno-upper-regs} option is used, it turns off both
19196 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19198 @item -mfloat-gprs=@var{yes/single/double/no}
19199 @itemx -mfloat-gprs
19200 @opindex mfloat-gprs
19201 This switch enables or disables the generation of floating-point
19202 operations on the general-purpose registers for architectures that
19205 The argument @samp{yes} or @samp{single} enables the use of
19206 single-precision floating-point operations.
19208 The argument @samp{double} enables the use of single and
19209 double-precision floating-point operations.
19211 The argument @samp{no} disables floating-point operations on the
19212 general-purpose registers.
19214 This option is currently only available on the MPC854x.
19220 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19221 targets (including GNU/Linux). The 32-bit environment sets int, long
19222 and pointer to 32 bits and generates code that runs on any PowerPC
19223 variant. The 64-bit environment sets int to 32 bits and long and
19224 pointer to 64 bits, and generates code for PowerPC64, as for
19225 @option{-mpowerpc64}.
19228 @itemx -mno-fp-in-toc
19229 @itemx -mno-sum-in-toc
19230 @itemx -mminimal-toc
19232 @opindex mno-fp-in-toc
19233 @opindex mno-sum-in-toc
19234 @opindex mminimal-toc
19235 Modify generation of the TOC (Table Of Contents), which is created for
19236 every executable file. The @option{-mfull-toc} option is selected by
19237 default. In that case, GCC allocates at least one TOC entry for
19238 each unique non-automatic variable reference in your program. GCC
19239 also places floating-point constants in the TOC@. However, only
19240 16,384 entries are available in the TOC@.
19242 If you receive a linker error message that saying you have overflowed
19243 the available TOC space, you can reduce the amount of TOC space used
19244 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19245 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19246 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19247 generate code to calculate the sum of an address and a constant at
19248 run time instead of putting that sum into the TOC@. You may specify one
19249 or both of these options. Each causes GCC to produce very slightly
19250 slower and larger code at the expense of conserving TOC space.
19252 If you still run out of space in the TOC even when you specify both of
19253 these options, specify @option{-mminimal-toc} instead. This option causes
19254 GCC to make only one TOC entry for every file. When you specify this
19255 option, GCC produces code that is slower and larger but which
19256 uses extremely little TOC space. You may wish to use this option
19257 only on files that contain less frequently-executed code.
19263 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19264 @code{long} type, and the infrastructure needed to support them.
19265 Specifying @option{-maix64} implies @option{-mpowerpc64},
19266 while @option{-maix32} disables the 64-bit ABI and
19267 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19270 @itemx -mno-xl-compat
19271 @opindex mxl-compat
19272 @opindex mno-xl-compat
19273 Produce code that conforms more closely to IBM XL compiler semantics
19274 when using AIX-compatible ABI@. Pass floating-point arguments to
19275 prototyped functions beyond the register save area (RSA) on the stack
19276 in addition to argument FPRs. Do not assume that most significant
19277 double in 128-bit long double value is properly rounded when comparing
19278 values and converting to double. Use XL symbol names for long double
19281 The AIX calling convention was extended but not initially documented to
19282 handle an obscure K&R C case of calling a function that takes the
19283 address of its arguments with fewer arguments than declared. IBM XL
19284 compilers access floating-point arguments that do not fit in the
19285 RSA from the stack when a subroutine is compiled without
19286 optimization. Because always storing floating-point arguments on the
19287 stack is inefficient and rarely needed, this option is not enabled by
19288 default and only is necessary when calling subroutines compiled by IBM
19289 XL compilers without optimization.
19293 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19294 application written to use message passing with special startup code to
19295 enable the application to run. The system must have PE installed in the
19296 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19297 must be overridden with the @option{-specs=} option to specify the
19298 appropriate directory location. The Parallel Environment does not
19299 support threads, so the @option{-mpe} option and the @option{-pthread}
19300 option are incompatible.
19302 @item -malign-natural
19303 @itemx -malign-power
19304 @opindex malign-natural
19305 @opindex malign-power
19306 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19307 @option{-malign-natural} overrides the ABI-defined alignment of larger
19308 types, such as floating-point doubles, on their natural size-based boundary.
19309 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19310 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19312 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19316 @itemx -mhard-float
19317 @opindex msoft-float
19318 @opindex mhard-float
19319 Generate code that does not use (uses) the floating-point register set.
19320 Software floating-point emulation is provided if you use the
19321 @option{-msoft-float} option, and pass the option to GCC when linking.
19323 @item -msingle-float
19324 @itemx -mdouble-float
19325 @opindex msingle-float
19326 @opindex mdouble-float
19327 Generate code for single- or double-precision floating-point operations.
19328 @option{-mdouble-float} implies @option{-msingle-float}.
19331 @opindex msimple-fpu
19332 Do not generate @code{sqrt} and @code{div} instructions for hardware
19333 floating-point unit.
19335 @item -mfpu=@var{name}
19337 Specify type of floating-point unit. Valid values for @var{name} are
19338 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19339 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19340 @samp{sp_full} (equivalent to @option{-msingle-float}),
19341 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19344 @opindex mxilinx-fpu
19345 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19348 @itemx -mno-multiple
19350 @opindex mno-multiple
19351 Generate code that uses (does not use) the load multiple word
19352 instructions and the store multiple word instructions. These
19353 instructions are generated by default on POWER systems, and not
19354 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19355 PowerPC systems, since those instructions do not work when the
19356 processor is in little-endian mode. The exceptions are PPC740 and
19357 PPC750 which permit these instructions in little-endian mode.
19362 @opindex mno-string
19363 Generate code that uses (does not use) the load string instructions
19364 and the store string word instructions to save multiple registers and
19365 do small block moves. These instructions are generated by default on
19366 POWER systems, and not generated on PowerPC systems. Do not use
19367 @option{-mstring} on little-endian PowerPC systems, since those
19368 instructions do not work when the processor is in little-endian mode.
19369 The exceptions are PPC740 and PPC750 which permit these instructions
19370 in little-endian mode.
19375 @opindex mno-update
19376 Generate code that uses (does not use) the load or store instructions
19377 that update the base register to the address of the calculated memory
19378 location. These instructions are generated by default. If you use
19379 @option{-mno-update}, there is a small window between the time that the
19380 stack pointer is updated and the address of the previous frame is
19381 stored, which means code that walks the stack frame across interrupts or
19382 signals may get corrupted data.
19384 @item -mavoid-indexed-addresses
19385 @itemx -mno-avoid-indexed-addresses
19386 @opindex mavoid-indexed-addresses
19387 @opindex mno-avoid-indexed-addresses
19388 Generate code that tries to avoid (not avoid) the use of indexed load
19389 or store instructions. These instructions can incur a performance
19390 penalty on Power6 processors in certain situations, such as when
19391 stepping through large arrays that cross a 16M boundary. This option
19392 is enabled by default when targeting Power6 and disabled otherwise.
19395 @itemx -mno-fused-madd
19396 @opindex mfused-madd
19397 @opindex mno-fused-madd
19398 Generate code that uses (does not use) the floating-point multiply and
19399 accumulate instructions. These instructions are generated by default
19400 if hardware floating point is used. The machine-dependent
19401 @option{-mfused-madd} option is now mapped to the machine-independent
19402 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19403 mapped to @option{-ffp-contract=off}.
19409 Generate code that uses (does not use) the half-word multiply and
19410 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19411 These instructions are generated by default when targeting those
19418 Generate code that uses (does not use) the string-search @samp{dlmzb}
19419 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19420 generated by default when targeting those processors.
19422 @item -mno-bit-align
19424 @opindex mno-bit-align
19425 @opindex mbit-align
19426 On System V.4 and embedded PowerPC systems do not (do) force structures
19427 and unions that contain bit-fields to be aligned to the base type of the
19430 For example, by default a structure containing nothing but 8
19431 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19432 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19433 the structure is aligned to a 1-byte boundary and is 1 byte in
19436 @item -mno-strict-align
19437 @itemx -mstrict-align
19438 @opindex mno-strict-align
19439 @opindex mstrict-align
19440 On System V.4 and embedded PowerPC systems do not (do) assume that
19441 unaligned memory references are handled by the system.
19443 @item -mrelocatable
19444 @itemx -mno-relocatable
19445 @opindex mrelocatable
19446 @opindex mno-relocatable
19447 Generate code that allows (does not allow) a static executable to be
19448 relocated to a different address at run time. A simple embedded
19449 PowerPC system loader should relocate the entire contents of
19450 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19451 a table of 32-bit addresses generated by this option. For this to
19452 work, all objects linked together must be compiled with
19453 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19454 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19456 @item -mrelocatable-lib
19457 @itemx -mno-relocatable-lib
19458 @opindex mrelocatable-lib
19459 @opindex mno-relocatable-lib
19460 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19461 @code{.fixup} section to allow static executables to be relocated at
19462 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19463 alignment of @option{-mrelocatable}. Objects compiled with
19464 @option{-mrelocatable-lib} may be linked with objects compiled with
19465 any combination of the @option{-mrelocatable} options.
19471 On System V.4 and embedded PowerPC systems do not (do) assume that
19472 register 2 contains a pointer to a global area pointing to the addresses
19473 used in the program.
19476 @itemx -mlittle-endian
19478 @opindex mlittle-endian
19479 On System V.4 and embedded PowerPC systems compile code for the
19480 processor in little-endian mode. The @option{-mlittle-endian} option is
19481 the same as @option{-mlittle}.
19484 @itemx -mbig-endian
19486 @opindex mbig-endian
19487 On System V.4 and embedded PowerPC systems compile code for the
19488 processor in big-endian mode. The @option{-mbig-endian} option is
19489 the same as @option{-mbig}.
19491 @item -mdynamic-no-pic
19492 @opindex mdynamic-no-pic
19493 On Darwin and Mac OS X systems, compile code so that it is not
19494 relocatable, but that its external references are relocatable. The
19495 resulting code is suitable for applications, but not shared
19498 @item -msingle-pic-base
19499 @opindex msingle-pic-base
19500 Treat the register used for PIC addressing as read-only, rather than
19501 loading it in the prologue for each function. The runtime system is
19502 responsible for initializing this register with an appropriate value
19503 before execution begins.
19505 @item -mprioritize-restricted-insns=@var{priority}
19506 @opindex mprioritize-restricted-insns
19507 This option controls the priority that is assigned to
19508 dispatch-slot restricted instructions during the second scheduling
19509 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19510 or @samp{2} to assign no, highest, or second-highest (respectively)
19511 priority to dispatch-slot restricted
19514 @item -msched-costly-dep=@var{dependence_type}
19515 @opindex msched-costly-dep
19516 This option controls which dependences are considered costly
19517 by the target during instruction scheduling. The argument
19518 @var{dependence_type} takes one of the following values:
19522 No dependence is costly.
19525 All dependences are costly.
19527 @item @samp{true_store_to_load}
19528 A true dependence from store to load is costly.
19530 @item @samp{store_to_load}
19531 Any dependence from store to load is costly.
19534 Any dependence for which the latency is greater than or equal to
19535 @var{number} is costly.
19538 @item -minsert-sched-nops=@var{scheme}
19539 @opindex minsert-sched-nops
19540 This option controls which NOP insertion scheme is used during
19541 the second scheduling pass. The argument @var{scheme} takes one of the
19549 Pad with NOPs any dispatch group that has vacant issue slots,
19550 according to the scheduler's grouping.
19552 @item @samp{regroup_exact}
19553 Insert NOPs to force costly dependent insns into
19554 separate groups. Insert exactly as many NOPs as needed to force an insn
19555 to a new group, according to the estimated processor grouping.
19558 Insert NOPs to force costly dependent insns into
19559 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19563 @opindex mcall-sysv
19564 On System V.4 and embedded PowerPC systems compile code using calling
19565 conventions that adhere to the March 1995 draft of the System V
19566 Application Binary Interface, PowerPC processor supplement. This is the
19567 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19569 @item -mcall-sysv-eabi
19571 @opindex mcall-sysv-eabi
19572 @opindex mcall-eabi
19573 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19575 @item -mcall-sysv-noeabi
19576 @opindex mcall-sysv-noeabi
19577 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19579 @item -mcall-aixdesc
19581 On System V.4 and embedded PowerPC systems compile code for the AIX
19585 @opindex mcall-linux
19586 On System V.4 and embedded PowerPC systems compile code for the
19587 Linux-based GNU system.
19589 @item -mcall-freebsd
19590 @opindex mcall-freebsd
19591 On System V.4 and embedded PowerPC systems compile code for the
19592 FreeBSD operating system.
19594 @item -mcall-netbsd
19595 @opindex mcall-netbsd
19596 On System V.4 and embedded PowerPC systems compile code for the
19597 NetBSD operating system.
19599 @item -mcall-openbsd
19600 @opindex mcall-netbsd
19601 On System V.4 and embedded PowerPC systems compile code for the
19602 OpenBSD operating system.
19604 @item -maix-struct-return
19605 @opindex maix-struct-return
19606 Return all structures in memory (as specified by the AIX ABI)@.
19608 @item -msvr4-struct-return
19609 @opindex msvr4-struct-return
19610 Return structures smaller than 8 bytes in registers (as specified by the
19613 @item -mabi=@var{abi-type}
19615 Extend the current ABI with a particular extension, or remove such extension.
19616 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19617 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19618 @samp{elfv1}, @samp{elfv2}@.
19622 Extend the current ABI with SPE ABI extensions. This does not change
19623 the default ABI, instead it adds the SPE ABI extensions to the current
19627 @opindex mabi=no-spe
19628 Disable Book-E SPE ABI extensions for the current ABI@.
19630 @item -mabi=ibmlongdouble
19631 @opindex mabi=ibmlongdouble
19632 Change the current ABI to use IBM extended-precision long double.
19633 This is a PowerPC 32-bit SYSV ABI option.
19635 @item -mabi=ieeelongdouble
19636 @opindex mabi=ieeelongdouble
19637 Change the current ABI to use IEEE extended-precision long double.
19638 This is a PowerPC 32-bit Linux ABI option.
19641 @opindex mabi=elfv1
19642 Change the current ABI to use the ELFv1 ABI.
19643 This is the default ABI for big-endian PowerPC 64-bit Linux.
19644 Overriding the default ABI requires special system support and is
19645 likely to fail in spectacular ways.
19648 @opindex mabi=elfv2
19649 Change the current ABI to use the ELFv2 ABI.
19650 This is the default ABI for little-endian PowerPC 64-bit Linux.
19651 Overriding the default ABI requires special system support and is
19652 likely to fail in spectacular ways.
19655 @itemx -mno-prototype
19656 @opindex mprototype
19657 @opindex mno-prototype
19658 On System V.4 and embedded PowerPC systems assume that all calls to
19659 variable argument functions are properly prototyped. Otherwise, the
19660 compiler must insert an instruction before every non-prototyped call to
19661 set or clear bit 6 of the condition code register (@code{CR}) to
19662 indicate whether floating-point values are passed in the floating-point
19663 registers in case the function takes variable arguments. With
19664 @option{-mprototype}, only calls to prototyped variable argument functions
19665 set or clear the bit.
19669 On embedded PowerPC systems, assume that the startup module is called
19670 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19671 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19676 On embedded PowerPC systems, assume that the startup module is called
19677 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19682 On embedded PowerPC systems, assume that the startup module is called
19683 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19686 @item -myellowknife
19687 @opindex myellowknife
19688 On embedded PowerPC systems, assume that the startup module is called
19689 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19694 On System V.4 and embedded PowerPC systems, specify that you are
19695 compiling for a VxWorks system.
19699 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19700 header to indicate that @samp{eabi} extended relocations are used.
19706 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19707 Embedded Applications Binary Interface (EABI), which is a set of
19708 modifications to the System V.4 specifications. Selecting @option{-meabi}
19709 means that the stack is aligned to an 8-byte boundary, a function
19710 @code{__eabi} is called from @code{main} to set up the EABI
19711 environment, and the @option{-msdata} option can use both @code{r2} and
19712 @code{r13} to point to two separate small data areas. Selecting
19713 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19714 no EABI initialization function is called from @code{main}, and the
19715 @option{-msdata} option only uses @code{r13} to point to a single
19716 small data area. The @option{-meabi} option is on by default if you
19717 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19720 @opindex msdata=eabi
19721 On System V.4 and embedded PowerPC systems, put small initialized
19722 @code{const} global and static data in the @code{.sdata2} section, which
19723 is pointed to by register @code{r2}. Put small initialized
19724 non-@code{const} global and static data in the @code{.sdata} section,
19725 which is pointed to by register @code{r13}. Put small uninitialized
19726 global and static data in the @code{.sbss} section, which is adjacent to
19727 the @code{.sdata} section. The @option{-msdata=eabi} option is
19728 incompatible with the @option{-mrelocatable} option. The
19729 @option{-msdata=eabi} option also sets the @option{-memb} option.
19732 @opindex msdata=sysv
19733 On System V.4 and embedded PowerPC systems, put small global and static
19734 data in the @code{.sdata} section, which is pointed to by register
19735 @code{r13}. Put small uninitialized global and static data in the
19736 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19737 The @option{-msdata=sysv} option is incompatible with the
19738 @option{-mrelocatable} option.
19740 @item -msdata=default
19742 @opindex msdata=default
19744 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19745 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19746 same as @option{-msdata=sysv}.
19749 @opindex msdata=data
19750 On System V.4 and embedded PowerPC systems, put small global
19751 data in the @code{.sdata} section. Put small uninitialized global
19752 data in the @code{.sbss} section. Do not use register @code{r13}
19753 to address small data however. This is the default behavior unless
19754 other @option{-msdata} options are used.
19758 @opindex msdata=none
19760 On embedded PowerPC systems, put all initialized global and static data
19761 in the @code{.data} section, and all uninitialized data in the
19762 @code{.bss} section.
19764 @item -mblock-move-inline-limit=@var{num}
19765 @opindex mblock-move-inline-limit
19766 Inline all block moves (such as calls to @code{memcpy} or structure
19767 copies) less than or equal to @var{num} bytes. The minimum value for
19768 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19769 targets. The default value is target-specific.
19773 @cindex smaller data references (PowerPC)
19774 @cindex .sdata/.sdata2 references (PowerPC)
19775 On embedded PowerPC systems, put global and static items less than or
19776 equal to @var{num} bytes into the small data or BSS sections instead of
19777 the normal data or BSS section. By default, @var{num} is 8. The
19778 @option{-G @var{num}} switch is also passed to the linker.
19779 All modules should be compiled with the same @option{-G @var{num}} value.
19782 @itemx -mno-regnames
19784 @opindex mno-regnames
19785 On System V.4 and embedded PowerPC systems do (do not) emit register
19786 names in the assembly language output using symbolic forms.
19789 @itemx -mno-longcall
19791 @opindex mno-longcall
19792 By default assume that all calls are far away so that a longer and more
19793 expensive calling sequence is required. This is required for calls
19794 farther than 32 megabytes (33,554,432 bytes) from the current location.
19795 A short call is generated if the compiler knows
19796 the call cannot be that far away. This setting can be overridden by
19797 the @code{shortcall} function attribute, or by @code{#pragma
19800 Some linkers are capable of detecting out-of-range calls and generating
19801 glue code on the fly. On these systems, long calls are unnecessary and
19802 generate slower code. As of this writing, the AIX linker can do this,
19803 as can the GNU linker for PowerPC/64. It is planned to add this feature
19804 to the GNU linker for 32-bit PowerPC systems as well.
19806 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
19807 callee, L42}, plus a @dfn{branch island} (glue code). The two target
19808 addresses represent the callee and the branch island. The
19809 Darwin/PPC linker prefers the first address and generates a @code{bl
19810 callee} if the PPC @code{bl} instruction reaches the callee directly;
19811 otherwise, the linker generates @code{bl L42} to call the branch
19812 island. The branch island is appended to the body of the
19813 calling function; it computes the full 32-bit address of the callee
19816 On Mach-O (Darwin) systems, this option directs the compiler emit to
19817 the glue for every direct call, and the Darwin linker decides whether
19818 to use or discard it.
19820 In the future, GCC may ignore all longcall specifications
19821 when the linker is known to generate glue.
19823 @item -mtls-markers
19824 @itemx -mno-tls-markers
19825 @opindex mtls-markers
19826 @opindex mno-tls-markers
19827 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
19828 specifying the function argument. The relocation allows the linker to
19829 reliably associate function call with argument setup instructions for
19830 TLS optimization, which in turn allows GCC to better schedule the
19835 Adds support for multithreading with the @dfn{pthreads} library.
19836 This option sets flags for both the preprocessor and linker.
19841 This option enables use of the reciprocal estimate and
19842 reciprocal square root estimate instructions with additional
19843 Newton-Raphson steps to increase precision instead of doing a divide or
19844 square root and divide for floating-point arguments. You should use
19845 the @option{-ffast-math} option when using @option{-mrecip} (or at
19846 least @option{-funsafe-math-optimizations},
19847 @option{-finite-math-only}, @option{-freciprocal-math} and
19848 @option{-fno-trapping-math}). Note that while the throughput of the
19849 sequence is generally higher than the throughput of the non-reciprocal
19850 instruction, the precision of the sequence can be decreased by up to 2
19851 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
19854 @item -mrecip=@var{opt}
19855 @opindex mrecip=opt
19856 This option controls which reciprocal estimate instructions
19857 may be used. @var{opt} is a comma-separated list of options, which may
19858 be preceded by a @code{!} to invert the option:
19863 Enable all estimate instructions.
19866 Enable the default instructions, equivalent to @option{-mrecip}.
19869 Disable all estimate instructions, equivalent to @option{-mno-recip}.
19872 Enable the reciprocal approximation instructions for both
19873 single and double precision.
19876 Enable the single-precision reciprocal approximation instructions.
19879 Enable the double-precision reciprocal approximation instructions.
19882 Enable the reciprocal square root approximation instructions for both
19883 single and double precision.
19886 Enable the single-precision reciprocal square root approximation instructions.
19889 Enable the double-precision reciprocal square root approximation instructions.
19893 So, for example, @option{-mrecip=all,!rsqrtd} enables
19894 all of the reciprocal estimate instructions, except for the
19895 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
19896 which handle the double-precision reciprocal square root calculations.
19898 @item -mrecip-precision
19899 @itemx -mno-recip-precision
19900 @opindex mrecip-precision
19901 Assume (do not assume) that the reciprocal estimate instructions
19902 provide higher-precision estimates than is mandated by the PowerPC
19903 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
19904 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
19905 The double-precision square root estimate instructions are not generated by
19906 default on low-precision machines, since they do not provide an
19907 estimate that converges after three steps.
19909 @item -mveclibabi=@var{type}
19910 @opindex mveclibabi
19911 Specifies the ABI type to use for vectorizing intrinsics using an
19912 external library. The only type supported at present is @samp{mass},
19913 which specifies to use IBM's Mathematical Acceleration Subsystem
19914 (MASS) libraries for vectorizing intrinsics using external libraries.
19915 GCC currently emits calls to @code{acosd2}, @code{acosf4},
19916 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
19917 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
19918 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
19919 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
19920 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
19921 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
19922 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
19923 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
19924 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
19925 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
19926 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
19927 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
19928 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
19929 for power7. Both @option{-ftree-vectorize} and
19930 @option{-funsafe-math-optimizations} must also be enabled. The MASS
19931 libraries must be specified at link time.
19936 Generate (do not generate) the @code{friz} instruction when the
19937 @option{-funsafe-math-optimizations} option is used to optimize
19938 rounding of floating-point values to 64-bit integer and back to floating
19939 point. The @code{friz} instruction does not return the same value if
19940 the floating-point number is too large to fit in an integer.
19942 @item -mpointers-to-nested-functions
19943 @itemx -mno-pointers-to-nested-functions
19944 @opindex mpointers-to-nested-functions
19945 Generate (do not generate) code to load up the static chain register
19946 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
19947 systems where a function pointer points to a 3-word descriptor giving
19948 the function address, TOC value to be loaded in register @code{r2}, and
19949 static chain value to be loaded in register @code{r11}. The
19950 @option{-mpointers-to-nested-functions} is on by default. You cannot
19951 call through pointers to nested functions or pointers
19952 to functions compiled in other languages that use the static chain if
19953 you use @option{-mno-pointers-to-nested-functions}.
19955 @item -msave-toc-indirect
19956 @itemx -mno-save-toc-indirect
19957 @opindex msave-toc-indirect
19958 Generate (do not generate) code to save the TOC value in the reserved
19959 stack location in the function prologue if the function calls through
19960 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
19961 saved in the prologue, it is saved just before the call through the
19962 pointer. The @option{-mno-save-toc-indirect} option is the default.
19964 @item -mcompat-align-parm
19965 @itemx -mno-compat-align-parm
19966 @opindex mcompat-align-parm
19967 Generate (do not generate) code to pass structure parameters with a
19968 maximum alignment of 64 bits, for compatibility with older versions
19971 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
19972 structure parameter on a 128-bit boundary when that structure contained
19973 a member requiring 128-bit alignment. This is corrected in more
19974 recent versions of GCC. This option may be used to generate code
19975 that is compatible with functions compiled with older versions of
19978 The @option{-mno-compat-align-parm} option is the default.
19982 @subsection RX Options
19985 These command-line options are defined for RX targets:
19988 @item -m64bit-doubles
19989 @itemx -m32bit-doubles
19990 @opindex m64bit-doubles
19991 @opindex m32bit-doubles
19992 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19993 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19994 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
19995 works on 32-bit values, which is why the default is
19996 @option{-m32bit-doubles}.
20002 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20003 floating-point hardware. The default is enabled for the RX600
20004 series and disabled for the RX200 series.
20006 Floating-point instructions are only generated for 32-bit floating-point
20007 values, however, so the FPU hardware is not used for doubles if the
20008 @option{-m64bit-doubles} option is used.
20010 @emph{Note} If the @option{-fpu} option is enabled then
20011 @option{-funsafe-math-optimizations} is also enabled automatically.
20012 This is because the RX FPU instructions are themselves unsafe.
20014 @item -mcpu=@var{name}
20016 Selects the type of RX CPU to be targeted. Currently three types are
20017 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20018 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20020 The only difference between @samp{RX600} and @samp{RX610} is that the
20021 @samp{RX610} does not support the @code{MVTIPL} instruction.
20023 The @samp{RX200} series does not have a hardware floating-point unit
20024 and so @option{-nofpu} is enabled by default when this type is
20027 @item -mbig-endian-data
20028 @itemx -mlittle-endian-data
20029 @opindex mbig-endian-data
20030 @opindex mlittle-endian-data
20031 Store data (but not code) in the big-endian format. The default is
20032 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20035 @item -msmall-data-limit=@var{N}
20036 @opindex msmall-data-limit
20037 Specifies the maximum size in bytes of global and static variables
20038 which can be placed into the small data area. Using the small data
20039 area can lead to smaller and faster code, but the size of area is
20040 limited and it is up to the programmer to ensure that the area does
20041 not overflow. Also when the small data area is used one of the RX's
20042 registers (usually @code{r13}) is reserved for use pointing to this
20043 area, so it is no longer available for use by the compiler. This
20044 could result in slower and/or larger code if variables are pushed onto
20045 the stack instead of being held in this register.
20047 Note, common variables (variables that have not been initialized) and
20048 constants are not placed into the small data area as they are assigned
20049 to other sections in the output executable.
20051 The default value is zero, which disables this feature. Note, this
20052 feature is not enabled by default with higher optimization levels
20053 (@option{-O2} etc) because of the potentially detrimental effects of
20054 reserving a register. It is up to the programmer to experiment and
20055 discover whether this feature is of benefit to their program. See the
20056 description of the @option{-mpid} option for a description of how the
20057 actual register to hold the small data area pointer is chosen.
20063 Use the simulator runtime. The default is to use the libgloss
20064 board-specific runtime.
20066 @item -mas100-syntax
20067 @itemx -mno-as100-syntax
20068 @opindex mas100-syntax
20069 @opindex mno-as100-syntax
20070 When generating assembler output use a syntax that is compatible with
20071 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20072 assembler, but it has some restrictions so it is not generated by default.
20074 @item -mmax-constant-size=@var{N}
20075 @opindex mmax-constant-size
20076 Specifies the maximum size, in bytes, of a constant that can be used as
20077 an operand in a RX instruction. Although the RX instruction set does
20078 allow constants of up to 4 bytes in length to be used in instructions,
20079 a longer value equates to a longer instruction. Thus in some
20080 circumstances it can be beneficial to restrict the size of constants
20081 that are used in instructions. Constants that are too big are instead
20082 placed into a constant pool and referenced via register indirection.
20084 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20085 or 4 means that constants of any size are allowed.
20089 Enable linker relaxation. Linker relaxation is a process whereby the
20090 linker attempts to reduce the size of a program by finding shorter
20091 versions of various instructions. Disabled by default.
20093 @item -mint-register=@var{N}
20094 @opindex mint-register
20095 Specify the number of registers to reserve for fast interrupt handler
20096 functions. The value @var{N} can be between 0 and 4. A value of 1
20097 means that register @code{r13} is reserved for the exclusive use
20098 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20099 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20100 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20101 A value of 0, the default, does not reserve any registers.
20103 @item -msave-acc-in-interrupts
20104 @opindex msave-acc-in-interrupts
20105 Specifies that interrupt handler functions should preserve the
20106 accumulator register. This is only necessary if normal code might use
20107 the accumulator register, for example because it performs 64-bit
20108 multiplications. The default is to ignore the accumulator as this
20109 makes the interrupt handlers faster.
20115 Enables the generation of position independent data. When enabled any
20116 access to constant data is done via an offset from a base address
20117 held in a register. This allows the location of constant data to be
20118 determined at run time without requiring the executable to be
20119 relocated, which is a benefit to embedded applications with tight
20120 memory constraints. Data that can be modified is not affected by this
20123 Note, using this feature reserves a register, usually @code{r13}, for
20124 the constant data base address. This can result in slower and/or
20125 larger code, especially in complicated functions.
20127 The actual register chosen to hold the constant data base address
20128 depends upon whether the @option{-msmall-data-limit} and/or the
20129 @option{-mint-register} command-line options are enabled. Starting
20130 with register @code{r13} and proceeding downwards, registers are
20131 allocated first to satisfy the requirements of @option{-mint-register},
20132 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20133 is possible for the small data area register to be @code{r8} if both
20134 @option{-mint-register=4} and @option{-mpid} are specified on the
20137 By default this feature is not enabled. The default can be restored
20138 via the @option{-mno-pid} command-line option.
20140 @item -mno-warn-multiple-fast-interrupts
20141 @itemx -mwarn-multiple-fast-interrupts
20142 @opindex mno-warn-multiple-fast-interrupts
20143 @opindex mwarn-multiple-fast-interrupts
20144 Prevents GCC from issuing a warning message if it finds more than one
20145 fast interrupt handler when it is compiling a file. The default is to
20146 issue a warning for each extra fast interrupt handler found, as the RX
20147 only supports one such interrupt.
20149 @item -mallow-string-insns
20150 @itemx -mno-allow-string-insns
20151 @opindex mallow-string-insns
20152 @opindex mno-allow-string-insns
20153 Enables or disables the use of the string manipulation instructions
20154 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20155 @code{SWHILE} and also the @code{RMPA} instruction. These
20156 instructions may prefetch data, which is not safe to do if accessing
20157 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20158 for more information).
20160 The default is to allow these instructions, but it is not possible for
20161 GCC to reliably detect all circumstances where a string instruction
20162 might be used to access an I/O register, so their use cannot be
20163 disabled automatically. Instead it is reliant upon the programmer to
20164 use the @option{-mno-allow-string-insns} option if their program
20165 accesses I/O space.
20167 When the instructions are enabled GCC defines the C preprocessor
20168 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20169 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20172 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20173 has special significance to the RX port when used with the
20174 @code{interrupt} function attribute. This attribute indicates a
20175 function intended to process fast interrupts. GCC ensures
20176 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20177 and/or @code{r13} and only provided that the normal use of the
20178 corresponding registers have been restricted via the
20179 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20182 @node S/390 and zSeries Options
20183 @subsection S/390 and zSeries Options
20184 @cindex S/390 and zSeries Options
20186 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20190 @itemx -msoft-float
20191 @opindex mhard-float
20192 @opindex msoft-float
20193 Use (do not use) the hardware floating-point instructions and registers
20194 for floating-point operations. When @option{-msoft-float} is specified,
20195 functions in @file{libgcc.a} are used to perform floating-point
20196 operations. When @option{-mhard-float} is specified, the compiler
20197 generates IEEE floating-point instructions. This is the default.
20200 @itemx -mno-hard-dfp
20202 @opindex mno-hard-dfp
20203 Use (do not use) the hardware decimal-floating-point instructions for
20204 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20205 specified, functions in @file{libgcc.a} are used to perform
20206 decimal-floating-point operations. When @option{-mhard-dfp} is
20207 specified, the compiler generates decimal-floating-point hardware
20208 instructions. This is the default for @option{-march=z9-ec} or higher.
20210 @item -mlong-double-64
20211 @itemx -mlong-double-128
20212 @opindex mlong-double-64
20213 @opindex mlong-double-128
20214 These switches control the size of @code{long double} type. A size
20215 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20216 type. This is the default.
20219 @itemx -mno-backchain
20220 @opindex mbackchain
20221 @opindex mno-backchain
20222 Store (do not store) the address of the caller's frame as backchain pointer
20223 into the callee's stack frame.
20224 A backchain may be needed to allow debugging using tools that do not understand
20225 DWARF 2 call frame information.
20226 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20227 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20228 the backchain is placed into the topmost word of the 96/160 byte register
20231 In general, code compiled with @option{-mbackchain} is call-compatible with
20232 code compiled with @option{-mmo-backchain}; however, use of the backchain
20233 for debugging purposes usually requires that the whole binary is built with
20234 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20235 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20236 to build a linux kernel use @option{-msoft-float}.
20238 The default is to not maintain the backchain.
20240 @item -mpacked-stack
20241 @itemx -mno-packed-stack
20242 @opindex mpacked-stack
20243 @opindex mno-packed-stack
20244 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20245 specified, the compiler uses the all fields of the 96/160 byte register save
20246 area only for their default purpose; unused fields still take up stack space.
20247 When @option{-mpacked-stack} is specified, register save slots are densely
20248 packed at the top of the register save area; unused space is reused for other
20249 purposes, allowing for more efficient use of the available stack space.
20250 However, when @option{-mbackchain} is also in effect, the topmost word of
20251 the save area is always used to store the backchain, and the return address
20252 register is always saved two words below the backchain.
20254 As long as the stack frame backchain is not used, code generated with
20255 @option{-mpacked-stack} is call-compatible with code generated with
20256 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20257 S/390 or zSeries generated code that uses the stack frame backchain at run
20258 time, not just for debugging purposes. Such code is not call-compatible
20259 with code compiled with @option{-mpacked-stack}. Also, note that the
20260 combination of @option{-mbackchain},
20261 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20262 to build a linux kernel use @option{-msoft-float}.
20264 The default is to not use the packed stack layout.
20267 @itemx -mno-small-exec
20268 @opindex msmall-exec
20269 @opindex mno-small-exec
20270 Generate (or do not generate) code using the @code{bras} instruction
20271 to do subroutine calls.
20272 This only works reliably if the total executable size does not
20273 exceed 64k. The default is to use the @code{basr} instruction instead,
20274 which does not have this limitation.
20280 When @option{-m31} is specified, generate code compliant to the
20281 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20282 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20283 particular to generate 64-bit instructions. For the @samp{s390}
20284 targets, the default is @option{-m31}, while the @samp{s390x}
20285 targets default to @option{-m64}.
20291 When @option{-mzarch} is specified, generate code using the
20292 instructions available on z/Architecture.
20293 When @option{-mesa} is specified, generate code using the
20294 instructions available on ESA/390. Note that @option{-mesa} is
20295 not possible with @option{-m64}.
20296 When generating code compliant to the GNU/Linux for S/390 ABI,
20297 the default is @option{-mesa}. When generating code compliant
20298 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20304 Generate (or do not generate) code using the @code{mvcle} instruction
20305 to perform block moves. When @option{-mno-mvcle} is specified,
20306 use a @code{mvc} loop instead. This is the default unless optimizing for
20313 Print (or do not print) additional debug information when compiling.
20314 The default is to not print debug information.
20316 @item -march=@var{cpu-type}
20318 Generate code that runs on @var{cpu-type}, which is the name of a system
20319 representing a certain processor type. Possible values for
20320 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20321 @samp{z9-109}, @samp{z9-ec}, @samp{z10}, @samp{z196}, and @samp{zEC12}.
20322 When generating code using the instructions available on z/Architecture,
20323 the default is @option{-march=z900}. Otherwise, the default is
20324 @option{-march=g5}.
20326 @item -mtune=@var{cpu-type}
20328 Tune to @var{cpu-type} everything applicable about the generated code,
20329 except for the ABI and the set of available instructions.
20330 The list of @var{cpu-type} values is the same as for @option{-march}.
20331 The default is the value used for @option{-march}.
20334 @itemx -mno-tpf-trace
20335 @opindex mtpf-trace
20336 @opindex mno-tpf-trace
20337 Generate code that adds (does not add) in TPF OS specific branches to trace
20338 routines in the operating system. This option is off by default, even
20339 when compiling for the TPF OS@.
20342 @itemx -mno-fused-madd
20343 @opindex mfused-madd
20344 @opindex mno-fused-madd
20345 Generate code that uses (does not use) the floating-point multiply and
20346 accumulate instructions. These instructions are generated by default if
20347 hardware floating point is used.
20349 @item -mwarn-framesize=@var{framesize}
20350 @opindex mwarn-framesize
20351 Emit a warning if the current function exceeds the given frame size. Because
20352 this is a compile-time check it doesn't need to be a real problem when the program
20353 runs. It is intended to identify functions that most probably cause
20354 a stack overflow. It is useful to be used in an environment with limited stack
20355 size e.g.@: the linux kernel.
20357 @item -mwarn-dynamicstack
20358 @opindex mwarn-dynamicstack
20359 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20360 arrays. This is generally a bad idea with a limited stack size.
20362 @item -mstack-guard=@var{stack-guard}
20363 @itemx -mstack-size=@var{stack-size}
20364 @opindex mstack-guard
20365 @opindex mstack-size
20366 If these options are provided the S/390 back end emits additional instructions in
20367 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20368 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20369 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20370 the frame size of the compiled function is chosen.
20371 These options are intended to be used to help debugging stack overflow problems.
20372 The additionally emitted code causes only little overhead and hence can also be
20373 used in production-like systems without greater performance degradation. The given
20374 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20375 @var{stack-guard} without exceeding 64k.
20376 In order to be efficient the extra code makes the assumption that the stack starts
20377 at an address aligned to the value given by @var{stack-size}.
20378 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20380 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20382 If the hotpatch option is enabled, a ``hot-patching'' function
20383 prologue is generated for all functions in the compilation unit.
20384 The funtion label is prepended with the given number of two-byte
20385 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20386 the label, 2 * @var{post-halfwords} bytes are appended, using the
20387 largest NOP like instructions the architecture allows (maximum
20390 If both arguments are zero, hotpatching is disabled.
20392 This option can be overridden for individual functions with the
20393 @code{hotpatch} attribute.
20396 @node Score Options
20397 @subsection Score Options
20398 @cindex Score Options
20400 These options are defined for Score implementations:
20405 Compile code for big-endian mode. This is the default.
20409 Compile code for little-endian mode.
20413 Disable generation of @code{bcnz} instructions.
20417 Enable generation of unaligned load and store instructions.
20421 Enable the use of multiply-accumulate instructions. Disabled by default.
20425 Specify the SCORE5 as the target architecture.
20429 Specify the SCORE5U of the target architecture.
20433 Specify the SCORE7 as the target architecture. This is the default.
20437 Specify the SCORE7D as the target architecture.
20441 @subsection SH Options
20443 These @samp{-m} options are defined for the SH implementations:
20448 Generate code for the SH1.
20452 Generate code for the SH2.
20455 Generate code for the SH2e.
20459 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20460 that the floating-point unit is not used.
20462 @item -m2a-single-only
20463 @opindex m2a-single-only
20464 Generate code for the SH2a-FPU, in such a way that no double-precision
20465 floating-point operations are used.
20468 @opindex m2a-single
20469 Generate code for the SH2a-FPU assuming the floating-point unit is in
20470 single-precision mode by default.
20474 Generate code for the SH2a-FPU assuming the floating-point unit is in
20475 double-precision mode by default.
20479 Generate code for the SH3.
20483 Generate code for the SH3e.
20487 Generate code for the SH4 without a floating-point unit.
20489 @item -m4-single-only
20490 @opindex m4-single-only
20491 Generate code for the SH4 with a floating-point unit that only
20492 supports single-precision arithmetic.
20496 Generate code for the SH4 assuming the floating-point unit is in
20497 single-precision mode by default.
20501 Generate code for the SH4.
20505 Generate code for SH4-100.
20507 @item -m4-100-nofpu
20508 @opindex m4-100-nofpu
20509 Generate code for SH4-100 in such a way that the
20510 floating-point unit is not used.
20512 @item -m4-100-single
20513 @opindex m4-100-single
20514 Generate code for SH4-100 assuming the floating-point unit is in
20515 single-precision mode by default.
20517 @item -m4-100-single-only
20518 @opindex m4-100-single-only
20519 Generate code for SH4-100 in such a way that no double-precision
20520 floating-point operations are used.
20524 Generate code for SH4-200.
20526 @item -m4-200-nofpu
20527 @opindex m4-200-nofpu
20528 Generate code for SH4-200 without in such a way that the
20529 floating-point unit is not used.
20531 @item -m4-200-single
20532 @opindex m4-200-single
20533 Generate code for SH4-200 assuming the floating-point unit is in
20534 single-precision mode by default.
20536 @item -m4-200-single-only
20537 @opindex m4-200-single-only
20538 Generate code for SH4-200 in such a way that no double-precision
20539 floating-point operations are used.
20543 Generate code for SH4-300.
20545 @item -m4-300-nofpu
20546 @opindex m4-300-nofpu
20547 Generate code for SH4-300 without in such a way that the
20548 floating-point unit is not used.
20550 @item -m4-300-single
20551 @opindex m4-300-single
20552 Generate code for SH4-300 in such a way that no double-precision
20553 floating-point operations are used.
20555 @item -m4-300-single-only
20556 @opindex m4-300-single-only
20557 Generate code for SH4-300 in such a way that no double-precision
20558 floating-point operations are used.
20562 Generate code for SH4-340 (no MMU, no FPU).
20566 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20571 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20572 floating-point unit is not used.
20574 @item -m4a-single-only
20575 @opindex m4a-single-only
20576 Generate code for the SH4a, in such a way that no double-precision
20577 floating-point operations are used.
20580 @opindex m4a-single
20581 Generate code for the SH4a assuming the floating-point unit is in
20582 single-precision mode by default.
20586 Generate code for the SH4a.
20590 Same as @option{-m4a-nofpu}, except that it implicitly passes
20591 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20592 instructions at the moment.
20595 @opindex m5-32media
20596 Generate 32-bit code for SHmedia.
20598 @item -m5-32media-nofpu
20599 @opindex m5-32media-nofpu
20600 Generate 32-bit code for SHmedia in such a way that the
20601 floating-point unit is not used.
20604 @opindex m5-64media
20605 Generate 64-bit code for SHmedia.
20607 @item -m5-64media-nofpu
20608 @opindex m5-64media-nofpu
20609 Generate 64-bit code for SHmedia in such a way that the
20610 floating-point unit is not used.
20613 @opindex m5-compact
20614 Generate code for SHcompact.
20616 @item -m5-compact-nofpu
20617 @opindex m5-compact-nofpu
20618 Generate code for SHcompact in such a way that the
20619 floating-point unit is not used.
20623 Compile code for the processor in big-endian mode.
20627 Compile code for the processor in little-endian mode.
20631 Align doubles at 64-bit boundaries. Note that this changes the calling
20632 conventions, and thus some functions from the standard C library do
20633 not work unless you recompile it first with @option{-mdalign}.
20637 Shorten some address references at link time, when possible; uses the
20638 linker option @option{-relax}.
20642 Use 32-bit offsets in @code{switch} tables. The default is to use
20647 Enable the use of bit manipulation instructions on SH2A.
20651 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20652 alignment constraints.
20656 Comply with the calling conventions defined by Renesas.
20659 @opindex mno-renesas
20660 Comply with the calling conventions defined for GCC before the Renesas
20661 conventions were available. This option is the default for all
20662 targets of the SH toolchain.
20665 @opindex mnomacsave
20666 Mark the @code{MAC} register as call-clobbered, even if
20667 @option{-mrenesas} is given.
20673 Control the IEEE compliance of floating-point comparisons, which affects the
20674 handling of cases where the result of a comparison is unordered. By default
20675 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20676 enabled @option{-mno-ieee} is implicitly set, which results in faster
20677 floating-point greater-equal and less-equal comparisons. The implcit settings
20678 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20680 @item -minline-ic_invalidate
20681 @opindex minline-ic_invalidate
20682 Inline code to invalidate instruction cache entries after setting up
20683 nested function trampolines.
20684 This option has no effect if @option{-musermode} is in effect and the selected
20685 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20687 If the selected code generation option does not allow the use of the @code{icbi}
20688 instruction, and @option{-musermode} is not in effect, the inlined code
20689 manipulates the instruction cache address array directly with an associative
20690 write. This not only requires privileged mode at run time, but it also
20691 fails if the cache line had been mapped via the TLB and has become unmapped.
20695 Dump instruction size and location in the assembly code.
20698 @opindex mpadstruct
20699 This option is deprecated. It pads structures to multiple of 4 bytes,
20700 which is incompatible with the SH ABI@.
20702 @item -matomic-model=@var{model}
20703 @opindex matomic-model=@var{model}
20704 Sets the model of atomic operations and additional parameters as a comma
20705 separated list. For details on the atomic built-in functions see
20706 @ref{__atomic Builtins}. The following models and parameters are supported:
20711 Disable compiler generated atomic sequences and emit library calls for atomic
20712 operations. This is the default if the target is not @code{sh*-*-linux*}.
20715 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20716 built-in functions. The generated atomic sequences require additional support
20717 from the interrupt/exception handling code of the system and are only suitable
20718 for SH3* and SH4* single-core systems. This option is enabled by default when
20719 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20720 this option also partially utilizes the hardware atomic instructions
20721 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20722 @samp{strict} is specified.
20725 Generate software atomic sequences that use a variable in the thread control
20726 block. This is a variation of the gUSA sequences which can also be used on
20727 SH1* and SH2* targets. The generated atomic sequences require additional
20728 support from the interrupt/exception handling code of the system and are only
20729 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20730 parameter has to be specified as well.
20733 Generate software atomic sequences that temporarily disable interrupts by
20734 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20735 in privileged mode and is only suitable for single-core systems. Additional
20736 support from the interrupt/exception handling code of the system is not
20737 required. This model is enabled by default when the target is
20738 @code{sh*-*-linux*} and SH1* or SH2*.
20741 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20742 instructions only. This is only available on SH4A and is suitable for
20743 multi-core systems. Since the hardware instructions support only 32 bit atomic
20744 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20745 Code compiled with this option is also compatible with other software
20746 atomic model interrupt/exception handling systems if executed on an SH4A
20747 system. Additional support from the interrupt/exception handling code of the
20748 system is not required for this model.
20751 This parameter specifies the offset in bytes of the variable in the thread
20752 control block structure that should be used by the generated atomic sequences
20753 when the @samp{soft-tcb} model has been selected. For other models this
20754 parameter is ignored. The specified value must be an integer multiple of four
20755 and in the range 0-1020.
20758 This parameter prevents mixed usage of multiple atomic models, even if they
20759 are compatible, and makes the compiler generate atomic sequences of the
20760 specified model only.
20766 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20767 Notice that depending on the particular hardware and software configuration
20768 this can degrade overall performance due to the operand cache line flushes
20769 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20770 processors the @code{tas.b} instruction must be used with caution since it
20771 can result in data corruption for certain cache configurations.
20774 @opindex mprefergot
20775 When generating position-independent code, emit function calls using
20776 the Global Offset Table instead of the Procedure Linkage Table.
20779 @itemx -mno-usermode
20781 @opindex mno-usermode
20782 Don't allow (allow) the compiler generating privileged mode code. Specifying
20783 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20784 inlined code would not work in user mode. @option{-musermode} is the default
20785 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
20786 @option{-musermode} has no effect, since there is no user mode.
20788 @item -multcost=@var{number}
20789 @opindex multcost=@var{number}
20790 Set the cost to assume for a multiply insn.
20792 @item -mdiv=@var{strategy}
20793 @opindex mdiv=@var{strategy}
20794 Set the division strategy to be used for integer division operations.
20795 For SHmedia @var{strategy} can be one of:
20800 Performs the operation in floating point. This has a very high latency,
20801 but needs only a few instructions, so it might be a good choice if
20802 your code has enough easily-exploitable ILP to allow the compiler to
20803 schedule the floating-point instructions together with other instructions.
20804 Division by zero causes a floating-point exception.
20807 Uses integer operations to calculate the inverse of the divisor,
20808 and then multiplies the dividend with the inverse. This strategy allows
20809 CSE and hoisting of the inverse calculation. Division by zero calculates
20810 an unspecified result, but does not trap.
20813 A variant of @samp{inv} where, if no CSE or hoisting opportunities
20814 have been found, or if the entire operation has been hoisted to the same
20815 place, the last stages of the inverse calculation are intertwined with the
20816 final multiply to reduce the overall latency, at the expense of using a few
20817 more instructions, and thus offering fewer scheduling opportunities with
20821 Calls a library function that usually implements the @samp{inv:minlat}
20823 This gives high code density for @code{m5-*media-nofpu} compilations.
20826 Uses a different entry point of the same library function, where it
20827 assumes that a pointer to a lookup table has already been set up, which
20828 exposes the pointer load to CSE and code hoisting optimizations.
20833 Use the @samp{inv} algorithm for initial
20834 code generation, but if the code stays unoptimized, revert to the @samp{call},
20835 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
20836 potentially-trapping side effect of division by zero is carried by a
20837 separate instruction, so it is possible that all the integer instructions
20838 are hoisted out, but the marker for the side effect stays where it is.
20839 A recombination to floating-point operations or a call is not possible
20844 Variants of the @samp{inv:minlat} strategy. In the case
20845 that the inverse calculation is not separated from the multiply, they speed
20846 up division where the dividend fits into 20 bits (plus sign where applicable)
20847 by inserting a test to skip a number of operations in this case; this test
20848 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
20849 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
20853 For targets other than SHmedia @var{strategy} can be one of:
20858 Calls a library function that uses the single-step division instruction
20859 @code{div1} to perform the operation. Division by zero calculates an
20860 unspecified result and does not trap. This is the default except for SH4,
20861 SH2A and SHcompact.
20864 Calls a library function that performs the operation in double precision
20865 floating point. Division by zero causes a floating-point exception. This is
20866 the default for SHcompact with FPU. Specifying this for targets that do not
20867 have a double precision FPU defaults to @code{call-div1}.
20870 Calls a library function that uses a lookup table for small divisors and
20871 the @code{div1} instruction with case distinction for larger divisors. Division
20872 by zero calculates an unspecified result and does not trap. This is the default
20873 for SH4. Specifying this for targets that do not have dynamic shift
20874 instructions defaults to @code{call-div1}.
20878 When a division strategy has not been specified the default strategy is
20879 selected based on the current target. For SH2A the default strategy is to
20880 use the @code{divs} and @code{divu} instructions instead of library function
20883 @item -maccumulate-outgoing-args
20884 @opindex maccumulate-outgoing-args
20885 Reserve space once for outgoing arguments in the function prologue rather
20886 than around each call. Generally beneficial for performance and size. Also
20887 needed for unwinding to avoid changing the stack frame around conditional code.
20889 @item -mdivsi3_libfunc=@var{name}
20890 @opindex mdivsi3_libfunc=@var{name}
20891 Set the name of the library function used for 32-bit signed division to
20893 This only affects the name used in the @samp{call} and @samp{inv:call}
20894 division strategies, and the compiler still expects the same
20895 sets of input/output/clobbered registers as if this option were not present.
20897 @item -mfixed-range=@var{register-range}
20898 @opindex mfixed-range
20899 Generate code treating the given register range as fixed registers.
20900 A fixed register is one that the register allocator can not use. This is
20901 useful when compiling kernel code. A register range is specified as
20902 two registers separated by a dash. Multiple register ranges can be
20903 specified separated by a comma.
20905 @item -mindexed-addressing
20906 @opindex mindexed-addressing
20907 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
20908 This is only safe if the hardware and/or OS implement 32-bit wrap-around
20909 semantics for the indexed addressing mode. The architecture allows the
20910 implementation of processors with 64-bit MMU, which the OS could use to
20911 get 32-bit addressing, but since no current hardware implementation supports
20912 this or any other way to make the indexed addressing mode safe to use in
20913 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
20915 @item -mgettrcost=@var{number}
20916 @opindex mgettrcost=@var{number}
20917 Set the cost assumed for the @code{gettr} instruction to @var{number}.
20918 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
20922 Assume @code{pt*} instructions won't trap. This generally generates
20923 better-scheduled code, but is unsafe on current hardware.
20924 The current architecture
20925 definition says that @code{ptabs} and @code{ptrel} trap when the target
20927 This has the unintentional effect of making it unsafe to schedule these
20928 instructions before a branch, or hoist them out of a loop. For example,
20929 @code{__do_global_ctors}, a part of @file{libgcc}
20930 that runs constructors at program
20931 startup, calls functions in a list which is delimited by @minus{}1. With the
20932 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
20933 That means that all the constructors run a bit more quickly, but when
20934 the loop comes to the end of the list, the program crashes because @code{ptabs}
20935 loads @minus{}1 into a target register.
20937 Since this option is unsafe for any
20938 hardware implementing the current architecture specification, the default
20939 is @option{-mno-pt-fixed}. Unless specified explicitly with
20940 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
20941 this deters register allocation from using target registers for storing
20944 @item -minvalid-symbols
20945 @opindex minvalid-symbols
20946 Assume symbols might be invalid. Ordinary function symbols generated by
20947 the compiler are always valid to load with
20948 @code{movi}/@code{shori}/@code{ptabs} or
20949 @code{movi}/@code{shori}/@code{ptrel},
20950 but with assembler and/or linker tricks it is possible
20951 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
20952 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
20953 It prevents cross-basic-block CSE, hoisting and most scheduling
20954 of symbol loads. The default is @option{-mno-invalid-symbols}.
20956 @item -mbranch-cost=@var{num}
20957 @opindex mbranch-cost=@var{num}
20958 Assume @var{num} to be the cost for a branch instruction. Higher numbers
20959 make the compiler try to generate more branch-free code if possible.
20960 If not specified the value is selected depending on the processor type that
20961 is being compiled for.
20964 @itemx -mno-zdcbranch
20965 @opindex mzdcbranch
20966 @opindex mno-zdcbranch
20967 Assume (do not assume) that zero displacement conditional branch instructions
20968 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
20969 compiler prefers zero displacement branch code sequences. This is
20970 enabled by default when generating code for SH4 and SH4A. It can be explicitly
20971 disabled by specifying @option{-mno-zdcbranch}.
20973 @item -mcbranch-force-delay-slot
20974 @opindex mcbranch-force-delay-slot
20975 Force the usage of delay slots for conditional branches, which stuffs the delay
20976 slot with a @code{nop} if a suitable instruction can't be found. By default
20977 this option is disabled. It can be enabled to work around hardware bugs as
20978 found in the original SH7055.
20981 @itemx -mno-fused-madd
20982 @opindex mfused-madd
20983 @opindex mno-fused-madd
20984 Generate code that uses (does not use) the floating-point multiply and
20985 accumulate instructions. These instructions are generated by default
20986 if hardware floating point is used. The machine-dependent
20987 @option{-mfused-madd} option is now mapped to the machine-independent
20988 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20989 mapped to @option{-ffp-contract=off}.
20995 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
20996 and cosine approximations. The option @option{-mfsca} must be used in
20997 combination with @option{-funsafe-math-optimizations}. It is enabled by default
20998 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
20999 approximations even if @option{-funsafe-math-optimizations} is in effect.
21005 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21006 reciprocal square root approximations. The option @option{-mfsrra} must be used
21007 in combination with @option{-funsafe-math-optimizations} and
21008 @option{-ffinite-math-only}. It is enabled by default when generating code for
21009 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21010 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21013 @item -mpretend-cmove
21014 @opindex mpretend-cmove
21015 Prefer zero-displacement conditional branches for conditional move instruction
21016 patterns. This can result in faster code on the SH4 processor.
21020 @node Solaris 2 Options
21021 @subsection Solaris 2 Options
21022 @cindex Solaris 2 options
21024 These @samp{-m} options are supported on Solaris 2:
21027 @item -mclear-hwcap
21028 @opindex mclear-hwcap
21029 @option{-mclear-hwcap} tells the compiler to remove the hardware
21030 capabilities generated by the Solaris assembler. This is only necessary
21031 when object files use ISA extensions not supported by the current
21032 machine, but check at runtime whether or not to use them.
21034 @item -mimpure-text
21035 @opindex mimpure-text
21036 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21037 the compiler to not pass @option{-z text} to the linker when linking a
21038 shared object. Using this option, you can link position-dependent
21039 code into a shared object.
21041 @option{-mimpure-text} suppresses the ``relocations remain against
21042 allocatable but non-writable sections'' linker error message.
21043 However, the necessary relocations trigger copy-on-write, and the
21044 shared object is not actually shared across processes. Instead of
21045 using @option{-mimpure-text}, you should compile all source code with
21046 @option{-fpic} or @option{-fPIC}.
21050 These switches are supported in addition to the above on Solaris 2:
21055 Add support for multithreading using the POSIX threads library. This
21056 option sets flags for both the preprocessor and linker. This option does
21057 not affect the thread safety of object code produced by the compiler or
21058 that of libraries supplied with it.
21062 This is a synonym for @option{-pthreads}.
21065 @node SPARC Options
21066 @subsection SPARC Options
21067 @cindex SPARC options
21069 These @samp{-m} options are supported on the SPARC:
21072 @item -mno-app-regs
21074 @opindex mno-app-regs
21076 Specify @option{-mapp-regs} to generate output using the global registers
21077 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21078 global register 1, each global register 2 through 4 is then treated as an
21079 allocable register that is clobbered by function calls. This is the default.
21081 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21082 specify @option{-mno-app-regs}. You should compile libraries and system
21083 software with this option.
21089 With @option{-mflat}, the compiler does not generate save/restore instructions
21090 and uses a ``flat'' or single register window model. This model is compatible
21091 with the regular register window model. The local registers and the input
21092 registers (0--5) are still treated as ``call-saved'' registers and are
21093 saved on the stack as needed.
21095 With @option{-mno-flat} (the default), the compiler generates save/restore
21096 instructions (except for leaf functions). This is the normal operating mode.
21099 @itemx -mhard-float
21101 @opindex mhard-float
21102 Generate output containing floating-point instructions. This is the
21106 @itemx -msoft-float
21108 @opindex msoft-float
21109 Generate output containing library calls for floating point.
21110 @strong{Warning:} the requisite libraries are not available for all SPARC
21111 targets. Normally the facilities of the machine's usual C compiler are
21112 used, but this cannot be done directly in cross-compilation. You must make
21113 your own arrangements to provide suitable library functions for
21114 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21115 @samp{sparclite-*-*} do provide software floating-point support.
21117 @option{-msoft-float} changes the calling convention in the output file;
21118 therefore, it is only useful if you compile @emph{all} of a program with
21119 this option. In particular, you need to compile @file{libgcc.a}, the
21120 library that comes with GCC, with @option{-msoft-float} in order for
21123 @item -mhard-quad-float
21124 @opindex mhard-quad-float
21125 Generate output containing quad-word (long double) floating-point
21128 @item -msoft-quad-float
21129 @opindex msoft-quad-float
21130 Generate output containing library calls for quad-word (long double)
21131 floating-point instructions. The functions called are those specified
21132 in the SPARC ABI@. This is the default.
21134 As of this writing, there are no SPARC implementations that have hardware
21135 support for the quad-word floating-point instructions. They all invoke
21136 a trap handler for one of these instructions, and then the trap handler
21137 emulates the effect of the instruction. Because of the trap handler overhead,
21138 this is much slower than calling the ABI library routines. Thus the
21139 @option{-msoft-quad-float} option is the default.
21141 @item -mno-unaligned-doubles
21142 @itemx -munaligned-doubles
21143 @opindex mno-unaligned-doubles
21144 @opindex munaligned-doubles
21145 Assume that doubles have 8-byte alignment. This is the default.
21147 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21148 alignment only if they are contained in another type, or if they have an
21149 absolute address. Otherwise, it assumes they have 4-byte alignment.
21150 Specifying this option avoids some rare compatibility problems with code
21151 generated by other compilers. It is not the default because it results
21152 in a performance loss, especially for floating-point code.
21155 @itemx -mno-user-mode
21156 @opindex muser-mode
21157 @opindex mno-user-mode
21158 Do not generate code that can only run in supervisor mode. This is relevant
21159 only for the @code{casa} instruction emitted for the LEON3 processor. The
21160 default is @option{-mno-user-mode}.
21162 @item -mno-faster-structs
21163 @itemx -mfaster-structs
21164 @opindex mno-faster-structs
21165 @opindex mfaster-structs
21166 With @option{-mfaster-structs}, the compiler assumes that structures
21167 should have 8-byte alignment. This enables the use of pairs of
21168 @code{ldd} and @code{std} instructions for copies in structure
21169 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21170 However, the use of this changed alignment directly violates the SPARC
21171 ABI@. Thus, it's intended only for use on targets where the developer
21172 acknowledges that their resulting code is not directly in line with
21173 the rules of the ABI@.
21175 @item -mcpu=@var{cpu_type}
21177 Set the instruction set, register set, and instruction scheduling parameters
21178 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21179 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21180 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21181 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21182 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21183 @samp{niagara3} and @samp{niagara4}.
21185 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21186 which selects the best architecture option for the host processor.
21187 @option{-mcpu=native} has no effect if GCC does not recognize
21190 Default instruction scheduling parameters are used for values that select
21191 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21192 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21194 Here is a list of each supported architecture and their supported
21202 supersparc, hypersparc, leon, leon3
21205 f930, f934, sparclite86x
21211 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21214 By default (unless configured otherwise), GCC generates code for the V7
21215 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21216 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21217 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21218 SPARCStation 1, 2, IPX etc.
21220 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21221 architecture. The only difference from V7 code is that the compiler emits
21222 the integer multiply and integer divide instructions which exist in SPARC-V8
21223 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21224 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21227 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21228 the SPARC architecture. This adds the integer multiply, integer divide step
21229 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21230 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21231 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21232 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21233 MB86934 chip, which is the more recent SPARClite with FPU@.
21235 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21236 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21237 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21238 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21239 optimizes it for the TEMIC SPARClet chip.
21241 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21242 architecture. This adds 64-bit integer and floating-point move instructions,
21243 3 additional floating-point condition code registers and conditional move
21244 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21245 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21246 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21247 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21248 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21249 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21250 additionally optimizes it for Sun UltraSPARC T2 chips. With
21251 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21252 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21253 additionally optimizes it for Sun UltraSPARC T4 chips.
21255 @item -mtune=@var{cpu_type}
21257 Set the instruction scheduling parameters for machine type
21258 @var{cpu_type}, but do not set the instruction set or register set that the
21259 option @option{-mcpu=@var{cpu_type}} does.
21261 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21262 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21263 that select a particular CPU implementation. Those are @samp{cypress},
21264 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21265 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21266 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21267 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21268 toolchains, @samp{native} can also be used.
21273 @opindex mno-v8plus
21274 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21275 difference from the V8 ABI is that the global and out registers are
21276 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21277 mode for all SPARC-V9 processors.
21283 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21284 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21290 With @option{-mvis2}, GCC generates code that takes advantage of
21291 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21292 default is @option{-mvis2} when targeting a cpu that supports such
21293 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21294 also sets @option{-mvis}.
21300 With @option{-mvis3}, GCC generates code that takes advantage of
21301 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21302 default is @option{-mvis3} when targeting a cpu that supports such
21303 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21304 also sets @option{-mvis2} and @option{-mvis}.
21309 @opindex mno-cbcond
21310 With @option{-mcbcond}, GCC generates code that takes advantage of
21311 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21312 The default is @option{-mcbcond} when targeting a cpu that supports such
21313 instructions, such as niagara-4 and later.
21319 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21320 population count instruction. The default is @option{-mpopc}
21321 when targeting a cpu that supports such instructions, such as Niagara-2 and
21328 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21329 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21330 when targeting a cpu that supports such instructions, such as Niagara-3 and
21334 @opindex mfix-at697f
21335 Enable the documented workaround for the single erratum of the Atmel AT697F
21336 processor (which corresponds to erratum #13 of the AT697E processor).
21339 @opindex mfix-ut699
21340 Enable the documented workarounds for the floating-point errata and the data
21341 cache nullify errata of the UT699 processor.
21344 These @samp{-m} options are supported in addition to the above
21345 on SPARC-V9 processors in 64-bit environments:
21352 Generate code for a 32-bit or 64-bit environment.
21353 The 32-bit environment sets int, long and pointer to 32 bits.
21354 The 64-bit environment sets int to 32 bits and long and pointer
21357 @item -mcmodel=@var{which}
21359 Set the code model to one of
21363 The Medium/Low code model: 64-bit addresses, programs
21364 must be linked in the low 32 bits of memory. Programs can be statically
21365 or dynamically linked.
21368 The Medium/Middle code model: 64-bit addresses, programs
21369 must be linked in the low 44 bits of memory, the text and data segments must
21370 be less than 2GB in size and the data segment must be located within 2GB of
21374 The Medium/Anywhere code model: 64-bit addresses, programs
21375 may be linked anywhere in memory, the text and data segments must be less
21376 than 2GB in size and the data segment must be located within 2GB of the
21380 The Medium/Anywhere code model for embedded systems:
21381 64-bit addresses, the text and data segments must be less than 2GB in
21382 size, both starting anywhere in memory (determined at link time). The
21383 global register %g4 points to the base of the data segment. Programs
21384 are statically linked and PIC is not supported.
21387 @item -mmemory-model=@var{mem-model}
21388 @opindex mmemory-model
21389 Set the memory model in force on the processor to one of
21393 The default memory model for the processor and operating system.
21396 Relaxed Memory Order
21399 Partial Store Order
21405 Sequential Consistency
21408 These memory models are formally defined in Appendix D of the Sparc V9
21409 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21412 @itemx -mno-stack-bias
21413 @opindex mstack-bias
21414 @opindex mno-stack-bias
21415 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21416 frame pointer if present, are offset by @minus{}2047 which must be added back
21417 when making stack frame references. This is the default in 64-bit mode.
21418 Otherwise, assume no such offset is present.
21422 @subsection SPU Options
21423 @cindex SPU options
21425 These @samp{-m} options are supported on the SPU:
21429 @itemx -merror-reloc
21430 @opindex mwarn-reloc
21431 @opindex merror-reloc
21433 The loader for SPU does not handle dynamic relocations. By default, GCC
21434 gives an error when it generates code that requires a dynamic
21435 relocation. @option{-mno-error-reloc} disables the error,
21436 @option{-mwarn-reloc} generates a warning instead.
21439 @itemx -munsafe-dma
21441 @opindex munsafe-dma
21443 Instructions that initiate or test completion of DMA must not be
21444 reordered with respect to loads and stores of the memory that is being
21446 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21447 memory accesses, but that can lead to inefficient code in places where the
21448 memory is known to not change. Rather than mark the memory as volatile,
21449 you can use @option{-msafe-dma} to tell the compiler to treat
21450 the DMA instructions as potentially affecting all memory.
21452 @item -mbranch-hints
21453 @opindex mbranch-hints
21455 By default, GCC generates a branch hint instruction to avoid
21456 pipeline stalls for always-taken or probably-taken branches. A hint
21457 is not generated closer than 8 instructions away from its branch.
21458 There is little reason to disable them, except for debugging purposes,
21459 or to make an object a little bit smaller.
21463 @opindex msmall-mem
21464 @opindex mlarge-mem
21466 By default, GCC generates code assuming that addresses are never larger
21467 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21468 a full 32-bit address.
21473 By default, GCC links against startup code that assumes the SPU-style
21474 main function interface (which has an unconventional parameter list).
21475 With @option{-mstdmain}, GCC links your program against startup
21476 code that assumes a C99-style interface to @code{main}, including a
21477 local copy of @code{argv} strings.
21479 @item -mfixed-range=@var{register-range}
21480 @opindex mfixed-range
21481 Generate code treating the given register range as fixed registers.
21482 A fixed register is one that the register allocator cannot use. This is
21483 useful when compiling kernel code. A register range is specified as
21484 two registers separated by a dash. Multiple register ranges can be
21485 specified separated by a comma.
21491 Compile code assuming that pointers to the PPU address space accessed
21492 via the @code{__ea} named address space qualifier are either 32 or 64
21493 bits wide. The default is 32 bits. As this is an ABI-changing option,
21494 all object code in an executable must be compiled with the same setting.
21496 @item -maddress-space-conversion
21497 @itemx -mno-address-space-conversion
21498 @opindex maddress-space-conversion
21499 @opindex mno-address-space-conversion
21500 Allow/disallow treating the @code{__ea} address space as superset
21501 of the generic address space. This enables explicit type casts
21502 between @code{__ea} and generic pointer as well as implicit
21503 conversions of generic pointers to @code{__ea} pointers. The
21504 default is to allow address space pointer conversions.
21506 @item -mcache-size=@var{cache-size}
21507 @opindex mcache-size
21508 This option controls the version of libgcc that the compiler links to an
21509 executable and selects a software-managed cache for accessing variables
21510 in the @code{__ea} address space with a particular cache size. Possible
21511 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21512 and @samp{128}. The default cache size is 64KB.
21514 @item -matomic-updates
21515 @itemx -mno-atomic-updates
21516 @opindex matomic-updates
21517 @opindex mno-atomic-updates
21518 This option controls the version of libgcc that the compiler links to an
21519 executable and selects whether atomic updates to the software-managed
21520 cache of PPU-side variables are used. If you use atomic updates, changes
21521 to a PPU variable from SPU code using the @code{__ea} named address space
21522 qualifier do not interfere with changes to other PPU variables residing
21523 in the same cache line from PPU code. If you do not use atomic updates,
21524 such interference may occur; however, writing back cache lines is
21525 more efficient. The default behavior is to use atomic updates.
21528 @itemx -mdual-nops=@var{n}
21529 @opindex mdual-nops
21530 By default, GCC inserts nops to increase dual issue when it expects
21531 it to increase performance. @var{n} can be a value from 0 to 10. A
21532 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21533 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21535 @item -mhint-max-nops=@var{n}
21536 @opindex mhint-max-nops
21537 Maximum number of nops to insert for a branch hint. A branch hint must
21538 be at least 8 instructions away from the branch it is affecting. GCC
21539 inserts up to @var{n} nops to enforce this, otherwise it does not
21540 generate the branch hint.
21542 @item -mhint-max-distance=@var{n}
21543 @opindex mhint-max-distance
21544 The encoding of the branch hint instruction limits the hint to be within
21545 256 instructions of the branch it is affecting. By default, GCC makes
21546 sure it is within 125.
21549 @opindex msafe-hints
21550 Work around a hardware bug that causes the SPU to stall indefinitely.
21551 By default, GCC inserts the @code{hbrp} instruction to make sure
21552 this stall won't happen.
21556 @node System V Options
21557 @subsection Options for System V
21559 These additional options are available on System V Release 4 for
21560 compatibility with other compilers on those systems:
21565 Create a shared object.
21566 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21570 Identify the versions of each tool used by the compiler, in a
21571 @code{.ident} assembler directive in the output.
21575 Refrain from adding @code{.ident} directives to the output file (this is
21578 @item -YP,@var{dirs}
21580 Search the directories @var{dirs}, and no others, for libraries
21581 specified with @option{-l}.
21583 @item -Ym,@var{dir}
21585 Look in the directory @var{dir} to find the M4 preprocessor.
21586 The assembler uses this option.
21587 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21588 @c the generic assembler that comes with Solaris takes just -Ym.
21591 @node TILE-Gx Options
21592 @subsection TILE-Gx Options
21593 @cindex TILE-Gx options
21595 These @samp{-m} options are supported on the TILE-Gx:
21598 @item -mcmodel=small
21599 @opindex mcmodel=small
21600 Generate code for the small model. The distance for direct calls is
21601 limited to 500M in either direction. PC-relative addresses are 32
21602 bits. Absolute addresses support the full address range.
21604 @item -mcmodel=large
21605 @opindex mcmodel=large
21606 Generate code for the large model. There is no limitation on call
21607 distance, pc-relative addresses, or absolute addresses.
21609 @item -mcpu=@var{name}
21611 Selects the type of CPU to be targeted. Currently the only supported
21612 type is @samp{tilegx}.
21618 Generate code for a 32-bit or 64-bit environment. The 32-bit
21619 environment sets int, long, and pointer to 32 bits. The 64-bit
21620 environment sets int to 32 bits and long and pointer to 64 bits.
21623 @itemx -mlittle-endian
21624 @opindex mbig-endian
21625 @opindex mlittle-endian
21626 Generate code in big/little endian mode, respectively.
21629 @node TILEPro Options
21630 @subsection TILEPro Options
21631 @cindex TILEPro options
21633 These @samp{-m} options are supported on the TILEPro:
21636 @item -mcpu=@var{name}
21638 Selects the type of CPU to be targeted. Currently the only supported
21639 type is @samp{tilepro}.
21643 Generate code for a 32-bit environment, which sets int, long, and
21644 pointer to 32 bits. This is the only supported behavior so the flag
21645 is essentially ignored.
21649 @subsection V850 Options
21650 @cindex V850 Options
21652 These @samp{-m} options are defined for V850 implementations:
21656 @itemx -mno-long-calls
21657 @opindex mlong-calls
21658 @opindex mno-long-calls
21659 Treat all calls as being far away (near). If calls are assumed to be
21660 far away, the compiler always loads the function's address into a
21661 register, and calls indirect through the pointer.
21667 Do not optimize (do optimize) basic blocks that use the same index
21668 pointer 4 or more times to copy pointer into the @code{ep} register, and
21669 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21670 option is on by default if you optimize.
21672 @item -mno-prolog-function
21673 @itemx -mprolog-function
21674 @opindex mno-prolog-function
21675 @opindex mprolog-function
21676 Do not use (do use) external functions to save and restore registers
21677 at the prologue and epilogue of a function. The external functions
21678 are slower, but use less code space if more than one function saves
21679 the same number of registers. The @option{-mprolog-function} option
21680 is on by default if you optimize.
21684 Try to make the code as small as possible. At present, this just turns
21685 on the @option{-mep} and @option{-mprolog-function} options.
21687 @item -mtda=@var{n}
21689 Put static or global variables whose size is @var{n} bytes or less into
21690 the tiny data area that register @code{ep} points to. The tiny data
21691 area can hold up to 256 bytes in total (128 bytes for byte references).
21693 @item -msda=@var{n}
21695 Put static or global variables whose size is @var{n} bytes or less into
21696 the small data area that register @code{gp} points to. The small data
21697 area can hold up to 64 kilobytes.
21699 @item -mzda=@var{n}
21701 Put static or global variables whose size is @var{n} bytes or less into
21702 the first 32 kilobytes of memory.
21706 Specify that the target processor is the V850.
21710 Specify that the target processor is the V850E3V5. The preprocessor
21711 constant @code{__v850e3v5__} is defined if this option is used.
21715 Specify that the target processor is the V850E3V5. This is an alias for
21716 the @option{-mv850e3v5} option.
21720 Specify that the target processor is the V850E2V3. The preprocessor
21721 constant @code{__v850e2v3__} is defined if this option is used.
21725 Specify that the target processor is the V850E2. The preprocessor
21726 constant @code{__v850e2__} is defined if this option is used.
21730 Specify that the target processor is the V850E1. The preprocessor
21731 constants @code{__v850e1__} and @code{__v850e__} are defined if
21732 this option is used.
21736 Specify that the target processor is the V850ES. This is an alias for
21737 the @option{-mv850e1} option.
21741 Specify that the target processor is the V850E@. The preprocessor
21742 constant @code{__v850e__} is defined if this option is used.
21744 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21745 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21746 are defined then a default target processor is chosen and the
21747 relevant @samp{__v850*__} preprocessor constant is defined.
21749 The preprocessor constants @code{__v850} and @code{__v851__} are always
21750 defined, regardless of which processor variant is the target.
21752 @item -mdisable-callt
21753 @itemx -mno-disable-callt
21754 @opindex mdisable-callt
21755 @opindex mno-disable-callt
21756 This option suppresses generation of the @code{CALLT} instruction for the
21757 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21760 This option is enabled by default when the RH850 ABI is
21761 in use (see @option{-mrh850-abi}), and disabled by default when the
21762 GCC ABI is in use. If @code{CALLT} instructions are being generated
21763 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21769 Pass on (or do not pass on) the @option{-mrelax} command-line option
21773 @itemx -mno-long-jumps
21774 @opindex mlong-jumps
21775 @opindex mno-long-jumps
21776 Disable (or re-enable) the generation of PC-relative jump instructions.
21779 @itemx -mhard-float
21780 @opindex msoft-float
21781 @opindex mhard-float
21782 Disable (or re-enable) the generation of hardware floating point
21783 instructions. This option is only significant when the target
21784 architecture is @samp{V850E2V3} or higher. If hardware floating point
21785 instructions are being generated then the C preprocessor symbol
21786 @code{__FPU_OK__} is defined, otherwise the symbol
21787 @code{__NO_FPU__} is defined.
21791 Enables the use of the e3v5 LOOP instruction. The use of this
21792 instruction is not enabled by default when the e3v5 architecture is
21793 selected because its use is still experimental.
21797 @opindex mrh850-abi
21799 Enables support for the RH850 version of the V850 ABI. This is the
21800 default. With this version of the ABI the following rules apply:
21804 Integer sized structures and unions are returned via a memory pointer
21805 rather than a register.
21808 Large structures and unions (more than 8 bytes in size) are passed by
21812 Functions are aligned to 16-bit boundaries.
21815 The @option{-m8byte-align} command-line option is supported.
21818 The @option{-mdisable-callt} command-line option is enabled by
21819 default. The @option{-mno-disable-callt} command-line option is not
21823 When this version of the ABI is enabled the C preprocessor symbol
21824 @code{__V850_RH850_ABI__} is defined.
21828 Enables support for the old GCC version of the V850 ABI. With this
21829 version of the ABI the following rules apply:
21833 Integer sized structures and unions are returned in register @code{r10}.
21836 Large structures and unions (more than 8 bytes in size) are passed by
21840 Functions are aligned to 32-bit boundaries, unless optimizing for
21844 The @option{-m8byte-align} command-line option is not supported.
21847 The @option{-mdisable-callt} command-line option is supported but not
21848 enabled by default.
21851 When this version of the ABI is enabled the C preprocessor symbol
21852 @code{__V850_GCC_ABI__} is defined.
21854 @item -m8byte-align
21855 @itemx -mno-8byte-align
21856 @opindex m8byte-align
21857 @opindex mno-8byte-align
21858 Enables support for @code{double} and @code{long long} types to be
21859 aligned on 8-byte boundaries. The default is to restrict the
21860 alignment of all objects to at most 4-bytes. When
21861 @option{-m8byte-align} is in effect the C preprocessor symbol
21862 @code{__V850_8BYTE_ALIGN__} is defined.
21865 @opindex mbig-switch
21866 Generate code suitable for big switch tables. Use this option only if
21867 the assembler/linker complain about out of range branches within a switch
21872 This option causes r2 and r5 to be used in the code generated by
21873 the compiler. This setting is the default.
21875 @item -mno-app-regs
21876 @opindex mno-app-regs
21877 This option causes r2 and r5 to be treated as fixed registers.
21882 @subsection VAX Options
21883 @cindex VAX options
21885 These @samp{-m} options are defined for the VAX:
21890 Do not output certain jump instructions (@code{aobleq} and so on)
21891 that the Unix assembler for the VAX cannot handle across long
21896 Do output those jump instructions, on the assumption that the
21897 GNU assembler is being used.
21901 Output code for G-format floating-point numbers instead of D-format.
21904 @node Visium Options
21905 @subsection Visium Options
21906 @cindex Visium options
21912 A program which performs file I/O and is destined to run on an MCM target
21913 should be linked with this option. It causes the libraries libc.a and
21914 libdebug.a to be linked. The program should be run on the target under
21915 the control of the GDB remote debugging stub.
21919 A program which performs file I/O and is destined to run on the simulator
21920 should be linked with option. This causes libraries libc.a and libsim.a to
21924 @itemx -mhard-float
21926 @opindex mhard-float
21927 Generate code containing floating-point instructions. This is the
21931 @itemx -msoft-float
21933 @opindex msoft-float
21934 Generate code containing library calls for floating-point.
21936 @option{-msoft-float} changes the calling convention in the output file;
21937 therefore, it is only useful if you compile @emph{all} of a program with
21938 this option. In particular, you need to compile @file{libgcc.a}, the
21939 library that comes with GCC, with @option{-msoft-float} in order for
21942 @item -mcpu=@var{cpu_type}
21944 Set the instruction set, register set, and instruction scheduling parameters
21945 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21946 @samp{mcm}, @samp{gr5} and @samp{gr6}.
21948 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
21950 By default (unless configured otherwise), GCC generates code for the GR5
21951 variant of the Visium architecture.
21953 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
21954 architecture. The only difference from GR5 code is that the compiler will
21955 generate block move instructions.
21957 @item -mtune=@var{cpu_type}
21959 Set the instruction scheduling parameters for machine type @var{cpu_type},
21960 but do not set the instruction set or register set that the option
21961 @option{-mcpu=@var{cpu_type}} would.
21965 Generate code for the supervisor mode, where there are no restrictions on
21966 the access to general registers. This is the default.
21969 @opindex muser-mode
21970 Generate code for the user mode, where the access to some general registers
21971 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
21972 mode; on the GR6, only registers r29 to r31 are affected.
21976 @subsection VMS Options
21978 These @samp{-m} options are defined for the VMS implementations:
21981 @item -mvms-return-codes
21982 @opindex mvms-return-codes
21983 Return VMS condition codes from @code{main}. The default is to return POSIX-style
21984 condition (e.g.@ error) codes.
21986 @item -mdebug-main=@var{prefix}
21987 @opindex mdebug-main=@var{prefix}
21988 Flag the first routine whose name starts with @var{prefix} as the main
21989 routine for the debugger.
21993 Default to 64-bit memory allocation routines.
21995 @item -mpointer-size=@var{size}
21996 @opindex mpointer-size=@var{size}
21997 Set the default size of pointers. Possible options for @var{size} are
21998 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
21999 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22000 The later option disables @code{pragma pointer_size}.
22003 @node VxWorks Options
22004 @subsection VxWorks Options
22005 @cindex VxWorks Options
22007 The options in this section are defined for all VxWorks targets.
22008 Options specific to the target hardware are listed with the other
22009 options for that target.
22014 GCC can generate code for both VxWorks kernels and real time processes
22015 (RTPs). This option switches from the former to the latter. It also
22016 defines the preprocessor macro @code{__RTP__}.
22019 @opindex non-static
22020 Link an RTP executable against shared libraries rather than static
22021 libraries. The options @option{-static} and @option{-shared} can
22022 also be used for RTPs (@pxref{Link Options}); @option{-static}
22029 These options are passed down to the linker. They are defined for
22030 compatibility with Diab.
22033 @opindex Xbind-lazy
22034 Enable lazy binding of function calls. This option is equivalent to
22035 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22039 Disable lazy binding of function calls. This option is the default and
22040 is defined for compatibility with Diab.
22044 @subsection x86 Options
22045 @cindex x86 Options
22047 These @samp{-m} options are defined for the x86 family of computers.
22051 @item -march=@var{cpu-type}
22053 Generate instructions for the machine type @var{cpu-type}. In contrast to
22054 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22055 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22056 to generate code that may not run at all on processors other than the one
22057 indicated. Specifying @option{-march=@var{cpu-type}} implies
22058 @option{-mtune=@var{cpu-type}}.
22060 The choices for @var{cpu-type} are:
22064 This selects the CPU to generate code for at compilation time by determining
22065 the processor type of the compiling machine. Using @option{-march=native}
22066 enables all instruction subsets supported by the local machine (hence
22067 the result might not run on different machines). Using @option{-mtune=native}
22068 produces code optimized for the local machine under the constraints
22069 of the selected instruction set.
22072 Original Intel i386 CPU@.
22075 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22079 Intel Pentium CPU with no MMX support.
22082 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22085 Intel Pentium Pro CPU@.
22088 When used with @option{-march}, the Pentium Pro
22089 instruction set is used, so the code runs on all i686 family chips.
22090 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22093 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22098 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22102 Intel Pentium M; low-power version of Intel Pentium III CPU
22103 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22107 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22110 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22114 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22115 SSE2 and SSE3 instruction set support.
22118 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22119 instruction set support.
22122 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22123 SSE4.1, SSE4.2 and POPCNT instruction set support.
22126 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22127 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22130 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22131 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22134 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22135 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22136 instruction set support.
22139 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22140 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22141 BMI, BMI2 and F16C instruction set support.
22144 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22145 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22146 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22149 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22150 instruction set support.
22153 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22154 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22157 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22158 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22159 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
22160 AVX512CD instruction set support.
22163 AMD K6 CPU with MMX instruction set support.
22167 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
22170 @itemx athlon-tbird
22171 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22177 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22178 instruction set support.
22184 Processors based on the AMD K8 core with x86-64 instruction set support,
22185 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22186 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22187 instruction set extensions.)
22190 @itemx opteron-sse3
22191 @itemx athlon64-sse3
22192 Improved versions of AMD K8 cores with SSE3 instruction set support.
22196 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22197 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22198 instruction set extensions.)
22201 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22202 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22203 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22205 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22206 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22207 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22210 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22211 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22212 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22213 64-bit instruction set extensions.
22215 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22216 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22217 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22218 SSE4.2, ABM and 64-bit instruction set extensions.
22221 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22222 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22223 instruction set extensions.)
22226 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22227 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22228 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22231 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22235 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22236 instruction set support.
22239 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22240 implemented for this chip.)
22243 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22245 implemented for this chip.)
22248 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22251 @item -mtune=@var{cpu-type}
22253 Tune to @var{cpu-type} everything applicable about the generated code, except
22254 for the ABI and the set of available instructions.
22255 While picking a specific @var{cpu-type} schedules things appropriately
22256 for that particular chip, the compiler does not generate any code that
22257 cannot run on the default machine type unless you use a
22258 @option{-march=@var{cpu-type}} option.
22259 For example, if GCC is configured for i686-pc-linux-gnu
22260 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22261 but still runs on i686 machines.
22263 The choices for @var{cpu-type} are the same as for @option{-march}.
22264 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22268 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22269 If you know the CPU on which your code will run, then you should use
22270 the corresponding @option{-mtune} or @option{-march} option instead of
22271 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22272 of your application will have, then you should use this option.
22274 As new processors are deployed in the marketplace, the behavior of this
22275 option will change. Therefore, if you upgrade to a newer version of
22276 GCC, code generation controlled by this option will change to reflect
22278 that are most common at the time that version of GCC is released.
22280 There is no @option{-march=generic} option because @option{-march}
22281 indicates the instruction set the compiler can use, and there is no
22282 generic instruction set applicable to all processors. In contrast,
22283 @option{-mtune} indicates the processor (or, in this case, collection of
22284 processors) for which the code is optimized.
22287 Produce code optimized for the most current Intel processors, which are
22288 Haswell and Silvermont for this version of GCC. If you know the CPU
22289 on which your code will run, then you should use the corresponding
22290 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22291 But, if you want your application performs better on both Haswell and
22292 Silvermont, then you should use this option.
22294 As new Intel processors are deployed in the marketplace, the behavior of
22295 this option will change. Therefore, if you upgrade to a newer version of
22296 GCC, code generation controlled by this option will change to reflect
22297 the most current Intel processors at the time that version of GCC is
22300 There is no @option{-march=intel} option because @option{-march} indicates
22301 the instruction set the compiler can use, and there is no common
22302 instruction set applicable to all processors. In contrast,
22303 @option{-mtune} indicates the processor (or, in this case, collection of
22304 processors) for which the code is optimized.
22307 @item -mcpu=@var{cpu-type}
22309 A deprecated synonym for @option{-mtune}.
22311 @item -mfpmath=@var{unit}
22313 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22314 for @var{unit} are:
22318 Use the standard 387 floating-point coprocessor present on the majority of chips and
22319 emulated otherwise. Code compiled with this option runs almost everywhere.
22320 The temporary results are computed in 80-bit precision instead of the precision
22321 specified by the type, resulting in slightly different results compared to most
22322 of other chips. See @option{-ffloat-store} for more detailed description.
22324 This is the default choice for x86-32 targets.
22327 Use scalar floating-point instructions present in the SSE instruction set.
22328 This instruction set is supported by Pentium III and newer chips,
22329 and in the AMD line
22330 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22331 instruction set supports only single-precision arithmetic, thus the double and
22332 extended-precision arithmetic are still done using 387. A later version, present
22333 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22336 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22337 or @option{-msse2} switches to enable SSE extensions and make this option
22338 effective. For the x86-64 compiler, these extensions are enabled by default.
22340 The resulting code should be considerably faster in the majority of cases and avoid
22341 the numerical instability problems of 387 code, but may break some existing
22342 code that expects temporaries to be 80 bits.
22344 This is the default choice for the x86-64 compiler.
22349 Attempt to utilize both instruction sets at once. This effectively doubles the
22350 amount of available registers, and on chips with separate execution units for
22351 387 and SSE the execution resources too. Use this option with care, as it is
22352 still experimental, because the GCC register allocator does not model separate
22353 functional units well, resulting in unstable performance.
22356 @item -masm=@var{dialect}
22357 @opindex masm=@var{dialect}
22358 Output assembly instructions using selected @var{dialect}. Also affects
22359 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22360 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22361 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22362 not support @samp{intel}.
22365 @itemx -mno-ieee-fp
22367 @opindex mno-ieee-fp
22368 Control whether or not the compiler uses IEEE floating-point
22369 comparisons. These correctly handle the case where the result of a
22370 comparison is unordered.
22373 @opindex msoft-float
22374 Generate output containing library calls for floating point.
22376 @strong{Warning:} the requisite libraries are not part of GCC@.
22377 Normally the facilities of the machine's usual C compiler are used, but
22378 this can't be done directly in cross-compilation. You must make your
22379 own arrangements to provide suitable library functions for
22382 On machines where a function returns floating-point results in the 80387
22383 register stack, some floating-point opcodes may be emitted even if
22384 @option{-msoft-float} is used.
22386 @item -mno-fp-ret-in-387
22387 @opindex mno-fp-ret-in-387
22388 Do not use the FPU registers for return values of functions.
22390 The usual calling convention has functions return values of types
22391 @code{float} and @code{double} in an FPU register, even if there
22392 is no FPU@. The idea is that the operating system should emulate
22395 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22396 in ordinary CPU registers instead.
22398 @item -mno-fancy-math-387
22399 @opindex mno-fancy-math-387
22400 Some 387 emulators do not support the @code{sin}, @code{cos} and
22401 @code{sqrt} instructions for the 387. Specify this option to avoid
22402 generating those instructions. This option is the default on FreeBSD,
22403 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22404 indicates that the target CPU always has an FPU and so the
22405 instruction does not need emulation. These
22406 instructions are not generated unless you also use the
22407 @option{-funsafe-math-optimizations} switch.
22409 @item -malign-double
22410 @itemx -mno-align-double
22411 @opindex malign-double
22412 @opindex mno-align-double
22413 Control whether GCC aligns @code{double}, @code{long double}, and
22414 @code{long long} variables on a two-word boundary or a one-word
22415 boundary. Aligning @code{double} variables on a two-word boundary
22416 produces code that runs somewhat faster on a Pentium at the
22417 expense of more memory.
22419 On x86-64, @option{-malign-double} is enabled by default.
22421 @strong{Warning:} if you use the @option{-malign-double} switch,
22422 structures containing the above types are aligned differently than
22423 the published application binary interface specifications for the x86-32
22424 and are not binary compatible with structures in code compiled
22425 without that switch.
22427 @item -m96bit-long-double
22428 @itemx -m128bit-long-double
22429 @opindex m96bit-long-double
22430 @opindex m128bit-long-double
22431 These switches control the size of @code{long double} type. The x86-32
22432 application binary interface specifies the size to be 96 bits,
22433 so @option{-m96bit-long-double} is the default in 32-bit mode.
22435 Modern architectures (Pentium and newer) prefer @code{long double}
22436 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22437 conforming to the ABI, this is not possible. So specifying
22438 @option{-m128bit-long-double} aligns @code{long double}
22439 to a 16-byte boundary by padding the @code{long double} with an additional
22442 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22443 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22445 Notice that neither of these options enable any extra precision over the x87
22446 standard of 80 bits for a @code{long double}.
22448 @strong{Warning:} if you override the default value for your target ABI, this
22449 changes the size of
22450 structures and arrays containing @code{long double} variables,
22451 as well as modifying the function calling convention for functions taking
22452 @code{long double}. Hence they are not binary-compatible
22453 with code compiled without that switch.
22455 @item -mlong-double-64
22456 @itemx -mlong-double-80
22457 @itemx -mlong-double-128
22458 @opindex mlong-double-64
22459 @opindex mlong-double-80
22460 @opindex mlong-double-128
22461 These switches control the size of @code{long double} type. A size
22462 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22463 type. This is the default for 32-bit Bionic C library. A size
22464 of 128 bits makes the @code{long double} type equivalent to the
22465 @code{__float128} type. This is the default for 64-bit Bionic C library.
22467 @strong{Warning:} if you override the default value for your target ABI, this
22468 changes the size of
22469 structures and arrays containing @code{long double} variables,
22470 as well as modifying the function calling convention for functions taking
22471 @code{long double}. Hence they are not binary-compatible
22472 with code compiled without that switch.
22474 @item -malign-data=@var{type}
22475 @opindex malign-data
22476 Control how GCC aligns variables. Supported values for @var{type} are
22477 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22478 and earlier, @samp{abi} uses alignment value as specified by the
22479 psABI, and @samp{cacheline} uses increased alignment value to match
22480 the cache line size. @samp{compat} is the default.
22482 @item -mlarge-data-threshold=@var{threshold}
22483 @opindex mlarge-data-threshold
22484 When @option{-mcmodel=medium} is specified, data objects larger than
22485 @var{threshold} are placed in the large data section. This value must be the
22486 same across all objects linked into the binary, and defaults to 65535.
22490 Use a different function-calling convention, in which functions that
22491 take a fixed number of arguments return with the @code{ret @var{num}}
22492 instruction, which pops their arguments while returning. This saves one
22493 instruction in the caller since there is no need to pop the arguments
22496 You can specify that an individual function is called with this calling
22497 sequence with the function attribute @code{stdcall}. You can also
22498 override the @option{-mrtd} option by using the function attribute
22499 @code{cdecl}. @xref{Function Attributes}.
22501 @strong{Warning:} this calling convention is incompatible with the one
22502 normally used on Unix, so you cannot use it if you need to call
22503 libraries compiled with the Unix compiler.
22505 Also, you must provide function prototypes for all functions that
22506 take variable numbers of arguments (including @code{printf});
22507 otherwise incorrect code is generated for calls to those
22510 In addition, seriously incorrect code results if you call a
22511 function with too many arguments. (Normally, extra arguments are
22512 harmlessly ignored.)
22514 @item -mregparm=@var{num}
22516 Control how many registers are used to pass integer arguments. By
22517 default, no registers are used to pass arguments, and at most 3
22518 registers can be used. You can control this behavior for a specific
22519 function by using the function attribute @code{regparm}.
22520 @xref{Function Attributes}.
22522 @strong{Warning:} if you use this switch, and
22523 @var{num} is nonzero, then you must build all modules with the same
22524 value, including any libraries. This includes the system libraries and
22528 @opindex msseregparm
22529 Use SSE register passing conventions for float and double arguments
22530 and return values. You can control this behavior for a specific
22531 function by using the function attribute @code{sseregparm}.
22532 @xref{Function Attributes}.
22534 @strong{Warning:} if you use this switch then you must build all
22535 modules with the same value, including any libraries. This includes
22536 the system libraries and startup modules.
22538 @item -mvect8-ret-in-mem
22539 @opindex mvect8-ret-in-mem
22540 Return 8-byte vectors in memory instead of MMX registers. This is the
22541 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22542 Studio compilers until version 12. Later compiler versions (starting
22543 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22544 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22545 you need to remain compatible with existing code produced by those
22546 previous compiler versions or older versions of GCC@.
22555 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22556 is specified, the significands of results of floating-point operations are
22557 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22558 significands of results of floating-point operations to 53 bits (double
22559 precision) and @option{-mpc80} rounds the significands of results of
22560 floating-point operations to 64 bits (extended double precision), which is
22561 the default. When this option is used, floating-point operations in higher
22562 precisions are not available to the programmer without setting the FPU
22563 control word explicitly.
22565 Setting the rounding of floating-point operations to less than the default
22566 80 bits can speed some programs by 2% or more. Note that some mathematical
22567 libraries assume that extended-precision (80-bit) floating-point operations
22568 are enabled by default; routines in such libraries could suffer significant
22569 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22570 when this option is used to set the precision to less than extended precision.
22572 @item -mstackrealign
22573 @opindex mstackrealign
22574 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22575 option generates an alternate prologue and epilogue that realigns the
22576 run-time stack if necessary. This supports mixing legacy codes that keep
22577 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22578 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22579 applicable to individual functions.
22581 @item -mpreferred-stack-boundary=@var{num}
22582 @opindex mpreferred-stack-boundary
22583 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22584 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22585 the default is 4 (16 bytes or 128 bits).
22587 @strong{Warning:} When generating code for the x86-64 architecture with
22588 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22589 used to keep the stack boundary aligned to 8 byte boundary. Since
22590 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22591 intended to be used in controlled environment where stack space is
22592 important limitation. This option leads to wrong code when functions
22593 compiled with 16 byte stack alignment (such as functions from a standard
22594 library) are called with misaligned stack. In this case, SSE
22595 instructions may lead to misaligned memory access traps. In addition,
22596 variable arguments are handled incorrectly for 16 byte aligned
22597 objects (including x87 long double and __int128), leading to wrong
22598 results. You must build all modules with
22599 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22600 includes the system libraries and startup modules.
22602 @item -mincoming-stack-boundary=@var{num}
22603 @opindex mincoming-stack-boundary
22604 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22605 boundary. If @option{-mincoming-stack-boundary} is not specified,
22606 the one specified by @option{-mpreferred-stack-boundary} is used.
22608 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22609 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22610 suffer significant run time performance penalties. On Pentium III, the
22611 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22612 properly if it is not 16-byte aligned.
22614 To ensure proper alignment of this values on the stack, the stack boundary
22615 must be as aligned as that required by any value stored on the stack.
22616 Further, every function must be generated such that it keeps the stack
22617 aligned. Thus calling a function compiled with a higher preferred
22618 stack boundary from a function compiled with a lower preferred stack
22619 boundary most likely misaligns the stack. It is recommended that
22620 libraries that use callbacks always use the default setting.
22622 This extra alignment does consume extra stack space, and generally
22623 increases code size. Code that is sensitive to stack space usage, such
22624 as embedded systems and operating system kernels, may want to reduce the
22625 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22671 @opindex mclfushopt
22689 @itemx -mprefetchwt1
22690 @opindex mprefetchwt1
22738 These switches enable the use of instructions in the MMX, SSE,
22739 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22740 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22741 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX or 3DNow!@:
22742 extended instruction sets. Each has a corresponding @option{-mno-} option
22743 to disable use of these instructions.
22745 These extensions are also available as built-in functions: see
22746 @ref{x86 Built-in Functions}, for details of the functions enabled and
22747 disabled by these switches.
22749 To generate SSE/SSE2 instructions automatically from floating-point
22750 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22752 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22753 generates new AVX instructions or AVX equivalence for all SSEx instructions
22756 These options enable GCC to use these extended instructions in
22757 generated code, even without @option{-mfpmath=sse}. Applications that
22758 perform run-time CPU detection must compile separate files for each
22759 supported architecture, using the appropriate flags. In particular,
22760 the file containing the CPU detection code should be compiled without
22763 @item -mdump-tune-features
22764 @opindex mdump-tune-features
22765 This option instructs GCC to dump the names of the x86 performance
22766 tuning features and default settings. The names can be used in
22767 @option{-mtune-ctrl=@var{feature-list}}.
22769 @item -mtune-ctrl=@var{feature-list}
22770 @opindex mtune-ctrl=@var{feature-list}
22771 This option is used to do fine grain control of x86 code generation features.
22772 @var{feature-list} is a comma separated list of @var{feature} names. See also
22773 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22774 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22775 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22776 developers. Using it may lead to code paths not covered by testing and can
22777 potentially result in compiler ICEs or runtime errors.
22780 @opindex mno-default
22781 This option instructs GCC to turn off all tunable features. See also
22782 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
22786 This option instructs GCC to emit a @code{cld} instruction in the prologue
22787 of functions that use string instructions. String instructions depend on
22788 the DF flag to select between autoincrement or autodecrement mode. While the
22789 ABI specifies the DF flag to be cleared on function entry, some operating
22790 systems violate this specification by not clearing the DF flag in their
22791 exception dispatchers. The exception handler can be invoked with the DF flag
22792 set, which leads to wrong direction mode when string instructions are used.
22793 This option can be enabled by default on 32-bit x86 targets by configuring
22794 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
22795 instructions can be suppressed with the @option{-mno-cld} compiler option
22799 @opindex mvzeroupper
22800 This option instructs GCC to emit a @code{vzeroupper} instruction
22801 before a transfer of control flow out of the function to minimize
22802 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
22805 @item -mprefer-avx128
22806 @opindex mprefer-avx128
22807 This option instructs GCC to use 128-bit AVX instructions instead of
22808 256-bit AVX instructions in the auto-vectorizer.
22812 This option enables GCC to generate @code{CMPXCHG16B} instructions.
22813 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
22814 (or oword) data types.
22815 This is useful for high-resolution counters that can be updated
22816 by multiple processors (or cores). This instruction is generated as part of
22817 atomic built-in functions: see @ref{__sync Builtins} or
22818 @ref{__atomic Builtins} for details.
22822 This option enables generation of @code{SAHF} instructions in 64-bit code.
22823 Early Intel Pentium 4 CPUs with Intel 64 support,
22824 prior to the introduction of Pentium 4 G1 step in December 2005,
22825 lacked the @code{LAHF} and @code{SAHF} instructions
22826 which are supported by AMD64.
22827 These are load and store instructions, respectively, for certain status flags.
22828 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
22829 @code{drem}, and @code{remainder} built-in functions;
22830 see @ref{Other Builtins} for details.
22834 This option enables use of the @code{movbe} instruction to implement
22835 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
22839 This option enables built-in functions @code{__builtin_ia32_crc32qi},
22840 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
22841 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
22845 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
22846 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
22847 with an additional Newton-Raphson step
22848 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
22849 (and their vectorized
22850 variants) for single-precision floating-point arguments. These instructions
22851 are generated only when @option{-funsafe-math-optimizations} is enabled
22852 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
22853 Note that while the throughput of the sequence is higher than the throughput
22854 of the non-reciprocal instruction, the precision of the sequence can be
22855 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
22857 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
22858 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
22859 combination), and doesn't need @option{-mrecip}.
22861 Also note that GCC emits the above sequence with additional Newton-Raphson step
22862 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
22863 already with @option{-ffast-math} (or the above option combination), and
22864 doesn't need @option{-mrecip}.
22866 @item -mrecip=@var{opt}
22867 @opindex mrecip=opt
22868 This option controls which reciprocal estimate instructions
22869 may be used. @var{opt} is a comma-separated list of options, which may
22870 be preceded by a @samp{!} to invert the option:
22874 Enable all estimate instructions.
22877 Enable the default instructions, equivalent to @option{-mrecip}.
22880 Disable all estimate instructions, equivalent to @option{-mno-recip}.
22883 Enable the approximation for scalar division.
22886 Enable the approximation for vectorized division.
22889 Enable the approximation for scalar square root.
22892 Enable the approximation for vectorized square root.
22895 So, for example, @option{-mrecip=all,!sqrt} enables
22896 all of the reciprocal approximations, except for square root.
22898 @item -mveclibabi=@var{type}
22899 @opindex mveclibabi
22900 Specifies the ABI type to use for vectorizing intrinsics using an
22901 external library. Supported values for @var{type} are @samp{svml}
22902 for the Intel short
22903 vector math library and @samp{acml} for the AMD math core library.
22904 To use this option, both @option{-ftree-vectorize} and
22905 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
22906 ABI-compatible library must be specified at link time.
22908 GCC currently emits calls to @code{vmldExp2},
22909 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
22910 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
22911 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
22912 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
22913 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
22914 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
22915 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
22916 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
22917 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
22918 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
22919 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
22920 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
22921 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
22922 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
22923 when @option{-mveclibabi=acml} is used.
22925 @item -mabi=@var{name}
22927 Generate code for the specified calling convention. Permissible values
22928 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
22929 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
22930 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
22931 You can control this behavior for specific functions by
22932 using the function attributes @code{ms_abi} and @code{sysv_abi}.
22933 @xref{Function Attributes}.
22935 @item -mtls-dialect=@var{type}
22936 @opindex mtls-dialect
22937 Generate code to access thread-local storage using the @samp{gnu} or
22938 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
22939 @samp{gnu2} is more efficient, but it may add compile- and run-time
22940 requirements that cannot be satisfied on all systems.
22943 @itemx -mno-push-args
22944 @opindex mpush-args
22945 @opindex mno-push-args
22946 Use PUSH operations to store outgoing parameters. This method is shorter
22947 and usually equally fast as method using SUB/MOV operations and is enabled
22948 by default. In some cases disabling it may improve performance because of
22949 improved scheduling and reduced dependencies.
22951 @item -maccumulate-outgoing-args
22952 @opindex maccumulate-outgoing-args
22953 If enabled, the maximum amount of space required for outgoing arguments is
22954 computed in the function prologue. This is faster on most modern CPUs
22955 because of reduced dependencies, improved scheduling and reduced stack usage
22956 when the preferred stack boundary is not equal to 2. The drawback is a notable
22957 increase in code size. This switch implies @option{-mno-push-args}.
22961 Support thread-safe exception handling on MinGW. Programs that rely
22962 on thread-safe exception handling must compile and link all code with the
22963 @option{-mthreads} option. When compiling, @option{-mthreads} defines
22964 @option{-D_MT}; when linking, it links in a special thread helper library
22965 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
22967 @item -mno-align-stringops
22968 @opindex mno-align-stringops
22969 Do not align the destination of inlined string operations. This switch reduces
22970 code size and improves performance in case the destination is already aligned,
22971 but GCC doesn't know about it.
22973 @item -minline-all-stringops
22974 @opindex minline-all-stringops
22975 By default GCC inlines string operations only when the destination is
22976 known to be aligned to least a 4-byte boundary.
22977 This enables more inlining and increases code
22978 size, but may improve performance of code that depends on fast
22979 @code{memcpy}, @code{strlen},
22980 and @code{memset} for short lengths.
22982 @item -minline-stringops-dynamically
22983 @opindex minline-stringops-dynamically
22984 For string operations of unknown size, use run-time checks with
22985 inline code for small blocks and a library call for large blocks.
22987 @item -mstringop-strategy=@var{alg}
22988 @opindex mstringop-strategy=@var{alg}
22989 Override the internal decision heuristic for the particular algorithm to use
22990 for inlining string operations. The allowed values for @var{alg} are:
22996 Expand using i386 @code{rep} prefix of the specified size.
23000 @itemx unrolled_loop
23001 Expand into an inline loop.
23004 Always use a library call.
23007 @item -mmemcpy-strategy=@var{strategy}
23008 @opindex mmemcpy-strategy=@var{strategy}
23009 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23010 should be inlined and what inline algorithm to use when the expected size
23011 of the copy operation is known. @var{strategy}
23012 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23013 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23014 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23015 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23016 in the list must be specified in increasing order. The minimal byte size for
23017 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23020 @item -mmemset-strategy=@var{strategy}
23021 @opindex mmemset-strategy=@var{strategy}
23022 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23023 @code{__builtin_memset} expansion.
23025 @item -momit-leaf-frame-pointer
23026 @opindex momit-leaf-frame-pointer
23027 Don't keep the frame pointer in a register for leaf functions. This
23028 avoids the instructions to save, set up, and restore frame pointers and
23029 makes an extra register available in leaf functions. The option
23030 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23031 which might make debugging harder.
23033 @item -mtls-direct-seg-refs
23034 @itemx -mno-tls-direct-seg-refs
23035 @opindex mtls-direct-seg-refs
23036 Controls whether TLS variables may be accessed with offsets from the
23037 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23038 or whether the thread base pointer must be added. Whether or not this
23039 is valid depends on the operating system, and whether it maps the
23040 segment to cover the entire TLS area.
23042 For systems that use the GNU C Library, the default is on.
23045 @itemx -mno-sse2avx
23047 Specify that the assembler should encode SSE instructions with VEX
23048 prefix. The option @option{-mavx} turns this on by default.
23053 If profiling is active (@option{-pg}), put the profiling
23054 counter call before the prologue.
23055 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23056 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23058 @item -mrecord-mcount
23059 @itemx -mno-record-mcount
23060 @opindex mrecord-mcount
23061 If profiling is active (@option{-pg}), generate a __mcount_loc section
23062 that contains pointers to each profiling call. This is useful for
23063 automatically patching and out calls.
23066 @itemx -mno-nop-mcount
23067 @opindex mnop-mcount
23068 If profiling is active (@option{-pg}), generate the calls to
23069 the profiling functions as nops. This is useful when they
23070 should be patched in later dynamically. This is likely only
23071 useful together with @option{-mrecord-mcount}.
23073 @item -mskip-rax-setup
23074 @itemx -mno-skip-rax-setup
23075 @opindex mskip-rax-setup
23076 When generating code for the x86-64 architecture with SSE extensions
23077 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23078 register when there are no variable arguments passed in vector registers.
23080 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23081 saving vector registers on stack when passing variable arguments, the
23082 impacts of this option are callees may waste some stack space,
23083 misbehave or jump to a random location. GCC 4.4 or newer don't have
23084 those issues, regardless the RAX register value.
23087 @itemx -mno-8bit-idiv
23088 @opindex m8bit-idiv
23089 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23090 much faster than 32-bit/64-bit integer divide. This option generates a
23091 run-time check. If both dividend and divisor are within range of 0
23092 to 255, 8-bit unsigned integer divide is used instead of
23093 32-bit/64-bit integer divide.
23095 @item -mavx256-split-unaligned-load
23096 @itemx -mavx256-split-unaligned-store
23097 @opindex mavx256-split-unaligned-load
23098 @opindex mavx256-split-unaligned-store
23099 Split 32-byte AVX unaligned load and store.
23101 @item -mstack-protector-guard=@var{guard}
23102 @opindex mstack-protector-guard=@var{guard}
23103 Generate stack protection code using canary at @var{guard}. Supported
23104 locations are @samp{global} for global canary or @samp{tls} for per-thread
23105 canary in the TLS block (the default). This option has effect only when
23106 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23110 These @samp{-m} switches are supported in addition to the above
23111 on x86-64 processors in 64-bit environments.
23122 Generate code for a 16-bit, 32-bit or 64-bit environment.
23123 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23125 generates code that runs on any i386 system.
23127 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23128 types to 64 bits, and generates code for the x86-64 architecture.
23129 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23130 and @option{-mdynamic-no-pic} options.
23132 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23134 generates code for the x86-64 architecture.
23136 The @option{-m16} option is the same as @option{-m32}, except for that
23137 it outputs the @code{.code16gcc} assembly directive at the beginning of
23138 the assembly output so that the binary can run in 16-bit mode.
23140 @item -mno-red-zone
23141 @opindex mno-red-zone
23142 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23143 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23144 stack pointer that is not modified by signal or interrupt handlers
23145 and therefore can be used for temporary data without adjusting the stack
23146 pointer. The flag @option{-mno-red-zone} disables this red zone.
23148 @item -mcmodel=small
23149 @opindex mcmodel=small
23150 Generate code for the small code model: the program and its symbols must
23151 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23152 Programs can be statically or dynamically linked. This is the default
23155 @item -mcmodel=kernel
23156 @opindex mcmodel=kernel
23157 Generate code for the kernel code model. The kernel runs in the
23158 negative 2 GB of the address space.
23159 This model has to be used for Linux kernel code.
23161 @item -mcmodel=medium
23162 @opindex mcmodel=medium
23163 Generate code for the medium model: the program is linked in the lower 2
23164 GB of the address space. Small symbols are also placed there. Symbols
23165 with sizes larger than @option{-mlarge-data-threshold} are put into
23166 large data or BSS sections and can be located above 2GB. Programs can
23167 be statically or dynamically linked.
23169 @item -mcmodel=large
23170 @opindex mcmodel=large
23171 Generate code for the large model. This model makes no assumptions
23172 about addresses and sizes of sections.
23174 @item -maddress-mode=long
23175 @opindex maddress-mode=long
23176 Generate code for long address mode. This is only supported for 64-bit
23177 and x32 environments. It is the default address mode for 64-bit
23180 @item -maddress-mode=short
23181 @opindex maddress-mode=short
23182 Generate code for short address mode. This is only supported for 32-bit
23183 and x32 environments. It is the default address mode for 32-bit and
23187 @node x86 Windows Options
23188 @subsection x86 Windows Options
23189 @cindex x86 Windows Options
23190 @cindex Windows Options for x86
23192 These additional options are available for Microsoft Windows targets:
23198 specifies that a console application is to be generated, by
23199 instructing the linker to set the PE header subsystem type
23200 required for console applications.
23201 This option is available for Cygwin and MinGW targets and is
23202 enabled by default on those targets.
23206 This option is available for Cygwin and MinGW targets. It
23207 specifies that a DLL---a dynamic link library---is to be
23208 generated, enabling the selection of the required runtime
23209 startup object and entry point.
23211 @item -mnop-fun-dllimport
23212 @opindex mnop-fun-dllimport
23213 This option is available for Cygwin and MinGW targets. It
23214 specifies that the @code{dllimport} attribute should be ignored.
23218 This option is available for MinGW targets. It specifies
23219 that MinGW-specific thread support is to be used.
23223 This option is available for MinGW-w64 targets. It causes
23224 the @code{UNICODE} preprocessor macro to be predefined, and
23225 chooses Unicode-capable runtime startup code.
23229 This option is available for Cygwin and MinGW targets. It
23230 specifies that the typical Microsoft Windows predefined macros are to
23231 be set in the pre-processor, but does not influence the choice
23232 of runtime library/startup code.
23236 This option is available for Cygwin and MinGW targets. It
23237 specifies that a GUI application is to be generated by
23238 instructing the linker to set the PE header subsystem type
23241 @item -fno-set-stack-executable
23242 @opindex fno-set-stack-executable
23243 This option is available for MinGW targets. It specifies that
23244 the executable flag for the stack used by nested functions isn't
23245 set. This is necessary for binaries running in kernel mode of
23246 Microsoft Windows, as there the User32 API, which is used to set executable
23247 privileges, isn't available.
23249 @item -fwritable-relocated-rdata
23250 @opindex fno-writable-relocated-rdata
23251 This option is available for MinGW and Cygwin targets. It specifies
23252 that relocated-data in read-only section is put into .data
23253 section. This is a necessary for older runtimes not supporting
23254 modification of .rdata sections for pseudo-relocation.
23256 @item -mpe-aligned-commons
23257 @opindex mpe-aligned-commons
23258 This option is available for Cygwin and MinGW targets. It
23259 specifies that the GNU extension to the PE file format that
23260 permits the correct alignment of COMMON variables should be
23261 used when generating code. It is enabled by default if
23262 GCC detects that the target assembler found during configuration
23263 supports the feature.
23266 See also under @ref{x86 Options} for standard options.
23268 @node Xstormy16 Options
23269 @subsection Xstormy16 Options
23270 @cindex Xstormy16 Options
23272 These options are defined for Xstormy16:
23277 Choose startup files and linker script suitable for the simulator.
23280 @node Xtensa Options
23281 @subsection Xtensa Options
23282 @cindex Xtensa Options
23284 These options are supported for Xtensa targets:
23288 @itemx -mno-const16
23290 @opindex mno-const16
23291 Enable or disable use of @code{CONST16} instructions for loading
23292 constant values. The @code{CONST16} instruction is currently not a
23293 standard option from Tensilica. When enabled, @code{CONST16}
23294 instructions are always used in place of the standard @code{L32R}
23295 instructions. The use of @code{CONST16} is enabled by default only if
23296 the @code{L32R} instruction is not available.
23299 @itemx -mno-fused-madd
23300 @opindex mfused-madd
23301 @opindex mno-fused-madd
23302 Enable or disable use of fused multiply/add and multiply/subtract
23303 instructions in the floating-point option. This has no effect if the
23304 floating-point option is not also enabled. Disabling fused multiply/add
23305 and multiply/subtract instructions forces the compiler to use separate
23306 instructions for the multiply and add/subtract operations. This may be
23307 desirable in some cases where strict IEEE 754-compliant results are
23308 required: the fused multiply add/subtract instructions do not round the
23309 intermediate result, thereby producing results with @emph{more} bits of
23310 precision than specified by the IEEE standard. Disabling fused multiply
23311 add/subtract instructions also ensures that the program output is not
23312 sensitive to the compiler's ability to combine multiply and add/subtract
23315 @item -mserialize-volatile
23316 @itemx -mno-serialize-volatile
23317 @opindex mserialize-volatile
23318 @opindex mno-serialize-volatile
23319 When this option is enabled, GCC inserts @code{MEMW} instructions before
23320 @code{volatile} memory references to guarantee sequential consistency.
23321 The default is @option{-mserialize-volatile}. Use
23322 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23324 @item -mforce-no-pic
23325 @opindex mforce-no-pic
23326 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23327 position-independent code (PIC), this option disables PIC for compiling
23330 @item -mtext-section-literals
23331 @itemx -mno-text-section-literals
23332 @opindex mtext-section-literals
23333 @opindex mno-text-section-literals
23334 These options control the treatment of literal pools. The default is
23335 @option{-mno-text-section-literals}, which places literals in a separate
23336 section in the output file. This allows the literal pool to be placed
23337 in a data RAM/ROM, and it also allows the linker to combine literal
23338 pools from separate object files to remove redundant literals and
23339 improve code size. With @option{-mtext-section-literals}, the literals
23340 are interspersed in the text section in order to keep them as close as
23341 possible to their references. This may be necessary for large assembly
23344 @item -mtarget-align
23345 @itemx -mno-target-align
23346 @opindex mtarget-align
23347 @opindex mno-target-align
23348 When this option is enabled, GCC instructs the assembler to
23349 automatically align instructions to reduce branch penalties at the
23350 expense of some code density. The assembler attempts to widen density
23351 instructions to align branch targets and the instructions following call
23352 instructions. If there are not enough preceding safe density
23353 instructions to align a target, no widening is performed. The
23354 default is @option{-mtarget-align}. These options do not affect the
23355 treatment of auto-aligned instructions like @code{LOOP}, which the
23356 assembler always aligns, either by widening density instructions or
23357 by inserting NOP instructions.
23360 @itemx -mno-longcalls
23361 @opindex mlongcalls
23362 @opindex mno-longcalls
23363 When this option is enabled, GCC instructs the assembler to translate
23364 direct calls to indirect calls unless it can determine that the target
23365 of a direct call is in the range allowed by the call instruction. This
23366 translation typically occurs for calls to functions in other source
23367 files. Specifically, the assembler translates a direct @code{CALL}
23368 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23369 The default is @option{-mno-longcalls}. This option should be used in
23370 programs where the call target can potentially be out of range. This
23371 option is implemented in the assembler, not the compiler, so the
23372 assembly code generated by GCC still shows direct call
23373 instructions---look at the disassembled object code to see the actual
23374 instructions. Note that the assembler uses an indirect call for
23375 every cross-file call, not just those that really are out of range.
23378 @node zSeries Options
23379 @subsection zSeries Options
23380 @cindex zSeries options
23382 These are listed under @xref{S/390 and zSeries Options}.
23384 @node Code Gen Options
23385 @section Options for Code Generation Conventions
23386 @cindex code generation conventions
23387 @cindex options, code generation
23388 @cindex run-time options
23390 These machine-independent options control the interface conventions
23391 used in code generation.
23393 Most of them have both positive and negative forms; the negative form
23394 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23395 one of the forms is listed---the one that is not the default. You
23396 can figure out the other form by either removing @samp{no-} or adding
23400 @item -fbounds-check
23401 @opindex fbounds-check
23402 For front ends that support it, generate additional code to check that
23403 indices used to access arrays are within the declared range. This is
23404 currently only supported by the Java and Fortran front ends, where
23405 this option defaults to true and false respectively.
23407 @item -fstack-reuse=@var{reuse-level}
23408 @opindex fstack_reuse
23409 This option controls stack space reuse for user declared local/auto variables
23410 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23411 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23412 local variables and temporaries, @samp{named_vars} enables the reuse only for
23413 user defined local variables with names, and @samp{none} disables stack reuse
23414 completely. The default value is @samp{all}. The option is needed when the
23415 program extends the lifetime of a scoped local variable or a compiler generated
23416 temporary beyond the end point defined by the language. When a lifetime of
23417 a variable ends, and if the variable lives in memory, the optimizing compiler
23418 has the freedom to reuse its stack space with other temporaries or scoped
23419 local variables whose live range does not overlap with it. Legacy code extending
23420 local lifetime is likely to break with the stack reuse optimization.
23439 if (*p == 10) // out of scope use of local1
23450 A(int k) : i(k), j(k) @{ @}
23457 void foo(const A& ar)
23464 foo(A(10)); // temp object's lifetime ends when foo returns
23470 ap->i+= 10; // ap references out of scope temp whose space
23471 // is reused with a. What is the value of ap->i?
23476 The lifetime of a compiler generated temporary is well defined by the C++
23477 standard. When a lifetime of a temporary ends, and if the temporary lives
23478 in memory, the optimizing compiler has the freedom to reuse its stack
23479 space with other temporaries or scoped local variables whose live range
23480 does not overlap with it. However some of the legacy code relies on
23481 the behavior of older compilers in which temporaries' stack space is
23482 not reused, the aggressive stack reuse can lead to runtime errors. This
23483 option is used to control the temporary stack reuse optimization.
23487 This option generates traps for signed overflow on addition, subtraction,
23488 multiplication operations.
23492 This option instructs the compiler to assume that signed arithmetic
23493 overflow of addition, subtraction and multiplication wraps around
23494 using twos-complement representation. This flag enables some optimizations
23495 and disables others. This option is enabled by default for the Java
23496 front end, as required by the Java language specification.
23499 @opindex fexceptions
23500 Enable exception handling. Generates extra code needed to propagate
23501 exceptions. For some targets, this implies GCC generates frame
23502 unwind information for all functions, which can produce significant data
23503 size overhead, although it does not affect execution. If you do not
23504 specify this option, GCC enables it by default for languages like
23505 C++ that normally require exception handling, and disables it for
23506 languages like C that do not normally require it. However, you may need
23507 to enable this option when compiling C code that needs to interoperate
23508 properly with exception handlers written in C++. You may also wish to
23509 disable this option if you are compiling older C++ programs that don't
23510 use exception handling.
23512 @item -fnon-call-exceptions
23513 @opindex fnon-call-exceptions
23514 Generate code that allows trapping instructions to throw exceptions.
23515 Note that this requires platform-specific runtime support that does
23516 not exist everywhere. Moreover, it only allows @emph{trapping}
23517 instructions to throw exceptions, i.e.@: memory references or floating-point
23518 instructions. It does not allow exceptions to be thrown from
23519 arbitrary signal handlers such as @code{SIGALRM}.
23521 @item -fdelete-dead-exceptions
23522 @opindex fdelete-dead-exceptions
23523 Consider that instructions that may throw exceptions but don't otherwise
23524 contribute to the execution of the program can be optimized away.
23525 This option is enabled by default for the Ada front end, as permitted by
23526 the Ada language specification.
23527 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23529 @item -funwind-tables
23530 @opindex funwind-tables
23531 Similar to @option{-fexceptions}, except that it just generates any needed
23532 static data, but does not affect the generated code in any other way.
23533 You normally do not need to enable this option; instead, a language processor
23534 that needs this handling enables it on your behalf.
23536 @item -fasynchronous-unwind-tables
23537 @opindex fasynchronous-unwind-tables
23538 Generate unwind table in DWARF 2 format, if supported by target machine. The
23539 table is exact at each instruction boundary, so it can be used for stack
23540 unwinding from asynchronous events (such as debugger or garbage collector).
23542 @item -fno-gnu-unique
23543 @opindex fno-gnu-unique
23544 On systems with recent GNU assembler and C library, the C++ compiler
23545 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23546 of template static data members and static local variables in inline
23547 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23548 is necessary to avoid problems with a library used by two different
23549 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23550 therefore disagreeing with the other one about the binding of the
23551 symbol. But this causes @code{dlclose} to be ignored for affected
23552 DSOs; if your program relies on reinitialization of a DSO via
23553 @code{dlclose} and @code{dlopen}, you can use
23554 @option{-fno-gnu-unique}.
23556 @item -fpcc-struct-return
23557 @opindex fpcc-struct-return
23558 Return ``short'' @code{struct} and @code{union} values in memory like
23559 longer ones, rather than in registers. This convention is less
23560 efficient, but it has the advantage of allowing intercallability between
23561 GCC-compiled files and files compiled with other compilers, particularly
23562 the Portable C Compiler (pcc).
23564 The precise convention for returning structures in memory depends
23565 on the target configuration macros.
23567 Short structures and unions are those whose size and alignment match
23568 that of some integer type.
23570 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23571 switch is not binary compatible with code compiled with the
23572 @option{-freg-struct-return} switch.
23573 Use it to conform to a non-default application binary interface.
23575 @item -freg-struct-return
23576 @opindex freg-struct-return
23577 Return @code{struct} and @code{union} values in registers when possible.
23578 This is more efficient for small structures than
23579 @option{-fpcc-struct-return}.
23581 If you specify neither @option{-fpcc-struct-return} nor
23582 @option{-freg-struct-return}, GCC defaults to whichever convention is
23583 standard for the target. If there is no standard convention, GCC
23584 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23585 the principal compiler. In those cases, we can choose the standard, and
23586 we chose the more efficient register return alternative.
23588 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23589 switch is not binary compatible with code compiled with the
23590 @option{-fpcc-struct-return} switch.
23591 Use it to conform to a non-default application binary interface.
23593 @item -fshort-enums
23594 @opindex fshort-enums
23595 Allocate to an @code{enum} type only as many bytes as it needs for the
23596 declared range of possible values. Specifically, the @code{enum} type
23597 is equivalent to the smallest integer type that has enough room.
23599 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23600 code that is not binary compatible with code generated without that switch.
23601 Use it to conform to a non-default application binary interface.
23603 @item -fshort-double
23604 @opindex fshort-double
23605 Use the same size for @code{double} as for @code{float}.
23607 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23608 code that is not binary compatible with code generated without that switch.
23609 Use it to conform to a non-default application binary interface.
23611 @item -fshort-wchar
23612 @opindex fshort-wchar
23613 Override the underlying type for @code{wchar_t} to be @code{short
23614 unsigned int} instead of the default for the target. This option is
23615 useful for building programs to run under WINE@.
23617 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23618 code that is not binary compatible with code generated without that switch.
23619 Use it to conform to a non-default application binary interface.
23622 @opindex fno-common
23623 In C code, controls the placement of uninitialized global variables.
23624 Unix C compilers have traditionally permitted multiple definitions of
23625 such variables in different compilation units by placing the variables
23627 This is the behavior specified by @option{-fcommon}, and is the default
23628 for GCC on most targets.
23629 On the other hand, this behavior is not required by ISO C, and on some
23630 targets may carry a speed or code size penalty on variable references.
23631 The @option{-fno-common} option specifies that the compiler should place
23632 uninitialized global variables in the data section of the object file,
23633 rather than generating them as common blocks.
23634 This has the effect that if the same variable is declared
23635 (without @code{extern}) in two different compilations,
23636 you get a multiple-definition error when you link them.
23637 In this case, you must compile with @option{-fcommon} instead.
23638 Compiling with @option{-fno-common} is useful on targets for which
23639 it provides better performance, or if you wish to verify that the
23640 program will work on other systems that always treat uninitialized
23641 variable declarations this way.
23645 Ignore the @code{#ident} directive.
23647 @item -finhibit-size-directive
23648 @opindex finhibit-size-directive
23649 Don't output a @code{.size} assembler directive, or anything else that
23650 would cause trouble if the function is split in the middle, and the
23651 two halves are placed at locations far apart in memory. This option is
23652 used when compiling @file{crtstuff.c}; you should not need to use it
23655 @item -fverbose-asm
23656 @opindex fverbose-asm
23657 Put extra commentary information in the generated assembly code to
23658 make it more readable. This option is generally only of use to those
23659 who actually need to read the generated assembly code (perhaps while
23660 debugging the compiler itself).
23662 @option{-fno-verbose-asm}, the default, causes the
23663 extra information to be omitted and is useful when comparing two assembler
23666 @item -frecord-gcc-switches
23667 @opindex frecord-gcc-switches
23668 This switch causes the command line used to invoke the
23669 compiler to be recorded into the object file that is being created.
23670 This switch is only implemented on some targets and the exact format
23671 of the recording is target and binary file format dependent, but it
23672 usually takes the form of a section containing ASCII text. This
23673 switch is related to the @option{-fverbose-asm} switch, but that
23674 switch only records information in the assembler output file as
23675 comments, so it never reaches the object file.
23676 See also @option{-grecord-gcc-switches} for another
23677 way of storing compiler options into the object file.
23681 @cindex global offset table
23683 Generate position-independent code (PIC) suitable for use in a shared
23684 library, if supported for the target machine. Such code accesses all
23685 constant addresses through a global offset table (GOT)@. The dynamic
23686 loader resolves the GOT entries when the program starts (the dynamic
23687 loader is not part of GCC; it is part of the operating system). If
23688 the GOT size for the linked executable exceeds a machine-specific
23689 maximum size, you get an error message from the linker indicating that
23690 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23691 instead. (These maximums are 8k on the SPARC and 32k
23692 on the m68k and RS/6000. The x86 has no such limit.)
23694 Position-independent code requires special support, and therefore works
23695 only on certain machines. For the x86, GCC supports PIC for System V
23696 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23697 position-independent.
23699 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23704 If supported for the target machine, emit position-independent code,
23705 suitable for dynamic linking and avoiding any limit on the size of the
23706 global offset table. This option makes a difference on the m68k,
23707 PowerPC and SPARC@.
23709 Position-independent code requires special support, and therefore works
23710 only on certain machines.
23712 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23719 These options are similar to @option{-fpic} and @option{-fPIC}, but
23720 generated position independent code can be only linked into executables.
23721 Usually these options are used when @option{-pie} GCC option is
23722 used during linking.
23724 @option{-fpie} and @option{-fPIE} both define the macros
23725 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23726 for @option{-fpie} and 2 for @option{-fPIE}.
23728 @item -fno-jump-tables
23729 @opindex fno-jump-tables
23730 Do not use jump tables for switch statements even where it would be
23731 more efficient than other code generation strategies. This option is
23732 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23733 building code that forms part of a dynamic linker and cannot
23734 reference the address of a jump table. On some targets, jump tables
23735 do not require a GOT and this option is not needed.
23737 @item -ffixed-@var{reg}
23739 Treat the register named @var{reg} as a fixed register; generated code
23740 should never refer to it (except perhaps as a stack pointer, frame
23741 pointer or in some other fixed role).
23743 @var{reg} must be the name of a register. The register names accepted
23744 are machine-specific and are defined in the @code{REGISTER_NAMES}
23745 macro in the machine description macro file.
23747 This flag does not have a negative form, because it specifies a
23750 @item -fcall-used-@var{reg}
23751 @opindex fcall-used
23752 Treat the register named @var{reg} as an allocable register that is
23753 clobbered by function calls. It may be allocated for temporaries or
23754 variables that do not live across a call. Functions compiled this way
23755 do not save and restore the register @var{reg}.
23757 It is an error to use this flag with the frame pointer or stack pointer.
23758 Use of this flag for other registers that have fixed pervasive roles in
23759 the machine's execution model produces disastrous results.
23761 This flag does not have a negative form, because it specifies a
23764 @item -fcall-saved-@var{reg}
23765 @opindex fcall-saved
23766 Treat the register named @var{reg} as an allocable register saved by
23767 functions. It may be allocated even for temporaries or variables that
23768 live across a call. Functions compiled this way save and restore
23769 the register @var{reg} if they use it.
23771 It is an error to use this flag with the frame pointer or stack pointer.
23772 Use of this flag for other registers that have fixed pervasive roles in
23773 the machine's execution model produces disastrous results.
23775 A different sort of disaster results from the use of this flag for
23776 a register in which function values may be returned.
23778 This flag does not have a negative form, because it specifies a
23781 @item -fpack-struct[=@var{n}]
23782 @opindex fpack-struct
23783 Without a value specified, pack all structure members together without
23784 holes. When a value is specified (which must be a small power of two), pack
23785 structure members according to this value, representing the maximum
23786 alignment (that is, objects with default alignment requirements larger than
23787 this are output potentially unaligned at the next fitting location.
23789 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
23790 code that is not binary compatible with code generated without that switch.
23791 Additionally, it makes the code suboptimal.
23792 Use it to conform to a non-default application binary interface.
23794 @item -finstrument-functions
23795 @opindex finstrument-functions
23796 Generate instrumentation calls for entry and exit to functions. Just
23797 after function entry and just before function exit, the following
23798 profiling functions are called with the address of the current
23799 function and its call site. (On some platforms,
23800 @code{__builtin_return_address} does not work beyond the current
23801 function, so the call site information may not be available to the
23802 profiling functions otherwise.)
23805 void __cyg_profile_func_enter (void *this_fn,
23807 void __cyg_profile_func_exit (void *this_fn,
23811 The first argument is the address of the start of the current function,
23812 which may be looked up exactly in the symbol table.
23814 This instrumentation is also done for functions expanded inline in other
23815 functions. The profiling calls indicate where, conceptually, the
23816 inline function is entered and exited. This means that addressable
23817 versions of such functions must be available. If all your uses of a
23818 function are expanded inline, this may mean an additional expansion of
23819 code size. If you use @code{extern inline} in your C code, an
23820 addressable version of such functions must be provided. (This is
23821 normally the case anyway, but if you get lucky and the optimizer always
23822 expands the functions inline, you might have gotten away without
23823 providing static copies.)
23825 A function may be given the attribute @code{no_instrument_function}, in
23826 which case this instrumentation is not done. This can be used, for
23827 example, for the profiling functions listed above, high-priority
23828 interrupt routines, and any functions from which the profiling functions
23829 cannot safely be called (perhaps signal handlers, if the profiling
23830 routines generate output or allocate memory).
23832 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
23833 @opindex finstrument-functions-exclude-file-list
23835 Set the list of functions that are excluded from instrumentation (see
23836 the description of @option{-finstrument-functions}). If the file that
23837 contains a function definition matches with one of @var{file}, then
23838 that function is not instrumented. The match is done on substrings:
23839 if the @var{file} parameter is a substring of the file name, it is
23840 considered to be a match.
23845 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
23849 excludes any inline function defined in files whose pathnames
23850 contain @file{/bits/stl} or @file{include/sys}.
23852 If, for some reason, you want to include letter @samp{,} in one of
23853 @var{sym}, write @samp{\,}. For example,
23854 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
23855 (note the single quote surrounding the option).
23857 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
23858 @opindex finstrument-functions-exclude-function-list
23860 This is similar to @option{-finstrument-functions-exclude-file-list},
23861 but this option sets the list of function names to be excluded from
23862 instrumentation. The function name to be matched is its user-visible
23863 name, such as @code{vector<int> blah(const vector<int> &)}, not the
23864 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
23865 match is done on substrings: if the @var{sym} parameter is a substring
23866 of the function name, it is considered to be a match. For C99 and C++
23867 extended identifiers, the function name must be given in UTF-8, not
23868 using universal character names.
23870 @item -fstack-check
23871 @opindex fstack-check
23872 Generate code to verify that you do not go beyond the boundary of the
23873 stack. You should specify this flag if you are running in an
23874 environment with multiple threads, but you only rarely need to specify it in
23875 a single-threaded environment since stack overflow is automatically
23876 detected on nearly all systems if there is only one stack.
23878 Note that this switch does not actually cause checking to be done; the
23879 operating system or the language runtime must do that. The switch causes
23880 generation of code to ensure that they see the stack being extended.
23882 You can additionally specify a string parameter: @samp{no} means no
23883 checking, @samp{generic} means force the use of old-style checking,
23884 @samp{specific} means use the best checking method and is equivalent
23885 to bare @option{-fstack-check}.
23887 Old-style checking is a generic mechanism that requires no specific
23888 target support in the compiler but comes with the following drawbacks:
23892 Modified allocation strategy for large objects: they are always
23893 allocated dynamically if their size exceeds a fixed threshold.
23896 Fixed limit on the size of the static frame of functions: when it is
23897 topped by a particular function, stack checking is not reliable and
23898 a warning is issued by the compiler.
23901 Inefficiency: because of both the modified allocation strategy and the
23902 generic implementation, code performance is hampered.
23905 Note that old-style stack checking is also the fallback method for
23906 @samp{specific} if no target support has been added in the compiler.
23908 @item -fstack-limit-register=@var{reg}
23909 @itemx -fstack-limit-symbol=@var{sym}
23910 @itemx -fno-stack-limit
23911 @opindex fstack-limit-register
23912 @opindex fstack-limit-symbol
23913 @opindex fno-stack-limit
23914 Generate code to ensure that the stack does not grow beyond a certain value,
23915 either the value of a register or the address of a symbol. If a larger
23916 stack is required, a signal is raised at run time. For most targets,
23917 the signal is raised before the stack overruns the boundary, so
23918 it is possible to catch the signal without taking special precautions.
23920 For instance, if the stack starts at absolute address @samp{0x80000000}
23921 and grows downwards, you can use the flags
23922 @option{-fstack-limit-symbol=__stack_limit} and
23923 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
23924 of 128KB@. Note that this may only work with the GNU linker.
23926 @item -fsplit-stack
23927 @opindex fsplit-stack
23928 Generate code to automatically split the stack before it overflows.
23929 The resulting program has a discontiguous stack which can only
23930 overflow if the program is unable to allocate any more memory. This
23931 is most useful when running threaded programs, as it is no longer
23932 necessary to calculate a good stack size to use for each thread. This
23933 is currently only implemented for the x86 targets running
23936 When code compiled with @option{-fsplit-stack} calls code compiled
23937 without @option{-fsplit-stack}, there may not be much stack space
23938 available for the latter code to run. If compiling all code,
23939 including library code, with @option{-fsplit-stack} is not an option,
23940 then the linker can fix up these calls so that the code compiled
23941 without @option{-fsplit-stack} always has a large stack. Support for
23942 this is implemented in the gold linker in GNU binutils release 2.21
23945 @item -fleading-underscore
23946 @opindex fleading-underscore
23947 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
23948 change the way C symbols are represented in the object file. One use
23949 is to help link with legacy assembly code.
23951 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
23952 generate code that is not binary compatible with code generated without that
23953 switch. Use it to conform to a non-default application binary interface.
23954 Not all targets provide complete support for this switch.
23956 @item -ftls-model=@var{model}
23957 @opindex ftls-model
23958 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
23959 The @var{model} argument should be one of @samp{global-dynamic},
23960 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
23961 Note that the choice is subject to optimization: the compiler may use
23962 a more efficient model for symbols not visible outside of the translation
23963 unit, or if @option{-fpic} is not given on the command line.
23965 The default without @option{-fpic} is @samp{initial-exec}; with
23966 @option{-fpic} the default is @samp{global-dynamic}.
23968 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
23969 @opindex fvisibility
23970 Set the default ELF image symbol visibility to the specified option---all
23971 symbols are marked with this unless overridden within the code.
23972 Using this feature can very substantially improve linking and
23973 load times of shared object libraries, produce more optimized
23974 code, provide near-perfect API export and prevent symbol clashes.
23975 It is @strong{strongly} recommended that you use this in any shared objects
23978 Despite the nomenclature, @samp{default} always means public; i.e.,
23979 available to be linked against from outside the shared object.
23980 @samp{protected} and @samp{internal} are pretty useless in real-world
23981 usage so the only other commonly used option is @samp{hidden}.
23982 The default if @option{-fvisibility} isn't specified is
23983 @samp{default}, i.e., make every symbol public.
23985 A good explanation of the benefits offered by ensuring ELF
23986 symbols have the correct visibility is given by ``How To Write
23987 Shared Libraries'' by Ulrich Drepper (which can be found at
23988 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
23989 solution made possible by this option to marking things hidden when
23990 the default is public is to make the default hidden and mark things
23991 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
23992 and @code{__attribute__ ((visibility("default")))} instead of
23993 @code{__declspec(dllexport)} you get almost identical semantics with
23994 identical syntax. This is a great boon to those working with
23995 cross-platform projects.
23997 For those adding visibility support to existing code, you may find
23998 @code{#pragma GCC visibility} of use. This works by you enclosing
23999 the declarations you wish to set visibility for with (for example)
24000 @code{#pragma GCC visibility push(hidden)} and
24001 @code{#pragma GCC visibility pop}.
24002 Bear in mind that symbol visibility should be viewed @strong{as
24003 part of the API interface contract} and thus all new code should
24004 always specify visibility when it is not the default; i.e., declarations
24005 only for use within the local DSO should @strong{always} be marked explicitly
24006 as hidden as so to avoid PLT indirection overheads---making this
24007 abundantly clear also aids readability and self-documentation of the code.
24008 Note that due to ISO C++ specification requirements, @code{operator new} and
24009 @code{operator delete} must always be of default visibility.
24011 Be aware that headers from outside your project, in particular system
24012 headers and headers from any other library you use, may not be
24013 expecting to be compiled with visibility other than the default. You
24014 may need to explicitly say @code{#pragma GCC visibility push(default)}
24015 before including any such headers.
24017 @code{extern} declarations are not affected by @option{-fvisibility}, so
24018 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24019 no modifications. However, this means that calls to @code{extern}
24020 functions with no explicit visibility use the PLT, so it is more
24021 effective to use @code{__attribute ((visibility))} and/or
24022 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24023 declarations should be treated as hidden.
24025 Note that @option{-fvisibility} does affect C++ vague linkage
24026 entities. This means that, for instance, an exception class that is
24027 be thrown between DSOs must be explicitly marked with default
24028 visibility so that the @samp{type_info} nodes are unified between
24031 An overview of these techniques, their benefits and how to use them
24032 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24034 @item -fstrict-volatile-bitfields
24035 @opindex fstrict-volatile-bitfields
24036 This option should be used if accesses to volatile bit-fields (or other
24037 structure fields, although the compiler usually honors those types
24038 anyway) should use a single access of the width of the
24039 field's type, aligned to a natural alignment if possible. For
24040 example, targets with memory-mapped peripheral registers might require
24041 all such accesses to be 16 bits wide; with this flag you can
24042 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24043 is 16 bits on these targets) to force GCC to use 16-bit accesses
24044 instead of, perhaps, a more efficient 32-bit access.
24046 If this option is disabled, the compiler uses the most efficient
24047 instruction. In the previous example, that might be a 32-bit load
24048 instruction, even though that accesses bytes that do not contain
24049 any portion of the bit-field, or memory-mapped registers unrelated to
24050 the one being updated.
24052 In some cases, such as when the @code{packed} attribute is applied to a
24053 structure field, it may not be possible to access the field with a single
24054 read or write that is correctly aligned for the target machine. In this
24055 case GCC falls back to generating multiple accesses rather than code that
24056 will fault or truncate the result at run time.
24058 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24059 not allowed to touch non bit-field members. It is therefore recommended
24060 to define all bits of the field's type as bit-field members.
24062 The default value of this option is determined by the application binary
24063 interface for the target processor.
24065 @item -fsync-libcalls
24066 @opindex fsync-libcalls
24067 This option controls whether any out-of-line instance of the @code{__sync}
24068 family of functions may be used to implement the C++11 @code{__atomic}
24069 family of functions.
24071 The default value of this option is enabled, thus the only useful form
24072 of the option is @option{-fno-sync-libcalls}. This option is used in
24073 the implementation of the @file{libatomic} runtime library.
24079 @node Environment Variables
24080 @section Environment Variables Affecting GCC
24081 @cindex environment variables
24083 @c man begin ENVIRONMENT
24084 This section describes several environment variables that affect how GCC
24085 operates. Some of them work by specifying directories or prefixes to use
24086 when searching for various kinds of files. Some are used to specify other
24087 aspects of the compilation environment.
24089 Note that you can also specify places to search using options such as
24090 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24091 take precedence over places specified using environment variables, which
24092 in turn take precedence over those specified by the configuration of GCC@.
24093 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24094 GNU Compiler Collection (GCC) Internals}.
24099 @c @itemx LC_COLLATE
24101 @c @itemx LC_MONETARY
24102 @c @itemx LC_NUMERIC
24107 @c @findex LC_COLLATE
24108 @findex LC_MESSAGES
24109 @c @findex LC_MONETARY
24110 @c @findex LC_NUMERIC
24114 These environment variables control the way that GCC uses
24115 localization information which allows GCC to work with different
24116 national conventions. GCC inspects the locale categories
24117 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24118 so. These locale categories can be set to any value supported by your
24119 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24120 Kingdom encoded in UTF-8.
24122 The @env{LC_CTYPE} environment variable specifies character
24123 classification. GCC uses it to determine the character boundaries in
24124 a string; this is needed for some multibyte encodings that contain quote
24125 and escape characters that are otherwise interpreted as a string
24128 The @env{LC_MESSAGES} environment variable specifies the language to
24129 use in diagnostic messages.
24131 If the @env{LC_ALL} environment variable is set, it overrides the value
24132 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24133 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24134 environment variable. If none of these variables are set, GCC
24135 defaults to traditional C English behavior.
24139 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24140 files. GCC uses temporary files to hold the output of one stage of
24141 compilation which is to be used as input to the next stage: for example,
24142 the output of the preprocessor, which is the input to the compiler
24145 @item GCC_COMPARE_DEBUG
24146 @findex GCC_COMPARE_DEBUG
24147 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24148 @option{-fcompare-debug} to the compiler driver. See the documentation
24149 of this option for more details.
24151 @item GCC_EXEC_PREFIX
24152 @findex GCC_EXEC_PREFIX
24153 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24154 names of the subprograms executed by the compiler. No slash is added
24155 when this prefix is combined with the name of a subprogram, but you can
24156 specify a prefix that ends with a slash if you wish.
24158 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24159 an appropriate prefix to use based on the pathname it is invoked with.
24161 If GCC cannot find the subprogram using the specified prefix, it
24162 tries looking in the usual places for the subprogram.
24164 The default value of @env{GCC_EXEC_PREFIX} is
24165 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24166 the installed compiler. In many cases @var{prefix} is the value
24167 of @code{prefix} when you ran the @file{configure} script.
24169 Other prefixes specified with @option{-B} take precedence over this prefix.
24171 This prefix is also used for finding files such as @file{crt0.o} that are
24174 In addition, the prefix is used in an unusual way in finding the
24175 directories to search for header files. For each of the standard
24176 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24177 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24178 replacing that beginning with the specified prefix to produce an
24179 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24180 @file{foo/bar} just before it searches the standard directory
24181 @file{/usr/local/lib/bar}.
24182 If a standard directory begins with the configured
24183 @var{prefix} then the value of @var{prefix} is replaced by
24184 @env{GCC_EXEC_PREFIX} when looking for header files.
24186 @item COMPILER_PATH
24187 @findex COMPILER_PATH
24188 The value of @env{COMPILER_PATH} is a colon-separated list of
24189 directories, much like @env{PATH}. GCC tries the directories thus
24190 specified when searching for subprograms, if it can't find the
24191 subprograms using @env{GCC_EXEC_PREFIX}.
24194 @findex LIBRARY_PATH
24195 The value of @env{LIBRARY_PATH} is a colon-separated list of
24196 directories, much like @env{PATH}. When configured as a native compiler,
24197 GCC tries the directories thus specified when searching for special
24198 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24199 using GCC also uses these directories when searching for ordinary
24200 libraries for the @option{-l} option (but directories specified with
24201 @option{-L} come first).
24205 @cindex locale definition
24206 This variable is used to pass locale information to the compiler. One way in
24207 which this information is used is to determine the character set to be used
24208 when character literals, string literals and comments are parsed in C and C++.
24209 When the compiler is configured to allow multibyte characters,
24210 the following values for @env{LANG} are recognized:
24214 Recognize JIS characters.
24216 Recognize SJIS characters.
24218 Recognize EUCJP characters.
24221 If @env{LANG} is not defined, or if it has some other value, then the
24222 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24223 recognize and translate multibyte characters.
24227 Some additional environment variables affect the behavior of the
24230 @include cppenv.texi
24234 @node Precompiled Headers
24235 @section Using Precompiled Headers
24236 @cindex precompiled headers
24237 @cindex speed of compilation
24239 Often large projects have many header files that are included in every
24240 source file. The time the compiler takes to process these header files
24241 over and over again can account for nearly all of the time required to
24242 build the project. To make builds faster, GCC allows you to
24243 @dfn{precompile} a header file.
24245 To create a precompiled header file, simply compile it as you would any
24246 other file, if necessary using the @option{-x} option to make the driver
24247 treat it as a C or C++ header file. You may want to use a
24248 tool like @command{make} to keep the precompiled header up-to-date when
24249 the headers it contains change.
24251 A precompiled header file is searched for when @code{#include} is
24252 seen in the compilation. As it searches for the included file
24253 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24254 compiler looks for a precompiled header in each directory just before it
24255 looks for the include file in that directory. The name searched for is
24256 the name specified in the @code{#include} with @samp{.gch} appended. If
24257 the precompiled header file can't be used, it is ignored.
24259 For instance, if you have @code{#include "all.h"}, and you have
24260 @file{all.h.gch} in the same directory as @file{all.h}, then the
24261 precompiled header file is used if possible, and the original
24262 header is used otherwise.
24264 Alternatively, you might decide to put the precompiled header file in a
24265 directory and use @option{-I} to ensure that directory is searched
24266 before (or instead of) the directory containing the original header.
24267 Then, if you want to check that the precompiled header file is always
24268 used, you can put a file of the same name as the original header in this
24269 directory containing an @code{#error} command.
24271 This also works with @option{-include}. So yet another way to use
24272 precompiled headers, good for projects not designed with precompiled
24273 header files in mind, is to simply take most of the header files used by
24274 a project, include them from another header file, precompile that header
24275 file, and @option{-include} the precompiled header. If the header files
24276 have guards against multiple inclusion, they are skipped because
24277 they've already been included (in the precompiled header).
24279 If you need to precompile the same header file for different
24280 languages, targets, or compiler options, you can instead make a
24281 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24282 header in the directory, perhaps using @option{-o}. It doesn't matter
24283 what you call the files in the directory; every precompiled header in
24284 the directory is considered. The first precompiled header
24285 encountered in the directory that is valid for this compilation is
24286 used; they're searched in no particular order.
24288 There are many other possibilities, limited only by your imagination,
24289 good sense, and the constraints of your build system.
24291 A precompiled header file can be used only when these conditions apply:
24295 Only one precompiled header can be used in a particular compilation.
24298 A precompiled header can't be used once the first C token is seen. You
24299 can have preprocessor directives before a precompiled header; you cannot
24300 include a precompiled header from inside another header.
24303 The precompiled header file must be produced for the same language as
24304 the current compilation. You can't use a C precompiled header for a C++
24308 The precompiled header file must have been produced by the same compiler
24309 binary as the current compilation is using.
24312 Any macros defined before the precompiled header is included must
24313 either be defined in the same way as when the precompiled header was
24314 generated, or must not affect the precompiled header, which usually
24315 means that they don't appear in the precompiled header at all.
24317 The @option{-D} option is one way to define a macro before a
24318 precompiled header is included; using a @code{#define} can also do it.
24319 There are also some options that define macros implicitly, like
24320 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
24323 @item If debugging information is output when using the precompiled
24324 header, using @option{-g} or similar, the same kind of debugging information
24325 must have been output when building the precompiled header. However,
24326 a precompiled header built using @option{-g} can be used in a compilation
24327 when no debugging information is being output.
24329 @item The same @option{-m} options must generally be used when building
24330 and using the precompiled header. @xref{Submodel Options},
24331 for any cases where this rule is relaxed.
24333 @item Each of the following options must be the same when building and using
24334 the precompiled header:
24336 @gccoptlist{-fexceptions}
24339 Some other command-line options starting with @option{-f},
24340 @option{-p}, or @option{-O} must be defined in the same way as when
24341 the precompiled header was generated. At present, it's not clear
24342 which options are safe to change and which are not; the safest choice
24343 is to use exactly the same options when generating and using the
24344 precompiled header. The following are known to be safe:
24346 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24347 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24348 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24353 For all of these except the last, the compiler automatically
24354 ignores the precompiled header if the conditions aren't met. If you
24355 find an option combination that doesn't work and doesn't cause the
24356 precompiled header to be ignored, please consider filing a bug report,
24359 If you do use differing options when generating and using the
24360 precompiled header, the actual behavior is a mixture of the
24361 behavior for the options. For instance, if you use @option{-g} to
24362 generate the precompiled header but not when using it, you may or may
24363 not get debugging information for routines in the precompiled header.