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 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
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 -Wmultiple-inheritance @gol
202 -Wnamespaces -Wnarrowing @gol
203 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
204 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
205 -Wno-non-template-friend -Wold-style-cast @gol
206 -Woverloaded-virtual -Wno-pmf-conversions @gol
207 -Wsign-promo -Wvirtual-inheritance}
209 @item Objective-C and Objective-C++ Language Options
210 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
211 Objective-C and Objective-C++ Dialects}.
212 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
213 -fgnu-runtime -fnext-runtime @gol
214 -fno-nil-receivers @gol
215 -fobjc-abi-version=@var{n} @gol
216 -fobjc-call-cxx-cdtors @gol
217 -fobjc-direct-dispatch @gol
218 -fobjc-exceptions @gol
221 -fobjc-std=objc1 @gol
222 -fno-local-ivars @gol
223 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
224 -freplace-objc-classes @gol
227 -Wassign-intercept @gol
228 -Wno-protocol -Wselector @gol
229 -Wstrict-selector-match @gol
230 -Wundeclared-selector}
232 @item Diagnostic Message Formatting Options
233 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
234 @gccoptlist{-fmessage-length=@var{n} @gol
235 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
236 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
237 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
239 @item Warning Options
240 @xref{Warning Options,,Options to Request or Suppress Warnings}.
241 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
242 -pedantic-errors @gol
243 -w -Wextra -Wall -Waddress -Waggregate-return @gol
244 -Waggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
245 -Wbool-compare -Wduplicated-cond -Wframe-address @gol
246 -Wno-attributes -Wno-builtin-macro-redefined @gol
247 -Wc90-c99-compat -Wc99-c11-compat @gol
248 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
249 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
250 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
251 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
252 -Wdisabled-optimization @gol
253 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
254 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
255 -Wno-endif-labels -Werror -Werror=* @gol
256 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
257 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
258 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
259 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
260 -Wignored-qualifiers -Wincompatible-pointer-types @gol
261 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
262 -Winit-self -Winline -Wno-int-conversion @gol
263 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
264 -Wnull-dereference @gol
265 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
266 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
267 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args @gol
268 -Wmisleading-indentation -Wmissing-braces @gol
269 -Wmissing-field-initializers -Wmissing-include-dirs @gol
270 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
271 -Wodr -Wno-overflow -Wopenmp-simd @gol
272 -Woverride-init-side-effects @gol
273 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
274 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
275 -Wpointer-arith -Wno-pointer-to-int-cast @gol
276 -Wredundant-decls -Wno-return-local-addr @gol
277 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
278 -Wshift-overflow -Wshift-overflow=@var{n} @gol
279 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
280 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
281 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
282 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
283 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
284 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
285 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
286 -Wmissing-format-attribute -Wsubobject-linkage @gol
287 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
288 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
289 -Wtype-limits -Wundef @gol
290 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
291 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
292 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
293 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
294 -Wunused-const-variable @gol
295 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
296 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
297 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
298 -Wzero-as-null-pointer-constant}
300 @item C and Objective-C-only Warning Options
301 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
302 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
303 -Wold-style-declaration -Wold-style-definition @gol
304 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
305 -Wdeclaration-after-statement -Wpointer-sign}
307 @item Debugging Options
308 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
309 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
310 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
311 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
312 -fsanitize-undefined-trap-on-error @gol
313 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
314 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
315 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
316 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
317 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
318 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
319 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
320 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
321 -fchkp-use-wrappers @gol
322 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
323 -fdisable-ipa-@var{pass_name} @gol
324 -fdisable-rtl-@var{pass_name} @gol
325 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
326 -fdisable-tree-@var{pass_name} @gol
327 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
328 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
329 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
330 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
331 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
333 -fdump-statistics @gol
335 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
337 -fdump-tree-cfg -fdump-tree-alias @gol
339 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
340 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
341 -fdump-tree-gimple@r{[}-raw@r{]} @gol
342 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
343 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
344 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
347 -fdump-tree-nrv -fdump-tree-vect @gol
348 -fdump-tree-sink @gol
349 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
350 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
351 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
352 -fdump-tree-vtable-verify @gol
353 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
354 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
355 -fdump-final-insns=@var{file} @gol
356 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
357 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
358 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
359 -fenable-@var{kind}-@var{pass} @gol
360 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
361 -fdebug-types-section -fmem-report-wpa @gol
362 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
364 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
365 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
366 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
367 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
368 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
369 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
370 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
371 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
372 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
373 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
374 -fdebug-prefix-map=@var{old}=@var{new} @gol
375 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
376 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
377 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
378 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
379 -print-prog-name=@var{program} -print-search-dirs -Q @gol
380 -print-sysroot -print-sysroot-headers-suffix @gol
381 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
383 @item Optimization Options
384 @xref{Optimize Options,,Options that Control Optimization}.
385 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
386 -falign-jumps[=@var{n}] @gol
387 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
388 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
389 -fauto-inc-dec -fbranch-probabilities @gol
390 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
391 -fbtr-bb-exclusive -fcaller-saves @gol
392 -fcombine-stack-adjustments -fconserve-stack @gol
393 -fcompare-elim -fcprop-registers -fcrossjumping @gol
394 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
395 -fcx-limited-range @gol
396 -fdata-sections -fdce -fdelayed-branch @gol
397 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
398 -fdevirtualize-at-ltrans -fdse @gol
399 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
400 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
401 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
402 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
403 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
404 -fif-conversion2 -findirect-inlining @gol
405 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
406 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
407 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
408 -fira-algorithm=@var{algorithm} @gol
409 -fira-region=@var{region} -fira-hoist-pressure @gol
410 -fira-loop-pressure -fno-ira-share-save-slots @gol
411 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
412 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
413 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
414 -fkeep-static-consts -flive-range-shrinkage @gol
415 -floop-block -floop-interchange -floop-strip-mine @gol
416 -floop-unroll-and-jam -floop-nest-optimize @gol
417 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
418 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
419 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
420 -fmove-loop-invariants -fno-branch-count-reg @gol
421 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
422 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
423 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
424 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
425 -fomit-frame-pointer -foptimize-sibling-calls @gol
426 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
427 -fprefetch-loop-arrays -fprofile-report @gol
428 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
429 -fprofile-generate=@var{path} @gol
430 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
431 -fprofile-reorder-functions @gol
432 -freciprocal-math -free -frename-registers -freorder-blocks @gol
433 -freorder-blocks-algorithm=@var{algorithm} @gol
434 -freorder-blocks-and-partition -freorder-functions @gol
435 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
436 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
437 -fsched-spec-load -fsched-spec-load-dangerous @gol
438 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
439 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
440 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
441 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
442 -fschedule-fusion @gol
443 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
444 -fselective-scheduling -fselective-scheduling2 @gol
445 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
446 -fsemantic-interposition @gol
447 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
448 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt @gol
449 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
450 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
451 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
452 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
453 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
454 -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
455 -ftree-loop-if-convert-stores -ftree-loop-im @gol
456 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
457 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
458 -ftree-loop-vectorize @gol
459 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
460 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
461 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
462 -ftree-vectorize -ftree-vrp @gol
463 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
464 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
465 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
466 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
467 --param @var{name}=@var{value}
468 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
470 @item Preprocessor Options
471 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
472 @gccoptlist{-A@var{question}=@var{answer} @gol
473 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
474 -C -dD -dI -dM -dN @gol
475 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
476 -idirafter @var{dir} @gol
477 -include @var{file} -imacros @var{file} @gol
478 -iprefix @var{file} -iwithprefix @var{dir} @gol
479 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
480 -imultilib @var{dir} -isysroot @var{dir} @gol
481 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
482 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
483 -remap -trigraphs -undef -U@var{macro} @gol
484 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
486 @item Assembler Option
487 @xref{Assembler Options,,Passing Options to the Assembler}.
488 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
491 @xref{Link Options,,Options for Linking}.
492 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
493 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
494 -s -static -static-libgcc -static-libstdc++ @gol
495 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
496 -static-libmpx -static-libmpxwrappers @gol
497 -shared -shared-libgcc -symbolic @gol
498 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
499 -u @var{symbol} -z @var{keyword}}
501 @item Directory Options
502 @xref{Directory Options,,Options for Directory Search}.
503 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
504 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
505 --sysroot=@var{dir} --no-sysroot-suffix}
507 @item Machine Dependent Options
508 @xref{Submodel Options,,Hardware Models and Configurations}.
509 @c This list is ordered alphanumerically by subsection name.
510 @c Try and put the significant identifier (CPU or system) first,
511 @c so users have a clue at guessing where the ones they want will be.
513 @emph{AArch64 Options}
514 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
515 -mgeneral-regs-only @gol
516 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
518 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
519 -mtls-dialect=desc -mtls-dialect=traditional @gol
520 -mtls-size=@var{size} @gol
521 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
522 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
523 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
525 @emph{Adapteva Epiphany Options}
526 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
527 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
528 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
529 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
530 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
531 -msplit-vecmove-early -m1reg-@var{reg}}
534 @gccoptlist{-mbarrel-shifter @gol
535 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
536 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
537 -mea -mno-mpy -mmul32x16 -mmul64 @gol
538 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
539 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
540 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
541 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
542 -mucb-mcount -mvolatile-cache @gol
543 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
544 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
545 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
546 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
547 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
548 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
551 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
552 -mabi=@var{name} @gol
553 -mapcs-stack-check -mno-apcs-stack-check @gol
554 -mapcs-float -mno-apcs-float @gol
555 -mapcs-reentrant -mno-apcs-reentrant @gol
556 -msched-prolog -mno-sched-prolog @gol
557 -mlittle-endian -mbig-endian @gol
558 -mfloat-abi=@var{name} @gol
559 -mfp16-format=@var{name}
560 -mthumb-interwork -mno-thumb-interwork @gol
561 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
562 -mtune=@var{name} -mprint-tune-info @gol
563 -mstructure-size-boundary=@var{n} @gol
564 -mabort-on-noreturn @gol
565 -mlong-calls -mno-long-calls @gol
566 -msingle-pic-base -mno-single-pic-base @gol
567 -mpic-register=@var{reg} @gol
568 -mnop-fun-dllimport @gol
569 -mpoke-function-name @gol
571 -mtpcs-frame -mtpcs-leaf-frame @gol
572 -mcaller-super-interworking -mcallee-super-interworking @gol
573 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
574 -mword-relocations @gol
575 -mfix-cortex-m3-ldrd @gol
576 -munaligned-access @gol
577 -mneon-for-64bits @gol
578 -mslow-flash-data @gol
579 -masm-syntax-unified @gol
583 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
584 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
585 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
587 @emph{Blackfin Options}
588 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
589 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
590 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
591 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
592 -mno-id-shared-library -mshared-library-id=@var{n} @gol
593 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
594 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
595 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
599 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
600 -msim -msdata=@var{sdata-type}}
603 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
604 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
605 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
606 -mstack-align -mdata-align -mconst-align @gol
607 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
608 -melf -maout -melinux -mlinux -sim -sim2 @gol
609 -mmul-bug-workaround -mno-mul-bug-workaround}
612 @gccoptlist{-mmac @gol
613 -mcr16cplus -mcr16c @gol
614 -msim -mint32 -mbit-ops
615 -mdata-model=@var{model}}
617 @emph{Darwin Options}
618 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
619 -arch_only -bind_at_load -bundle -bundle_loader @gol
620 -client_name -compatibility_version -current_version @gol
622 -dependency-file -dylib_file -dylinker_install_name @gol
623 -dynamic -dynamiclib -exported_symbols_list @gol
624 -filelist -flat_namespace -force_cpusubtype_ALL @gol
625 -force_flat_namespace -headerpad_max_install_names @gol
627 -image_base -init -install_name -keep_private_externs @gol
628 -multi_module -multiply_defined -multiply_defined_unused @gol
629 -noall_load -no_dead_strip_inits_and_terms @gol
630 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
631 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
632 -private_bundle -read_only_relocs -sectalign @gol
633 -sectobjectsymbols -whyload -seg1addr @gol
634 -sectcreate -sectobjectsymbols -sectorder @gol
635 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
636 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
637 -segprot -segs_read_only_addr -segs_read_write_addr @gol
638 -single_module -static -sub_library -sub_umbrella @gol
639 -twolevel_namespace -umbrella -undefined @gol
640 -unexported_symbols_list -weak_reference_mismatches @gol
641 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
642 -mkernel -mone-byte-bool}
644 @emph{DEC Alpha Options}
645 @gccoptlist{-mno-fp-regs -msoft-float @gol
646 -mieee -mieee-with-inexact -mieee-conformant @gol
647 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
648 -mtrap-precision=@var{mode} -mbuild-constants @gol
649 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
650 -mbwx -mmax -mfix -mcix @gol
651 -mfloat-vax -mfloat-ieee @gol
652 -mexplicit-relocs -msmall-data -mlarge-data @gol
653 -msmall-text -mlarge-text @gol
654 -mmemory-latency=@var{time}}
657 @gccoptlist{-msmall-model -mno-lsim}
660 @gccoptlist{-msim -mlra}
663 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
664 -mhard-float -msoft-float @gol
665 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
666 -mdouble -mno-double @gol
667 -mmedia -mno-media -mmuladd -mno-muladd @gol
668 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
669 -mlinked-fp -mlong-calls -malign-labels @gol
670 -mlibrary-pic -macc-4 -macc-8 @gol
671 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
672 -moptimize-membar -mno-optimize-membar @gol
673 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
674 -mvliw-branch -mno-vliw-branch @gol
675 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
676 -mno-nested-cond-exec -mtomcat-stats @gol
680 @emph{GNU/Linux Options}
681 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
682 -tno-android-cc -tno-android-ld}
684 @emph{H8/300 Options}
685 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
688 @gccoptlist{-march=@var{architecture-type} @gol
689 -mdisable-fpregs -mdisable-indexing @gol
690 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
691 -mfixed-range=@var{register-range} @gol
692 -mjump-in-delay -mlinker-opt -mlong-calls @gol
693 -mlong-load-store -mno-disable-fpregs @gol
694 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
695 -mno-jump-in-delay -mno-long-load-store @gol
696 -mno-portable-runtime -mno-soft-float @gol
697 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
698 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
699 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
700 -munix=@var{unix-std} -nolibdld -static -threads}
703 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
704 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
705 -mconstant-gp -mauto-pic -mfused-madd @gol
706 -minline-float-divide-min-latency @gol
707 -minline-float-divide-max-throughput @gol
708 -mno-inline-float-divide @gol
709 -minline-int-divide-min-latency @gol
710 -minline-int-divide-max-throughput @gol
711 -mno-inline-int-divide @gol
712 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
713 -mno-inline-sqrt @gol
714 -mdwarf2-asm -mearly-stop-bits @gol
715 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
716 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
717 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
718 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
719 -msched-spec-ldc -msched-spec-control-ldc @gol
720 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
721 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
722 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
723 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
726 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
727 -msign-extend-enabled -muser-enabled}
729 @emph{M32R/D Options}
730 @gccoptlist{-m32r2 -m32rx -m32r @gol
732 -malign-loops -mno-align-loops @gol
733 -missue-rate=@var{number} @gol
734 -mbranch-cost=@var{number} @gol
735 -mmodel=@var{code-size-model-type} @gol
736 -msdata=@var{sdata-type} @gol
737 -mno-flush-func -mflush-func=@var{name} @gol
738 -mno-flush-trap -mflush-trap=@var{number} @gol
742 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
744 @emph{M680x0 Options}
745 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
746 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
747 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
748 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
749 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
750 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
751 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
752 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
756 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
757 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
758 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
759 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
760 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
763 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
764 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
765 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
766 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
769 @emph{MicroBlaze Options}
770 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
771 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
772 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
773 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
774 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
777 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
778 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
779 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
780 -mips16 -mno-mips16 -mflip-mips16 @gol
781 -minterlink-compressed -mno-interlink-compressed @gol
782 -minterlink-mips16 -mno-interlink-mips16 @gol
783 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
784 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
785 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
786 -mno-float -msingle-float -mdouble-float @gol
787 -modd-spreg -mno-odd-spreg @gol
788 -mcompact-branches=@var{policy} @gol
789 -mabs=@var{mode} -mnan=@var{encoding} @gol
790 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
793 -mvirt -mno-virt @gol
795 -mmicromips -mno-micromips @gol
796 -mfpu=@var{fpu-type} @gol
797 -msmartmips -mno-smartmips @gol
798 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
799 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
800 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
801 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
802 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
803 -membedded-data -mno-embedded-data @gol
804 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
805 -mcode-readable=@var{setting} @gol
806 -msplit-addresses -mno-split-addresses @gol
807 -mexplicit-relocs -mno-explicit-relocs @gol
808 -mcheck-zero-division -mno-check-zero-division @gol
809 -mdivide-traps -mdivide-breaks @gol
810 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
811 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
812 -mfix-24k -mno-fix-24k @gol
813 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
814 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
815 -mfix-vr4120 -mno-fix-vr4120 @gol
816 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
817 -mflush-func=@var{func} -mno-flush-func @gol
818 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
819 -mfp-exceptions -mno-fp-exceptions @gol
820 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
821 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
822 -mframe-header-opt -mno-frame-header-opt}
825 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
826 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
827 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
828 -mno-base-addresses -msingle-exit -mno-single-exit}
830 @emph{MN10300 Options}
831 @gccoptlist{-mmult-bug -mno-mult-bug @gol
832 -mno-am33 -mam33 -mam33-2 -mam34 @gol
833 -mtune=@var{cpu-type} @gol
834 -mreturn-pointer-on-d0 @gol
835 -mno-crt0 -mrelax -mliw -msetlb}
838 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
840 @emph{MSP430 Options}
841 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
842 -mcode-region= -mdata-region= @gol
846 @gccoptlist{-mbig-endian -mlittle-endian @gol
847 -mreduced-regs -mfull-regs @gol
848 -mcmov -mno-cmov @gol
849 -mperf-ext -mno-perf-ext @gol
850 -mv3push -mno-v3push @gol
851 -m16bit -mno-16bit @gol
852 -misr-vector-size=@var{num} @gol
853 -mcache-block-size=@var{num} @gol
854 -march=@var{arch} @gol
855 -mcmodel=@var{code-model} @gol
858 @emph{Nios II Options}
859 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
861 -mno-bypass-cache -mbypass-cache @gol
862 -mno-cache-volatile -mcache-volatile @gol
863 -mno-fast-sw-div -mfast-sw-div @gol
864 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
865 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
866 -mcustom-fpu-cfg=@var{name} @gol
867 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
868 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
870 @emph{Nvidia PTX Options}
871 @gccoptlist{-m32 -m64 -mmainkernel}
873 @emph{PDP-11 Options}
874 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
875 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
876 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
877 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
878 -mbranch-expensive -mbranch-cheap @gol
879 -munix-asm -mdec-asm}
881 @emph{picoChip Options}
882 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
883 -msymbol-as-address -mno-inefficient-warnings}
885 @emph{PowerPC Options}
886 See RS/6000 and PowerPC Options.
889 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
890 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
891 -m64bit-doubles -m32bit-doubles}
893 @emph{RS/6000 and PowerPC Options}
894 @gccoptlist{-mcpu=@var{cpu-type} @gol
895 -mtune=@var{cpu-type} @gol
896 -mcmodel=@var{code-model} @gol
898 -maltivec -mno-altivec @gol
899 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
900 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
901 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
902 -mfprnd -mno-fprnd @gol
903 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
904 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
905 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
906 -malign-power -malign-natural @gol
907 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
908 -msingle-float -mdouble-float -msimple-fpu @gol
909 -mstring -mno-string -mupdate -mno-update @gol
910 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
911 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
912 -mstrict-align -mno-strict-align -mrelocatable @gol
913 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
914 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
915 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
916 -mprioritize-restricted-insns=@var{priority} @gol
917 -msched-costly-dep=@var{dependence_type} @gol
918 -minsert-sched-nops=@var{scheme} @gol
919 -mcall-sysv -mcall-netbsd @gol
920 -maix-struct-return -msvr4-struct-return @gol
921 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
922 -mblock-move-inline-limit=@var{num} @gol
923 -misel -mno-isel @gol
924 -misel=yes -misel=no @gol
926 -mspe=yes -mspe=no @gol
928 -mgen-cell-microcode -mwarn-cell-microcode @gol
929 -mvrsave -mno-vrsave @gol
930 -mmulhw -mno-mulhw @gol
931 -mdlmzb -mno-dlmzb @gol
932 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
933 -mprototype -mno-prototype @gol
934 -msim -mmvme -mads -myellowknife -memb -msdata @gol
935 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
936 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
937 -mno-recip-precision @gol
938 -mveclibabi=@var{type} -mfriz -mno-friz @gol
939 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
940 -msave-toc-indirect -mno-save-toc-indirect @gol
941 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
942 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
943 -mquad-memory -mno-quad-memory @gol
944 -mquad-memory-atomic -mno-quad-memory-atomic @gol
945 -mcompat-align-parm -mno-compat-align-parm @gol
946 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
947 -mupper-regs -mno-upper-regs}
950 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
952 -mbig-endian-data -mlittle-endian-data @gol
955 -mas100-syntax -mno-as100-syntax@gol
957 -mmax-constant-size=@gol
960 -mallow-string-insns -mno-allow-string-insns@gol
961 -mno-warn-multiple-fast-interrupts@gol
962 -msave-acc-in-interrupts}
964 @emph{S/390 and zSeries Options}
965 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
966 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
967 -mlong-double-64 -mlong-double-128 @gol
968 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
969 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
970 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
971 -mhtm -mvx -mzvector @gol
972 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
973 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
974 -mhotpatch=@var{halfwords},@var{halfwords}}
977 @gccoptlist{-meb -mel @gol
981 -mscore5 -mscore5u -mscore7 -mscore7d}
984 @gccoptlist{-m1 -m2 -m2e @gol
985 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
987 -m4-nofpu -m4-single-only -m4-single -m4 @gol
988 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
989 -mb -ml -mdalign -mrelax @gol
990 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
991 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
992 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
993 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
994 -maccumulate-outgoing-args @gol
995 -matomic-model=@var{atomic-model} @gol
996 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
997 -mcbranch-force-delay-slot @gol
998 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
999 -mpretend-cmove -mtas}
1001 @emph{Solaris 2 Options}
1002 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1005 @emph{SPARC Options}
1006 @gccoptlist{-mcpu=@var{cpu-type} @gol
1007 -mtune=@var{cpu-type} @gol
1008 -mcmodel=@var{code-model} @gol
1009 -mmemory-model=@var{mem-model} @gol
1010 -m32 -m64 -mapp-regs -mno-app-regs @gol
1011 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1012 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1013 -mhard-quad-float -msoft-quad-float @gol
1014 -mstack-bias -mno-stack-bias @gol
1015 -munaligned-doubles -mno-unaligned-doubles @gol
1016 -muser-mode -mno-user-mode @gol
1017 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1018 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1019 -mcbcond -mno-cbcond @gol
1020 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1021 -mfix-at697f -mfix-ut699}
1024 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1025 -msafe-dma -munsafe-dma @gol
1027 -msmall-mem -mlarge-mem -mstdmain @gol
1028 -mfixed-range=@var{register-range} @gol
1030 -maddress-space-conversion -mno-address-space-conversion @gol
1031 -mcache-size=@var{cache-size} @gol
1032 -matomic-updates -mno-atomic-updates}
1034 @emph{System V Options}
1035 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1037 @emph{TILE-Gx Options}
1038 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1039 -mcmodel=@var{code-model}}
1041 @emph{TILEPro Options}
1042 @gccoptlist{-mcpu=@var{cpu} -m32}
1045 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1046 -mprolog-function -mno-prolog-function -mspace @gol
1047 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1048 -mapp-regs -mno-app-regs @gol
1049 -mdisable-callt -mno-disable-callt @gol
1050 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1051 -mv850e -mv850 -mv850e3v5 @gol
1062 @gccoptlist{-mg -mgnu -munix}
1064 @emph{Visium Options}
1065 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1066 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1069 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1070 -mpointer-size=@var{size}}
1072 @emph{VxWorks Options}
1073 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1074 -Xbind-lazy -Xbind-now}
1077 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1078 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1079 -mfpmath=@var{unit} @gol
1080 -masm=@var{dialect} -mno-fancy-math-387 @gol
1081 -mno-fp-ret-in-387 -msoft-float @gol
1082 -mno-wide-multiply -mrtd -malign-double @gol
1083 -mpreferred-stack-boundary=@var{num} @gol
1084 -mincoming-stack-boundary=@var{num} @gol
1085 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1086 -mrecip -mrecip=@var{opt} @gol
1087 -mvzeroupper -mprefer-avx128 @gol
1088 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1089 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1090 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1091 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1092 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1093 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1094 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mthreads @gol
1095 -mno-align-stringops -minline-all-stringops @gol
1096 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1097 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1098 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1099 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1100 -mregparm=@var{num} -msseregparm @gol
1101 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1102 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1103 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1104 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1105 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1106 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1107 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1108 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1110 @emph{x86 Windows Options}
1111 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1112 -mnop-fun-dllimport -mthread @gol
1113 -municode -mwin32 -mwindows -fno-set-stack-executable}
1115 @emph{Xstormy16 Options}
1118 @emph{Xtensa Options}
1119 @gccoptlist{-mconst16 -mno-const16 @gol
1120 -mfused-madd -mno-fused-madd @gol
1122 -mserialize-volatile -mno-serialize-volatile @gol
1123 -mtext-section-literals -mno-text-section-literals @gol
1124 -mauto-litpools -mno-auto-litpools @gol
1125 -mtarget-align -mno-target-align @gol
1126 -mlongcalls -mno-longcalls}
1128 @emph{zSeries Options}
1129 See S/390 and zSeries Options.
1131 @item Code Generation Options
1132 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1133 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1134 -ffixed-@var{reg} -fexceptions @gol
1135 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1136 -fasynchronous-unwind-tables @gol
1137 -fno-gnu-unique @gol
1138 -finhibit-size-directive -finstrument-functions @gol
1139 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1140 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1141 -fno-common -fno-ident @gol
1142 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
1143 -fno-jump-tables @gol
1144 -frecord-gcc-switches @gol
1145 -freg-struct-return -fshort-enums @gol
1146 -fshort-double -fshort-wchar @gol
1147 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1148 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1149 -fno-stack-limit -fsplit-stack @gol
1150 -fleading-underscore -ftls-model=@var{model} @gol
1151 -fstack-reuse=@var{reuse_level} @gol
1152 -ftrapv -fwrapv -fbounds-check @gol
1153 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1154 -fstrict-volatile-bitfields -fsync-libcalls}
1158 @node Overall Options
1159 @section Options Controlling the Kind of Output
1161 Compilation can involve up to four stages: preprocessing, compilation
1162 proper, assembly and linking, always in that order. GCC is capable of
1163 preprocessing and compiling several files either into several
1164 assembler input files, or into one assembler input file; then each
1165 assembler input file produces an object file, and linking combines all
1166 the object files (those newly compiled, and those specified as input)
1167 into an executable file.
1169 @cindex file name suffix
1170 For any given input file, the file name suffix determines what kind of
1171 compilation is done:
1175 C source code that must be preprocessed.
1178 C source code that should not be preprocessed.
1181 C++ source code that should not be preprocessed.
1184 Objective-C source code. Note that you must link with the @file{libobjc}
1185 library to make an Objective-C program work.
1188 Objective-C source code that should not be preprocessed.
1192 Objective-C++ source code. Note that you must link with the @file{libobjc}
1193 library to make an Objective-C++ program work. Note that @samp{.M} refers
1194 to a literal capital M@.
1196 @item @var{file}.mii
1197 Objective-C++ source code that should not be preprocessed.
1200 C, C++, Objective-C or Objective-C++ header file to be turned into a
1201 precompiled header (default), or C, C++ header file to be turned into an
1202 Ada spec (via the @option{-fdump-ada-spec} switch).
1205 @itemx @var{file}.cp
1206 @itemx @var{file}.cxx
1207 @itemx @var{file}.cpp
1208 @itemx @var{file}.CPP
1209 @itemx @var{file}.c++
1211 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1212 the last two letters must both be literally @samp{x}. Likewise,
1213 @samp{.C} refers to a literal capital C@.
1217 Objective-C++ source code that must be preprocessed.
1219 @item @var{file}.mii
1220 Objective-C++ source code that should not be preprocessed.
1224 @itemx @var{file}.hp
1225 @itemx @var{file}.hxx
1226 @itemx @var{file}.hpp
1227 @itemx @var{file}.HPP
1228 @itemx @var{file}.h++
1229 @itemx @var{file}.tcc
1230 C++ header file to be turned into a precompiled header or Ada spec.
1233 @itemx @var{file}.for
1234 @itemx @var{file}.ftn
1235 Fixed form Fortran source code that should not be preprocessed.
1238 @itemx @var{file}.FOR
1239 @itemx @var{file}.fpp
1240 @itemx @var{file}.FPP
1241 @itemx @var{file}.FTN
1242 Fixed form Fortran source code that must be preprocessed (with the traditional
1245 @item @var{file}.f90
1246 @itemx @var{file}.f95
1247 @itemx @var{file}.f03
1248 @itemx @var{file}.f08
1249 Free form Fortran source code that should not be preprocessed.
1251 @item @var{file}.F90
1252 @itemx @var{file}.F95
1253 @itemx @var{file}.F03
1254 @itemx @var{file}.F08
1255 Free form Fortran source code that must be preprocessed (with the
1256 traditional preprocessor).
1261 @c FIXME: Descriptions of Java file types.
1267 @item @var{file}.ads
1268 Ada source code file that contains a library unit declaration (a
1269 declaration of a package, subprogram, or generic, or a generic
1270 instantiation), or a library unit renaming declaration (a package,
1271 generic, or subprogram renaming declaration). Such files are also
1274 @item @var{file}.adb
1275 Ada source code file containing a library unit body (a subprogram or
1276 package body). Such files are also called @dfn{bodies}.
1278 @c GCC also knows about some suffixes for languages not yet included:
1289 @itemx @var{file}.sx
1290 Assembler code that must be preprocessed.
1293 An object file to be fed straight into linking.
1294 Any file name with no recognized suffix is treated this way.
1298 You can specify the input language explicitly with the @option{-x} option:
1301 @item -x @var{language}
1302 Specify explicitly the @var{language} for the following input files
1303 (rather than letting the compiler choose a default based on the file
1304 name suffix). This option applies to all following input files until
1305 the next @option{-x} option. Possible values for @var{language} are:
1307 c c-header cpp-output
1308 c++ c++-header c++-cpp-output
1309 objective-c objective-c-header objective-c-cpp-output
1310 objective-c++ objective-c++-header objective-c++-cpp-output
1311 assembler assembler-with-cpp
1313 f77 f77-cpp-input f95 f95-cpp-input
1319 Turn off any specification of a language, so that subsequent files are
1320 handled according to their file name suffixes (as they are if @option{-x}
1321 has not been used at all).
1323 @item -pass-exit-codes
1324 @opindex pass-exit-codes
1325 Normally the @command{gcc} program exits with the code of 1 if any
1326 phase of the compiler returns a non-success return code. If you specify
1327 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1328 the numerically highest error produced by any phase returning an error
1329 indication. The C, C++, and Fortran front ends return 4 if an internal
1330 compiler error is encountered.
1333 If you only want some of the stages of compilation, you can use
1334 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1335 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1336 @command{gcc} is to stop. Note that some combinations (for example,
1337 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1342 Compile or assemble the source files, but do not link. The linking
1343 stage simply is not done. The ultimate output is in the form of an
1344 object file for each source file.
1346 By default, the object file name for a source file is made by replacing
1347 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1349 Unrecognized input files, not requiring compilation or assembly, are
1354 Stop after the stage of compilation proper; do not assemble. The output
1355 is in the form of an assembler code file for each non-assembler input
1358 By default, the assembler file name for a source file is made by
1359 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1361 Input files that don't require compilation are ignored.
1365 Stop after the preprocessing stage; do not run the compiler proper. The
1366 output is in the form of preprocessed source code, which is sent to the
1369 Input files that don't require preprocessing are ignored.
1371 @cindex output file option
1374 Place output in file @var{file}. This applies to whatever
1375 sort of output is being produced, whether it be an executable file,
1376 an object file, an assembler file or preprocessed C code.
1378 If @option{-o} is not specified, the default is to put an executable
1379 file in @file{a.out}, the object file for
1380 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1381 assembler file in @file{@var{source}.s}, a precompiled header file in
1382 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1387 Print (on standard error output) the commands executed to run the stages
1388 of compilation. Also print the version number of the compiler driver
1389 program and of the preprocessor and the compiler proper.
1393 Like @option{-v} except the commands are not executed and arguments
1394 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1395 This is useful for shell scripts to capture the driver-generated command lines.
1399 Use pipes rather than temporary files for communication between the
1400 various stages of compilation. This fails to work on some systems where
1401 the assembler is unable to read from a pipe; but the GNU assembler has
1406 Print (on the standard output) a description of the command-line options
1407 understood by @command{gcc}. If the @option{-v} option is also specified
1408 then @option{--help} is also passed on to the various processes
1409 invoked by @command{gcc}, so that they can display the command-line options
1410 they accept. If the @option{-Wextra} option has also been specified
1411 (prior to the @option{--help} option), then command-line options that
1412 have no documentation associated with them are also displayed.
1415 @opindex target-help
1416 Print (on the standard output) a description of target-specific command-line
1417 options for each tool. For some targets extra target-specific
1418 information may also be printed.
1420 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1421 Print (on the standard output) a description of the command-line
1422 options understood by the compiler that fit into all specified classes
1423 and qualifiers. These are the supported classes:
1426 @item @samp{optimizers}
1427 Display all of the optimization options supported by the
1430 @item @samp{warnings}
1431 Display all of the options controlling warning messages
1432 produced by the compiler.
1435 Display target-specific options. Unlike the
1436 @option{--target-help} option however, target-specific options of the
1437 linker and assembler are not displayed. This is because those
1438 tools do not currently support the extended @option{--help=} syntax.
1441 Display the values recognized by the @option{--param}
1444 @item @var{language}
1445 Display the options supported for @var{language}, where
1446 @var{language} is the name of one of the languages supported in this
1450 Display the options that are common to all languages.
1453 These are the supported qualifiers:
1456 @item @samp{undocumented}
1457 Display only those options that are undocumented.
1460 Display options taking an argument that appears after an equal
1461 sign in the same continuous piece of text, such as:
1462 @samp{--help=target}.
1464 @item @samp{separate}
1465 Display options taking an argument that appears as a separate word
1466 following the original option, such as: @samp{-o output-file}.
1469 Thus for example to display all the undocumented target-specific
1470 switches supported by the compiler, use:
1473 --help=target,undocumented
1476 The sense of a qualifier can be inverted by prefixing it with the
1477 @samp{^} character, so for example to display all binary warning
1478 options (i.e., ones that are either on or off and that do not take an
1479 argument) that have a description, use:
1482 --help=warnings,^joined,^undocumented
1485 The argument to @option{--help=} should not consist solely of inverted
1488 Combining several classes is possible, although this usually
1489 restricts the output so much that there is nothing to display. One
1490 case where it does work, however, is when one of the classes is
1491 @var{target}. For example, to display all the target-specific
1492 optimization options, use:
1495 --help=target,optimizers
1498 The @option{--help=} option can be repeated on the command line. Each
1499 successive use displays its requested class of options, skipping
1500 those that have already been displayed.
1502 If the @option{-Q} option appears on the command line before the
1503 @option{--help=} option, then the descriptive text displayed by
1504 @option{--help=} is changed. Instead of describing the displayed
1505 options, an indication is given as to whether the option is enabled,
1506 disabled or set to a specific value (assuming that the compiler
1507 knows this at the point where the @option{--help=} option is used).
1509 Here is a truncated example from the ARM port of @command{gcc}:
1512 % gcc -Q -mabi=2 --help=target -c
1513 The following options are target specific:
1515 -mabort-on-noreturn [disabled]
1519 The output is sensitive to the effects of previous command-line
1520 options, so for example it is possible to find out which optimizations
1521 are enabled at @option{-O2} by using:
1524 -Q -O2 --help=optimizers
1527 Alternatively you can discover which binary optimizations are enabled
1528 by @option{-O3} by using:
1531 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1532 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1533 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1536 @item -no-canonical-prefixes
1537 @opindex no-canonical-prefixes
1538 Do not expand any symbolic links, resolve references to @samp{/../}
1539 or @samp{/./}, or make the path absolute when generating a relative
1544 Display the version number and copyrights of the invoked GCC@.
1548 Invoke all subcommands under a wrapper program. The name of the
1549 wrapper program and its parameters are passed as a comma separated
1553 gcc -c t.c -wrapper gdb,--args
1557 This invokes all subprograms of @command{gcc} under
1558 @samp{gdb --args}, thus the invocation of @command{cc1} is
1559 @samp{gdb --args cc1 @dots{}}.
1561 @item -fplugin=@var{name}.so
1563 Load the plugin code in file @var{name}.so, assumed to be a
1564 shared object to be dlopen'd by the compiler. The base name of
1565 the shared object file is used to identify the plugin for the
1566 purposes of argument parsing (See
1567 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1568 Each plugin should define the callback functions specified in the
1571 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1572 @opindex fplugin-arg
1573 Define an argument called @var{key} with a value of @var{value}
1574 for the plugin called @var{name}.
1576 @item -fdump-ada-spec@r{[}-slim@r{]}
1577 @opindex fdump-ada-spec
1578 For C and C++ source and include files, generate corresponding Ada specs.
1579 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1580 GNAT User's Guide}, which provides detailed documentation on this feature.
1582 @item -fada-spec-parent=@var{unit}
1583 @opindex fada-spec-parent
1584 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1585 Ada specs as child units of parent @var{unit}.
1587 @item -fdump-go-spec=@var{file}
1588 @opindex fdump-go-spec
1589 For input files in any language, generate corresponding Go
1590 declarations in @var{file}. This generates Go @code{const},
1591 @code{type}, @code{var}, and @code{func} declarations which may be a
1592 useful way to start writing a Go interface to code written in some
1595 @include @value{srcdir}/../libiberty/at-file.texi
1599 @section Compiling C++ Programs
1601 @cindex suffixes for C++ source
1602 @cindex C++ source file suffixes
1603 C++ source files conventionally use one of the suffixes @samp{.C},
1604 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1605 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1606 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1607 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1608 files with these names and compiles them as C++ programs even if you
1609 call the compiler the same way as for compiling C programs (usually
1610 with the name @command{gcc}).
1614 However, the use of @command{gcc} does not add the C++ library.
1615 @command{g++} is a program that calls GCC and automatically specifies linking
1616 against the C++ library. It treats @samp{.c},
1617 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1618 files unless @option{-x} is used. This program is also useful when
1619 precompiling a C header file with a @samp{.h} extension for use in C++
1620 compilations. On many systems, @command{g++} is also installed with
1621 the name @command{c++}.
1623 @cindex invoking @command{g++}
1624 When you compile C++ programs, you may specify many of the same
1625 command-line options that you use for compiling programs in any
1626 language; or command-line options meaningful for C and related
1627 languages; or options that are meaningful only for C++ programs.
1628 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1629 explanations of options for languages related to C@.
1630 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1631 explanations of options that are meaningful only for C++ programs.
1633 @node C Dialect Options
1634 @section Options Controlling C Dialect
1635 @cindex dialect options
1636 @cindex language dialect options
1637 @cindex options, dialect
1639 The following options control the dialect of C (or languages derived
1640 from C, such as C++, Objective-C and Objective-C++) that the compiler
1644 @cindex ANSI support
1648 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1649 equivalent to @option{-std=c++98}.
1651 This turns off certain features of GCC that are incompatible with ISO
1652 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1653 such as the @code{asm} and @code{typeof} keywords, and
1654 predefined macros such as @code{unix} and @code{vax} that identify the
1655 type of system you are using. It also enables the undesirable and
1656 rarely used ISO trigraph feature. For the C compiler,
1657 it disables recognition of C++ style @samp{//} comments as well as
1658 the @code{inline} keyword.
1660 The alternate keywords @code{__asm__}, @code{__extension__},
1661 @code{__inline__} and @code{__typeof__} continue to work despite
1662 @option{-ansi}. You would not want to use them in an ISO C program, of
1663 course, but it is useful to put them in header files that might be included
1664 in compilations done with @option{-ansi}. Alternate predefined macros
1665 such as @code{__unix__} and @code{__vax__} are also available, with or
1666 without @option{-ansi}.
1668 The @option{-ansi} option does not cause non-ISO programs to be
1669 rejected gratuitously. For that, @option{-Wpedantic} is required in
1670 addition to @option{-ansi}. @xref{Warning Options}.
1672 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1673 option is used. Some header files may notice this macro and refrain
1674 from declaring certain functions or defining certain macros that the
1675 ISO standard doesn't call for; this is to avoid interfering with any
1676 programs that might use these names for other things.
1678 Functions that are normally built in but do not have semantics
1679 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1680 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1681 built-in functions provided by GCC}, for details of the functions
1686 Determine the language standard. @xref{Standards,,Language Standards
1687 Supported by GCC}, for details of these standard versions. This option
1688 is currently only supported when compiling C or C++.
1690 The compiler can accept several base standards, such as @samp{c90} or
1691 @samp{c++98}, and GNU dialects of those standards, such as
1692 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1693 compiler accepts all programs following that standard plus those
1694 using GNU extensions that do not contradict it. For example,
1695 @option{-std=c90} turns off certain features of GCC that are
1696 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1697 keywords, but not other GNU extensions that do not have a meaning in
1698 ISO C90, such as omitting the middle term of a @code{?:}
1699 expression. On the other hand, when a GNU dialect of a standard is
1700 specified, all features supported by the compiler are enabled, even when
1701 those features change the meaning of the base standard. As a result, some
1702 strict-conforming programs may be rejected. The particular standard
1703 is used by @option{-Wpedantic} to identify which features are GNU
1704 extensions given that version of the standard. For example
1705 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1706 comments, while @option{-std=gnu99 -Wpedantic} does not.
1708 A value for this option must be provided; possible values are
1714 Support all ISO C90 programs (certain GNU extensions that conflict
1715 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1717 @item iso9899:199409
1718 ISO C90 as modified in amendment 1.
1724 ISO C99. This standard is substantially completely supported, modulo
1725 bugs and floating-point issues
1726 (mainly but not entirely relating to optional C99 features from
1727 Annexes F and G). See
1728 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1729 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1734 ISO C11, the 2011 revision of the ISO C standard. This standard is
1735 substantially completely supported, modulo bugs, floating-point issues
1736 (mainly but not entirely relating to optional C11 features from
1737 Annexes F and G) and the optional Annexes K (Bounds-checking
1738 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1742 GNU dialect of ISO C90 (including some C99 features).
1746 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1750 GNU dialect of ISO C11. This is the default for C code.
1751 The name @samp{gnu1x} is deprecated.
1755 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1756 additional defect reports. Same as @option{-ansi} for C++ code.
1760 GNU dialect of @option{-std=c++98}. This is the default for
1765 The 2011 ISO C++ standard plus amendments.
1766 The name @samp{c++0x} is deprecated.
1770 GNU dialect of @option{-std=c++11}.
1771 The name @samp{gnu++0x} is deprecated.
1775 The 2014 ISO C++ standard plus amendments.
1776 The name @samp{c++1y} is deprecated.
1780 GNU dialect of @option{-std=c++14}.
1781 The name @samp{gnu++1y} is deprecated.
1784 The next revision of the ISO C++ standard, tentatively planned for
1785 2017. Support is highly experimental, and will almost certainly
1786 change in incompatible ways in future releases.
1789 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1790 and will almost certainly change in incompatible ways in future
1794 @item -fgnu89-inline
1795 @opindex fgnu89-inline
1796 The option @option{-fgnu89-inline} tells GCC to use the traditional
1797 GNU semantics for @code{inline} functions when in C99 mode.
1798 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1799 Using this option is roughly equivalent to adding the
1800 @code{gnu_inline} function attribute to all inline functions
1801 (@pxref{Function Attributes}).
1803 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1804 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1805 specifies the default behavior).
1806 This option is not supported in @option{-std=c90} or
1807 @option{-std=gnu90} mode.
1809 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1810 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1811 in effect for @code{inline} functions. @xref{Common Predefined
1812 Macros,,,cpp,The C Preprocessor}.
1814 @item -aux-info @var{filename}
1816 Output to the given filename prototyped declarations for all functions
1817 declared and/or defined in a translation unit, including those in header
1818 files. This option is silently ignored in any language other than C@.
1820 Besides declarations, the file indicates, in comments, the origin of
1821 each declaration (source file and line), whether the declaration was
1822 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1823 @samp{O} for old, respectively, in the first character after the line
1824 number and the colon), and whether it came from a declaration or a
1825 definition (@samp{C} or @samp{F}, respectively, in the following
1826 character). In the case of function definitions, a K&R-style list of
1827 arguments followed by their declarations is also provided, inside
1828 comments, after the declaration.
1830 @item -fallow-parameterless-variadic-functions
1831 @opindex fallow-parameterless-variadic-functions
1832 Accept variadic functions without named parameters.
1834 Although it is possible to define such a function, this is not very
1835 useful as it is not possible to read the arguments. This is only
1836 supported for C as this construct is allowed by C++.
1840 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1841 keyword, so that code can use these words as identifiers. You can use
1842 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1843 instead. @option{-ansi} implies @option{-fno-asm}.
1845 In C++, this switch only affects the @code{typeof} keyword, since
1846 @code{asm} and @code{inline} are standard keywords. You may want to
1847 use the @option{-fno-gnu-keywords} flag instead, which has the same
1848 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1849 switch only affects the @code{asm} and @code{typeof} keywords, since
1850 @code{inline} is a standard keyword in ISO C99.
1853 @itemx -fno-builtin-@var{function}
1854 @opindex fno-builtin
1855 @cindex built-in functions
1856 Don't recognize built-in functions that do not begin with
1857 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1858 functions provided by GCC}, for details of the functions affected,
1859 including those which are not built-in functions when @option{-ansi} or
1860 @option{-std} options for strict ISO C conformance are used because they
1861 do not have an ISO standard meaning.
1863 GCC normally generates special code to handle certain built-in functions
1864 more efficiently; for instance, calls to @code{alloca} may become single
1865 instructions which adjust the stack directly, and calls to @code{memcpy}
1866 may become inline copy loops. The resulting code is often both smaller
1867 and faster, but since the function calls no longer appear as such, you
1868 cannot set a breakpoint on those calls, nor can you change the behavior
1869 of the functions by linking with a different library. In addition,
1870 when a function is recognized as a built-in function, GCC may use
1871 information about that function to warn about problems with calls to
1872 that function, or to generate more efficient code, even if the
1873 resulting code still contains calls to that function. For example,
1874 warnings are given with @option{-Wformat} for bad calls to
1875 @code{printf} when @code{printf} is built in and @code{strlen} is
1876 known not to modify global memory.
1878 With the @option{-fno-builtin-@var{function}} option
1879 only the built-in function @var{function} is
1880 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1881 function is named that is not built-in in this version of GCC, this
1882 option is ignored. There is no corresponding
1883 @option{-fbuiltin-@var{function}} option; if you wish to enable
1884 built-in functions selectively when using @option{-fno-builtin} or
1885 @option{-ffreestanding}, you may define macros such as:
1888 #define abs(n) __builtin_abs ((n))
1889 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1894 @cindex hosted environment
1896 Assert that compilation targets a hosted environment. This implies
1897 @option{-fbuiltin}. A hosted environment is one in which the
1898 entire standard library is available, and in which @code{main} has a return
1899 type of @code{int}. Examples are nearly everything except a kernel.
1900 This is equivalent to @option{-fno-freestanding}.
1902 @item -ffreestanding
1903 @opindex ffreestanding
1904 @cindex hosted environment
1906 Assert that compilation targets a freestanding environment. This
1907 implies @option{-fno-builtin}. A freestanding environment
1908 is one in which the standard library may not exist, and program startup may
1909 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1910 This is equivalent to @option{-fno-hosted}.
1912 @xref{Standards,,Language Standards Supported by GCC}, for details of
1913 freestanding and hosted environments.
1917 @cindex OpenACC accelerator programming
1918 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1919 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1920 compiler generates accelerated code according to the OpenACC Application
1921 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1922 implies @option{-pthread}, and thus is only supported on targets that
1923 have support for @option{-pthread}.
1925 Note that this is an experimental feature, incomplete, and subject to
1926 change in future versions of GCC. See
1927 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1931 @cindex OpenMP parallel
1932 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1933 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1934 compiler generates parallel code according to the OpenMP Application
1935 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1936 implies @option{-pthread}, and thus is only supported on targets that
1937 have support for @option{-pthread}. @option{-fopenmp} implies
1938 @option{-fopenmp-simd}.
1941 @opindex fopenmp-simd
1944 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1945 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1950 @cindex Enable Cilk Plus
1951 Enable the usage of Cilk Plus language extension features for C/C++.
1952 When the option @option{-fcilkplus} is specified, enable the usage of
1953 the Cilk Plus Language extension features for C/C++. The present
1954 implementation follows ABI version 1.2. This is an experimental
1955 feature that is only partially complete, and whose interface may
1956 change in future versions of GCC as the official specification
1957 changes. Currently, all features but @code{_Cilk_for} have been
1962 When the option @option{-fgnu-tm} is specified, the compiler
1963 generates code for the Linux variant of Intel's current Transactional
1964 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1965 an experimental feature whose interface may change in future versions
1966 of GCC, as the official specification changes. Please note that not
1967 all architectures are supported for this feature.
1969 For more information on GCC's support for transactional memory,
1970 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1971 Transactional Memory Library}.
1973 Note that the transactional memory feature is not supported with
1974 non-call exceptions (@option{-fnon-call-exceptions}).
1976 @item -fms-extensions
1977 @opindex fms-extensions
1978 Accept some non-standard constructs used in Microsoft header files.
1980 In C++ code, this allows member names in structures to be similar
1981 to previous types declarations.
1990 Some cases of unnamed fields in structures and unions are only
1991 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1992 fields within structs/unions}, for details.
1994 Note that this option is off for all targets but x86
1995 targets using ms-abi.
1997 @item -fplan9-extensions
1998 @opindex fplan9-extensions
1999 Accept some non-standard constructs used in Plan 9 code.
2001 This enables @option{-fms-extensions}, permits passing pointers to
2002 structures with anonymous fields to functions that expect pointers to
2003 elements of the type of the field, and permits referring to anonymous
2004 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2005 struct/union fields within structs/unions}, for details. This is only
2006 supported for C, not C++.
2010 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2011 options for strict ISO C conformance) implies @option{-trigraphs}.
2013 @cindex traditional C language
2014 @cindex C language, traditional
2016 @itemx -traditional-cpp
2017 @opindex traditional-cpp
2018 @opindex traditional
2019 Formerly, these options caused GCC to attempt to emulate a pre-standard
2020 C compiler. They are now only supported with the @option{-E} switch.
2021 The preprocessor continues to support a pre-standard mode. See the GNU
2022 CPP manual for details.
2024 @item -fcond-mismatch
2025 @opindex fcond-mismatch
2026 Allow conditional expressions with mismatched types in the second and
2027 third arguments. The value of such an expression is void. This option
2028 is not supported for C++.
2030 @item -flax-vector-conversions
2031 @opindex flax-vector-conversions
2032 Allow implicit conversions between vectors with differing numbers of
2033 elements and/or incompatible element types. This option should not be
2036 @item -funsigned-char
2037 @opindex funsigned-char
2038 Let the type @code{char} be unsigned, like @code{unsigned char}.
2040 Each kind of machine has a default for what @code{char} should
2041 be. It is either like @code{unsigned char} by default or like
2042 @code{signed char} by default.
2044 Ideally, a portable program should always use @code{signed char} or
2045 @code{unsigned char} when it depends on the signedness of an object.
2046 But many programs have been written to use plain @code{char} and
2047 expect it to be signed, or expect it to be unsigned, depending on the
2048 machines they were written for. This option, and its inverse, let you
2049 make such a program work with the opposite default.
2051 The type @code{char} is always a distinct type from each of
2052 @code{signed char} or @code{unsigned char}, even though its behavior
2053 is always just like one of those two.
2056 @opindex fsigned-char
2057 Let the type @code{char} be signed, like @code{signed char}.
2059 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2060 the negative form of @option{-funsigned-char}. Likewise, the option
2061 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2063 @item -fsigned-bitfields
2064 @itemx -funsigned-bitfields
2065 @itemx -fno-signed-bitfields
2066 @itemx -fno-unsigned-bitfields
2067 @opindex fsigned-bitfields
2068 @opindex funsigned-bitfields
2069 @opindex fno-signed-bitfields
2070 @opindex fno-unsigned-bitfields
2071 These options control whether a bit-field is signed or unsigned, when the
2072 declaration does not use either @code{signed} or @code{unsigned}. By
2073 default, such a bit-field is signed, because this is consistent: the
2074 basic integer types such as @code{int} are signed types.
2077 @node C++ Dialect Options
2078 @section Options Controlling C++ Dialect
2080 @cindex compiler options, C++
2081 @cindex C++ options, command-line
2082 @cindex options, C++
2083 This section describes the command-line options that are only meaningful
2084 for C++ programs. You can also use most of the GNU compiler options
2085 regardless of what language your program is in. For example, you
2086 might compile a file @file{firstClass.C} like this:
2089 g++ -g -fstrict-enums -O -c firstClass.C
2093 In this example, only @option{-fstrict-enums} is an option meant
2094 only for C++ programs; you can use the other options with any
2095 language supported by GCC@.
2097 Here is a list of options that are @emph{only} for compiling C++ programs:
2101 @item -fabi-version=@var{n}
2102 @opindex fabi-version
2103 Use version @var{n} of the C++ ABI@. The default is version 0.
2105 Version 0 refers to the version conforming most closely to
2106 the C++ ABI specification. Therefore, the ABI obtained using version 0
2107 will change in different versions of G++ as ABI bugs are fixed.
2109 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2111 Version 2 is the version of the C++ ABI that first appeared in G++
2112 3.4, and was the default through G++ 4.9.
2114 Version 3 corrects an error in mangling a constant address as a
2117 Version 4, which first appeared in G++ 4.5, implements a standard
2118 mangling for vector types.
2120 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2121 attribute const/volatile on function pointer types, decltype of a
2122 plain decl, and use of a function parameter in the declaration of
2125 Version 6, which first appeared in G++ 4.7, corrects the promotion
2126 behavior of C++11 scoped enums and the mangling of template argument
2127 packs, const/static_cast, prefix ++ and --, and a class scope function
2128 used as a template argument.
2130 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2131 builtin type and corrects the mangling of lambdas in default argument
2134 Version 8, which first appeared in G++ 4.9, corrects the substitution
2135 behavior of function types with function-cv-qualifiers.
2137 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2140 Version 10, which first appeared in G++ 6.1, adds mangling of
2141 attributes that affect type identity, such as ia32 calling convention
2142 attributes (e.g. @samp{stdcall}).
2144 See also @option{-Wabi}.
2146 @item -fabi-compat-version=@var{n}
2147 @opindex fabi-compat-version
2148 On targets that support strong aliases, G++
2149 works around mangling changes by creating an alias with the correct
2150 mangled name when defining a symbol with an incorrect mangled name.
2151 This switch specifies which ABI version to use for the alias.
2153 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2154 compatibility). If another ABI version is explicitly selected, this
2155 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2156 use @option{-fabi-compat-version=2}.
2158 If this option is not provided but @option{-Wabi=@var{n}} is, that
2159 version is used for compatibility aliases. If this option is provided
2160 along with @option{-Wabi} (without the version), the version from this
2161 option is used for the warning.
2163 @item -fno-access-control
2164 @opindex fno-access-control
2165 Turn off all access checking. This switch is mainly useful for working
2166 around bugs in the access control code.
2170 Check that the pointer returned by @code{operator new} is non-null
2171 before attempting to modify the storage allocated. This check is
2172 normally unnecessary because the C++ standard specifies that
2173 @code{operator new} only returns @code{0} if it is declared
2174 @code{throw()}, in which case the compiler always checks the
2175 return value even without this option. In all other cases, when
2176 @code{operator new} has a non-empty exception specification, memory
2177 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2178 @samp{new (nothrow)}.
2180 @item -fconstexpr-depth=@var{n}
2181 @opindex fconstexpr-depth
2182 Set the maximum nested evaluation depth for C++11 constexpr functions
2183 to @var{n}. A limit is needed to detect endless recursion during
2184 constant expression evaluation. The minimum specified by the standard
2187 @item -fdeduce-init-list
2188 @opindex fdeduce-init-list
2189 Enable deduction of a template type parameter as
2190 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2193 template <class T> auto forward(T t) -> decltype (realfn (t))
2200 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2204 This deduction was implemented as a possible extension to the
2205 originally proposed semantics for the C++11 standard, but was not part
2206 of the final standard, so it is disabled by default. This option is
2207 deprecated, and may be removed in a future version of G++.
2209 @item -ffriend-injection
2210 @opindex ffriend-injection
2211 Inject friend functions into the enclosing namespace, so that they are
2212 visible outside the scope of the class in which they are declared.
2213 Friend functions were documented to work this way in the old Annotated
2214 C++ Reference Manual.
2215 However, in ISO C++ a friend function that is not declared
2216 in an enclosing scope can only be found using argument dependent
2217 lookup. GCC defaults to the standard behavior.
2219 This option is for compatibility, and may be removed in a future
2222 @item -fno-elide-constructors
2223 @opindex fno-elide-constructors
2224 The C++ standard allows an implementation to omit creating a temporary
2225 that is only used to initialize another object of the same type.
2226 Specifying this option disables that optimization, and forces G++ to
2227 call the copy constructor in all cases.
2229 @item -fno-enforce-eh-specs
2230 @opindex fno-enforce-eh-specs
2231 Don't generate code to check for violation of exception specifications
2232 at run time. This option violates the C++ standard, but may be useful
2233 for reducing code size in production builds, much like defining
2234 @code{NDEBUG}. This does not give user code permission to throw
2235 exceptions in violation of the exception specifications; the compiler
2236 still optimizes based on the specifications, so throwing an
2237 unexpected exception results in undefined behavior at run time.
2239 @item -fextern-tls-init
2240 @itemx -fno-extern-tls-init
2241 @opindex fextern-tls-init
2242 @opindex fno-extern-tls-init
2243 The C++11 and OpenMP standards allow @code{thread_local} and
2244 @code{threadprivate} variables to have dynamic (runtime)
2245 initialization. To support this, any use of such a variable goes
2246 through a wrapper function that performs any necessary initialization.
2247 When the use and definition of the variable are in the same
2248 translation unit, this overhead can be optimized away, but when the
2249 use is in a different translation unit there is significant overhead
2250 even if the variable doesn't actually need dynamic initialization. If
2251 the programmer can be sure that no use of the variable in a
2252 non-defining TU needs to trigger dynamic initialization (either
2253 because the variable is statically initialized, or a use of the
2254 variable in the defining TU will be executed before any uses in
2255 another TU), they can avoid this overhead with the
2256 @option{-fno-extern-tls-init} option.
2258 On targets that support symbol aliases, the default is
2259 @option{-fextern-tls-init}. On targets that do not support symbol
2260 aliases, the default is @option{-fno-extern-tls-init}.
2263 @itemx -fno-for-scope
2265 @opindex fno-for-scope
2266 If @option{-ffor-scope} is specified, the scope of variables declared in
2267 a @i{for-init-statement} is limited to the @code{for} loop itself,
2268 as specified by the C++ standard.
2269 If @option{-fno-for-scope} is specified, the scope of variables declared in
2270 a @i{for-init-statement} extends to the end of the enclosing scope,
2271 as was the case in old versions of G++, and other (traditional)
2272 implementations of C++.
2274 If neither flag is given, the default is to follow the standard,
2275 but to allow and give a warning for old-style code that would
2276 otherwise be invalid, or have different behavior.
2278 @item -fno-gnu-keywords
2279 @opindex fno-gnu-keywords
2280 Do not recognize @code{typeof} as a keyword, so that code can use this
2281 word as an identifier. You can use the keyword @code{__typeof__} instead.
2282 @option{-ansi} implies @option{-fno-gnu-keywords}.
2284 @item -fno-implicit-templates
2285 @opindex fno-implicit-templates
2286 Never emit code for non-inline templates that are instantiated
2287 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2288 @xref{Template Instantiation}, for more information.
2290 @item -fno-implicit-inline-templates
2291 @opindex fno-implicit-inline-templates
2292 Don't emit code for implicit instantiations of inline templates, either.
2293 The default is to handle inlines differently so that compiles with and
2294 without optimization need the same set of explicit instantiations.
2296 @item -fno-implement-inlines
2297 @opindex fno-implement-inlines
2298 To save space, do not emit out-of-line copies of inline functions
2299 controlled by @code{#pragma implementation}. This causes linker
2300 errors if these functions are not inlined everywhere they are called.
2302 @item -fms-extensions
2303 @opindex fms-extensions
2304 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2305 int and getting a pointer to member function via non-standard syntax.
2307 @item -fno-nonansi-builtins
2308 @opindex fno-nonansi-builtins
2309 Disable built-in declarations of functions that are not mandated by
2310 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2311 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2314 @opindex fnothrow-opt
2315 Treat a @code{throw()} exception specification as if it were a
2316 @code{noexcept} specification to reduce or eliminate the text size
2317 overhead relative to a function with no exception specification. If
2318 the function has local variables of types with non-trivial
2319 destructors, the exception specification actually makes the
2320 function smaller because the EH cleanups for those variables can be
2321 optimized away. The semantic effect is that an exception thrown out of
2322 a function with such an exception specification results in a call
2323 to @code{terminate} rather than @code{unexpected}.
2325 @item -fno-operator-names
2326 @opindex fno-operator-names
2327 Do not treat the operator name keywords @code{and}, @code{bitand},
2328 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2329 synonyms as keywords.
2331 @item -fno-optional-diags
2332 @opindex fno-optional-diags
2333 Disable diagnostics that the standard says a compiler does not need to
2334 issue. Currently, the only such diagnostic issued by G++ is the one for
2335 a name having multiple meanings within a class.
2338 @opindex fpermissive
2339 Downgrade some diagnostics about nonconformant code from errors to
2340 warnings. Thus, using @option{-fpermissive} allows some
2341 nonconforming code to compile.
2343 @item -fno-pretty-templates
2344 @opindex fno-pretty-templates
2345 When an error message refers to a specialization of a function
2346 template, the compiler normally prints the signature of the
2347 template followed by the template arguments and any typedefs or
2348 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2349 rather than @code{void f(int)}) so that it's clear which template is
2350 involved. When an error message refers to a specialization of a class
2351 template, the compiler omits any template arguments that match
2352 the default template arguments for that template. If either of these
2353 behaviors make it harder to understand the error message rather than
2354 easier, you can use @option{-fno-pretty-templates} to disable them.
2358 Enable automatic template instantiation at link time. This option also
2359 implies @option{-fno-implicit-templates}. @xref{Template
2360 Instantiation}, for more information.
2364 Disable generation of information about every class with virtual
2365 functions for use by the C++ run-time type identification features
2366 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2367 of the language, you can save some space by using this flag. Note that
2368 exception handling uses the same information, but G++ generates it as
2369 needed. The @code{dynamic_cast} operator can still be used for casts that
2370 do not require run-time type information, i.e.@: casts to @code{void *} or to
2371 unambiguous base classes.
2373 @item -fsized-deallocation
2374 @opindex fsized-deallocation
2375 Enable the built-in global declarations
2377 void operator delete (void *, std::size_t) noexcept;
2378 void operator delete[] (void *, std::size_t) noexcept;
2380 as introduced in C++14. This is useful for user-defined replacement
2381 deallocation functions that, for example, use the size of the object
2382 to make deallocation faster. Enabled by default under
2383 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2384 warns about places that might want to add a definition.
2388 Emit statistics about front-end processing at the end of the compilation.
2389 This information is generally only useful to the G++ development team.
2391 @item -fstrict-enums
2392 @opindex fstrict-enums
2393 Allow the compiler to optimize using the assumption that a value of
2394 enumerated type can only be one of the values of the enumeration (as
2395 defined in the C++ standard; basically, a value that can be
2396 represented in the minimum number of bits needed to represent all the
2397 enumerators). This assumption may not be valid if the program uses a
2398 cast to convert an arbitrary integer value to the enumerated type.
2400 @item -ftemplate-backtrace-limit=@var{n}
2401 @opindex ftemplate-backtrace-limit
2402 Set the maximum number of template instantiation notes for a single
2403 warning or error to @var{n}. The default value is 10.
2405 @item -ftemplate-depth=@var{n}
2406 @opindex ftemplate-depth
2407 Set the maximum instantiation depth for template classes to @var{n}.
2408 A limit on the template instantiation depth is needed to detect
2409 endless recursions during template class instantiation. ANSI/ISO C++
2410 conforming programs must not rely on a maximum depth greater than 17
2411 (changed to 1024 in C++11). The default value is 900, as the compiler
2412 can run out of stack space before hitting 1024 in some situations.
2414 @item -fno-threadsafe-statics
2415 @opindex fno-threadsafe-statics
2416 Do not emit the extra code to use the routines specified in the C++
2417 ABI for thread-safe initialization of local statics. You can use this
2418 option to reduce code size slightly in code that doesn't need to be
2421 @item -fuse-cxa-atexit
2422 @opindex fuse-cxa-atexit
2423 Register destructors for objects with static storage duration with the
2424 @code{__cxa_atexit} function rather than the @code{atexit} function.
2425 This option is required for fully standards-compliant handling of static
2426 destructors, but only works if your C library supports
2427 @code{__cxa_atexit}.
2429 @item -fno-use-cxa-get-exception-ptr
2430 @opindex fno-use-cxa-get-exception-ptr
2431 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2432 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2433 if the runtime routine is not available.
2435 @item -fvisibility-inlines-hidden
2436 @opindex fvisibility-inlines-hidden
2437 This switch declares that the user does not attempt to compare
2438 pointers to inline functions or methods where the addresses of the two functions
2439 are taken in different shared objects.
2441 The effect of this is that GCC may, effectively, mark inline methods with
2442 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2443 appear in the export table of a DSO and do not require a PLT indirection
2444 when used within the DSO@. Enabling this option can have a dramatic effect
2445 on load and link times of a DSO as it massively reduces the size of the
2446 dynamic export table when the library makes heavy use of templates.
2448 The behavior of this switch is not quite the same as marking the
2449 methods as hidden directly, because it does not affect static variables
2450 local to the function or cause the compiler to deduce that
2451 the function is defined in only one shared object.
2453 You may mark a method as having a visibility explicitly to negate the
2454 effect of the switch for that method. For example, if you do want to
2455 compare pointers to a particular inline method, you might mark it as
2456 having default visibility. Marking the enclosing class with explicit
2457 visibility has no effect.
2459 Explicitly instantiated inline methods are unaffected by this option
2460 as their linkage might otherwise cross a shared library boundary.
2461 @xref{Template Instantiation}.
2463 @item -fvisibility-ms-compat
2464 @opindex fvisibility-ms-compat
2465 This flag attempts to use visibility settings to make GCC's C++
2466 linkage model compatible with that of Microsoft Visual Studio.
2468 The flag makes these changes to GCC's linkage model:
2472 It sets the default visibility to @code{hidden}, like
2473 @option{-fvisibility=hidden}.
2476 Types, but not their members, are not hidden by default.
2479 The One Definition Rule is relaxed for types without explicit
2480 visibility specifications that are defined in more than one
2481 shared object: those declarations are permitted if they are
2482 permitted when this option is not used.
2485 In new code it is better to use @option{-fvisibility=hidden} and
2486 export those classes that are intended to be externally visible.
2487 Unfortunately it is possible for code to rely, perhaps accidentally,
2488 on the Visual Studio behavior.
2490 Among the consequences of these changes are that static data members
2491 of the same type with the same name but defined in different shared
2492 objects are different, so changing one does not change the other;
2493 and that pointers to function members defined in different shared
2494 objects may not compare equal. When this flag is given, it is a
2495 violation of the ODR to define types with the same name differently.
2497 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2498 @opindex fvtable-verify
2499 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2500 feature that verifies at run time, for every virtual call, that
2501 the vtable pointer through which the call is made is valid for the type of
2502 the object, and has not been corrupted or overwritten. If an invalid vtable
2503 pointer is detected at run time, an error is reported and execution of the
2504 program is immediately halted.
2506 This option causes run-time data structures to be built at program startup,
2507 which are used for verifying the vtable pointers.
2508 The options @samp{std} and @samp{preinit}
2509 control the timing of when these data structures are built. In both cases the
2510 data structures are built before execution reaches @code{main}. Using
2511 @option{-fvtable-verify=std} causes the data structures to be built after
2512 shared libraries have been loaded and initialized.
2513 @option{-fvtable-verify=preinit} causes them to be built before shared
2514 libraries have been loaded and initialized.
2516 If this option appears multiple times in the command line with different
2517 values specified, @samp{none} takes highest priority over both @samp{std} and
2518 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2522 When used in conjunction with @option{-fvtable-verify=std} or
2523 @option{-fvtable-verify=preinit}, causes debug versions of the
2524 runtime functions for the vtable verification feature to be called.
2525 This flag also causes the compiler to log information about which
2526 vtable pointers it finds for each class.
2527 This information is written to a file named @file{vtv_set_ptr_data.log}
2528 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2529 if that is defined or the current working directory otherwise.
2531 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2532 file, be sure to delete any existing one.
2535 @opindex fvtv-counts
2536 This is a debugging flag. When used in conjunction with
2537 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2538 causes the compiler to keep track of the total number of virtual calls
2539 it encounters and the number of verifications it inserts. It also
2540 counts the number of calls to certain run-time library functions
2541 that it inserts and logs this information for each compilation unit.
2542 The compiler writes this information to a file named
2543 @file{vtv_count_data.log} in the directory named by the environment
2544 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2545 directory otherwise. It also counts the size of the vtable pointer sets
2546 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2547 in the same directory.
2549 Note: This feature @emph{appends} data to the log files. To get fresh log
2550 files, be sure to delete any existing ones.
2554 Do not use weak symbol support, even if it is provided by the linker.
2555 By default, G++ uses weak symbols if they are available. This
2556 option exists only for testing, and should not be used by end-users;
2557 it results in inferior code and has no benefits. This option may
2558 be removed in a future release of G++.
2562 Do not search for header files in the standard directories specific to
2563 C++, but do still search the other standard directories. (This option
2564 is used when building the C++ library.)
2567 In addition, these optimization, warning, and code generation options
2568 have meanings only for C++ programs:
2571 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2574 Warn when G++ it generates code that is probably not compatible with
2575 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2576 ABI with each major release, normally @option{-Wabi} will warn only if
2577 there is a check added later in a release series for an ABI issue
2578 discovered since the initial release. @option{-Wabi} will warn about
2579 more things if an older ABI version is selected (with
2580 @option{-fabi-version=@var{n}}).
2582 @option{-Wabi} can also be used with an explicit version number to
2583 warn about compatibility with a particular @option{-fabi-version}
2584 level, e.g. @option{-Wabi=2} to warn about changes relative to
2585 @option{-fabi-version=2}.
2587 If an explicit version number is provided and
2588 @option{-fabi-compat-version} is not specified, the version number
2589 from this option is used for compatibility aliases. If no explicit
2590 version number is provided with this option, but
2591 @option{-fabi-compat-version} is specified, that version number is
2592 used for ABI warnings.
2594 Although an effort has been made to warn about
2595 all such cases, there are probably some cases that are not warned about,
2596 even though G++ is generating incompatible code. There may also be
2597 cases where warnings are emitted even though the code that is generated
2600 You should rewrite your code to avoid these warnings if you are
2601 concerned about the fact that code generated by G++ may not be binary
2602 compatible with code generated by other compilers.
2604 Known incompatibilities in @option{-fabi-version=2} (which was the
2605 default from GCC 3.4 to 4.9) include:
2610 A template with a non-type template parameter of reference type was
2611 mangled incorrectly:
2614 template <int &> struct S @{@};
2618 This was fixed in @option{-fabi-version=3}.
2621 SIMD vector types declared using @code{__attribute ((vector_size))} were
2622 mangled in a non-standard way that does not allow for overloading of
2623 functions taking vectors of different sizes.
2625 The mangling was changed in @option{-fabi-version=4}.
2628 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2629 qualifiers, and @code{decltype} of a plain declaration was folded away.
2631 These mangling issues were fixed in @option{-fabi-version=5}.
2634 Scoped enumerators passed as arguments to a variadic function are
2635 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2636 On most targets this does not actually affect the parameter passing
2637 ABI, as there is no way to pass an argument smaller than @code{int}.
2639 Also, the ABI changed the mangling of template argument packs,
2640 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2641 a class scope function used as a template argument.
2643 These issues were corrected in @option{-fabi-version=6}.
2646 Lambdas in default argument scope were mangled incorrectly, and the
2647 ABI changed the mangling of @code{nullptr_t}.
2649 These issues were corrected in @option{-fabi-version=7}.
2652 When mangling a function type with function-cv-qualifiers, the
2653 un-qualified function type was incorrectly treated as a substitution
2656 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2659 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2660 unaligned accesses. Note that this did not affect the ABI of a
2661 function with a @code{nullptr_t} parameter, as parameters have a
2664 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2667 Target-specific attributes that affect the identity of a type, such as
2668 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2669 did not affect the mangled name, leading to name collisions when
2670 function pointers were used as template arguments.
2672 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2676 It also warns about psABI-related changes. The known psABI changes at this
2682 For SysV/x86-64, unions with @code{long double} members are
2683 passed in memory as specified in psABI. For example:
2693 @code{union U} is always passed in memory.
2697 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2700 Warn when a type with an ABI tag is used in a context that does not
2701 have that ABI tag. See @ref{C++ Attributes} for more information
2704 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2705 @opindex Wctor-dtor-privacy
2706 @opindex Wno-ctor-dtor-privacy
2707 Warn when a class seems unusable because all the constructors or
2708 destructors in that class are private, and it has neither friends nor
2709 public static member functions. Also warn if there are no non-private
2710 methods, and there's at least one private member function that isn't
2711 a constructor or destructor.
2713 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2714 @opindex Wdelete-non-virtual-dtor
2715 @opindex Wno-delete-non-virtual-dtor
2716 Warn when @code{delete} is used to destroy an instance of a class that
2717 has virtual functions and non-virtual destructor. It is unsafe to delete
2718 an instance of a derived class through a pointer to a base class if the
2719 base class does not have a virtual destructor. This warning is enabled
2722 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2723 @opindex Wliteral-suffix
2724 @opindex Wno-literal-suffix
2725 Warn when a string or character literal is followed by a ud-suffix which does
2726 not begin with an underscore. As a conforming extension, GCC treats such
2727 suffixes as separate preprocessing tokens in order to maintain backwards
2728 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2732 #define __STDC_FORMAT_MACROS
2733 #include <inttypes.h>
2738 printf("My int64: %" PRId64"\n", i64);
2742 In this case, @code{PRId64} is treated as a separate preprocessing token.
2744 This warning is enabled by default.
2746 @item -Wlto-type-mismatch
2747 @opindex Wlto-type-mismatch
2748 @opindex Wno-lto-type-mistmach
2750 During the link-time optimization warn about type mismatches in between
2751 global declarations from different compilation units.
2752 Requires @option{-flto} to be enabled. Enabled by default.
2754 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2756 @opindex Wno-narrowing
2757 Warn when a narrowing conversion prohibited by C++11 occurs within
2761 int i = @{ 2.2 @}; // error: narrowing from double to int
2764 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2766 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2767 required by the standard. Note that this does not affect the meaning
2768 of well-formed code; narrowing conversions are still considered
2769 ill-formed in SFINAE context.
2771 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2773 @opindex Wno-noexcept
2774 Warn when a noexcept-expression evaluates to false because of a call
2775 to a function that does not have a non-throwing exception
2776 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2777 the compiler to never throw an exception.
2779 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2780 @opindex Wnon-virtual-dtor
2781 @opindex Wno-non-virtual-dtor
2782 Warn when a class has virtual functions and an accessible non-virtual
2783 destructor itself or in an accessible polymorphic base class, in which
2784 case it is possible but unsafe to delete an instance of a derived
2785 class through a pointer to the class itself or base class. This
2786 warning is automatically enabled if @option{-Weffc++} is specified.
2788 @item -Wreorder @r{(C++ and Objective-C++ only)}
2790 @opindex Wno-reorder
2791 @cindex reordering, warning
2792 @cindex warning for reordering of member initializers
2793 Warn when the order of member initializers given in the code does not
2794 match the order in which they must be executed. For instance:
2800 A(): j (0), i (1) @{ @}
2805 The compiler rearranges the member initializers for @code{i}
2806 and @code{j} to match the declaration order of the members, emitting
2807 a warning to that effect. This warning is enabled by @option{-Wall}.
2809 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2810 @opindex fext-numeric-literals
2811 @opindex fno-ext-numeric-literals
2812 Accept imaginary, fixed-point, or machine-defined
2813 literal number suffixes as GNU extensions.
2814 When this option is turned off these suffixes are treated
2815 as C++11 user-defined literal numeric suffixes.
2816 This is on by default for all pre-C++11 dialects and all GNU dialects:
2817 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2818 @option{-std=gnu++14}.
2819 This option is off by default
2820 for ISO C++11 onwards (@option{-std=c++11}, ...).
2823 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2826 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2829 Warn about violations of the following style guidelines from Scott Meyers'
2830 @cite{Effective C++} series of books:
2834 Define a copy constructor and an assignment operator for classes
2835 with dynamically-allocated memory.
2838 Prefer initialization to assignment in constructors.
2841 Have @code{operator=} return a reference to @code{*this}.
2844 Don't try to return a reference when you must return an object.
2847 Distinguish between prefix and postfix forms of increment and
2848 decrement operators.
2851 Never overload @code{&&}, @code{||}, or @code{,}.
2855 This option also enables @option{-Wnon-virtual-dtor}, which is also
2856 one of the effective C++ recommendations. However, the check is
2857 extended to warn about the lack of virtual destructor in accessible
2858 non-polymorphic bases classes too.
2860 When selecting this option, be aware that the standard library
2861 headers do not obey all of these guidelines; use @samp{grep -v}
2862 to filter out those warnings.
2864 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2865 @opindex Wstrict-null-sentinel
2866 @opindex Wno-strict-null-sentinel
2867 Warn about the use of an uncasted @code{NULL} as sentinel. When
2868 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2869 to @code{__null}. Although it is a null pointer constant rather than a
2870 null pointer, it is guaranteed to be of the same size as a pointer.
2871 But this use is not portable across different compilers.
2873 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2874 @opindex Wno-non-template-friend
2875 @opindex Wnon-template-friend
2876 Disable warnings when non-templatized friend functions are declared
2877 within a template. Since the advent of explicit template specification
2878 support in G++, if the name of the friend is an unqualified-id (i.e.,
2879 @samp{friend foo(int)}), the C++ language specification demands that the
2880 friend declare or define an ordinary, nontemplate function. (Section
2881 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2882 could be interpreted as a particular specialization of a templatized
2883 function. Because this non-conforming behavior is no longer the default
2884 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2885 check existing code for potential trouble spots and is on by default.
2886 This new compiler behavior can be turned off with
2887 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2888 but disables the helpful warning.
2890 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2891 @opindex Wold-style-cast
2892 @opindex Wno-old-style-cast
2893 Warn if an old-style (C-style) cast to a non-void type is used within
2894 a C++ program. The new-style casts (@code{dynamic_cast},
2895 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2896 less vulnerable to unintended effects and much easier to search for.
2898 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2899 @opindex Woverloaded-virtual
2900 @opindex Wno-overloaded-virtual
2901 @cindex overloaded virtual function, warning
2902 @cindex warning for overloaded virtual function
2903 Warn when a function declaration hides virtual functions from a
2904 base class. For example, in:
2911 struct B: public A @{
2916 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2927 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2928 @opindex Wno-pmf-conversions
2929 @opindex Wpmf-conversions
2930 Disable the diagnostic for converting a bound pointer to member function
2933 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2934 @opindex Wsign-promo
2935 @opindex Wno-sign-promo
2936 Warn when overload resolution chooses a promotion from unsigned or
2937 enumerated type to a signed type, over a conversion to an unsigned type of
2938 the same size. Previous versions of G++ tried to preserve
2939 unsignedness, but the standard mandates the current behavior.
2941 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2943 Warn when a primary template declaration is encountered. Some coding
2944 rules disallow templates, and this may be used to enforce that rule.
2945 The warning is inactive inside a system header file, such as the STL, so
2946 one can still use the STL. One may also instantiate or specialize
2949 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2950 @opindex Wmultiple-inheritance
2951 Warn when a class is defined with multiple direct base classes. Some
2952 coding rules disallow multiple inheritance, and this may be used to
2953 enforce that rule. The warning is inactive inside a system header file,
2954 such as the STL, so one can still use the STL. One may also define
2955 classes that indirectly use multiple inheritance.
2957 @item -Wvirtual-inheritance
2958 @opindex Wvirtual-inheritance
2959 Warn when a class is defined with a virtual direct base classe. Some
2960 coding rules disallow multiple inheritance, and this may be used to
2961 enforce that rule. The warning is inactive inside a system header file,
2962 such as the STL, so one can still use the STL. One may also define
2963 classes that indirectly use virtual inheritance.
2966 @opindex Wnamespaces
2967 Warn when a namespace definition is opened. Some coding rules disallow
2968 namespaces, and this may be used to enforce that rule. The warning is
2969 inactive inside a system header file, such as the STL, so one can still
2970 use the STL. One may also use using directives and qualified names.
2972 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2974 @opindex Wno-terminate
2975 Disable the warning about a throw-expression that will immediately
2976 result in a call to @code{terminate}.
2979 @node Objective-C and Objective-C++ Dialect Options
2980 @section Options Controlling Objective-C and Objective-C++ Dialects
2982 @cindex compiler options, Objective-C and Objective-C++
2983 @cindex Objective-C and Objective-C++ options, command-line
2984 @cindex options, Objective-C and Objective-C++
2985 (NOTE: This manual does not describe the Objective-C and Objective-C++
2986 languages themselves. @xref{Standards,,Language Standards
2987 Supported by GCC}, for references.)
2989 This section describes the command-line options that are only meaningful
2990 for Objective-C and Objective-C++ programs. You can also use most of
2991 the language-independent GNU compiler options.
2992 For example, you might compile a file @file{some_class.m} like this:
2995 gcc -g -fgnu-runtime -O -c some_class.m
2999 In this example, @option{-fgnu-runtime} is an option meant only for
3000 Objective-C and Objective-C++ programs; you can use the other options with
3001 any language supported by GCC@.
3003 Note that since Objective-C is an extension of the C language, Objective-C
3004 compilations may also use options specific to the C front-end (e.g.,
3005 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3006 C++-specific options (e.g., @option{-Wabi}).
3008 Here is a list of options that are @emph{only} for compiling Objective-C
3009 and Objective-C++ programs:
3012 @item -fconstant-string-class=@var{class-name}
3013 @opindex fconstant-string-class
3014 Use @var{class-name} as the name of the class to instantiate for each
3015 literal string specified with the syntax @code{@@"@dots{}"}. The default
3016 class name is @code{NXConstantString} if the GNU runtime is being used, and
3017 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3018 @option{-fconstant-cfstrings} option, if also present, overrides the
3019 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3020 to be laid out as constant CoreFoundation strings.
3023 @opindex fgnu-runtime
3024 Generate object code compatible with the standard GNU Objective-C
3025 runtime. This is the default for most types of systems.
3027 @item -fnext-runtime
3028 @opindex fnext-runtime
3029 Generate output compatible with the NeXT runtime. This is the default
3030 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3031 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3034 @item -fno-nil-receivers
3035 @opindex fno-nil-receivers
3036 Assume that all Objective-C message dispatches (@code{[receiver
3037 message:arg]}) in this translation unit ensure that the receiver is
3038 not @code{nil}. This allows for more efficient entry points in the
3039 runtime to be used. This option is only available in conjunction with
3040 the NeXT runtime and ABI version 0 or 1.
3042 @item -fobjc-abi-version=@var{n}
3043 @opindex fobjc-abi-version
3044 Use version @var{n} of the Objective-C ABI for the selected runtime.
3045 This option is currently supported only for the NeXT runtime. In that
3046 case, Version 0 is the traditional (32-bit) ABI without support for
3047 properties and other Objective-C 2.0 additions. Version 1 is the
3048 traditional (32-bit) ABI with support for properties and other
3049 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3050 nothing is specified, the default is Version 0 on 32-bit target
3051 machines, and Version 2 on 64-bit target machines.
3053 @item -fobjc-call-cxx-cdtors
3054 @opindex fobjc-call-cxx-cdtors
3055 For each Objective-C class, check if any of its instance variables is a
3056 C++ object with a non-trivial default constructor. If so, synthesize a
3057 special @code{- (id) .cxx_construct} instance method which runs
3058 non-trivial default constructors on any such instance variables, in order,
3059 and then return @code{self}. Similarly, check if any instance variable
3060 is a C++ object with a non-trivial destructor, and if so, synthesize a
3061 special @code{- (void) .cxx_destruct} method which runs
3062 all such default destructors, in reverse order.
3064 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3065 methods thusly generated only operate on instance variables
3066 declared in the current Objective-C class, and not those inherited
3067 from superclasses. It is the responsibility of the Objective-C
3068 runtime to invoke all such methods in an object's inheritance
3069 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3070 by the runtime immediately after a new object instance is allocated;
3071 the @code{- (void) .cxx_destruct} methods are invoked immediately
3072 before the runtime deallocates an object instance.
3074 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3075 support for invoking the @code{- (id) .cxx_construct} and
3076 @code{- (void) .cxx_destruct} methods.
3078 @item -fobjc-direct-dispatch
3079 @opindex fobjc-direct-dispatch
3080 Allow fast jumps to the message dispatcher. On Darwin this is
3081 accomplished via the comm page.
3083 @item -fobjc-exceptions
3084 @opindex fobjc-exceptions
3085 Enable syntactic support for structured exception handling in
3086 Objective-C, similar to what is offered by C++ and Java. This option
3087 is required to use the Objective-C keywords @code{@@try},
3088 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3089 @code{@@synchronized}. This option is available with both the GNU
3090 runtime and the NeXT runtime (but not available in conjunction with
3091 the NeXT runtime on Mac OS X 10.2 and earlier).
3095 Enable garbage collection (GC) in Objective-C and Objective-C++
3096 programs. This option is only available with the NeXT runtime; the
3097 GNU runtime has a different garbage collection implementation that
3098 does not require special compiler flags.
3100 @item -fobjc-nilcheck
3101 @opindex fobjc-nilcheck
3102 For the NeXT runtime with version 2 of the ABI, check for a nil
3103 receiver in method invocations before doing the actual method call.
3104 This is the default and can be disabled using
3105 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3106 checked for nil in this way no matter what this flag is set to.
3107 Currently this flag does nothing when the GNU runtime, or an older
3108 version of the NeXT runtime ABI, is used.
3110 @item -fobjc-std=objc1
3112 Conform to the language syntax of Objective-C 1.0, the language
3113 recognized by GCC 4.0. This only affects the Objective-C additions to
3114 the C/C++ language; it does not affect conformance to C/C++ standards,
3115 which is controlled by the separate C/C++ dialect option flags. When
3116 this option is used with the Objective-C or Objective-C++ compiler,
3117 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3118 This is useful if you need to make sure that your Objective-C code can
3119 be compiled with older versions of GCC@.
3121 @item -freplace-objc-classes
3122 @opindex freplace-objc-classes
3123 Emit a special marker instructing @command{ld(1)} not to statically link in
3124 the resulting object file, and allow @command{dyld(1)} to load it in at
3125 run time instead. This is used in conjunction with the Fix-and-Continue
3126 debugging mode, where the object file in question may be recompiled and
3127 dynamically reloaded in the course of program execution, without the need
3128 to restart the program itself. Currently, Fix-and-Continue functionality
3129 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3134 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3135 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3136 compile time) with static class references that get initialized at load time,
3137 which improves run-time performance. Specifying the @option{-fzero-link} flag
3138 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3139 to be retained. This is useful in Zero-Link debugging mode, since it allows
3140 for individual class implementations to be modified during program execution.
3141 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3142 regardless of command-line options.
3144 @item -fno-local-ivars
3145 @opindex fno-local-ivars
3146 @opindex flocal-ivars
3147 By default instance variables in Objective-C can be accessed as if
3148 they were local variables from within the methods of the class they're
3149 declared in. This can lead to shadowing between instance variables
3150 and other variables declared either locally inside a class method or
3151 globally with the same name. Specifying the @option{-fno-local-ivars}
3152 flag disables this behavior thus avoiding variable shadowing issues.
3154 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3155 @opindex fivar-visibility
3156 Set the default instance variable visibility to the specified option
3157 so that instance variables declared outside the scope of any access
3158 modifier directives default to the specified visibility.
3162 Dump interface declarations for all classes seen in the source file to a
3163 file named @file{@var{sourcename}.decl}.
3165 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3166 @opindex Wassign-intercept
3167 @opindex Wno-assign-intercept
3168 Warn whenever an Objective-C assignment is being intercepted by the
3171 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3172 @opindex Wno-protocol
3174 If a class is declared to implement a protocol, a warning is issued for
3175 every method in the protocol that is not implemented by the class. The
3176 default behavior is to issue a warning for every method not explicitly
3177 implemented in the class, even if a method implementation is inherited
3178 from the superclass. If you use the @option{-Wno-protocol} option, then
3179 methods inherited from the superclass are considered to be implemented,
3180 and no warning is issued for them.
3182 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3184 @opindex Wno-selector
3185 Warn if multiple methods of different types for the same selector are
3186 found during compilation. The check is performed on the list of methods
3187 in the final stage of compilation. Additionally, a check is performed
3188 for each selector appearing in a @code{@@selector(@dots{})}
3189 expression, and a corresponding method for that selector has been found
3190 during compilation. Because these checks scan the method table only at
3191 the end of compilation, these warnings are not produced if the final
3192 stage of compilation is not reached, for example because an error is
3193 found during compilation, or because the @option{-fsyntax-only} option is
3196 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3197 @opindex Wstrict-selector-match
3198 @opindex Wno-strict-selector-match
3199 Warn if multiple methods with differing argument and/or return types are
3200 found for a given selector when attempting to send a message using this
3201 selector to a receiver of type @code{id} or @code{Class}. When this flag
3202 is off (which is the default behavior), the compiler omits such warnings
3203 if any differences found are confined to types that share the same size
3206 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3207 @opindex Wundeclared-selector
3208 @opindex Wno-undeclared-selector
3209 Warn if a @code{@@selector(@dots{})} expression referring to an
3210 undeclared selector is found. A selector is considered undeclared if no
3211 method with that name has been declared before the
3212 @code{@@selector(@dots{})} expression, either explicitly in an
3213 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3214 an @code{@@implementation} section. This option always performs its
3215 checks as soon as a @code{@@selector(@dots{})} expression is found,
3216 while @option{-Wselector} only performs its checks in the final stage of
3217 compilation. This also enforces the coding style convention
3218 that methods and selectors must be declared before being used.
3220 @item -print-objc-runtime-info
3221 @opindex print-objc-runtime-info
3222 Generate C header describing the largest structure that is passed by
3227 @node Diagnostic Message Formatting Options
3228 @section Options to Control Diagnostic Messages Formatting
3229 @cindex options to control diagnostics formatting
3230 @cindex diagnostic messages
3231 @cindex message formatting
3233 Traditionally, diagnostic messages have been formatted irrespective of
3234 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3235 options described below
3236 to control the formatting algorithm for diagnostic messages,
3237 e.g.@: how many characters per line, how often source location
3238 information should be reported. Note that some language front ends may not
3239 honor these options.
3242 @item -fmessage-length=@var{n}
3243 @opindex fmessage-length
3244 Try to format error messages so that they fit on lines of about
3245 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3246 done; each error message appears on a single line. This is the
3247 default for all front ends.
3249 @item -fdiagnostics-show-location=once
3250 @opindex fdiagnostics-show-location
3251 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3252 reporter to emit source location information @emph{once}; that is, in
3253 case the message is too long to fit on a single physical line and has to
3254 be wrapped, the source location won't be emitted (as prefix) again,
3255 over and over, in subsequent continuation lines. This is the default
3258 @item -fdiagnostics-show-location=every-line
3259 Only meaningful in line-wrapping mode. Instructs the diagnostic
3260 messages reporter to emit the same source location information (as
3261 prefix) for physical lines that result from the process of breaking
3262 a message which is too long to fit on a single line.
3264 @item -fdiagnostics-color[=@var{WHEN}]
3265 @itemx -fno-diagnostics-color
3266 @opindex fdiagnostics-color
3267 @cindex highlight, color, colour
3268 @vindex GCC_COLORS @r{environment variable}
3269 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3270 or @samp{auto}. The default depends on how the compiler has been configured,
3271 it can be any of the above @var{WHEN} options or also @samp{never}
3272 if @env{GCC_COLORS} environment variable isn't present in the environment,
3273 and @samp{auto} otherwise.
3274 @samp{auto} means to use color only when the standard error is a terminal.
3275 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3276 aliases for @option{-fdiagnostics-color=always} and
3277 @option{-fdiagnostics-color=never}, respectively.
3279 The colors are defined by the environment variable @env{GCC_COLORS}.
3280 Its value is a colon-separated list of capabilities and Select Graphic
3281 Rendition (SGR) substrings. SGR commands are interpreted by the
3282 terminal or terminal emulator. (See the section in the documentation
3283 of your text terminal for permitted values and their meanings as
3284 character attributes.) These substring values are integers in decimal
3285 representation and can be concatenated with semicolons.
3286 Common values to concatenate include
3288 @samp{4} for underline,
3290 @samp{7} for inverse,
3291 @samp{39} for default foreground color,
3292 @samp{30} to @samp{37} for foreground colors,
3293 @samp{90} to @samp{97} for 16-color mode foreground colors,
3294 @samp{38;5;0} to @samp{38;5;255}
3295 for 88-color and 256-color modes foreground colors,
3296 @samp{49} for default background color,
3297 @samp{40} to @samp{47} for background colors,
3298 @samp{100} to @samp{107} for 16-color mode background colors,
3299 and @samp{48;5;0} to @samp{48;5;255}
3300 for 88-color and 256-color modes background colors.
3302 The default @env{GCC_COLORS} is
3304 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3307 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3308 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3309 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3310 string disables colors.
3311 Supported capabilities are as follows.
3315 @vindex error GCC_COLORS @r{capability}
3316 SGR substring for error: markers.
3319 @vindex warning GCC_COLORS @r{capability}
3320 SGR substring for warning: markers.
3323 @vindex note GCC_COLORS @r{capability}
3324 SGR substring for note: markers.
3327 @vindex caret GCC_COLORS @r{capability}
3328 SGR substring for caret line.
3331 @vindex locus GCC_COLORS @r{capability}
3332 SGR substring for location information, @samp{file:line} or
3333 @samp{file:line:column} etc.
3336 @vindex quote GCC_COLORS @r{capability}
3337 SGR substring for information printed within quotes.
3340 @item -fno-diagnostics-show-option
3341 @opindex fno-diagnostics-show-option
3342 @opindex fdiagnostics-show-option
3343 By default, each diagnostic emitted includes text indicating the
3344 command-line option that directly controls the diagnostic (if such an
3345 option is known to the diagnostic machinery). Specifying the
3346 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3348 @item -fno-diagnostics-show-caret
3349 @opindex fno-diagnostics-show-caret
3350 @opindex fdiagnostics-show-caret
3351 By default, each diagnostic emitted includes the original source line
3352 and a caret '^' indicating the column. This option suppresses this
3353 information. The source line is truncated to @var{n} characters, if
3354 the @option{-fmessage-length=n} option is given. When the output is done
3355 to the terminal, the width is limited to the width given by the
3356 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3360 @node Warning Options
3361 @section Options to Request or Suppress Warnings
3362 @cindex options to control warnings
3363 @cindex warning messages
3364 @cindex messages, warning
3365 @cindex suppressing warnings
3367 Warnings are diagnostic messages that report constructions that
3368 are not inherently erroneous but that are risky or suggest there
3369 may have been an error.
3371 The following language-independent options do not enable specific
3372 warnings but control the kinds of diagnostics produced by GCC@.
3375 @cindex syntax checking
3377 @opindex fsyntax-only
3378 Check the code for syntax errors, but don't do anything beyond that.
3380 @item -fmax-errors=@var{n}
3381 @opindex fmax-errors
3382 Limits the maximum number of error messages to @var{n}, at which point
3383 GCC bails out rather than attempting to continue processing the source
3384 code. If @var{n} is 0 (the default), there is no limit on the number
3385 of error messages produced. If @option{-Wfatal-errors} is also
3386 specified, then @option{-Wfatal-errors} takes precedence over this
3391 Inhibit all warning messages.
3396 Make all warnings into errors.
3401 Make the specified warning into an error. The specifier for a warning
3402 is appended; for example @option{-Werror=switch} turns the warnings
3403 controlled by @option{-Wswitch} into errors. This switch takes a
3404 negative form, to be used to negate @option{-Werror} for specific
3405 warnings; for example @option{-Wno-error=switch} makes
3406 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3409 The warning message for each controllable warning includes the
3410 option that controls the warning. That option can then be used with
3411 @option{-Werror=} and @option{-Wno-error=} as described above.
3412 (Printing of the option in the warning message can be disabled using the
3413 @option{-fno-diagnostics-show-option} flag.)
3415 Note that specifying @option{-Werror=}@var{foo} automatically implies
3416 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3419 @item -Wfatal-errors
3420 @opindex Wfatal-errors
3421 @opindex Wno-fatal-errors
3422 This option causes the compiler to abort compilation on the first error
3423 occurred rather than trying to keep going and printing further error
3428 You can request many specific warnings with options beginning with
3429 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3430 implicit declarations. Each of these specific warning options also
3431 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3432 example, @option{-Wno-implicit}. This manual lists only one of the
3433 two forms, whichever is not the default. For further
3434 language-specific options also refer to @ref{C++ Dialect Options} and
3435 @ref{Objective-C and Objective-C++ Dialect Options}.
3437 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3438 options, such as @option{-Wunused}, which may turn on further options,
3439 such as @option{-Wunused-value}. The combined effect of positive and
3440 negative forms is that more specific options have priority over less
3441 specific ones, independently of their position in the command-line. For
3442 options of the same specificity, the last one takes effect. Options
3443 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3444 as if they appeared at the end of the command-line.
3446 When an unrecognized warning option is requested (e.g.,
3447 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3448 that the option is not recognized. However, if the @option{-Wno-} form
3449 is used, the behavior is slightly different: no diagnostic is
3450 produced for @option{-Wno-unknown-warning} unless other diagnostics
3451 are being produced. This allows the use of new @option{-Wno-} options
3452 with old compilers, but if something goes wrong, the compiler
3453 warns that an unrecognized option is present.
3460 Issue all the warnings demanded by strict ISO C and ISO C++;
3461 reject all programs that use forbidden extensions, and some other
3462 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3463 version of the ISO C standard specified by any @option{-std} option used.
3465 Valid ISO C and ISO C++ programs should compile properly with or without
3466 this option (though a rare few require @option{-ansi} or a
3467 @option{-std} option specifying the required version of ISO C)@. However,
3468 without this option, certain GNU extensions and traditional C and C++
3469 features are supported as well. With this option, they are rejected.
3471 @option{-Wpedantic} does not cause warning messages for use of the
3472 alternate keywords whose names begin and end with @samp{__}. Pedantic
3473 warnings are also disabled in the expression that follows
3474 @code{__extension__}. However, only system header files should use
3475 these escape routes; application programs should avoid them.
3476 @xref{Alternate Keywords}.
3478 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3479 C conformance. They soon find that it does not do quite what they want:
3480 it finds some non-ISO practices, but not all---only those for which
3481 ISO C @emph{requires} a diagnostic, and some others for which
3482 diagnostics have been added.
3484 A feature to report any failure to conform to ISO C might be useful in
3485 some instances, but would require considerable additional work and would
3486 be quite different from @option{-Wpedantic}. We don't have plans to
3487 support such a feature in the near future.
3489 Where the standard specified with @option{-std} represents a GNU
3490 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3491 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3492 extended dialect is based. Warnings from @option{-Wpedantic} are given
3493 where they are required by the base standard. (It does not make sense
3494 for such warnings to be given only for features not in the specified GNU
3495 C dialect, since by definition the GNU dialects of C include all
3496 features the compiler supports with the given option, and there would be
3497 nothing to warn about.)
3499 @item -pedantic-errors
3500 @opindex pedantic-errors
3501 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3502 requires a diagnostic, in some cases where there is undefined behavior
3503 at compile-time and in some other cases that do not prevent compilation
3504 of programs that are valid according to the standard. This is not
3505 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3506 by this option and not enabled by the latter and vice versa.
3511 This enables all the warnings about constructions that some users
3512 consider questionable, and that are easy to avoid (or modify to
3513 prevent the warning), even in conjunction with macros. This also
3514 enables some language-specific warnings described in @ref{C++ Dialect
3515 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3517 @option{-Wall} turns on the following warning flags:
3519 @gccoptlist{-Waddress @gol
3520 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3521 -Wc++11-compat -Wc++14-compat@gol
3522 -Wchar-subscripts @gol
3523 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3524 -Wimplicit-int @r{(C and Objective-C only)} @gol
3525 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3527 -Wduplicated-cond @gol
3530 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3531 -Wmaybe-uninitialized @gol
3532 -Wmissing-braces @r{(only for C/ObjC)} @gol
3539 -Wsequence-point @gol
3540 -Wsign-compare @r{(only in C++)} @gol
3541 -Wstrict-aliasing @gol
3542 -Wstrict-overflow=1 @gol
3544 -Wtautological-compare @gol
3546 -Wuninitialized @gol
3547 -Wunknown-pragmas @gol
3548 -Wunused-function @gol
3551 -Wunused-variable @gol
3552 -Wvolatile-register-var @gol
3555 Note that some warning flags are not implied by @option{-Wall}. Some of
3556 them warn about constructions that users generally do not consider
3557 questionable, but which occasionally you might wish to check for;
3558 others warn about constructions that are necessary or hard to avoid in
3559 some cases, and there is no simple way to modify the code to suppress
3560 the warning. Some of them are enabled by @option{-Wextra} but many of
3561 them must be enabled individually.
3567 This enables some extra warning flags that are not enabled by
3568 @option{-Wall}. (This option used to be called @option{-W}. The older
3569 name is still supported, but the newer name is more descriptive.)
3571 @gccoptlist{-Wclobbered @gol
3573 -Wignored-qualifiers @gol
3574 -Wmissing-field-initializers @gol
3575 -Wmissing-parameter-type @r{(C only)} @gol
3576 -Wold-style-declaration @r{(C only)} @gol
3577 -Woverride-init @gol
3580 -Wuninitialized @gol
3581 -Wshift-negative-value @gol
3582 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3583 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3586 The option @option{-Wextra} also prints warning messages for the
3592 A pointer is compared against integer zero with @code{<}, @code{<=},
3593 @code{>}, or @code{>=}.
3596 (C++ only) An enumerator and a non-enumerator both appear in a
3597 conditional expression.
3600 (C++ only) Ambiguous virtual bases.
3603 (C++ only) Subscripting an array that has been declared @code{register}.
3606 (C++ only) Taking the address of a variable that has been declared
3610 (C++ only) A base class is not initialized in a derived class's copy
3615 @item -Wchar-subscripts
3616 @opindex Wchar-subscripts
3617 @opindex Wno-char-subscripts
3618 Warn if an array subscript has type @code{char}. This is a common cause
3619 of error, as programmers often forget that this type is signed on some
3621 This warning is enabled by @option{-Wall}.
3625 @opindex Wno-comment
3626 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3627 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3628 This warning is enabled by @option{-Wall}.
3630 @item -Wno-coverage-mismatch
3631 @opindex Wno-coverage-mismatch
3632 Warn if feedback profiles do not match when using the
3633 @option{-fprofile-use} option.
3634 If a source file is changed between compiling with @option{-fprofile-gen} and
3635 with @option{-fprofile-use}, the files with the profile feedback can fail
3636 to match the source file and GCC cannot use the profile feedback
3637 information. By default, this warning is enabled and is treated as an
3638 error. @option{-Wno-coverage-mismatch} can be used to disable the
3639 warning or @option{-Wno-error=coverage-mismatch} can be used to
3640 disable the error. Disabling the error for this warning can result in
3641 poorly optimized code and is useful only in the
3642 case of very minor changes such as bug fixes to an existing code-base.
3643 Completely disabling the warning is not recommended.
3646 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3648 Suppress warning messages emitted by @code{#warning} directives.
3650 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3651 @opindex Wdouble-promotion
3652 @opindex Wno-double-promotion
3653 Give a warning when a value of type @code{float} is implicitly
3654 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3655 floating-point unit implement @code{float} in hardware, but emulate
3656 @code{double} in software. On such a machine, doing computations
3657 using @code{double} values is much more expensive because of the
3658 overhead required for software emulation.
3660 It is easy to accidentally do computations with @code{double} because
3661 floating-point literals are implicitly of type @code{double}. For
3665 float area(float radius)
3667 return 3.14159 * radius * radius;
3671 the compiler performs the entire computation with @code{double}
3672 because the floating-point literal is a @code{double}.
3675 @itemx -Wformat=@var{n}
3678 @opindex ffreestanding
3679 @opindex fno-builtin
3681 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3682 the arguments supplied have types appropriate to the format string
3683 specified, and that the conversions specified in the format string make
3684 sense. This includes standard functions, and others specified by format
3685 attributes (@pxref{Function Attributes}), in the @code{printf},
3686 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3687 not in the C standard) families (or other target-specific families).
3688 Which functions are checked without format attributes having been
3689 specified depends on the standard version selected, and such checks of
3690 functions without the attribute specified are disabled by
3691 @option{-ffreestanding} or @option{-fno-builtin}.
3693 The formats are checked against the format features supported by GNU
3694 libc version 2.2. These include all ISO C90 and C99 features, as well
3695 as features from the Single Unix Specification and some BSD and GNU
3696 extensions. Other library implementations may not support all these
3697 features; GCC does not support warning about features that go beyond a
3698 particular library's limitations. However, if @option{-Wpedantic} is used
3699 with @option{-Wformat}, warnings are given about format features not
3700 in the selected standard version (but not for @code{strfmon} formats,
3701 since those are not in any version of the C standard). @xref{C Dialect
3702 Options,,Options Controlling C Dialect}.
3709 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3710 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3711 @option{-Wformat} also checks for null format arguments for several
3712 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3713 aspects of this level of format checking can be disabled by the
3714 options: @option{-Wno-format-contains-nul},
3715 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3716 @option{-Wformat} is enabled by @option{-Wall}.
3718 @item -Wno-format-contains-nul
3719 @opindex Wno-format-contains-nul
3720 @opindex Wformat-contains-nul
3721 If @option{-Wformat} is specified, do not warn about format strings that
3724 @item -Wno-format-extra-args
3725 @opindex Wno-format-extra-args
3726 @opindex Wformat-extra-args
3727 If @option{-Wformat} is specified, do not warn about excess arguments to a
3728 @code{printf} or @code{scanf} format function. The C standard specifies
3729 that such arguments are ignored.
3731 Where the unused arguments lie between used arguments that are
3732 specified with @samp{$} operand number specifications, normally
3733 warnings are still given, since the implementation could not know what
3734 type to pass to @code{va_arg} to skip the unused arguments. However,
3735 in the case of @code{scanf} formats, this option suppresses the
3736 warning if the unused arguments are all pointers, since the Single
3737 Unix Specification says that such unused arguments are allowed.
3739 @item -Wno-format-zero-length
3740 @opindex Wno-format-zero-length
3741 @opindex Wformat-zero-length
3742 If @option{-Wformat} is specified, do not warn about zero-length formats.
3743 The C standard specifies that zero-length formats are allowed.
3748 Enable @option{-Wformat} plus additional format checks. Currently
3749 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3752 @item -Wformat-nonliteral
3753 @opindex Wformat-nonliteral
3754 @opindex Wno-format-nonliteral
3755 If @option{-Wformat} is specified, also warn if the format string is not a
3756 string literal and so cannot be checked, unless the format function
3757 takes its format arguments as a @code{va_list}.
3759 @item -Wformat-security
3760 @opindex Wformat-security
3761 @opindex Wno-format-security
3762 If @option{-Wformat} is specified, also warn about uses of format
3763 functions that represent possible security problems. At present, this
3764 warns about calls to @code{printf} and @code{scanf} functions where the
3765 format string is not a string literal and there are no format arguments,
3766 as in @code{printf (foo);}. This may be a security hole if the format
3767 string came from untrusted input and contains @samp{%n}. (This is
3768 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3769 in future warnings may be added to @option{-Wformat-security} that are not
3770 included in @option{-Wformat-nonliteral}.)
3772 @item -Wformat-signedness
3773 @opindex Wformat-signedness
3774 @opindex Wno-format-signedness
3775 If @option{-Wformat} is specified, also warn if the format string
3776 requires an unsigned argument and the argument is signed and vice versa.
3779 @opindex Wformat-y2k
3780 @opindex Wno-format-y2k
3781 If @option{-Wformat} is specified, also warn about @code{strftime}
3782 formats that may yield only a two-digit year.
3787 @opindex Wno-nonnull
3788 Warn about passing a null pointer for arguments marked as
3789 requiring a non-null value by the @code{nonnull} function attribute.
3791 Also warns when comparing an argument marked with the @code{nonnull}
3792 function attribute against null inside the function.
3794 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3795 can be disabled with the @option{-Wno-nonnull} option.
3797 @item -Wnull-dereference
3798 @opindex Wnull-dereference
3799 @opindex Wno-null-dereference
3800 Warn if the compiler detects paths that trigger erroneous or
3801 undefined behavior due to dereferencing a null pointer. This option
3802 is only active when @option{-fdelete-null-pointer-checks} is active,
3803 which is enabled by optimizations in most targets. The precision of
3804 the warnings depends on the optimization options used.
3806 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3808 @opindex Wno-init-self
3809 Warn about uninitialized variables that are initialized with themselves.
3810 Note this option can only be used with the @option{-Wuninitialized} option.
3812 For example, GCC warns about @code{i} being uninitialized in the
3813 following snippet only when @option{-Winit-self} has been specified:
3824 This warning is enabled by @option{-Wall} in C++.
3826 @item -Wimplicit-int @r{(C and Objective-C only)}
3827 @opindex Wimplicit-int
3828 @opindex Wno-implicit-int
3829 Warn when a declaration does not specify a type.
3830 This warning is enabled by @option{-Wall}.
3832 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3833 @opindex Wimplicit-function-declaration
3834 @opindex Wno-implicit-function-declaration
3835 Give a warning whenever a function is used before being declared. In
3836 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3837 enabled by default and it is made into an error by
3838 @option{-pedantic-errors}. This warning is also enabled by
3841 @item -Wimplicit @r{(C and Objective-C only)}
3843 @opindex Wno-implicit
3844 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3845 This warning is enabled by @option{-Wall}.
3847 @item -Wignored-qualifiers @r{(C and C++ only)}
3848 @opindex Wignored-qualifiers
3849 @opindex Wno-ignored-qualifiers
3850 Warn if the return type of a function has a type qualifier
3851 such as @code{const}. For ISO C such a type qualifier has no effect,
3852 since the value returned by a function is not an lvalue.
3853 For C++, the warning is only emitted for scalar types or @code{void}.
3854 ISO C prohibits qualified @code{void} return types on function
3855 definitions, so such return types always receive a warning
3856 even without this option.
3858 This warning is also enabled by @option{-Wextra}.
3863 Warn if the type of @code{main} is suspicious. @code{main} should be
3864 a function with external linkage, returning int, taking either zero
3865 arguments, two, or three arguments of appropriate types. This warning
3866 is enabled by default in C++ and is enabled by either @option{-Wall}
3867 or @option{-Wpedantic}.
3869 @item -Wmisleading-indentation @r{(C and C++ only)}
3870 @opindex Wmisleading-indentation
3871 @opindex Wno-misleading-indentation
3872 Warn when the indentation of the code does not reflect the block structure.
3873 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3874 @code{for} clauses with a guarded statement that does not use braces,
3875 followed by an unguarded statement with the same indentation.
3877 This warning is disabled by default.
3879 In the following example, the call to ``bar'' is misleadingly indented as
3880 if it were guarded by the ``if'' conditional.
3883 if (some_condition ())
3885 bar (); /* Gotcha: this is not guarded by the "if". */
3888 In the case of mixed tabs and spaces, the warning uses the
3889 @option{-ftabstop=} option to determine if the statements line up
3892 The warning is not issued for code involving multiline preprocessor logic
3893 such as the following example.
3898 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3904 The warning is not issued after a @code{#line} directive, since this
3905 typically indicates autogenerated code, and no assumptions can be made
3906 about the layout of the file that the directive references.
3908 @item -Wmissing-braces
3909 @opindex Wmissing-braces
3910 @opindex Wno-missing-braces
3911 Warn if an aggregate or union initializer is not fully bracketed. In
3912 the following example, the initializer for @code{a} is not fully
3913 bracketed, but that for @code{b} is fully bracketed. This warning is
3914 enabled by @option{-Wall} in C.
3917 int a[2][2] = @{ 0, 1, 2, 3 @};
3918 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3921 This warning is enabled by @option{-Wall}.
3923 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3924 @opindex Wmissing-include-dirs
3925 @opindex Wno-missing-include-dirs
3926 Warn if a user-supplied include directory does not exist.
3929 @opindex Wparentheses
3930 @opindex Wno-parentheses
3931 Warn if parentheses are omitted in certain contexts, such
3932 as when there is an assignment in a context where a truth value
3933 is expected, or when operators are nested whose precedence people
3934 often get confused about.
3936 Also warn if a comparison like @code{x<=y<=z} appears; this is
3937 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3938 interpretation from that of ordinary mathematical notation.
3940 Also warn about constructions where there may be confusion to which
3941 @code{if} statement an @code{else} branch belongs. Here is an example of
3956 In C/C++, every @code{else} branch belongs to the innermost possible
3957 @code{if} statement, which in this example is @code{if (b)}. This is
3958 often not what the programmer expected, as illustrated in the above
3959 example by indentation the programmer chose. When there is the
3960 potential for this confusion, GCC issues a warning when this flag
3961 is specified. To eliminate the warning, add explicit braces around
3962 the innermost @code{if} statement so there is no way the @code{else}
3963 can belong to the enclosing @code{if}. The resulting code
3980 Also warn for dangerous uses of the GNU extension to
3981 @code{?:} with omitted middle operand. When the condition
3982 in the @code{?}: operator is a boolean expression, the omitted value is
3983 always 1. Often programmers expect it to be a value computed
3984 inside the conditional expression instead.
3986 This warning is enabled by @option{-Wall}.
3988 @item -Wsequence-point
3989 @opindex Wsequence-point
3990 @opindex Wno-sequence-point
3991 Warn about code that may have undefined semantics because of violations
3992 of sequence point rules in the C and C++ standards.
3994 The C and C++ standards define the order in which expressions in a C/C++
3995 program are evaluated in terms of @dfn{sequence points}, which represent
3996 a partial ordering between the execution of parts of the program: those
3997 executed before the sequence point, and those executed after it. These
3998 occur after the evaluation of a full expression (one which is not part
3999 of a larger expression), after the evaluation of the first operand of a
4000 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4001 function is called (but after the evaluation of its arguments and the
4002 expression denoting the called function), and in certain other places.
4003 Other than as expressed by the sequence point rules, the order of
4004 evaluation of subexpressions of an expression is not specified. All
4005 these rules describe only a partial order rather than a total order,
4006 since, for example, if two functions are called within one expression
4007 with no sequence point between them, the order in which the functions
4008 are called is not specified. However, the standards committee have
4009 ruled that function calls do not overlap.
4011 It is not specified when between sequence points modifications to the
4012 values of objects take effect. Programs whose behavior depends on this
4013 have undefined behavior; the C and C++ standards specify that ``Between
4014 the previous and next sequence point an object shall have its stored
4015 value modified at most once by the evaluation of an expression.
4016 Furthermore, the prior value shall be read only to determine the value
4017 to be stored.''. If a program breaks these rules, the results on any
4018 particular implementation are entirely unpredictable.
4020 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4021 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4022 diagnosed by this option, and it may give an occasional false positive
4023 result, but in general it has been found fairly effective at detecting
4024 this sort of problem in programs.
4026 The standard is worded confusingly, therefore there is some debate
4027 over the precise meaning of the sequence point rules in subtle cases.
4028 Links to discussions of the problem, including proposed formal
4029 definitions, may be found on the GCC readings page, at
4030 @uref{http://gcc.gnu.org/@/readings.html}.
4032 This warning is enabled by @option{-Wall} for C and C++.
4034 @item -Wno-return-local-addr
4035 @opindex Wno-return-local-addr
4036 @opindex Wreturn-local-addr
4037 Do not warn about returning a pointer (or in C++, a reference) to a
4038 variable that goes out of scope after the function returns.
4041 @opindex Wreturn-type
4042 @opindex Wno-return-type
4043 Warn whenever a function is defined with a return type that defaults
4044 to @code{int}. Also warn about any @code{return} statement with no
4045 return value in a function whose return type is not @code{void}
4046 (falling off the end of the function body is considered returning
4047 without a value), and about a @code{return} statement with an
4048 expression in a function whose return type is @code{void}.
4050 For C++, a function without return type always produces a diagnostic
4051 message, even when @option{-Wno-return-type} is specified. The only
4052 exceptions are @code{main} and functions defined in system headers.
4054 This warning is enabled by @option{-Wall}.
4056 @item -Wshift-count-negative
4057 @opindex Wshift-count-negative
4058 @opindex Wno-shift-count-negative
4059 Warn if shift count is negative. This warning is enabled by default.
4061 @item -Wshift-count-overflow
4062 @opindex Wshift-count-overflow
4063 @opindex Wno-shift-count-overflow
4064 Warn if shift count >= width of type. This warning is enabled by default.
4066 @item -Wshift-negative-value
4067 @opindex Wshift-negative-value
4068 @opindex Wno-shift-negative-value
4069 Warn if left shifting a negative value. This warning is enabled by
4070 @option{-Wextra} in C99 and C++11 modes (and newer).
4072 @item -Wshift-overflow
4073 @itemx -Wshift-overflow=@var{n}
4074 @opindex Wshift-overflow
4075 @opindex Wno-shift-overflow
4076 Warn about left shift overflows. This warning is enabled by
4077 default in C99 and C++11 modes (and newer).
4080 @item -Wshift-overflow=1
4081 This is the warning level of @option{-Wshift-overflow} and is enabled
4082 by default in C99 and C++11 modes (and newer). This warning level does
4083 not warn about left-shifting 1 into the sign bit. (However, in C, such
4084 an overflow is still rejected in contexts where an integer constant expression
4087 @item -Wshift-overflow=2
4088 This warning level also warns about left-shifting 1 into the sign bit,
4089 unless C++14 mode is active.
4095 Warn whenever a @code{switch} statement has an index of enumerated type
4096 and lacks a @code{case} for one or more of the named codes of that
4097 enumeration. (The presence of a @code{default} label prevents this
4098 warning.) @code{case} labels outside the enumeration range also
4099 provoke warnings when this option is used (even if there is a
4100 @code{default} label).
4101 This warning is enabled by @option{-Wall}.
4103 @item -Wswitch-default
4104 @opindex Wswitch-default
4105 @opindex Wno-switch-default
4106 Warn whenever a @code{switch} statement does not have a @code{default}
4110 @opindex Wswitch-enum
4111 @opindex Wno-switch-enum
4112 Warn whenever a @code{switch} statement has an index of enumerated type
4113 and lacks a @code{case} for one or more of the named codes of that
4114 enumeration. @code{case} labels outside the enumeration range also
4115 provoke warnings when this option is used. The only difference
4116 between @option{-Wswitch} and this option is that this option gives a
4117 warning about an omitted enumeration code even if there is a
4118 @code{default} label.
4121 @opindex Wswitch-bool
4122 @opindex Wno-switch-bool
4123 Warn whenever a @code{switch} statement has an index of boolean type
4124 and the case values are outside the range of a boolean type.
4125 It is possible to suppress this warning by casting the controlling
4126 expression to a type other than @code{bool}. For example:
4129 switch ((int) (a == 4))
4135 This warning is enabled by default for C and C++ programs.
4137 @item -Wsync-nand @r{(C and C++ only)}
4139 @opindex Wno-sync-nand
4140 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4141 built-in functions are used. These functions changed semantics in GCC 4.4.
4145 @opindex Wno-trigraphs
4146 Warn if any trigraphs are encountered that might change the meaning of
4147 the program (trigraphs within comments are not warned about).
4148 This warning is enabled by @option{-Wall}.
4150 @item -Wunused-but-set-parameter
4151 @opindex Wunused-but-set-parameter
4152 @opindex Wno-unused-but-set-parameter
4153 Warn whenever a function parameter is assigned to, but otherwise unused
4154 (aside from its declaration).
4156 To suppress this warning use the @code{unused} attribute
4157 (@pxref{Variable Attributes}).
4159 This warning is also enabled by @option{-Wunused} together with
4162 @item -Wunused-but-set-variable
4163 @opindex Wunused-but-set-variable
4164 @opindex Wno-unused-but-set-variable
4165 Warn whenever a local variable is assigned to, but otherwise unused
4166 (aside from its declaration).
4167 This warning is enabled by @option{-Wall}.
4169 To suppress this warning use the @code{unused} attribute
4170 (@pxref{Variable Attributes}).
4172 This warning is also enabled by @option{-Wunused}, which is enabled
4175 @item -Wunused-function
4176 @opindex Wunused-function
4177 @opindex Wno-unused-function
4178 Warn whenever a static function is declared but not defined or a
4179 non-inline static function is unused.
4180 This warning is enabled by @option{-Wall}.
4182 @item -Wunused-label
4183 @opindex Wunused-label
4184 @opindex Wno-unused-label
4185 Warn whenever a label is declared but not used.
4186 This warning is enabled by @option{-Wall}.
4188 To suppress this warning use the @code{unused} attribute
4189 (@pxref{Variable Attributes}).
4191 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4192 @opindex Wunused-local-typedefs
4193 Warn when a typedef locally defined in a function is not used.
4194 This warning is enabled by @option{-Wall}.
4196 @item -Wunused-parameter
4197 @opindex Wunused-parameter
4198 @opindex Wno-unused-parameter
4199 Warn whenever a function parameter is unused aside from its declaration.
4201 To suppress this warning use the @code{unused} attribute
4202 (@pxref{Variable Attributes}).
4204 @item -Wno-unused-result
4205 @opindex Wunused-result
4206 @opindex Wno-unused-result
4207 Do not warn if a caller of a function marked with attribute
4208 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4209 its return value. The default is @option{-Wunused-result}.
4211 @item -Wunused-variable
4212 @opindex Wunused-variable
4213 @opindex Wno-unused-variable
4214 Warn whenever a local or static variable is unused aside from its
4215 declaration. This option implies @option{-Wunused-const-variable} for C,
4216 but not for C++. This warning is enabled by @option{-Wall}.
4218 To suppress this warning use the @code{unused} attribute
4219 (@pxref{Variable Attributes}).
4221 @item -Wunused-const-variable
4222 @opindex Wunused-const-variable
4223 @opindex Wno-unused-const-variable
4224 Warn whenever a constant static variable is unused aside from its declaration.
4225 This warning is enabled by @option{-Wunused-variable} for C, but not for C++.
4226 In C++ this is normally not an error since const variables take the place of
4227 @code{#define}s in C++.
4229 To suppress this warning use the @code{unused} attribute
4230 (@pxref{Variable Attributes}).
4232 @item -Wunused-value
4233 @opindex Wunused-value
4234 @opindex Wno-unused-value
4235 Warn whenever a statement computes a result that is explicitly not
4236 used. To suppress this warning cast the unused expression to
4237 @code{void}. This includes an expression-statement or the left-hand
4238 side of a comma expression that contains no side effects. For example,
4239 an expression such as @code{x[i,j]} causes a warning, while
4240 @code{x[(void)i,j]} does not.
4242 This warning is enabled by @option{-Wall}.
4247 All the above @option{-Wunused} options combined.
4249 In order to get a warning about an unused function parameter, you must
4250 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4251 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4253 @item -Wuninitialized
4254 @opindex Wuninitialized
4255 @opindex Wno-uninitialized
4256 Warn if an automatic variable is used without first being initialized
4257 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4258 warn if a non-static reference or non-static @code{const} member
4259 appears in a class without constructors.
4261 If you want to warn about code that uses the uninitialized value of the
4262 variable in its own initializer, use the @option{-Winit-self} option.
4264 These warnings occur for individual uninitialized or clobbered
4265 elements of structure, union or array variables as well as for
4266 variables that are uninitialized or clobbered as a whole. They do
4267 not occur for variables or elements declared @code{volatile}. Because
4268 these warnings depend on optimization, the exact variables or elements
4269 for which there are warnings depends on the precise optimization
4270 options and version of GCC used.
4272 Note that there may be no warning about a variable that is used only
4273 to compute a value that itself is never used, because such
4274 computations may be deleted by data flow analysis before the warnings
4277 @item -Wmaybe-uninitialized
4278 @opindex Wmaybe-uninitialized
4279 @opindex Wno-maybe-uninitialized
4280 For an automatic variable, if there exists a path from the function
4281 entry to a use of the variable that is initialized, but there exist
4282 some other paths for which the variable is not initialized, the compiler
4283 emits a warning if it cannot prove the uninitialized paths are not
4284 executed at run time. These warnings are made optional because GCC is
4285 not smart enough to see all the reasons why the code might be correct
4286 in spite of appearing to have an error. Here is one example of how
4307 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4308 always initialized, but GCC doesn't know this. To suppress the
4309 warning, you need to provide a default case with assert(0) or
4312 @cindex @code{longjmp} warnings
4313 This option also warns when a non-volatile automatic variable might be
4314 changed by a call to @code{longjmp}. These warnings as well are possible
4315 only in optimizing compilation.
4317 The compiler sees only the calls to @code{setjmp}. It cannot know
4318 where @code{longjmp} will be called; in fact, a signal handler could
4319 call it at any point in the code. As a result, you may get a warning
4320 even when there is in fact no problem because @code{longjmp} cannot
4321 in fact be called at the place that would cause a problem.
4323 Some spurious warnings can be avoided if you declare all the functions
4324 you use that never return as @code{noreturn}. @xref{Function
4327 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4329 @item -Wunknown-pragmas
4330 @opindex Wunknown-pragmas
4331 @opindex Wno-unknown-pragmas
4332 @cindex warning for unknown pragmas
4333 @cindex unknown pragmas, warning
4334 @cindex pragmas, warning of unknown
4335 Warn when a @code{#pragma} directive is encountered that is not understood by
4336 GCC@. If this command-line option is used, warnings are even issued
4337 for unknown pragmas in system header files. This is not the case if
4338 the warnings are only enabled by the @option{-Wall} command-line option.
4341 @opindex Wno-pragmas
4343 Do not warn about misuses of pragmas, such as incorrect parameters,
4344 invalid syntax, or conflicts between pragmas. See also
4345 @option{-Wunknown-pragmas}.
4347 @item -Wstrict-aliasing
4348 @opindex Wstrict-aliasing
4349 @opindex Wno-strict-aliasing
4350 This option is only active when @option{-fstrict-aliasing} is active.
4351 It warns about code that might break the strict aliasing rules that the
4352 compiler is using for optimization. The warning does not catch all
4353 cases, but does attempt to catch the more common pitfalls. It is
4354 included in @option{-Wall}.
4355 It is equivalent to @option{-Wstrict-aliasing=3}
4357 @item -Wstrict-aliasing=n
4358 @opindex Wstrict-aliasing=n
4359 This option is only active when @option{-fstrict-aliasing} is active.
4360 It warns about code that might break the strict aliasing rules that the
4361 compiler is using for optimization.
4362 Higher levels correspond to higher accuracy (fewer false positives).
4363 Higher levels also correspond to more effort, similar to the way @option{-O}
4365 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4367 Level 1: Most aggressive, quick, least accurate.
4368 Possibly useful when higher levels
4369 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4370 false negatives. However, it has many false positives.
4371 Warns for all pointer conversions between possibly incompatible types,
4372 even if never dereferenced. Runs in the front end only.
4374 Level 2: Aggressive, quick, not too precise.
4375 May still have many false positives (not as many as level 1 though),
4376 and few false negatives (but possibly more than level 1).
4377 Unlike level 1, it only warns when an address is taken. Warns about
4378 incomplete types. Runs in the front end only.
4380 Level 3 (default for @option{-Wstrict-aliasing}):
4381 Should have very few false positives and few false
4382 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4383 Takes care of the common pun+dereference pattern in the front end:
4384 @code{*(int*)&some_float}.
4385 If optimization is enabled, it also runs in the back end, where it deals
4386 with multiple statement cases using flow-sensitive points-to information.
4387 Only warns when the converted pointer is dereferenced.
4388 Does not warn about incomplete types.
4390 @item -Wstrict-overflow
4391 @itemx -Wstrict-overflow=@var{n}
4392 @opindex Wstrict-overflow
4393 @opindex Wno-strict-overflow
4394 This option is only active when @option{-fstrict-overflow} is active.
4395 It warns about cases where the compiler optimizes based on the
4396 assumption that signed overflow does not occur. Note that it does not
4397 warn about all cases where the code might overflow: it only warns
4398 about cases where the compiler implements some optimization. Thus
4399 this warning depends on the optimization level.
4401 An optimization that assumes that signed overflow does not occur is
4402 perfectly safe if the values of the variables involved are such that
4403 overflow never does, in fact, occur. Therefore this warning can
4404 easily give a false positive: a warning about code that is not
4405 actually a problem. To help focus on important issues, several
4406 warning levels are defined. No warnings are issued for the use of
4407 undefined signed overflow when estimating how many iterations a loop
4408 requires, in particular when determining whether a loop will be
4412 @item -Wstrict-overflow=1
4413 Warn about cases that are both questionable and easy to avoid. For
4414 example, with @option{-fstrict-overflow}, the compiler simplifies
4415 @code{x + 1 > x} to @code{1}. This level of
4416 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4417 are not, and must be explicitly requested.
4419 @item -Wstrict-overflow=2
4420 Also warn about other cases where a comparison is simplified to a
4421 constant. For example: @code{abs (x) >= 0}. This can only be
4422 simplified when @option{-fstrict-overflow} is in effect, because
4423 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4424 zero. @option{-Wstrict-overflow} (with no level) is the same as
4425 @option{-Wstrict-overflow=2}.
4427 @item -Wstrict-overflow=3
4428 Also warn about other cases where a comparison is simplified. For
4429 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4431 @item -Wstrict-overflow=4
4432 Also warn about other simplifications not covered by the above cases.
4433 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4435 @item -Wstrict-overflow=5
4436 Also warn about cases where the compiler reduces the magnitude of a
4437 constant involved in a comparison. For example: @code{x + 2 > y} is
4438 simplified to @code{x + 1 >= y}. This is reported only at the
4439 highest warning level because this simplification applies to many
4440 comparisons, so this warning level gives a very large number of
4444 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4445 @opindex Wsuggest-attribute=
4446 @opindex Wno-suggest-attribute=
4447 Warn for cases where adding an attribute may be beneficial. The
4448 attributes currently supported are listed below.
4451 @item -Wsuggest-attribute=pure
4452 @itemx -Wsuggest-attribute=const
4453 @itemx -Wsuggest-attribute=noreturn
4454 @opindex Wsuggest-attribute=pure
4455 @opindex Wno-suggest-attribute=pure
4456 @opindex Wsuggest-attribute=const
4457 @opindex Wno-suggest-attribute=const
4458 @opindex Wsuggest-attribute=noreturn
4459 @opindex Wno-suggest-attribute=noreturn
4461 Warn about functions that might be candidates for attributes
4462 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4463 functions visible in other compilation units or (in the case of @code{pure} and
4464 @code{const}) if it cannot prove that the function returns normally. A function
4465 returns normally if it doesn't contain an infinite loop or return abnormally
4466 by throwing, calling @code{abort} or trapping. This analysis requires option
4467 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4468 higher. Higher optimization levels improve the accuracy of the analysis.
4470 @item -Wsuggest-attribute=format
4471 @itemx -Wmissing-format-attribute
4472 @opindex Wsuggest-attribute=format
4473 @opindex Wmissing-format-attribute
4474 @opindex Wno-suggest-attribute=format
4475 @opindex Wno-missing-format-attribute
4479 Warn about function pointers that might be candidates for @code{format}
4480 attributes. Note these are only possible candidates, not absolute ones.
4481 GCC guesses that function pointers with @code{format} attributes that
4482 are used in assignment, initialization, parameter passing or return
4483 statements should have a corresponding @code{format} attribute in the
4484 resulting type. I.e.@: the left-hand side of the assignment or
4485 initialization, the type of the parameter variable, or the return type
4486 of the containing function respectively should also have a @code{format}
4487 attribute to avoid the warning.
4489 GCC also warns about function definitions that might be
4490 candidates for @code{format} attributes. Again, these are only
4491 possible candidates. GCC guesses that @code{format} attributes
4492 might be appropriate for any function that calls a function like
4493 @code{vprintf} or @code{vscanf}, but this might not always be the
4494 case, and some functions for which @code{format} attributes are
4495 appropriate may not be detected.
4498 @item -Wsuggest-final-types
4499 @opindex Wno-suggest-final-types
4500 @opindex Wsuggest-final-types
4501 Warn about types with virtual methods where code quality would be improved
4502 if the type were declared with the C++11 @code{final} specifier,
4504 declared in an anonymous namespace. This allows GCC to more aggressively
4505 devirtualize the polymorphic calls. This warning is more effective with link
4506 time optimization, where the information about the class hierarchy graph is
4509 @item -Wsuggest-final-methods
4510 @opindex Wno-suggest-final-methods
4511 @opindex Wsuggest-final-methods
4512 Warn about virtual methods where code quality would be improved if the method
4513 were declared with the C++11 @code{final} specifier,
4514 or, if possible, its type were
4515 declared in an anonymous namespace or with the @code{final} specifier.
4517 more effective with link time optimization, where the information about the
4518 class hierarchy graph is more complete. It is recommended to first consider
4519 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4522 @item -Wsuggest-override
4523 Warn about overriding virtual functions that are not marked with the override
4526 @item -Warray-bounds
4527 @itemx -Warray-bounds=@var{n}
4528 @opindex Wno-array-bounds
4529 @opindex Warray-bounds
4530 This option is only active when @option{-ftree-vrp} is active
4531 (default for @option{-O2} and above). It warns about subscripts to arrays
4532 that are always out of bounds. This warning is enabled by @option{-Wall}.
4535 @item -Warray-bounds=1
4536 This is the warning level of @option{-Warray-bounds} and is enabled
4537 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4539 @item -Warray-bounds=2
4540 This warning level also warns about out of bounds access for
4541 arrays at the end of a struct and for arrays accessed through
4542 pointers. This warning level may give a larger number of
4543 false positives and is deactivated by default.
4546 @item -Wbool-compare
4547 @opindex Wno-bool-compare
4548 @opindex Wbool-compare
4549 Warn about boolean expression compared with an integer value different from
4550 @code{true}/@code{false}. For instance, the following comparison is
4555 if ((n > 1) == 2) @{ @dots{} @}
4557 This warning is enabled by @option{-Wall}.
4559 @item -Wduplicated-cond
4560 @opindex Wno-duplicated-cond
4561 @opindex Wduplicated-cond
4562 Warn about duplicated conditions in an if-else-if chain. For instance,
4563 warn for the following code:
4565 if (p->q != NULL) @{ @dots{} @}
4566 else if (p->q != NULL) @{ @dots{} @}
4568 This warning is enabled by @option{-Wall}.
4570 @item -Wframe-address
4571 @opindex Wno-frame-address
4572 @opindex Wframe-address
4573 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4574 is called with an argument greater than 0. Such calls may return indeterminate
4575 values or crash the program. The warning is included in @option{-Wall}.
4577 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4578 @opindex Wno-discarded-qualifiers
4579 @opindex Wdiscarded-qualifiers
4580 Do not warn if type qualifiers on pointers are being discarded.
4581 Typically, the compiler warns if a @code{const char *} variable is
4582 passed to a function that takes a @code{char *} parameter. This option
4583 can be used to suppress such a warning.
4585 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4586 @opindex Wno-discarded-array-qualifiers
4587 @opindex Wdiscarded-array-qualifiers
4588 Do not warn if type qualifiers on arrays which are pointer targets
4589 are being discarded. Typically, the compiler warns if a
4590 @code{const int (*)[]} variable is passed to a function that
4591 takes a @code{int (*)[]} parameter. This option can be used to
4592 suppress such a warning.
4594 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4595 @opindex Wno-incompatible-pointer-types
4596 @opindex Wincompatible-pointer-types
4597 Do not warn when there is a conversion between pointers that have incompatible
4598 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4599 which warns for pointer argument passing or assignment with different
4602 @item -Wno-int-conversion @r{(C and Objective-C only)}
4603 @opindex Wno-int-conversion
4604 @opindex Wint-conversion
4605 Do not warn about incompatible integer to pointer and pointer to integer
4606 conversions. This warning is about implicit conversions; for explicit
4607 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4608 @option{-Wno-pointer-to-int-cast} may be used.
4610 @item -Wno-div-by-zero
4611 @opindex Wno-div-by-zero
4612 @opindex Wdiv-by-zero
4613 Do not warn about compile-time integer division by zero. Floating-point
4614 division by zero is not warned about, as it can be a legitimate way of
4615 obtaining infinities and NaNs.
4617 @item -Wsystem-headers
4618 @opindex Wsystem-headers
4619 @opindex Wno-system-headers
4620 @cindex warnings from system headers
4621 @cindex system headers, warnings from
4622 Print warning messages for constructs found in system header files.
4623 Warnings from system headers are normally suppressed, on the assumption
4624 that they usually do not indicate real problems and would only make the
4625 compiler output harder to read. Using this command-line option tells
4626 GCC to emit warnings from system headers as if they occurred in user
4627 code. However, note that using @option{-Wall} in conjunction with this
4628 option does @emph{not} warn about unknown pragmas in system
4629 headers---for that, @option{-Wunknown-pragmas} must also be used.
4631 @item -Wtautological-compare
4632 @opindex Wtautological-compare
4633 @opindex Wno-tautological-compare
4634 Warn if a self-comparison always evaluates to true or false. This
4635 warning detects various mistakes such as:
4639 if (i > i) @{ @dots{} @}
4641 This warning is enabled by @option{-Wall}.
4644 @opindex Wtrampolines
4645 @opindex Wno-trampolines
4646 Warn about trampolines generated for pointers to nested functions.
4647 A trampoline is a small piece of data or code that is created at run
4648 time on the stack when the address of a nested function is taken, and is
4649 used to call the nested function indirectly. For some targets, it is
4650 made up of data only and thus requires no special treatment. But, for
4651 most targets, it is made up of code and thus requires the stack to be
4652 made executable in order for the program to work properly.
4655 @opindex Wfloat-equal
4656 @opindex Wno-float-equal
4657 Warn if floating-point values are used in equality comparisons.
4659 The idea behind this is that sometimes it is convenient (for the
4660 programmer) to consider floating-point values as approximations to
4661 infinitely precise real numbers. If you are doing this, then you need
4662 to compute (by analyzing the code, or in some other way) the maximum or
4663 likely maximum error that the computation introduces, and allow for it
4664 when performing comparisons (and when producing output, but that's a
4665 different problem). In particular, instead of testing for equality, you
4666 should check to see whether the two values have ranges that overlap; and
4667 this is done with the relational operators, so equality comparisons are
4670 @item -Wtraditional @r{(C and Objective-C only)}
4671 @opindex Wtraditional
4672 @opindex Wno-traditional
4673 Warn about certain constructs that behave differently in traditional and
4674 ISO C@. Also warn about ISO C constructs that have no traditional C
4675 equivalent, and/or problematic constructs that should be avoided.
4679 Macro parameters that appear within string literals in the macro body.
4680 In traditional C macro replacement takes place within string literals,
4681 but in ISO C it does not.
4684 In traditional C, some preprocessor directives did not exist.
4685 Traditional preprocessors only considered a line to be a directive
4686 if the @samp{#} appeared in column 1 on the line. Therefore
4687 @option{-Wtraditional} warns about directives that traditional C
4688 understands but ignores because the @samp{#} does not appear as the
4689 first character on the line. It also suggests you hide directives like
4690 @code{#pragma} not understood by traditional C by indenting them. Some
4691 traditional implementations do not recognize @code{#elif}, so this option
4692 suggests avoiding it altogether.
4695 A function-like macro that appears without arguments.
4698 The unary plus operator.
4701 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4702 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4703 constants.) Note, these suffixes appear in macros defined in the system
4704 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4705 Use of these macros in user code might normally lead to spurious
4706 warnings, however GCC's integrated preprocessor has enough context to
4707 avoid warning in these cases.
4710 A function declared external in one block and then used after the end of
4714 A @code{switch} statement has an operand of type @code{long}.
4717 A non-@code{static} function declaration follows a @code{static} one.
4718 This construct is not accepted by some traditional C compilers.
4721 The ISO type of an integer constant has a different width or
4722 signedness from its traditional type. This warning is only issued if
4723 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4724 typically represent bit patterns, are not warned about.
4727 Usage of ISO string concatenation is detected.
4730 Initialization of automatic aggregates.
4733 Identifier conflicts with labels. Traditional C lacks a separate
4734 namespace for labels.
4737 Initialization of unions. If the initializer is zero, the warning is
4738 omitted. This is done under the assumption that the zero initializer in
4739 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4740 initializer warnings and relies on default initialization to zero in the
4744 Conversions by prototypes between fixed/floating-point values and vice
4745 versa. The absence of these prototypes when compiling with traditional
4746 C causes serious problems. This is a subset of the possible
4747 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4750 Use of ISO C style function definitions. This warning intentionally is
4751 @emph{not} issued for prototype declarations or variadic functions
4752 because these ISO C features appear in your code when using
4753 libiberty's traditional C compatibility macros, @code{PARAMS} and
4754 @code{VPARAMS}. This warning is also bypassed for nested functions
4755 because that feature is already a GCC extension and thus not relevant to
4756 traditional C compatibility.
4759 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4760 @opindex Wtraditional-conversion
4761 @opindex Wno-traditional-conversion
4762 Warn if a prototype causes a type conversion that is different from what
4763 would happen to the same argument in the absence of a prototype. This
4764 includes conversions of fixed point to floating and vice versa, and
4765 conversions changing the width or signedness of a fixed-point argument
4766 except when the same as the default promotion.
4768 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4769 @opindex Wdeclaration-after-statement
4770 @opindex Wno-declaration-after-statement
4771 Warn when a declaration is found after a statement in a block. This
4772 construct, known from C++, was introduced with ISO C99 and is by default
4773 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4778 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4780 @item -Wno-endif-labels
4781 @opindex Wno-endif-labels
4782 @opindex Wendif-labels
4783 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4788 Warn whenever a local variable or type declaration shadows another
4789 variable, parameter, type, class member (in C++), or instance variable
4790 (in Objective-C) or whenever a built-in function is shadowed. Note
4791 that in C++, the compiler warns if a local variable shadows an
4792 explicit typedef, but not if it shadows a struct/class/enum.
4794 @item -Wno-shadow-ivar @r{(Objective-C only)}
4795 @opindex Wno-shadow-ivar
4796 @opindex Wshadow-ivar
4797 Do not warn whenever a local variable shadows an instance variable in an
4800 @item -Wlarger-than=@var{len}
4801 @opindex Wlarger-than=@var{len}
4802 @opindex Wlarger-than-@var{len}
4803 Warn whenever an object of larger than @var{len} bytes is defined.
4805 @item -Wframe-larger-than=@var{len}
4806 @opindex Wframe-larger-than
4807 Warn if the size of a function frame is larger than @var{len} bytes.
4808 The computation done to determine the stack frame size is approximate
4809 and not conservative.
4810 The actual requirements may be somewhat greater than @var{len}
4811 even if you do not get a warning. In addition, any space allocated
4812 via @code{alloca}, variable-length arrays, or related constructs
4813 is not included by the compiler when determining
4814 whether or not to issue a warning.
4816 @item -Wno-free-nonheap-object
4817 @opindex Wno-free-nonheap-object
4818 @opindex Wfree-nonheap-object
4819 Do not warn when attempting to free an object that was not allocated
4822 @item -Wstack-usage=@var{len}
4823 @opindex Wstack-usage
4824 Warn if the stack usage of a function might be larger than @var{len} bytes.
4825 The computation done to determine the stack usage is conservative.
4826 Any space allocated via @code{alloca}, variable-length arrays, or related
4827 constructs is included by the compiler when determining whether or not to
4830 The message is in keeping with the output of @option{-fstack-usage}.
4834 If the stack usage is fully static but exceeds the specified amount, it's:
4837 warning: stack usage is 1120 bytes
4840 If the stack usage is (partly) dynamic but bounded, it's:
4843 warning: stack usage might be 1648 bytes
4846 If the stack usage is (partly) dynamic and not bounded, it's:
4849 warning: stack usage might be unbounded
4853 @item -Wunsafe-loop-optimizations
4854 @opindex Wunsafe-loop-optimizations
4855 @opindex Wno-unsafe-loop-optimizations
4856 Warn if the loop cannot be optimized because the compiler cannot
4857 assume anything on the bounds of the loop indices. With
4858 @option{-funsafe-loop-optimizations} warn if the compiler makes
4861 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4862 @opindex Wno-pedantic-ms-format
4863 @opindex Wpedantic-ms-format
4864 When used in combination with @option{-Wformat}
4865 and @option{-pedantic} without GNU extensions, this option
4866 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4867 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4868 which depend on the MS runtime.
4870 @item -Wpointer-arith
4871 @opindex Wpointer-arith
4872 @opindex Wno-pointer-arith
4873 Warn about anything that depends on the ``size of'' a function type or
4874 of @code{void}. GNU C assigns these types a size of 1, for
4875 convenience in calculations with @code{void *} pointers and pointers
4876 to functions. In C++, warn also when an arithmetic operation involves
4877 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4880 @opindex Wtype-limits
4881 @opindex Wno-type-limits
4882 Warn if a comparison is always true or always false due to the limited
4883 range of the data type, but do not warn for constant expressions. For
4884 example, warn if an unsigned variable is compared against zero with
4885 @code{<} or @code{>=}. This warning is also enabled by
4888 @item -Wbad-function-cast @r{(C and Objective-C only)}
4889 @opindex Wbad-function-cast
4890 @opindex Wno-bad-function-cast
4891 Warn when a function call is cast to a non-matching type.
4892 For example, warn if a call to a function returning an integer type
4893 is cast to a pointer type.
4895 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4896 @opindex Wc90-c99-compat
4897 @opindex Wno-c90-c99-compat
4898 Warn about features not present in ISO C90, but present in ISO C99.
4899 For instance, warn about use of variable length arrays, @code{long long}
4900 type, @code{bool} type, compound literals, designated initializers, and so
4901 on. This option is independent of the standards mode. Warnings are disabled
4902 in the expression that follows @code{__extension__}.
4904 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4905 @opindex Wc99-c11-compat
4906 @opindex Wno-c99-c11-compat
4907 Warn about features not present in ISO C99, but present in ISO C11.
4908 For instance, warn about use of anonymous structures and unions,
4909 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4910 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4911 and so on. This option is independent of the standards mode. Warnings are
4912 disabled in the expression that follows @code{__extension__}.
4914 @item -Wc++-compat @r{(C and Objective-C only)}
4915 @opindex Wc++-compat
4916 Warn about ISO C constructs that are outside of the common subset of
4917 ISO C and ISO C++, e.g.@: request for implicit conversion from
4918 @code{void *} to a pointer to non-@code{void} type.
4920 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4921 @opindex Wc++11-compat
4922 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4923 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4924 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4925 enabled by @option{-Wall}.
4927 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4928 @opindex Wc++14-compat
4929 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4930 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4934 @opindex Wno-cast-qual
4935 Warn whenever a pointer is cast so as to remove a type qualifier from
4936 the target type. For example, warn if a @code{const char *} is cast
4937 to an ordinary @code{char *}.
4939 Also warn when making a cast that introduces a type qualifier in an
4940 unsafe way. For example, casting @code{char **} to @code{const char **}
4941 is unsafe, as in this example:
4944 /* p is char ** value. */
4945 const char **q = (const char **) p;
4946 /* Assignment of readonly string to const char * is OK. */
4948 /* Now char** pointer points to read-only memory. */
4953 @opindex Wcast-align
4954 @opindex Wno-cast-align
4955 Warn whenever a pointer is cast such that the required alignment of the
4956 target is increased. For example, warn if a @code{char *} is cast to
4957 an @code{int *} on machines where integers can only be accessed at
4958 two- or four-byte boundaries.
4960 @item -Wwrite-strings
4961 @opindex Wwrite-strings
4962 @opindex Wno-write-strings
4963 When compiling C, give string constants the type @code{const
4964 char[@var{length}]} so that copying the address of one into a
4965 non-@code{const} @code{char *} pointer produces a warning. These
4966 warnings help you find at compile time code that can try to write
4967 into a string constant, but only if you have been very careful about
4968 using @code{const} in declarations and prototypes. Otherwise, it is
4969 just a nuisance. This is why we did not make @option{-Wall} request
4972 When compiling C++, warn about the deprecated conversion from string
4973 literals to @code{char *}. This warning is enabled by default for C++
4978 @opindex Wno-clobbered
4979 Warn for variables that might be changed by @code{longjmp} or
4980 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4982 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4983 @opindex Wconditionally-supported
4984 @opindex Wno-conditionally-supported
4985 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4988 @opindex Wconversion
4989 @opindex Wno-conversion
4990 Warn for implicit conversions that may alter a value. This includes
4991 conversions between real and integer, like @code{abs (x)} when
4992 @code{x} is @code{double}; conversions between signed and unsigned,
4993 like @code{unsigned ui = -1}; and conversions to smaller types, like
4994 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4995 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4996 changed by the conversion like in @code{abs (2.0)}. Warnings about
4997 conversions between signed and unsigned integers can be disabled by
4998 using @option{-Wno-sign-conversion}.
5000 For C++, also warn for confusing overload resolution for user-defined
5001 conversions; and conversions that never use a type conversion
5002 operator: conversions to @code{void}, the same type, a base class or a
5003 reference to them. Warnings about conversions between signed and
5004 unsigned integers are disabled by default in C++ unless
5005 @option{-Wsign-conversion} is explicitly enabled.
5007 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5008 @opindex Wconversion-null
5009 @opindex Wno-conversion-null
5010 Do not warn for conversions between @code{NULL} and non-pointer
5011 types. @option{-Wconversion-null} is enabled by default.
5013 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5014 @opindex Wzero-as-null-pointer-constant
5015 @opindex Wno-zero-as-null-pointer-constant
5016 Warn when a literal '0' is used as null pointer constant. This can
5017 be useful to facilitate the conversion to @code{nullptr} in C++11.
5019 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5020 @opindex Wsubobject-linkage
5021 @opindex Wno-subobject-linkage
5022 Warn if a class type has a base or a field whose type uses the anonymous
5023 namespace or depends on a type with no linkage. If a type A depends on
5024 a type B with no or internal linkage, defining it in multiple
5025 translation units would be an ODR violation because the meaning of B
5026 is different in each translation unit. If A only appears in a single
5027 translation unit, the best way to silence the warning is to give it
5028 internal linkage by putting it in an anonymous namespace as well. The
5029 compiler doesn't give this warning for types defined in the main .C
5030 file, as those are unlikely to have multiple definitions.
5031 @option{-Wsubobject-linkage} is enabled by default.
5035 @opindex Wno-date-time
5036 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5037 are encountered as they might prevent bit-wise-identical reproducible
5040 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5041 @opindex Wdelete-incomplete
5042 @opindex Wno-delete-incomplete
5043 Warn when deleting a pointer to incomplete type, which may cause
5044 undefined behavior at runtime. This warning is enabled by default.
5046 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5047 @opindex Wuseless-cast
5048 @opindex Wno-useless-cast
5049 Warn when an expression is casted to its own type.
5052 @opindex Wempty-body
5053 @opindex Wno-empty-body
5054 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5055 while} statement. This warning is also enabled by @option{-Wextra}.
5057 @item -Wenum-compare
5058 @opindex Wenum-compare
5059 @opindex Wno-enum-compare
5060 Warn about a comparison between values of different enumerated types.
5061 In C++ enumeral mismatches in conditional expressions are also
5062 diagnosed and the warning is enabled by default. In C this warning is
5063 enabled by @option{-Wall}.
5065 @item -Wjump-misses-init @r{(C, Objective-C only)}
5066 @opindex Wjump-misses-init
5067 @opindex Wno-jump-misses-init
5068 Warn if a @code{goto} statement or a @code{switch} statement jumps
5069 forward across the initialization of a variable, or jumps backward to a
5070 label after the variable has been initialized. This only warns about
5071 variables that are initialized when they are declared. This warning is
5072 only supported for C and Objective-C; in C++ this sort of branch is an
5075 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5076 can be disabled with the @option{-Wno-jump-misses-init} option.
5078 @item -Wsign-compare
5079 @opindex Wsign-compare
5080 @opindex Wno-sign-compare
5081 @cindex warning for comparison of signed and unsigned values
5082 @cindex comparison of signed and unsigned values, warning
5083 @cindex signed and unsigned values, comparison warning
5084 Warn when a comparison between signed and unsigned values could produce
5085 an incorrect result when the signed value is converted to unsigned.
5086 This warning is also enabled by @option{-Wextra}; to get the other warnings
5087 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
5089 @item -Wsign-conversion
5090 @opindex Wsign-conversion
5091 @opindex Wno-sign-conversion
5092 Warn for implicit conversions that may change the sign of an integer
5093 value, like assigning a signed integer expression to an unsigned
5094 integer variable. An explicit cast silences the warning. In C, this
5095 option is enabled also by @option{-Wconversion}.
5097 @item -Wfloat-conversion
5098 @opindex Wfloat-conversion
5099 @opindex Wno-float-conversion
5100 Warn for implicit conversions that reduce the precision of a real value.
5101 This includes conversions from real to integer, and from higher precision
5102 real to lower precision real values. This option is also enabled by
5103 @option{-Wconversion}.
5105 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5106 @opindex Wsized-deallocation
5107 @opindex Wno-sized-deallocation
5108 Warn about a definition of an unsized deallocation function
5110 void operator delete (void *) noexcept;
5111 void operator delete[] (void *) noexcept;
5113 without a definition of the corresponding sized deallocation function
5115 void operator delete (void *, std::size_t) noexcept;
5116 void operator delete[] (void *, std::size_t) noexcept;
5118 or vice versa. Enabled by @option{-Wextra} along with
5119 @option{-fsized-deallocation}.
5121 @item -Wsizeof-pointer-memaccess
5122 @opindex Wsizeof-pointer-memaccess
5123 @opindex Wno-sizeof-pointer-memaccess
5124 Warn for suspicious length parameters to certain string and memory built-in
5125 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5126 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5127 but a pointer, and suggests a possible fix, or about
5128 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5131 @item -Wsizeof-array-argument
5132 @opindex Wsizeof-array-argument
5133 @opindex Wno-sizeof-array-argument
5134 Warn when the @code{sizeof} operator is applied to a parameter that is
5135 declared as an array in a function definition. This warning is enabled by
5136 default for C and C++ programs.
5138 @item -Wmemset-transposed-args
5139 @opindex Wmemset-transposed-args
5140 @opindex Wno-memset-transposed-args
5141 Warn for suspicious calls to the @code{memset} built-in function, if the
5142 second argument is not zero and the third argument is zero. This warns e.g.@
5143 about @code{memset (buf, sizeof buf, 0)} where most probably
5144 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5145 is only emitted if the third argument is literal zero. If it is some
5146 expression that is folded to zero, a cast of zero to some type, etc.,
5147 it is far less likely that the user has mistakenly exchanged the arguments
5148 and no warning is emitted. This warning is enabled by @option{-Wall}.
5152 @opindex Wno-address
5153 Warn about suspicious uses of memory addresses. These include using
5154 the address of a function in a conditional expression, such as
5155 @code{void func(void); if (func)}, and comparisons against the memory
5156 address of a string literal, such as @code{if (x == "abc")}. Such
5157 uses typically indicate a programmer error: the address of a function
5158 always evaluates to true, so their use in a conditional usually
5159 indicate that the programmer forgot the parentheses in a function
5160 call; and comparisons against string literals result in unspecified
5161 behavior and are not portable in C, so they usually indicate that the
5162 programmer intended to use @code{strcmp}. This warning is enabled by
5166 @opindex Wlogical-op
5167 @opindex Wno-logical-op
5168 Warn about suspicious uses of logical operators in expressions.
5169 This includes using logical operators in contexts where a
5170 bit-wise operator is likely to be expected. Also warns when
5171 the operands of a logical operator are the same:
5174 if (a < 0 && a < 0) @{ @dots{} @}
5177 @item -Wlogical-not-parentheses
5178 @opindex Wlogical-not-parentheses
5179 @opindex Wno-logical-not-parentheses
5180 Warn about logical not used on the left hand side operand of a comparison.
5181 This option does not warn if the RHS operand is of a boolean type. Its
5182 purpose is to detect suspicious code like the following:
5186 if (!a > 1) @{ @dots{} @}
5189 It is possible to suppress the warning by wrapping the LHS into
5192 if ((!a) > 1) @{ @dots{} @}
5195 This warning is enabled by @option{-Wall}.
5197 @item -Waggregate-return
5198 @opindex Waggregate-return
5199 @opindex Wno-aggregate-return
5200 Warn if any functions that return structures or unions are defined or
5201 called. (In languages where you can return an array, this also elicits
5204 @item -Wno-aggressive-loop-optimizations
5205 @opindex Wno-aggressive-loop-optimizations
5206 @opindex Waggressive-loop-optimizations
5207 Warn if in a loop with constant number of iterations the compiler detects
5208 undefined behavior in some statement during one or more of the iterations.
5210 @item -Wno-attributes
5211 @opindex Wno-attributes
5212 @opindex Wattributes
5213 Do not warn if an unexpected @code{__attribute__} is used, such as
5214 unrecognized attributes, function attributes applied to variables,
5215 etc. This does not stop errors for incorrect use of supported
5218 @item -Wno-builtin-macro-redefined
5219 @opindex Wno-builtin-macro-redefined
5220 @opindex Wbuiltin-macro-redefined
5221 Do not warn if certain built-in macros are redefined. This suppresses
5222 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5223 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5225 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5226 @opindex Wstrict-prototypes
5227 @opindex Wno-strict-prototypes
5228 Warn if a function is declared or defined without specifying the
5229 argument types. (An old-style function definition is permitted without
5230 a warning if preceded by a declaration that specifies the argument
5233 @item -Wold-style-declaration @r{(C and Objective-C only)}
5234 @opindex Wold-style-declaration
5235 @opindex Wno-old-style-declaration
5236 Warn for obsolescent usages, according to the C Standard, in a
5237 declaration. For example, warn if storage-class specifiers like
5238 @code{static} are not the first things in a declaration. This warning
5239 is also enabled by @option{-Wextra}.
5241 @item -Wold-style-definition @r{(C and Objective-C only)}
5242 @opindex Wold-style-definition
5243 @opindex Wno-old-style-definition
5244 Warn if an old-style function definition is used. A warning is given
5245 even if there is a previous prototype.
5247 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5248 @opindex Wmissing-parameter-type
5249 @opindex Wno-missing-parameter-type
5250 A function parameter is declared without a type specifier in K&R-style
5257 This warning is also enabled by @option{-Wextra}.
5259 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5260 @opindex Wmissing-prototypes
5261 @opindex Wno-missing-prototypes
5262 Warn if a global function is defined without a previous prototype
5263 declaration. This warning is issued even if the definition itself
5264 provides a prototype. Use this option to detect global functions
5265 that do not have a matching prototype declaration in a header file.
5266 This option is not valid for C++ because all function declarations
5267 provide prototypes and a non-matching declaration declares an
5268 overload rather than conflict with an earlier declaration.
5269 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5271 @item -Wmissing-declarations
5272 @opindex Wmissing-declarations
5273 @opindex Wno-missing-declarations
5274 Warn if a global function is defined without a previous declaration.
5275 Do so even if the definition itself provides a prototype.
5276 Use this option to detect global functions that are not declared in
5277 header files. In C, no warnings are issued for functions with previous
5278 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5279 missing prototypes. In C++, no warnings are issued for function templates,
5280 or for inline functions, or for functions in anonymous namespaces.
5282 @item -Wmissing-field-initializers
5283 @opindex Wmissing-field-initializers
5284 @opindex Wno-missing-field-initializers
5288 Warn if a structure's initializer has some fields missing. For
5289 example, the following code causes such a warning, because
5290 @code{x.h} is implicitly zero:
5293 struct s @{ int f, g, h; @};
5294 struct s x = @{ 3, 4 @};
5297 This option does not warn about designated initializers, so the following
5298 modification does not trigger a warning:
5301 struct s @{ int f, g, h; @};
5302 struct s x = @{ .f = 3, .g = 4 @};
5305 In C++ this option does not warn either about the empty @{ @}
5306 initializer, for example:
5309 struct s @{ int f, g, h; @};
5313 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5314 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5316 @item -Wno-multichar
5317 @opindex Wno-multichar
5319 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5320 Usually they indicate a typo in the user's code, as they have
5321 implementation-defined values, and should not be used in portable code.
5323 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5324 @opindex Wnormalized=
5325 @opindex Wnormalized
5326 @opindex Wno-normalized
5329 @cindex character set, input normalization
5330 In ISO C and ISO C++, two identifiers are different if they are
5331 different sequences of characters. However, sometimes when characters
5332 outside the basic ASCII character set are used, you can have two
5333 different character sequences that look the same. To avoid confusion,
5334 the ISO 10646 standard sets out some @dfn{normalization rules} which
5335 when applied ensure that two sequences that look the same are turned into
5336 the same sequence. GCC can warn you if you are using identifiers that
5337 have not been normalized; this option controls that warning.
5339 There are four levels of warning supported by GCC@. The default is
5340 @option{-Wnormalized=nfc}, which warns about any identifier that is
5341 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5342 recommended form for most uses. It is equivalent to
5343 @option{-Wnormalized}.
5345 Unfortunately, there are some characters allowed in identifiers by
5346 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5347 identifiers. That is, there's no way to use these symbols in portable
5348 ISO C or C++ and have all your identifiers in NFC@.
5349 @option{-Wnormalized=id} suppresses the warning for these characters.
5350 It is hoped that future versions of the standards involved will correct
5351 this, which is why this option is not the default.
5353 You can switch the warning off for all characters by writing
5354 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5355 only do this if you are using some other normalization scheme (like
5356 ``D''), because otherwise you can easily create bugs that are
5357 literally impossible to see.
5359 Some characters in ISO 10646 have distinct meanings but look identical
5360 in some fonts or display methodologies, especially once formatting has
5361 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5362 LETTER N'', displays just like a regular @code{n} that has been
5363 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5364 normalization scheme to convert all these into a standard form as
5365 well, and GCC warns if your code is not in NFKC if you use
5366 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5367 about every identifier that contains the letter O because it might be
5368 confused with the digit 0, and so is not the default, but may be
5369 useful as a local coding convention if the programming environment
5370 cannot be fixed to display these characters distinctly.
5372 @item -Wno-deprecated
5373 @opindex Wno-deprecated
5374 @opindex Wdeprecated
5375 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5377 @item -Wno-deprecated-declarations
5378 @opindex Wno-deprecated-declarations
5379 @opindex Wdeprecated-declarations
5380 Do not warn about uses of functions (@pxref{Function Attributes}),
5381 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5382 Attributes}) marked as deprecated by using the @code{deprecated}
5386 @opindex Wno-overflow
5388 Do not warn about compile-time overflow in constant expressions.
5393 Warn about One Definition Rule violations during link-time optimization.
5394 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5397 @opindex Wopenm-simd
5398 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5399 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5400 option can be used to relax the cost model.
5402 @item -Woverride-init @r{(C and Objective-C only)}
5403 @opindex Woverride-init
5404 @opindex Wno-override-init
5408 Warn if an initialized field without side effects is overridden when
5409 using designated initializers (@pxref{Designated Inits, , Designated
5412 This warning is included in @option{-Wextra}. To get other
5413 @option{-Wextra} warnings without this one, use @option{-Wextra
5414 -Wno-override-init}.
5416 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5417 @opindex Woverride-init-side-effects
5418 @opindex Wno-override-init-side-effects
5419 Warn if an initialized field with side effects is overridden when
5420 using designated initializers (@pxref{Designated Inits, , Designated
5421 Initializers}). This warning is enabled by default.
5426 Warn if a structure is given the packed attribute, but the packed
5427 attribute has no effect on the layout or size of the structure.
5428 Such structures may be mis-aligned for little benefit. For
5429 instance, in this code, the variable @code{f.x} in @code{struct bar}
5430 is misaligned even though @code{struct bar} does not itself
5431 have the packed attribute:
5438 @} __attribute__((packed));
5446 @item -Wpacked-bitfield-compat
5447 @opindex Wpacked-bitfield-compat
5448 @opindex Wno-packed-bitfield-compat
5449 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5450 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5451 the change can lead to differences in the structure layout. GCC
5452 informs you when the offset of such a field has changed in GCC 4.4.
5453 For example there is no longer a 4-bit padding between field @code{a}
5454 and @code{b} in this structure:
5461 @} __attribute__ ((packed));
5464 This warning is enabled by default. Use
5465 @option{-Wno-packed-bitfield-compat} to disable this warning.
5470 Warn if padding is included in a structure, either to align an element
5471 of the structure or to align the whole structure. Sometimes when this
5472 happens it is possible to rearrange the fields of the structure to
5473 reduce the padding and so make the structure smaller.
5475 @item -Wredundant-decls
5476 @opindex Wredundant-decls
5477 @opindex Wno-redundant-decls
5478 Warn if anything is declared more than once in the same scope, even in
5479 cases where multiple declaration is valid and changes nothing.
5481 @item -Wnested-externs @r{(C and Objective-C only)}
5482 @opindex Wnested-externs
5483 @opindex Wno-nested-externs
5484 Warn if an @code{extern} declaration is encountered within a function.
5486 @item -Wno-inherited-variadic-ctor
5487 @opindex Winherited-variadic-ctor
5488 @opindex Wno-inherited-variadic-ctor
5489 Suppress warnings about use of C++11 inheriting constructors when the
5490 base class inherited from has a C variadic constructor; the warning is
5491 on by default because the ellipsis is not inherited.
5496 Warn if a function that is declared as inline cannot be inlined.
5497 Even with this option, the compiler does not warn about failures to
5498 inline functions declared in system headers.
5500 The compiler uses a variety of heuristics to determine whether or not
5501 to inline a function. For example, the compiler takes into account
5502 the size of the function being inlined and the amount of inlining
5503 that has already been done in the current function. Therefore,
5504 seemingly insignificant changes in the source program can cause the
5505 warnings produced by @option{-Winline} to appear or disappear.
5507 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5508 @opindex Wno-invalid-offsetof
5509 @opindex Winvalid-offsetof
5510 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5511 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5512 to a non-standard-layout type is undefined. In existing C++ implementations,
5513 however, @code{offsetof} typically gives meaningful results.
5514 This flag is for users who are aware that they are
5515 writing nonportable code and who have deliberately chosen to ignore the
5518 The restrictions on @code{offsetof} may be relaxed in a future version
5519 of the C++ standard.
5521 @item -Wno-int-to-pointer-cast
5522 @opindex Wno-int-to-pointer-cast
5523 @opindex Wint-to-pointer-cast
5524 Suppress warnings from casts to pointer type of an integer of a
5525 different size. In C++, casting to a pointer type of smaller size is
5526 an error. @option{Wint-to-pointer-cast} is enabled by default.
5529 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5530 @opindex Wno-pointer-to-int-cast
5531 @opindex Wpointer-to-int-cast
5532 Suppress warnings from casts from a pointer to an integer type of a
5536 @opindex Winvalid-pch
5537 @opindex Wno-invalid-pch
5538 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5539 the search path but can't be used.
5543 @opindex Wno-long-long
5544 Warn if @code{long long} type is used. This is enabled by either
5545 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5546 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5548 @item -Wvariadic-macros
5549 @opindex Wvariadic-macros
5550 @opindex Wno-variadic-macros
5551 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5552 alternate syntax is used in ISO C99 mode. This is enabled by either
5553 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5554 messages, use @option{-Wno-variadic-macros}.
5558 @opindex Wno-varargs
5559 Warn upon questionable usage of the macros used to handle variable
5560 arguments like @code{va_start}. This is default. To inhibit the
5561 warning messages, use @option{-Wno-varargs}.
5563 @item -Wvector-operation-performance
5564 @opindex Wvector-operation-performance
5565 @opindex Wno-vector-operation-performance
5566 Warn if vector operation is not implemented via SIMD capabilities of the
5567 architecture. Mainly useful for the performance tuning.
5568 Vector operation can be implemented @code{piecewise}, which means that the
5569 scalar operation is performed on every vector element;
5570 @code{in parallel}, which means that the vector operation is implemented
5571 using scalars of wider type, which normally is more performance efficient;
5572 and @code{as a single scalar}, which means that vector fits into a
5575 @item -Wno-virtual-move-assign
5576 @opindex Wvirtual-move-assign
5577 @opindex Wno-virtual-move-assign
5578 Suppress warnings about inheriting from a virtual base with a
5579 non-trivial C++11 move assignment operator. This is dangerous because
5580 if the virtual base is reachable along more than one path, it is
5581 moved multiple times, which can mean both objects end up in the
5582 moved-from state. If the move assignment operator is written to avoid
5583 moving from a moved-from object, this warning can be disabled.
5588 Warn if variable length array is used in the code.
5589 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5590 the variable length array.
5592 @item -Wvolatile-register-var
5593 @opindex Wvolatile-register-var
5594 @opindex Wno-volatile-register-var
5595 Warn if a register variable is declared volatile. The volatile
5596 modifier does not inhibit all optimizations that may eliminate reads
5597 and/or writes to register variables. This warning is enabled by
5600 @item -Wdisabled-optimization
5601 @opindex Wdisabled-optimization
5602 @opindex Wno-disabled-optimization
5603 Warn if a requested optimization pass is disabled. This warning does
5604 not generally indicate that there is anything wrong with your code; it
5605 merely indicates that GCC's optimizers are unable to handle the code
5606 effectively. Often, the problem is that your code is too big or too
5607 complex; GCC refuses to optimize programs when the optimization
5608 itself is likely to take inordinate amounts of time.
5610 @item -Wpointer-sign @r{(C and Objective-C only)}
5611 @opindex Wpointer-sign
5612 @opindex Wno-pointer-sign
5613 Warn for pointer argument passing or assignment with different signedness.
5614 This option is only supported for C and Objective-C@. It is implied by
5615 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5616 @option{-Wno-pointer-sign}.
5618 @item -Wstack-protector
5619 @opindex Wstack-protector
5620 @opindex Wno-stack-protector
5621 This option is only active when @option{-fstack-protector} is active. It
5622 warns about functions that are not protected against stack smashing.
5624 @item -Woverlength-strings
5625 @opindex Woverlength-strings
5626 @opindex Wno-overlength-strings
5627 Warn about string constants that are longer than the ``minimum
5628 maximum'' length specified in the C standard. Modern compilers
5629 generally allow string constants that are much longer than the
5630 standard's minimum limit, but very portable programs should avoid
5631 using longer strings.
5633 The limit applies @emph{after} string constant concatenation, and does
5634 not count the trailing NUL@. In C90, the limit was 509 characters; in
5635 C99, it was raised to 4095. C++98 does not specify a normative
5636 minimum maximum, so we do not diagnose overlength strings in C++@.
5638 This option is implied by @option{-Wpedantic}, and can be disabled with
5639 @option{-Wno-overlength-strings}.
5641 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5642 @opindex Wunsuffixed-float-constants
5644 Issue a warning for any floating constant that does not have
5645 a suffix. When used together with @option{-Wsystem-headers} it
5646 warns about such constants in system header files. This can be useful
5647 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5648 from the decimal floating-point extension to C99.
5650 @item -Wno-designated-init @r{(C and Objective-C only)}
5651 Suppress warnings when a positional initializer is used to initialize
5652 a structure that has been marked with the @code{designated_init}
5657 @node Debugging Options
5658 @section Options for Debugging Your Program or GCC
5659 @cindex options, debugging
5660 @cindex debugging information options
5662 GCC has various special options that are used for debugging
5663 either your program or GCC:
5668 Produce debugging information in the operating system's native format
5669 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5672 On most systems that use stabs format, @option{-g} enables use of extra
5673 debugging information that only GDB can use; this extra information
5674 makes debugging work better in GDB but probably makes other debuggers
5676 refuse to read the program. If you want to control for certain whether
5677 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5678 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5680 GCC allows you to use @option{-g} with
5681 @option{-O}. The shortcuts taken by optimized code may occasionally
5682 produce surprising results: some variables you declared may not exist
5683 at all; flow of control may briefly move where you did not expect it;
5684 some statements may not be executed because they compute constant
5685 results or their values are already at hand; some statements may
5686 execute in different places because they have been moved out of loops.
5688 Nevertheless it proves possible to debug optimized output. This makes
5689 it reasonable to use the optimizer for programs that might have bugs.
5691 The following options are useful when GCC is generated with the
5692 capability for more than one debugging format.
5695 @opindex gsplit-dwarf
5696 Separate as much dwarf debugging information as possible into a
5697 separate output file with the extension .dwo. This option allows
5698 the build system to avoid linking files with debug information. To
5699 be useful, this option requires a debugger capable of reading .dwo
5704 Produce debugging information for use by GDB@. This means to use the
5705 most expressive format available (DWARF 2, stabs, or the native format
5706 if neither of those are supported), including GDB extensions if at all
5711 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5713 @item -ggnu-pubnames
5714 @opindex ggnu-pubnames
5715 Generate .debug_pubnames and .debug_pubtypes sections in a format
5716 suitable for conversion into a GDB@ index. This option is only useful
5717 with a linker that can produce GDB@ index version 7.
5721 Produce debugging information in stabs format (if that is supported),
5722 without GDB extensions. This is the format used by DBX on most BSD
5723 systems. On MIPS, Alpha and System V Release 4 systems this option
5724 produces stabs debugging output that is not understood by DBX or SDB@.
5725 On System V Release 4 systems this option requires the GNU assembler.
5727 @item -feliminate-unused-debug-symbols
5728 @opindex feliminate-unused-debug-symbols
5729 Produce debugging information in stabs format (if that is supported),
5730 for only symbols that are actually used.
5732 @item -femit-class-debug-always
5733 @opindex femit-class-debug-always
5734 Instead of emitting debugging information for a C++ class in only one
5735 object file, emit it in all object files using the class. This option
5736 should be used only with debuggers that are unable to handle the way GCC
5737 normally emits debugging information for classes because using this
5738 option increases the size of debugging information by as much as a
5741 @item -fdebug-types-section
5742 @opindex fdebug-types-section
5743 @opindex fno-debug-types-section
5744 When using DWARF Version 4 or higher, type DIEs can be put into
5745 their own @code{.debug_types} section instead of making them part of the
5746 @code{.debug_info} section. It is more efficient to put them in a separate
5747 comdat sections since the linker can then remove duplicates.
5748 But not all DWARF consumers support @code{.debug_types} sections yet
5749 and on some objects @code{.debug_types} produces larger instead of smaller
5750 debugging information.
5754 Produce debugging information in stabs format (if that is supported),
5755 using GNU extensions understood only by the GNU debugger (GDB)@. The
5756 use of these extensions is likely to make other debuggers crash or
5757 refuse to read the program.
5761 Produce debugging information in COFF format (if that is supported).
5762 This is the format used by SDB on most System V systems prior to
5767 Produce debugging information in XCOFF format (if that is supported).
5768 This is the format used by the DBX debugger on IBM RS/6000 systems.
5772 Produce debugging information in XCOFF format (if that is supported),
5773 using GNU extensions understood only by the GNU debugger (GDB)@. The
5774 use of these extensions is likely to make other debuggers crash or
5775 refuse to read the program, and may cause assemblers other than the GNU
5776 assembler (GAS) to fail with an error.
5778 @item -gdwarf-@var{version}
5779 @opindex gdwarf-@var{version}
5780 Produce debugging information in DWARF format (if that is supported).
5781 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5782 for most targets is 4. DWARF Version 5 is only experimental.
5784 Note that with DWARF Version 2, some ports require and always
5785 use some non-conflicting DWARF 3 extensions in the unwind tables.
5787 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5788 for maximum benefit.
5790 @item -grecord-gcc-switches
5791 @opindex grecord-gcc-switches
5792 This switch causes the command-line options used to invoke the
5793 compiler that may affect code generation to be appended to the
5794 DW_AT_producer attribute in DWARF debugging information. The options
5795 are concatenated with spaces separating them from each other and from
5796 the compiler version. See also @option{-frecord-gcc-switches} for another
5797 way of storing compiler options into the object file. This is the default.
5799 @item -gno-record-gcc-switches
5800 @opindex gno-record-gcc-switches
5801 Disallow appending command-line options to the DW_AT_producer attribute
5802 in DWARF debugging information.
5804 @item -gstrict-dwarf
5805 @opindex gstrict-dwarf
5806 Disallow using extensions of later DWARF standard version than selected
5807 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5808 DWARF extensions from later standard versions is allowed.
5810 @item -gno-strict-dwarf
5811 @opindex gno-strict-dwarf
5812 Allow using extensions of later DWARF standard version than selected with
5813 @option{-gdwarf-@var{version}}.
5815 @item -gz@r{[}=@var{type}@r{]}
5817 Produce compressed debug sections in DWARF format, if that is supported.
5818 If @var{type} is not given, the default type depends on the capabilities
5819 of the assembler and linker used. @var{type} may be one of
5820 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5821 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5822 compression in traditional GNU format). If the linker doesn't support
5823 writing compressed debug sections, the option is rejected. Otherwise,
5824 if the assembler does not support them, @option{-gz} is silently ignored
5825 when producing object files.
5829 Produce debugging information in Alpha/VMS debug format (if that is
5830 supported). This is the format used by DEBUG on Alpha/VMS systems.
5833 @itemx -ggdb@var{level}
5834 @itemx -gstabs@var{level}
5835 @itemx -gcoff@var{level}
5836 @itemx -gxcoff@var{level}
5837 @itemx -gvms@var{level}
5838 Request debugging information and also use @var{level} to specify how
5839 much information. The default level is 2.
5841 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5844 Level 1 produces minimal information, enough for making backtraces in
5845 parts of the program that you don't plan to debug. This includes
5846 descriptions of functions and external variables, and line number
5847 tables, but no information about local variables.
5849 Level 3 includes extra information, such as all the macro definitions
5850 present in the program. Some debuggers support macro expansion when
5851 you use @option{-g3}.
5853 @option{-gdwarf-2} does not accept a concatenated debug level, because
5854 GCC used to support an option @option{-gdwarf} that meant to generate
5855 debug information in version 1 of the DWARF format (which is very
5856 different from version 2), and it would have been too confusing. That
5857 debug format is long obsolete, but the option cannot be changed now.
5858 Instead use an additional @option{-g@var{level}} option to change the
5859 debug level for DWARF.
5863 Turn off generation of debug info, if leaving out this option
5864 generates it, or turn it on at level 2 otherwise. The position of this
5865 argument in the command line does not matter; it takes effect after all
5866 other options are processed, and it does so only once, no matter how
5867 many times it is given. This is mainly intended to be used with
5868 @option{-fcompare-debug}.
5870 @item -fsanitize=address
5871 @opindex fsanitize=address
5872 Enable AddressSanitizer, a fast memory error detector.
5873 Memory access instructions are instrumented to detect
5874 out-of-bounds and use-after-free bugs.
5875 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
5876 more details. The run-time behavior can be influenced using the
5877 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
5878 the available options are shown at startup of the instrumended program. See
5879 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
5880 for a list of supported options.
5882 @item -fsanitize=kernel-address
5883 @opindex fsanitize=kernel-address
5884 Enable AddressSanitizer for Linux kernel.
5885 See @uref{https://github.com/google/kasan/wiki} for more details.
5887 @item -fsanitize=thread
5888 @opindex fsanitize=thread
5889 Enable ThreadSanitizer, a fast data race detector.
5890 Memory access instructions are instrumented to detect
5891 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
5892 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5893 environment variable; see
5894 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
5897 @item -fsanitize=leak
5898 @opindex fsanitize=leak
5899 Enable LeakSanitizer, a memory leak detector.
5900 This option only matters for linking of executables and if neither
5901 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5902 case the executable is linked against a library that overrides @code{malloc}
5903 and other allocator functions. See
5904 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
5905 details. The run-time behavior can be influenced using the
5906 @env{LSAN_OPTIONS} environment variable.
5908 @item -fsanitize=undefined
5909 @opindex fsanitize=undefined
5910 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5911 Various computations are instrumented to detect undefined behavior
5912 at runtime. Current suboptions are:
5916 @item -fsanitize=shift
5917 @opindex fsanitize=shift
5918 This option enables checking that the result of a shift operation is
5919 not undefined. Note that what exactly is considered undefined differs
5920 slightly between C and C++, as well as between ISO C90 and C99, etc.
5922 @item -fsanitize=integer-divide-by-zero
5923 @opindex fsanitize=integer-divide-by-zero
5924 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5926 @item -fsanitize=unreachable
5927 @opindex fsanitize=unreachable
5928 With this option, the compiler turns the @code{__builtin_unreachable}
5929 call into a diagnostics message call instead. When reaching the
5930 @code{__builtin_unreachable} call, the behavior is undefined.
5932 @item -fsanitize=vla-bound
5933 @opindex fsanitize=vla-bound
5934 This option instructs the compiler to check that the size of a variable
5935 length array is positive.
5937 @item -fsanitize=null
5938 @opindex fsanitize=null
5939 This option enables pointer checking. Particularly, the application
5940 built with this option turned on will issue an error message when it
5941 tries to dereference a NULL pointer, or if a reference (possibly an
5942 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5943 on an object pointed by a NULL pointer.
5945 @item -fsanitize=return
5946 @opindex fsanitize=return
5947 This option enables return statement checking. Programs
5948 built with this option turned on will issue an error message
5949 when the end of a non-void function is reached without actually
5950 returning a value. This option works in C++ only.
5952 @item -fsanitize=signed-integer-overflow
5953 @opindex fsanitize=signed-integer-overflow
5954 This option enables signed integer overflow checking. We check that
5955 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5956 does not overflow in the signed arithmetics. Note, integer promotion
5957 rules must be taken into account. That is, the following is not an
5960 signed char a = SCHAR_MAX;
5964 @item -fsanitize=bounds
5965 @opindex fsanitize=bounds
5966 This option enables instrumentation of array bounds. Various out of bounds
5967 accesses are detected. Flexible array members, flexible array member-like
5968 arrays, and initializers of variables with static storage are not instrumented.
5970 @item -fsanitize=bounds-strict
5971 @opindex fsanitize=bounds-strict
5972 This option enables strict instrumentation of array bounds. Most out of bounds
5973 accesses are detected, including flexible array members and flexible array
5974 member-like arrays. Initializers of variables with static storage are not
5977 @item -fsanitize=alignment
5978 @opindex fsanitize=alignment
5980 This option enables checking of alignment of pointers when they are
5981 dereferenced, or when a reference is bound to insufficiently aligned target,
5982 or when a method or constructor is invoked on insufficiently aligned object.
5984 @item -fsanitize=object-size
5985 @opindex fsanitize=object-size
5986 This option enables instrumentation of memory references using the
5987 @code{__builtin_object_size} function. Various out of bounds pointer
5988 accesses are detected.
5990 @item -fsanitize=float-divide-by-zero
5991 @opindex fsanitize=float-divide-by-zero
5992 Detect floating-point division by zero. Unlike other similar options,
5993 @option{-fsanitize=float-divide-by-zero} is not enabled by
5994 @option{-fsanitize=undefined}, since floating-point division by zero can
5995 be a legitimate way of obtaining infinities and NaNs.
5997 @item -fsanitize=float-cast-overflow
5998 @opindex fsanitize=float-cast-overflow
5999 This option enables floating-point type to integer conversion checking.
6000 We check that the result of the conversion does not overflow.
6001 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
6002 not enabled by @option{-fsanitize=undefined}.
6003 This option does not work well with @code{FE_INVALID} exceptions enabled.
6005 @item -fsanitize=nonnull-attribute
6006 @opindex fsanitize=nonnull-attribute
6008 This option enables instrumentation of calls, checking whether null values
6009 are not passed to arguments marked as requiring a non-null value by the
6010 @code{nonnull} function attribute.
6012 @item -fsanitize=returns-nonnull-attribute
6013 @opindex fsanitize=returns-nonnull-attribute
6015 This option enables instrumentation of return statements in functions
6016 marked with @code{returns_nonnull} function attribute, to detect returning
6017 of null values from such functions.
6019 @item -fsanitize=bool
6020 @opindex fsanitize=bool
6022 This option enables instrumentation of loads from bool. If a value other
6023 than 0/1 is loaded, a run-time error is issued.
6025 @item -fsanitize=enum
6026 @opindex fsanitize=enum
6028 This option enables instrumentation of loads from an enum type. If
6029 a value outside the range of values for the enum type is loaded,
6030 a run-time error is issued.
6032 @item -fsanitize=vptr
6033 @opindex fsanitize=vptr
6035 This option enables instrumentation of C++ member function calls, member
6036 accesses and some conversions between pointers to base and derived classes,
6037 to verify the referenced object has the correct dynamic type.
6041 While @option{-ftrapv} causes traps for signed overflows to be emitted,
6042 @option{-fsanitize=undefined} gives a diagnostic message.
6043 This currently works only for the C family of languages.
6045 @item -fno-sanitize=all
6046 @opindex fno-sanitize=all
6048 This option disables all previously enabled sanitizers.
6049 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
6052 @item -fasan-shadow-offset=@var{number}
6053 @opindex fasan-shadow-offset
6054 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
6055 It is useful for experimenting with different shadow memory layouts in
6056 Kernel AddressSanitizer.
6058 @item -fsanitize-sections=@var{s1},@var{s2},...
6059 @opindex fsanitize-sections
6060 Sanitize global variables in selected user-defined sections. @var{si} may
6063 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
6064 @opindex fsanitize-recover
6065 @opindex fno-sanitize-recover
6066 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
6067 mentioned in comma-separated list of @var{opts}. Enabling this option
6068 for a sanitizer component causes it to attempt to continue
6069 running the program as if no error happened. This means multiple
6070 runtime errors can be reported in a single program run, and the exit
6071 code of the program may indicate success even when errors
6072 have been reported. The @option{-fno-sanitize-recover=} option
6073 can be used to alter
6074 this behavior: only the first detected error is reported
6075 and program then exits with a non-zero exit code.
6077 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
6078 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
6079 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
6080 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
6081 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
6082 accepted, the former enables recovery for all sanitizers that support it,
6083 the latter disables recovery for all sanitizers that support it.
6085 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
6087 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6090 Similarly @option{-fno-sanitize-recover} is equivalent to
6092 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6095 @item -fsanitize-undefined-trap-on-error
6096 @opindex fsanitize-undefined-trap-on-error
6097 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
6098 report undefined behavior using @code{__builtin_trap} rather than
6099 a @code{libubsan} library routine. The advantage of this is that the
6100 @code{libubsan} library is not needed and is not linked in, so this
6101 is usable even in freestanding environments.
6103 @item -fcheck-pointer-bounds
6104 @opindex fcheck-pointer-bounds
6105 @opindex fno-check-pointer-bounds
6106 @cindex Pointer Bounds Checker options
6107 Enable Pointer Bounds Checker instrumentation. Each memory reference
6108 is instrumented with checks of the pointer used for memory access against
6109 bounds associated with that pointer.
6112 is only an implementation for Intel MPX available, thus x86 target
6113 and @option{-mmpx} are required to enable this feature.
6114 MPX-based instrumentation requires
6115 a runtime library to enable MPX in hardware and handle bounds
6116 violation signals. By default when @option{-fcheck-pointer-bounds}
6117 and @option{-mmpx} options are used to link a program, the GCC driver
6118 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
6119 library. It also passes '-z bndplt' to a linker in case it supports this
6120 option (which is checked on libmpx configuration). Note that old versions
6121 of linker may ignore option. Gold linker doesn't support '-z bndplt'
6122 option. With no '-z bndplt' support in linker all calls to dynamic libraries
6123 lose passed bounds reducing overall protection level. It's highly
6124 recommended to use linker with '-z bndplt' support. In case such linker
6125 is not available it is adviced to always use @option{-static-libmpxwrappers}
6126 for better protection level or use @option{-static} to completely avoid
6127 external calls to dynamic libraries. MPX-based instrumentation
6128 may be used for debugging and also may be included in production code
6129 to increase program security. Depending on usage, you may
6130 have different requirements for the runtime library. The current version
6131 of the MPX runtime library is more oriented for use as a debugging
6132 tool. MPX runtime library usage implies @option{-lpthread}. See
6133 also @option{-static-libmpx}. The runtime library behavior can be
6134 influenced using various @env{CHKP_RT_*} environment variables. See
6135 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
6138 Generated instrumentation may be controlled by various
6139 @option{-fchkp-*} options and by the @code{bnd_variable_size}
6140 structure field attribute (@pxref{Type Attributes}) and
6141 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
6142 (@pxref{Function Attributes}). GCC also provides a number of built-in
6143 functions for controlling the Pointer Bounds Checker. @xref{Pointer
6144 Bounds Checker builtins}, for more information.
6146 @item -fchkp-check-incomplete-type
6147 @opindex fchkp-check-incomplete-type
6148 @opindex fno-chkp-check-incomplete-type
6149 Generate pointer bounds checks for variables with incomplete type.
6152 @item -fchkp-narrow-bounds
6153 @opindex fchkp-narrow-bounds
6154 @opindex fno-chkp-narrow-bounds
6155 Controls bounds used by Pointer Bounds Checker for pointers to object
6156 fields. If narrowing is enabled then field bounds are used. Otherwise
6157 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
6158 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
6160 @item -fchkp-first-field-has-own-bounds
6161 @opindex fchkp-first-field-has-own-bounds
6162 @opindex fno-chkp-first-field-has-own-bounds
6163 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
6164 first field in the structure. By default a pointer to the first field has
6165 the same bounds as a pointer to the whole structure.
6167 @item -fchkp-narrow-to-innermost-array
6168 @opindex fchkp-narrow-to-innermost-array
6169 @opindex fno-chkp-narrow-to-innermost-array
6170 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
6171 case of nested static array access. By default this option is disabled and
6172 bounds of the outermost array are used.
6174 @item -fchkp-optimize
6175 @opindex fchkp-optimize
6176 @opindex fno-chkp-optimize
6177 Enables Pointer Bounds Checker optimizations. Enabled by default at
6178 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
6180 @item -fchkp-use-fast-string-functions
6181 @opindex fchkp-use-fast-string-functions
6182 @opindex fno-chkp-use-fast-string-functions
6183 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
6184 by Pointer Bounds Checker. Disabled by default.
6186 @item -fchkp-use-nochk-string-functions
6187 @opindex fchkp-use-nochk-string-functions
6188 @opindex fno-chkp-use-nochk-string-functions
6189 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
6190 by Pointer Bounds Checker. Disabled by default.
6192 @item -fchkp-use-static-bounds
6193 @opindex fchkp-use-static-bounds
6194 @opindex fno-chkp-use-static-bounds
6195 Allow Pointer Bounds Checker to generate static bounds holding
6196 bounds of static variables. Enabled by default.
6198 @item -fchkp-use-static-const-bounds
6199 @opindex fchkp-use-static-const-bounds
6200 @opindex fno-chkp-use-static-const-bounds
6201 Use statically-initialized bounds for constant bounds instead of
6202 generating them each time they are required. By default enabled when
6203 @option{-fchkp-use-static-bounds} is enabled.
6205 @item -fchkp-treat-zero-dynamic-size-as-infinite
6206 @opindex fchkp-treat-zero-dynamic-size-as-infinite
6207 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
6208 With this option, objects with incomplete type whose
6209 dynamically-obtained size is zero are treated as having infinite size
6210 instead by Pointer Bounds
6211 Checker. This option may be helpful if a program is linked with a library
6212 missing size information for some symbols. Disabled by default.
6214 @item -fchkp-check-read
6215 @opindex fchkp-check-read
6216 @opindex fno-chkp-check-read
6217 Instructs Pointer Bounds Checker to generate checks for all read
6218 accesses to memory. Enabled by default.
6220 @item -fchkp-check-write
6221 @opindex fchkp-check-write
6222 @opindex fno-chkp-check-write
6223 Instructs Pointer Bounds Checker to generate checks for all write
6224 accesses to memory. Enabled by default.
6226 @item -fchkp-store-bounds
6227 @opindex fchkp-store-bounds
6228 @opindex fno-chkp-store-bounds
6229 Instructs Pointer Bounds Checker to generate bounds stores for
6230 pointer writes. Enabled by default.
6232 @item -fchkp-instrument-calls
6233 @opindex fchkp-instrument-calls
6234 @opindex fno-chkp-instrument-calls
6235 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
6238 @item -fchkp-instrument-marked-only
6239 @opindex fchkp-instrument-marked-only
6240 @opindex fno-chkp-instrument-marked-only
6241 Instructs Pointer Bounds Checker to instrument only functions
6242 marked with the @code{bnd_instrument} attribute
6243 (@pxref{Function Attributes}). Disabled by default.
6245 @item -fchkp-use-wrappers
6246 @opindex fchkp-use-wrappers
6247 @opindex fno-chkp-use-wrappers
6248 Allows Pointer Bounds Checker to replace calls to built-in functions
6249 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6250 is used to link a program, the GCC driver automatically links
6251 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6254 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6255 @opindex fdump-final-insns
6256 Dump the final internal representation (RTL) to @var{file}. If the
6257 optional argument is omitted (or if @var{file} is @code{.}), the name
6258 of the dump file is determined by appending @code{.gkd} to the
6259 compilation output file name.
6261 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6262 @opindex fcompare-debug
6263 @opindex fno-compare-debug
6264 If no error occurs during compilation, run the compiler a second time,
6265 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6266 passed to the second compilation. Dump the final internal
6267 representation in both compilations, and print an error if they differ.
6269 If the equal sign is omitted, the default @option{-gtoggle} is used.
6271 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6272 and nonzero, implicitly enables @option{-fcompare-debug}. If
6273 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6274 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6277 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6278 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6279 of the final representation and the second compilation, preventing even
6280 @env{GCC_COMPARE_DEBUG} from taking effect.
6282 To verify full coverage during @option{-fcompare-debug} testing, set
6283 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6284 which GCC rejects as an invalid option in any actual compilation
6285 (rather than preprocessing, assembly or linking). To get just a
6286 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6287 not overridden} will do.
6289 @item -fcompare-debug-second
6290 @opindex fcompare-debug-second
6291 This option is implicitly passed to the compiler for the second
6292 compilation requested by @option{-fcompare-debug}, along with options to
6293 silence warnings, and omitting other options that would cause
6294 side-effect compiler outputs to files or to the standard output. Dump
6295 files and preserved temporary files are renamed so as to contain the
6296 @code{.gk} additional extension during the second compilation, to avoid
6297 overwriting those generated by the first.
6299 When this option is passed to the compiler driver, it causes the
6300 @emph{first} compilation to be skipped, which makes it useful for little
6301 other than debugging the compiler proper.
6303 @item -feliminate-dwarf2-dups
6304 @opindex feliminate-dwarf2-dups
6305 Compress DWARF 2 debugging information by eliminating duplicated
6306 information about each symbol. This option only makes sense when
6307 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6309 @item -femit-struct-debug-baseonly
6310 @opindex femit-struct-debug-baseonly
6311 Emit debug information for struct-like types
6312 only when the base name of the compilation source file
6313 matches the base name of file in which the struct is defined.
6315 This option substantially reduces the size of debugging information,
6316 but at significant potential loss in type information to the debugger.
6317 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6318 See @option{-femit-struct-debug-detailed} for more detailed control.
6320 This option works only with DWARF 2.
6322 @item -femit-struct-debug-reduced
6323 @opindex femit-struct-debug-reduced
6324 Emit debug information for struct-like types
6325 only when the base name of the compilation source file
6326 matches the base name of file in which the type is defined,
6327 unless the struct is a template or defined in a system header.
6329 This option significantly reduces the size of debugging information,
6330 with some potential loss in type information to the debugger.
6331 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6332 See @option{-femit-struct-debug-detailed} for more detailed control.
6334 This option works only with DWARF 2.
6336 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6337 @opindex femit-struct-debug-detailed
6338 Specify the struct-like types
6339 for which the compiler generates debug information.
6340 The intent is to reduce duplicate struct debug information
6341 between different object files within the same program.
6343 This option is a detailed version of
6344 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6345 which serves for most needs.
6347 A specification has the syntax@*
6348 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6350 The optional first word limits the specification to
6351 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6352 A struct type is used directly when it is the type of a variable, member.
6353 Indirect uses arise through pointers to structs.
6354 That is, when use of an incomplete struct is valid, the use is indirect.
6356 @samp{struct one direct; struct two * indirect;}.
6358 The optional second word limits the specification to
6359 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6360 Generic structs are a bit complicated to explain.
6361 For C++, these are non-explicit specializations of template classes,
6362 or non-template classes within the above.
6363 Other programming languages have generics,
6364 but @option{-femit-struct-debug-detailed} does not yet implement them.
6366 The third word specifies the source files for those
6367 structs for which the compiler should emit debug information.
6368 The values @samp{none} and @samp{any} have the normal meaning.
6369 The value @samp{base} means that
6370 the base of name of the file in which the type declaration appears
6371 must match the base of the name of the main compilation file.
6372 In practice, this means that when compiling @file{foo.c}, debug information
6373 is generated for types declared in that file and @file{foo.h},
6374 but not other header files.
6375 The value @samp{sys} means those types satisfying @samp{base}
6376 or declared in system or compiler headers.
6378 You may need to experiment to determine the best settings for your application.
6380 The default is @option{-femit-struct-debug-detailed=all}.
6382 This option works only with DWARF 2.
6384 @item -fno-merge-debug-strings
6385 @opindex fmerge-debug-strings
6386 @opindex fno-merge-debug-strings
6387 Direct the linker to not merge together strings in the debugging
6388 information that are identical in different object files. Merging is
6389 not supported by all assemblers or linkers. Merging decreases the size
6390 of the debug information in the output file at the cost of increasing
6391 link processing time. Merging is enabled by default.
6393 @item -fdebug-prefix-map=@var{old}=@var{new}
6394 @opindex fdebug-prefix-map
6395 When compiling files in directory @file{@var{old}}, record debugging
6396 information describing them as in @file{@var{new}} instead.
6398 @item -fno-dwarf2-cfi-asm
6399 @opindex fdwarf2-cfi-asm
6400 @opindex fno-dwarf2-cfi-asm
6401 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6402 instead of using GAS @code{.cfi_*} directives.
6404 @cindex @command{prof}
6407 Generate extra code to write profile information suitable for the
6408 analysis program @command{prof}. You must use this option when compiling
6409 the source files you want data about, and you must also use it when
6412 @cindex @command{gprof}
6415 Generate extra code to write profile information suitable for the
6416 analysis program @command{gprof}. You must use this option when compiling
6417 the source files you want data about, and you must also use it when
6422 Makes the compiler print out each function name as it is compiled, and
6423 print some statistics about each pass when it finishes.
6426 @opindex ftime-report
6427 Makes the compiler print some statistics about the time consumed by each
6428 pass when it finishes.
6431 @opindex fmem-report
6432 Makes the compiler print some statistics about permanent memory
6433 allocation when it finishes.
6435 @item -fmem-report-wpa
6436 @opindex fmem-report-wpa
6437 Makes the compiler print some statistics about permanent memory
6438 allocation for the WPA phase only.
6440 @item -fpre-ipa-mem-report
6441 @opindex fpre-ipa-mem-report
6442 @item -fpost-ipa-mem-report
6443 @opindex fpost-ipa-mem-report
6444 Makes the compiler print some statistics about permanent memory
6445 allocation before or after interprocedural optimization.
6447 @item -fprofile-report
6448 @opindex fprofile-report
6449 Makes the compiler print some statistics about consistency of the
6450 (estimated) profile and effect of individual passes.
6453 @opindex fstack-usage
6454 Makes the compiler output stack usage information for the program, on a
6455 per-function basis. The filename for the dump is made by appending
6456 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6457 the output file, if explicitly specified and it is not an executable,
6458 otherwise it is the basename of the source file. An entry is made up
6463 The name of the function.
6467 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6470 The qualifier @code{static} means that the function manipulates the stack
6471 statically: a fixed number of bytes are allocated for the frame on function
6472 entry and released on function exit; no stack adjustments are otherwise made
6473 in the function. The second field is this fixed number of bytes.
6475 The qualifier @code{dynamic} means that the function manipulates the stack
6476 dynamically: in addition to the static allocation described above, stack
6477 adjustments are made in the body of the function, for example to push/pop
6478 arguments around function calls. If the qualifier @code{bounded} is also
6479 present, the amount of these adjustments is bounded at compile time and
6480 the second field is an upper bound of the total amount of stack used by
6481 the function. If it is not present, the amount of these adjustments is
6482 not bounded at compile time and the second field only represents the
6485 @item -fprofile-arcs
6486 @opindex fprofile-arcs
6487 Add code so that program flow @dfn{arcs} are instrumented. During
6488 execution the program records how many times each branch and call is
6489 executed and how many times it is taken or returns. When the compiled
6490 program exits it saves this data to a file called
6491 @file{@var{auxname}.gcda} for each source file. The data may be used for
6492 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6493 test coverage analysis (@option{-ftest-coverage}). Each object file's
6494 @var{auxname} is generated from the name of the output file, if
6495 explicitly specified and it is not the final executable, otherwise it is
6496 the basename of the source file. In both cases any suffix is removed
6497 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6498 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6499 @xref{Cross-profiling}.
6501 @cindex @command{gcov}
6505 This option is used to compile and link code instrumented for coverage
6506 analysis. The option is a synonym for @option{-fprofile-arcs}
6507 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6508 linking). See the documentation for those options for more details.
6513 Compile the source files with @option{-fprofile-arcs} plus optimization
6514 and code generation options. For test coverage analysis, use the
6515 additional @option{-ftest-coverage} option. You do not need to profile
6516 every source file in a program.
6519 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6520 (the latter implies the former).
6523 Run the program on a representative workload to generate the arc profile
6524 information. This may be repeated any number of times. You can run
6525 concurrent instances of your program, and provided that the file system
6526 supports locking, the data files will be correctly updated. Also
6527 @code{fork} calls are detected and correctly handled (double counting
6531 For profile-directed optimizations, compile the source files again with
6532 the same optimization and code generation options plus
6533 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6534 Control Optimization}).
6537 For test coverage analysis, use @command{gcov} to produce human readable
6538 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6539 @command{gcov} documentation for further information.
6543 With @option{-fprofile-arcs}, for each function of your program GCC
6544 creates a program flow graph, then finds a spanning tree for the graph.
6545 Only arcs that are not on the spanning tree have to be instrumented: the
6546 compiler adds code to count the number of times that these arcs are
6547 executed. When an arc is the only exit or only entrance to a block, the
6548 instrumentation code can be added to the block; otherwise, a new basic
6549 block must be created to hold the instrumentation code.
6552 @item -ftest-coverage
6553 @opindex ftest-coverage
6554 Produce a notes file that the @command{gcov} code-coverage utility
6555 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6556 show program coverage. Each source file's note file is called
6557 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6558 above for a description of @var{auxname} and instructions on how to
6559 generate test coverage data. Coverage data matches the source files
6560 more closely if you do not optimize.
6562 @item -fdbg-cnt-list
6563 @opindex fdbg-cnt-list
6564 Print the name and the counter upper bound for all debug counters.
6567 @item -fdbg-cnt=@var{counter-value-list}
6569 Set the internal debug counter upper bound. @var{counter-value-list}
6570 is a comma-separated list of @var{name}:@var{value} pairs
6571 which sets the upper bound of each debug counter @var{name} to @var{value}.
6572 All debug counters have the initial upper bound of @code{UINT_MAX};
6573 thus @code{dbg_cnt} returns true always unless the upper bound
6574 is set by this option.
6575 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6576 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6578 @item -fenable-@var{kind}-@var{pass}
6579 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6583 This is a set of options that are used to explicitly disable/enable
6584 optimization passes. These options are intended for use for debugging GCC.
6585 Compiler users should use regular options for enabling/disabling
6590 @item -fdisable-ipa-@var{pass}
6591 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6592 statically invoked in the compiler multiple times, the pass name should be
6593 appended with a sequential number starting from 1.
6595 @item -fdisable-rtl-@var{pass}
6596 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6597 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6598 statically invoked in the compiler multiple times, the pass name should be
6599 appended with a sequential number starting from 1. @var{range-list} is a
6600 comma-separated list of function ranges or assembler names. Each range is a number
6601 pair separated by a colon. The range is inclusive in both ends. If the range
6602 is trivial, the number pair can be simplified as a single number. If the
6603 function's call graph node's @var{uid} falls within one of the specified ranges,
6604 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6605 function header of a dump file, and the pass names can be dumped by using
6606 option @option{-fdump-passes}.
6608 @item -fdisable-tree-@var{pass}
6609 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6610 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6613 @item -fenable-ipa-@var{pass}
6614 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6615 statically invoked in the compiler multiple times, the pass name should be
6616 appended with a sequential number starting from 1.
6618 @item -fenable-rtl-@var{pass}
6619 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6620 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6621 description and examples.
6623 @item -fenable-tree-@var{pass}
6624 @itemx -fenable-tree-@var{pass}=@var{range-list}
6625 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6626 of option arguments.
6630 Here are some examples showing uses of these options.
6634 # disable ccp1 for all functions
6636 # disable complete unroll for function whose cgraph node uid is 1
6637 -fenable-tree-cunroll=1
6638 # disable gcse2 for functions at the following ranges [1,1],
6639 # [300,400], and [400,1000]
6640 # disable gcse2 for functions foo and foo2
6641 -fdisable-rtl-gcse2=foo,foo2
6642 # disable early inlining
6643 -fdisable-tree-einline
6644 # disable ipa inlining
6645 -fdisable-ipa-inline
6646 # enable tree full unroll
6647 -fenable-tree-unroll
6651 @item -d@var{letters}
6652 @itemx -fdump-rtl-@var{pass}
6653 @itemx -fdump-rtl-@var{pass}=@var{filename}
6655 @opindex fdump-rtl-@var{pass}
6656 Says to make debugging dumps during compilation at times specified by
6657 @var{letters}. This is used for debugging the RTL-based passes of the
6658 compiler. The file names for most of the dumps are made by appending
6659 a pass number and a word to the @var{dumpname}, and the files are
6660 created in the directory of the output file. In case of
6661 @option{=@var{filename}} option, the dump is output on the given file
6662 instead of the pass numbered dump files. Note that the pass number is
6663 computed statically as passes get registered into the pass manager.
6664 Thus the numbering is not related to the dynamic order of execution of
6665 passes. In particular, a pass installed by a plugin could have a
6666 number over 200 even if it executed quite early. @var{dumpname} is
6667 generated from the name of the output file, if explicitly specified
6668 and it is not an executable, otherwise it is the basename of the
6669 source file. These switches may have different effects when
6670 @option{-E} is used for preprocessing.
6672 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6673 @option{-d} option @var{letters}. Here are the possible
6674 letters for use in @var{pass} and @var{letters}, and their meanings:
6678 @item -fdump-rtl-alignments
6679 @opindex fdump-rtl-alignments
6680 Dump after branch alignments have been computed.
6682 @item -fdump-rtl-asmcons
6683 @opindex fdump-rtl-asmcons
6684 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6686 @item -fdump-rtl-auto_inc_dec
6687 @opindex fdump-rtl-auto_inc_dec
6688 Dump after auto-inc-dec discovery. This pass is only run on
6689 architectures that have auto inc or auto dec instructions.
6691 @item -fdump-rtl-barriers
6692 @opindex fdump-rtl-barriers
6693 Dump after cleaning up the barrier instructions.
6695 @item -fdump-rtl-bbpart
6696 @opindex fdump-rtl-bbpart
6697 Dump after partitioning hot and cold basic blocks.
6699 @item -fdump-rtl-bbro
6700 @opindex fdump-rtl-bbro
6701 Dump after block reordering.
6703 @item -fdump-rtl-btl1
6704 @itemx -fdump-rtl-btl2
6705 @opindex fdump-rtl-btl2
6706 @opindex fdump-rtl-btl2
6707 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6708 after the two branch
6709 target load optimization passes.
6711 @item -fdump-rtl-bypass
6712 @opindex fdump-rtl-bypass
6713 Dump after jump bypassing and control flow optimizations.
6715 @item -fdump-rtl-combine
6716 @opindex fdump-rtl-combine
6717 Dump after the RTL instruction combination pass.
6719 @item -fdump-rtl-compgotos
6720 @opindex fdump-rtl-compgotos
6721 Dump after duplicating the computed gotos.
6723 @item -fdump-rtl-ce1
6724 @itemx -fdump-rtl-ce2
6725 @itemx -fdump-rtl-ce3
6726 @opindex fdump-rtl-ce1
6727 @opindex fdump-rtl-ce2
6728 @opindex fdump-rtl-ce3
6729 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6730 @option{-fdump-rtl-ce3} enable dumping after the three
6731 if conversion passes.
6733 @item -fdump-rtl-cprop_hardreg
6734 @opindex fdump-rtl-cprop_hardreg
6735 Dump after hard register copy propagation.
6737 @item -fdump-rtl-csa
6738 @opindex fdump-rtl-csa
6739 Dump after combining stack adjustments.
6741 @item -fdump-rtl-cse1
6742 @itemx -fdump-rtl-cse2
6743 @opindex fdump-rtl-cse1
6744 @opindex fdump-rtl-cse2
6745 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6746 the two common subexpression elimination passes.
6748 @item -fdump-rtl-dce
6749 @opindex fdump-rtl-dce
6750 Dump after the standalone dead code elimination passes.
6752 @item -fdump-rtl-dbr
6753 @opindex fdump-rtl-dbr
6754 Dump after delayed branch scheduling.
6756 @item -fdump-rtl-dce1
6757 @itemx -fdump-rtl-dce2
6758 @opindex fdump-rtl-dce1
6759 @opindex fdump-rtl-dce2
6760 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6761 the two dead store elimination passes.
6764 @opindex fdump-rtl-eh
6765 Dump after finalization of EH handling code.
6767 @item -fdump-rtl-eh_ranges
6768 @opindex fdump-rtl-eh_ranges
6769 Dump after conversion of EH handling range regions.
6771 @item -fdump-rtl-expand
6772 @opindex fdump-rtl-expand
6773 Dump after RTL generation.
6775 @item -fdump-rtl-fwprop1
6776 @itemx -fdump-rtl-fwprop2
6777 @opindex fdump-rtl-fwprop1
6778 @opindex fdump-rtl-fwprop2
6779 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6780 dumping after the two forward propagation passes.
6782 @item -fdump-rtl-gcse1
6783 @itemx -fdump-rtl-gcse2
6784 @opindex fdump-rtl-gcse1
6785 @opindex fdump-rtl-gcse2
6786 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6787 after global common subexpression elimination.
6789 @item -fdump-rtl-init-regs
6790 @opindex fdump-rtl-init-regs
6791 Dump after the initialization of the registers.
6793 @item -fdump-rtl-initvals
6794 @opindex fdump-rtl-initvals
6795 Dump after the computation of the initial value sets.
6797 @item -fdump-rtl-into_cfglayout
6798 @opindex fdump-rtl-into_cfglayout
6799 Dump after converting to cfglayout mode.
6801 @item -fdump-rtl-ira
6802 @opindex fdump-rtl-ira
6803 Dump after iterated register allocation.
6805 @item -fdump-rtl-jump
6806 @opindex fdump-rtl-jump
6807 Dump after the second jump optimization.
6809 @item -fdump-rtl-loop2
6810 @opindex fdump-rtl-loop2
6811 @option{-fdump-rtl-loop2} enables dumping after the rtl
6812 loop optimization passes.
6814 @item -fdump-rtl-mach
6815 @opindex fdump-rtl-mach
6816 Dump after performing the machine dependent reorganization pass, if that
6819 @item -fdump-rtl-mode_sw
6820 @opindex fdump-rtl-mode_sw
6821 Dump after removing redundant mode switches.
6823 @item -fdump-rtl-rnreg
6824 @opindex fdump-rtl-rnreg
6825 Dump after register renumbering.
6827 @item -fdump-rtl-outof_cfglayout
6828 @opindex fdump-rtl-outof_cfglayout
6829 Dump after converting from cfglayout mode.
6831 @item -fdump-rtl-peephole2
6832 @opindex fdump-rtl-peephole2
6833 Dump after the peephole pass.
6835 @item -fdump-rtl-postreload
6836 @opindex fdump-rtl-postreload
6837 Dump after post-reload optimizations.
6839 @item -fdump-rtl-pro_and_epilogue
6840 @opindex fdump-rtl-pro_and_epilogue
6841 Dump after generating the function prologues and epilogues.
6843 @item -fdump-rtl-sched1
6844 @itemx -fdump-rtl-sched2
6845 @opindex fdump-rtl-sched1
6846 @opindex fdump-rtl-sched2
6847 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6848 after the basic block scheduling passes.
6850 @item -fdump-rtl-ree
6851 @opindex fdump-rtl-ree
6852 Dump after sign/zero extension elimination.
6854 @item -fdump-rtl-seqabstr
6855 @opindex fdump-rtl-seqabstr
6856 Dump after common sequence discovery.
6858 @item -fdump-rtl-shorten
6859 @opindex fdump-rtl-shorten
6860 Dump after shortening branches.
6862 @item -fdump-rtl-sibling
6863 @opindex fdump-rtl-sibling
6864 Dump after sibling call optimizations.
6866 @item -fdump-rtl-split1
6867 @itemx -fdump-rtl-split2
6868 @itemx -fdump-rtl-split3
6869 @itemx -fdump-rtl-split4
6870 @itemx -fdump-rtl-split5
6871 @opindex fdump-rtl-split1
6872 @opindex fdump-rtl-split2
6873 @opindex fdump-rtl-split3
6874 @opindex fdump-rtl-split4
6875 @opindex fdump-rtl-split5
6876 These options enable dumping after five rounds of
6877 instruction splitting.
6879 @item -fdump-rtl-sms
6880 @opindex fdump-rtl-sms
6881 Dump after modulo scheduling. This pass is only run on some
6884 @item -fdump-rtl-stack
6885 @opindex fdump-rtl-stack
6886 Dump after conversion from GCC's ``flat register file'' registers to the
6887 x87's stack-like registers. This pass is only run on x86 variants.
6889 @item -fdump-rtl-subreg1
6890 @itemx -fdump-rtl-subreg2
6891 @opindex fdump-rtl-subreg1
6892 @opindex fdump-rtl-subreg2
6893 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6894 the two subreg expansion passes.
6896 @item -fdump-rtl-unshare
6897 @opindex fdump-rtl-unshare
6898 Dump after all rtl has been unshared.
6900 @item -fdump-rtl-vartrack
6901 @opindex fdump-rtl-vartrack
6902 Dump after variable tracking.
6904 @item -fdump-rtl-vregs
6905 @opindex fdump-rtl-vregs
6906 Dump after converting virtual registers to hard registers.
6908 @item -fdump-rtl-web
6909 @opindex fdump-rtl-web
6910 Dump after live range splitting.
6912 @item -fdump-rtl-regclass
6913 @itemx -fdump-rtl-subregs_of_mode_init
6914 @itemx -fdump-rtl-subregs_of_mode_finish
6915 @itemx -fdump-rtl-dfinit
6916 @itemx -fdump-rtl-dfinish
6917 @opindex fdump-rtl-regclass
6918 @opindex fdump-rtl-subregs_of_mode_init
6919 @opindex fdump-rtl-subregs_of_mode_finish
6920 @opindex fdump-rtl-dfinit
6921 @opindex fdump-rtl-dfinish
6922 These dumps are defined but always produce empty files.
6925 @itemx -fdump-rtl-all
6927 @opindex fdump-rtl-all
6928 Produce all the dumps listed above.
6932 Annotate the assembler output with miscellaneous debugging information.
6936 Dump all macro definitions, at the end of preprocessing, in addition to
6941 Produce a core dump whenever an error occurs.
6945 Annotate the assembler output with a comment indicating which
6946 pattern and alternative is used. The length of each instruction is
6951 Dump the RTL in the assembler output as a comment before each instruction.
6952 Also turns on @option{-dp} annotation.
6956 Just generate RTL for a function instead of compiling it. Usually used
6957 with @option{-fdump-rtl-expand}.
6961 @opindex fdump-noaddr
6962 When doing debugging dumps, suppress address output. This makes it more
6963 feasible to use diff on debugging dumps for compiler invocations with
6964 different compiler binaries and/or different
6965 text / bss / data / heap / stack / dso start locations.
6968 @opindex freport-bug
6969 Collect and dump debug information into temporary file if ICE in C/C++
6972 @item -fdump-unnumbered
6973 @opindex fdump-unnumbered
6974 When doing debugging dumps, suppress instruction numbers and address output.
6975 This makes it more feasible to use diff on debugging dumps for compiler
6976 invocations with different options, in particular with and without
6979 @item -fdump-unnumbered-links
6980 @opindex fdump-unnumbered-links
6981 When doing debugging dumps (see @option{-d} option above), suppress
6982 instruction numbers for the links to the previous and next instructions
6985 @item -fdump-translation-unit @r{(C++ only)}
6986 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6987 @opindex fdump-translation-unit
6988 Dump a representation of the tree structure for the entire translation
6989 unit to a file. The file name is made by appending @file{.tu} to the
6990 source file name, and the file is created in the same directory as the
6991 output file. If the @samp{-@var{options}} form is used, @var{options}
6992 controls the details of the dump as described for the
6993 @option{-fdump-tree} options.
6995 @item -fdump-class-hierarchy @r{(C++ only)}
6996 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6997 @opindex fdump-class-hierarchy
6998 Dump a representation of each class's hierarchy and virtual function
6999 table layout to a file. The file name is made by appending
7000 @file{.class} to the source file name, and the file is created in the
7001 same directory as the output file. If the @samp{-@var{options}} form
7002 is used, @var{options} controls the details of the dump as described
7003 for the @option{-fdump-tree} options.
7005 @item -fdump-ipa-@var{switch}
7007 Control the dumping at various stages of inter-procedural analysis
7008 language tree to a file. The file name is generated by appending a
7009 switch specific suffix to the source file name, and the file is created
7010 in the same directory as the output file. The following dumps are
7015 Enables all inter-procedural analysis dumps.
7018 Dumps information about call-graph optimization, unused function removal,
7019 and inlining decisions.
7022 Dump after function inlining.
7027 @opindex fdump-passes
7028 Dump the list of optimization passes that are turned on and off by
7029 the current command-line options.
7031 @item -fdump-statistics-@var{option}
7032 @opindex fdump-statistics
7033 Enable and control dumping of pass statistics in a separate file. The
7034 file name is generated by appending a suffix ending in
7035 @samp{.statistics} to the source file name, and the file is created in
7036 the same directory as the output file. If the @samp{-@var{option}}
7037 form is used, @samp{-stats} causes counters to be summed over the
7038 whole compilation unit while @samp{-details} dumps every event as
7039 the passes generate them. The default with no option is to sum
7040 counters for each function compiled.
7042 @item -fdump-tree-@var{switch}
7043 @itemx -fdump-tree-@var{switch}-@var{options}
7044 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
7046 Control the dumping at various stages of processing the intermediate
7047 language tree to a file. The file name is generated by appending a
7048 switch-specific suffix to the source file name, and the file is
7049 created in the same directory as the output file. In case of
7050 @option{=@var{filename}} option, the dump is output on the given file
7051 instead of the auto named dump files. If the @samp{-@var{options}}
7052 form is used, @var{options} is a list of @samp{-} separated options
7053 which control the details of the dump. Not all options are applicable
7054 to all dumps; those that are not meaningful are ignored. The
7055 following options are available
7059 Print the address of each node. Usually this is not meaningful as it
7060 changes according to the environment and source file. Its primary use
7061 is for tying up a dump file with a debug environment.
7063 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
7064 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
7065 use working backward from mangled names in the assembly file.
7067 When dumping front-end intermediate representations, inhibit dumping
7068 of members of a scope or body of a function merely because that scope
7069 has been reached. Only dump such items when they are directly reachable
7072 When dumping pretty-printed trees, this option inhibits dumping the
7073 bodies of control structures.
7075 When dumping RTL, print the RTL in slim (condensed) form instead of
7076 the default LISP-like representation.
7078 Print a raw representation of the tree. By default, trees are
7079 pretty-printed into a C-like representation.
7081 Enable more detailed dumps (not honored by every dump option). Also
7082 include information from the optimization passes.
7084 Enable dumping various statistics about the pass (not honored by every dump
7087 Enable showing basic block boundaries (disabled in raw dumps).
7089 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
7090 dump a representation of the control flow graph suitable for viewing with
7091 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
7092 the file is pretty-printed as a subgraph, so that GraphViz can render them
7093 all in a single plot.
7095 This option currently only works for RTL dumps, and the RTL is always
7096 dumped in slim form.
7098 Enable showing virtual operands for every statement.
7100 Enable showing line numbers for statements.
7102 Enable showing the unique ID (@code{DECL_UID}) for each variable.
7104 Enable showing the tree dump for each statement.
7106 Enable showing the EH region number holding each statement.
7108 Enable showing scalar evolution analysis details.
7110 Enable showing optimization information (only available in certain
7113 Enable showing missed optimization information (only available in certain
7116 Enable other detailed optimization information (only available in
7118 @item =@var{filename}
7119 Instead of an auto named dump file, output into the given file
7120 name. The file names @file{stdout} and @file{stderr} are treated
7121 specially and are considered already open standard streams. For
7125 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
7126 -fdump-tree-pre=stderr file.c
7129 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
7130 output on to @file{stderr}. If two conflicting dump filenames are
7131 given for the same pass, then the latter option overrides the earlier
7135 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
7136 and @option{lineno}.
7139 Turn on all optimization options, i.e., @option{optimized},
7140 @option{missed}, and @option{note}.
7143 The following tree dumps are possible:
7147 @opindex fdump-tree-original
7148 Dump before any tree based optimization, to @file{@var{file}.original}.
7151 @opindex fdump-tree-optimized
7152 Dump after all tree based optimization, to @file{@var{file}.optimized}.
7155 @opindex fdump-tree-gimple
7156 Dump each function before and after the gimplification pass to a file. The
7157 file name is made by appending @file{.gimple} to the source file name.
7160 @opindex fdump-tree-cfg
7161 Dump the control flow graph of each function to a file. The file name is
7162 made by appending @file{.cfg} to the source file name.
7165 @opindex fdump-tree-ch
7166 Dump each function after copying loop headers. The file name is made by
7167 appending @file{.ch} to the source file name.
7170 @opindex fdump-tree-ssa
7171 Dump SSA related information to a file. The file name is made by appending
7172 @file{.ssa} to the source file name.
7175 @opindex fdump-tree-alias
7176 Dump aliasing information for each function. The file name is made by
7177 appending @file{.alias} to the source file name.
7180 @opindex fdump-tree-ccp
7181 Dump each function after CCP@. The file name is made by appending
7182 @file{.ccp} to the source file name.
7185 @opindex fdump-tree-storeccp
7186 Dump each function after STORE-CCP@. The file name is made by appending
7187 @file{.storeccp} to the source file name.
7190 @opindex fdump-tree-pre
7191 Dump trees after partial redundancy elimination. The file name is made
7192 by appending @file{.pre} to the source file name.
7195 @opindex fdump-tree-fre
7196 Dump trees after full redundancy elimination. The file name is made
7197 by appending @file{.fre} to the source file name.
7200 @opindex fdump-tree-copyprop
7201 Dump trees after copy propagation. The file name is made
7202 by appending @file{.copyprop} to the source file name.
7204 @item store_copyprop
7205 @opindex fdump-tree-store_copyprop
7206 Dump trees after store copy-propagation. The file name is made
7207 by appending @file{.store_copyprop} to the source file name.
7210 @opindex fdump-tree-dce
7211 Dump each function after dead code elimination. The file name is made by
7212 appending @file{.dce} to the source file name.
7215 @opindex fdump-tree-sra
7216 Dump each function after performing scalar replacement of aggregates. The
7217 file name is made by appending @file{.sra} to the source file name.
7220 @opindex fdump-tree-sink
7221 Dump each function after performing code sinking. The file name is made
7222 by appending @file{.sink} to the source file name.
7225 @opindex fdump-tree-dom
7226 Dump each function after applying dominator tree optimizations. The file
7227 name is made by appending @file{.dom} to the source file name.
7230 @opindex fdump-tree-dse
7231 Dump each function after applying dead store elimination. The file
7232 name is made by appending @file{.dse} to the source file name.
7235 @opindex fdump-tree-phiopt
7236 Dump each function after optimizing PHI nodes into straightline code. The file
7237 name is made by appending @file{.phiopt} to the source file name.
7240 @opindex fdump-tree-forwprop
7241 Dump each function after forward propagating single use variables. The file
7242 name is made by appending @file{.forwprop} to the source file name.
7245 @opindex fdump-tree-nrv
7246 Dump each function after applying the named return value optimization on
7247 generic trees. The file name is made by appending @file{.nrv} to the source
7251 @opindex fdump-tree-vect
7252 Dump each function after applying vectorization of loops. The file name is
7253 made by appending @file{.vect} to the source file name.
7256 @opindex fdump-tree-slp
7257 Dump each function after applying vectorization of basic blocks. The file name
7258 is made by appending @file{.slp} to the source file name.
7261 @opindex fdump-tree-vrp
7262 Dump each function after Value Range Propagation (VRP). The file name
7263 is made by appending @file{.vrp} to the source file name.
7266 @opindex fdump-tree-oaccdevlow
7267 Dump each function after applying device-specific OpenACC transformations.
7268 The file name is made by appending @file{.oaccdevlow} to the source file name.
7271 @opindex fdump-tree-all
7272 Enable all the available tree dumps with the flags provided in this option.
7276 @itemx -fopt-info-@var{options}
7277 @itemx -fopt-info-@var{options}=@var{filename}
7279 Controls optimization dumps from various optimization passes. If the
7280 @samp{-@var{options}} form is used, @var{options} is a list of
7281 @samp{-} separated option keywords to select the dump details and
7284 The @var{options} can be divided into two groups: options describing the
7285 verbosity of the dump, and options describing which optimizations
7286 should be included. The options from both the groups can be freely
7287 mixed as they are non-overlapping. However, in case of any conflicts,
7288 the later options override the earlier options on the command
7291 The following options control the dump verbosity:
7295 Print information when an optimization is successfully applied. It is
7296 up to a pass to decide which information is relevant. For example, the
7297 vectorizer passes print the source location of loops which are
7298 successfully vectorized.
7300 Print information about missed optimizations. Individual passes
7301 control which information to include in the output.
7303 Print verbose information about optimizations, such as certain
7304 transformations, more detailed messages about decisions etc.
7306 Print detailed optimization information. This includes
7307 @samp{optimized}, @samp{missed}, and @samp{note}.
7310 One or more of the following option keywords can be used to describe a
7311 group of optimizations:
7315 Enable dumps from all interprocedural optimizations.
7317 Enable dumps from all loop optimizations.
7319 Enable dumps from all inlining optimizations.
7321 Enable dumps from all vectorization optimizations.
7323 Enable dumps from all optimizations. This is a superset of
7324 the optimization groups listed above.
7328 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7329 info about successful optimizations from all the passes.
7331 If the @var{filename} is provided, then the dumps from all the
7332 applicable optimizations are concatenated into the @var{filename}.
7333 Otherwise the dump is output onto @file{stderr}. Though multiple
7334 @option{-fopt-info} options are accepted, only one of them can include
7335 a @var{filename}. If other filenames are provided then all but the
7336 first such option are ignored.
7338 Note that the output @var{filename} is overwritten
7339 in case of multiple translation units. If a combined output from
7340 multiple translation units is desired, @file{stderr} should be used
7343 In the following example, the optimization info is output to
7352 gcc -O3 -fopt-info-missed=missed.all
7356 outputs missed optimization report from all the passes into
7357 @file{missed.all}, and this one:
7360 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7364 prints information about missed optimization opportunities from
7365 vectorization passes on @file{stderr}.
7366 Note that @option{-fopt-info-vec-missed} is equivalent to
7367 @option{-fopt-info-missed-vec}.
7371 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7375 outputs information about missed optimizations as well as
7376 optimized locations from all the inlining passes into
7382 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7386 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7387 in conflict since only one output file is allowed. In this case, only
7388 the first option takes effect and the subsequent options are
7389 ignored. Thus only @file{vec.miss} is produced which contains
7390 dumps from the vectorizer about missed opportunities.
7392 @item -frandom-seed=@var{number}
7393 @opindex frandom-seed
7394 This option provides a seed that GCC uses in place of
7395 random numbers in generating certain symbol names
7396 that have to be different in every compiled file. It is also used to
7397 place unique stamps in coverage data files and the object files that
7398 produce them. You can use the @option{-frandom-seed} option to produce
7399 reproducibly identical object files.
7401 The @var{number} should be different for every file you compile.
7403 @item -fsched-verbose=@var{n}
7404 @opindex fsched-verbose
7405 On targets that use instruction scheduling, this option controls the
7406 amount of debugging output the scheduler prints. This information is
7407 written to standard error, unless @option{-fdump-rtl-sched1} or
7408 @option{-fdump-rtl-sched2} is specified, in which case it is output
7409 to the usual dump listing file, @file{.sched1} or @file{.sched2}
7410 respectively. However for @var{n} greater than nine, the output is
7411 always printed to standard error.
7413 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7414 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7415 For @var{n} greater than one, it also output basic block probabilities,
7416 detailed ready list information and unit/insn info. For @var{n} greater
7417 than two, it includes RTL at abort point, control-flow and regions info.
7418 And for @var{n} over four, @option{-fsched-verbose} also includes
7422 @itemx -save-temps=cwd
7424 Store the usual ``temporary'' intermediate files permanently; place them
7425 in the current directory and name them based on the source file. Thus,
7426 compiling @file{foo.c} with @option{-c -save-temps} produces files
7427 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7428 preprocessed @file{foo.i} output file even though the compiler now
7429 normally uses an integrated preprocessor.
7431 When used in combination with the @option{-x} command-line option,
7432 @option{-save-temps} is sensible enough to avoid over writing an
7433 input source file with the same extension as an intermediate file.
7434 The corresponding intermediate file may be obtained by renaming the
7435 source file before using @option{-save-temps}.
7437 If you invoke GCC in parallel, compiling several different source
7438 files that share a common base name in different subdirectories or the
7439 same source file compiled for multiple output destinations, it is
7440 likely that the different parallel compilers will interfere with each
7441 other, and overwrite the temporary files. For instance:
7444 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7445 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7448 may result in @file{foo.i} and @file{foo.o} being written to
7449 simultaneously by both compilers.
7451 @item -save-temps=obj
7452 @opindex save-temps=obj
7453 Store the usual ``temporary'' intermediate files permanently. If the
7454 @option{-o} option is used, the temporary files are based on the
7455 object file. If the @option{-o} option is not used, the
7456 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7461 gcc -save-temps=obj -c foo.c
7462 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7463 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7467 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7468 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7469 @file{dir2/yfoobar.o}.
7471 @item -time@r{[}=@var{file}@r{]}
7473 Report the CPU time taken by each subprocess in the compilation
7474 sequence. For C source files, this is the compiler proper and assembler
7475 (plus the linker if linking is done).
7477 Without the specification of an output file, the output looks like this:
7484 The first number on each line is the ``user time'', that is time spent
7485 executing the program itself. The second number is ``system time'',
7486 time spent executing operating system routines on behalf of the program.
7487 Both numbers are in seconds.
7489 With the specification of an output file, the output is appended to the
7490 named file, and it looks like this:
7493 0.12 0.01 cc1 @var{options}
7494 0.00 0.01 as @var{options}
7497 The ``user time'' and the ``system time'' are moved before the program
7498 name, and the options passed to the program are displayed, so that one
7499 can later tell what file was being compiled, and with which options.
7501 @item -fvar-tracking
7502 @opindex fvar-tracking
7503 Run variable tracking pass. It computes where variables are stored at each
7504 position in code. Better debugging information is then generated
7505 (if the debugging information format supports this information).
7507 It is enabled by default when compiling with optimization (@option{-Os},
7508 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7509 the debug info format supports it.
7511 @item -fvar-tracking-assignments
7512 @opindex fvar-tracking-assignments
7513 @opindex fno-var-tracking-assignments
7514 Annotate assignments to user variables early in the compilation and
7515 attempt to carry the annotations over throughout the compilation all the
7516 way to the end, in an attempt to improve debug information while
7517 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7519 It can be enabled even if var-tracking is disabled, in which case
7520 annotations are created and maintained, but discarded at the end.
7521 By default, this flag is enabled together with @option{-fvar-tracking},
7522 except when selective scheduling is enabled.
7524 @item -fvar-tracking-assignments-toggle
7525 @opindex fvar-tracking-assignments-toggle
7526 @opindex fno-var-tracking-assignments-toggle
7527 Toggle @option{-fvar-tracking-assignments}, in the same way that
7528 @option{-gtoggle} toggles @option{-g}.
7530 @item -print-file-name=@var{library}
7531 @opindex print-file-name
7532 Print the full absolute name of the library file @var{library} that
7533 would be used when linking---and don't do anything else. With this
7534 option, GCC does not compile or link anything; it just prints the
7537 @item -print-multi-directory
7538 @opindex print-multi-directory
7539 Print the directory name corresponding to the multilib selected by any
7540 other switches present in the command line. This directory is supposed
7541 to exist in @env{GCC_EXEC_PREFIX}.
7543 @item -print-multi-lib
7544 @opindex print-multi-lib
7545 Print the mapping from multilib directory names to compiler switches
7546 that enable them. The directory name is separated from the switches by
7547 @samp{;}, and each switch starts with an @samp{@@} instead of the
7548 @samp{-}, without spaces between multiple switches. This is supposed to
7549 ease shell processing.
7551 @item -print-multi-os-directory
7552 @opindex print-multi-os-directory
7553 Print the path to OS libraries for the selected
7554 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7555 present in the @file{lib} subdirectory and no multilibs are used, this is
7556 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7557 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7558 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7559 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7561 @item -print-multiarch
7562 @opindex print-multiarch
7563 Print the path to OS libraries for the selected multiarch,
7564 relative to some @file{lib} subdirectory.
7566 @item -print-prog-name=@var{program}
7567 @opindex print-prog-name
7568 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7570 @item -print-libgcc-file-name
7571 @opindex print-libgcc-file-name
7572 Same as @option{-print-file-name=libgcc.a}.
7574 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7575 but you do want to link with @file{libgcc.a}. You can do:
7578 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7581 @item -print-search-dirs
7582 @opindex print-search-dirs
7583 Print the name of the configured installation directory and a list of
7584 program and library directories @command{gcc} searches---and don't do anything else.
7586 This is useful when @command{gcc} prints the error message
7587 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7588 To resolve this you either need to put @file{cpp0} and the other compiler
7589 components where @command{gcc} expects to find them, or you can set the environment
7590 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7591 Don't forget the trailing @samp{/}.
7592 @xref{Environment Variables}.
7594 @item -print-sysroot
7595 @opindex print-sysroot
7596 Print the target sysroot directory that is used during
7597 compilation. This is the target sysroot specified either at configure
7598 time or using the @option{--sysroot} option, possibly with an extra
7599 suffix that depends on compilation options. If no target sysroot is
7600 specified, the option prints nothing.
7602 @item -print-sysroot-headers-suffix
7603 @opindex print-sysroot-headers-suffix
7604 Print the suffix added to the target sysroot when searching for
7605 headers, or give an error if the compiler is not configured with such
7606 a suffix---and don't do anything else.
7609 @opindex dumpmachine
7610 Print the compiler's target machine (for example,
7611 @samp{i686-pc-linux-gnu})---and don't do anything else.
7614 @opindex dumpversion
7615 Print the compiler version (for example, @code{3.0})---and don't do
7620 Print the compiler's built-in specs---and don't do anything else. (This
7621 is used when GCC itself is being built.) @xref{Spec Files}.
7623 @item -fno-eliminate-unused-debug-types
7624 @opindex feliminate-unused-debug-types
7625 @opindex fno-eliminate-unused-debug-types
7626 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7627 output for types that are nowhere used in the source file being compiled.
7628 Sometimes it is useful to have GCC emit debugging
7629 information for all types declared in a compilation
7630 unit, regardless of whether or not they are actually used
7631 in that compilation unit, for example
7632 if, in the debugger, you want to cast a value to a type that is
7633 not actually used in your program (but is declared). More often,
7634 however, this results in a significant amount of wasted space.
7637 @node Optimize Options
7638 @section Options That Control Optimization
7639 @cindex optimize options
7640 @cindex options, optimization
7642 These options control various sorts of optimizations.
7644 Without any optimization option, the compiler's goal is to reduce the
7645 cost of compilation and to make debugging produce the expected
7646 results. Statements are independent: if you stop the program with a
7647 breakpoint between statements, you can then assign a new value to any
7648 variable or change the program counter to any other statement in the
7649 function and get exactly the results you expect from the source
7652 Turning on optimization flags makes the compiler attempt to improve
7653 the performance and/or code size at the expense of compilation time
7654 and possibly the ability to debug the program.
7656 The compiler performs optimization based on the knowledge it has of the
7657 program. Compiling multiple files at once to a single output file mode allows
7658 the compiler to use information gained from all of the files when compiling
7661 Not all optimizations are controlled directly by a flag. Only
7662 optimizations that have a flag are listed in this section.
7664 Most optimizations are only enabled if an @option{-O} level is set on
7665 the command line. Otherwise they are disabled, even if individual
7666 optimization flags are specified.
7668 Depending on the target and how GCC was configured, a slightly different
7669 set of optimizations may be enabled at each @option{-O} level than
7670 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7671 to find out the exact set of optimizations that are enabled at each level.
7672 @xref{Overall Options}, for examples.
7679 Optimize. Optimizing compilation takes somewhat more time, and a lot
7680 more memory for a large function.
7682 With @option{-O}, the compiler tries to reduce code size and execution
7683 time, without performing any optimizations that take a great deal of
7686 @option{-O} turns on the following optimization flags:
7689 -fbranch-count-reg @gol
7690 -fcombine-stack-adjustments @gol
7692 -fcprop-registers @gol
7695 -fdelayed-branch @gol
7697 -fforward-propagate @gol
7698 -fguess-branch-probability @gol
7699 -fif-conversion2 @gol
7700 -fif-conversion @gol
7701 -finline-functions-called-once @gol
7702 -fipa-pure-const @gol
7704 -fipa-reference @gol
7705 -fmerge-constants @gol
7706 -fmove-loop-invariants @gol
7707 -freorder-blocks @gol
7709 -fsplit-wide-types @gol
7714 -ftree-coalesce-vars @gol
7715 -ftree-copy-prop @gol
7717 -ftree-dominator-opts @gol
7719 -ftree-forwprop @gol
7729 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7730 where doing so does not interfere with debugging.
7734 Optimize even more. GCC performs nearly all supported optimizations
7735 that do not involve a space-speed tradeoff.
7736 As compared to @option{-O}, this option increases both compilation time
7737 and the performance of the generated code.
7739 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7740 also turns on the following optimization flags:
7741 @gccoptlist{-fthread-jumps @gol
7742 -falign-functions -falign-jumps @gol
7743 -falign-loops -falign-labels @gol
7746 -fcse-follow-jumps -fcse-skip-blocks @gol
7747 -fdelete-null-pointer-checks @gol
7748 -fdevirtualize -fdevirtualize-speculatively @gol
7749 -fexpensive-optimizations @gol
7750 -fgcse -fgcse-lm @gol
7751 -fhoist-adjacent-loads @gol
7752 -finline-small-functions @gol
7753 -findirect-inlining @gol
7755 -fipa-cp-alignment @gol
7758 -fisolate-erroneous-paths-dereference @gol
7760 -foptimize-sibling-calls @gol
7761 -foptimize-strlen @gol
7762 -fpartial-inlining @gol
7764 -freorder-blocks-algorithm=stc @gol
7765 -freorder-blocks-and-partition -freorder-functions @gol
7766 -frerun-cse-after-loop @gol
7767 -fsched-interblock -fsched-spec @gol
7768 -fschedule-insns -fschedule-insns2 @gol
7769 -fstrict-aliasing -fstrict-overflow @gol
7770 -ftree-builtin-call-dce @gol
7771 -ftree-switch-conversion -ftree-tail-merge @gol
7776 Please note the warning under @option{-fgcse} about
7777 invoking @option{-O2} on programs that use computed gotos.
7781 Optimize yet more. @option{-O3} turns on all optimizations specified
7782 by @option{-O2} and also turns on the @option{-finline-functions},
7783 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7784 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7785 @option{-ftree-loop-distribute-patterns},
7786 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7787 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7791 Reduce compilation time and make debugging produce the expected
7792 results. This is the default.
7796 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7797 do not typically increase code size. It also performs further
7798 optimizations designed to reduce code size.
7800 @option{-Os} disables the following optimization flags:
7801 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7802 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7803 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7807 Disregard strict standards compliance. @option{-Ofast} enables all
7808 @option{-O3} optimizations. It also enables optimizations that are not
7809 valid for all standard-compliant programs.
7810 It turns on @option{-ffast-math} and the Fortran-specific
7811 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7815 Optimize debugging experience. @option{-Og} enables optimizations
7816 that do not interfere with debugging. It should be the optimization
7817 level of choice for the standard edit-compile-debug cycle, offering
7818 a reasonable level of optimization while maintaining fast compilation
7819 and a good debugging experience.
7821 If you use multiple @option{-O} options, with or without level numbers,
7822 the last such option is the one that is effective.
7825 Options of the form @option{-f@var{flag}} specify machine-independent
7826 flags. Most flags have both positive and negative forms; the negative
7827 form of @option{-ffoo} is @option{-fno-foo}. In the table
7828 below, only one of the forms is listed---the one you typically
7829 use. You can figure out the other form by either removing @samp{no-}
7832 The following options control specific optimizations. They are either
7833 activated by @option{-O} options or are related to ones that are. You
7834 can use the following flags in the rare cases when ``fine-tuning'' of
7835 optimizations to be performed is desired.
7838 @item -fno-defer-pop
7839 @opindex fno-defer-pop
7840 Always pop the arguments to each function call as soon as that function
7841 returns. For machines that must pop arguments after a function call,
7842 the compiler normally lets arguments accumulate on the stack for several
7843 function calls and pops them all at once.
7845 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7847 @item -fforward-propagate
7848 @opindex fforward-propagate
7849 Perform a forward propagation pass on RTL@. The pass tries to combine two
7850 instructions and checks if the result can be simplified. If loop unrolling
7851 is active, two passes are performed and the second is scheduled after
7854 This option is enabled by default at optimization levels @option{-O},
7855 @option{-O2}, @option{-O3}, @option{-Os}.
7857 @item -ffp-contract=@var{style}
7858 @opindex ffp-contract
7859 @option{-ffp-contract=off} disables floating-point expression contraction.
7860 @option{-ffp-contract=fast} enables floating-point expression contraction
7861 such as forming of fused multiply-add operations if the target has
7862 native support for them.
7863 @option{-ffp-contract=on} enables floating-point expression contraction
7864 if allowed by the language standard. This is currently not implemented
7865 and treated equal to @option{-ffp-contract=off}.
7867 The default is @option{-ffp-contract=fast}.
7869 @item -fomit-frame-pointer
7870 @opindex fomit-frame-pointer
7871 Don't keep the frame pointer in a register for functions that
7872 don't need one. This avoids the instructions to save, set up and
7873 restore frame pointers; it also makes an extra register available
7874 in many functions. @strong{It also makes debugging impossible on
7877 On some machines, such as the VAX, this flag has no effect, because
7878 the standard calling sequence automatically handles the frame pointer
7879 and nothing is saved by pretending it doesn't exist. The
7880 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7881 whether a target machine supports this flag. @xref{Registers,,Register
7882 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7884 The default setting (when not optimizing for
7885 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7886 @option{-fomit-frame-pointer}. You can configure GCC with the
7887 @option{--enable-frame-pointer} configure option to change the default.
7889 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7891 @item -foptimize-sibling-calls
7892 @opindex foptimize-sibling-calls
7893 Optimize sibling and tail recursive calls.
7895 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7897 @item -foptimize-strlen
7898 @opindex foptimize-strlen
7899 Optimize various standard C string functions (e.g. @code{strlen},
7900 @code{strchr} or @code{strcpy}) and
7901 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7903 Enabled at levels @option{-O2}, @option{-O3}.
7907 Do not expand any functions inline apart from those marked with
7908 the @code{always_inline} attribute. This is the default when not
7911 Single functions can be exempted from inlining by marking them
7912 with the @code{noinline} attribute.
7914 @item -finline-small-functions
7915 @opindex finline-small-functions
7916 Integrate functions into their callers when their body is smaller than expected
7917 function call code (so overall size of program gets smaller). The compiler
7918 heuristically decides which functions are simple enough to be worth integrating
7919 in this way. This inlining applies to all functions, even those not declared
7922 Enabled at level @option{-O2}.
7924 @item -findirect-inlining
7925 @opindex findirect-inlining
7926 Inline also indirect calls that are discovered to be known at compile
7927 time thanks to previous inlining. This option has any effect only
7928 when inlining itself is turned on by the @option{-finline-functions}
7929 or @option{-finline-small-functions} options.
7931 Enabled at level @option{-O2}.
7933 @item -finline-functions
7934 @opindex finline-functions
7935 Consider all functions for inlining, even if they are not declared inline.
7936 The compiler heuristically decides which functions are worth integrating
7939 If all calls to a given function are integrated, and the function is
7940 declared @code{static}, then the function is normally not output as
7941 assembler code in its own right.
7943 Enabled at level @option{-O3}.
7945 @item -finline-functions-called-once
7946 @opindex finline-functions-called-once
7947 Consider all @code{static} functions called once for inlining into their
7948 caller even if they are not marked @code{inline}. If a call to a given
7949 function is integrated, then the function is not output as assembler code
7952 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7954 @item -fearly-inlining
7955 @opindex fearly-inlining
7956 Inline functions marked by @code{always_inline} and functions whose body seems
7957 smaller than the function call overhead early before doing
7958 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7959 makes profiling significantly cheaper and usually inlining faster on programs
7960 having large chains of nested wrapper functions.
7966 Perform interprocedural scalar replacement of aggregates, removal of
7967 unused parameters and replacement of parameters passed by reference
7968 by parameters passed by value.
7970 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7972 @item -finline-limit=@var{n}
7973 @opindex finline-limit
7974 By default, GCC limits the size of functions that can be inlined. This flag
7975 allows coarse control of this limit. @var{n} is the size of functions that
7976 can be inlined in number of pseudo instructions.
7978 Inlining is actually controlled by a number of parameters, which may be
7979 specified individually by using @option{--param @var{name}=@var{value}}.
7980 The @option{-finline-limit=@var{n}} option sets some of these parameters
7984 @item max-inline-insns-single
7985 is set to @var{n}/2.
7986 @item max-inline-insns-auto
7987 is set to @var{n}/2.
7990 See below for a documentation of the individual
7991 parameters controlling inlining and for the defaults of these parameters.
7993 @emph{Note:} there may be no value to @option{-finline-limit} that results
7994 in default behavior.
7996 @emph{Note:} pseudo instruction represents, in this particular context, an
7997 abstract measurement of function's size. In no way does it represent a count
7998 of assembly instructions and as such its exact meaning might change from one
7999 release to an another.
8001 @item -fno-keep-inline-dllexport
8002 @opindex fno-keep-inline-dllexport
8003 This is a more fine-grained version of @option{-fkeep-inline-functions},
8004 which applies only to functions that are declared using the @code{dllexport}
8005 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
8008 @item -fkeep-inline-functions
8009 @opindex fkeep-inline-functions
8010 In C, emit @code{static} functions that are declared @code{inline}
8011 into the object file, even if the function has been inlined into all
8012 of its callers. This switch does not affect functions using the
8013 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8014 inline functions into the object file.
8016 @item -fkeep-static-functions
8017 @opindex fkeep-static-functions
8018 Emit @code{static} functions into the object file, even if the function
8021 @item -fkeep-static-consts
8022 @opindex fkeep-static-consts
8023 Emit variables declared @code{static const} when optimization isn't turned
8024 on, even if the variables aren't referenced.
8026 GCC enables this option by default. If you want to force the compiler to
8027 check if a variable is referenced, regardless of whether or not
8028 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8030 @item -fmerge-constants
8031 @opindex fmerge-constants
8032 Attempt to merge identical constants (string constants and floating-point
8033 constants) across compilation units.
8035 This option is the default for optimized compilation if the assembler and
8036 linker support it. Use @option{-fno-merge-constants} to inhibit this
8039 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8041 @item -fmerge-all-constants
8042 @opindex fmerge-all-constants
8043 Attempt to merge identical constants and identical variables.
8045 This option implies @option{-fmerge-constants}. In addition to
8046 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8047 arrays or initialized constant variables with integral or floating-point
8048 types. Languages like C or C++ require each variable, including multiple
8049 instances of the same variable in recursive calls, to have distinct locations,
8050 so using this option results in non-conforming
8053 @item -fmodulo-sched
8054 @opindex fmodulo-sched
8055 Perform swing modulo scheduling immediately before the first scheduling
8056 pass. This pass looks at innermost loops and reorders their
8057 instructions by overlapping different iterations.
8059 @item -fmodulo-sched-allow-regmoves
8060 @opindex fmodulo-sched-allow-regmoves
8061 Perform more aggressive SMS-based modulo scheduling with register moves
8062 allowed. By setting this flag certain anti-dependences edges are
8063 deleted, which triggers the generation of reg-moves based on the
8064 life-range analysis. This option is effective only with
8065 @option{-fmodulo-sched} enabled.
8067 @item -fno-branch-count-reg
8068 @opindex fno-branch-count-reg
8069 Do not use ``decrement and branch'' instructions on a count register,
8070 but instead generate a sequence of instructions that decrement a
8071 register, compare it against zero, then branch based upon the result.
8072 This option is only meaningful on architectures that support such
8073 instructions, which include x86, PowerPC, IA-64 and S/390.
8075 Enabled by default at @option{-O1} and higher.
8077 The default is @option{-fbranch-count-reg}.
8079 @item -fno-function-cse
8080 @opindex fno-function-cse
8081 Do not put function addresses in registers; make each instruction that
8082 calls a constant function contain the function's address explicitly.
8084 This option results in less efficient code, but some strange hacks
8085 that alter the assembler output may be confused by the optimizations
8086 performed when this option is not used.
8088 The default is @option{-ffunction-cse}
8090 @item -fno-zero-initialized-in-bss
8091 @opindex fno-zero-initialized-in-bss
8092 If the target supports a BSS section, GCC by default puts variables that
8093 are initialized to zero into BSS@. This can save space in the resulting
8096 This option turns off this behavior because some programs explicitly
8097 rely on variables going to the data section---e.g., so that the
8098 resulting executable can find the beginning of that section and/or make
8099 assumptions based on that.
8101 The default is @option{-fzero-initialized-in-bss}.
8103 @item -fthread-jumps
8104 @opindex fthread-jumps
8105 Perform optimizations that check to see if a jump branches to a
8106 location where another comparison subsumed by the first is found. If
8107 so, the first branch is redirected to either the destination of the
8108 second branch or a point immediately following it, depending on whether
8109 the condition is known to be true or false.
8111 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8113 @item -fsplit-wide-types
8114 @opindex fsplit-wide-types
8115 When using a type that occupies multiple registers, such as @code{long
8116 long} on a 32-bit system, split the registers apart and allocate them
8117 independently. This normally generates better code for those types,
8118 but may make debugging more difficult.
8120 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8123 @item -fcse-follow-jumps
8124 @opindex fcse-follow-jumps
8125 In common subexpression elimination (CSE), scan through jump instructions
8126 when the target of the jump is not reached by any other path. For
8127 example, when CSE encounters an @code{if} statement with an
8128 @code{else} clause, CSE follows the jump when the condition
8131 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8133 @item -fcse-skip-blocks
8134 @opindex fcse-skip-blocks
8135 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8136 follow jumps that conditionally skip over blocks. When CSE
8137 encounters a simple @code{if} statement with no else clause,
8138 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8139 body of the @code{if}.
8141 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8143 @item -frerun-cse-after-loop
8144 @opindex frerun-cse-after-loop
8145 Re-run common subexpression elimination after loop optimizations are
8148 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8152 Perform a global common subexpression elimination pass.
8153 This pass also performs global constant and copy propagation.
8155 @emph{Note:} When compiling a program using computed gotos, a GCC
8156 extension, you may get better run-time performance if you disable
8157 the global common subexpression elimination pass by adding
8158 @option{-fno-gcse} to the command line.
8160 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8164 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8165 attempts to move loads that are only killed by stores into themselves. This
8166 allows a loop containing a load/store sequence to be changed to a load outside
8167 the loop, and a copy/store within the loop.
8169 Enabled by default when @option{-fgcse} is enabled.
8173 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8174 global common subexpression elimination. This pass attempts to move
8175 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8176 loops containing a load/store sequence can be changed to a load before
8177 the loop and a store after the loop.
8179 Not enabled at any optimization level.
8183 When @option{-fgcse-las} is enabled, the global common subexpression
8184 elimination pass eliminates redundant loads that come after stores to the
8185 same memory location (both partial and full redundancies).
8187 Not enabled at any optimization level.
8189 @item -fgcse-after-reload
8190 @opindex fgcse-after-reload
8191 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8192 pass is performed after reload. The purpose of this pass is to clean up
8195 @item -faggressive-loop-optimizations
8196 @opindex faggressive-loop-optimizations
8197 This option tells the loop optimizer to use language constraints to
8198 derive bounds for the number of iterations of a loop. This assumes that
8199 loop code does not invoke undefined behavior by for example causing signed
8200 integer overflows or out-of-bound array accesses. The bounds for the
8201 number of iterations of a loop are used to guide loop unrolling and peeling
8202 and loop exit test optimizations.
8203 This option is enabled by default.
8205 @item -funsafe-loop-optimizations
8206 @opindex funsafe-loop-optimizations
8207 This option tells the loop optimizer to assume that loop indices do not
8208 overflow, and that loops with nontrivial exit condition are not
8209 infinite. This enables a wider range of loop optimizations even if
8210 the loop optimizer itself cannot prove that these assumptions are valid.
8211 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
8212 if it finds this kind of loop.
8214 @item -fcrossjumping
8215 @opindex fcrossjumping
8216 Perform cross-jumping transformation.
8217 This transformation unifies equivalent code and saves code size. The
8218 resulting code may or may not perform better than without cross-jumping.
8220 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8222 @item -fauto-inc-dec
8223 @opindex fauto-inc-dec
8224 Combine increments or decrements of addresses with memory accesses.
8225 This pass is always skipped on architectures that do not have
8226 instructions to support this. Enabled by default at @option{-O} and
8227 higher on architectures that support this.
8231 Perform dead code elimination (DCE) on RTL@.
8232 Enabled by default at @option{-O} and higher.
8236 Perform dead store elimination (DSE) on RTL@.
8237 Enabled by default at @option{-O} and higher.
8239 @item -fif-conversion
8240 @opindex fif-conversion
8241 Attempt to transform conditional jumps into branch-less equivalents. This
8242 includes use of conditional moves, min, max, set flags and abs instructions, and
8243 some tricks doable by standard arithmetics. The use of conditional execution
8244 on chips where it is available is controlled by @option{-fif-conversion2}.
8246 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8248 @item -fif-conversion2
8249 @opindex fif-conversion2
8250 Use conditional execution (where available) to transform conditional jumps into
8251 branch-less equivalents.
8253 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8255 @item -fdeclone-ctor-dtor
8256 @opindex fdeclone-ctor-dtor
8257 The C++ ABI requires multiple entry points for constructors and
8258 destructors: one for a base subobject, one for a complete object, and
8259 one for a virtual destructor that calls operator delete afterwards.
8260 For a hierarchy with virtual bases, the base and complete variants are
8261 clones, which means two copies of the function. With this option, the
8262 base and complete variants are changed to be thunks that call a common
8265 Enabled by @option{-Os}.
8267 @item -fdelete-null-pointer-checks
8268 @opindex fdelete-null-pointer-checks
8269 Assume that programs cannot safely dereference null pointers, and that
8270 no code or data element resides at address zero.
8271 This option enables simple constant
8272 folding optimizations at all optimization levels. In addition, other
8273 optimization passes in GCC use this flag to control global dataflow
8274 analyses that eliminate useless checks for null pointers; these assume
8275 that a memory access to address zero always results in a trap, so
8276 that if a pointer is checked after it has already been dereferenced,
8279 Note however that in some environments this assumption is not true.
8280 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8281 for programs that depend on that behavior.
8283 This option is enabled by default on most targets. On Nios II ELF, it
8284 defaults to off. On AVR and CR16, this option is completely disabled.
8286 Passes that use the dataflow information
8287 are enabled independently at different optimization levels.
8289 @item -fdevirtualize
8290 @opindex fdevirtualize
8291 Attempt to convert calls to virtual functions to direct calls. This
8292 is done both within a procedure and interprocedurally as part of
8293 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8294 propagation (@option{-fipa-cp}).
8295 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8297 @item -fdevirtualize-speculatively
8298 @opindex fdevirtualize-speculatively
8299 Attempt to convert calls to virtual functions to speculative direct calls.
8300 Based on the analysis of the type inheritance graph, determine for a given call
8301 the set of likely targets. If the set is small, preferably of size 1, change
8302 the call into a conditional deciding between direct and indirect calls. The
8303 speculative calls enable more optimizations, such as inlining. When they seem
8304 useless after further optimization, they are converted back into original form.
8306 @item -fdevirtualize-at-ltrans
8307 @opindex fdevirtualize-at-ltrans
8308 Stream extra information needed for aggressive devirtualization when running
8309 the link-time optimizer in local transformation mode.
8310 This option enables more devirtualization but
8311 significantly increases the size of streamed data. For this reason it is
8312 disabled by default.
8314 @item -fexpensive-optimizations
8315 @opindex fexpensive-optimizations
8316 Perform a number of minor optimizations that are relatively expensive.
8318 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8322 Attempt to remove redundant extension instructions. This is especially
8323 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8324 registers after writing to their lower 32-bit half.
8326 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8327 @option{-O3}, @option{-Os}.
8329 @item -fno-lifetime-dse
8330 @opindex fno-lifetime-dse
8331 In C++ the value of an object is only affected by changes within its
8332 lifetime: when the constructor begins, the object has an indeterminate
8333 value, and any changes during the lifetime of the object are dead when
8334 the object is destroyed. Normally dead store elimination will take
8335 advantage of this; if your code relies on the value of the object
8336 storage persisting beyond the lifetime of the object, you can use this
8337 flag to disable this optimization.
8339 @item -flive-range-shrinkage
8340 @opindex flive-range-shrinkage
8341 Attempt to decrease register pressure through register live range
8342 shrinkage. This is helpful for fast processors with small or moderate
8345 @item -fira-algorithm=@var{algorithm}
8346 @opindex fira-algorithm
8347 Use the specified coloring algorithm for the integrated register
8348 allocator. The @var{algorithm} argument can be @samp{priority}, which
8349 specifies Chow's priority coloring, or @samp{CB}, which specifies
8350 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8351 for all architectures, but for those targets that do support it, it is
8352 the default because it generates better code.
8354 @item -fira-region=@var{region}
8355 @opindex fira-region
8356 Use specified regions for the integrated register allocator. The
8357 @var{region} argument should be one of the following:
8362 Use all loops as register allocation regions.
8363 This can give the best results for machines with a small and/or
8364 irregular register set.
8367 Use all loops except for loops with small register pressure
8368 as the regions. This value usually gives
8369 the best results in most cases and for most architectures,
8370 and is enabled by default when compiling with optimization for speed
8371 (@option{-O}, @option{-O2}, @dots{}).
8374 Use all functions as a single region.
8375 This typically results in the smallest code size, and is enabled by default for
8376 @option{-Os} or @option{-O0}.
8380 @item -fira-hoist-pressure
8381 @opindex fira-hoist-pressure
8382 Use IRA to evaluate register pressure in the code hoisting pass for
8383 decisions to hoist expressions. This option usually results in smaller
8384 code, but it can slow the compiler down.
8386 This option is enabled at level @option{-Os} for all targets.
8388 @item -fira-loop-pressure
8389 @opindex fira-loop-pressure
8390 Use IRA to evaluate register pressure in loops for decisions to move
8391 loop invariants. This option usually results in generation
8392 of faster and smaller code on machines with large register files (>= 32
8393 registers), but it can slow the compiler down.
8395 This option is enabled at level @option{-O3} for some targets.
8397 @item -fno-ira-share-save-slots
8398 @opindex fno-ira-share-save-slots
8399 Disable sharing of stack slots used for saving call-used hard
8400 registers living through a call. Each hard register gets a
8401 separate stack slot, and as a result function stack frames are
8404 @item -fno-ira-share-spill-slots
8405 @opindex fno-ira-share-spill-slots
8406 Disable sharing of stack slots allocated for pseudo-registers. Each
8407 pseudo-register that does not get a hard register gets a separate
8408 stack slot, and as a result function stack frames are larger.
8410 @item -fira-verbose=@var{n}
8411 @opindex fira-verbose
8412 Control the verbosity of the dump file for the integrated register allocator.
8413 The default value is 5. If the value @var{n} is greater or equal to 10,
8414 the dump output is sent to stderr using the same format as @var{n} minus 10.
8418 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8419 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8420 values if it is profitable.
8422 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8424 @item -fdelayed-branch
8425 @opindex fdelayed-branch
8426 If supported for the target machine, attempt to reorder instructions
8427 to exploit instruction slots available after delayed branch
8430 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8432 @item -fschedule-insns
8433 @opindex fschedule-insns
8434 If supported for the target machine, attempt to reorder instructions to
8435 eliminate execution stalls due to required data being unavailable. This
8436 helps machines that have slow floating point or memory load instructions
8437 by allowing other instructions to be issued until the result of the load
8438 or floating-point instruction is required.
8440 Enabled at levels @option{-O2}, @option{-O3}.
8442 @item -fschedule-insns2
8443 @opindex fschedule-insns2
8444 Similar to @option{-fschedule-insns}, but requests an additional pass of
8445 instruction scheduling after register allocation has been done. This is
8446 especially useful on machines with a relatively small number of
8447 registers and where memory load instructions take more than one cycle.
8449 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8451 @item -fno-sched-interblock
8452 @opindex fno-sched-interblock
8453 Don't schedule instructions across basic blocks. This is normally
8454 enabled by default when scheduling before register allocation, i.e.@:
8455 with @option{-fschedule-insns} or at @option{-O2} or higher.
8457 @item -fno-sched-spec
8458 @opindex fno-sched-spec
8459 Don't allow speculative motion of non-load instructions. This is normally
8460 enabled by default when scheduling before register allocation, i.e.@:
8461 with @option{-fschedule-insns} or at @option{-O2} or higher.
8463 @item -fsched-pressure
8464 @opindex fsched-pressure
8465 Enable register pressure sensitive insn scheduling before register
8466 allocation. This only makes sense when scheduling before register
8467 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8468 @option{-O2} or higher. Usage of this option can improve the
8469 generated code and decrease its size by preventing register pressure
8470 increase above the number of available hard registers and subsequent
8471 spills in register allocation.
8473 @item -fsched-spec-load
8474 @opindex fsched-spec-load
8475 Allow speculative motion of some load instructions. This only makes
8476 sense when scheduling before register allocation, i.e.@: with
8477 @option{-fschedule-insns} or at @option{-O2} or higher.
8479 @item -fsched-spec-load-dangerous
8480 @opindex fsched-spec-load-dangerous
8481 Allow speculative motion of more load instructions. This only makes
8482 sense when scheduling before register allocation, i.e.@: with
8483 @option{-fschedule-insns} or at @option{-O2} or higher.
8485 @item -fsched-stalled-insns
8486 @itemx -fsched-stalled-insns=@var{n}
8487 @opindex fsched-stalled-insns
8488 Define how many insns (if any) can be moved prematurely from the queue
8489 of stalled insns into the ready list during the second scheduling pass.
8490 @option{-fno-sched-stalled-insns} means that no insns are moved
8491 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8492 on how many queued insns can be moved prematurely.
8493 @option{-fsched-stalled-insns} without a value is equivalent to
8494 @option{-fsched-stalled-insns=1}.
8496 @item -fsched-stalled-insns-dep
8497 @itemx -fsched-stalled-insns-dep=@var{n}
8498 @opindex fsched-stalled-insns-dep
8499 Define how many insn groups (cycles) are examined for a dependency
8500 on a stalled insn that is a candidate for premature removal from the queue
8501 of stalled insns. This has an effect only during the second scheduling pass,
8502 and only if @option{-fsched-stalled-insns} is used.
8503 @option{-fno-sched-stalled-insns-dep} is equivalent to
8504 @option{-fsched-stalled-insns-dep=0}.
8505 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8506 @option{-fsched-stalled-insns-dep=1}.
8508 @item -fsched2-use-superblocks
8509 @opindex fsched2-use-superblocks
8510 When scheduling after register allocation, use superblock scheduling.
8511 This allows motion across basic block boundaries,
8512 resulting in faster schedules. This option is experimental, as not all machine
8513 descriptions used by GCC model the CPU closely enough to avoid unreliable
8514 results from the algorithm.
8516 This only makes sense when scheduling after register allocation, i.e.@: with
8517 @option{-fschedule-insns2} or at @option{-O2} or higher.
8519 @item -fsched-group-heuristic
8520 @opindex fsched-group-heuristic
8521 Enable the group heuristic in the scheduler. This heuristic favors
8522 the instruction that belongs to a schedule group. This is enabled
8523 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8524 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8526 @item -fsched-critical-path-heuristic
8527 @opindex fsched-critical-path-heuristic
8528 Enable the critical-path heuristic in the scheduler. This heuristic favors
8529 instructions on the critical path. This is enabled by default when
8530 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8531 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8533 @item -fsched-spec-insn-heuristic
8534 @opindex fsched-spec-insn-heuristic
8535 Enable the speculative instruction heuristic in the scheduler. This
8536 heuristic favors speculative instructions with greater dependency weakness.
8537 This is enabled by default when scheduling is enabled, i.e.@:
8538 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8539 or at @option{-O2} or higher.
8541 @item -fsched-rank-heuristic
8542 @opindex fsched-rank-heuristic
8543 Enable the rank heuristic in the scheduler. This heuristic favors
8544 the instruction belonging to a basic block with greater size or frequency.
8545 This is enabled by default when scheduling is enabled, i.e.@:
8546 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8547 at @option{-O2} or higher.
8549 @item -fsched-last-insn-heuristic
8550 @opindex fsched-last-insn-heuristic
8551 Enable the last-instruction heuristic in the scheduler. This heuristic
8552 favors the instruction that is less dependent on the last instruction
8553 scheduled. This is enabled by default when scheduling is enabled,
8554 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8555 at @option{-O2} or higher.
8557 @item -fsched-dep-count-heuristic
8558 @opindex fsched-dep-count-heuristic
8559 Enable the dependent-count heuristic in the scheduler. This heuristic
8560 favors the instruction that has more instructions depending on it.
8561 This is enabled by default when scheduling is enabled, i.e.@:
8562 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8563 at @option{-O2} or higher.
8565 @item -freschedule-modulo-scheduled-loops
8566 @opindex freschedule-modulo-scheduled-loops
8567 Modulo scheduling is performed before traditional scheduling. If a loop
8568 is modulo scheduled, later scheduling passes may change its schedule.
8569 Use this option to control that behavior.
8571 @item -fselective-scheduling
8572 @opindex fselective-scheduling
8573 Schedule instructions using selective scheduling algorithm. Selective
8574 scheduling runs instead of the first scheduler pass.
8576 @item -fselective-scheduling2
8577 @opindex fselective-scheduling2
8578 Schedule instructions using selective scheduling algorithm. Selective
8579 scheduling runs instead of the second scheduler pass.
8581 @item -fsel-sched-pipelining
8582 @opindex fsel-sched-pipelining
8583 Enable software pipelining of innermost loops during selective scheduling.
8584 This option has no effect unless one of @option{-fselective-scheduling} or
8585 @option{-fselective-scheduling2} is turned on.
8587 @item -fsel-sched-pipelining-outer-loops
8588 @opindex fsel-sched-pipelining-outer-loops
8589 When pipelining loops during selective scheduling, also pipeline outer loops.
8590 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8592 @item -fsemantic-interposition
8593 @opindex fsemantic-interposition
8594 Some object formats, like ELF, allow interposing of symbols by the
8596 This means that for symbols exported from the DSO, the compiler cannot perform
8597 interprocedural propagation, inlining and other optimizations in anticipation
8598 that the function or variable in question may change. While this feature is
8599 useful, for example, to rewrite memory allocation functions by a debugging
8600 implementation, it is expensive in the terms of code quality.
8601 With @option{-fno-semantic-interposition} the compiler assumes that
8602 if interposition happens for functions the overwriting function will have
8603 precisely the same semantics (and side effects).
8604 Similarly if interposition happens
8605 for variables, the constructor of the variable will be the same. The flag
8606 has no effect for functions explicitly declared inline
8607 (where it is never allowed for interposition to change semantics)
8608 and for symbols explicitly declared weak.
8611 @opindex fshrink-wrap
8612 Emit function prologues only before parts of the function that need it,
8613 rather than at the top of the function. This flag is enabled by default at
8614 @option{-O} and higher.
8616 @item -fcaller-saves
8617 @opindex fcaller-saves
8618 Enable allocation of values to registers that are clobbered by
8619 function calls, by emitting extra instructions to save and restore the
8620 registers around such calls. Such allocation is done only when it
8621 seems to result in better code.
8623 This option is always enabled by default on certain machines, usually
8624 those which have no call-preserved registers to use instead.
8626 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8628 @item -fcombine-stack-adjustments
8629 @opindex fcombine-stack-adjustments
8630 Tracks stack adjustments (pushes and pops) and stack memory references
8631 and then tries to find ways to combine them.
8633 Enabled by default at @option{-O1} and higher.
8637 Use caller save registers for allocation if those registers are not used by
8638 any called function. In that case it is not necessary to save and restore
8639 them around calls. This is only possible if called functions are part of
8640 same compilation unit as current function and they are compiled before it.
8642 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8644 @item -fconserve-stack
8645 @opindex fconserve-stack
8646 Attempt to minimize stack usage. The compiler attempts to use less
8647 stack space, even if that makes the program slower. This option
8648 implies setting the @option{large-stack-frame} parameter to 100
8649 and the @option{large-stack-frame-growth} parameter to 400.
8651 @item -ftree-reassoc
8652 @opindex ftree-reassoc
8653 Perform reassociation on trees. This flag is enabled by default
8654 at @option{-O} and higher.
8658 Perform partial redundancy elimination (PRE) on trees. This flag is
8659 enabled by default at @option{-O2} and @option{-O3}.
8661 @item -ftree-partial-pre
8662 @opindex ftree-partial-pre
8663 Make partial redundancy elimination (PRE) more aggressive. This flag is
8664 enabled by default at @option{-O3}.
8666 @item -ftree-forwprop
8667 @opindex ftree-forwprop
8668 Perform forward propagation on trees. This flag is enabled by default
8669 at @option{-O} and higher.
8673 Perform full redundancy elimination (FRE) on trees. The difference
8674 between FRE and PRE is that FRE only considers expressions
8675 that are computed on all paths leading to the redundant computation.
8676 This analysis is faster than PRE, though it exposes fewer redundancies.
8677 This flag is enabled by default at @option{-O} and higher.
8679 @item -ftree-phiprop
8680 @opindex ftree-phiprop
8681 Perform hoisting of loads from conditional pointers on trees. This
8682 pass is enabled by default at @option{-O} and higher.
8684 @item -fhoist-adjacent-loads
8685 @opindex fhoist-adjacent-loads
8686 Speculatively hoist loads from both branches of an if-then-else if the
8687 loads are from adjacent locations in the same structure and the target
8688 architecture has a conditional move instruction. This flag is enabled
8689 by default at @option{-O2} and higher.
8691 @item -ftree-copy-prop
8692 @opindex ftree-copy-prop
8693 Perform copy propagation on trees. This pass eliminates unnecessary
8694 copy operations. This flag is enabled by default at @option{-O} and
8697 @item -fipa-pure-const
8698 @opindex fipa-pure-const
8699 Discover which functions are pure or constant.
8700 Enabled by default at @option{-O} and higher.
8702 @item -fipa-reference
8703 @opindex fipa-reference
8704 Discover which static variables do not escape the
8706 Enabled by default at @option{-O} and higher.
8710 Perform interprocedural pointer analysis and interprocedural modification
8711 and reference analysis. This option can cause excessive memory and
8712 compile-time usage on large compilation units. It is not enabled by
8713 default at any optimization level.
8716 @opindex fipa-profile
8717 Perform interprocedural profile propagation. The functions called only from
8718 cold functions are marked as cold. Also functions executed once (such as
8719 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8720 functions and loop less parts of functions executed once are then optimized for
8722 Enabled by default at @option{-O} and higher.
8726 Perform interprocedural constant propagation.
8727 This optimization analyzes the program to determine when values passed
8728 to functions are constants and then optimizes accordingly.
8729 This optimization can substantially increase performance
8730 if the application has constants passed to functions.
8731 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8733 @item -fipa-cp-clone
8734 @opindex fipa-cp-clone
8735 Perform function cloning to make interprocedural constant propagation stronger.
8736 When enabled, interprocedural constant propagation performs function cloning
8737 when externally visible function can be called with constant arguments.
8738 Because this optimization can create multiple copies of functions,
8739 it may significantly increase code size
8740 (see @option{--param ipcp-unit-growth=@var{value}}).
8741 This flag is enabled by default at @option{-O3}.
8743 @item -fipa-cp-alignment
8744 @opindex -fipa-cp-alignment
8745 When enabled, this optimization propagates alignment of function
8746 parameters to support better vectorization and string operations.
8748 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8749 requires that @option{-fipa-cp} is enabled.
8753 Perform Identical Code Folding for functions and read-only variables.
8754 The optimization reduces code size and may disturb unwind stacks by replacing
8755 a function by equivalent one with a different name. The optimization works
8756 more effectively with link time optimization enabled.
8758 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8759 works on different levels and thus the optimizations are not same - there are
8760 equivalences that are found only by GCC and equivalences found only by Gold.
8762 This flag is enabled by default at @option{-O2} and @option{-Os}.
8764 @item -fisolate-erroneous-paths-dereference
8765 @opindex fisolate-erroneous-paths-dereference
8766 Detect paths that trigger erroneous or undefined behavior due to
8767 dereferencing a null pointer. Isolate those paths from the main control
8768 flow and turn the statement with erroneous or undefined behavior into a trap.
8769 This flag is enabled by default at @option{-O2} and higher and depends on
8770 @option{-fdelete-null-pointer-checks} also being enabled.
8772 @item -fisolate-erroneous-paths-attribute
8773 @opindex fisolate-erroneous-paths-attribute
8774 Detect paths that trigger erroneous or undefined behavior due a null value
8775 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8776 attribute. Isolate those paths from the main control flow and turn the
8777 statement with erroneous or undefined behavior into a trap. This is not
8778 currently enabled, but may be enabled by @option{-O2} in the future.
8782 Perform forward store motion on trees. This flag is
8783 enabled by default at @option{-O} and higher.
8785 @item -ftree-bit-ccp
8786 @opindex ftree-bit-ccp
8787 Perform sparse conditional bit constant propagation on trees and propagate
8788 pointer alignment information.
8789 This pass only operates on local scalar variables and is enabled by default
8790 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8794 Perform sparse conditional constant propagation (CCP) on trees. This
8795 pass only operates on local scalar variables and is enabled by default
8796 at @option{-O} and higher.
8799 @opindex fssa-phiopt
8800 Perform pattern matching on SSA PHI nodes to optimize conditional
8801 code. This pass is enabled by default at @option{-O} and higher.
8803 @item -ftree-switch-conversion
8804 @opindex ftree-switch-conversion
8805 Perform conversion of simple initializations in a switch to
8806 initializations from a scalar array. This flag is enabled by default
8807 at @option{-O2} and higher.
8809 @item -ftree-tail-merge
8810 @opindex ftree-tail-merge
8811 Look for identical code sequences. When found, replace one with a jump to the
8812 other. This optimization is known as tail merging or cross jumping. This flag
8813 is enabled by default at @option{-O2} and higher. The compilation time
8815 be limited using @option{max-tail-merge-comparisons} parameter and
8816 @option{max-tail-merge-iterations} parameter.
8820 Perform dead code elimination (DCE) on trees. This flag is enabled by
8821 default at @option{-O} and higher.
8823 @item -ftree-builtin-call-dce
8824 @opindex ftree-builtin-call-dce
8825 Perform conditional dead code elimination (DCE) for calls to built-in functions
8826 that may set @code{errno} but are otherwise side-effect free. This flag is
8827 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8830 @item -ftree-dominator-opts
8831 @opindex ftree-dominator-opts
8832 Perform a variety of simple scalar cleanups (constant/copy
8833 propagation, redundancy elimination, range propagation and expression
8834 simplification) based on a dominator tree traversal. This also
8835 performs jump threading (to reduce jumps to jumps). This flag is
8836 enabled by default at @option{-O} and higher.
8840 Perform dead store elimination (DSE) on trees. A dead store is a store into
8841 a memory location that is later overwritten by another store without
8842 any intervening loads. In this case the earlier store can be deleted. This
8843 flag is enabled by default at @option{-O} and higher.
8847 Perform loop header copying on trees. This is beneficial since it increases
8848 effectiveness of code motion optimizations. It also saves one jump. This flag
8849 is enabled by default at @option{-O} and higher. It is not enabled
8850 for @option{-Os}, since it usually increases code size.
8852 @item -ftree-loop-optimize
8853 @opindex ftree-loop-optimize
8854 Perform loop optimizations on trees. This flag is enabled by default
8855 at @option{-O} and higher.
8857 @item -ftree-loop-linear
8858 @itemx -floop-interchange
8859 @itemx -floop-strip-mine
8861 @itemx -floop-unroll-and-jam
8862 @opindex ftree-loop-linear
8863 @opindex floop-interchange
8864 @opindex floop-strip-mine
8865 @opindex floop-block
8866 @opindex floop-unroll-and-jam
8867 Perform loop nest optimizations. Same as
8868 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8869 to be configured with @option{--with-isl} to enable the Graphite loop
8870 transformation infrastructure.
8872 @item -fgraphite-identity
8873 @opindex fgraphite-identity
8874 Enable the identity transformation for graphite. For every SCoP we generate
8875 the polyhedral representation and transform it back to gimple. Using
8876 @option{-fgraphite-identity} we can check the costs or benefits of the
8877 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8878 are also performed by the code generator ISL, like index splitting and
8879 dead code elimination in loops.
8881 @item -floop-nest-optimize
8882 @opindex floop-nest-optimize
8883 Enable the ISL based loop nest optimizer. This is a generic loop nest
8884 optimizer based on the Pluto optimization algorithms. It calculates a loop
8885 structure optimized for data-locality and parallelism. This option
8888 @item -floop-parallelize-all
8889 @opindex floop-parallelize-all
8890 Use the Graphite data dependence analysis to identify loops that can
8891 be parallelized. Parallelize all the loops that can be analyzed to
8892 not contain loop carried dependences without checking that it is
8893 profitable to parallelize the loops.
8895 @item -ftree-coalesce-vars
8896 @opindex ftree-coalesce-vars
8897 While transforming the program out of the SSA representation, attempt to
8898 reduce copying by coalescing versions of different user-defined
8899 variables, instead of just compiler temporaries. This may severely
8900 limit the ability to debug an optimized program compiled with
8901 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8902 prevents SSA coalescing of user variables. This option is enabled by
8903 default if optimization is enabled, and it does very little otherwise.
8905 @item -ftree-loop-if-convert
8906 @opindex ftree-loop-if-convert
8907 Attempt to transform conditional jumps in the innermost loops to
8908 branch-less equivalents. The intent is to remove control-flow from
8909 the innermost loops in order to improve the ability of the
8910 vectorization pass to handle these loops. This is enabled by default
8911 if vectorization is enabled.
8913 @item -ftree-loop-if-convert-stores
8914 @opindex ftree-loop-if-convert-stores
8915 Attempt to also if-convert conditional jumps containing memory writes.
8916 This transformation can be unsafe for multi-threaded programs as it
8917 transforms conditional memory writes into unconditional memory writes.
8920 for (i = 0; i < N; i++)
8926 for (i = 0; i < N; i++)
8927 A[i] = cond ? expr : A[i];
8929 potentially producing data races.
8931 @item -ftree-loop-distribution
8932 @opindex ftree-loop-distribution
8933 Perform loop distribution. This flag can improve cache performance on
8934 big loop bodies and allow further loop optimizations, like
8935 parallelization or vectorization, to take place. For example, the loop
8952 @item -ftree-loop-distribute-patterns
8953 @opindex ftree-loop-distribute-patterns
8954 Perform loop distribution of patterns that can be code generated with
8955 calls to a library. This flag is enabled by default at @option{-O3}.
8957 This pass distributes the initialization loops and generates a call to
8958 memset zero. For example, the loop
8974 and the initialization loop is transformed into a call to memset zero.
8976 @item -ftree-loop-im
8977 @opindex ftree-loop-im
8978 Perform loop invariant motion on trees. This pass moves only invariants that
8979 are hard to handle at RTL level (function calls, operations that expand to
8980 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8981 operands of conditions that are invariant out of the loop, so that we can use
8982 just trivial invariantness analysis in loop unswitching. The pass also includes
8985 @item -ftree-loop-ivcanon
8986 @opindex ftree-loop-ivcanon
8987 Create a canonical counter for number of iterations in loops for which
8988 determining number of iterations requires complicated analysis. Later
8989 optimizations then may determine the number easily. Useful especially
8990 in connection with unrolling.
8994 Perform induction variable optimizations (strength reduction, induction
8995 variable merging and induction variable elimination) on trees.
8997 @item -ftree-parallelize-loops=n
8998 @opindex ftree-parallelize-loops
8999 Parallelize loops, i.e., split their iteration space to run in n threads.
9000 This is only possible for loops whose iterations are independent
9001 and can be arbitrarily reordered. The optimization is only
9002 profitable on multiprocessor machines, for loops that are CPU-intensive,
9003 rather than constrained e.g.@: by memory bandwidth. This option
9004 implies @option{-pthread}, and thus is only supported on targets
9005 that have support for @option{-pthread}.
9009 Perform function-local points-to analysis on trees. This flag is
9010 enabled by default at @option{-O} and higher.
9014 Perform scalar replacement of aggregates. This pass replaces structure
9015 references with scalars to prevent committing structures to memory too
9016 early. This flag is enabled by default at @option{-O} and higher.
9020 Perform temporary expression replacement during the SSA->normal phase. Single
9021 use/single def temporaries are replaced at their use location with their
9022 defining expression. This results in non-GIMPLE code, but gives the expanders
9023 much more complex trees to work on resulting in better RTL generation. This is
9024 enabled by default at @option{-O} and higher.
9028 Perform straight-line strength reduction on trees. This recognizes related
9029 expressions involving multiplications and replaces them by less expensive
9030 calculations when possible. This is enabled by default at @option{-O} and
9033 @item -ftree-vectorize
9034 @opindex ftree-vectorize
9035 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9036 and @option{-ftree-slp-vectorize} if not explicitly specified.
9038 @item -ftree-loop-vectorize
9039 @opindex ftree-loop-vectorize
9040 Perform loop vectorization on trees. This flag is enabled by default at
9041 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9043 @item -ftree-slp-vectorize
9044 @opindex ftree-slp-vectorize
9045 Perform basic block vectorization on trees. This flag is enabled by default at
9046 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9048 @item -fvect-cost-model=@var{model}
9049 @opindex fvect-cost-model
9050 Alter the cost model used for vectorization. The @var{model} argument
9051 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9052 With the @samp{unlimited} model the vectorized code-path is assumed
9053 to be profitable while with the @samp{dynamic} model a runtime check
9054 guards the vectorized code-path to enable it only for iteration
9055 counts that will likely execute faster than when executing the original
9056 scalar loop. The @samp{cheap} model disables vectorization of
9057 loops where doing so would be cost prohibitive for example due to
9058 required runtime checks for data dependence or alignment but otherwise
9059 is equal to the @samp{dynamic} model.
9060 The default cost model depends on other optimization flags and is
9061 either @samp{dynamic} or @samp{cheap}.
9063 @item -fsimd-cost-model=@var{model}
9064 @opindex fsimd-cost-model
9065 Alter the cost model used for vectorization of loops marked with the OpenMP
9066 or Cilk Plus simd directive. The @var{model} argument should be one of
9067 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9068 have the same meaning as described in @option{-fvect-cost-model} and by
9069 default a cost model defined with @option{-fvect-cost-model} is used.
9073 Perform Value Range Propagation on trees. This is similar to the
9074 constant propagation pass, but instead of values, ranges of values are
9075 propagated. This allows the optimizers to remove unnecessary range
9076 checks like array bound checks and null pointer checks. This is
9077 enabled by default at @option{-O2} and higher. Null pointer check
9078 elimination is only done if @option{-fdelete-null-pointer-checks} is
9081 @item -fsplit-ivs-in-unroller
9082 @opindex fsplit-ivs-in-unroller
9083 Enables expression of values of induction variables in later iterations
9084 of the unrolled loop using the value in the first iteration. This breaks
9085 long dependency chains, thus improving efficiency of the scheduling passes.
9087 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9088 same effect. However, that is not reliable in cases where the loop body
9089 is more complicated than a single basic block. It also does not work at all
9090 on some architectures due to restrictions in the CSE pass.
9092 This optimization is enabled by default.
9094 @item -fvariable-expansion-in-unroller
9095 @opindex fvariable-expansion-in-unroller
9096 With this option, the compiler creates multiple copies of some
9097 local variables when unrolling a loop, which can result in superior code.
9099 @item -fpartial-inlining
9100 @opindex fpartial-inlining
9101 Inline parts of functions. This option has any effect only
9102 when inlining itself is turned on by the @option{-finline-functions}
9103 or @option{-finline-small-functions} options.
9105 Enabled at level @option{-O2}.
9107 @item -fpredictive-commoning
9108 @opindex fpredictive-commoning
9109 Perform predictive commoning optimization, i.e., reusing computations
9110 (especially memory loads and stores) performed in previous
9111 iterations of loops.
9113 This option is enabled at level @option{-O3}.
9115 @item -fprefetch-loop-arrays
9116 @opindex fprefetch-loop-arrays
9117 If supported by the target machine, generate instructions to prefetch
9118 memory to improve the performance of loops that access large arrays.
9120 This option may generate better or worse code; results are highly
9121 dependent on the structure of loops within the source code.
9123 Disabled at level @option{-Os}.
9126 @itemx -fno-peephole2
9127 @opindex fno-peephole
9128 @opindex fno-peephole2
9129 Disable any machine-specific peephole optimizations. The difference
9130 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9131 are implemented in the compiler; some targets use one, some use the
9132 other, a few use both.
9134 @option{-fpeephole} is enabled by default.
9135 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9137 @item -fno-guess-branch-probability
9138 @opindex fno-guess-branch-probability
9139 Do not guess branch probabilities using heuristics.
9141 GCC uses heuristics to guess branch probabilities if they are
9142 not provided by profiling feedback (@option{-fprofile-arcs}). These
9143 heuristics are based on the control flow graph. If some branch probabilities
9144 are specified by @code{__builtin_expect}, then the heuristics are
9145 used to guess branch probabilities for the rest of the control flow graph,
9146 taking the @code{__builtin_expect} info into account. The interactions
9147 between the heuristics and @code{__builtin_expect} can be complex, and in
9148 some cases, it may be useful to disable the heuristics so that the effects
9149 of @code{__builtin_expect} are easier to understand.
9151 The default is @option{-fguess-branch-probability} at levels
9152 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9154 @item -freorder-blocks
9155 @opindex freorder-blocks
9156 Reorder basic blocks in the compiled function in order to reduce number of
9157 taken branches and improve code locality.
9159 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9161 @item -freorder-blocks-algorithm=@var{algorithm}
9162 @opindex freorder-blocks-algorithm
9163 Use the specified algorithm for basic block reordering. The
9164 @var{algorithm} argument can be @samp{simple}, which does not increase
9165 code size (except sometimes due to secondary effects like alignment),
9166 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9167 put all often executed code together, minimizing the number of branches
9168 executed by making extra copies of code.
9170 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9171 @samp{stc} at levels @option{-O2}, @option{-O3}.
9173 @item -freorder-blocks-and-partition
9174 @opindex freorder-blocks-and-partition
9175 In addition to reordering basic blocks in the compiled function, in order
9176 to reduce number of taken branches, partitions hot and cold basic blocks
9177 into separate sections of the assembly and .o files, to improve
9178 paging and cache locality performance.
9180 This optimization is automatically turned off in the presence of
9181 exception handling, for linkonce sections, for functions with a user-defined
9182 section attribute and on any architecture that does not support named
9185 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9187 @item -freorder-functions
9188 @opindex freorder-functions
9189 Reorder functions in the object file in order to
9190 improve code locality. This is implemented by using special
9191 subsections @code{.text.hot} for most frequently executed functions and
9192 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9193 the linker so object file format must support named sections and linker must
9194 place them in a reasonable way.
9196 Also profile feedback must be available to make this option effective. See
9197 @option{-fprofile-arcs} for details.
9199 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9201 @item -fstrict-aliasing
9202 @opindex fstrict-aliasing
9203 Allow the compiler to assume the strictest aliasing rules applicable to
9204 the language being compiled. For C (and C++), this activates
9205 optimizations based on the type of expressions. In particular, an
9206 object of one type is assumed never to reside at the same address as an
9207 object of a different type, unless the types are almost the same. For
9208 example, an @code{unsigned int} can alias an @code{int}, but not a
9209 @code{void*} or a @code{double}. A character type may alias any other
9212 @anchor{Type-punning}Pay special attention to code like this:
9225 The practice of reading from a different union member than the one most
9226 recently written to (called ``type-punning'') is common. Even with
9227 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9228 is accessed through the union type. So, the code above works as
9229 expected. @xref{Structures unions enumerations and bit-fields
9230 implementation}. However, this code might not:
9241 Similarly, access by taking the address, casting the resulting pointer
9242 and dereferencing the result has undefined behavior, even if the cast
9243 uses a union type, e.g.:
9247 return ((union a_union *) &d)->i;
9251 The @option{-fstrict-aliasing} option is enabled at levels
9252 @option{-O2}, @option{-O3}, @option{-Os}.
9254 @item -fstrict-overflow
9255 @opindex fstrict-overflow
9256 Allow the compiler to assume strict signed overflow rules, depending
9257 on the language being compiled. For C (and C++) this means that
9258 overflow when doing arithmetic with signed numbers is undefined, which
9259 means that the compiler may assume that it does not happen. This
9260 permits various optimizations. For example, the compiler assumes
9261 that an expression like @code{i + 10 > i} is always true for
9262 signed @code{i}. This assumption is only valid if signed overflow is
9263 undefined, as the expression is false if @code{i + 10} overflows when
9264 using twos complement arithmetic. When this option is in effect any
9265 attempt to determine whether an operation on signed numbers
9266 overflows must be written carefully to not actually involve overflow.
9268 This option also allows the compiler to assume strict pointer
9269 semantics: given a pointer to an object, if adding an offset to that
9270 pointer does not produce a pointer to the same object, the addition is
9271 undefined. This permits the compiler to conclude that @code{p + u >
9272 p} is always true for a pointer @code{p} and unsigned integer
9273 @code{u}. This assumption is only valid because pointer wraparound is
9274 undefined, as the expression is false if @code{p + u} overflows using
9275 twos complement arithmetic.
9277 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9278 that integer signed overflow is fully defined: it wraps. When
9279 @option{-fwrapv} is used, there is no difference between
9280 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9281 integers. With @option{-fwrapv} certain types of overflow are
9282 permitted. For example, if the compiler gets an overflow when doing
9283 arithmetic on constants, the overflowed value can still be used with
9284 @option{-fwrapv}, but not otherwise.
9286 The @option{-fstrict-overflow} option is enabled at levels
9287 @option{-O2}, @option{-O3}, @option{-Os}.
9289 @item -falign-functions
9290 @itemx -falign-functions=@var{n}
9291 @opindex falign-functions
9292 Align the start of functions to the next power-of-two greater than
9293 @var{n}, skipping up to @var{n} bytes. For instance,
9294 @option{-falign-functions=32} aligns functions to the next 32-byte
9295 boundary, but @option{-falign-functions=24} aligns to the next
9296 32-byte boundary only if this can be done by skipping 23 bytes or less.
9298 @option{-fno-align-functions} and @option{-falign-functions=1} are
9299 equivalent and mean that functions are not aligned.
9301 Some assemblers only support this flag when @var{n} is a power of two;
9302 in that case, it is rounded up.
9304 If @var{n} is not specified or is zero, use a machine-dependent default.
9306 Enabled at levels @option{-O2}, @option{-O3}.
9308 @item -falign-labels
9309 @itemx -falign-labels=@var{n}
9310 @opindex falign-labels
9311 Align all branch targets to a power-of-two boundary, skipping up to
9312 @var{n} bytes like @option{-falign-functions}. This option can easily
9313 make code slower, because it must insert dummy operations for when the
9314 branch target is reached in the usual flow of the code.
9316 @option{-fno-align-labels} and @option{-falign-labels=1} are
9317 equivalent and mean that labels are not aligned.
9319 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9320 are greater than this value, then their values are used instead.
9322 If @var{n} is not specified or is zero, use a machine-dependent default
9323 which is very likely to be @samp{1}, meaning no alignment.
9325 Enabled at levels @option{-O2}, @option{-O3}.
9328 @itemx -falign-loops=@var{n}
9329 @opindex falign-loops
9330 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9331 like @option{-falign-functions}. If the loops are
9332 executed many times, this makes up for any execution of the dummy
9335 @option{-fno-align-loops} and @option{-falign-loops=1} are
9336 equivalent and mean that loops are not aligned.
9338 If @var{n} is not specified or is zero, use a machine-dependent default.
9340 Enabled at levels @option{-O2}, @option{-O3}.
9343 @itemx -falign-jumps=@var{n}
9344 @opindex falign-jumps
9345 Align branch targets to a power-of-two boundary, for branch targets
9346 where the targets can only be reached by jumping, skipping up to @var{n}
9347 bytes like @option{-falign-functions}. In this case, no dummy operations
9350 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9351 equivalent and mean that loops are not aligned.
9353 If @var{n} is not specified or is zero, use a machine-dependent default.
9355 Enabled at levels @option{-O2}, @option{-O3}.
9357 @item -funit-at-a-time
9358 @opindex funit-at-a-time
9359 This option is left for compatibility reasons. @option{-funit-at-a-time}
9360 has no effect, while @option{-fno-unit-at-a-time} implies
9361 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9365 @item -fno-toplevel-reorder
9366 @opindex fno-toplevel-reorder
9367 Do not reorder top-level functions, variables, and @code{asm}
9368 statements. Output them in the same order that they appear in the
9369 input file. When this option is used, unreferenced static variables
9370 are not removed. This option is intended to support existing code
9371 that relies on a particular ordering. For new code, it is better to
9372 use attributes when possible.
9374 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9375 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9380 Constructs webs as commonly used for register allocation purposes and assign
9381 each web individual pseudo register. This allows the register allocation pass
9382 to operate on pseudos directly, but also strengthens several other optimization
9383 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9384 however, make debugging impossible, since variables no longer stay in a
9387 Enabled by default with @option{-funroll-loops}.
9389 @item -fwhole-program
9390 @opindex fwhole-program
9391 Assume that the current compilation unit represents the whole program being
9392 compiled. All public functions and variables with the exception of @code{main}
9393 and those merged by attribute @code{externally_visible} become static functions
9394 and in effect are optimized more aggressively by interprocedural optimizers.
9396 This option should not be used in combination with @option{-flto}.
9397 Instead relying on a linker plugin should provide safer and more precise
9400 @item -flto[=@var{n}]
9402 This option runs the standard link-time optimizer. When invoked
9403 with source code, it generates GIMPLE (one of GCC's internal
9404 representations) and writes it to special ELF sections in the object
9405 file. When the object files are linked together, all the function
9406 bodies are read from these ELF sections and instantiated as if they
9407 had been part of the same translation unit.
9409 To use the link-time optimizer, @option{-flto} and optimization
9410 options should be specified at compile time and during the final link.
9414 gcc -c -O2 -flto foo.c
9415 gcc -c -O2 -flto bar.c
9416 gcc -o myprog -flto -O2 foo.o bar.o
9419 The first two invocations to GCC save a bytecode representation
9420 of GIMPLE into special ELF sections inside @file{foo.o} and
9421 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9422 @file{foo.o} and @file{bar.o}, merges the two files into a single
9423 internal image, and compiles the result as usual. Since both
9424 @file{foo.o} and @file{bar.o} are merged into a single image, this
9425 causes all the interprocedural analyses and optimizations in GCC to
9426 work across the two files as if they were a single one. This means,
9427 for example, that the inliner is able to inline functions in
9428 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9430 Another (simpler) way to enable link-time optimization is:
9433 gcc -o myprog -flto -O2 foo.c bar.c
9436 The above generates bytecode for @file{foo.c} and @file{bar.c},
9437 merges them together into a single GIMPLE representation and optimizes
9438 them as usual to produce @file{myprog}.
9440 The only important thing to keep in mind is that to enable link-time
9441 optimizations you need to use the GCC driver to perform the link-step.
9442 GCC then automatically performs link-time optimization if any of the
9443 objects involved were compiled with the @option{-flto} command-line option.
9445 should specify the optimization options to be used for link-time
9446 optimization though GCC tries to be clever at guessing an
9447 optimization level to use from the options used at compile-time
9448 if you fail to specify one at link-time. You can always override
9449 the automatic decision to do link-time optimization at link-time
9450 by passing @option{-fno-lto} to the link command.
9452 To make whole program optimization effective, it is necessary to make
9453 certain whole program assumptions. The compiler needs to know
9454 what functions and variables can be accessed by libraries and runtime
9455 outside of the link-time optimized unit. When supported by the linker,
9456 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9457 to the compiler about used and externally visible symbols. When
9458 the linker plugin is not available, @option{-fwhole-program} should be
9459 used to allow the compiler to make these assumptions, which leads
9460 to more aggressive optimization decisions.
9462 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9463 compiled with @option{-flto}, the generated object file is larger than
9464 a regular object file because it contains GIMPLE bytecodes and the usual
9465 final code (see @option{-ffat-lto-objects}. This means that
9466 object files with LTO information can be linked as normal object
9467 files; if @option{-fno-lto} is passed to the linker, no
9468 interprocedural optimizations are applied. Note that when
9469 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9470 but you cannot perform a regular, non-LTO link on them.
9472 Additionally, the optimization flags used to compile individual files
9473 are not necessarily related to those used at link time. For instance,
9476 gcc -c -O0 -ffat-lto-objects -flto foo.c
9477 gcc -c -O0 -ffat-lto-objects -flto bar.c
9478 gcc -o myprog -O3 foo.o bar.o
9481 This produces individual object files with unoptimized assembler
9482 code, but the resulting binary @file{myprog} is optimized at
9483 @option{-O3}. If, instead, the final binary is generated with
9484 @option{-fno-lto}, then @file{myprog} is not optimized.
9486 When producing the final binary, GCC only
9487 applies link-time optimizations to those files that contain bytecode.
9488 Therefore, you can mix and match object files and libraries with
9489 GIMPLE bytecodes and final object code. GCC automatically selects
9490 which files to optimize in LTO mode and which files to link without
9493 There are some code generation flags preserved by GCC when
9494 generating bytecodes, as they need to be used during the final link
9495 stage. Generally options specified at link-time override those
9496 specified at compile-time.
9498 If you do not specify an optimization level option @option{-O} at
9499 link-time then GCC computes one based on the optimization levels
9500 used when compiling the object files. The highest optimization
9503 Currently, the following options and their setting are take from
9504 the first object file that explicitely specified it:
9505 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9506 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9507 and all the @option{-m} target flags.
9509 Certain ABI changing flags are required to match in all compilation-units
9510 and trying to override this at link-time with a conflicting value
9511 is ignored. This includes options such as @option{-freg-struct-return}
9512 and @option{-fpcc-struct-return}.
9514 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9515 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9516 are passed through to the link stage and merged conservatively for
9517 conflicting translation units. Specifically
9518 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9519 precedence and for example @option{-ffp-contract=off} takes precedence
9520 over @option{-ffp-contract=fast}. You can override them at linke-time.
9522 It is recommended that you compile all the files participating in the
9523 same link with the same options and also specify those options at
9526 If LTO encounters objects with C linkage declared with incompatible
9527 types in separate translation units to be linked together (undefined
9528 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9529 issued. The behavior is still undefined at run time. Similar
9530 diagnostics may be raised for other languages.
9532 Another feature of LTO is that it is possible to apply interprocedural
9533 optimizations on files written in different languages:
9538 gfortran -c -flto baz.f90
9539 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9542 Notice that the final link is done with @command{g++} to get the C++
9543 runtime libraries and @option{-lgfortran} is added to get the Fortran
9544 runtime libraries. In general, when mixing languages in LTO mode, you
9545 should use the same link command options as when mixing languages in a
9546 regular (non-LTO) compilation.
9548 If object files containing GIMPLE bytecode are stored in a library archive, say
9549 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9550 are using a linker with plugin support. To create static libraries suitable
9551 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9552 and @command{ranlib};
9553 to show the symbols of object files with GIMPLE bytecode, use
9554 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9555 and @command{nm} have been compiled with plugin support. At link time, use the the
9556 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9557 the LTO optimization process:
9560 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9563 With the linker plugin enabled, the linker extracts the needed
9564 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9565 to make them part of the aggregated GIMPLE image to be optimized.
9567 If you are not using a linker with plugin support and/or do not
9568 enable the linker plugin, then the objects inside @file{libfoo.a}
9569 are extracted and linked as usual, but they do not participate
9570 in the LTO optimization process. In order to make a static library suitable
9571 for both LTO optimization and usual linkage, compile its object files with
9572 @option{-flto} @option{-ffat-lto-objects}.
9574 Link-time optimizations do not require the presence of the whole program to
9575 operate. If the program does not require any symbols to be exported, it is
9576 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9577 the interprocedural optimizers to use more aggressive assumptions which may
9578 lead to improved optimization opportunities.
9579 Use of @option{-fwhole-program} is not needed when linker plugin is
9580 active (see @option{-fuse-linker-plugin}).
9582 The current implementation of LTO makes no
9583 attempt to generate bytecode that is portable between different
9584 types of hosts. The bytecode files are versioned and there is a
9585 strict version check, so bytecode files generated in one version of
9586 GCC do not work with an older or newer version of GCC.
9588 Link-time optimization does not work well with generation of debugging
9589 information. Combining @option{-flto} with
9590 @option{-g} is currently experimental and expected to produce unexpected
9593 If you specify the optional @var{n}, the optimization and code
9594 generation done at link time is executed in parallel using @var{n}
9595 parallel jobs by utilizing an installed @command{make} program. The
9596 environment variable @env{MAKE} may be used to override the program
9597 used. The default value for @var{n} is 1.
9599 You can also specify @option{-flto=jobserver} to use GNU make's
9600 job server mode to determine the number of parallel jobs. This
9601 is useful when the Makefile calling GCC is already executing in parallel.
9602 You must prepend a @samp{+} to the command recipe in the parent Makefile
9603 for this to work. This option likely only works if @env{MAKE} is
9606 @item -flto-partition=@var{alg}
9607 @opindex flto-partition
9608 Specify the partitioning algorithm used by the link-time optimizer.
9609 The value is either @samp{1to1} to specify a partitioning mirroring
9610 the original source files or @samp{balanced} to specify partitioning
9611 into equally sized chunks (whenever possible) or @samp{max} to create
9612 new partition for every symbol where possible. Specifying @samp{none}
9613 as an algorithm disables partitioning and streaming completely.
9614 The default value is @samp{balanced}. While @samp{1to1} can be used
9615 as an workaround for various code ordering issues, the @samp{max}
9616 partitioning is intended for internal testing only.
9617 The value @samp{one} specifies that exactly one partition should be
9618 used while the value @samp{none} bypasses partitioning and executes
9619 the link-time optimization step directly from the WPA phase.
9621 @item -flto-odr-type-merging
9622 @opindex flto-odr-type-merging
9623 Enable streaming of mangled types names of C++ types and their unification
9624 at linktime. This increases size of LTO object files, but enable
9625 diagnostics about One Definition Rule violations.
9627 @item -flto-compression-level=@var{n}
9628 @opindex flto-compression-level
9629 This option specifies the level of compression used for intermediate
9630 language written to LTO object files, and is only meaningful in
9631 conjunction with LTO mode (@option{-flto}). Valid
9632 values are 0 (no compression) to 9 (maximum compression). Values
9633 outside this range are clamped to either 0 or 9. If the option is not
9634 given, a default balanced compression setting is used.
9637 @opindex flto-report
9638 Prints a report with internal details on the workings of the link-time
9639 optimizer. The contents of this report vary from version to version.
9640 It is meant to be useful to GCC developers when processing object
9641 files in LTO mode (via @option{-flto}).
9643 Disabled by default.
9645 @item -flto-report-wpa
9646 @opindex flto-report-wpa
9647 Like @option{-flto-report}, but only print for the WPA phase of Link
9650 @item -fuse-linker-plugin
9651 @opindex fuse-linker-plugin
9652 Enables the use of a linker plugin during link-time optimization. This
9653 option relies on plugin support in the linker, which is available in gold
9654 or in GNU ld 2.21 or newer.
9656 This option enables the extraction of object files with GIMPLE bytecode out
9657 of library archives. This improves the quality of optimization by exposing
9658 more code to the link-time optimizer. This information specifies what
9659 symbols can be accessed externally (by non-LTO object or during dynamic
9660 linking). Resulting code quality improvements on binaries (and shared
9661 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9662 See @option{-flto} for a description of the effect of this flag and how to
9665 This option is enabled by default when LTO support in GCC is enabled
9666 and GCC was configured for use with
9667 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9669 @item -ffat-lto-objects
9670 @opindex ffat-lto-objects
9671 Fat LTO objects are object files that contain both the intermediate language
9672 and the object code. This makes them usable for both LTO linking and normal
9673 linking. This option is effective only when compiling with @option{-flto}
9674 and is ignored at link time.
9676 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9677 requires the complete toolchain to be aware of LTO. It requires a linker with
9678 linker plugin support for basic functionality. Additionally,
9679 @command{nm}, @command{ar} and @command{ranlib}
9680 need to support linker plugins to allow a full-featured build environment
9681 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9682 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9683 to these tools. With non fat LTO makefiles need to be modified to use them.
9685 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9688 @item -fcompare-elim
9689 @opindex fcompare-elim
9690 After register allocation and post-register allocation instruction splitting,
9691 identify arithmetic instructions that compute processor flags similar to a
9692 comparison operation based on that arithmetic. If possible, eliminate the
9693 explicit comparison operation.
9695 This pass only applies to certain targets that cannot explicitly represent
9696 the comparison operation before register allocation is complete.
9698 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9700 @item -fcprop-registers
9701 @opindex fcprop-registers
9702 After register allocation and post-register allocation instruction splitting,
9703 perform a copy-propagation pass to try to reduce scheduling dependencies
9704 and occasionally eliminate the copy.
9706 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9708 @item -fprofile-correction
9709 @opindex fprofile-correction
9710 Profiles collected using an instrumented binary for multi-threaded programs may
9711 be inconsistent due to missed counter updates. When this option is specified,
9712 GCC uses heuristics to correct or smooth out such inconsistencies. By
9713 default, GCC emits an error message when an inconsistent profile is detected.
9715 @item -fprofile-dir=@var{path}
9716 @opindex fprofile-dir
9718 Set the directory to search for the profile data files in to @var{path}.
9719 This option affects only the profile data generated by
9720 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9721 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9722 and its related options. Both absolute and relative paths can be used.
9723 By default, GCC uses the current directory as @var{path}, thus the
9724 profile data file appears in the same directory as the object file.
9726 @item -fprofile-generate
9727 @itemx -fprofile-generate=@var{path}
9728 @opindex fprofile-generate
9730 Enable options usually used for instrumenting application to produce
9731 profile useful for later recompilation with profile feedback based
9732 optimization. You must use @option{-fprofile-generate} both when
9733 compiling and when linking your program.
9735 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9737 If @var{path} is specified, GCC looks at the @var{path} to find
9738 the profile feedback data files. See @option{-fprofile-dir}.
9741 @itemx -fprofile-use=@var{path}
9742 @opindex fprofile-use
9743 Enable profile feedback-directed optimizations,
9744 and the following optimizations
9745 which are generally profitable only with profile feedback available:
9746 @option{-fbranch-probabilities}, @option{-fvpt},
9747 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9748 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9750 By default, GCC emits an error message if the feedback profiles do not
9751 match the source code. This error can be turned into a warning by using
9752 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9755 If @var{path} is specified, GCC looks at the @var{path} to find
9756 the profile feedback data files. See @option{-fprofile-dir}.
9758 @item -fauto-profile
9759 @itemx -fauto-profile=@var{path}
9760 @opindex fauto-profile
9761 Enable sampling-based feedback-directed optimizations,
9762 and the following optimizations
9763 which are generally profitable only with profile feedback available:
9764 @option{-fbranch-probabilities}, @option{-fvpt},
9765 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9766 @option{-ftree-vectorize},
9767 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9768 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9769 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9771 @var{path} is the name of a file containing AutoFDO profile information.
9772 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9774 Producing an AutoFDO profile data file requires running your program
9775 with the @command{perf} utility on a supported GNU/Linux target system.
9776 For more information, see @uref{https://perf.wiki.kernel.org/}.
9780 perf record -e br_inst_retired:near_taken -b -o perf.data \
9784 Then use the @command{create_gcov} tool to convert the raw profile data
9785 to a format that can be used by GCC.@ You must also supply the
9786 unstripped binary for your program to this tool.
9787 See @uref{https://github.com/google/autofdo}.
9791 create_gcov --binary=your_program.unstripped --profile=perf.data \
9796 The following options control compiler behavior regarding floating-point
9797 arithmetic. These options trade off between speed and
9798 correctness. All must be specifically enabled.
9802 @opindex ffloat-store
9803 Do not store floating-point variables in registers, and inhibit other
9804 options that might change whether a floating-point value is taken from a
9807 @cindex floating-point precision
9808 This option prevents undesirable excess precision on machines such as
9809 the 68000 where the floating registers (of the 68881) keep more
9810 precision than a @code{double} is supposed to have. Similarly for the
9811 x86 architecture. For most programs, the excess precision does only
9812 good, but a few programs rely on the precise definition of IEEE floating
9813 point. Use @option{-ffloat-store} for such programs, after modifying
9814 them to store all pertinent intermediate computations into variables.
9816 @item -fexcess-precision=@var{style}
9817 @opindex fexcess-precision
9818 This option allows further control over excess precision on machines
9819 where floating-point registers have more precision than the IEEE
9820 @code{float} and @code{double} types and the processor does not
9821 support operations rounding to those types. By default,
9822 @option{-fexcess-precision=fast} is in effect; this means that
9823 operations are carried out in the precision of the registers and that
9824 it is unpredictable when rounding to the types specified in the source
9825 code takes place. When compiling C, if
9826 @option{-fexcess-precision=standard} is specified then excess
9827 precision follows the rules specified in ISO C99; in particular,
9828 both casts and assignments cause values to be rounded to their
9829 semantic types (whereas @option{-ffloat-store} only affects
9830 assignments). This option is enabled by default for C if a strict
9831 conformance option such as @option{-std=c99} is used.
9834 @option{-fexcess-precision=standard} is not implemented for languages
9835 other than C, and has no effect if
9836 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9837 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9838 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9839 semantics apply without excess precision, and in the latter, rounding
9844 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9845 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9846 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9848 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9850 This option is not turned on by any @option{-O} option besides
9851 @option{-Ofast} since it can result in incorrect output for programs
9852 that depend on an exact implementation of IEEE or ISO rules/specifications
9853 for math functions. It may, however, yield faster code for programs
9854 that do not require the guarantees of these specifications.
9856 @item -fno-math-errno
9857 @opindex fno-math-errno
9858 Do not set @code{errno} after calling math functions that are executed
9859 with a single instruction, e.g., @code{sqrt}. A program that relies on
9860 IEEE exceptions for math error handling may want to use this flag
9861 for speed while maintaining IEEE arithmetic compatibility.
9863 This option is not turned on by any @option{-O} option since
9864 it can result in incorrect output for programs that depend on
9865 an exact implementation of IEEE or ISO rules/specifications for
9866 math functions. It may, however, yield faster code for programs
9867 that do not require the guarantees of these specifications.
9869 The default is @option{-fmath-errno}.
9871 On Darwin systems, the math library never sets @code{errno}. There is
9872 therefore no reason for the compiler to consider the possibility that
9873 it might, and @option{-fno-math-errno} is the default.
9875 @item -funsafe-math-optimizations
9876 @opindex funsafe-math-optimizations
9878 Allow optimizations for floating-point arithmetic that (a) assume
9879 that arguments and results are valid and (b) may violate IEEE or
9880 ANSI standards. When used at link-time, it may include libraries
9881 or startup files that change the default FPU control word or other
9882 similar optimizations.
9884 This option is not turned on by any @option{-O} option since
9885 it can result in incorrect output for programs that depend on
9886 an exact implementation of IEEE or ISO rules/specifications for
9887 math functions. It may, however, yield faster code for programs
9888 that do not require the guarantees of these specifications.
9889 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9890 @option{-fassociative-math} and @option{-freciprocal-math}.
9892 The default is @option{-fno-unsafe-math-optimizations}.
9894 @item -fassociative-math
9895 @opindex fassociative-math
9897 Allow re-association of operands in series of floating-point operations.
9898 This violates the ISO C and C++ language standard by possibly changing
9899 computation result. NOTE: re-ordering may change the sign of zero as
9900 well as ignore NaNs and inhibit or create underflow or overflow (and
9901 thus cannot be used on code that relies on rounding behavior like
9902 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9903 and thus may not be used when ordered comparisons are required.
9904 This option requires that both @option{-fno-signed-zeros} and
9905 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9906 much sense with @option{-frounding-math}. For Fortran the option
9907 is automatically enabled when both @option{-fno-signed-zeros} and
9908 @option{-fno-trapping-math} are in effect.
9910 The default is @option{-fno-associative-math}.
9912 @item -freciprocal-math
9913 @opindex freciprocal-math
9915 Allow the reciprocal of a value to be used instead of dividing by
9916 the value if this enables optimizations. For example @code{x / y}
9917 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9918 is subject to common subexpression elimination. Note that this loses
9919 precision and increases the number of flops operating on the value.
9921 The default is @option{-fno-reciprocal-math}.
9923 @item -ffinite-math-only
9924 @opindex ffinite-math-only
9925 Allow optimizations for floating-point arithmetic that assume
9926 that arguments and results are not NaNs or +-Infs.
9928 This option is not turned on by any @option{-O} option since
9929 it can result in incorrect output for programs that depend on
9930 an exact implementation of IEEE or ISO rules/specifications for
9931 math functions. It may, however, yield faster code for programs
9932 that do not require the guarantees of these specifications.
9934 The default is @option{-fno-finite-math-only}.
9936 @item -fno-signed-zeros
9937 @opindex fno-signed-zeros
9938 Allow optimizations for floating-point arithmetic that ignore the
9939 signedness of zero. IEEE arithmetic specifies the behavior of
9940 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9941 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9942 This option implies that the sign of a zero result isn't significant.
9944 The default is @option{-fsigned-zeros}.
9946 @item -fno-trapping-math
9947 @opindex fno-trapping-math
9948 Compile code assuming that floating-point operations cannot generate
9949 user-visible traps. These traps include division by zero, overflow,
9950 underflow, inexact result and invalid operation. This option requires
9951 that @option{-fno-signaling-nans} be in effect. Setting this option may
9952 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9954 This option should never be turned on by any @option{-O} option since
9955 it can result in incorrect output for programs that depend on
9956 an exact implementation of IEEE or ISO rules/specifications for
9959 The default is @option{-ftrapping-math}.
9961 @item -frounding-math
9962 @opindex frounding-math
9963 Disable transformations and optimizations that assume default floating-point
9964 rounding behavior. This is round-to-zero for all floating point
9965 to integer conversions, and round-to-nearest for all other arithmetic
9966 truncations. This option should be specified for programs that change
9967 the FP rounding mode dynamically, or that may be executed with a
9968 non-default rounding mode. This option disables constant folding of
9969 floating-point expressions at compile time (which may be affected by
9970 rounding mode) and arithmetic transformations that are unsafe in the
9971 presence of sign-dependent rounding modes.
9973 The default is @option{-fno-rounding-math}.
9975 This option is experimental and does not currently guarantee to
9976 disable all GCC optimizations that are affected by rounding mode.
9977 Future versions of GCC may provide finer control of this setting
9978 using C99's @code{FENV_ACCESS} pragma. This command-line option
9979 will be used to specify the default state for @code{FENV_ACCESS}.
9981 @item -fsignaling-nans
9982 @opindex fsignaling-nans
9983 Compile code assuming that IEEE signaling NaNs may generate user-visible
9984 traps during floating-point operations. Setting this option disables
9985 optimizations that may change the number of exceptions visible with
9986 signaling NaNs. This option implies @option{-ftrapping-math}.
9988 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9991 The default is @option{-fno-signaling-nans}.
9993 This option is experimental and does not currently guarantee to
9994 disable all GCC optimizations that affect signaling NaN behavior.
9996 @item -fsingle-precision-constant
9997 @opindex fsingle-precision-constant
9998 Treat floating-point constants as single precision instead of
9999 implicitly converting them to double-precision constants.
10001 @item -fcx-limited-range
10002 @opindex fcx-limited-range
10003 When enabled, this option states that a range reduction step is not
10004 needed when performing complex division. Also, there is no checking
10005 whether the result of a complex multiplication or division is @code{NaN
10006 + I*NaN}, with an attempt to rescue the situation in that case. The
10007 default is @option{-fno-cx-limited-range}, but is enabled by
10008 @option{-ffast-math}.
10010 This option controls the default setting of the ISO C99
10011 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10014 @item -fcx-fortran-rules
10015 @opindex fcx-fortran-rules
10016 Complex multiplication and division follow Fortran rules. Range
10017 reduction is done as part of complex division, but there is no checking
10018 whether the result of a complex multiplication or division is @code{NaN
10019 + I*NaN}, with an attempt to rescue the situation in that case.
10021 The default is @option{-fno-cx-fortran-rules}.
10025 The following options control optimizations that may improve
10026 performance, but are not enabled by any @option{-O} options. This
10027 section includes experimental options that may produce broken code.
10030 @item -fbranch-probabilities
10031 @opindex fbranch-probabilities
10032 After running a program compiled with @option{-fprofile-arcs}
10033 (@pxref{Debugging Options,, Options for Debugging Your Program or
10034 @command{gcc}}), you can compile it a second time using
10035 @option{-fbranch-probabilities}, to improve optimizations based on
10036 the number of times each branch was taken. When a program
10037 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10038 counts to a file called @file{@var{sourcename}.gcda} for each source
10039 file. The information in this data file is very dependent on the
10040 structure of the generated code, so you must use the same source code
10041 and the same optimization options for both compilations.
10043 With @option{-fbranch-probabilities}, GCC puts a
10044 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10045 These can be used to improve optimization. Currently, they are only
10046 used in one place: in @file{reorg.c}, instead of guessing which path a
10047 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10048 exactly determine which path is taken more often.
10050 @item -fprofile-values
10051 @opindex fprofile-values
10052 If combined with @option{-fprofile-arcs}, it adds code so that some
10053 data about values of expressions in the program is gathered.
10055 With @option{-fbranch-probabilities}, it reads back the data gathered
10056 from profiling values of expressions for usage in optimizations.
10058 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
10060 @item -fprofile-reorder-functions
10061 @opindex fprofile-reorder-functions
10062 Function reordering based on profile instrumentation collects
10063 first time of execution of a function and orders these functions
10064 in ascending order.
10066 Enabled with @option{-fprofile-use}.
10070 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10071 to add code to gather information about values of expressions.
10073 With @option{-fbranch-probabilities}, it reads back the data gathered
10074 and actually performs the optimizations based on them.
10075 Currently the optimizations include specialization of division operations
10076 using the knowledge about the value of the denominator.
10078 @item -frename-registers
10079 @opindex frename-registers
10080 Attempt to avoid false dependencies in scheduled code by making use
10081 of registers left over after register allocation. This optimization
10082 most benefits processors with lots of registers. Depending on the
10083 debug information format adopted by the target, however, it can
10084 make debugging impossible, since variables no longer stay in
10085 a ``home register''.
10087 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
10089 @item -fschedule-fusion
10090 @opindex fschedule-fusion
10091 Performs a target dependent pass over the instruction stream to schedule
10092 instructions of same type together because target machine can execute them
10093 more efficiently if they are adjacent to each other in the instruction flow.
10095 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10099 Perform tail duplication to enlarge superblock size. This transformation
10100 simplifies the control flow of the function allowing other optimizations to do
10103 Enabled with @option{-fprofile-use}.
10105 @item -funroll-loops
10106 @opindex funroll-loops
10107 Unroll loops whose number of iterations can be determined at compile time or
10108 upon entry to the loop. @option{-funroll-loops} implies
10109 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10110 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10111 a small constant number of iterations). This option makes code larger, and may
10112 or may not make it run faster.
10114 Enabled with @option{-fprofile-use}.
10116 @item -funroll-all-loops
10117 @opindex funroll-all-loops
10118 Unroll all loops, even if their number of iterations is uncertain when
10119 the loop is entered. This usually makes programs run more slowly.
10120 @option{-funroll-all-loops} implies the same options as
10121 @option{-funroll-loops}.
10124 @opindex fpeel-loops
10125 Peels loops for which there is enough information that they do not
10126 roll much (from profile feedback). It also turns on complete loop peeling
10127 (i.e.@: complete removal of loops with small constant number of iterations).
10129 Enabled with @option{-fprofile-use}.
10131 @item -fmove-loop-invariants
10132 @opindex fmove-loop-invariants
10133 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10134 at level @option{-O1}
10136 @item -funswitch-loops
10137 @opindex funswitch-loops
10138 Move branches with loop invariant conditions out of the loop, with duplicates
10139 of the loop on both branches (modified according to result of the condition).
10141 @item -ffunction-sections
10142 @itemx -fdata-sections
10143 @opindex ffunction-sections
10144 @opindex fdata-sections
10145 Place each function or data item into its own section in the output
10146 file if the target supports arbitrary sections. The name of the
10147 function or the name of the data item determines the section's name
10148 in the output file.
10150 Use these options on systems where the linker can perform optimizations
10151 to improve locality of reference in the instruction space. Most systems
10152 using the ELF object format and SPARC processors running Solaris 2 have
10153 linkers with such optimizations. AIX may have these optimizations in
10156 Only use these options when there are significant benefits from doing
10157 so. When you specify these options, the assembler and linker
10158 create larger object and executable files and are also slower.
10159 You cannot use @command{gprof} on all systems if you
10160 specify this option, and you may have problems with debugging if
10161 you specify both this option and @option{-g}.
10163 @item -fbranch-target-load-optimize
10164 @opindex fbranch-target-load-optimize
10165 Perform branch target register load optimization before prologue / epilogue
10167 The use of target registers can typically be exposed only during reload,
10168 thus hoisting loads out of loops and doing inter-block scheduling needs
10169 a separate optimization pass.
10171 @item -fbranch-target-load-optimize2
10172 @opindex fbranch-target-load-optimize2
10173 Perform branch target register load optimization after prologue / epilogue
10176 @item -fbtr-bb-exclusive
10177 @opindex fbtr-bb-exclusive
10178 When performing branch target register load optimization, don't reuse
10179 branch target registers within any basic block.
10181 @item -fstack-protector
10182 @opindex fstack-protector
10183 Emit extra code to check for buffer overflows, such as stack smashing
10184 attacks. This is done by adding a guard variable to functions with
10185 vulnerable objects. This includes functions that call @code{alloca}, and
10186 functions with buffers larger than 8 bytes. The guards are initialized
10187 when a function is entered and then checked when the function exits.
10188 If a guard check fails, an error message is printed and the program exits.
10190 @item -fstack-protector-all
10191 @opindex fstack-protector-all
10192 Like @option{-fstack-protector} except that all functions are protected.
10194 @item -fstack-protector-strong
10195 @opindex fstack-protector-strong
10196 Like @option{-fstack-protector} but includes additional functions to
10197 be protected --- those that have local array definitions, or have
10198 references to local frame addresses.
10200 @item -fstack-protector-explicit
10201 @opindex fstack-protector-explicit
10202 Like @option{-fstack-protector} but only protects those functions which
10203 have the @code{stack_protect} attribute
10206 @opindex fstdarg-opt
10207 Optimize the prologue of variadic argument functions with respect to usage of
10210 @item -fsection-anchors
10211 @opindex fsection-anchors
10212 Try to reduce the number of symbolic address calculations by using
10213 shared ``anchor'' symbols to address nearby objects. This transformation
10214 can help to reduce the number of GOT entries and GOT accesses on some
10217 For example, the implementation of the following function @code{foo}:
10220 static int a, b, c;
10221 int foo (void) @{ return a + b + c; @}
10225 usually calculates the addresses of all three variables, but if you
10226 compile it with @option{-fsection-anchors}, it accesses the variables
10227 from a common anchor point instead. The effect is similar to the
10228 following pseudocode (which isn't valid C):
10233 register int *xr = &x;
10234 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10238 Not all targets support this option.
10240 @item --param @var{name}=@var{value}
10242 In some places, GCC uses various constants to control the amount of
10243 optimization that is done. For example, GCC does not inline functions
10244 that contain more than a certain number of instructions. You can
10245 control some of these constants on the command line using the
10246 @option{--param} option.
10248 The names of specific parameters, and the meaning of the values, are
10249 tied to the internals of the compiler, and are subject to change
10250 without notice in future releases.
10252 In each case, the @var{value} is an integer. The allowable choices for
10256 @item predictable-branch-outcome
10257 When branch is predicted to be taken with probability lower than this threshold
10258 (in percent), then it is considered well predictable. The default is 10.
10260 @item max-crossjump-edges
10261 The maximum number of incoming edges to consider for cross-jumping.
10262 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10263 the number of edges incoming to each block. Increasing values mean
10264 more aggressive optimization, making the compilation time increase with
10265 probably small improvement in executable size.
10267 @item min-crossjump-insns
10268 The minimum number of instructions that must be matched at the end
10269 of two blocks before cross-jumping is performed on them. This
10270 value is ignored in the case where all instructions in the block being
10271 cross-jumped from are matched. The default value is 5.
10273 @item max-grow-copy-bb-insns
10274 The maximum code size expansion factor when copying basic blocks
10275 instead of jumping. The expansion is relative to a jump instruction.
10276 The default value is 8.
10278 @item max-goto-duplication-insns
10279 The maximum number of instructions to duplicate to a block that jumps
10280 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10281 passes, GCC factors computed gotos early in the compilation process,
10282 and unfactors them as late as possible. Only computed jumps at the
10283 end of a basic blocks with no more than max-goto-duplication-insns are
10284 unfactored. The default value is 8.
10286 @item max-delay-slot-insn-search
10287 The maximum number of instructions to consider when looking for an
10288 instruction to fill a delay slot. If more than this arbitrary number of
10289 instructions are searched, the time savings from filling the delay slot
10290 are minimal, so stop searching. Increasing values mean more
10291 aggressive optimization, making the compilation time increase with probably
10292 small improvement in execution time.
10294 @item max-delay-slot-live-search
10295 When trying to fill delay slots, the maximum number of instructions to
10296 consider when searching for a block with valid live register
10297 information. Increasing this arbitrarily chosen value means more
10298 aggressive optimization, increasing the compilation time. This parameter
10299 should be removed when the delay slot code is rewritten to maintain the
10300 control-flow graph.
10302 @item max-gcse-memory
10303 The approximate maximum amount of memory that can be allocated in
10304 order to perform the global common subexpression elimination
10305 optimization. If more memory than specified is required, the
10306 optimization is not done.
10308 @item max-gcse-insertion-ratio
10309 If the ratio of expression insertions to deletions is larger than this value
10310 for any expression, then RTL PRE inserts or removes the expression and thus
10311 leaves partially redundant computations in the instruction stream. The default value is 20.
10313 @item max-pending-list-length
10314 The maximum number of pending dependencies scheduling allows
10315 before flushing the current state and starting over. Large functions
10316 with few branches or calls can create excessively large lists which
10317 needlessly consume memory and resources.
10319 @item max-modulo-backtrack-attempts
10320 The maximum number of backtrack attempts the scheduler should make
10321 when modulo scheduling a loop. Larger values can exponentially increase
10324 @item max-inline-insns-single
10325 Several parameters control the tree inliner used in GCC@.
10326 This number sets the maximum number of instructions (counted in GCC's
10327 internal representation) in a single function that the tree inliner
10328 considers for inlining. This only affects functions declared
10329 inline and methods implemented in a class declaration (C++).
10330 The default value is 400.
10332 @item max-inline-insns-auto
10333 When you use @option{-finline-functions} (included in @option{-O3}),
10334 a lot of functions that would otherwise not be considered for inlining
10335 by the compiler are investigated. To those functions, a different
10336 (more restrictive) limit compared to functions declared inline can
10338 The default value is 40.
10340 @item inline-min-speedup
10341 When estimated performance improvement of caller + callee runtime exceeds this
10342 threshold (in precent), the function can be inlined regardless the limit on
10343 @option{--param max-inline-insns-single} and @option{--param
10344 max-inline-insns-auto}.
10346 @item large-function-insns
10347 The limit specifying really large functions. For functions larger than this
10348 limit after inlining, inlining is constrained by
10349 @option{--param large-function-growth}. This parameter is useful primarily
10350 to avoid extreme compilation time caused by non-linear algorithms used by the
10352 The default value is 2700.
10354 @item large-function-growth
10355 Specifies maximal growth of large function caused by inlining in percents.
10356 The default value is 100 which limits large function growth to 2.0 times
10359 @item large-unit-insns
10360 The limit specifying large translation unit. Growth caused by inlining of
10361 units larger than this limit is limited by @option{--param inline-unit-growth}.
10362 For small units this might be too tight.
10363 For example, consider a unit consisting of function A
10364 that is inline and B that just calls A three times. If B is small relative to
10365 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10366 large units consisting of small inlineable functions, however, the overall unit
10367 growth limit is needed to avoid exponential explosion of code size. Thus for
10368 smaller units, the size is increased to @option{--param large-unit-insns}
10369 before applying @option{--param inline-unit-growth}. The default is 10000.
10371 @item inline-unit-growth
10372 Specifies maximal overall growth of the compilation unit caused by inlining.
10373 The default value is 20 which limits unit growth to 1.2 times the original
10374 size. Cold functions (either marked cold via an attribute or by profile
10375 feedback) are not accounted into the unit size.
10377 @item ipcp-unit-growth
10378 Specifies maximal overall growth of the compilation unit caused by
10379 interprocedural constant propagation. The default value is 10 which limits
10380 unit growth to 1.1 times the original size.
10382 @item large-stack-frame
10383 The limit specifying large stack frames. While inlining the algorithm is trying
10384 to not grow past this limit too much. The default value is 256 bytes.
10386 @item large-stack-frame-growth
10387 Specifies maximal growth of large stack frames caused by inlining in percents.
10388 The default value is 1000 which limits large stack frame growth to 11 times
10391 @item max-inline-insns-recursive
10392 @itemx max-inline-insns-recursive-auto
10393 Specifies the maximum number of instructions an out-of-line copy of a
10394 self-recursive inline
10395 function can grow into by performing recursive inlining.
10397 @option{--param max-inline-insns-recursive} applies to functions
10399 For functions not declared inline, recursive inlining
10400 happens only when @option{-finline-functions} (included in @option{-O3}) is
10401 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10402 default value is 450.
10404 @item max-inline-recursive-depth
10405 @itemx max-inline-recursive-depth-auto
10406 Specifies the maximum recursion depth used for recursive inlining.
10408 @option{--param max-inline-recursive-depth} applies to functions
10409 declared inline. For functions not declared inline, recursive inlining
10410 happens only when @option{-finline-functions} (included in @option{-O3}) is
10411 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10412 default value is 8.
10414 @item min-inline-recursive-probability
10415 Recursive inlining is profitable only for function having deep recursion
10416 in average and can hurt for function having little recursion depth by
10417 increasing the prologue size or complexity of function body to other
10420 When profile feedback is available (see @option{-fprofile-generate}) the actual
10421 recursion depth can be guessed from probability that function recurses via a
10422 given call expression. This parameter limits inlining only to call expressions
10423 whose probability exceeds the given threshold (in percents).
10424 The default value is 10.
10426 @item early-inlining-insns
10427 Specify growth that the early inliner can make. In effect it increases
10428 the amount of inlining for code having a large abstraction penalty.
10429 The default value is 14.
10431 @item max-early-inliner-iterations
10432 Limit of iterations of the early inliner. This basically bounds
10433 the number of nested indirect calls the early inliner can resolve.
10434 Deeper chains are still handled by late inlining.
10436 @item comdat-sharing-probability
10437 Probability (in percent) that C++ inline function with comdat visibility
10438 are shared across multiple compilation units. The default value is 20.
10440 @item profile-func-internal-id
10441 A parameter to control whether to use function internal id in profile
10442 database lookup. If the value is 0, the compiler uses an id that
10443 is based on function assembler name and filename, which makes old profile
10444 data more tolerant to source changes such as function reordering etc.
10445 The default value is 0.
10447 @item min-vect-loop-bound
10448 The minimum number of iterations under which loops are not vectorized
10449 when @option{-ftree-vectorize} is used. The number of iterations after
10450 vectorization needs to be greater than the value specified by this option
10451 to allow vectorization. The default value is 0.
10453 @item gcse-cost-distance-ratio
10454 Scaling factor in calculation of maximum distance an expression
10455 can be moved by GCSE optimizations. This is currently supported only in the
10456 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10457 is with simple expressions, i.e., the expressions that have cost
10458 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10459 hoisting of simple expressions. The default value is 10.
10461 @item gcse-unrestricted-cost
10462 Cost, roughly measured as the cost of a single typical machine
10463 instruction, at which GCSE optimizations do not constrain
10464 the distance an expression can travel. This is currently
10465 supported only in the code hoisting pass. The lesser the cost,
10466 the more aggressive code hoisting is. Specifying 0
10467 allows all expressions to travel unrestricted distances.
10468 The default value is 3.
10470 @item max-hoist-depth
10471 The depth of search in the dominator tree for expressions to hoist.
10472 This is used to avoid quadratic behavior in hoisting algorithm.
10473 The value of 0 does not limit on the search, but may slow down compilation
10474 of huge functions. The default value is 30.
10476 @item max-tail-merge-comparisons
10477 The maximum amount of similar bbs to compare a bb with. This is used to
10478 avoid quadratic behavior in tree tail merging. The default value is 10.
10480 @item max-tail-merge-iterations
10481 The maximum amount of iterations of the pass over the function. This is used to
10482 limit compilation time in tree tail merging. The default value is 2.
10484 @item max-unrolled-insns
10485 The maximum number of instructions that a loop may have to be unrolled.
10486 If a loop is unrolled, this parameter also determines how many times
10487 the loop code is unrolled.
10489 @item max-average-unrolled-insns
10490 The maximum number of instructions biased by probabilities of their execution
10491 that a loop may have to be unrolled. If a loop is unrolled,
10492 this parameter also determines how many times the loop code is unrolled.
10494 @item max-unroll-times
10495 The maximum number of unrollings of a single loop.
10497 @item max-peeled-insns
10498 The maximum number of instructions that a loop may have to be peeled.
10499 If a loop is peeled, this parameter also determines how many times
10500 the loop code is peeled.
10502 @item max-peel-times
10503 The maximum number of peelings of a single loop.
10505 @item max-peel-branches
10506 The maximum number of branches on the hot path through the peeled sequence.
10508 @item max-completely-peeled-insns
10509 The maximum number of insns of a completely peeled loop.
10511 @item max-completely-peel-times
10512 The maximum number of iterations of a loop to be suitable for complete peeling.
10514 @item max-completely-peel-loop-nest-depth
10515 The maximum depth of a loop nest suitable for complete peeling.
10517 @item max-unswitch-insns
10518 The maximum number of insns of an unswitched loop.
10520 @item max-unswitch-level
10521 The maximum number of branches unswitched in a single loop.
10523 @item lim-expensive
10524 The minimum cost of an expensive expression in the loop invariant motion.
10526 @item iv-consider-all-candidates-bound
10527 Bound on number of candidates for induction variables, below which
10528 all candidates are considered for each use in induction variable
10529 optimizations. If there are more candidates than this,
10530 only the most relevant ones are considered to avoid quadratic time complexity.
10532 @item iv-max-considered-uses
10533 The induction variable optimizations give up on loops that contain more
10534 induction variable uses.
10536 @item iv-always-prune-cand-set-bound
10537 If the number of candidates in the set is smaller than this value,
10538 always try to remove unnecessary ivs from the set
10539 when adding a new one.
10541 @item scev-max-expr-size
10542 Bound on size of expressions used in the scalar evolutions analyzer.
10543 Large expressions slow the analyzer.
10545 @item scev-max-expr-complexity
10546 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10547 Complex expressions slow the analyzer.
10549 @item vect-max-version-for-alignment-checks
10550 The maximum number of run-time checks that can be performed when
10551 doing loop versioning for alignment in the vectorizer.
10553 @item vect-max-version-for-alias-checks
10554 The maximum number of run-time checks that can be performed when
10555 doing loop versioning for alias in the vectorizer.
10557 @item vect-max-peeling-for-alignment
10558 The maximum number of loop peels to enhance access alignment
10559 for vectorizer. Value -1 means 'no limit'.
10561 @item max-iterations-to-track
10562 The maximum number of iterations of a loop the brute-force algorithm
10563 for analysis of the number of iterations of the loop tries to evaluate.
10565 @item hot-bb-count-ws-permille
10566 A basic block profile count is considered hot if it contributes to
10567 the given permillage (i.e. 0...1000) of the entire profiled execution.
10569 @item hot-bb-frequency-fraction
10570 Select fraction of the entry block frequency of executions of basic block in
10571 function given basic block needs to have to be considered hot.
10573 @item max-predicted-iterations
10574 The maximum number of loop iterations we predict statically. This is useful
10575 in cases where a function contains a single loop with known bound and
10576 another loop with unknown bound.
10577 The known number of iterations is predicted correctly, while
10578 the unknown number of iterations average to roughly 10. This means that the
10579 loop without bounds appears artificially cold relative to the other one.
10581 @item builtin-expect-probability
10582 Control the probability of the expression having the specified value. This
10583 parameter takes a percentage (i.e. 0 ... 100) as input.
10584 The default probability of 90 is obtained empirically.
10586 @item align-threshold
10588 Select fraction of the maximal frequency of executions of a basic block in
10589 a function to align the basic block.
10591 @item align-loop-iterations
10593 A loop expected to iterate at least the selected number of iterations is
10596 @item tracer-dynamic-coverage
10597 @itemx tracer-dynamic-coverage-feedback
10599 This value is used to limit superblock formation once the given percentage of
10600 executed instructions is covered. This limits unnecessary code size
10603 The @option{tracer-dynamic-coverage-feedback} parameter
10604 is used only when profile
10605 feedback is available. The real profiles (as opposed to statically estimated
10606 ones) are much less balanced allowing the threshold to be larger value.
10608 @item tracer-max-code-growth
10609 Stop tail duplication once code growth has reached given percentage. This is
10610 a rather artificial limit, as most of the duplicates are eliminated later in
10611 cross jumping, so it may be set to much higher values than is the desired code
10614 @item tracer-min-branch-ratio
10616 Stop reverse growth when the reverse probability of best edge is less than this
10617 threshold (in percent).
10619 @item tracer-min-branch-ratio
10620 @itemx tracer-min-branch-ratio-feedback
10622 Stop forward growth if the best edge has probability lower than this
10625 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10626 compilation for profile feedback and one for compilation without. The value
10627 for compilation with profile feedback needs to be more conservative (higher) in
10628 order to make tracer effective.
10630 @item max-cse-path-length
10632 The maximum number of basic blocks on path that CSE considers.
10635 @item max-cse-insns
10636 The maximum number of instructions CSE processes before flushing.
10637 The default is 1000.
10639 @item ggc-min-expand
10641 GCC uses a garbage collector to manage its own memory allocation. This
10642 parameter specifies the minimum percentage by which the garbage
10643 collector's heap should be allowed to expand between collections.
10644 Tuning this may improve compilation speed; it has no effect on code
10647 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10648 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10649 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10650 GCC is not able to calculate RAM on a particular platform, the lower
10651 bound of 30% is used. Setting this parameter and
10652 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10653 every opportunity. This is extremely slow, but can be useful for
10656 @item ggc-min-heapsize
10658 Minimum size of the garbage collector's heap before it begins bothering
10659 to collect garbage. The first collection occurs after the heap expands
10660 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10661 tuning this may improve compilation speed, and has no effect on code
10664 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10665 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10666 with a lower bound of 4096 (four megabytes) and an upper bound of
10667 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10668 particular platform, the lower bound is used. Setting this parameter
10669 very large effectively disables garbage collection. Setting this
10670 parameter and @option{ggc-min-expand} to zero causes a full collection
10671 to occur at every opportunity.
10673 @item max-reload-search-insns
10674 The maximum number of instruction reload should look backward for equivalent
10675 register. Increasing values mean more aggressive optimization, making the
10676 compilation time increase with probably slightly better performance.
10677 The default value is 100.
10679 @item max-cselib-memory-locations
10680 The maximum number of memory locations cselib should take into account.
10681 Increasing values mean more aggressive optimization, making the compilation time
10682 increase with probably slightly better performance. The default value is 500.
10684 @item reorder-blocks-duplicate
10685 @itemx reorder-blocks-duplicate-feedback
10687 Used by the basic block reordering pass to decide whether to use unconditional
10688 branch or duplicate the code on its destination. Code is duplicated when its
10689 estimated size is smaller than this value multiplied by the estimated size of
10690 unconditional jump in the hot spots of the program.
10692 The @option{reorder-block-duplicate-feedback} parameter
10693 is used only when profile
10694 feedback is available. It may be set to higher values than
10695 @option{reorder-block-duplicate} since information about the hot spots is more
10698 @item max-sched-ready-insns
10699 The maximum number of instructions ready to be issued the scheduler should
10700 consider at any given time during the first scheduling pass. Increasing
10701 values mean more thorough searches, making the compilation time increase
10702 with probably little benefit. The default value is 100.
10704 @item max-sched-region-blocks
10705 The maximum number of blocks in a region to be considered for
10706 interblock scheduling. The default value is 10.
10708 @item max-pipeline-region-blocks
10709 The maximum number of blocks in a region to be considered for
10710 pipelining in the selective scheduler. The default value is 15.
10712 @item max-sched-region-insns
10713 The maximum number of insns in a region to be considered for
10714 interblock scheduling. The default value is 100.
10716 @item max-pipeline-region-insns
10717 The maximum number of insns in a region to be considered for
10718 pipelining in the selective scheduler. The default value is 200.
10720 @item min-spec-prob
10721 The minimum probability (in percents) of reaching a source block
10722 for interblock speculative scheduling. The default value is 40.
10724 @item max-sched-extend-regions-iters
10725 The maximum number of iterations through CFG to extend regions.
10726 A value of 0 (the default) disables region extensions.
10728 @item max-sched-insn-conflict-delay
10729 The maximum conflict delay for an insn to be considered for speculative motion.
10730 The default value is 3.
10732 @item sched-spec-prob-cutoff
10733 The minimal probability of speculation success (in percents), so that
10734 speculative insns are scheduled.
10735 The default value is 40.
10737 @item sched-spec-state-edge-prob-cutoff
10738 The minimum probability an edge must have for the scheduler to save its
10740 The default value is 10.
10742 @item sched-mem-true-dep-cost
10743 Minimal distance (in CPU cycles) between store and load targeting same
10744 memory locations. The default value is 1.
10746 @item selsched-max-lookahead
10747 The maximum size of the lookahead window of selective scheduling. It is a
10748 depth of search for available instructions.
10749 The default value is 50.
10751 @item selsched-max-sched-times
10752 The maximum number of times that an instruction is scheduled during
10753 selective scheduling. This is the limit on the number of iterations
10754 through which the instruction may be pipelined. The default value is 2.
10756 @item selsched-max-insns-to-rename
10757 The maximum number of best instructions in the ready list that are considered
10758 for renaming in the selective scheduler. The default value is 2.
10761 The minimum value of stage count that swing modulo scheduler
10762 generates. The default value is 2.
10764 @item max-last-value-rtl
10765 The maximum size measured as number of RTLs that can be recorded in an expression
10766 in combiner for a pseudo register as last known value of that register. The default
10769 @item max-combine-insns
10770 The maximum number of instructions the RTL combiner tries to combine.
10771 The default value is 2 at @option{-Og} and 4 otherwise.
10773 @item integer-share-limit
10774 Small integer constants can use a shared data structure, reducing the
10775 compiler's memory usage and increasing its speed. This sets the maximum
10776 value of a shared integer constant. The default value is 256.
10778 @item ssp-buffer-size
10779 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10780 protection when @option{-fstack-protection} is used.
10782 @item min-size-for-stack-sharing
10783 The minimum size of variables taking part in stack slot sharing when not
10784 optimizing. The default value is 32.
10786 @item max-jump-thread-duplication-stmts
10787 Maximum number of statements allowed in a block that needs to be
10788 duplicated when threading jumps.
10790 @item max-fields-for-field-sensitive
10791 Maximum number of fields in a structure treated in
10792 a field sensitive manner during pointer analysis. The default is zero
10793 for @option{-O0} and @option{-O1},
10794 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10796 @item prefetch-latency
10797 Estimate on average number of instructions that are executed before
10798 prefetch finishes. The distance prefetched ahead is proportional
10799 to this constant. Increasing this number may also lead to less
10800 streams being prefetched (see @option{simultaneous-prefetches}).
10802 @item simultaneous-prefetches
10803 Maximum number of prefetches that can run at the same time.
10805 @item l1-cache-line-size
10806 The size of cache line in L1 cache, in bytes.
10808 @item l1-cache-size
10809 The size of L1 cache, in kilobytes.
10811 @item l2-cache-size
10812 The size of L2 cache, in kilobytes.
10814 @item min-insn-to-prefetch-ratio
10815 The minimum ratio between the number of instructions and the
10816 number of prefetches to enable prefetching in a loop.
10818 @item prefetch-min-insn-to-mem-ratio
10819 The minimum ratio between the number of instructions and the
10820 number of memory references to enable prefetching in a loop.
10822 @item use-canonical-types
10823 Whether the compiler should use the ``canonical'' type system. By
10824 default, this should always be 1, which uses a more efficient internal
10825 mechanism for comparing types in C++ and Objective-C++. However, if
10826 bugs in the canonical type system are causing compilation failures,
10827 set this value to 0 to disable canonical types.
10829 @item switch-conversion-max-branch-ratio
10830 Switch initialization conversion refuses to create arrays that are
10831 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10832 branches in the switch.
10834 @item max-partial-antic-length
10835 Maximum length of the partial antic set computed during the tree
10836 partial redundancy elimination optimization (@option{-ftree-pre}) when
10837 optimizing at @option{-O3} and above. For some sorts of source code
10838 the enhanced partial redundancy elimination optimization can run away,
10839 consuming all of the memory available on the host machine. This
10840 parameter sets a limit on the length of the sets that are computed,
10841 which prevents the runaway behavior. Setting a value of 0 for
10842 this parameter allows an unlimited set length.
10844 @item sccvn-max-scc-size
10845 Maximum size of a strongly connected component (SCC) during SCCVN
10846 processing. If this limit is hit, SCCVN processing for the whole
10847 function is not done and optimizations depending on it are
10848 disabled. The default maximum SCC size is 10000.
10850 @item sccvn-max-alias-queries-per-access
10851 Maximum number of alias-oracle queries we perform when looking for
10852 redundancies for loads and stores. If this limit is hit the search
10853 is aborted and the load or store is not considered redundant. The
10854 number of queries is algorithmically limited to the number of
10855 stores on all paths from the load to the function entry.
10856 The default maxmimum number of queries is 1000.
10858 @item ira-max-loops-num
10859 IRA uses regional register allocation by default. If a function
10860 contains more loops than the number given by this parameter, only at most
10861 the given number of the most frequently-executed loops form regions
10862 for regional register allocation. The default value of the
10865 @item ira-max-conflict-table-size
10866 Although IRA uses a sophisticated algorithm to compress the conflict
10867 table, the table can still require excessive amounts of memory for
10868 huge functions. If the conflict table for a function could be more
10869 than the size in MB given by this parameter, the register allocator
10870 instead uses a faster, simpler, and lower-quality
10871 algorithm that does not require building a pseudo-register conflict table.
10872 The default value of the parameter is 2000.
10874 @item ira-loop-reserved-regs
10875 IRA can be used to evaluate more accurate register pressure in loops
10876 for decisions to move loop invariants (see @option{-O3}). The number
10877 of available registers reserved for some other purposes is given
10878 by this parameter. The default value of the parameter is 2, which is
10879 the minimal number of registers needed by typical instructions.
10880 This value is the best found from numerous experiments.
10882 @item lra-inheritance-ebb-probability-cutoff
10883 LRA tries to reuse values reloaded in registers in subsequent insns.
10884 This optimization is called inheritance. EBB is used as a region to
10885 do this optimization. The parameter defines a minimal fall-through
10886 edge probability in percentage used to add BB to inheritance EBB in
10887 LRA. The default value of the parameter is 40. The value was chosen
10888 from numerous runs of SPEC2000 on x86-64.
10890 @item loop-invariant-max-bbs-in-loop
10891 Loop invariant motion can be very expensive, both in compilation time and
10892 in amount of needed compile-time memory, with very large loops. Loops
10893 with more basic blocks than this parameter won't have loop invariant
10894 motion optimization performed on them. The default value of the
10895 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10897 @item loop-max-datarefs-for-datadeps
10898 Building data dapendencies is expensive for very large loops. This
10899 parameter limits the number of data references in loops that are
10900 considered for data dependence analysis. These large loops are no
10901 handled by the optimizations using loop data dependencies.
10902 The default value is 1000.
10904 @item max-vartrack-size
10905 Sets a maximum number of hash table slots to use during variable
10906 tracking dataflow analysis of any function. If this limit is exceeded
10907 with variable tracking at assignments enabled, analysis for that
10908 function is retried without it, after removing all debug insns from
10909 the function. If the limit is exceeded even without debug insns, var
10910 tracking analysis is completely disabled for the function. Setting
10911 the parameter to zero makes it unlimited.
10913 @item max-vartrack-expr-depth
10914 Sets a maximum number of recursion levels when attempting to map
10915 variable names or debug temporaries to value expressions. This trades
10916 compilation time for more complete debug information. If this is set too
10917 low, value expressions that are available and could be represented in
10918 debug information may end up not being used; setting this higher may
10919 enable the compiler to find more complex debug expressions, but compile
10920 time and memory use may grow. The default is 12.
10922 @item min-nondebug-insn-uid
10923 Use uids starting at this parameter for nondebug insns. The range below
10924 the parameter is reserved exclusively for debug insns created by
10925 @option{-fvar-tracking-assignments}, but debug insns may get
10926 (non-overlapping) uids above it if the reserved range is exhausted.
10928 @item ipa-sra-ptr-growth-factor
10929 IPA-SRA replaces a pointer to an aggregate with one or more new
10930 parameters only when their cumulative size is less or equal to
10931 @option{ipa-sra-ptr-growth-factor} times the size of the original
10934 @item sra-max-scalarization-size-Ospeed
10935 @item sra-max-scalarization-size-Osize
10936 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10937 replace scalar parts of aggregates with uses of independent scalar
10938 variables. These parameters control the maximum size, in storage units,
10939 of aggregate which is considered for replacement when compiling for
10941 (@option{sra-max-scalarization-size-Ospeed}) or size
10942 (@option{sra-max-scalarization-size-Osize}) respectively.
10944 @item tm-max-aggregate-size
10945 When making copies of thread-local variables in a transaction, this
10946 parameter specifies the size in bytes after which variables are
10947 saved with the logging functions as opposed to save/restore code
10948 sequence pairs. This option only applies when using
10951 @item graphite-max-nb-scop-params
10952 To avoid exponential effects in the Graphite loop transforms, the
10953 number of parameters in a Static Control Part (SCoP) is bounded. The
10954 default value is 10 parameters. A variable whose value is unknown at
10955 compilation time and defined outside a SCoP is a parameter of the SCoP.
10957 @item graphite-max-bbs-per-function
10958 To avoid exponential effects in the detection of SCoPs, the size of
10959 the functions analyzed by Graphite is bounded. The default value is
10962 @item loop-block-tile-size
10963 Loop blocking or strip mining transforms, enabled with
10964 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10965 loop in the loop nest by a given number of iterations. The strip
10966 length can be changed using the @option{loop-block-tile-size}
10967 parameter. The default value is 51 iterations.
10969 @item loop-unroll-jam-size
10970 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10971 default value is 4.
10973 @item loop-unroll-jam-depth
10974 Specify the dimension to be unrolled (counting from the most inner loop)
10975 for the @option{-floop-unroll-and-jam}. The default value is 2.
10977 @item ipa-cp-value-list-size
10978 IPA-CP attempts to track all possible values and types passed to a function's
10979 parameter in order to propagate them and perform devirtualization.
10980 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10981 stores per one formal parameter of a function.
10983 @item ipa-cp-eval-threshold
10984 IPA-CP calculates its own score of cloning profitability heuristics
10985 and performs those cloning opportunities with scores that exceed
10986 @option{ipa-cp-eval-threshold}.
10988 @item ipa-cp-recursion-penalty
10989 Percentage penalty the recursive functions will receive when they
10990 are evaluated for cloning.
10992 @item ipa-cp-single-call-penalty
10993 Percentage penalty functions containg a single call to another
10994 function will receive when they are evaluated for cloning.
10997 @item ipa-max-agg-items
10998 IPA-CP is also capable to propagate a number of scalar values passed
10999 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11000 number of such values per one parameter.
11002 @item ipa-cp-loop-hint-bonus
11003 When IPA-CP determines that a cloning candidate would make the number
11004 of iterations of a loop known, it adds a bonus of
11005 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11008 @item ipa-cp-array-index-hint-bonus
11009 When IPA-CP determines that a cloning candidate would make the index of
11010 an array access known, it adds a bonus of
11011 @option{ipa-cp-array-index-hint-bonus} to the profitability
11012 score of the candidate.
11014 @item ipa-max-aa-steps
11015 During its analysis of function bodies, IPA-CP employs alias analysis
11016 in order to track values pointed to by function parameters. In order
11017 not spend too much time analyzing huge functions, it gives up and
11018 consider all memory clobbered after examining
11019 @option{ipa-max-aa-steps} statements modifying memory.
11021 @item lto-partitions
11022 Specify desired number of partitions produced during WHOPR compilation.
11023 The number of partitions should exceed the number of CPUs used for compilation.
11024 The default value is 32.
11026 @item lto-minpartition
11027 Size of minimal partition for WHOPR (in estimated instructions).
11028 This prevents expenses of splitting very small programs into too many
11031 @item cxx-max-namespaces-for-diagnostic-help
11032 The maximum number of namespaces to consult for suggestions when C++
11033 name lookup fails for an identifier. The default is 1000.
11035 @item sink-frequency-threshold
11036 The maximum relative execution frequency (in percents) of the target block
11037 relative to a statement's original block to allow statement sinking of a
11038 statement. Larger numbers result in more aggressive statement sinking.
11039 The default value is 75. A small positive adjustment is applied for
11040 statements with memory operands as those are even more profitable so sink.
11042 @item max-stores-to-sink
11043 The maximum number of conditional stores paires that can be sunk. Set to 0
11044 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11045 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
11047 @item allow-store-data-races
11048 Allow optimizers to introduce new data races on stores.
11049 Set to 1 to allow, otherwise to 0. This option is enabled by default
11050 at optimization level @option{-Ofast}.
11052 @item case-values-threshold
11053 The smallest number of different values for which it is best to use a
11054 jump-table instead of a tree of conditional branches. If the value is
11055 0, use the default for the machine. The default is 0.
11057 @item tree-reassoc-width
11058 Set the maximum number of instructions executed in parallel in
11059 reassociated tree. This parameter overrides target dependent
11060 heuristics used by default if has non zero value.
11062 @item sched-pressure-algorithm
11063 Choose between the two available implementations of
11064 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11065 and is the more likely to prevent instructions from being reordered.
11066 Algorithm 2 was designed to be a compromise between the relatively
11067 conservative approach taken by algorithm 1 and the rather aggressive
11068 approach taken by the default scheduler. It relies more heavily on
11069 having a regular register file and accurate register pressure classes.
11070 See @file{haifa-sched.c} in the GCC sources for more details.
11072 The default choice depends on the target.
11074 @item max-slsr-cand-scan
11075 Set the maximum number of existing candidates that are considered when
11076 seeking a basis for a new straight-line strength reduction candidate.
11079 Enable buffer overflow detection for global objects. This kind
11080 of protection is enabled by default if you are using
11081 @option{-fsanitize=address} option.
11082 To disable global objects protection use @option{--param asan-globals=0}.
11085 Enable buffer overflow detection for stack objects. This kind of
11086 protection is enabled by default when using @option{-fsanitize=address}.
11087 To disable stack protection use @option{--param asan-stack=0} option.
11089 @item asan-instrument-reads
11090 Enable buffer overflow detection for memory reads. This kind of
11091 protection is enabled by default when using @option{-fsanitize=address}.
11092 To disable memory reads protection use
11093 @option{--param asan-instrument-reads=0}.
11095 @item asan-instrument-writes
11096 Enable buffer overflow detection for memory writes. This kind of
11097 protection is enabled by default when using @option{-fsanitize=address}.
11098 To disable memory writes protection use
11099 @option{--param asan-instrument-writes=0} option.
11101 @item asan-memintrin
11102 Enable detection for built-in functions. This kind of protection
11103 is enabled by default when using @option{-fsanitize=address}.
11104 To disable built-in functions protection use
11105 @option{--param asan-memintrin=0}.
11107 @item asan-use-after-return
11108 Enable detection of use-after-return. This kind of protection
11109 is enabled by default when using @option{-fsanitize=address} option.
11110 To disable use-after-return detection use
11111 @option{--param asan-use-after-return=0}.
11113 @item asan-instrumentation-with-call-threshold
11114 If number of memory accesses in function being instrumented
11115 is greater or equal to this number, use callbacks instead of inline checks.
11116 E.g. to disable inline code use
11117 @option{--param asan-instrumentation-with-call-threshold=0}.
11119 @item chkp-max-ctor-size
11120 Static constructors generated by Pointer Bounds Checker may become very
11121 large and significantly increase compile time at optimization level
11122 @option{-O1} and higher. This parameter is a maximum nubmer of statements
11123 in a single generated constructor. Default value is 5000.
11125 @item max-fsm-thread-path-insns
11126 Maximum number of instructions to copy when duplicating blocks on a
11127 finite state automaton jump thread path. The default is 100.
11129 @item max-fsm-thread-length
11130 Maximum number of basic blocks on a finite state automaton jump thread
11131 path. The default is 10.
11133 @item max-fsm-thread-paths
11134 Maximum number of new jump thread paths to create for a finite state
11135 automaton. The default is 50.
11137 @item parloops-chunk-size
11138 Chunk size of omp schedule for loops parallelized by parloops. The default
11141 @item parloops-schedule
11142 Schedule type of omp schedule for loops parallelized by parloops (static,
11143 dynamic, guided, auto, runtime). The default is static.
11145 @item max-ssa-name-query-depth
11146 Maximum depth of recursion when querying properties of SSA names in things
11147 like fold routines. One level of recursion corresponds to following a
11152 @node Preprocessor Options
11153 @section Options Controlling the Preprocessor
11154 @cindex preprocessor options
11155 @cindex options, preprocessor
11157 These options control the C preprocessor, which is run on each C source
11158 file before actual compilation.
11160 If you use the @option{-E} option, nothing is done except preprocessing.
11161 Some of these options make sense only together with @option{-E} because
11162 they cause the preprocessor output to be unsuitable for actual
11166 @item -Wp,@var{option}
11168 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11169 and pass @var{option} directly through to the preprocessor. If
11170 @var{option} contains commas, it is split into multiple options at the
11171 commas. However, many options are modified, translated or interpreted
11172 by the compiler driver before being passed to the preprocessor, and
11173 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11174 interface is undocumented and subject to change, so whenever possible
11175 you should avoid using @option{-Wp} and let the driver handle the
11178 @item -Xpreprocessor @var{option}
11179 @opindex Xpreprocessor
11180 Pass @var{option} as an option to the preprocessor. You can use this to
11181 supply system-specific preprocessor options that GCC does not
11184 If you want to pass an option that takes an argument, you must use
11185 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11187 @item -no-integrated-cpp
11188 @opindex no-integrated-cpp
11189 Perform preprocessing as a separate pass before compilation.
11190 By default, GCC performs preprocessing as an integrated part of
11191 input tokenization and parsing.
11192 If this option is provided, the appropriate language front end
11193 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11194 and Objective-C, respectively) is instead invoked twice,
11195 once for preprocessing only and once for actual compilation
11196 of the preprocessed input.
11197 This option may be useful in conjunction with the @option{-B} or
11198 @option{-wrapper} options to specify an alternate preprocessor or
11199 perform additional processing of the program source between
11200 normal preprocessing and compilation.
11203 @include cppopts.texi
11205 @node Assembler Options
11206 @section Passing Options to the Assembler
11208 @c prevent bad page break with this line
11209 You can pass options to the assembler.
11212 @item -Wa,@var{option}
11214 Pass @var{option} as an option to the assembler. If @var{option}
11215 contains commas, it is split into multiple options at the commas.
11217 @item -Xassembler @var{option}
11218 @opindex Xassembler
11219 Pass @var{option} as an option to the assembler. You can use this to
11220 supply system-specific assembler options that GCC does not
11223 If you want to pass an option that takes an argument, you must use
11224 @option{-Xassembler} twice, once for the option and once for the argument.
11229 @section Options for Linking
11230 @cindex link options
11231 @cindex options, linking
11233 These options come into play when the compiler links object files into
11234 an executable output file. They are meaningless if the compiler is
11235 not doing a link step.
11239 @item @var{object-file-name}
11240 A file name that does not end in a special recognized suffix is
11241 considered to name an object file or library. (Object files are
11242 distinguished from libraries by the linker according to the file
11243 contents.) If linking is done, these object files are used as input
11252 If any of these options is used, then the linker is not run, and
11253 object file names should not be used as arguments. @xref{Overall
11257 @opindex fuse-ld=bfd
11258 Use the @command{bfd} linker instead of the default linker.
11260 @item -fuse-ld=gold
11261 @opindex fuse-ld=gold
11262 Use the @command{gold} linker instead of the default linker.
11265 @item -l@var{library}
11266 @itemx -l @var{library}
11268 Search the library named @var{library} when linking. (The second
11269 alternative with the library as a separate argument is only for
11270 POSIX compliance and is not recommended.)
11272 It makes a difference where in the command you write this option; the
11273 linker searches and processes libraries and object files in the order they
11274 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11275 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11276 to functions in @samp{z}, those functions may not be loaded.
11278 The linker searches a standard list of directories for the library,
11279 which is actually a file named @file{lib@var{library}.a}. The linker
11280 then uses this file as if it had been specified precisely by name.
11282 The directories searched include several standard system directories
11283 plus any that you specify with @option{-L}.
11285 Normally the files found this way are library files---archive files
11286 whose members are object files. The linker handles an archive file by
11287 scanning through it for members which define symbols that have so far
11288 been referenced but not defined. But if the file that is found is an
11289 ordinary object file, it is linked in the usual fashion. The only
11290 difference between using an @option{-l} option and specifying a file name
11291 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11292 and searches several directories.
11296 You need this special case of the @option{-l} option in order to
11297 link an Objective-C or Objective-C++ program.
11299 @item -nostartfiles
11300 @opindex nostartfiles
11301 Do not use the standard system startup files when linking.
11302 The standard system libraries are used normally, unless @option{-nostdlib}
11303 or @option{-nodefaultlibs} is used.
11305 @item -nodefaultlibs
11306 @opindex nodefaultlibs
11307 Do not use the standard system libraries when linking.
11308 Only the libraries you specify are passed to the linker, and options
11309 specifying linkage of the system libraries, such as @option{-static-libgcc}
11310 or @option{-shared-libgcc}, are ignored.
11311 The standard startup files are used normally, unless @option{-nostartfiles}
11314 The compiler may generate calls to @code{memcmp},
11315 @code{memset}, @code{memcpy} and @code{memmove}.
11316 These entries are usually resolved by entries in
11317 libc. These entry points should be supplied through some other
11318 mechanism when this option is specified.
11322 Do not use the standard system startup files or libraries when linking.
11323 No startup files and only the libraries you specify are passed to
11324 the linker, and options specifying linkage of the system libraries, such as
11325 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11327 The compiler may generate calls to @code{memcmp}, @code{memset},
11328 @code{memcpy} and @code{memmove}.
11329 These entries are usually resolved by entries in
11330 libc. These entry points should be supplied through some other
11331 mechanism when this option is specified.
11333 @cindex @option{-lgcc}, use with @option{-nostdlib}
11334 @cindex @option{-nostdlib} and unresolved references
11335 @cindex unresolved references and @option{-nostdlib}
11336 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11337 @cindex @option{-nodefaultlibs} and unresolved references
11338 @cindex unresolved references and @option{-nodefaultlibs}
11339 One of the standard libraries bypassed by @option{-nostdlib} and
11340 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11341 which GCC uses to overcome shortcomings of particular machines, or special
11342 needs for some languages.
11343 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11344 Collection (GCC) Internals},
11345 for more discussion of @file{libgcc.a}.)
11346 In most cases, you need @file{libgcc.a} even when you want to avoid
11347 other standard libraries. In other words, when you specify @option{-nostdlib}
11348 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11349 This ensures that you have no unresolved references to internal GCC
11350 library subroutines.
11351 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11352 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11353 GNU Compiler Collection (GCC) Internals}.)
11357 Produce a position independent executable on targets that support it.
11358 For predictable results, you must also specify the same set of options
11359 used for compilation (@option{-fpie}, @option{-fPIE},
11360 or model suboptions) when you specify this linker option.
11364 Don't produce a position independent executable.
11368 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11369 that support it. This instructs the linker to add all symbols, not
11370 only used ones, to the dynamic symbol table. This option is needed
11371 for some uses of @code{dlopen} or to allow obtaining backtraces
11372 from within a program.
11376 Remove all symbol table and relocation information from the executable.
11380 On systems that support dynamic linking, this prevents linking with the shared
11381 libraries. On other systems, this option has no effect.
11385 Produce a shared object which can then be linked with other objects to
11386 form an executable. Not all systems support this option. For predictable
11387 results, you must also specify the same set of options used for compilation
11388 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11389 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11390 needs to build supplementary stub code for constructors to work. On
11391 multi-libbed systems, @samp{gcc -shared} must select the correct support
11392 libraries to link against. Failing to supply the correct flags may lead
11393 to subtle defects. Supplying them in cases where they are not necessary
11396 @item -shared-libgcc
11397 @itemx -static-libgcc
11398 @opindex shared-libgcc
11399 @opindex static-libgcc
11400 On systems that provide @file{libgcc} as a shared library, these options
11401 force the use of either the shared or static version, respectively.
11402 If no shared version of @file{libgcc} was built when the compiler was
11403 configured, these options have no effect.
11405 There are several situations in which an application should use the
11406 shared @file{libgcc} instead of the static version. The most common
11407 of these is when the application wishes to throw and catch exceptions
11408 across different shared libraries. In that case, each of the libraries
11409 as well as the application itself should use the shared @file{libgcc}.
11411 Therefore, the G++ and GCJ drivers automatically add
11412 @option{-shared-libgcc} whenever you build a shared library or a main
11413 executable, because C++ and Java programs typically use exceptions, so
11414 this is the right thing to do.
11416 If, instead, you use the GCC driver to create shared libraries, you may
11417 find that they are not always linked with the shared @file{libgcc}.
11418 If GCC finds, at its configuration time, that you have a non-GNU linker
11419 or a GNU linker that does not support option @option{--eh-frame-hdr},
11420 it links the shared version of @file{libgcc} into shared libraries
11421 by default. Otherwise, it takes advantage of the linker and optimizes
11422 away the linking with the shared version of @file{libgcc}, linking with
11423 the static version of libgcc by default. This allows exceptions to
11424 propagate through such shared libraries, without incurring relocation
11425 costs at library load time.
11427 However, if a library or main executable is supposed to throw or catch
11428 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11429 for the languages used in the program, or using the option
11430 @option{-shared-libgcc}, such that it is linked with the shared
11433 @item -static-libasan
11434 @opindex static-libasan
11435 When the @option{-fsanitize=address} option is used to link a program,
11436 the GCC driver automatically links against @option{libasan}. If
11437 @file{libasan} is available as a shared library, and the @option{-static}
11438 option is not used, then this links against the shared version of
11439 @file{libasan}. The @option{-static-libasan} option directs the GCC
11440 driver to link @file{libasan} statically, without necessarily linking
11441 other libraries statically.
11443 @item -static-libtsan
11444 @opindex static-libtsan
11445 When the @option{-fsanitize=thread} option is used to link a program,
11446 the GCC driver automatically links against @option{libtsan}. If
11447 @file{libtsan} is available as a shared library, and the @option{-static}
11448 option is not used, then this links against the shared version of
11449 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11450 driver to link @file{libtsan} statically, without necessarily linking
11451 other libraries statically.
11453 @item -static-liblsan
11454 @opindex static-liblsan
11455 When the @option{-fsanitize=leak} option is used to link a program,
11456 the GCC driver automatically links against @option{liblsan}. If
11457 @file{liblsan} is available as a shared library, and the @option{-static}
11458 option is not used, then this links against the shared version of
11459 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11460 driver to link @file{liblsan} statically, without necessarily linking
11461 other libraries statically.
11463 @item -static-libubsan
11464 @opindex static-libubsan
11465 When the @option{-fsanitize=undefined} option is used to link a program,
11466 the GCC driver automatically links against @option{libubsan}. If
11467 @file{libubsan} is available as a shared library, and the @option{-static}
11468 option is not used, then this links against the shared version of
11469 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11470 driver to link @file{libubsan} statically, without necessarily linking
11471 other libraries statically.
11473 @item -static-libmpx
11474 @opindex static-libmpx
11475 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11476 used to link a program, the GCC driver automatically links against
11477 @file{libmpx}. If @file{libmpx} is available as a shared library,
11478 and the @option{-static} option is not used, then this links against
11479 the shared version of @file{libmpx}. The @option{-static-libmpx}
11480 option directs the GCC driver to link @file{libmpx} statically,
11481 without necessarily linking other libraries statically.
11483 @item -static-libmpxwrappers
11484 @opindex static-libmpxwrappers
11485 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11486 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11487 GCC driver automatically links against @file{libmpxwrappers}. If
11488 @file{libmpxwrappers} is available as a shared library, and the
11489 @option{-static} option is not used, then this links against the shared
11490 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11491 option directs the GCC driver to link @file{libmpxwrappers} statically,
11492 without necessarily linking other libraries statically.
11494 @item -static-libstdc++
11495 @opindex static-libstdc++
11496 When the @command{g++} program is used to link a C++ program, it
11497 normally automatically links against @option{libstdc++}. If
11498 @file{libstdc++} is available as a shared library, and the
11499 @option{-static} option is not used, then this links against the
11500 shared version of @file{libstdc++}. That is normally fine. However, it
11501 is sometimes useful to freeze the version of @file{libstdc++} used by
11502 the program without going all the way to a fully static link. The
11503 @option{-static-libstdc++} option directs the @command{g++} driver to
11504 link @file{libstdc++} statically, without necessarily linking other
11505 libraries statically.
11509 Bind references to global symbols when building a shared object. Warn
11510 about any unresolved references (unless overridden by the link editor
11511 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11514 @item -T @var{script}
11516 @cindex linker script
11517 Use @var{script} as the linker script. This option is supported by most
11518 systems using the GNU linker. On some targets, such as bare-board
11519 targets without an operating system, the @option{-T} option may be required
11520 when linking to avoid references to undefined symbols.
11522 @item -Xlinker @var{option}
11524 Pass @var{option} as an option to the linker. You can use this to
11525 supply system-specific linker options that GCC does not recognize.
11527 If you want to pass an option that takes a separate argument, you must use
11528 @option{-Xlinker} twice, once for the option and once for the argument.
11529 For example, to pass @option{-assert definitions}, you must write
11530 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11531 @option{-Xlinker "-assert definitions"}, because this passes the entire
11532 string as a single argument, which is not what the linker expects.
11534 When using the GNU linker, it is usually more convenient to pass
11535 arguments to linker options using the @option{@var{option}=@var{value}}
11536 syntax than as separate arguments. For example, you can specify
11537 @option{-Xlinker -Map=output.map} rather than
11538 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11539 this syntax for command-line options.
11541 @item -Wl,@var{option}
11543 Pass @var{option} as an option to the linker. If @var{option} contains
11544 commas, it is split into multiple options at the commas. You can use this
11545 syntax to pass an argument to the option.
11546 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11547 linker. When using the GNU linker, you can also get the same effect with
11548 @option{-Wl,-Map=output.map}.
11550 @item -u @var{symbol}
11552 Pretend the symbol @var{symbol} is undefined, to force linking of
11553 library modules to define it. You can use @option{-u} multiple times with
11554 different symbols to force loading of additional library modules.
11556 @item -z @var{keyword}
11558 @option{-z} is passed directly on to the linker along with the keyword
11559 @var{keyword}. See the section in the documentation of your linker for
11560 permitted values and their meanings.
11563 @node Directory Options
11564 @section Options for Directory Search
11565 @cindex directory options
11566 @cindex options, directory search
11567 @cindex search path
11569 These options specify directories to search for header files, for
11570 libraries and for parts of the compiler:
11575 Add the directory @var{dir} to the head of the list of directories to be
11576 searched for header files. This can be used to override a system header
11577 file, substituting your own version, since these directories are
11578 searched before the system header file directories. However, you should
11579 not use this option to add directories that contain vendor-supplied
11580 system header files (use @option{-isystem} for that). If you use more than
11581 one @option{-I} option, the directories are scanned in left-to-right
11582 order; the standard system directories come after.
11584 If a standard system include directory, or a directory specified with
11585 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11586 option is ignored. The directory is still searched but as a
11587 system directory at its normal position in the system include chain.
11588 This is to ensure that GCC's procedure to fix buggy system headers and
11589 the ordering for the @code{include_next} directive are not inadvertently changed.
11590 If you really need to change the search order for system directories,
11591 use the @option{-nostdinc} and/or @option{-isystem} options.
11593 @item -iplugindir=@var{dir}
11594 @opindex iplugindir=
11595 Set the directory to search for plugins that are passed
11596 by @option{-fplugin=@var{name}} instead of
11597 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11598 to be used by the user, but only passed by the driver.
11600 @item -iquote@var{dir}
11602 Add the directory @var{dir} to the head of the list of directories to
11603 be searched for header files only for the case of @code{#include
11604 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11605 otherwise just like @option{-I}.
11609 Add directory @var{dir} to the list of directories to be searched
11612 @item -B@var{prefix}
11614 This option specifies where to find the executables, libraries,
11615 include files, and data files of the compiler itself.
11617 The compiler driver program runs one or more of the subprograms
11618 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11619 @var{prefix} as a prefix for each program it tries to run, both with and
11620 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11622 For each subprogram to be run, the compiler driver first tries the
11623 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11624 is not specified, the driver tries two standard prefixes,
11625 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11626 those results in a file name that is found, the unmodified program
11627 name is searched for using the directories specified in your
11628 @env{PATH} environment variable.
11630 The compiler checks to see if the path provided by @option{-B}
11631 refers to a directory, and if necessary it adds a directory
11632 separator character at the end of the path.
11634 @option{-B} prefixes that effectively specify directory names also apply
11635 to libraries in the linker, because the compiler translates these
11636 options into @option{-L} options for the linker. They also apply to
11637 include files in the preprocessor, because the compiler translates these
11638 options into @option{-isystem} options for the preprocessor. In this case,
11639 the compiler appends @samp{include} to the prefix.
11641 The runtime support file @file{libgcc.a} can also be searched for using
11642 the @option{-B} prefix, if needed. If it is not found there, the two
11643 standard prefixes above are tried, and that is all. The file is left
11644 out of the link if it is not found by those means.
11646 Another way to specify a prefix much like the @option{-B} prefix is to use
11647 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11650 As a special kludge, if the path provided by @option{-B} is
11651 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11652 9, then it is replaced by @file{[dir/]include}. This is to help
11653 with boot-strapping the compiler.
11655 @item -specs=@var{file}
11657 Process @var{file} after the compiler reads in the standard @file{specs}
11658 file, in order to override the defaults which the @command{gcc} driver
11659 program uses when determining what switches to pass to @command{cc1},
11660 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11661 @option{-specs=@var{file}} can be specified on the command line, and they
11662 are processed in order, from left to right.
11664 @item --sysroot=@var{dir}
11666 Use @var{dir} as the logical root directory for headers and libraries.
11667 For example, if the compiler normally searches for headers in
11668 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11669 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11671 If you use both this option and the @option{-isysroot} option, then
11672 the @option{--sysroot} option applies to libraries, but the
11673 @option{-isysroot} option applies to header files.
11675 The GNU linker (beginning with version 2.16) has the necessary support
11676 for this option. If your linker does not support this option, the
11677 header file aspect of @option{--sysroot} still works, but the
11678 library aspect does not.
11680 @item --no-sysroot-suffix
11681 @opindex no-sysroot-suffix
11682 For some targets, a suffix is added to the root directory specified
11683 with @option{--sysroot}, depending on the other options used, so that
11684 headers may for example be found in
11685 @file{@var{dir}/@var{suffix}/usr/include} instead of
11686 @file{@var{dir}/usr/include}. This option disables the addition of
11691 This option has been deprecated. Please use @option{-iquote} instead for
11692 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11694 Any directories you specify with @option{-I} options before the @option{-I-}
11695 option are searched only for the case of @code{#include "@var{file}"};
11696 they are not searched for @code{#include <@var{file}>}.
11698 If additional directories are specified with @option{-I} options after
11699 the @option{-I-} option, these directories are searched for all @code{#include}
11700 directives. (Ordinarily @emph{all} @option{-I} directories are used
11703 In addition, the @option{-I-} option inhibits the use of the current
11704 directory (where the current input file came from) as the first search
11705 directory for @code{#include "@var{file}"}. There is no way to
11706 override this effect of @option{-I-}. With @option{-I.} you can specify
11707 searching the directory that is current when the compiler is
11708 invoked. That is not exactly the same as what the preprocessor does
11709 by default, but it is often satisfactory.
11711 @option{-I-} does not inhibit the use of the standard system directories
11712 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11719 @section Specifying Subprocesses and the Switches to Pass to Them
11722 @command{gcc} is a driver program. It performs its job by invoking a
11723 sequence of other programs to do the work of compiling, assembling and
11724 linking. GCC interprets its command-line parameters and uses these to
11725 deduce which programs it should invoke, and which command-line options
11726 it ought to place on their command lines. This behavior is controlled
11727 by @dfn{spec strings}. In most cases there is one spec string for each
11728 program that GCC can invoke, but a few programs have multiple spec
11729 strings to control their behavior. The spec strings built into GCC can
11730 be overridden by using the @option{-specs=} command-line switch to specify
11733 @dfn{Spec files} are plaintext files that are used to construct spec
11734 strings. They consist of a sequence of directives separated by blank
11735 lines. The type of directive is determined by the first non-whitespace
11736 character on the line, which can be one of the following:
11739 @item %@var{command}
11740 Issues a @var{command} to the spec file processor. The commands that can
11744 @item %include <@var{file}>
11745 @cindex @code{%include}
11746 Search for @var{file} and insert its text at the current point in the
11749 @item %include_noerr <@var{file}>
11750 @cindex @code{%include_noerr}
11751 Just like @samp{%include}, but do not generate an error message if the include
11752 file cannot be found.
11754 @item %rename @var{old_name} @var{new_name}
11755 @cindex @code{%rename}
11756 Rename the spec string @var{old_name} to @var{new_name}.
11760 @item *[@var{spec_name}]:
11761 This tells the compiler to create, override or delete the named spec
11762 string. All lines after this directive up to the next directive or
11763 blank line are considered to be the text for the spec string. If this
11764 results in an empty string then the spec is deleted. (Or, if the
11765 spec did not exist, then nothing happens.) Otherwise, if the spec
11766 does not currently exist a new spec is created. If the spec does
11767 exist then its contents are overridden by the text of this
11768 directive, unless the first character of that text is the @samp{+}
11769 character, in which case the text is appended to the spec.
11771 @item [@var{suffix}]:
11772 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11773 and up to the next directive or blank line are considered to make up the
11774 spec string for the indicated suffix. When the compiler encounters an
11775 input file with the named suffix, it processes the spec string in
11776 order to work out how to compile that file. For example:
11780 z-compile -input %i
11783 This says that any input file whose name ends in @samp{.ZZ} should be
11784 passed to the program @samp{z-compile}, which should be invoked with the
11785 command-line switch @option{-input} and with the result of performing the
11786 @samp{%i} substitution. (See below.)
11788 As an alternative to providing a spec string, the text following a
11789 suffix directive can be one of the following:
11792 @item @@@var{language}
11793 This says that the suffix is an alias for a known @var{language}. This is
11794 similar to using the @option{-x} command-line switch to GCC to specify a
11795 language explicitly. For example:
11802 Says that .ZZ files are, in fact, C++ source files.
11805 This causes an error messages saying:
11808 @var{name} compiler not installed on this system.
11812 GCC already has an extensive list of suffixes built into it.
11813 This directive adds an entry to the end of the list of suffixes, but
11814 since the list is searched from the end backwards, it is effectively
11815 possible to override earlier entries using this technique.
11819 GCC has the following spec strings built into it. Spec files can
11820 override these strings or create their own. Note that individual
11821 targets can also add their own spec strings to this list.
11824 asm Options to pass to the assembler
11825 asm_final Options to pass to the assembler post-processor
11826 cpp Options to pass to the C preprocessor
11827 cc1 Options to pass to the C compiler
11828 cc1plus Options to pass to the C++ compiler
11829 endfile Object files to include at the end of the link
11830 link Options to pass to the linker
11831 lib Libraries to include on the command line to the linker
11832 libgcc Decides which GCC support library to pass to the linker
11833 linker Sets the name of the linker
11834 predefines Defines to be passed to the C preprocessor
11835 signed_char Defines to pass to CPP to say whether @code{char} is signed
11837 startfile Object files to include at the start of the link
11840 Here is a small example of a spec file:
11843 %rename lib old_lib
11846 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11849 This example renames the spec called @samp{lib} to @samp{old_lib} and
11850 then overrides the previous definition of @samp{lib} with a new one.
11851 The new definition adds in some extra command-line options before
11852 including the text of the old definition.
11854 @dfn{Spec strings} are a list of command-line options to be passed to their
11855 corresponding program. In addition, the spec strings can contain
11856 @samp{%}-prefixed sequences to substitute variable text or to
11857 conditionally insert text into the command line. Using these constructs
11858 it is possible to generate quite complex command lines.
11860 Here is a table of all defined @samp{%}-sequences for spec
11861 strings. Note that spaces are not generated automatically around the
11862 results of expanding these sequences. Therefore you can concatenate them
11863 together or combine them with constant text in a single argument.
11867 Substitute one @samp{%} into the program name or argument.
11870 Substitute the name of the input file being processed.
11873 Substitute the basename of the input file being processed.
11874 This is the substring up to (and not including) the last period
11875 and not including the directory.
11878 This is the same as @samp{%b}, but include the file suffix (text after
11882 Marks the argument containing or following the @samp{%d} as a
11883 temporary file name, so that that file is deleted if GCC exits
11884 successfully. Unlike @samp{%g}, this contributes no text to the
11887 @item %g@var{suffix}
11888 Substitute a file name that has suffix @var{suffix} and is chosen
11889 once per compilation, and mark the argument in the same way as
11890 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11891 name is now chosen in a way that is hard to predict even when previously
11892 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11893 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11894 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11895 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11896 was simply substituted with a file name chosen once per compilation,
11897 without regard to any appended suffix (which was therefore treated
11898 just like ordinary text), making such attacks more likely to succeed.
11900 @item %u@var{suffix}
11901 Like @samp{%g}, but generates a new temporary file name
11902 each time it appears instead of once per compilation.
11904 @item %U@var{suffix}
11905 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11906 new one if there is no such last file name. In the absence of any
11907 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11908 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11909 involves the generation of two distinct file names, one
11910 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11911 simply substituted with a file name chosen for the previous @samp{%u},
11912 without regard to any appended suffix.
11914 @item %j@var{suffix}
11915 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11916 writable, and if @option{-save-temps} is not used;
11917 otherwise, substitute the name
11918 of a temporary file, just like @samp{%u}. This temporary file is not
11919 meant for communication between processes, but rather as a junk
11920 disposal mechanism.
11922 @item %|@var{suffix}
11923 @itemx %m@var{suffix}
11924 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11925 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11926 all. These are the two most common ways to instruct a program that it
11927 should read from standard input or write to standard output. If you
11928 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11929 construct: see for example @file{f/lang-specs.h}.
11931 @item %.@var{SUFFIX}
11932 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11933 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11934 terminated by the next space or %.
11937 Marks the argument containing or following the @samp{%w} as the
11938 designated output file of this compilation. This puts the argument
11939 into the sequence of arguments that @samp{%o} substitutes.
11942 Substitutes the names of all the output files, with spaces
11943 automatically placed around them. You should write spaces
11944 around the @samp{%o} as well or the results are undefined.
11945 @samp{%o} is for use in the specs for running the linker.
11946 Input files whose names have no recognized suffix are not compiled
11947 at all, but they are included among the output files, so they are
11951 Substitutes the suffix for object files. Note that this is
11952 handled specially when it immediately follows @samp{%g, %u, or %U},
11953 because of the need for those to form complete file names. The
11954 handling is such that @samp{%O} is treated exactly as if it had already
11955 been substituted, except that @samp{%g, %u, and %U} do not currently
11956 support additional @var{suffix} characters following @samp{%O} as they do
11957 following, for example, @samp{.o}.
11960 Substitutes the standard macro predefinitions for the
11961 current target machine. Use this when running @command{cpp}.
11964 Like @samp{%p}, but puts @samp{__} before and after the name of each
11965 predefined macro, except for macros that start with @samp{__} or with
11966 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11970 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11971 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11972 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11973 and @option{-imultilib} as necessary.
11976 Current argument is the name of a library or startup file of some sort.
11977 Search for that file in a standard list of directories and substitute
11978 the full name found. The current working directory is included in the
11979 list of directories scanned.
11982 Current argument is the name of a linker script. Search for that file
11983 in the current list of directories to scan for libraries. If the file
11984 is located insert a @option{--script} option into the command line
11985 followed by the full path name found. If the file is not found then
11986 generate an error message. Note: the current working directory is not
11990 Print @var{str} as an error message. @var{str} is terminated by a newline.
11991 Use this when inconsistent options are detected.
11993 @item %(@var{name})
11994 Substitute the contents of spec string @var{name} at this point.
11996 @item %x@{@var{option}@}
11997 Accumulate an option for @samp{%X}.
12000 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
12004 Output the accumulated assembler options specified by @option{-Wa}.
12007 Output the accumulated preprocessor options specified by @option{-Wp}.
12010 Process the @code{asm} spec. This is used to compute the
12011 switches to be passed to the assembler.
12014 Process the @code{asm_final} spec. This is a spec string for
12015 passing switches to an assembler post-processor, if such a program is
12019 Process the @code{link} spec. This is the spec for computing the
12020 command line passed to the linker. Typically it makes use of the
12021 @samp{%L %G %S %D and %E} sequences.
12024 Dump out a @option{-L} option for each directory that GCC believes might
12025 contain startup files. If the target supports multilibs then the
12026 current multilib directory is prepended to each of these paths.
12029 Process the @code{lib} spec. This is a spec string for deciding which
12030 libraries are included on the command line to the linker.
12033 Process the @code{libgcc} spec. This is a spec string for deciding
12034 which GCC support library is included on the command line to the linker.
12037 Process the @code{startfile} spec. This is a spec for deciding which
12038 object files are the first ones passed to the linker. Typically
12039 this might be a file named @file{crt0.o}.
12042 Process the @code{endfile} spec. This is a spec string that specifies
12043 the last object files that are passed to the linker.
12046 Process the @code{cpp} spec. This is used to construct the arguments
12047 to be passed to the C preprocessor.
12050 Process the @code{cc1} spec. This is used to construct the options to be
12051 passed to the actual C compiler (@command{cc1}).
12054 Process the @code{cc1plus} spec. This is used to construct the options to be
12055 passed to the actual C++ compiler (@command{cc1plus}).
12058 Substitute the variable part of a matched option. See below.
12059 Note that each comma in the substituted string is replaced by
12063 Remove all occurrences of @code{-S} from the command line. Note---this
12064 command is position dependent. @samp{%} commands in the spec string
12065 before this one see @code{-S}, @samp{%} commands in the spec string
12066 after this one do not.
12068 @item %:@var{function}(@var{args})
12069 Call the named function @var{function}, passing it @var{args}.
12070 @var{args} is first processed as a nested spec string, then split
12071 into an argument vector in the usual fashion. The function returns
12072 a string which is processed as if it had appeared literally as part
12073 of the current spec.
12075 The following built-in spec functions are provided:
12078 @item @code{getenv}
12079 The @code{getenv} spec function takes two arguments: an environment
12080 variable name and a string. If the environment variable is not
12081 defined, a fatal error is issued. Otherwise, the return value is the
12082 value of the environment variable concatenated with the string. For
12083 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
12086 %:getenv(TOPDIR /include)
12089 expands to @file{/path/to/top/include}.
12091 @item @code{if-exists}
12092 The @code{if-exists} spec function takes one argument, an absolute
12093 pathname to a file. If the file exists, @code{if-exists} returns the
12094 pathname. Here is a small example of its usage:
12098 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
12101 @item @code{if-exists-else}
12102 The @code{if-exists-else} spec function is similar to the @code{if-exists}
12103 spec function, except that it takes two arguments. The first argument is
12104 an absolute pathname to a file. If the file exists, @code{if-exists-else}
12105 returns the pathname. If it does not exist, it returns the second argument.
12106 This way, @code{if-exists-else} can be used to select one file or another,
12107 based on the existence of the first. Here is a small example of its usage:
12111 crt0%O%s %:if-exists(crti%O%s) \
12112 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
12115 @item @code{replace-outfile}
12116 The @code{replace-outfile} spec function takes two arguments. It looks for the
12117 first argument in the outfiles array and replaces it with the second argument. Here
12118 is a small example of its usage:
12121 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
12124 @item @code{remove-outfile}
12125 The @code{remove-outfile} spec function takes one argument. It looks for the
12126 first argument in the outfiles array and removes it. Here is a small example
12130 %:remove-outfile(-lm)
12133 @item @code{pass-through-libs}
12134 The @code{pass-through-libs} spec function takes any number of arguments. It
12135 finds any @option{-l} options and any non-options ending in @file{.a} (which it
12136 assumes are the names of linker input library archive files) and returns a
12137 result containing all the found arguments each prepended by
12138 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
12139 intended to be passed to the LTO linker plugin.
12142 %:pass-through-libs(%G %L %G)
12145 @item @code{print-asm-header}
12146 The @code{print-asm-header} function takes no arguments and simply
12147 prints a banner like:
12153 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12156 It is used to separate compiler options from assembler options
12157 in the @option{--target-help} output.
12160 @item %@{@code{S}@}
12161 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12162 If that switch is not specified, this substitutes nothing. Note that
12163 the leading dash is omitted when specifying this option, and it is
12164 automatically inserted if the substitution is performed. Thus the spec
12165 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12166 and outputs the command-line option @option{-foo}.
12168 @item %W@{@code{S}@}
12169 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12170 deleted on failure.
12172 @item %@{@code{S}*@}
12173 Substitutes all the switches specified to GCC whose names start
12174 with @code{-S}, but which also take an argument. This is used for
12175 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12176 GCC considers @option{-o foo} as being
12177 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12178 text, including the space. Thus two arguments are generated.
12180 @item %@{@code{S}*&@code{T}*@}
12181 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12182 (the order of @code{S} and @code{T} in the spec is not significant).
12183 There can be any number of ampersand-separated variables; for each the
12184 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12186 @item %@{@code{S}:@code{X}@}
12187 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12189 @item %@{!@code{S}:@code{X}@}
12190 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12192 @item %@{@code{S}*:@code{X}@}
12193 Substitutes @code{X} if one or more switches whose names start with
12194 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12195 once, no matter how many such switches appeared. However, if @code{%*}
12196 appears somewhere in @code{X}, then @code{X} is substituted once
12197 for each matching switch, with the @code{%*} replaced by the part of
12198 that switch matching the @code{*}.
12200 If @code{%*} appears as the last part of a spec sequence then a space
12201 is added after the end of the last substitution. If there is more
12202 text in the sequence, however, then a space is not generated. This
12203 allows the @code{%*} substitution to be used as part of a larger
12204 string. For example, a spec string like this:
12207 %@{mcu=*:--script=%*/memory.ld@}
12211 when matching an option like @option{-mcu=newchip} produces:
12214 --script=newchip/memory.ld
12217 @item %@{.@code{S}:@code{X}@}
12218 Substitutes @code{X}, if processing a file with suffix @code{S}.
12220 @item %@{!.@code{S}:@code{X}@}
12221 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12223 @item %@{,@code{S}:@code{X}@}
12224 Substitutes @code{X}, if processing a file for language @code{S}.
12226 @item %@{!,@code{S}:@code{X}@}
12227 Substitutes @code{X}, if not processing a file for language @code{S}.
12229 @item %@{@code{S}|@code{P}:@code{X}@}
12230 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12231 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12232 @code{*} sequences as well, although they have a stronger binding than
12233 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12234 alternatives must be starred, and only the first matching alternative
12237 For example, a spec string like this:
12240 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12244 outputs the following command-line options from the following input
12245 command-line options:
12250 -d fred.c -foo -baz -boggle
12251 -d jim.d -bar -baz -boggle
12254 @item %@{S:X; T:Y; :D@}
12256 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12257 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12258 be as many clauses as you need. This may be combined with @code{.},
12259 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12264 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12265 construct may contain other nested @samp{%} constructs or spaces, or
12266 even newlines. They are processed as usual, as described above.
12267 Trailing white space in @code{X} is ignored. White space may also
12268 appear anywhere on the left side of the colon in these constructs,
12269 except between @code{.} or @code{*} and the corresponding word.
12271 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12272 handled specifically in these constructs. If another value of
12273 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12274 @option{-W} switch is found later in the command line, the earlier
12275 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12276 just one letter, which passes all matching options.
12278 The character @samp{|} at the beginning of the predicate text is used to
12279 indicate that a command should be piped to the following command, but
12280 only if @option{-pipe} is specified.
12282 It is built into GCC which switches take arguments and which do not.
12283 (You might think it would be useful to generalize this to allow each
12284 compiler's spec to say which switches take arguments. But this cannot
12285 be done in a consistent fashion. GCC cannot even decide which input
12286 files have been specified without knowing which switches take arguments,
12287 and it must know which input files to compile in order to tell which
12290 GCC also knows implicitly that arguments starting in @option{-l} are to be
12291 treated as compiler output files, and passed to the linker in their
12292 proper position among the other output files.
12294 @c man begin OPTIONS
12296 @node Target Options
12297 @section Specifying Target Machine and Compiler Version
12298 @cindex target options
12299 @cindex cross compiling
12300 @cindex specifying machine version
12301 @cindex specifying compiler version and target machine
12302 @cindex compiler version, specifying
12303 @cindex target machine, specifying
12305 The usual way to run GCC is to run the executable called @command{gcc}, or
12306 @command{@var{machine}-gcc} when cross-compiling, or
12307 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12308 one that was installed last.
12310 @node Submodel Options
12311 @section Hardware Models and Configurations
12312 @cindex submodel options
12313 @cindex specifying hardware config
12314 @cindex hardware models and configurations, specifying
12315 @cindex machine dependent options
12317 Each target machine types can have its own
12318 special options, starting with @samp{-m}, to choose among various
12319 hardware models or configurations---for example, 68010 vs 68020,
12320 floating coprocessor or none. A single installed version of the
12321 compiler can compile for any model or configuration, according to the
12324 Some configurations of the compiler also support additional special
12325 options, usually for compatibility with other compilers on the same
12328 @c This list is ordered alphanumerically by subsection name.
12329 @c It should be the same order and spelling as these options are listed
12330 @c in Machine Dependent Options
12333 * AArch64 Options::
12334 * Adapteva Epiphany Options::
12338 * Blackfin Options::
12343 * DEC Alpha Options::
12347 * GNU/Linux Options::
12357 * MicroBlaze Options::
12360 * MN10300 Options::
12364 * Nios II Options::
12365 * Nvidia PTX Options::
12367 * picoChip Options::
12368 * PowerPC Options::
12370 * RS/6000 and PowerPC Options::
12372 * S/390 and zSeries Options::
12375 * Solaris 2 Options::
12378 * System V Options::
12379 * TILE-Gx Options::
12380 * TILEPro Options::
12385 * VxWorks Options::
12387 * x86 Windows Options::
12388 * Xstormy16 Options::
12390 * zSeries Options::
12393 @node AArch64 Options
12394 @subsection AArch64 Options
12395 @cindex AArch64 Options
12397 These options are defined for AArch64 implementations:
12401 @item -mabi=@var{name}
12403 Generate code for the specified data model. Permissible values
12404 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12405 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12406 but long int and pointer are 64-bit.
12408 The default depends on the specific target configuration. Note that
12409 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12410 entire program with the same ABI, and link with a compatible set of libraries.
12413 @opindex mbig-endian
12414 Generate big-endian code. This is the default when GCC is configured for an
12415 @samp{aarch64_be-*-*} target.
12417 @item -mgeneral-regs-only
12418 @opindex mgeneral-regs-only
12419 Generate code which uses only the general-purpose registers. This is equivalent
12420 to feature modifier @option{nofp} of @option{-march} or @option{-mcpu}, except
12421 that @option{-mgeneral-regs-only} takes precedence over any conflicting feature
12422 modifier regardless of sequence.
12424 @item -mlittle-endian
12425 @opindex mlittle-endian
12426 Generate little-endian code. This is the default when GCC is configured for an
12427 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12429 @item -mcmodel=tiny
12430 @opindex mcmodel=tiny
12431 Generate code for the tiny code model. The program and its statically defined
12432 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12433 be statically or dynamically linked. This model is not fully implemented and
12434 mostly treated as @samp{small}.
12436 @item -mcmodel=small
12437 @opindex mcmodel=small
12438 Generate code for the small code model. The program and its statically defined
12439 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12440 be statically or dynamically linked. This is the default code model.
12442 @item -mcmodel=large
12443 @opindex mcmodel=large
12444 Generate code for the large code model. This makes no assumptions about
12445 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12446 statically linked only.
12448 @item -mstrict-align
12449 @opindex mstrict-align
12450 Do not assume that unaligned memory references are handled by the system.
12452 @item -momit-leaf-frame-pointer
12453 @itemx -mno-omit-leaf-frame-pointer
12454 @opindex momit-leaf-frame-pointer
12455 @opindex mno-omit-leaf-frame-pointer
12456 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12459 @item -mtls-dialect=desc
12460 @opindex mtls-dialect=desc
12461 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12462 of TLS variables. This is the default.
12464 @item -mtls-dialect=traditional
12465 @opindex mtls-dialect=traditional
12466 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12469 @item -mtls-size=@var{size}
12471 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
12472 This option depends on binutils higher than 2.25.
12474 @item -mfix-cortex-a53-835769
12475 @itemx -mno-fix-cortex-a53-835769
12476 @opindex mfix-cortex-a53-835769
12477 @opindex mno-fix-cortex-a53-835769
12478 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12479 This involves inserting a NOP instruction between memory instructions and
12480 64-bit integer multiply-accumulate instructions.
12482 @item -mfix-cortex-a53-843419
12483 @itemx -mno-fix-cortex-a53-843419
12484 @opindex mfix-cortex-a53-843419
12485 @opindex mno-fix-cortex-a53-843419
12486 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12487 This erratum workaround is made at link time and this will only pass the
12488 corresponding flag to the linker.
12490 @item -march=@var{name}
12492 Specify the name of the target architecture, optionally suffixed by one or
12493 more feature modifiers. This option has the form
12494 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12496 The permissible values for @var{arch} are @samp{armv8-a} or
12499 For the permissible values for @var{feature}, see the sub-section on
12500 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12501 Feature Modifiers}. Where conflicting feature modifiers are
12502 specified, the right-most feature is used.
12504 Additionally on native AArch64 GNU/Linux systems the value
12505 @samp{native} is available. This option causes the compiler to pick the
12506 architecture of the host system. If the compiler is unable to recognize the
12507 architecture of the host system this option has no effect.
12509 GCC uses @var{name} to determine what kind of instructions it can emit
12510 when generating assembly code. If @option{-march} is specified
12511 without either of @option{-mtune} or @option{-mcpu} also being
12512 specified, the code is tuned to perform well across a range of target
12513 processors implementing the target architecture.
12515 @item -mtune=@var{name}
12517 Specify the name of the target processor for which GCC should tune the
12518 performance of the code. Permissible values for this option are:
12519 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
12520 @samp{exynos-m1}, @samp{thunderx}, @samp{xgene1}.
12522 Additionally, this option can specify that GCC should tune the performance
12523 of the code for a big.LITTLE system. Permissible values for this
12524 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12526 Additionally on native AArch64 GNU/Linux systems the value
12527 @samp{native} is available. This option causes the compiler to pick
12528 the architecture of and tune the performance of the code for the
12529 processor of the host system. If the compiler is unable to recognize
12530 the processor of the host system this option has no effect.
12532 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12533 are specified, the code is tuned to perform well across a range
12534 of target processors.
12536 This option cannot be suffixed by feature modifiers.
12538 @item -mcpu=@var{name}
12540 Specify the name of the target processor, optionally suffixed by one
12541 or more feature modifiers. This option has the form
12542 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
12543 the permissible values for @var{cpu} are the same as those available
12544 for @option{-mtune}. The permissible values for @var{feature} are
12545 documented in the sub-section on
12546 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12547 Feature Modifiers}. Where conflicting feature modifiers are
12548 specified, the right-most feature is used.
12550 Additionally on native AArch64 GNU/Linux systems the value
12551 @samp{native} is available. This option causes the compiler to tune
12552 the performance of the code for the processor of the host system. If
12553 the compiler is unable to recognize the processor of the host system
12554 this option has no effect.
12556 GCC uses @var{name} to determine what kind of instructions it can emit when
12557 generating assembly code (as if by @option{-march}) and to determine
12558 the target processor for which to tune for performance (as if
12559 by @option{-mtune}). Where this option is used in conjunction
12560 with @option{-march} or @option{-mtune}, those options take precedence
12561 over the appropriate part of this option.
12563 @item -moverride=@var{string}
12565 Override tuning decisions made by the back-end in response to a
12566 @option{-mtune=} switch. The syntax, semantics, and accepted values
12567 for @var{string} in this option are not guaranteed to be consistent
12570 This option is only intended to be useful when developing GCC.
12572 @item -mpc-relative-literal-loads
12573 @opindex mpcrelativeliteralloads
12574 Enable PC relative literal loads. If this option is used, literal
12575 pools are assumed to have a range of up to 1MiB and an appropriate
12576 instruction sequence is used. This option has no impact when used
12577 with @option{-mcmodel=tiny}.
12581 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12582 @anchor{aarch64-feature-modifiers}
12583 @cindex @option{-march} feature modifiers
12584 @cindex @option{-mcpu} feature modifiers
12585 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
12586 the following and their inverses @option{no@var{feature}}:
12590 Enable CRC extension.
12592 Enable Crypto extension. This also enables Advanced SIMD and floating-point
12595 Enable floating-point instructions. This is on by default for all possible
12596 values for options @option{-march} and @option{-mcpu}.
12598 Enable Advanced SIMD instructions. This also enables floating-point
12599 instructions. This is on by default for all possible values for options
12600 @option{-march} and @option{-mcpu}.
12602 Enable Large System Extension instructions.
12604 Enable Privileged Access Never support.
12606 Enable Limited Ordering Regions support.
12608 Enable ARMv8.1 Advanced SIMD instructions. This implies Advanced SIMD
12613 That is, @option{crypto} implies @option{simd} implies @option{fp}.
12614 Conversely, @option{nofp} (or equivalently, @option{-mgeneral-regs-only})
12615 implies @option{nosimd} implies @option{nocrypto}.
12617 @node Adapteva Epiphany Options
12618 @subsection Adapteva Epiphany Options
12620 These @samp{-m} options are defined for Adapteva Epiphany:
12623 @item -mhalf-reg-file
12624 @opindex mhalf-reg-file
12625 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12626 That allows code to run on hardware variants that lack these registers.
12628 @item -mprefer-short-insn-regs
12629 @opindex mprefer-short-insn-regs
12630 Preferrentially allocate registers that allow short instruction generation.
12631 This can result in increased instruction count, so this may either reduce or
12632 increase overall code size.
12634 @item -mbranch-cost=@var{num}
12635 @opindex mbranch-cost
12636 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12637 This cost is only a heuristic and is not guaranteed to produce
12638 consistent results across releases.
12642 Enable the generation of conditional moves.
12644 @item -mnops=@var{num}
12646 Emit @var{num} NOPs before every other generated instruction.
12648 @item -mno-soft-cmpsf
12649 @opindex mno-soft-cmpsf
12650 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12651 and test the flags. This is faster than a software comparison, but can
12652 get incorrect results in the presence of NaNs, or when two different small
12653 numbers are compared such that their difference is calculated as zero.
12654 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12655 software comparisons.
12657 @item -mstack-offset=@var{num}
12658 @opindex mstack-offset
12659 Set the offset between the top of the stack and the stack pointer.
12660 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12661 can be used by leaf functions without stack allocation.
12662 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12663 Note also that this option changes the ABI; compiling a program with a
12664 different stack offset than the libraries have been compiled with
12665 generally does not work.
12666 This option can be useful if you want to evaluate if a different stack
12667 offset would give you better code, but to actually use a different stack
12668 offset to build working programs, it is recommended to configure the
12669 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12671 @item -mno-round-nearest
12672 @opindex mno-round-nearest
12673 Make the scheduler assume that the rounding mode has been set to
12674 truncating. The default is @option{-mround-nearest}.
12677 @opindex mlong-calls
12678 If not otherwise specified by an attribute, assume all calls might be beyond
12679 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12680 function address into a register before performing a (otherwise direct) call.
12681 This is the default.
12683 @item -mshort-calls
12684 @opindex short-calls
12685 If not otherwise specified by an attribute, assume all direct calls are
12686 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12687 for direct calls. The default is @option{-mlong-calls}.
12691 Assume addresses can be loaded as 16-bit unsigned values. This does not
12692 apply to function addresses for which @option{-mlong-calls} semantics
12695 @item -mfp-mode=@var{mode}
12697 Set the prevailing mode of the floating-point unit.
12698 This determines the floating-point mode that is provided and expected
12699 at function call and return time. Making this mode match the mode you
12700 predominantly need at function start can make your programs smaller and
12701 faster by avoiding unnecessary mode switches.
12703 @var{mode} can be set to one the following values:
12707 Any mode at function entry is valid, and retained or restored when
12708 the function returns, and when it calls other functions.
12709 This mode is useful for compiling libraries or other compilation units
12710 you might want to incorporate into different programs with different
12711 prevailing FPU modes, and the convenience of being able to use a single
12712 object file outweighs the size and speed overhead for any extra
12713 mode switching that might be needed, compared with what would be needed
12714 with a more specific choice of prevailing FPU mode.
12717 This is the mode used for floating-point calculations with
12718 truncating (i.e.@: round towards zero) rounding mode. That includes
12719 conversion from floating point to integer.
12721 @item round-nearest
12722 This is the mode used for floating-point calculations with
12723 round-to-nearest-or-even rounding mode.
12726 This is the mode used to perform integer calculations in the FPU, e.g.@:
12727 integer multiply, or integer multiply-and-accumulate.
12730 The default is @option{-mfp-mode=caller}
12732 @item -mnosplit-lohi
12733 @itemx -mno-postinc
12734 @itemx -mno-postmodify
12735 @opindex mnosplit-lohi
12736 @opindex mno-postinc
12737 @opindex mno-postmodify
12738 Code generation tweaks that disable, respectively, splitting of 32-bit
12739 loads, generation of post-increment addresses, and generation of
12740 post-modify addresses. The defaults are @option{msplit-lohi},
12741 @option{-mpost-inc}, and @option{-mpost-modify}.
12743 @item -mnovect-double
12744 @opindex mno-vect-double
12745 Change the preferred SIMD mode to SImode. The default is
12746 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12748 @item -max-vect-align=@var{num}
12749 @opindex max-vect-align
12750 The maximum alignment for SIMD vector mode types.
12751 @var{num} may be 4 or 8. The default is 8.
12752 Note that this is an ABI change, even though many library function
12753 interfaces are unaffected if they don't use SIMD vector modes
12754 in places that affect size and/or alignment of relevant types.
12756 @item -msplit-vecmove-early
12757 @opindex msplit-vecmove-early
12758 Split vector moves into single word moves before reload. In theory this
12759 can give better register allocation, but so far the reverse seems to be
12760 generally the case.
12762 @item -m1reg-@var{reg}
12764 Specify a register to hold the constant @minus{}1, which makes loading small negative
12765 constants and certain bitmasks faster.
12766 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12767 which specify use of that register as a fixed register,
12768 and @samp{none}, which means that no register is used for this
12769 purpose. The default is @option{-m1reg-none}.
12774 @subsection ARC Options
12775 @cindex ARC options
12777 The following options control the architecture variant for which code
12780 @c architecture variants
12783 @item -mbarrel-shifter
12784 @opindex mbarrel-shifter
12785 Generate instructions supported by barrel shifter. This is the default
12786 unless @option{-mcpu=ARC601} is in effect.
12788 @item -mcpu=@var{cpu}
12790 Set architecture type, register usage, and instruction scheduling
12791 parameters for @var{cpu}. There are also shortcut alias options
12792 available for backward compatibility and convenience. Supported
12793 values for @var{cpu} are
12799 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12803 Compile for ARC601. Alias: @option{-mARC601}.
12808 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12809 This is the default when configured with @option{--with-cpu=arc700}@.
12814 @itemx -mdpfp-compact
12815 @opindex mdpfp-compact
12816 FPX: Generate Double Precision FPX instructions, tuned for the compact
12820 @opindex mdpfp-fast
12821 FPX: Generate Double Precision FPX instructions, tuned for the fast
12824 @item -mno-dpfp-lrsr
12825 @opindex mno-dpfp-lrsr
12826 Disable LR and SR instructions from using FPX extension aux registers.
12830 Generate Extended arithmetic instructions. Currently only
12831 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12832 supported. This is always enabled for @option{-mcpu=ARC700}.
12836 Do not generate mpy instructions for ARC700.
12840 Generate 32x16 bit multiply and mac instructions.
12844 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12848 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12853 @itemx -mspfp-compact
12854 @opindex mspfp-compact
12855 FPX: Generate Single Precision FPX instructions, tuned for the compact
12859 @opindex mspfp-fast
12860 FPX: Generate Single Precision FPX instructions, tuned for the fast
12865 Enable generation of ARC SIMD instructions via target-specific
12866 builtins. Only valid for @option{-mcpu=ARC700}.
12869 @opindex msoft-float
12870 This option ignored; it is provided for compatibility purposes only.
12871 Software floating point code is emitted by default, and this default
12872 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12873 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12874 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12878 Generate swap instructions.
12882 The following options are passed through to the assembler, and also
12883 define preprocessor macro symbols.
12885 @c Flags used by the assembler, but for which we define preprocessor
12886 @c macro symbols as well.
12889 @opindex mdsp-packa
12890 Passed down to the assembler to enable the DSP Pack A extensions.
12891 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12895 Passed down to the assembler to enable the dual viterbi butterfly
12896 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12898 @c ARC700 4.10 extension instruction
12901 Passed down to the assembler to enable the Locked Load/Store
12902 Conditional extension. Also sets the preprocessor symbol
12907 Passed down to the assembler. Also sets the preprocessor symbol
12908 @code{__Xxmac_d16}.
12912 Passed down to the assembler. Also sets the preprocessor symbol
12915 @c ARC700 4.10 extension instruction
12918 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12919 extension instruction. Also sets the preprocessor symbol
12922 @c ARC700 4.10 extension instruction
12925 Passed down to the assembler to enable the swap byte ordering
12926 extension instruction. Also sets the preprocessor symbol
12930 @opindex mtelephony
12931 Passed down to the assembler to enable dual and single operand
12932 instructions for telephony. Also sets the preprocessor symbol
12933 @code{__Xtelephony}.
12937 Passed down to the assembler to enable the XY Memory extension. Also
12938 sets the preprocessor symbol @code{__Xxy}.
12942 The following options control how the assembly code is annotated:
12944 @c Assembly annotation options
12948 Annotate assembler instructions with estimated addresses.
12950 @item -mannotate-align
12951 @opindex mannotate-align
12952 Explain what alignment considerations lead to the decision to make an
12953 instruction short or long.
12957 The following options are passed through to the linker:
12959 @c options passed through to the linker
12963 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12964 This option is enabled by default in tool chains built for
12965 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12966 when profiling is not requested.
12968 @item -marclinux_prof
12969 @opindex marclinux_prof
12970 Passed through to the linker, to specify use of the
12971 @code{arclinux_prof} emulation. This option is enabled by default in
12972 tool chains built for @w{@code{arc-linux-uclibc}} and
12973 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12977 The following options control the semantics of generated code:
12979 @c semantically relevant code generation options
12981 @item -mepilogue-cfi
12982 @opindex mepilogue-cfi
12983 Enable generation of call frame information for epilogues.
12985 @item -mno-epilogue-cfi
12986 @opindex mno-epilogue-cfi
12987 Disable generation of call frame information for epilogues.
12990 @opindex mlong-calls
12991 Generate call insns as register indirect calls, thus providing access
12992 to the full 32-bit address range.
12994 @item -mmedium-calls
12995 @opindex mmedium-calls
12996 Don't use less than 25 bit addressing range for calls, which is the
12997 offset available for an unconditional branch-and-link
12998 instruction. Conditional execution of function calls is suppressed, to
12999 allow use of the 25-bit range, rather than the 21-bit range with
13000 conditional branch-and-link. This is the default for tool chains built
13001 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13005 Do not generate sdata references. This is the default for tool chains
13006 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13010 @opindex mucb-mcount
13011 Instrument with mcount calls as used in UCB code. I.e. do the
13012 counting in the callee, not the caller. By default ARC instrumentation
13013 counts in the caller.
13015 @item -mvolatile-cache
13016 @opindex mvolatile-cache
13017 Use ordinarily cached memory accesses for volatile references. This is the
13020 @item -mno-volatile-cache
13021 @opindex mno-volatile-cache
13022 Enable cache bypass for volatile references.
13026 The following options fine tune code generation:
13027 @c code generation tuning options
13030 @opindex malign-call
13031 Do alignment optimizations for call instructions.
13033 @item -mauto-modify-reg
13034 @opindex mauto-modify-reg
13035 Enable the use of pre/post modify with register displacement.
13037 @item -mbbit-peephole
13038 @opindex mbbit-peephole
13039 Enable bbit peephole2.
13043 This option disables a target-specific pass in @file{arc_reorg} to
13044 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13045 generation driven by the combiner pass.
13047 @item -mcase-vector-pcrel
13048 @opindex mcase-vector-pcrel
13049 Use pc-relative switch case tables - this enables case table shortening.
13050 This is the default for @option{-Os}.
13052 @item -mcompact-casesi
13053 @opindex mcompact-casesi
13054 Enable compact casesi pattern.
13055 This is the default for @option{-Os}.
13057 @item -mno-cond-exec
13058 @opindex mno-cond-exec
13059 Disable ARCompact specific pass to generate conditional execution instructions.
13060 Due to delay slot scheduling and interactions between operand numbers,
13061 literal sizes, instruction lengths, and the support for conditional execution,
13062 the target-independent pass to generate conditional execution is often lacking,
13063 so the ARC port has kept a special pass around that tries to find more
13064 conditional execution generating opportunities after register allocation,
13065 branch shortening, and delay slot scheduling have been done. This pass
13066 generally, but not always, improves performance and code size, at the cost of
13067 extra compilation time, which is why there is an option to switch it off.
13068 If you have a problem with call instructions exceeding their allowable
13069 offset range because they are conditionalized, you should consider using
13070 @option{-mmedium-calls} instead.
13072 @item -mearly-cbranchsi
13073 @opindex mearly-cbranchsi
13074 Enable pre-reload use of the cbranchsi pattern.
13076 @item -mexpand-adddi
13077 @opindex mexpand-adddi
13078 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13079 @code{add.f}, @code{adc} etc.
13081 @item -mindexed-loads
13082 @opindex mindexed-loads
13083 Enable the use of indexed loads. This can be problematic because some
13084 optimizers then assume that indexed stores exist, which is not
13089 Enable Local Register Allocation. This is still experimental for ARC,
13090 so by default the compiler uses standard reload
13091 (i.e. @option{-mno-lra}).
13093 @item -mlra-priority-none
13094 @opindex mlra-priority-none
13095 Don't indicate any priority for target registers.
13097 @item -mlra-priority-compact
13098 @opindex mlra-priority-compact
13099 Indicate target register priority for r0..r3 / r12..r15.
13101 @item -mlra-priority-noncompact
13102 @opindex mlra-priority-noncompact
13103 Reduce target regsiter priority for r0..r3 / r12..r15.
13105 @item -mno-millicode
13106 @opindex mno-millicode
13107 When optimizing for size (using @option{-Os}), prologues and epilogues
13108 that have to save or restore a large number of registers are often
13109 shortened by using call to a special function in libgcc; this is
13110 referred to as a @emph{millicode} call. As these calls can pose
13111 performance issues, and/or cause linking issues when linking in a
13112 nonstandard way, this option is provided to turn off millicode call
13116 @opindex mmixed-code
13117 Tweak register allocation to help 16-bit instruction generation.
13118 This generally has the effect of decreasing the average instruction size
13119 while increasing the instruction count.
13123 Enable 'q' instruction alternatives.
13124 This is the default for @option{-Os}.
13128 Enable Rcq constraint handling - most short code generation depends on this.
13129 This is the default.
13133 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13134 This is the default.
13136 @item -msize-level=@var{level}
13137 @opindex msize-level
13138 Fine-tune size optimization with regards to instruction lengths and alignment.
13139 The recognized values for @var{level} are:
13142 No size optimization. This level is deprecated and treated like @samp{1}.
13145 Short instructions are used opportunistically.
13148 In addition, alignment of loops and of code after barriers are dropped.
13151 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13155 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13156 the behavior when this is not set is equivalent to level @samp{1}.
13158 @item -mtune=@var{cpu}
13160 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13161 by @option{-mcpu=}.
13163 Supported values for @var{cpu} are
13167 Tune for ARC600 cpu.
13170 Tune for ARC601 cpu.
13173 Tune for ARC700 cpu with standard multiplier block.
13176 Tune for ARC700 cpu with XMAC block.
13179 Tune for ARC725D cpu.
13182 Tune for ARC750D cpu.
13186 @item -mmultcost=@var{num}
13188 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13189 normal instruction.
13191 @item -munalign-prob-threshold=@var{probability}
13192 @opindex munalign-prob-threshold
13193 Set probability threshold for unaligning branches.
13194 When tuning for @samp{ARC700} and optimizing for speed, branches without
13195 filled delay slot are preferably emitted unaligned and long, unless
13196 profiling indicates that the probability for the branch to be taken
13197 is below @var{probability}. @xref{Cross-profiling}.
13198 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13202 The following options are maintained for backward compatibility, but
13203 are now deprecated and will be removed in a future release:
13205 @c Deprecated options
13213 @opindex mbig-endian
13216 Compile code for big endian targets. Use of these options is now
13217 deprecated. Users wanting big-endian code, should use the
13218 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13219 building the tool chain, for which big-endian is the default.
13221 @item -mlittle-endian
13222 @opindex mlittle-endian
13225 Compile code for little endian targets. Use of these options is now
13226 deprecated. Users wanting little-endian code should use the
13227 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13228 building the tool chain, for which little-endian is the default.
13230 @item -mbarrel_shifter
13231 @opindex mbarrel_shifter
13232 Replaced by @option{-mbarrel-shifter}.
13234 @item -mdpfp_compact
13235 @opindex mdpfp_compact
13236 Replaced by @option{-mdpfp-compact}.
13239 @opindex mdpfp_fast
13240 Replaced by @option{-mdpfp-fast}.
13243 @opindex mdsp_packa
13244 Replaced by @option{-mdsp-packa}.
13248 Replaced by @option{-mea}.
13252 Replaced by @option{-mmac-24}.
13256 Replaced by @option{-mmac-d16}.
13258 @item -mspfp_compact
13259 @opindex mspfp_compact
13260 Replaced by @option{-mspfp-compact}.
13263 @opindex mspfp_fast
13264 Replaced by @option{-mspfp-fast}.
13266 @item -mtune=@var{cpu}
13268 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13269 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13270 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13272 @item -multcost=@var{num}
13274 Replaced by @option{-mmultcost}.
13279 @subsection ARM Options
13280 @cindex ARM options
13282 These @samp{-m} options are defined for the ARM port:
13285 @item -mabi=@var{name}
13287 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13288 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13291 @opindex mapcs-frame
13292 Generate a stack frame that is compliant with the ARM Procedure Call
13293 Standard for all functions, even if this is not strictly necessary for
13294 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13295 with this option causes the stack frames not to be generated for
13296 leaf functions. The default is @option{-mno-apcs-frame}.
13297 This option is deprecated.
13301 This is a synonym for @option{-mapcs-frame} and is deprecated.
13304 @c not currently implemented
13305 @item -mapcs-stack-check
13306 @opindex mapcs-stack-check
13307 Generate code to check the amount of stack space available upon entry to
13308 every function (that actually uses some stack space). If there is
13309 insufficient space available then either the function
13310 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13311 called, depending upon the amount of stack space required. The runtime
13312 system is required to provide these functions. The default is
13313 @option{-mno-apcs-stack-check}, since this produces smaller code.
13315 @c not currently implemented
13317 @opindex mapcs-float
13318 Pass floating-point arguments using the floating-point registers. This is
13319 one of the variants of the APCS@. This option is recommended if the
13320 target hardware has a floating-point unit or if a lot of floating-point
13321 arithmetic is going to be performed by the code. The default is
13322 @option{-mno-apcs-float}, since the size of integer-only code is
13323 slightly increased if @option{-mapcs-float} is used.
13325 @c not currently implemented
13326 @item -mapcs-reentrant
13327 @opindex mapcs-reentrant
13328 Generate reentrant, position-independent code. The default is
13329 @option{-mno-apcs-reentrant}.
13332 @item -mthumb-interwork
13333 @opindex mthumb-interwork
13334 Generate code that supports calling between the ARM and Thumb
13335 instruction sets. Without this option, on pre-v5 architectures, the
13336 two instruction sets cannot be reliably used inside one program. The
13337 default is @option{-mno-thumb-interwork}, since slightly larger code
13338 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13339 configurations this option is meaningless.
13341 @item -mno-sched-prolog
13342 @opindex mno-sched-prolog
13343 Prevent the reordering of instructions in the function prologue, or the
13344 merging of those instruction with the instructions in the function's
13345 body. This means that all functions start with a recognizable set
13346 of instructions (or in fact one of a choice from a small set of
13347 different function prologues), and this information can be used to
13348 locate the start of functions inside an executable piece of code. The
13349 default is @option{-msched-prolog}.
13351 @item -mfloat-abi=@var{name}
13352 @opindex mfloat-abi
13353 Specifies which floating-point ABI to use. Permissible values
13354 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13356 Specifying @samp{soft} causes GCC to generate output containing
13357 library calls for floating-point operations.
13358 @samp{softfp} allows the generation of code using hardware floating-point
13359 instructions, but still uses the soft-float calling conventions.
13360 @samp{hard} allows generation of floating-point instructions
13361 and uses FPU-specific calling conventions.
13363 The default depends on the specific target configuration. Note that
13364 the hard-float and soft-float ABIs are not link-compatible; you must
13365 compile your entire program with the same ABI, and link with a
13366 compatible set of libraries.
13368 @item -mlittle-endian
13369 @opindex mlittle-endian
13370 Generate code for a processor running in little-endian mode. This is
13371 the default for all standard configurations.
13374 @opindex mbig-endian
13375 Generate code for a processor running in big-endian mode; the default is
13376 to compile code for a little-endian processor.
13378 @item -march=@var{name}
13380 This specifies the name of the target ARM architecture. GCC uses this
13381 name to determine what kind of instructions it can emit when generating
13382 assembly code. This option can be used in conjunction with or instead
13383 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13384 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13385 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13386 @samp{armv6}, @samp{armv6j},
13387 @samp{armv6t2}, @samp{armv6z}, @samp{armv6kz}, @samp{armv6-m},
13388 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13389 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
13390 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13392 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13395 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13396 architecture together with the optional CRC32 extensions.
13398 @option{-march=native} causes the compiler to auto-detect the architecture
13399 of the build computer. At present, this feature is only supported on
13400 GNU/Linux, and not all architectures are recognized. If the auto-detect
13401 is unsuccessful the option has no effect.
13403 @item -mtune=@var{name}
13405 This option specifies the name of the target ARM processor for
13406 which GCC should tune the performance of the code.
13407 For some ARM implementations better performance can be obtained by using
13409 Permissible names are: @samp{arm2}, @samp{arm250},
13410 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13411 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13412 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13413 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13415 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13416 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13417 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13418 @samp{strongarm1110},
13419 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13420 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13421 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13422 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13423 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13424 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13425 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13426 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
13427 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
13428 @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
13430 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13435 @samp{cortex-m0plus},
13436 @samp{cortex-m1.small-multiply},
13437 @samp{cortex-m0.small-multiply},
13438 @samp{cortex-m0plus.small-multiply},
13440 @samp{marvell-pj4},
13441 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13442 @samp{fa526}, @samp{fa626},
13443 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13446 Additionally, this option can specify that GCC should tune the performance
13447 of the code for a big.LITTLE system. Permissible names are:
13448 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
13449 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
13451 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13452 performance for a blend of processors within architecture @var{arch}.
13453 The aim is to generate code that run well on the current most popular
13454 processors, balancing between optimizations that benefit some CPUs in the
13455 range, and avoiding performance pitfalls of other CPUs. The effects of
13456 this option may change in future GCC versions as CPU models come and go.
13458 @option{-mtune=native} causes the compiler to auto-detect the CPU
13459 of the build computer. At present, this feature is only supported on
13460 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13461 unsuccessful the option has no effect.
13463 @item -mcpu=@var{name}
13465 This specifies the name of the target ARM processor. GCC uses this name
13466 to derive the name of the target ARM architecture (as if specified
13467 by @option{-march}) and the ARM processor type for which to tune for
13468 performance (as if specified by @option{-mtune}). Where this option
13469 is used in conjunction with @option{-march} or @option{-mtune},
13470 those options take precedence over the appropriate part of this option.
13472 Permissible names for this option are the same as those for
13475 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13476 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13477 See @option{-mtune} for more information.
13479 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13480 of the build computer. At present, this feature is only supported on
13481 GNU/Linux, and not all architectures are recognized. If the auto-detect
13482 is unsuccessful the option has no effect.
13484 @item -mfpu=@var{name}
13486 This specifies what floating-point hardware (or hardware emulation) is
13487 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13488 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13489 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13490 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13491 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13492 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13494 If @option{-msoft-float} is specified this specifies the format of
13495 floating-point values.
13497 If the selected floating-point hardware includes the NEON extension
13498 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13499 operations are not generated by GCC's auto-vectorization pass unless
13500 @option{-funsafe-math-optimizations} is also specified. This is
13501 because NEON hardware does not fully implement the IEEE 754 standard for
13502 floating-point arithmetic (in particular denormal values are treated as
13503 zero), so the use of NEON instructions may lead to a loss of precision.
13505 @item -mfp16-format=@var{name}
13506 @opindex mfp16-format
13507 Specify the format of the @code{__fp16} half-precision floating-point type.
13508 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13509 the default is @samp{none}, in which case the @code{__fp16} type is not
13510 defined. @xref{Half-Precision}, for more information.
13512 @item -mstructure-size-boundary=@var{n}
13513 @opindex mstructure-size-boundary
13514 The sizes of all structures and unions are rounded up to a multiple
13515 of the number of bits set by this option. Permissible values are 8, 32
13516 and 64. The default value varies for different toolchains. For the COFF
13517 targeted toolchain the default value is 8. A value of 64 is only allowed
13518 if the underlying ABI supports it.
13520 Specifying a larger number can produce faster, more efficient code, but
13521 can also increase the size of the program. Different values are potentially
13522 incompatible. Code compiled with one value cannot necessarily expect to
13523 work with code or libraries compiled with another value, if they exchange
13524 information using structures or unions.
13526 @item -mabort-on-noreturn
13527 @opindex mabort-on-noreturn
13528 Generate a call to the function @code{abort} at the end of a
13529 @code{noreturn} function. It is executed if the function tries to
13533 @itemx -mno-long-calls
13534 @opindex mlong-calls
13535 @opindex mno-long-calls
13536 Tells the compiler to perform function calls by first loading the
13537 address of the function into a register and then performing a subroutine
13538 call on this register. This switch is needed if the target function
13539 lies outside of the 64-megabyte addressing range of the offset-based
13540 version of subroutine call instruction.
13542 Even if this switch is enabled, not all function calls are turned
13543 into long calls. The heuristic is that static functions, functions
13544 that have the @code{short_call} attribute, functions that are inside
13545 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13546 definitions have already been compiled within the current compilation
13547 unit are not turned into long calls. The exceptions to this rule are
13548 that weak function definitions, functions with the @code{long_call}
13549 attribute or the @code{section} attribute, and functions that are within
13550 the scope of a @code{#pragma long_calls} directive are always
13551 turned into long calls.
13553 This feature is not enabled by default. Specifying
13554 @option{-mno-long-calls} restores the default behavior, as does
13555 placing the function calls within the scope of a @code{#pragma
13556 long_calls_off} directive. Note these switches have no effect on how
13557 the compiler generates code to handle function calls via function
13560 @item -msingle-pic-base
13561 @opindex msingle-pic-base
13562 Treat the register used for PIC addressing as read-only, rather than
13563 loading it in the prologue for each function. The runtime system is
13564 responsible for initializing this register with an appropriate value
13565 before execution begins.
13567 @item -mpic-register=@var{reg}
13568 @opindex mpic-register
13569 Specify the register to be used for PIC addressing.
13570 For standard PIC base case, the default is any suitable register
13571 determined by compiler. For single PIC base case, the default is
13572 @samp{R9} if target is EABI based or stack-checking is enabled,
13573 otherwise the default is @samp{R10}.
13575 @item -mpic-data-is-text-relative
13576 @opindex mpic-data-is-text-relative
13577 Assume that each data segments are relative to text segment at load time.
13578 Therefore, it permits addressing data using PC-relative operations.
13579 This option is on by default for targets other than VxWorks RTP.
13581 @item -mpoke-function-name
13582 @opindex mpoke-function-name
13583 Write the name of each function into the text section, directly
13584 preceding the function prologue. The generated code is similar to this:
13588 .ascii "arm_poke_function_name", 0
13591 .word 0xff000000 + (t1 - t0)
13592 arm_poke_function_name
13594 stmfd sp!, @{fp, ip, lr, pc@}
13598 When performing a stack backtrace, code can inspect the value of
13599 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13600 location @code{pc - 12} and the top 8 bits are set, then we know that
13601 there is a function name embedded immediately preceding this location
13602 and has length @code{((pc[-3]) & 0xff000000)}.
13609 Select between generating code that executes in ARM and Thumb
13610 states. The default for most configurations is to generate code
13611 that executes in ARM state, but the default can be changed by
13612 configuring GCC with the @option{--with-mode=}@var{state}
13615 You can also override the ARM and Thumb mode for each function
13616 by using the @code{target("thumb")} and @code{target("arm")} function attributes
13617 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13620 @opindex mtpcs-frame
13621 Generate a stack frame that is compliant with the Thumb Procedure Call
13622 Standard for all non-leaf functions. (A leaf function is one that does
13623 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13625 @item -mtpcs-leaf-frame
13626 @opindex mtpcs-leaf-frame
13627 Generate a stack frame that is compliant with the Thumb Procedure Call
13628 Standard for all leaf functions. (A leaf function is one that does
13629 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13631 @item -mcallee-super-interworking
13632 @opindex mcallee-super-interworking
13633 Gives all externally visible functions in the file being compiled an ARM
13634 instruction set header which switches to Thumb mode before executing the
13635 rest of the function. This allows these functions to be called from
13636 non-interworking code. This option is not valid in AAPCS configurations
13637 because interworking is enabled by default.
13639 @item -mcaller-super-interworking
13640 @opindex mcaller-super-interworking
13641 Allows calls via function pointers (including virtual functions) to
13642 execute correctly regardless of whether the target code has been
13643 compiled for interworking or not. There is a small overhead in the cost
13644 of executing a function pointer if this option is enabled. This option
13645 is not valid in AAPCS configurations because interworking is enabled
13648 @item -mtp=@var{name}
13650 Specify the access model for the thread local storage pointer. The valid
13651 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13652 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13653 (supported in the arm6k architecture), and @samp{auto}, which uses the
13654 best available method for the selected processor. The default setting is
13657 @item -mtls-dialect=@var{dialect}
13658 @opindex mtls-dialect
13659 Specify the dialect to use for accessing thread local storage. Two
13660 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13661 @samp{gnu} dialect selects the original GNU scheme for supporting
13662 local and global dynamic TLS models. The @samp{gnu2} dialect
13663 selects the GNU descriptor scheme, which provides better performance
13664 for shared libraries. The GNU descriptor scheme is compatible with
13665 the original scheme, but does require new assembler, linker and
13666 library support. Initial and local exec TLS models are unaffected by
13667 this option and always use the original scheme.
13669 @item -mword-relocations
13670 @opindex mword-relocations
13671 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13672 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13673 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13676 @item -mfix-cortex-m3-ldrd
13677 @opindex mfix-cortex-m3-ldrd
13678 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13679 with overlapping destination and base registers are used. This option avoids
13680 generating these instructions. This option is enabled by default when
13681 @option{-mcpu=cortex-m3} is specified.
13683 @item -munaligned-access
13684 @itemx -mno-unaligned-access
13685 @opindex munaligned-access
13686 @opindex mno-unaligned-access
13687 Enables (or disables) reading and writing of 16- and 32- bit values
13688 from addresses that are not 16- or 32- bit aligned. By default
13689 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13690 architectures, and enabled for all other architectures. If unaligned
13691 access is not enabled then words in packed data structures are
13692 accessed a byte at a time.
13694 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13695 generated object file to either true or false, depending upon the
13696 setting of this option. If unaligned access is enabled then the
13697 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13700 @item -mneon-for-64bits
13701 @opindex mneon-for-64bits
13702 Enables using Neon to handle scalar 64-bits operations. This is
13703 disabled by default since the cost of moving data from core registers
13706 @item -mslow-flash-data
13707 @opindex mslow-flash-data
13708 Assume loading data from flash is slower than fetching instruction.
13709 Therefore literal load is minimized for better performance.
13710 This option is only supported when compiling for ARMv7 M-profile and
13713 @item -masm-syntax-unified
13714 @opindex masm-syntax-unified
13715 Assume inline assembler is using unified asm syntax. The default is
13716 currently off which implies divided syntax. Currently this option is
13717 available only for Thumb1 and has no effect on ARM state and Thumb2.
13718 However, this may change in future releases of GCC. Divided syntax
13719 should be considered deprecated.
13721 @item -mrestrict-it
13722 @opindex mrestrict-it
13723 Restricts generation of IT blocks to conform to the rules of ARMv8.
13724 IT blocks can only contain a single 16-bit instruction from a select
13725 set of instructions. This option is on by default for ARMv8 Thumb mode.
13727 @item -mprint-tune-info
13728 @opindex mprint-tune-info
13729 Print CPU tuning information as comment in assembler file. This is
13730 an option used only for regression testing of the compiler and not
13731 intended for ordinary use in compiling code. This option is disabled
13736 @subsection AVR Options
13737 @cindex AVR Options
13739 These options are defined for AVR implementations:
13742 @item -mmcu=@var{mcu}
13744 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13746 The default for this option is@tie{}@samp{avr2}.
13748 GCC supports the following AVR devices and ISAs:
13750 @include avr-mmcu.texi
13752 @item -maccumulate-args
13753 @opindex maccumulate-args
13754 Accumulate outgoing function arguments and acquire/release the needed
13755 stack space for outgoing function arguments once in function
13756 prologue/epilogue. Without this option, outgoing arguments are pushed
13757 before calling a function and popped afterwards.
13759 Popping the arguments after the function call can be expensive on
13760 AVR so that accumulating the stack space might lead to smaller
13761 executables because arguments need not to be removed from the
13762 stack after such a function call.
13764 This option can lead to reduced code size for functions that perform
13765 several calls to functions that get their arguments on the stack like
13766 calls to printf-like functions.
13768 @item -mbranch-cost=@var{cost}
13769 @opindex mbranch-cost
13770 Set the branch costs for conditional branch instructions to
13771 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13772 integers. The default branch cost is 0.
13774 @item -mcall-prologues
13775 @opindex mcall-prologues
13776 Functions prologues/epilogues are expanded as calls to appropriate
13777 subroutines. Code size is smaller.
13781 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13782 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13783 and @code{long long} is 4 bytes. Please note that this option does not
13784 conform to the C standards, but it results in smaller code
13787 @item -mn-flash=@var{num}
13789 Assume that the flash memory has a size of
13790 @var{num} times 64@tie{}KiB.
13792 @item -mno-interrupts
13793 @opindex mno-interrupts
13794 Generated code is not compatible with hardware interrupts.
13795 Code size is smaller.
13799 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13800 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13801 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13802 the assembler's command line and the @option{--relax} option to the
13803 linker's command line.
13805 Jump relaxing is performed by the linker because jump offsets are not
13806 known before code is located. Therefore, the assembler code generated by the
13807 compiler is the same, but the instructions in the executable may
13808 differ from instructions in the assembler code.
13810 Relaxing must be turned on if linker stubs are needed, see the
13811 section on @code{EIND} and linker stubs below.
13815 Assume that the device supports the Read-Modify-Write
13816 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13820 Treat the stack pointer register as an 8-bit register,
13821 i.e.@: assume the high byte of the stack pointer is zero.
13822 In general, you don't need to set this option by hand.
13824 This option is used internally by the compiler to select and
13825 build multilibs for architectures @code{avr2} and @code{avr25}.
13826 These architectures mix devices with and without @code{SPH}.
13827 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13828 the compiler driver adds or removes this option from the compiler
13829 proper's command line, because the compiler then knows if the device
13830 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13835 Use address register @code{X} in a way proposed by the hardware. This means
13836 that @code{X} is only used in indirect, post-increment or
13837 pre-decrement addressing.
13839 Without this option, the @code{X} register may be used in the same way
13840 as @code{Y} or @code{Z} which then is emulated by additional
13842 For example, loading a value with @code{X+const} addressing with a
13843 small non-negative @code{const < 64} to a register @var{Rn} is
13847 adiw r26, const ; X += const
13848 ld @var{Rn}, X ; @var{Rn} = *X
13849 sbiw r26, const ; X -= const
13853 @opindex mtiny-stack
13854 Only change the lower 8@tie{}bits of the stack pointer.
13857 @opindex nodevicelib
13858 Don't link against AVR-LibC's device specific library @code{libdev.a}.
13860 @item -Waddr-space-convert
13861 @opindex Waddr-space-convert
13862 Warn about conversions between address spaces in the case where the
13863 resulting address space is not contained in the incoming address space.
13866 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13867 @cindex @code{EIND}
13868 Pointers in the implementation are 16@tie{}bits wide.
13869 The address of a function or label is represented as word address so
13870 that indirect jumps and calls can target any code address in the
13871 range of 64@tie{}Ki words.
13873 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13874 bytes of program memory space, there is a special function register called
13875 @code{EIND} that serves as most significant part of the target address
13876 when @code{EICALL} or @code{EIJMP} instructions are used.
13878 Indirect jumps and calls on these devices are handled as follows by
13879 the compiler and are subject to some limitations:
13884 The compiler never sets @code{EIND}.
13887 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13888 instructions or might read @code{EIND} directly in order to emulate an
13889 indirect call/jump by means of a @code{RET} instruction.
13892 The compiler assumes that @code{EIND} never changes during the startup
13893 code or during the application. In particular, @code{EIND} is not
13894 saved/restored in function or interrupt service routine
13898 For indirect calls to functions and computed goto, the linker
13899 generates @emph{stubs}. Stubs are jump pads sometimes also called
13900 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13901 The stub contains a direct jump to the desired address.
13904 Linker relaxation must be turned on so that the linker generates
13905 the stubs correctly in all situations. See the compiler option
13906 @option{-mrelax} and the linker option @option{--relax}.
13907 There are corner cases where the linker is supposed to generate stubs
13908 but aborts without relaxation and without a helpful error message.
13911 The default linker script is arranged for code with @code{EIND = 0}.
13912 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13913 linker script has to be used in order to place the sections whose
13914 name start with @code{.trampolines} into the segment where @code{EIND}
13918 The startup code from libgcc never sets @code{EIND}.
13919 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13920 For the impact of AVR-LibC on @code{EIND}, see the
13921 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13924 It is legitimate for user-specific startup code to set up @code{EIND}
13925 early, for example by means of initialization code located in
13926 section @code{.init3}. Such code runs prior to general startup code
13927 that initializes RAM and calls constructors, but after the bit
13928 of startup code from AVR-LibC that sets @code{EIND} to the segment
13929 where the vector table is located.
13931 #include <avr/io.h>
13934 __attribute__((section(".init3"),naked,used,no_instrument_function))
13935 init3_set_eind (void)
13937 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13938 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13943 The @code{__trampolines_start} symbol is defined in the linker script.
13946 Stubs are generated automatically by the linker if
13947 the following two conditions are met:
13950 @item The address of a label is taken by means of the @code{gs} modifier
13951 (short for @emph{generate stubs}) like so:
13953 LDI r24, lo8(gs(@var{func}))
13954 LDI r25, hi8(gs(@var{func}))
13956 @item The final location of that label is in a code segment
13957 @emph{outside} the segment where the stubs are located.
13961 The compiler emits such @code{gs} modifiers for code labels in the
13962 following situations:
13964 @item Taking address of a function or code label.
13965 @item Computed goto.
13966 @item If prologue-save function is used, see @option{-mcall-prologues}
13967 command-line option.
13968 @item Switch/case dispatch tables. If you do not want such dispatch
13969 tables you can specify the @option{-fno-jump-tables} command-line option.
13970 @item C and C++ constructors/destructors called during startup/shutdown.
13971 @item If the tools hit a @code{gs()} modifier explained above.
13975 Jumping to non-symbolic addresses like so is @emph{not} supported:
13980 /* Call function at word address 0x2 */
13981 return ((int(*)(void)) 0x2)();
13985 Instead, a stub has to be set up, i.e.@: the function has to be called
13986 through a symbol (@code{func_4} in the example):
13991 extern int func_4 (void);
13993 /* Call function at byte address 0x4 */
13998 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13999 Alternatively, @code{func_4} can be defined in the linker script.
14002 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14003 @cindex @code{RAMPD}
14004 @cindex @code{RAMPX}
14005 @cindex @code{RAMPY}
14006 @cindex @code{RAMPZ}
14007 Some AVR devices support memories larger than the 64@tie{}KiB range
14008 that can be accessed with 16-bit pointers. To access memory locations
14009 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14010 register is used as high part of the address:
14011 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14012 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14013 register, respectively, to get a wide address. Similarly,
14014 @code{RAMPD} is used together with direct addressing.
14018 The startup code initializes the @code{RAMP} special function
14019 registers with zero.
14022 If a @ref{AVR Named Address Spaces,named address space} other than
14023 generic or @code{__flash} is used, then @code{RAMPZ} is set
14024 as needed before the operation.
14027 If the device supports RAM larger than 64@tie{}KiB and the compiler
14028 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14029 is reset to zero after the operation.
14032 If the device comes with a specific @code{RAMP} register, the ISR
14033 prologue/epilogue saves/restores that SFR and initializes it with
14034 zero in case the ISR code might (implicitly) use it.
14037 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14038 If you use inline assembler to read from locations outside the
14039 16-bit address range and change one of the @code{RAMP} registers,
14040 you must reset it to zero after the access.
14044 @subsubsection AVR Built-in Macros
14046 GCC defines several built-in macros so that the user code can test
14047 for the presence or absence of features. Almost any of the following
14048 built-in macros are deduced from device capabilities and thus
14049 triggered by the @option{-mmcu=} command-line option.
14051 For even more AVR-specific built-in macros see
14052 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14057 Build-in macro that resolves to a decimal number that identifies the
14058 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14059 Possible values are:
14061 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14062 @code{4}, @code{5}, @code{51}, @code{6}
14064 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14065 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14069 @code{100}, @code{102}, @code{104},
14070 @code{105}, @code{106}, @code{107}
14072 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14073 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14074 If @var{mcu} specifies a device, this built-in macro is set
14075 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14076 defined to @code{4}.
14078 @item __AVR_@var{Device}__
14079 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14080 the device's name. For example, @option{-mmcu=atmega8} defines the
14081 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14082 @code{__AVR_ATtiny261A__}, etc.
14084 The built-in macros' names follow
14085 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14086 the device name as from the AVR user manual. The difference between
14087 @var{Device} in the built-in macro and @var{device} in
14088 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14090 If @var{device} is not a device but only a core architecture like
14091 @samp{avr51}, this macro is not defined.
14093 @item __AVR_DEVICE_NAME__
14094 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14095 the device's name. For example, with @option{-mmcu=atmega8} the macro
14096 is defined to @code{atmega8}.
14098 If @var{device} is not a device but only a core architecture like
14099 @samp{avr51}, this macro is not defined.
14101 @item __AVR_XMEGA__
14102 The device / architecture belongs to the XMEGA family of devices.
14104 @item __AVR_HAVE_ELPM__
14105 The device has the @code{ELPM} instruction.
14107 @item __AVR_HAVE_ELPMX__
14108 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14109 R@var{n},Z+} instructions.
14111 @item __AVR_HAVE_MOVW__
14112 The device has the @code{MOVW} instruction to perform 16-bit
14113 register-register moves.
14115 @item __AVR_HAVE_LPMX__
14116 The device has the @code{LPM R@var{n},Z} and
14117 @code{LPM R@var{n},Z+} instructions.
14119 @item __AVR_HAVE_MUL__
14120 The device has a hardware multiplier.
14122 @item __AVR_HAVE_JMP_CALL__
14123 The device has the @code{JMP} and @code{CALL} instructions.
14124 This is the case for devices with at least 16@tie{}KiB of program
14127 @item __AVR_HAVE_EIJMP_EICALL__
14128 @itemx __AVR_3_BYTE_PC__
14129 The device has the @code{EIJMP} and @code{EICALL} instructions.
14130 This is the case for devices with more than 128@tie{}KiB of program memory.
14131 This also means that the program counter
14132 (PC) is 3@tie{}bytes wide.
14134 @item __AVR_2_BYTE_PC__
14135 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14136 with up to 128@tie{}KiB of program memory.
14138 @item __AVR_HAVE_8BIT_SP__
14139 @itemx __AVR_HAVE_16BIT_SP__
14140 The stack pointer (SP) register is treated as 8-bit respectively
14141 16-bit register by the compiler.
14142 The definition of these macros is affected by @option{-mtiny-stack}.
14144 @item __AVR_HAVE_SPH__
14146 The device has the SPH (high part of stack pointer) special function
14147 register or has an 8-bit stack pointer, respectively.
14148 The definition of these macros is affected by @option{-mmcu=} and
14149 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14152 @item __AVR_HAVE_RAMPD__
14153 @itemx __AVR_HAVE_RAMPX__
14154 @itemx __AVR_HAVE_RAMPY__
14155 @itemx __AVR_HAVE_RAMPZ__
14156 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14157 @code{RAMPZ} special function register, respectively.
14159 @item __NO_INTERRUPTS__
14160 This macro reflects the @option{-mno-interrupts} command-line option.
14162 @item __AVR_ERRATA_SKIP__
14163 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14164 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14165 instructions because of a hardware erratum. Skip instructions are
14166 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14167 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14170 @item __AVR_ISA_RMW__
14171 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14173 @item __AVR_SFR_OFFSET__=@var{offset}
14174 Instructions that can address I/O special function registers directly
14175 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14176 address as if addressed by an instruction to access RAM like @code{LD}
14177 or @code{STS}. This offset depends on the device architecture and has
14178 to be subtracted from the RAM address in order to get the
14179 respective I/O@tie{}address.
14181 @item __WITH_AVRLIBC__
14182 The compiler is configured to be used together with AVR-Libc.
14183 See the @option{--with-avrlibc} configure option.
14187 @node Blackfin Options
14188 @subsection Blackfin Options
14189 @cindex Blackfin Options
14192 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14194 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14195 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14196 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14197 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14198 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14199 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14200 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14201 @samp{bf561}, @samp{bf592}.
14203 The optional @var{sirevision} specifies the silicon revision of the target
14204 Blackfin processor. Any workarounds available for the targeted silicon revision
14205 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14206 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14207 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14208 hexadecimal digits representing the major and minor numbers in the silicon
14209 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14210 is not defined. If @var{sirevision} is @samp{any}, the
14211 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14212 If this optional @var{sirevision} is not used, GCC assumes the latest known
14213 silicon revision of the targeted Blackfin processor.
14215 GCC defines a preprocessor macro for the specified @var{cpu}.
14216 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14217 provided by libgloss to be linked in if @option{-msim} is not given.
14219 Without this option, @samp{bf532} is used as the processor by default.
14221 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14222 only the preprocessor macro is defined.
14226 Specifies that the program will be run on the simulator. This causes
14227 the simulator BSP provided by libgloss to be linked in. This option
14228 has effect only for @samp{bfin-elf} toolchain.
14229 Certain other options, such as @option{-mid-shared-library} and
14230 @option{-mfdpic}, imply @option{-msim}.
14232 @item -momit-leaf-frame-pointer
14233 @opindex momit-leaf-frame-pointer
14234 Don't keep the frame pointer in a register for leaf functions. This
14235 avoids the instructions to save, set up and restore frame pointers and
14236 makes an extra register available in leaf functions. The option
14237 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14238 which might make debugging harder.
14240 @item -mspecld-anomaly
14241 @opindex mspecld-anomaly
14242 When enabled, the compiler ensures that the generated code does not
14243 contain speculative loads after jump instructions. If this option is used,
14244 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14246 @item -mno-specld-anomaly
14247 @opindex mno-specld-anomaly
14248 Don't generate extra code to prevent speculative loads from occurring.
14250 @item -mcsync-anomaly
14251 @opindex mcsync-anomaly
14252 When enabled, the compiler ensures that the generated code does not
14253 contain CSYNC or SSYNC instructions too soon after conditional branches.
14254 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14256 @item -mno-csync-anomaly
14257 @opindex mno-csync-anomaly
14258 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14259 occurring too soon after a conditional branch.
14263 When enabled, the compiler is free to take advantage of the knowledge that
14264 the entire program fits into the low 64k of memory.
14267 @opindex mno-low-64k
14268 Assume that the program is arbitrarily large. This is the default.
14270 @item -mstack-check-l1
14271 @opindex mstack-check-l1
14272 Do stack checking using information placed into L1 scratchpad memory by the
14275 @item -mid-shared-library
14276 @opindex mid-shared-library
14277 Generate code that supports shared libraries via the library ID method.
14278 This allows for execute in place and shared libraries in an environment
14279 without virtual memory management. This option implies @option{-fPIC}.
14280 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14282 @item -mno-id-shared-library
14283 @opindex mno-id-shared-library
14284 Generate code that doesn't assume ID-based shared libraries are being used.
14285 This is the default.
14287 @item -mleaf-id-shared-library
14288 @opindex mleaf-id-shared-library
14289 Generate code that supports shared libraries via the library ID method,
14290 but assumes that this library or executable won't link against any other
14291 ID shared libraries. That allows the compiler to use faster code for jumps
14294 @item -mno-leaf-id-shared-library
14295 @opindex mno-leaf-id-shared-library
14296 Do not assume that the code being compiled won't link against any ID shared
14297 libraries. Slower code is generated for jump and call insns.
14299 @item -mshared-library-id=n
14300 @opindex mshared-library-id
14301 Specifies the identification number of the ID-based shared library being
14302 compiled. Specifying a value of 0 generates more compact code; specifying
14303 other values forces the allocation of that number to the current
14304 library but is no more space- or time-efficient than omitting this option.
14308 Generate code that allows the data segment to be located in a different
14309 area of memory from the text segment. This allows for execute in place in
14310 an environment without virtual memory management by eliminating relocations
14311 against the text section.
14313 @item -mno-sep-data
14314 @opindex mno-sep-data
14315 Generate code that assumes that the data segment follows the text segment.
14316 This is the default.
14319 @itemx -mno-long-calls
14320 @opindex mlong-calls
14321 @opindex mno-long-calls
14322 Tells the compiler to perform function calls by first loading the
14323 address of the function into a register and then performing a subroutine
14324 call on this register. This switch is needed if the target function
14325 lies outside of the 24-bit addressing range of the offset-based
14326 version of subroutine call instruction.
14328 This feature is not enabled by default. Specifying
14329 @option{-mno-long-calls} restores the default behavior. Note these
14330 switches have no effect on how the compiler generates code to handle
14331 function calls via function pointers.
14335 Link with the fast floating-point library. This library relaxes some of
14336 the IEEE floating-point standard's rules for checking inputs against
14337 Not-a-Number (NAN), in the interest of performance.
14340 @opindex minline-plt
14341 Enable inlining of PLT entries in function calls to functions that are
14342 not known to bind locally. It has no effect without @option{-mfdpic}.
14345 @opindex mmulticore
14346 Build a standalone application for multicore Blackfin processors.
14347 This option causes proper start files and link scripts supporting
14348 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14349 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14351 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14352 selects the one-application-per-core programming model. Without
14353 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14354 programming model is used. In this model, the main function of Core B
14355 should be named as @code{coreb_main}.
14357 If this option is not used, the single-core application programming
14362 Build a standalone application for Core A of BF561 when using
14363 the one-application-per-core programming model. Proper start files
14364 and link scripts are used to support Core A, and the macro
14365 @code{__BFIN_COREA} is defined.
14366 This option can only be used in conjunction with @option{-mmulticore}.
14370 Build a standalone application for Core B of BF561 when using
14371 the one-application-per-core programming model. Proper start files
14372 and link scripts are used to support Core B, and the macro
14373 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14374 should be used instead of @code{main}.
14375 This option can only be used in conjunction with @option{-mmulticore}.
14379 Build a standalone application for SDRAM. Proper start files and
14380 link scripts are used to put the application into SDRAM, and the macro
14381 @code{__BFIN_SDRAM} is defined.
14382 The loader should initialize SDRAM before loading the application.
14386 Assume that ICPLBs are enabled at run time. This has an effect on certain
14387 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14388 are enabled; for standalone applications the default is off.
14392 @subsection C6X Options
14393 @cindex C6X Options
14396 @item -march=@var{name}
14398 This specifies the name of the target architecture. GCC uses this
14399 name to determine what kind of instructions it can emit when generating
14400 assembly code. Permissible names are: @samp{c62x},
14401 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14404 @opindex mbig-endian
14405 Generate code for a big-endian target.
14407 @item -mlittle-endian
14408 @opindex mlittle-endian
14409 Generate code for a little-endian target. This is the default.
14413 Choose startup files and linker script suitable for the simulator.
14415 @item -msdata=default
14416 @opindex msdata=default
14417 Put small global and static data in the @code{.neardata} section,
14418 which is pointed to by register @code{B14}. Put small uninitialized
14419 global and static data in the @code{.bss} section, which is adjacent
14420 to the @code{.neardata} section. Put small read-only data into the
14421 @code{.rodata} section. The corresponding sections used for large
14422 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14425 @opindex msdata=all
14426 Put all data, not just small objects, into the sections reserved for
14427 small data, and use addressing relative to the @code{B14} register to
14431 @opindex msdata=none
14432 Make no use of the sections reserved for small data, and use absolute
14433 addresses to access all data. Put all initialized global and static
14434 data in the @code{.fardata} section, and all uninitialized data in the
14435 @code{.far} section. Put all constant data into the @code{.const}
14440 @subsection CRIS Options
14441 @cindex CRIS Options
14443 These options are defined specifically for the CRIS ports.
14446 @item -march=@var{architecture-type}
14447 @itemx -mcpu=@var{architecture-type}
14450 Generate code for the specified architecture. The choices for
14451 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14452 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14453 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14456 @item -mtune=@var{architecture-type}
14458 Tune to @var{architecture-type} everything applicable about the generated
14459 code, except for the ABI and the set of available instructions. The
14460 choices for @var{architecture-type} are the same as for
14461 @option{-march=@var{architecture-type}}.
14463 @item -mmax-stack-frame=@var{n}
14464 @opindex mmax-stack-frame
14465 Warn when the stack frame of a function exceeds @var{n} bytes.
14471 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14472 @option{-march=v3} and @option{-march=v8} respectively.
14474 @item -mmul-bug-workaround
14475 @itemx -mno-mul-bug-workaround
14476 @opindex mmul-bug-workaround
14477 @opindex mno-mul-bug-workaround
14478 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14479 models where it applies. This option is active by default.
14483 Enable CRIS-specific verbose debug-related information in the assembly
14484 code. This option also has the effect of turning off the @samp{#NO_APP}
14485 formatted-code indicator to the assembler at the beginning of the
14490 Do not use condition-code results from previous instruction; always emit
14491 compare and test instructions before use of condition codes.
14493 @item -mno-side-effects
14494 @opindex mno-side-effects
14495 Do not emit instructions with side effects in addressing modes other than
14498 @item -mstack-align
14499 @itemx -mno-stack-align
14500 @itemx -mdata-align
14501 @itemx -mno-data-align
14502 @itemx -mconst-align
14503 @itemx -mno-const-align
14504 @opindex mstack-align
14505 @opindex mno-stack-align
14506 @opindex mdata-align
14507 @opindex mno-data-align
14508 @opindex mconst-align
14509 @opindex mno-const-align
14510 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14511 stack frame, individual data and constants to be aligned for the maximum
14512 single data access size for the chosen CPU model. The default is to
14513 arrange for 32-bit alignment. ABI details such as structure layout are
14514 not affected by these options.
14522 Similar to the stack- data- and const-align options above, these options
14523 arrange for stack frame, writable data and constants to all be 32-bit,
14524 16-bit or 8-bit aligned. The default is 32-bit alignment.
14526 @item -mno-prologue-epilogue
14527 @itemx -mprologue-epilogue
14528 @opindex mno-prologue-epilogue
14529 @opindex mprologue-epilogue
14530 With @option{-mno-prologue-epilogue}, the normal function prologue and
14531 epilogue which set up the stack frame are omitted and no return
14532 instructions or return sequences are generated in the code. Use this
14533 option only together with visual inspection of the compiled code: no
14534 warnings or errors are generated when call-saved registers must be saved,
14535 or storage for local variables needs to be allocated.
14539 @opindex mno-gotplt
14541 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14542 instruction sequences that load addresses for functions from the PLT part
14543 of the GOT rather than (traditional on other architectures) calls to the
14544 PLT@. The default is @option{-mgotplt}.
14548 Legacy no-op option only recognized with the cris-axis-elf and
14549 cris-axis-linux-gnu targets.
14553 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14557 This option, recognized for the cris-axis-elf, arranges
14558 to link with input-output functions from a simulator library. Code,
14559 initialized data and zero-initialized data are allocated consecutively.
14563 Like @option{-sim}, but pass linker options to locate initialized data at
14564 0x40000000 and zero-initialized data at 0x80000000.
14568 @subsection CR16 Options
14569 @cindex CR16 Options
14571 These options are defined specifically for the CR16 ports.
14577 Enable the use of multiply-accumulate instructions. Disabled by default.
14581 @opindex mcr16cplus
14583 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14588 Links the library libsim.a which is in compatible with simulator. Applicable
14589 to ELF compiler only.
14593 Choose integer type as 32-bit wide.
14597 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14599 @item -mdata-model=@var{model}
14600 @opindex mdata-model
14601 Choose a data model. The choices for @var{model} are @samp{near},
14602 @samp{far} or @samp{medium}. @samp{medium} is default.
14603 However, @samp{far} is not valid with @option{-mcr16c}, as the
14604 CR16C architecture does not support the far data model.
14607 @node Darwin Options
14608 @subsection Darwin Options
14609 @cindex Darwin options
14611 These options are defined for all architectures running the Darwin operating
14614 FSF GCC on Darwin does not create ``fat'' object files; it creates
14615 an object file for the single architecture that GCC was built to
14616 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14617 @option{-arch} options are used; it does so by running the compiler or
14618 linker multiple times and joining the results together with
14621 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14622 @samp{i686}) is determined by the flags that specify the ISA
14623 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14624 @option{-force_cpusubtype_ALL} option can be used to override this.
14626 The Darwin tools vary in their behavior when presented with an ISA
14627 mismatch. The assembler, @file{as}, only permits instructions to
14628 be used that are valid for the subtype of the file it is generating,
14629 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14630 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14631 and prints an error if asked to create a shared library with a less
14632 restrictive subtype than its input files (for instance, trying to put
14633 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14634 for executables, @command{ld}, quietly gives the executable the most
14635 restrictive subtype of any of its input files.
14640 Add the framework directory @var{dir} to the head of the list of
14641 directories to be searched for header files. These directories are
14642 interleaved with those specified by @option{-I} options and are
14643 scanned in a left-to-right order.
14645 A framework directory is a directory with frameworks in it. A
14646 framework is a directory with a @file{Headers} and/or
14647 @file{PrivateHeaders} directory contained directly in it that ends
14648 in @file{.framework}. The name of a framework is the name of this
14649 directory excluding the @file{.framework}. Headers associated with
14650 the framework are found in one of those two directories, with
14651 @file{Headers} being searched first. A subframework is a framework
14652 directory that is in a framework's @file{Frameworks} directory.
14653 Includes of subframework headers can only appear in a header of a
14654 framework that contains the subframework, or in a sibling subframework
14655 header. Two subframeworks are siblings if they occur in the same
14656 framework. A subframework should not have the same name as a
14657 framework; a warning is issued if this is violated. Currently a
14658 subframework cannot have subframeworks; in the future, the mechanism
14659 may be extended to support this. The standard frameworks can be found
14660 in @file{/System/Library/Frameworks} and
14661 @file{/Library/Frameworks}. An example include looks like
14662 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14663 the name of the framework and @file{header.h} is found in the
14664 @file{PrivateHeaders} or @file{Headers} directory.
14666 @item -iframework@var{dir}
14667 @opindex iframework
14668 Like @option{-F} except the directory is a treated as a system
14669 directory. The main difference between this @option{-iframework} and
14670 @option{-F} is that with @option{-iframework} the compiler does not
14671 warn about constructs contained within header files found via
14672 @var{dir}. This option is valid only for the C family of languages.
14676 Emit debugging information for symbols that are used. For stabs
14677 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14678 This is by default ON@.
14682 Emit debugging information for all symbols and types.
14684 @item -mmacosx-version-min=@var{version}
14685 The earliest version of MacOS X that this executable will run on
14686 is @var{version}. Typical values of @var{version} include @code{10.1},
14687 @code{10.2}, and @code{10.3.9}.
14689 If the compiler was built to use the system's headers by default,
14690 then the default for this option is the system version on which the
14691 compiler is running, otherwise the default is to make choices that
14692 are compatible with as many systems and code bases as possible.
14696 Enable kernel development mode. The @option{-mkernel} option sets
14697 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14698 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14699 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14700 applicable. This mode also sets @option{-mno-altivec},
14701 @option{-msoft-float}, @option{-fno-builtin} and
14702 @option{-mlong-branch} for PowerPC targets.
14704 @item -mone-byte-bool
14705 @opindex mone-byte-bool
14706 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14707 By default @code{sizeof(bool)} is @code{4} when compiling for
14708 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14709 option has no effect on x86.
14711 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14712 to generate code that is not binary compatible with code generated
14713 without that switch. Using this switch may require recompiling all
14714 other modules in a program, including system libraries. Use this
14715 switch to conform to a non-default data model.
14717 @item -mfix-and-continue
14718 @itemx -ffix-and-continue
14719 @itemx -findirect-data
14720 @opindex mfix-and-continue
14721 @opindex ffix-and-continue
14722 @opindex findirect-data
14723 Generate code suitable for fast turnaround development, such as to
14724 allow GDB to dynamically load @file{.o} files into already-running
14725 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14726 are provided for backwards compatibility.
14730 Loads all members of static archive libraries.
14731 See man ld(1) for more information.
14733 @item -arch_errors_fatal
14734 @opindex arch_errors_fatal
14735 Cause the errors having to do with files that have the wrong architecture
14738 @item -bind_at_load
14739 @opindex bind_at_load
14740 Causes the output file to be marked such that the dynamic linker will
14741 bind all undefined references when the file is loaded or launched.
14745 Produce a Mach-o bundle format file.
14746 See man ld(1) for more information.
14748 @item -bundle_loader @var{executable}
14749 @opindex bundle_loader
14750 This option specifies the @var{executable} that will load the build
14751 output file being linked. See man ld(1) for more information.
14754 @opindex dynamiclib
14755 When passed this option, GCC produces a dynamic library instead of
14756 an executable when linking, using the Darwin @file{libtool} command.
14758 @item -force_cpusubtype_ALL
14759 @opindex force_cpusubtype_ALL
14760 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14761 one controlled by the @option{-mcpu} or @option{-march} option.
14763 @item -allowable_client @var{client_name}
14764 @itemx -client_name
14765 @itemx -compatibility_version
14766 @itemx -current_version
14768 @itemx -dependency-file
14770 @itemx -dylinker_install_name
14772 @itemx -exported_symbols_list
14775 @itemx -flat_namespace
14776 @itemx -force_flat_namespace
14777 @itemx -headerpad_max_install_names
14780 @itemx -install_name
14781 @itemx -keep_private_externs
14782 @itemx -multi_module
14783 @itemx -multiply_defined
14784 @itemx -multiply_defined_unused
14787 @itemx -no_dead_strip_inits_and_terms
14788 @itemx -nofixprebinding
14789 @itemx -nomultidefs
14791 @itemx -noseglinkedit
14792 @itemx -pagezero_size
14794 @itemx -prebind_all_twolevel_modules
14795 @itemx -private_bundle
14797 @itemx -read_only_relocs
14799 @itemx -sectobjectsymbols
14803 @itemx -sectobjectsymbols
14806 @itemx -segs_read_only_addr
14808 @itemx -segs_read_write_addr
14809 @itemx -seg_addr_table
14810 @itemx -seg_addr_table_filename
14811 @itemx -seglinkedit
14813 @itemx -segs_read_only_addr
14814 @itemx -segs_read_write_addr
14815 @itemx -single_module
14817 @itemx -sub_library
14819 @itemx -sub_umbrella
14820 @itemx -twolevel_namespace
14823 @itemx -unexported_symbols_list
14824 @itemx -weak_reference_mismatches
14825 @itemx -whatsloaded
14826 @opindex allowable_client
14827 @opindex client_name
14828 @opindex compatibility_version
14829 @opindex current_version
14830 @opindex dead_strip
14831 @opindex dependency-file
14832 @opindex dylib_file
14833 @opindex dylinker_install_name
14835 @opindex exported_symbols_list
14837 @opindex flat_namespace
14838 @opindex force_flat_namespace
14839 @opindex headerpad_max_install_names
14840 @opindex image_base
14842 @opindex install_name
14843 @opindex keep_private_externs
14844 @opindex multi_module
14845 @opindex multiply_defined
14846 @opindex multiply_defined_unused
14847 @opindex noall_load
14848 @opindex no_dead_strip_inits_and_terms
14849 @opindex nofixprebinding
14850 @opindex nomultidefs
14852 @opindex noseglinkedit
14853 @opindex pagezero_size
14855 @opindex prebind_all_twolevel_modules
14856 @opindex private_bundle
14857 @opindex read_only_relocs
14859 @opindex sectobjectsymbols
14862 @opindex sectcreate
14863 @opindex sectobjectsymbols
14866 @opindex segs_read_only_addr
14867 @opindex segs_read_write_addr
14868 @opindex seg_addr_table
14869 @opindex seg_addr_table_filename
14870 @opindex seglinkedit
14872 @opindex segs_read_only_addr
14873 @opindex segs_read_write_addr
14874 @opindex single_module
14876 @opindex sub_library
14877 @opindex sub_umbrella
14878 @opindex twolevel_namespace
14881 @opindex unexported_symbols_list
14882 @opindex weak_reference_mismatches
14883 @opindex whatsloaded
14884 These options are passed to the Darwin linker. The Darwin linker man page
14885 describes them in detail.
14888 @node DEC Alpha Options
14889 @subsection DEC Alpha Options
14891 These @samp{-m} options are defined for the DEC Alpha implementations:
14894 @item -mno-soft-float
14895 @itemx -msoft-float
14896 @opindex mno-soft-float
14897 @opindex msoft-float
14898 Use (do not use) the hardware floating-point instructions for
14899 floating-point operations. When @option{-msoft-float} is specified,
14900 functions in @file{libgcc.a} are used to perform floating-point
14901 operations. Unless they are replaced by routines that emulate the
14902 floating-point operations, or compiled in such a way as to call such
14903 emulations routines, these routines issue floating-point
14904 operations. If you are compiling for an Alpha without floating-point
14905 operations, you must ensure that the library is built so as not to call
14908 Note that Alpha implementations without floating-point operations are
14909 required to have floating-point registers.
14912 @itemx -mno-fp-regs
14914 @opindex mno-fp-regs
14915 Generate code that uses (does not use) the floating-point register set.
14916 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14917 register set is not used, floating-point operands are passed in integer
14918 registers as if they were integers and floating-point results are passed
14919 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14920 so any function with a floating-point argument or return value called by code
14921 compiled with @option{-mno-fp-regs} must also be compiled with that
14924 A typical use of this option is building a kernel that does not use,
14925 and hence need not save and restore, any floating-point registers.
14929 The Alpha architecture implements floating-point hardware optimized for
14930 maximum performance. It is mostly compliant with the IEEE floating-point
14931 standard. However, for full compliance, software assistance is
14932 required. This option generates code fully IEEE-compliant code
14933 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14934 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14935 defined during compilation. The resulting code is less efficient but is
14936 able to correctly support denormalized numbers and exceptional IEEE
14937 values such as not-a-number and plus/minus infinity. Other Alpha
14938 compilers call this option @option{-ieee_with_no_inexact}.
14940 @item -mieee-with-inexact
14941 @opindex mieee-with-inexact
14942 This is like @option{-mieee} except the generated code also maintains
14943 the IEEE @var{inexact-flag}. Turning on this option causes the
14944 generated code to implement fully-compliant IEEE math. In addition to
14945 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14946 macro. On some Alpha implementations the resulting code may execute
14947 significantly slower than the code generated by default. Since there is
14948 very little code that depends on the @var{inexact-flag}, you should
14949 normally not specify this option. Other Alpha compilers call this
14950 option @option{-ieee_with_inexact}.
14952 @item -mfp-trap-mode=@var{trap-mode}
14953 @opindex mfp-trap-mode
14954 This option controls what floating-point related traps are enabled.
14955 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14956 The trap mode can be set to one of four values:
14960 This is the default (normal) setting. The only traps that are enabled
14961 are the ones that cannot be disabled in software (e.g., division by zero
14965 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14969 Like @samp{u}, but the instructions are marked to be safe for software
14970 completion (see Alpha architecture manual for details).
14973 Like @samp{su}, but inexact traps are enabled as well.
14976 @item -mfp-rounding-mode=@var{rounding-mode}
14977 @opindex mfp-rounding-mode
14978 Selects the IEEE rounding mode. Other Alpha compilers call this option
14979 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14984 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14985 the nearest machine number or towards the even machine number in case
14989 Round towards minus infinity.
14992 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14995 Dynamic rounding mode. A field in the floating-point control register
14996 (@var{fpcr}, see Alpha architecture reference manual) controls the
14997 rounding mode in effect. The C library initializes this register for
14998 rounding towards plus infinity. Thus, unless your program modifies the
14999 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15002 @item -mtrap-precision=@var{trap-precision}
15003 @opindex mtrap-precision
15004 In the Alpha architecture, floating-point traps are imprecise. This
15005 means without software assistance it is impossible to recover from a
15006 floating trap and program execution normally needs to be terminated.
15007 GCC can generate code that can assist operating system trap handlers
15008 in determining the exact location that caused a floating-point trap.
15009 Depending on the requirements of an application, different levels of
15010 precisions can be selected:
15014 Program precision. This option is the default and means a trap handler
15015 can only identify which program caused a floating-point exception.
15018 Function precision. The trap handler can determine the function that
15019 caused a floating-point exception.
15022 Instruction precision. The trap handler can determine the exact
15023 instruction that caused a floating-point exception.
15026 Other Alpha compilers provide the equivalent options called
15027 @option{-scope_safe} and @option{-resumption_safe}.
15029 @item -mieee-conformant
15030 @opindex mieee-conformant
15031 This option marks the generated code as IEEE conformant. You must not
15032 use this option unless you also specify @option{-mtrap-precision=i} and either
15033 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15034 is to emit the line @samp{.eflag 48} in the function prologue of the
15035 generated assembly file.
15037 @item -mbuild-constants
15038 @opindex mbuild-constants
15039 Normally GCC examines a 32- or 64-bit integer constant to
15040 see if it can construct it from smaller constants in two or three
15041 instructions. If it cannot, it outputs the constant as a literal and
15042 generates code to load it from the data segment at run time.
15044 Use this option to require GCC to construct @emph{all} integer constants
15045 using code, even if it takes more instructions (the maximum is six).
15047 You typically use this option to build a shared library dynamic
15048 loader. Itself a shared library, it must relocate itself in memory
15049 before it can find the variables and constants in its own data segment.
15067 Indicate whether GCC should generate code to use the optional BWX,
15068 CIX, FIX and MAX instruction sets. The default is to use the instruction
15069 sets supported by the CPU type specified via @option{-mcpu=} option or that
15070 of the CPU on which GCC was built if none is specified.
15073 @itemx -mfloat-ieee
15074 @opindex mfloat-vax
15075 @opindex mfloat-ieee
15076 Generate code that uses (does not use) VAX F and G floating-point
15077 arithmetic instead of IEEE single and double precision.
15079 @item -mexplicit-relocs
15080 @itemx -mno-explicit-relocs
15081 @opindex mexplicit-relocs
15082 @opindex mno-explicit-relocs
15083 Older Alpha assemblers provided no way to generate symbol relocations
15084 except via assembler macros. Use of these macros does not allow
15085 optimal instruction scheduling. GNU binutils as of version 2.12
15086 supports a new syntax that allows the compiler to explicitly mark
15087 which relocations should apply to which instructions. This option
15088 is mostly useful for debugging, as GCC detects the capabilities of
15089 the assembler when it is built and sets the default accordingly.
15092 @itemx -mlarge-data
15093 @opindex msmall-data
15094 @opindex mlarge-data
15095 When @option{-mexplicit-relocs} is in effect, static data is
15096 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15097 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15098 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15099 16-bit relocations off of the @code{$gp} register. This limits the
15100 size of the small data area to 64KB, but allows the variables to be
15101 directly accessed via a single instruction.
15103 The default is @option{-mlarge-data}. With this option the data area
15104 is limited to just below 2GB@. Programs that require more than 2GB of
15105 data must use @code{malloc} or @code{mmap} to allocate the data in the
15106 heap instead of in the program's data segment.
15108 When generating code for shared libraries, @option{-fpic} implies
15109 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15112 @itemx -mlarge-text
15113 @opindex msmall-text
15114 @opindex mlarge-text
15115 When @option{-msmall-text} is used, the compiler assumes that the
15116 code of the entire program (or shared library) fits in 4MB, and is
15117 thus reachable with a branch instruction. When @option{-msmall-data}
15118 is used, the compiler can assume that all local symbols share the
15119 same @code{$gp} value, and thus reduce the number of instructions
15120 required for a function call from 4 to 1.
15122 The default is @option{-mlarge-text}.
15124 @item -mcpu=@var{cpu_type}
15126 Set the instruction set and instruction scheduling parameters for
15127 machine type @var{cpu_type}. You can specify either the @samp{EV}
15128 style name or the corresponding chip number. GCC supports scheduling
15129 parameters for the EV4, EV5 and EV6 family of processors and
15130 chooses the default values for the instruction set from the processor
15131 you specify. If you do not specify a processor type, GCC defaults
15132 to the processor on which the compiler was built.
15134 Supported values for @var{cpu_type} are
15140 Schedules as an EV4 and has no instruction set extensions.
15144 Schedules as an EV5 and has no instruction set extensions.
15148 Schedules as an EV5 and supports the BWX extension.
15153 Schedules as an EV5 and supports the BWX and MAX extensions.
15157 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15161 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15164 Native toolchains also support the value @samp{native},
15165 which selects the best architecture option for the host processor.
15166 @option{-mcpu=native} has no effect if GCC does not recognize
15169 @item -mtune=@var{cpu_type}
15171 Set only the instruction scheduling parameters for machine type
15172 @var{cpu_type}. The instruction set is not changed.
15174 Native toolchains also support the value @samp{native},
15175 which selects the best architecture option for the host processor.
15176 @option{-mtune=native} has no effect if GCC does not recognize
15179 @item -mmemory-latency=@var{time}
15180 @opindex mmemory-latency
15181 Sets the latency the scheduler should assume for typical memory
15182 references as seen by the application. This number is highly
15183 dependent on the memory access patterns used by the application
15184 and the size of the external cache on the machine.
15186 Valid options for @var{time} are
15190 A decimal number representing clock cycles.
15196 The compiler contains estimates of the number of clock cycles for
15197 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15198 (also called Dcache, Scache, and Bcache), as well as to main memory.
15199 Note that L3 is only valid for EV5.
15205 @subsection FR30 Options
15206 @cindex FR30 Options
15208 These options are defined specifically for the FR30 port.
15212 @item -msmall-model
15213 @opindex msmall-model
15214 Use the small address space model. This can produce smaller code, but
15215 it does assume that all symbolic values and addresses fit into a
15220 Assume that runtime support has been provided and so there is no need
15221 to include the simulator library (@file{libsim.a}) on the linker
15227 @subsection FT32 Options
15228 @cindex FT32 Options
15230 These options are defined specifically for the FT32 port.
15236 Specifies that the program will be run on the simulator. This causes
15237 an alternate runtime startup and library to be linked.
15238 You must not use this option when generating programs that will run on
15239 real hardware; you must provide your own runtime library for whatever
15240 I/O functions are needed.
15244 Enable Local Register Allocation. This is still experimental for FT32,
15245 so by default the compiler uses standard reload.
15250 @subsection FRV Options
15251 @cindex FRV Options
15257 Only use the first 32 general-purpose registers.
15262 Use all 64 general-purpose registers.
15267 Use only the first 32 floating-point registers.
15272 Use all 64 floating-point registers.
15275 @opindex mhard-float
15277 Use hardware instructions for floating-point operations.
15280 @opindex msoft-float
15282 Use library routines for floating-point operations.
15287 Dynamically allocate condition code registers.
15292 Do not try to dynamically allocate condition code registers, only
15293 use @code{icc0} and @code{fcc0}.
15298 Change ABI to use double word insns.
15303 Do not use double word instructions.
15308 Use floating-point double instructions.
15311 @opindex mno-double
15313 Do not use floating-point double instructions.
15318 Use media instructions.
15323 Do not use media instructions.
15328 Use multiply and add/subtract instructions.
15331 @opindex mno-muladd
15333 Do not use multiply and add/subtract instructions.
15338 Select the FDPIC ABI, which uses function descriptors to represent
15339 pointers to functions. Without any PIC/PIE-related options, it
15340 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15341 assumes GOT entries and small data are within a 12-bit range from the
15342 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15343 are computed with 32 bits.
15344 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15347 @opindex minline-plt
15349 Enable inlining of PLT entries in function calls to functions that are
15350 not known to bind locally. It has no effect without @option{-mfdpic}.
15351 It's enabled by default if optimizing for speed and compiling for
15352 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15353 optimization option such as @option{-O3} or above is present in the
15359 Assume a large TLS segment when generating thread-local code.
15364 Do not assume a large TLS segment when generating thread-local code.
15369 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15370 that is known to be in read-only sections. It's enabled by default,
15371 except for @option{-fpic} or @option{-fpie}: even though it may help
15372 make the global offset table smaller, it trades 1 instruction for 4.
15373 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15374 one of which may be shared by multiple symbols, and it avoids the need
15375 for a GOT entry for the referenced symbol, so it's more likely to be a
15376 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15378 @item -multilib-library-pic
15379 @opindex multilib-library-pic
15381 Link with the (library, not FD) pic libraries. It's implied by
15382 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15383 @option{-fpic} without @option{-mfdpic}. You should never have to use
15387 @opindex mlinked-fp
15389 Follow the EABI requirement of always creating a frame pointer whenever
15390 a stack frame is allocated. This option is enabled by default and can
15391 be disabled with @option{-mno-linked-fp}.
15394 @opindex mlong-calls
15396 Use indirect addressing to call functions outside the current
15397 compilation unit. This allows the functions to be placed anywhere
15398 within the 32-bit address space.
15400 @item -malign-labels
15401 @opindex malign-labels
15403 Try to align labels to an 8-byte boundary by inserting NOPs into the
15404 previous packet. This option only has an effect when VLIW packing
15405 is enabled. It doesn't create new packets; it merely adds NOPs to
15408 @item -mlibrary-pic
15409 @opindex mlibrary-pic
15411 Generate position-independent EABI code.
15416 Use only the first four media accumulator registers.
15421 Use all eight media accumulator registers.
15426 Pack VLIW instructions.
15431 Do not pack VLIW instructions.
15434 @opindex mno-eflags
15436 Do not mark ABI switches in e_flags.
15439 @opindex mcond-move
15441 Enable the use of conditional-move instructions (default).
15443 This switch is mainly for debugging the compiler and will likely be removed
15444 in a future version.
15446 @item -mno-cond-move
15447 @opindex mno-cond-move
15449 Disable the use of conditional-move instructions.
15451 This switch is mainly for debugging the compiler and will likely be removed
15452 in a future version.
15457 Enable the use of conditional set instructions (default).
15459 This switch is mainly for debugging the compiler and will likely be removed
15460 in a future version.
15465 Disable the use of conditional set instructions.
15467 This switch is mainly for debugging the compiler and will likely be removed
15468 in a future version.
15471 @opindex mcond-exec
15473 Enable the use of conditional execution (default).
15475 This switch is mainly for debugging the compiler and will likely be removed
15476 in a future version.
15478 @item -mno-cond-exec
15479 @opindex mno-cond-exec
15481 Disable the use of conditional execution.
15483 This switch is mainly for debugging the compiler and will likely be removed
15484 in a future version.
15486 @item -mvliw-branch
15487 @opindex mvliw-branch
15489 Run a pass to pack branches into VLIW instructions (default).
15491 This switch is mainly for debugging the compiler and will likely be removed
15492 in a future version.
15494 @item -mno-vliw-branch
15495 @opindex mno-vliw-branch
15497 Do not run a pass to pack branches into VLIW instructions.
15499 This switch is mainly for debugging the compiler and will likely be removed
15500 in a future version.
15502 @item -mmulti-cond-exec
15503 @opindex mmulti-cond-exec
15505 Enable optimization of @code{&&} and @code{||} in conditional execution
15508 This switch is mainly for debugging the compiler and will likely be removed
15509 in a future version.
15511 @item -mno-multi-cond-exec
15512 @opindex mno-multi-cond-exec
15514 Disable optimization of @code{&&} and @code{||} in conditional execution.
15516 This switch is mainly for debugging the compiler and will likely be removed
15517 in a future version.
15519 @item -mnested-cond-exec
15520 @opindex mnested-cond-exec
15522 Enable nested conditional execution optimizations (default).
15524 This switch is mainly for debugging the compiler and will likely be removed
15525 in a future version.
15527 @item -mno-nested-cond-exec
15528 @opindex mno-nested-cond-exec
15530 Disable nested conditional execution optimizations.
15532 This switch is mainly for debugging the compiler and will likely be removed
15533 in a future version.
15535 @item -moptimize-membar
15536 @opindex moptimize-membar
15538 This switch removes redundant @code{membar} instructions from the
15539 compiler-generated code. It is enabled by default.
15541 @item -mno-optimize-membar
15542 @opindex mno-optimize-membar
15544 This switch disables the automatic removal of redundant @code{membar}
15545 instructions from the generated code.
15547 @item -mtomcat-stats
15548 @opindex mtomcat-stats
15550 Cause gas to print out tomcat statistics.
15552 @item -mcpu=@var{cpu}
15555 Select the processor type for which to generate code. Possible values are
15556 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15557 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15561 @node GNU/Linux Options
15562 @subsection GNU/Linux Options
15564 These @samp{-m} options are defined for GNU/Linux targets:
15569 Use the GNU C library. This is the default except
15570 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
15571 @samp{*-*-linux-*android*} targets.
15575 Use uClibc C library. This is the default on
15576 @samp{*-*-linux-*uclibc*} targets.
15580 Use the musl C library. This is the default on
15581 @samp{*-*-linux-*musl*} targets.
15585 Use Bionic C library. This is the default on
15586 @samp{*-*-linux-*android*} targets.
15590 Compile code compatible with Android platform. This is the default on
15591 @samp{*-*-linux-*android*} targets.
15593 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15594 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15595 this option makes the GCC driver pass Android-specific options to the linker.
15596 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15599 @item -tno-android-cc
15600 @opindex tno-android-cc
15601 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15602 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15603 @option{-fno-rtti} by default.
15605 @item -tno-android-ld
15606 @opindex tno-android-ld
15607 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15608 linking options to the linker.
15612 @node H8/300 Options
15613 @subsection H8/300 Options
15615 These @samp{-m} options are defined for the H8/300 implementations:
15620 Shorten some address references at link time, when possible; uses the
15621 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15622 ld, Using ld}, for a fuller description.
15626 Generate code for the H8/300H@.
15630 Generate code for the H8S@.
15634 Generate code for the H8S and H8/300H in the normal mode. This switch
15635 must be used either with @option{-mh} or @option{-ms}.
15639 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15643 Extended registers are stored on stack before execution of function
15644 with monitor attribute. Default option is @option{-mexr}.
15645 This option is valid only for H8S targets.
15649 Extended registers are not stored on stack before execution of function
15650 with monitor attribute. Default option is @option{-mno-exr}.
15651 This option is valid only for H8S targets.
15655 Make @code{int} data 32 bits by default.
15658 @opindex malign-300
15659 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15660 The default for the H8/300H and H8S is to align longs and floats on
15662 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15663 This option has no effect on the H8/300.
15667 @subsection HPPA Options
15668 @cindex HPPA Options
15670 These @samp{-m} options are defined for the HPPA family of computers:
15673 @item -march=@var{architecture-type}
15675 Generate code for the specified architecture. The choices for
15676 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15677 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15678 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15679 architecture option for your machine. Code compiled for lower numbered
15680 architectures runs on higher numbered architectures, but not the
15683 @item -mpa-risc-1-0
15684 @itemx -mpa-risc-1-1
15685 @itemx -mpa-risc-2-0
15686 @opindex mpa-risc-1-0
15687 @opindex mpa-risc-1-1
15688 @opindex mpa-risc-2-0
15689 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15691 @item -mjump-in-delay
15692 @opindex mjump-in-delay
15693 This option is ignored and provided for compatibility purposes only.
15695 @item -mdisable-fpregs
15696 @opindex mdisable-fpregs
15697 Prevent floating-point registers from being used in any manner. This is
15698 necessary for compiling kernels that perform lazy context switching of
15699 floating-point registers. If you use this option and attempt to perform
15700 floating-point operations, the compiler aborts.
15702 @item -mdisable-indexing
15703 @opindex mdisable-indexing
15704 Prevent the compiler from using indexing address modes. This avoids some
15705 rather obscure problems when compiling MIG generated code under MACH@.
15707 @item -mno-space-regs
15708 @opindex mno-space-regs
15709 Generate code that assumes the target has no space registers. This allows
15710 GCC to generate faster indirect calls and use unscaled index address modes.
15712 Such code is suitable for level 0 PA systems and kernels.
15714 @item -mfast-indirect-calls
15715 @opindex mfast-indirect-calls
15716 Generate code that assumes calls never cross space boundaries. This
15717 allows GCC to emit code that performs faster indirect calls.
15719 This option does not work in the presence of shared libraries or nested
15722 @item -mfixed-range=@var{register-range}
15723 @opindex mfixed-range
15724 Generate code treating the given register range as fixed registers.
15725 A fixed register is one that the register allocator cannot use. This is
15726 useful when compiling kernel code. A register range is specified as
15727 two registers separated by a dash. Multiple register ranges can be
15728 specified separated by a comma.
15730 @item -mlong-load-store
15731 @opindex mlong-load-store
15732 Generate 3-instruction load and store sequences as sometimes required by
15733 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15736 @item -mportable-runtime
15737 @opindex mportable-runtime
15738 Use the portable calling conventions proposed by HP for ELF systems.
15742 Enable the use of assembler directives only GAS understands.
15744 @item -mschedule=@var{cpu-type}
15746 Schedule code according to the constraints for the machine type
15747 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15748 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15749 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15750 proper scheduling option for your machine. The default scheduling is
15754 @opindex mlinker-opt
15755 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15756 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15757 linkers in which they give bogus error messages when linking some programs.
15760 @opindex msoft-float
15761 Generate output containing library calls for floating point.
15762 @strong{Warning:} the requisite libraries are not available for all HPPA
15763 targets. Normally the facilities of the machine's usual C compiler are
15764 used, but this cannot be done directly in cross-compilation. You must make
15765 your own arrangements to provide suitable library functions for
15768 @option{-msoft-float} changes the calling convention in the output file;
15769 therefore, it is only useful if you compile @emph{all} of a program with
15770 this option. In particular, you need to compile @file{libgcc.a}, the
15771 library that comes with GCC, with @option{-msoft-float} in order for
15776 Generate the predefine, @code{_SIO}, for server IO@. The default is
15777 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15778 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15779 options are available under HP-UX and HI-UX@.
15783 Use options specific to GNU @command{ld}.
15784 This passes @option{-shared} to @command{ld} when
15785 building a shared library. It is the default when GCC is configured,
15786 explicitly or implicitly, with the GNU linker. This option does not
15787 affect which @command{ld} is called; it only changes what parameters
15788 are passed to that @command{ld}.
15789 The @command{ld} that is called is determined by the
15790 @option{--with-ld} configure option, GCC's program search path, and
15791 finally by the user's @env{PATH}. The linker used by GCC can be printed
15792 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15793 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15797 Use options specific to HP @command{ld}.
15798 This passes @option{-b} to @command{ld} when building
15799 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15800 links. It is the default when GCC is configured, explicitly or
15801 implicitly, with the HP linker. This option does not affect
15802 which @command{ld} is called; it only changes what parameters are passed to that
15804 The @command{ld} that is called is determined by the @option{--with-ld}
15805 configure option, GCC's program search path, and finally by the user's
15806 @env{PATH}. The linker used by GCC can be printed using @samp{which
15807 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15808 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15811 @opindex mno-long-calls
15812 Generate code that uses long call sequences. This ensures that a call
15813 is always able to reach linker generated stubs. The default is to generate
15814 long calls only when the distance from the call site to the beginning
15815 of the function or translation unit, as the case may be, exceeds a
15816 predefined limit set by the branch type being used. The limits for
15817 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15818 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15821 Distances are measured from the beginning of functions when using the
15822 @option{-ffunction-sections} option, or when using the @option{-mgas}
15823 and @option{-mno-portable-runtime} options together under HP-UX with
15826 It is normally not desirable to use this option as it degrades
15827 performance. However, it may be useful in large applications,
15828 particularly when partial linking is used to build the application.
15830 The types of long calls used depends on the capabilities of the
15831 assembler and linker, and the type of code being generated. The
15832 impact on systems that support long absolute calls, and long pic
15833 symbol-difference or pc-relative calls should be relatively small.
15834 However, an indirect call is used on 32-bit ELF systems in pic code
15835 and it is quite long.
15837 @item -munix=@var{unix-std}
15839 Generate compiler predefines and select a startfile for the specified
15840 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15841 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15842 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15843 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15844 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15847 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15848 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15849 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15850 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15851 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15852 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15854 It is @emph{important} to note that this option changes the interfaces
15855 for various library routines. It also affects the operational behavior
15856 of the C library. Thus, @emph{extreme} care is needed in using this
15859 Library code that is intended to operate with more than one UNIX
15860 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15861 as appropriate. Most GNU software doesn't provide this capability.
15865 Suppress the generation of link options to search libdld.sl when the
15866 @option{-static} option is specified on HP-UX 10 and later.
15870 The HP-UX implementation of setlocale in libc has a dependency on
15871 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15872 when the @option{-static} option is specified, special link options
15873 are needed to resolve this dependency.
15875 On HP-UX 10 and later, the GCC driver adds the necessary options to
15876 link with libdld.sl when the @option{-static} option is specified.
15877 This causes the resulting binary to be dynamic. On the 64-bit port,
15878 the linkers generate dynamic binaries by default in any case. The
15879 @option{-nolibdld} option can be used to prevent the GCC driver from
15880 adding these link options.
15884 Add support for multithreading with the @dfn{dce thread} library
15885 under HP-UX@. This option sets flags for both the preprocessor and
15889 @node IA-64 Options
15890 @subsection IA-64 Options
15891 @cindex IA-64 Options
15893 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15897 @opindex mbig-endian
15898 Generate code for a big-endian target. This is the default for HP-UX@.
15900 @item -mlittle-endian
15901 @opindex mlittle-endian
15902 Generate code for a little-endian target. This is the default for AIX5
15908 @opindex mno-gnu-as
15909 Generate (or don't) code for the GNU assembler. This is the default.
15910 @c Also, this is the default if the configure option @option{--with-gnu-as}
15916 @opindex mno-gnu-ld
15917 Generate (or don't) code for the GNU linker. This is the default.
15918 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15923 Generate code that does not use a global pointer register. The result
15924 is not position independent code, and violates the IA-64 ABI@.
15926 @item -mvolatile-asm-stop
15927 @itemx -mno-volatile-asm-stop
15928 @opindex mvolatile-asm-stop
15929 @opindex mno-volatile-asm-stop
15930 Generate (or don't) a stop bit immediately before and after volatile asm
15933 @item -mregister-names
15934 @itemx -mno-register-names
15935 @opindex mregister-names
15936 @opindex mno-register-names
15937 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15938 the stacked registers. This may make assembler output more readable.
15944 Disable (or enable) optimizations that use the small data section. This may
15945 be useful for working around optimizer bugs.
15947 @item -mconstant-gp
15948 @opindex mconstant-gp
15949 Generate code that uses a single constant global pointer value. This is
15950 useful when compiling kernel code.
15954 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15955 This is useful when compiling firmware code.
15957 @item -minline-float-divide-min-latency
15958 @opindex minline-float-divide-min-latency
15959 Generate code for inline divides of floating-point values
15960 using the minimum latency algorithm.
15962 @item -minline-float-divide-max-throughput
15963 @opindex minline-float-divide-max-throughput
15964 Generate code for inline divides of floating-point values
15965 using the maximum throughput algorithm.
15967 @item -mno-inline-float-divide
15968 @opindex mno-inline-float-divide
15969 Do not generate inline code for divides of floating-point values.
15971 @item -minline-int-divide-min-latency
15972 @opindex minline-int-divide-min-latency
15973 Generate code for inline divides of integer values
15974 using the minimum latency algorithm.
15976 @item -minline-int-divide-max-throughput
15977 @opindex minline-int-divide-max-throughput
15978 Generate code for inline divides of integer values
15979 using the maximum throughput algorithm.
15981 @item -mno-inline-int-divide
15982 @opindex mno-inline-int-divide
15983 Do not generate inline code for divides of integer values.
15985 @item -minline-sqrt-min-latency
15986 @opindex minline-sqrt-min-latency
15987 Generate code for inline square roots
15988 using the minimum latency algorithm.
15990 @item -minline-sqrt-max-throughput
15991 @opindex minline-sqrt-max-throughput
15992 Generate code for inline square roots
15993 using the maximum throughput algorithm.
15995 @item -mno-inline-sqrt
15996 @opindex mno-inline-sqrt
15997 Do not generate inline code for @code{sqrt}.
16000 @itemx -mno-fused-madd
16001 @opindex mfused-madd
16002 @opindex mno-fused-madd
16003 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16004 instructions. The default is to use these instructions.
16006 @item -mno-dwarf2-asm
16007 @itemx -mdwarf2-asm
16008 @opindex mno-dwarf2-asm
16009 @opindex mdwarf2-asm
16010 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16011 info. This may be useful when not using the GNU assembler.
16013 @item -mearly-stop-bits
16014 @itemx -mno-early-stop-bits
16015 @opindex mearly-stop-bits
16016 @opindex mno-early-stop-bits
16017 Allow stop bits to be placed earlier than immediately preceding the
16018 instruction that triggered the stop bit. This can improve instruction
16019 scheduling, but does not always do so.
16021 @item -mfixed-range=@var{register-range}
16022 @opindex mfixed-range
16023 Generate code treating the given register range as fixed registers.
16024 A fixed register is one that the register allocator cannot use. This is
16025 useful when compiling kernel code. A register range is specified as
16026 two registers separated by a dash. Multiple register ranges can be
16027 specified separated by a comma.
16029 @item -mtls-size=@var{tls-size}
16031 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16034 @item -mtune=@var{cpu-type}
16036 Tune the instruction scheduling for a particular CPU, Valid values are
16037 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16038 and @samp{mckinley}.
16044 Generate code for a 32-bit or 64-bit environment.
16045 The 32-bit environment sets int, long and pointer to 32 bits.
16046 The 64-bit environment sets int to 32 bits and long and pointer
16047 to 64 bits. These are HP-UX specific flags.
16049 @item -mno-sched-br-data-spec
16050 @itemx -msched-br-data-spec
16051 @opindex mno-sched-br-data-spec
16052 @opindex msched-br-data-spec
16053 (Dis/En)able data speculative scheduling before reload.
16054 This results in generation of @code{ld.a} instructions and
16055 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16056 The default is 'disable'.
16058 @item -msched-ar-data-spec
16059 @itemx -mno-sched-ar-data-spec
16060 @opindex msched-ar-data-spec
16061 @opindex mno-sched-ar-data-spec
16062 (En/Dis)able data speculative scheduling after reload.
16063 This results in generation of @code{ld.a} instructions and
16064 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16065 The default is 'enable'.
16067 @item -mno-sched-control-spec
16068 @itemx -msched-control-spec
16069 @opindex mno-sched-control-spec
16070 @opindex msched-control-spec
16071 (Dis/En)able control speculative scheduling. This feature is
16072 available only during region scheduling (i.e.@: before reload).
16073 This results in generation of the @code{ld.s} instructions and
16074 the corresponding check instructions @code{chk.s}.
16075 The default is 'disable'.
16077 @item -msched-br-in-data-spec
16078 @itemx -mno-sched-br-in-data-spec
16079 @opindex msched-br-in-data-spec
16080 @opindex mno-sched-br-in-data-spec
16081 (En/Dis)able speculative scheduling of the instructions that
16082 are dependent on the data speculative loads before reload.
16083 This is effective only with @option{-msched-br-data-spec} enabled.
16084 The default is 'enable'.
16086 @item -msched-ar-in-data-spec
16087 @itemx -mno-sched-ar-in-data-spec
16088 @opindex msched-ar-in-data-spec
16089 @opindex mno-sched-ar-in-data-spec
16090 (En/Dis)able speculative scheduling of the instructions that
16091 are dependent on the data speculative loads after reload.
16092 This is effective only with @option{-msched-ar-data-spec} enabled.
16093 The default is 'enable'.
16095 @item -msched-in-control-spec
16096 @itemx -mno-sched-in-control-spec
16097 @opindex msched-in-control-spec
16098 @opindex mno-sched-in-control-spec
16099 (En/Dis)able speculative scheduling of the instructions that
16100 are dependent on the control speculative loads.
16101 This is effective only with @option{-msched-control-spec} enabled.
16102 The default is 'enable'.
16104 @item -mno-sched-prefer-non-data-spec-insns
16105 @itemx -msched-prefer-non-data-spec-insns
16106 @opindex mno-sched-prefer-non-data-spec-insns
16107 @opindex msched-prefer-non-data-spec-insns
16108 If enabled, data-speculative instructions are chosen for schedule
16109 only if there are no other choices at the moment. This makes
16110 the use of the data speculation much more conservative.
16111 The default is 'disable'.
16113 @item -mno-sched-prefer-non-control-spec-insns
16114 @itemx -msched-prefer-non-control-spec-insns
16115 @opindex mno-sched-prefer-non-control-spec-insns
16116 @opindex msched-prefer-non-control-spec-insns
16117 If enabled, control-speculative instructions are chosen for schedule
16118 only if there are no other choices at the moment. This makes
16119 the use of the control speculation much more conservative.
16120 The default is 'disable'.
16122 @item -mno-sched-count-spec-in-critical-path
16123 @itemx -msched-count-spec-in-critical-path
16124 @opindex mno-sched-count-spec-in-critical-path
16125 @opindex msched-count-spec-in-critical-path
16126 If enabled, speculative dependencies are considered during
16127 computation of the instructions priorities. This makes the use of the
16128 speculation a bit more conservative.
16129 The default is 'disable'.
16131 @item -msched-spec-ldc
16132 @opindex msched-spec-ldc
16133 Use a simple data speculation check. This option is on by default.
16135 @item -msched-control-spec-ldc
16136 @opindex msched-spec-ldc
16137 Use a simple check for control speculation. This option is on by default.
16139 @item -msched-stop-bits-after-every-cycle
16140 @opindex msched-stop-bits-after-every-cycle
16141 Place a stop bit after every cycle when scheduling. This option is on
16144 @item -msched-fp-mem-deps-zero-cost
16145 @opindex msched-fp-mem-deps-zero-cost
16146 Assume that floating-point stores and loads are not likely to cause a conflict
16147 when placed into the same instruction group. This option is disabled by
16150 @item -msel-sched-dont-check-control-spec
16151 @opindex msel-sched-dont-check-control-spec
16152 Generate checks for control speculation in selective scheduling.
16153 This flag is disabled by default.
16155 @item -msched-max-memory-insns=@var{max-insns}
16156 @opindex msched-max-memory-insns
16157 Limit on the number of memory insns per instruction group, giving lower
16158 priority to subsequent memory insns attempting to schedule in the same
16159 instruction group. Frequently useful to prevent cache bank conflicts.
16160 The default value is 1.
16162 @item -msched-max-memory-insns-hard-limit
16163 @opindex msched-max-memory-insns-hard-limit
16164 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16165 disallowing more than that number in an instruction group.
16166 Otherwise, the limit is ``soft'', meaning that non-memory operations
16167 are preferred when the limit is reached, but memory operations may still
16173 @subsection LM32 Options
16174 @cindex LM32 options
16176 These @option{-m} options are defined for the LatticeMico32 architecture:
16179 @item -mbarrel-shift-enabled
16180 @opindex mbarrel-shift-enabled
16181 Enable barrel-shift instructions.
16183 @item -mdivide-enabled
16184 @opindex mdivide-enabled
16185 Enable divide and modulus instructions.
16187 @item -mmultiply-enabled
16188 @opindex multiply-enabled
16189 Enable multiply instructions.
16191 @item -msign-extend-enabled
16192 @opindex msign-extend-enabled
16193 Enable sign extend instructions.
16195 @item -muser-enabled
16196 @opindex muser-enabled
16197 Enable user-defined instructions.
16202 @subsection M32C Options
16203 @cindex M32C options
16206 @item -mcpu=@var{name}
16208 Select the CPU for which code is generated. @var{name} may be one of
16209 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16210 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16211 the M32C/80 series.
16215 Specifies that the program will be run on the simulator. This causes
16216 an alternate runtime library to be linked in which supports, for
16217 example, file I/O@. You must not use this option when generating
16218 programs that will run on real hardware; you must provide your own
16219 runtime library for whatever I/O functions are needed.
16221 @item -memregs=@var{number}
16223 Specifies the number of memory-based pseudo-registers GCC uses
16224 during code generation. These pseudo-registers are used like real
16225 registers, so there is a tradeoff between GCC's ability to fit the
16226 code into available registers, and the performance penalty of using
16227 memory instead of registers. Note that all modules in a program must
16228 be compiled with the same value for this option. Because of that, you
16229 must not use this option with GCC's default runtime libraries.
16233 @node M32R/D Options
16234 @subsection M32R/D Options
16235 @cindex M32R/D options
16237 These @option{-m} options are defined for Renesas M32R/D architectures:
16242 Generate code for the M32R/2@.
16246 Generate code for the M32R/X@.
16250 Generate code for the M32R@. This is the default.
16252 @item -mmodel=small
16253 @opindex mmodel=small
16254 Assume all objects live in the lower 16MB of memory (so that their addresses
16255 can be loaded with the @code{ld24} instruction), and assume all subroutines
16256 are reachable with the @code{bl} instruction.
16257 This is the default.
16259 The addressability of a particular object can be set with the
16260 @code{model} attribute.
16262 @item -mmodel=medium
16263 @opindex mmodel=medium
16264 Assume objects may be anywhere in the 32-bit address space (the compiler
16265 generates @code{seth/add3} instructions to load their addresses), and
16266 assume all subroutines are reachable with the @code{bl} instruction.
16268 @item -mmodel=large
16269 @opindex mmodel=large
16270 Assume objects may be anywhere in the 32-bit address space (the compiler
16271 generates @code{seth/add3} instructions to load their addresses), and
16272 assume subroutines may not be reachable with the @code{bl} instruction
16273 (the compiler generates the much slower @code{seth/add3/jl}
16274 instruction sequence).
16277 @opindex msdata=none
16278 Disable use of the small data area. Variables are put into
16279 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16280 @code{section} attribute has been specified).
16281 This is the default.
16283 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16284 Objects may be explicitly put in the small data area with the
16285 @code{section} attribute using one of these sections.
16287 @item -msdata=sdata
16288 @opindex msdata=sdata
16289 Put small global and static data in the small data area, but do not
16290 generate special code to reference them.
16293 @opindex msdata=use
16294 Put small global and static data in the small data area, and generate
16295 special instructions to reference them.
16299 @cindex smaller data references
16300 Put global and static objects less than or equal to @var{num} bytes
16301 into the small data or BSS sections instead of the normal data or BSS
16302 sections. The default value of @var{num} is 8.
16303 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16304 for this option to have any effect.
16306 All modules should be compiled with the same @option{-G @var{num}} value.
16307 Compiling with different values of @var{num} may or may not work; if it
16308 doesn't the linker gives an error message---incorrect code is not
16313 Makes the M32R-specific code in the compiler display some statistics
16314 that might help in debugging programs.
16316 @item -malign-loops
16317 @opindex malign-loops
16318 Align all loops to a 32-byte boundary.
16320 @item -mno-align-loops
16321 @opindex mno-align-loops
16322 Do not enforce a 32-byte alignment for loops. This is the default.
16324 @item -missue-rate=@var{number}
16325 @opindex missue-rate=@var{number}
16326 Issue @var{number} instructions per cycle. @var{number} can only be 1
16329 @item -mbranch-cost=@var{number}
16330 @opindex mbranch-cost=@var{number}
16331 @var{number} can only be 1 or 2. If it is 1 then branches are
16332 preferred over conditional code, if it is 2, then the opposite applies.
16334 @item -mflush-trap=@var{number}
16335 @opindex mflush-trap=@var{number}
16336 Specifies the trap number to use to flush the cache. The default is
16337 12. Valid numbers are between 0 and 15 inclusive.
16339 @item -mno-flush-trap
16340 @opindex mno-flush-trap
16341 Specifies that the cache cannot be flushed by using a trap.
16343 @item -mflush-func=@var{name}
16344 @opindex mflush-func=@var{name}
16345 Specifies the name of the operating system function to call to flush
16346 the cache. The default is @samp{_flush_cache}, but a function call
16347 is only used if a trap is not available.
16349 @item -mno-flush-func
16350 @opindex mno-flush-func
16351 Indicates that there is no OS function for flushing the cache.
16355 @node M680x0 Options
16356 @subsection M680x0 Options
16357 @cindex M680x0 options
16359 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16360 The default settings depend on which architecture was selected when
16361 the compiler was configured; the defaults for the most common choices
16365 @item -march=@var{arch}
16367 Generate code for a specific M680x0 or ColdFire instruction set
16368 architecture. Permissible values of @var{arch} for M680x0
16369 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16370 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16371 architectures are selected according to Freescale's ISA classification
16372 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16373 @samp{isab} and @samp{isac}.
16375 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16376 code for a ColdFire target. The @var{arch} in this macro is one of the
16377 @option{-march} arguments given above.
16379 When used together, @option{-march} and @option{-mtune} select code
16380 that runs on a family of similar processors but that is optimized
16381 for a particular microarchitecture.
16383 @item -mcpu=@var{cpu}
16385 Generate code for a specific M680x0 or ColdFire processor.
16386 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16387 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16388 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16389 below, which also classifies the CPUs into families:
16391 @multitable @columnfractions 0.20 0.80
16392 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16393 @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}
16394 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16395 @item @samp{5206e} @tab @samp{5206e}
16396 @item @samp{5208} @tab @samp{5207} @samp{5208}
16397 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16398 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16399 @item @samp{5216} @tab @samp{5214} @samp{5216}
16400 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16401 @item @samp{5225} @tab @samp{5224} @samp{5225}
16402 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16403 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16404 @item @samp{5249} @tab @samp{5249}
16405 @item @samp{5250} @tab @samp{5250}
16406 @item @samp{5271} @tab @samp{5270} @samp{5271}
16407 @item @samp{5272} @tab @samp{5272}
16408 @item @samp{5275} @tab @samp{5274} @samp{5275}
16409 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16410 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16411 @item @samp{5307} @tab @samp{5307}
16412 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16413 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16414 @item @samp{5407} @tab @samp{5407}
16415 @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}
16418 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16419 @var{arch} is compatible with @var{cpu}. Other combinations of
16420 @option{-mcpu} and @option{-march} are rejected.
16422 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16423 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16424 where the value of @var{family} is given by the table above.
16426 @item -mtune=@var{tune}
16428 Tune the code for a particular microarchitecture within the
16429 constraints set by @option{-march} and @option{-mcpu}.
16430 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16431 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16432 and @samp{cpu32}. The ColdFire microarchitectures
16433 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16435 You can also use @option{-mtune=68020-40} for code that needs
16436 to run relatively well on 68020, 68030 and 68040 targets.
16437 @option{-mtune=68020-60} is similar but includes 68060 targets
16438 as well. These two options select the same tuning decisions as
16439 @option{-m68020-40} and @option{-m68020-60} respectively.
16441 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16442 when tuning for 680x0 architecture @var{arch}. It also defines
16443 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16444 option is used. If GCC is tuning for a range of architectures,
16445 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16446 it defines the macros for every architecture in the range.
16448 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16449 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16450 of the arguments given above.
16456 Generate output for a 68000. This is the default
16457 when the compiler is configured for 68000-based systems.
16458 It is equivalent to @option{-march=68000}.
16460 Use this option for microcontrollers with a 68000 or EC000 core,
16461 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16465 Generate output for a 68010. This is the default
16466 when the compiler is configured for 68010-based systems.
16467 It is equivalent to @option{-march=68010}.
16473 Generate output for a 68020. This is the default
16474 when the compiler is configured for 68020-based systems.
16475 It is equivalent to @option{-march=68020}.
16479 Generate output for a 68030. This is the default when the compiler is
16480 configured for 68030-based systems. It is equivalent to
16481 @option{-march=68030}.
16485 Generate output for a 68040. This is the default when the compiler is
16486 configured for 68040-based systems. It is equivalent to
16487 @option{-march=68040}.
16489 This option inhibits the use of 68881/68882 instructions that have to be
16490 emulated by software on the 68040. Use this option if your 68040 does not
16491 have code to emulate those instructions.
16495 Generate output for a 68060. This is the default when the compiler is
16496 configured for 68060-based systems. It is equivalent to
16497 @option{-march=68060}.
16499 This option inhibits the use of 68020 and 68881/68882 instructions that
16500 have to be emulated by software on the 68060. Use this option if your 68060
16501 does not have code to emulate those instructions.
16505 Generate output for a CPU32. This is the default
16506 when the compiler is configured for CPU32-based systems.
16507 It is equivalent to @option{-march=cpu32}.
16509 Use this option for microcontrollers with a
16510 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16511 68336, 68340, 68341, 68349 and 68360.
16515 Generate output for a 520X ColdFire CPU@. This is the default
16516 when the compiler is configured for 520X-based systems.
16517 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16518 in favor of that option.
16520 Use this option for microcontroller with a 5200 core, including
16521 the MCF5202, MCF5203, MCF5204 and MCF5206.
16525 Generate output for a 5206e ColdFire CPU@. The option is now
16526 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16530 Generate output for a member of the ColdFire 528X family.
16531 The option is now deprecated in favor of the equivalent
16532 @option{-mcpu=528x}.
16536 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16537 in favor of the equivalent @option{-mcpu=5307}.
16541 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16542 in favor of the equivalent @option{-mcpu=5407}.
16546 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16547 This includes use of hardware floating-point instructions.
16548 The option is equivalent to @option{-mcpu=547x}, and is now
16549 deprecated in favor of that option.
16553 Generate output for a 68040, without using any of the new instructions.
16554 This results in code that can run relatively efficiently on either a
16555 68020/68881 or a 68030 or a 68040. The generated code does use the
16556 68881 instructions that are emulated on the 68040.
16558 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16562 Generate output for a 68060, without using any of the new instructions.
16563 This results in code that can run relatively efficiently on either a
16564 68020/68881 or a 68030 or a 68040. The generated code does use the
16565 68881 instructions that are emulated on the 68060.
16567 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16571 @opindex mhard-float
16573 Generate floating-point instructions. This is the default for 68020
16574 and above, and for ColdFire devices that have an FPU@. It defines the
16575 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16576 on ColdFire targets.
16579 @opindex msoft-float
16580 Do not generate floating-point instructions; use library calls instead.
16581 This is the default for 68000, 68010, and 68832 targets. It is also
16582 the default for ColdFire devices that have no FPU.
16588 Generate (do not generate) ColdFire hardware divide and remainder
16589 instructions. If @option{-march} is used without @option{-mcpu},
16590 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16591 architectures. Otherwise, the default is taken from the target CPU
16592 (either the default CPU, or the one specified by @option{-mcpu}). For
16593 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16594 @option{-mcpu=5206e}.
16596 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16600 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16601 Additionally, parameters passed on the stack are also aligned to a
16602 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16606 Do not consider type @code{int} to be 16 bits wide. This is the default.
16609 @itemx -mno-bitfield
16610 @opindex mnobitfield
16611 @opindex mno-bitfield
16612 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16613 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16617 Do use the bit-field instructions. The @option{-m68020} option implies
16618 @option{-mbitfield}. This is the default if you use a configuration
16619 designed for a 68020.
16623 Use a different function-calling convention, in which functions
16624 that take a fixed number of arguments return with the @code{rtd}
16625 instruction, which pops their arguments while returning. This
16626 saves one instruction in the caller since there is no need to pop
16627 the arguments there.
16629 This calling convention is incompatible with the one normally
16630 used on Unix, so you cannot use it if you need to call libraries
16631 compiled with the Unix compiler.
16633 Also, you must provide function prototypes for all functions that
16634 take variable numbers of arguments (including @code{printf});
16635 otherwise incorrect code is generated for calls to those
16638 In addition, seriously incorrect code results if you call a
16639 function with too many arguments. (Normally, extra arguments are
16640 harmlessly ignored.)
16642 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16643 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16647 Do not use the calling conventions selected by @option{-mrtd}.
16648 This is the default.
16651 @itemx -mno-align-int
16652 @opindex malign-int
16653 @opindex mno-align-int
16654 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16655 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16656 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16657 Aligning variables on 32-bit boundaries produces code that runs somewhat
16658 faster on processors with 32-bit busses at the expense of more memory.
16660 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16661 aligns structures containing the above types differently than
16662 most published application binary interface specifications for the m68k.
16666 Use the pc-relative addressing mode of the 68000 directly, instead of
16667 using a global offset table. At present, this option implies @option{-fpic},
16668 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16669 not presently supported with @option{-mpcrel}, though this could be supported for
16670 68020 and higher processors.
16672 @item -mno-strict-align
16673 @itemx -mstrict-align
16674 @opindex mno-strict-align
16675 @opindex mstrict-align
16676 Do not (do) assume that unaligned memory references are handled by
16680 Generate code that allows the data segment to be located in a different
16681 area of memory from the text segment. This allows for execute-in-place in
16682 an environment without virtual memory management. This option implies
16685 @item -mno-sep-data
16686 Generate code that assumes that the data segment follows the text segment.
16687 This is the default.
16689 @item -mid-shared-library
16690 Generate code that supports shared libraries via the library ID method.
16691 This allows for execute-in-place and shared libraries in an environment
16692 without virtual memory management. This option implies @option{-fPIC}.
16694 @item -mno-id-shared-library
16695 Generate code that doesn't assume ID-based shared libraries are being used.
16696 This is the default.
16698 @item -mshared-library-id=n
16699 Specifies the identification number of the ID-based shared library being
16700 compiled. Specifying a value of 0 generates more compact code; specifying
16701 other values forces the allocation of that number to the current
16702 library, but is no more space- or time-efficient than omitting this option.
16708 When generating position-independent code for ColdFire, generate code
16709 that works if the GOT has more than 8192 entries. This code is
16710 larger and slower than code generated without this option. On M680x0
16711 processors, this option is not needed; @option{-fPIC} suffices.
16713 GCC normally uses a single instruction to load values from the GOT@.
16714 While this is relatively efficient, it only works if the GOT
16715 is smaller than about 64k. Anything larger causes the linker
16716 to report an error such as:
16718 @cindex relocation truncated to fit (ColdFire)
16720 relocation truncated to fit: R_68K_GOT16O foobar
16723 If this happens, you should recompile your code with @option{-mxgot}.
16724 It should then work with very large GOTs. However, code generated with
16725 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16726 the value of a global symbol.
16728 Note that some linkers, including newer versions of the GNU linker,
16729 can create multiple GOTs and sort GOT entries. If you have such a linker,
16730 you should only need to use @option{-mxgot} when compiling a single
16731 object file that accesses more than 8192 GOT entries. Very few do.
16733 These options have no effect unless GCC is generating
16734 position-independent code.
16738 @node MCore Options
16739 @subsection MCore Options
16740 @cindex MCore options
16742 These are the @samp{-m} options defined for the Motorola M*Core
16748 @itemx -mno-hardlit
16750 @opindex mno-hardlit
16751 Inline constants into the code stream if it can be done in two
16752 instructions or less.
16758 Use the divide instruction. (Enabled by default).
16760 @item -mrelax-immediate
16761 @itemx -mno-relax-immediate
16762 @opindex mrelax-immediate
16763 @opindex mno-relax-immediate
16764 Allow arbitrary-sized immediates in bit operations.
16766 @item -mwide-bitfields
16767 @itemx -mno-wide-bitfields
16768 @opindex mwide-bitfields
16769 @opindex mno-wide-bitfields
16770 Always treat bit-fields as @code{int}-sized.
16772 @item -m4byte-functions
16773 @itemx -mno-4byte-functions
16774 @opindex m4byte-functions
16775 @opindex mno-4byte-functions
16776 Force all functions to be aligned to a 4-byte boundary.
16778 @item -mcallgraph-data
16779 @itemx -mno-callgraph-data
16780 @opindex mcallgraph-data
16781 @opindex mno-callgraph-data
16782 Emit callgraph information.
16785 @itemx -mno-slow-bytes
16786 @opindex mslow-bytes
16787 @opindex mno-slow-bytes
16788 Prefer word access when reading byte quantities.
16790 @item -mlittle-endian
16791 @itemx -mbig-endian
16792 @opindex mlittle-endian
16793 @opindex mbig-endian
16794 Generate code for a little-endian target.
16800 Generate code for the 210 processor.
16804 Assume that runtime support has been provided and so omit the
16805 simulator library (@file{libsim.a)} from the linker command line.
16807 @item -mstack-increment=@var{size}
16808 @opindex mstack-increment
16809 Set the maximum amount for a single stack increment operation. Large
16810 values can increase the speed of programs that contain functions
16811 that need a large amount of stack space, but they can also trigger a
16812 segmentation fault if the stack is extended too much. The default
16818 @subsection MeP Options
16819 @cindex MeP options
16825 Enables the @code{abs} instruction, which is the absolute difference
16826 between two registers.
16830 Enables all the optional instructions---average, multiply, divide, bit
16831 operations, leading zero, absolute difference, min/max, clip, and
16837 Enables the @code{ave} instruction, which computes the average of two
16840 @item -mbased=@var{n}
16842 Variables of size @var{n} bytes or smaller are placed in the
16843 @code{.based} section by default. Based variables use the @code{$tp}
16844 register as a base register, and there is a 128-byte limit to the
16845 @code{.based} section.
16849 Enables the bit operation instructions---bit test (@code{btstm}), set
16850 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16851 test-and-set (@code{tas}).
16853 @item -mc=@var{name}
16855 Selects which section constant data is placed in. @var{name} may
16856 be @samp{tiny}, @samp{near}, or @samp{far}.
16860 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16861 useful unless you also provide @option{-mminmax}.
16863 @item -mconfig=@var{name}
16865 Selects one of the built-in core configurations. Each MeP chip has
16866 one or more modules in it; each module has a core CPU and a variety of
16867 coprocessors, optional instructions, and peripherals. The
16868 @code{MeP-Integrator} tool, not part of GCC, provides these
16869 configurations through this option; using this option is the same as
16870 using all the corresponding command-line options. The default
16871 configuration is @samp{default}.
16875 Enables the coprocessor instructions. By default, this is a 32-bit
16876 coprocessor. Note that the coprocessor is normally enabled via the
16877 @option{-mconfig=} option.
16881 Enables the 32-bit coprocessor's instructions.
16885 Enables the 64-bit coprocessor's instructions.
16889 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16893 Causes constant variables to be placed in the @code{.near} section.
16897 Enables the @code{div} and @code{divu} instructions.
16901 Generate big-endian code.
16905 Generate little-endian code.
16907 @item -mio-volatile
16908 @opindex mio-volatile
16909 Tells the compiler that any variable marked with the @code{io}
16910 attribute is to be considered volatile.
16914 Causes variables to be assigned to the @code{.far} section by default.
16918 Enables the @code{leadz} (leading zero) instruction.
16922 Causes variables to be assigned to the @code{.near} section by default.
16926 Enables the @code{min} and @code{max} instructions.
16930 Enables the multiplication and multiply-accumulate instructions.
16934 Disables all the optional instructions enabled by @option{-mall-opts}.
16938 Enables the @code{repeat} and @code{erepeat} instructions, used for
16939 low-overhead looping.
16943 Causes all variables to default to the @code{.tiny} section. Note
16944 that there is a 65536-byte limit to this section. Accesses to these
16945 variables use the @code{%gp} base register.
16949 Enables the saturation instructions. Note that the compiler does not
16950 currently generate these itself, but this option is included for
16951 compatibility with other tools, like @code{as}.
16955 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16959 Link the simulator run-time libraries.
16963 Link the simulator runtime libraries, excluding built-in support
16964 for reset and exception vectors and tables.
16968 Causes all functions to default to the @code{.far} section. Without
16969 this option, functions default to the @code{.near} section.
16971 @item -mtiny=@var{n}
16973 Variables that are @var{n} bytes or smaller are allocated to the
16974 @code{.tiny} section. These variables use the @code{$gp} base
16975 register. The default for this option is 4, but note that there's a
16976 65536-byte limit to the @code{.tiny} section.
16980 @node MicroBlaze Options
16981 @subsection MicroBlaze Options
16982 @cindex MicroBlaze Options
16987 @opindex msoft-float
16988 Use software emulation for floating point (default).
16991 @opindex mhard-float
16992 Use hardware floating-point instructions.
16996 Do not optimize block moves, use @code{memcpy}.
16998 @item -mno-clearbss
16999 @opindex mno-clearbss
17000 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17002 @item -mcpu=@var{cpu-type}
17004 Use features of, and schedule code for, the given CPU.
17005 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17006 where @var{X} is a major version, @var{YY} is the minor version, and
17007 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17008 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17010 @item -mxl-soft-mul
17011 @opindex mxl-soft-mul
17012 Use software multiply emulation (default).
17014 @item -mxl-soft-div
17015 @opindex mxl-soft-div
17016 Use software emulation for divides (default).
17018 @item -mxl-barrel-shift
17019 @opindex mxl-barrel-shift
17020 Use the hardware barrel shifter.
17022 @item -mxl-pattern-compare
17023 @opindex mxl-pattern-compare
17024 Use pattern compare instructions.
17026 @item -msmall-divides
17027 @opindex msmall-divides
17028 Use table lookup optimization for small signed integer divisions.
17030 @item -mxl-stack-check
17031 @opindex mxl-stack-check
17032 This option is deprecated. Use @option{-fstack-check} instead.
17035 @opindex mxl-gp-opt
17036 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17038 @item -mxl-multiply-high
17039 @opindex mxl-multiply-high
17040 Use multiply high instructions for high part of 32x32 multiply.
17042 @item -mxl-float-convert
17043 @opindex mxl-float-convert
17044 Use hardware floating-point conversion instructions.
17046 @item -mxl-float-sqrt
17047 @opindex mxl-float-sqrt
17048 Use hardware floating-point square root instruction.
17051 @opindex mbig-endian
17052 Generate code for a big-endian target.
17054 @item -mlittle-endian
17055 @opindex mlittle-endian
17056 Generate code for a little-endian target.
17059 @opindex mxl-reorder
17060 Use reorder instructions (swap and byte reversed load/store).
17062 @item -mxl-mode-@var{app-model}
17063 Select application model @var{app-model}. Valid models are
17066 normal executable (default), uses startup code @file{crt0.o}.
17069 for use with Xilinx Microprocessor Debugger (XMD) based
17070 software intrusive debug agent called xmdstub. This uses startup file
17071 @file{crt1.o} and sets the start address of the program to 0x800.
17074 for applications that are loaded using a bootloader.
17075 This model uses startup file @file{crt2.o} which does not contain a processor
17076 reset vector handler. This is suitable for transferring control on a
17077 processor reset to the bootloader rather than the application.
17080 for applications that do not require any of the
17081 MicroBlaze vectors. This option may be useful for applications running
17082 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17085 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17086 @option{-mxl-mode-@var{app-model}}.
17091 @subsection MIPS Options
17092 @cindex MIPS options
17098 Generate big-endian code.
17102 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17105 @item -march=@var{arch}
17107 Generate code that runs on @var{arch}, which can be the name of a
17108 generic MIPS ISA, or the name of a particular processor.
17110 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17111 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17112 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17113 @samp{mips64r5} and @samp{mips64r6}.
17114 The processor names are:
17115 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17116 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17117 @samp{5kc}, @samp{5kf},
17119 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17120 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17121 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17122 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17123 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17126 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17128 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17129 @samp{m5100}, @samp{m5101},
17130 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17133 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17134 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17135 @samp{rm7000}, @samp{rm9000},
17136 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17139 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17140 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17141 @samp{xlr} and @samp{xlp}.
17142 The special value @samp{from-abi} selects the
17143 most compatible architecture for the selected ABI (that is,
17144 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17146 The native Linux/GNU toolchain also supports the value @samp{native},
17147 which selects the best architecture option for the host processor.
17148 @option{-march=native} has no effect if GCC does not recognize
17151 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17152 (for example, @option{-march=r2k}). Prefixes are optional, and
17153 @samp{vr} may be written @samp{r}.
17155 Names of the form @samp{@var{n}f2_1} refer to processors with
17156 FPUs clocked at half the rate of the core, names of the form
17157 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17158 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17159 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17160 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17161 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17162 accepted as synonyms for @samp{@var{n}f1_1}.
17164 GCC defines two macros based on the value of this option. The first
17165 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17166 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17167 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17168 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17169 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17171 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17172 above. In other words, it has the full prefix and does not
17173 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17174 the macro names the resolved architecture (either @code{"mips1"} or
17175 @code{"mips3"}). It names the default architecture when no
17176 @option{-march} option is given.
17178 @item -mtune=@var{arch}
17180 Optimize for @var{arch}. Among other things, this option controls
17181 the way instructions are scheduled, and the perceived cost of arithmetic
17182 operations. The list of @var{arch} values is the same as for
17185 When this option is not used, GCC optimizes for the processor
17186 specified by @option{-march}. By using @option{-march} and
17187 @option{-mtune} together, it is possible to generate code that
17188 runs on a family of processors, but optimize the code for one
17189 particular member of that family.
17191 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17192 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17193 @option{-march} ones described above.
17197 Equivalent to @option{-march=mips1}.
17201 Equivalent to @option{-march=mips2}.
17205 Equivalent to @option{-march=mips3}.
17209 Equivalent to @option{-march=mips4}.
17213 Equivalent to @option{-march=mips32}.
17217 Equivalent to @option{-march=mips32r3}.
17221 Equivalent to @option{-march=mips32r5}.
17225 Equivalent to @option{-march=mips32r6}.
17229 Equivalent to @option{-march=mips64}.
17233 Equivalent to @option{-march=mips64r2}.
17237 Equivalent to @option{-march=mips64r3}.
17241 Equivalent to @option{-march=mips64r5}.
17245 Equivalent to @option{-march=mips64r6}.
17250 @opindex mno-mips16
17251 Generate (do not generate) MIPS16 code. If GCC is targeting a
17252 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17254 MIPS16 code generation can also be controlled on a per-function basis
17255 by means of @code{mips16} and @code{nomips16} attributes.
17256 @xref{Function Attributes}, for more information.
17258 @item -mflip-mips16
17259 @opindex mflip-mips16
17260 Generate MIPS16 code on alternating functions. This option is provided
17261 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17262 not intended for ordinary use in compiling user code.
17264 @item -minterlink-compressed
17265 @item -mno-interlink-compressed
17266 @opindex minterlink-compressed
17267 @opindex mno-interlink-compressed
17268 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17269 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17271 For example, code using the standard ISA encoding cannot jump directly
17272 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17273 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17274 knows that the target of the jump is not compressed.
17276 @item -minterlink-mips16
17277 @itemx -mno-interlink-mips16
17278 @opindex minterlink-mips16
17279 @opindex mno-interlink-mips16
17280 Aliases of @option{-minterlink-compressed} and
17281 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17282 and are retained for backwards compatibility.
17294 Generate code for the given ABI@.
17296 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17297 generates 64-bit code when you select a 64-bit architecture, but you
17298 can use @option{-mgp32} to get 32-bit code instead.
17300 For information about the O64 ABI, see
17301 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17303 GCC supports a variant of the o32 ABI in which floating-point registers
17304 are 64 rather than 32 bits wide. You can select this combination with
17305 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17306 and @code{mfhc1} instructions and is therefore only supported for
17307 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17309 The register assignments for arguments and return values remain the
17310 same, but each scalar value is passed in a single 64-bit register
17311 rather than a pair of 32-bit registers. For example, scalar
17312 floating-point values are returned in @samp{$f0} only, not a
17313 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17314 remains the same in that the even-numbered double-precision registers
17317 Two additional variants of the o32 ABI are supported to enable
17318 a transition from 32-bit to 64-bit registers. These are FPXX
17319 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17320 The FPXX extension mandates that all code must execute correctly
17321 when run using 32-bit or 64-bit registers. The code can be interlinked
17322 with either FP32 or FP64, but not both.
17323 The FP64A extension is similar to the FP64 extension but forbids the
17324 use of odd-numbered single-precision registers. This can be used
17325 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17326 processors and allows both FP32 and FP64A code to interlink and
17327 run in the same process without changing FPU modes.
17330 @itemx -mno-abicalls
17332 @opindex mno-abicalls
17333 Generate (do not generate) code that is suitable for SVR4-style
17334 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17339 Generate (do not generate) code that is fully position-independent,
17340 and that can therefore be linked into shared libraries. This option
17341 only affects @option{-mabicalls}.
17343 All @option{-mabicalls} code has traditionally been position-independent,
17344 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17345 as an extension, the GNU toolchain allows executables to use absolute
17346 accesses for locally-binding symbols. It can also use shorter GP
17347 initialization sequences and generate direct calls to locally-defined
17348 functions. This mode is selected by @option{-mno-shared}.
17350 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17351 objects that can only be linked by the GNU linker. However, the option
17352 does not affect the ABI of the final executable; it only affects the ABI
17353 of relocatable objects. Using @option{-mno-shared} generally makes
17354 executables both smaller and quicker.
17356 @option{-mshared} is the default.
17362 Assume (do not assume) that the static and dynamic linkers
17363 support PLTs and copy relocations. This option only affects
17364 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17365 has no effect without @option{-msym32}.
17367 You can make @option{-mplt} the default by configuring
17368 GCC with @option{--with-mips-plt}. The default is
17369 @option{-mno-plt} otherwise.
17375 Lift (do not lift) the usual restrictions on the size of the global
17378 GCC normally uses a single instruction to load values from the GOT@.
17379 While this is relatively efficient, it only works if the GOT
17380 is smaller than about 64k. Anything larger causes the linker
17381 to report an error such as:
17383 @cindex relocation truncated to fit (MIPS)
17385 relocation truncated to fit: R_MIPS_GOT16 foobar
17388 If this happens, you should recompile your code with @option{-mxgot}.
17389 This works with very large GOTs, although the code is also
17390 less efficient, since it takes three instructions to fetch the
17391 value of a global symbol.
17393 Note that some linkers can create multiple GOTs. If you have such a
17394 linker, you should only need to use @option{-mxgot} when a single object
17395 file accesses more than 64k's worth of GOT entries. Very few do.
17397 These options have no effect unless GCC is generating position
17402 Assume that general-purpose registers are 32 bits wide.
17406 Assume that general-purpose registers are 64 bits wide.
17410 Assume that floating-point registers are 32 bits wide.
17414 Assume that floating-point registers are 64 bits wide.
17418 Do not assume the width of floating-point registers.
17421 @opindex mhard-float
17422 Use floating-point coprocessor instructions.
17425 @opindex msoft-float
17426 Do not use floating-point coprocessor instructions. Implement
17427 floating-point calculations using library calls instead.
17431 Equivalent to @option{-msoft-float}, but additionally asserts that the
17432 program being compiled does not perform any floating-point operations.
17433 This option is presently supported only by some bare-metal MIPS
17434 configurations, where it may select a special set of libraries
17435 that lack all floating-point support (including, for example, the
17436 floating-point @code{printf} formats).
17437 If code compiled with @option{-mno-float} accidentally contains
17438 floating-point operations, it is likely to suffer a link-time
17439 or run-time failure.
17441 @item -msingle-float
17442 @opindex msingle-float
17443 Assume that the floating-point coprocessor only supports single-precision
17446 @item -mdouble-float
17447 @opindex mdouble-float
17448 Assume that the floating-point coprocessor supports double-precision
17449 operations. This is the default.
17452 @itemx -mno-odd-spreg
17453 @opindex modd-spreg
17454 @opindex mno-odd-spreg
17455 Enable the use of odd-numbered single-precision floating-point registers
17456 for the o32 ABI. This is the default for processors that are known to
17457 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17460 @item -mcompact-branches=never
17461 @itemx -mcompact-branches=optimal
17462 @itemx -mcompact-branches=always
17463 @opindex mcompact-branches=never
17464 @opindex mcompact-branches=optimal
17465 @opindex mcompact-branches=always
17466 These options control which form of branches will be generated. The
17467 default is @option{-mcompact-branches=optimal}.
17469 The @option{-mcompact-branches=never} option ensures that compact branch
17470 instructions will never be generated.
17472 The @option{-mcompact-branches=always} option ensures that a compact
17473 branch instruction will be generated if available. If a compact branch
17474 instruction is not available, a delay slot form of the branch will be
17477 This option is supported from MIPS Release 6 onwards.
17479 The @option{-mcompact-branches=optimal} option will cause a delay slot
17480 branch to be used if one is available in the current ISA and the delay
17481 slot is successfully filled. If the delay slot is not filled, a compact
17482 branch will be chosen if one is available.
17485 @itemx -mabs=legacy
17487 @opindex mabs=legacy
17488 These options control the treatment of the special not-a-number (NaN)
17489 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17490 @code{neg.@i{fmt}} machine instructions.
17492 By default or when @option{-mabs=legacy} is used the legacy
17493 treatment is selected. In this case these instructions are considered
17494 arithmetic and avoided where correct operation is required and the
17495 input operand might be a NaN. A longer sequence of instructions that
17496 manipulate the sign bit of floating-point datum manually is used
17497 instead unless the @option{-ffinite-math-only} option has also been
17500 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17501 this case these instructions are considered non-arithmetic and therefore
17502 operating correctly in all cases, including in particular where the
17503 input operand is a NaN. These instructions are therefore always used
17504 for the respective operations.
17507 @itemx -mnan=legacy
17509 @opindex mnan=legacy
17510 These options control the encoding of the special not-a-number (NaN)
17511 IEEE 754 floating-point data.
17513 The @option{-mnan=legacy} option selects the legacy encoding. In this
17514 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17515 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17516 by the first bit of their trailing significand field being 1.
17518 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17519 this case qNaNs are denoted by the first bit of their trailing
17520 significand field being 1, whereas sNaNs are denoted by the first bit of
17521 their trailing significand field being 0.
17523 The default is @option{-mnan=legacy} unless GCC has been configured with
17524 @option{--with-nan=2008}.
17530 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17531 implement atomic memory built-in functions. When neither option is
17532 specified, GCC uses the instructions if the target architecture
17535 @option{-mllsc} is useful if the runtime environment can emulate the
17536 instructions and @option{-mno-llsc} can be useful when compiling for
17537 nonstandard ISAs. You can make either option the default by
17538 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17539 respectively. @option{--with-llsc} is the default for some
17540 configurations; see the installation documentation for details.
17546 Use (do not use) revision 1 of the MIPS DSP ASE@.
17547 @xref{MIPS DSP Built-in Functions}. This option defines the
17548 preprocessor macro @code{__mips_dsp}. It also defines
17549 @code{__mips_dsp_rev} to 1.
17555 Use (do not use) revision 2 of the MIPS DSP ASE@.
17556 @xref{MIPS DSP Built-in Functions}. This option defines the
17557 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17558 It also defines @code{__mips_dsp_rev} to 2.
17561 @itemx -mno-smartmips
17562 @opindex msmartmips
17563 @opindex mno-smartmips
17564 Use (do not use) the MIPS SmartMIPS ASE.
17566 @item -mpaired-single
17567 @itemx -mno-paired-single
17568 @opindex mpaired-single
17569 @opindex mno-paired-single
17570 Use (do not use) paired-single floating-point instructions.
17571 @xref{MIPS Paired-Single Support}. This option requires
17572 hardware floating-point support to be enabled.
17578 Use (do not use) MIPS Digital Media Extension instructions.
17579 This option can only be used when generating 64-bit code and requires
17580 hardware floating-point support to be enabled.
17585 @opindex mno-mips3d
17586 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17587 The option @option{-mips3d} implies @option{-mpaired-single}.
17590 @itemx -mno-micromips
17591 @opindex mmicromips
17592 @opindex mno-mmicromips
17593 Generate (do not generate) microMIPS code.
17595 MicroMIPS code generation can also be controlled on a per-function basis
17596 by means of @code{micromips} and @code{nomicromips} attributes.
17597 @xref{Function Attributes}, for more information.
17603 Use (do not use) MT Multithreading instructions.
17609 Use (do not use) the MIPS MCU ASE instructions.
17615 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17621 Use (do not use) the MIPS Virtualization Application Specific instructions.
17627 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17631 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17632 an explanation of the default and the way that the pointer size is
17637 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17639 The default size of @code{int}s, @code{long}s and pointers depends on
17640 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17641 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17642 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17643 or the same size as integer registers, whichever is smaller.
17649 Assume (do not assume) that all symbols have 32-bit values, regardless
17650 of the selected ABI@. This option is useful in combination with
17651 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17652 to generate shorter and faster references to symbolic addresses.
17656 Put definitions of externally-visible data in a small data section
17657 if that data is no bigger than @var{num} bytes. GCC can then generate
17658 more efficient accesses to the data; see @option{-mgpopt} for details.
17660 The default @option{-G} option depends on the configuration.
17662 @item -mlocal-sdata
17663 @itemx -mno-local-sdata
17664 @opindex mlocal-sdata
17665 @opindex mno-local-sdata
17666 Extend (do not extend) the @option{-G} behavior to local data too,
17667 such as to static variables in C@. @option{-mlocal-sdata} is the
17668 default for all configurations.
17670 If the linker complains that an application is using too much small data,
17671 you might want to try rebuilding the less performance-critical parts with
17672 @option{-mno-local-sdata}. You might also want to build large
17673 libraries with @option{-mno-local-sdata}, so that the libraries leave
17674 more room for the main program.
17676 @item -mextern-sdata
17677 @itemx -mno-extern-sdata
17678 @opindex mextern-sdata
17679 @opindex mno-extern-sdata
17680 Assume (do not assume) that externally-defined data is in
17681 a small data section if the size of that data is within the @option{-G} limit.
17682 @option{-mextern-sdata} is the default for all configurations.
17684 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17685 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17686 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17687 is placed in a small data section. If @var{Var} is defined by another
17688 module, you must either compile that module with a high-enough
17689 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17690 definition. If @var{Var} is common, you must link the application
17691 with a high-enough @option{-G} setting.
17693 The easiest way of satisfying these restrictions is to compile
17694 and link every module with the same @option{-G} option. However,
17695 you may wish to build a library that supports several different
17696 small data limits. You can do this by compiling the library with
17697 the highest supported @option{-G} setting and additionally using
17698 @option{-mno-extern-sdata} to stop the library from making assumptions
17699 about externally-defined data.
17705 Use (do not use) GP-relative accesses for symbols that are known to be
17706 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17707 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17710 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17711 might not hold the value of @code{_gp}. For example, if the code is
17712 part of a library that might be used in a boot monitor, programs that
17713 call boot monitor routines pass an unknown value in @code{$gp}.
17714 (In such situations, the boot monitor itself is usually compiled
17715 with @option{-G0}.)
17717 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17718 @option{-mno-extern-sdata}.
17720 @item -membedded-data
17721 @itemx -mno-embedded-data
17722 @opindex membedded-data
17723 @opindex mno-embedded-data
17724 Allocate variables to the read-only data section first if possible, then
17725 next in the small data section if possible, otherwise in data. This gives
17726 slightly slower code than the default, but reduces the amount of RAM required
17727 when executing, and thus may be preferred for some embedded systems.
17729 @item -muninit-const-in-rodata
17730 @itemx -mno-uninit-const-in-rodata
17731 @opindex muninit-const-in-rodata
17732 @opindex mno-uninit-const-in-rodata
17733 Put uninitialized @code{const} variables in the read-only data section.
17734 This option is only meaningful in conjunction with @option{-membedded-data}.
17736 @item -mcode-readable=@var{setting}
17737 @opindex mcode-readable
17738 Specify whether GCC may generate code that reads from executable sections.
17739 There are three possible settings:
17742 @item -mcode-readable=yes
17743 Instructions may freely access executable sections. This is the
17746 @item -mcode-readable=pcrel
17747 MIPS16 PC-relative load instructions can access executable sections,
17748 but other instructions must not do so. This option is useful on 4KSc
17749 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17750 It is also useful on processors that can be configured to have a dual
17751 instruction/data SRAM interface and that, like the M4K, automatically
17752 redirect PC-relative loads to the instruction RAM.
17754 @item -mcode-readable=no
17755 Instructions must not access executable sections. This option can be
17756 useful on targets that are configured to have a dual instruction/data
17757 SRAM interface but that (unlike the M4K) do not automatically redirect
17758 PC-relative loads to the instruction RAM.
17761 @item -msplit-addresses
17762 @itemx -mno-split-addresses
17763 @opindex msplit-addresses
17764 @opindex mno-split-addresses
17765 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17766 relocation operators. This option has been superseded by
17767 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17769 @item -mexplicit-relocs
17770 @itemx -mno-explicit-relocs
17771 @opindex mexplicit-relocs
17772 @opindex mno-explicit-relocs
17773 Use (do not use) assembler relocation operators when dealing with symbolic
17774 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17775 is to use assembler macros instead.
17777 @option{-mexplicit-relocs} is the default if GCC was configured
17778 to use an assembler that supports relocation operators.
17780 @item -mcheck-zero-division
17781 @itemx -mno-check-zero-division
17782 @opindex mcheck-zero-division
17783 @opindex mno-check-zero-division
17784 Trap (do not trap) on integer division by zero.
17786 The default is @option{-mcheck-zero-division}.
17788 @item -mdivide-traps
17789 @itemx -mdivide-breaks
17790 @opindex mdivide-traps
17791 @opindex mdivide-breaks
17792 MIPS systems check for division by zero by generating either a
17793 conditional trap or a break instruction. Using traps results in
17794 smaller code, but is only supported on MIPS II and later. Also, some
17795 versions of the Linux kernel have a bug that prevents trap from
17796 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17797 allow conditional traps on architectures that support them and
17798 @option{-mdivide-breaks} to force the use of breaks.
17800 The default is usually @option{-mdivide-traps}, but this can be
17801 overridden at configure time using @option{--with-divide=breaks}.
17802 Divide-by-zero checks can be completely disabled using
17803 @option{-mno-check-zero-division}.
17808 @opindex mno-memcpy
17809 Force (do not force) the use of @code{memcpy} for non-trivial block
17810 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17811 most constant-sized copies.
17814 @itemx -mno-long-calls
17815 @opindex mlong-calls
17816 @opindex mno-long-calls
17817 Disable (do not disable) use of the @code{jal} instruction. Calling
17818 functions using @code{jal} is more efficient but requires the caller
17819 and callee to be in the same 256 megabyte segment.
17821 This option has no effect on abicalls code. The default is
17822 @option{-mno-long-calls}.
17828 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17829 instructions, as provided by the R4650 ISA@.
17835 Enable (disable) use of the @code{madd} and @code{msub} integer
17836 instructions. The default is @option{-mimadd} on architectures
17837 that support @code{madd} and @code{msub} except for the 74k
17838 architecture where it was found to generate slower code.
17841 @itemx -mno-fused-madd
17842 @opindex mfused-madd
17843 @opindex mno-fused-madd
17844 Enable (disable) use of the floating-point multiply-accumulate
17845 instructions, when they are available. The default is
17846 @option{-mfused-madd}.
17848 On the R8000 CPU when multiply-accumulate instructions are used,
17849 the intermediate product is calculated to infinite precision
17850 and is not subject to the FCSR Flush to Zero bit. This may be
17851 undesirable in some circumstances. On other processors the result
17852 is numerically identical to the equivalent computation using
17853 separate multiply, add, subtract and negate instructions.
17857 Tell the MIPS assembler to not run its preprocessor over user
17858 assembler files (with a @samp{.s} suffix) when assembling them.
17863 @opindex mno-fix-24k
17864 Work around the 24K E48 (lost data on stores during refill) errata.
17865 The workarounds are implemented by the assembler rather than by GCC@.
17868 @itemx -mno-fix-r4000
17869 @opindex mfix-r4000
17870 @opindex mno-fix-r4000
17871 Work around certain R4000 CPU errata:
17874 A double-word or a variable shift may give an incorrect result if executed
17875 immediately after starting an integer division.
17877 A double-word or a variable shift may give an incorrect result if executed
17878 while an integer multiplication is in progress.
17880 An integer division may give an incorrect result if started in a delay slot
17881 of a taken branch or a jump.
17885 @itemx -mno-fix-r4400
17886 @opindex mfix-r4400
17887 @opindex mno-fix-r4400
17888 Work around certain R4400 CPU errata:
17891 A double-word or a variable shift may give an incorrect result if executed
17892 immediately after starting an integer division.
17896 @itemx -mno-fix-r10000
17897 @opindex mfix-r10000
17898 @opindex mno-fix-r10000
17899 Work around certain R10000 errata:
17902 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17903 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17906 This option can only be used if the target architecture supports
17907 branch-likely instructions. @option{-mfix-r10000} is the default when
17908 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17912 @itemx -mno-fix-rm7000
17913 @opindex mfix-rm7000
17914 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17915 workarounds are implemented by the assembler rather than by GCC@.
17918 @itemx -mno-fix-vr4120
17919 @opindex mfix-vr4120
17920 Work around certain VR4120 errata:
17923 @code{dmultu} does not always produce the correct result.
17925 @code{div} and @code{ddiv} do not always produce the correct result if one
17926 of the operands is negative.
17928 The workarounds for the division errata rely on special functions in
17929 @file{libgcc.a}. At present, these functions are only provided by
17930 the @code{mips64vr*-elf} configurations.
17932 Other VR4120 errata require a NOP to be inserted between certain pairs of
17933 instructions. These errata are handled by the assembler, not by GCC itself.
17936 @opindex mfix-vr4130
17937 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17938 workarounds are implemented by the assembler rather than by GCC,
17939 although GCC avoids using @code{mflo} and @code{mfhi} if the
17940 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17941 instructions are available instead.
17944 @itemx -mno-fix-sb1
17946 Work around certain SB-1 CPU core errata.
17947 (This flag currently works around the SB-1 revision 2
17948 ``F1'' and ``F2'' floating-point errata.)
17950 @item -mr10k-cache-barrier=@var{setting}
17951 @opindex mr10k-cache-barrier
17952 Specify whether GCC should insert cache barriers to avoid the
17953 side-effects of speculation on R10K processors.
17955 In common with many processors, the R10K tries to predict the outcome
17956 of a conditional branch and speculatively executes instructions from
17957 the ``taken'' branch. It later aborts these instructions if the
17958 predicted outcome is wrong. However, on the R10K, even aborted
17959 instructions can have side effects.
17961 This problem only affects kernel stores and, depending on the system,
17962 kernel loads. As an example, a speculatively-executed store may load
17963 the target memory into cache and mark the cache line as dirty, even if
17964 the store itself is later aborted. If a DMA operation writes to the
17965 same area of memory before the ``dirty'' line is flushed, the cached
17966 data overwrites the DMA-ed data. See the R10K processor manual
17967 for a full description, including other potential problems.
17969 One workaround is to insert cache barrier instructions before every memory
17970 access that might be speculatively executed and that might have side
17971 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17972 controls GCC's implementation of this workaround. It assumes that
17973 aborted accesses to any byte in the following regions does not have
17978 the memory occupied by the current function's stack frame;
17981 the memory occupied by an incoming stack argument;
17984 the memory occupied by an object with a link-time-constant address.
17987 It is the kernel's responsibility to ensure that speculative
17988 accesses to these regions are indeed safe.
17990 If the input program contains a function declaration such as:
17996 then the implementation of @code{foo} must allow @code{j foo} and
17997 @code{jal foo} to be executed speculatively. GCC honors this
17998 restriction for functions it compiles itself. It expects non-GCC
17999 functions (such as hand-written assembly code) to do the same.
18001 The option has three forms:
18004 @item -mr10k-cache-barrier=load-store
18005 Insert a cache barrier before a load or store that might be
18006 speculatively executed and that might have side effects even
18009 @item -mr10k-cache-barrier=store
18010 Insert a cache barrier before a store that might be speculatively
18011 executed and that might have side effects even if aborted.
18013 @item -mr10k-cache-barrier=none
18014 Disable the insertion of cache barriers. This is the default setting.
18017 @item -mflush-func=@var{func}
18018 @itemx -mno-flush-func
18019 @opindex mflush-func
18020 Specifies the function to call to flush the I and D caches, or to not
18021 call any such function. If called, the function must take the same
18022 arguments as the common @code{_flush_func}, that is, the address of the
18023 memory range for which the cache is being flushed, the size of the
18024 memory range, and the number 3 (to flush both caches). The default
18025 depends on the target GCC was configured for, but commonly is either
18026 @code{_flush_func} or @code{__cpu_flush}.
18028 @item mbranch-cost=@var{num}
18029 @opindex mbranch-cost
18030 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18031 This cost is only a heuristic and is not guaranteed to produce
18032 consistent results across releases. A zero cost redundantly selects
18033 the default, which is based on the @option{-mtune} setting.
18035 @item -mbranch-likely
18036 @itemx -mno-branch-likely
18037 @opindex mbranch-likely
18038 @opindex mno-branch-likely
18039 Enable or disable use of Branch Likely instructions, regardless of the
18040 default for the selected architecture. By default, Branch Likely
18041 instructions may be generated if they are supported by the selected
18042 architecture. An exception is for the MIPS32 and MIPS64 architectures
18043 and processors that implement those architectures; for those, Branch
18044 Likely instructions are not be generated by default because the MIPS32
18045 and MIPS64 architectures specifically deprecate their use.
18047 @item -mfp-exceptions
18048 @itemx -mno-fp-exceptions
18049 @opindex mfp-exceptions
18050 Specifies whether FP exceptions are enabled. This affects how
18051 FP instructions are scheduled for some processors.
18052 The default is that FP exceptions are
18055 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18056 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18059 @item -mvr4130-align
18060 @itemx -mno-vr4130-align
18061 @opindex mvr4130-align
18062 The VR4130 pipeline is two-way superscalar, but can only issue two
18063 instructions together if the first one is 8-byte aligned. When this
18064 option is enabled, GCC aligns pairs of instructions that it
18065 thinks should execute in parallel.
18067 This option only has an effect when optimizing for the VR4130.
18068 It normally makes code faster, but at the expense of making it bigger.
18069 It is enabled by default at optimization level @option{-O3}.
18074 Enable (disable) generation of @code{synci} instructions on
18075 architectures that support it. The @code{synci} instructions (if
18076 enabled) are generated when @code{__builtin___clear_cache} is
18079 This option defaults to @option{-mno-synci}, but the default can be
18080 overridden by configuring GCC with @option{--with-synci}.
18082 When compiling code for single processor systems, it is generally safe
18083 to use @code{synci}. However, on many multi-core (SMP) systems, it
18084 does not invalidate the instruction caches on all cores and may lead
18085 to undefined behavior.
18087 @item -mrelax-pic-calls
18088 @itemx -mno-relax-pic-calls
18089 @opindex mrelax-pic-calls
18090 Try to turn PIC calls that are normally dispatched via register
18091 @code{$25} into direct calls. This is only possible if the linker can
18092 resolve the destination at link-time and if the destination is within
18093 range for a direct call.
18095 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18096 an assembler and a linker that support the @code{.reloc} assembly
18097 directive and @option{-mexplicit-relocs} is in effect. With
18098 @option{-mno-explicit-relocs}, this optimization can be performed by the
18099 assembler and the linker alone without help from the compiler.
18101 @item -mmcount-ra-address
18102 @itemx -mno-mcount-ra-address
18103 @opindex mmcount-ra-address
18104 @opindex mno-mcount-ra-address
18105 Emit (do not emit) code that allows @code{_mcount} to modify the
18106 calling function's return address. When enabled, this option extends
18107 the usual @code{_mcount} interface with a new @var{ra-address}
18108 parameter, which has type @code{intptr_t *} and is passed in register
18109 @code{$12}. @code{_mcount} can then modify the return address by
18110 doing both of the following:
18113 Returning the new address in register @code{$31}.
18115 Storing the new address in @code{*@var{ra-address}},
18116 if @var{ra-address} is nonnull.
18119 The default is @option{-mno-mcount-ra-address}.
18121 @item -mframe-header-opt
18122 @itemx -mno-frame-header-opt
18123 @opindex mframe-header-opt
18124 Enable (disable) frame header optimization in the o32 ABI. When using the
18125 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18126 function to write out register arguments. When enabled, this optimization
18127 will suppress the allocation of the frame header if it can be determined that
18130 This optimization is off by default at all optimization levels.
18135 @subsection MMIX Options
18136 @cindex MMIX Options
18138 These options are defined for the MMIX:
18142 @itemx -mno-libfuncs
18144 @opindex mno-libfuncs
18145 Specify that intrinsic library functions are being compiled, passing all
18146 values in registers, no matter the size.
18149 @itemx -mno-epsilon
18151 @opindex mno-epsilon
18152 Generate floating-point comparison instructions that compare with respect
18153 to the @code{rE} epsilon register.
18155 @item -mabi=mmixware
18157 @opindex mabi=mmixware
18159 Generate code that passes function parameters and return values that (in
18160 the called function) are seen as registers @code{$0} and up, as opposed to
18161 the GNU ABI which uses global registers @code{$231} and up.
18163 @item -mzero-extend
18164 @itemx -mno-zero-extend
18165 @opindex mzero-extend
18166 @opindex mno-zero-extend
18167 When reading data from memory in sizes shorter than 64 bits, use (do not
18168 use) zero-extending load instructions by default, rather than
18169 sign-extending ones.
18172 @itemx -mno-knuthdiv
18174 @opindex mno-knuthdiv
18175 Make the result of a division yielding a remainder have the same sign as
18176 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18177 remainder follows the sign of the dividend. Both methods are
18178 arithmetically valid, the latter being almost exclusively used.
18180 @item -mtoplevel-symbols
18181 @itemx -mno-toplevel-symbols
18182 @opindex mtoplevel-symbols
18183 @opindex mno-toplevel-symbols
18184 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18185 code can be used with the @code{PREFIX} assembly directive.
18189 Generate an executable in the ELF format, rather than the default
18190 @samp{mmo} format used by the @command{mmix} simulator.
18192 @item -mbranch-predict
18193 @itemx -mno-branch-predict
18194 @opindex mbranch-predict
18195 @opindex mno-branch-predict
18196 Use (do not use) the probable-branch instructions, when static branch
18197 prediction indicates a probable branch.
18199 @item -mbase-addresses
18200 @itemx -mno-base-addresses
18201 @opindex mbase-addresses
18202 @opindex mno-base-addresses
18203 Generate (do not generate) code that uses @emph{base addresses}. Using a
18204 base address automatically generates a request (handled by the assembler
18205 and the linker) for a constant to be set up in a global register. The
18206 register is used for one or more base address requests within the range 0
18207 to 255 from the value held in the register. The generally leads to short
18208 and fast code, but the number of different data items that can be
18209 addressed is limited. This means that a program that uses lots of static
18210 data may require @option{-mno-base-addresses}.
18212 @item -msingle-exit
18213 @itemx -mno-single-exit
18214 @opindex msingle-exit
18215 @opindex mno-single-exit
18216 Force (do not force) generated code to have a single exit point in each
18220 @node MN10300 Options
18221 @subsection MN10300 Options
18222 @cindex MN10300 options
18224 These @option{-m} options are defined for Matsushita MN10300 architectures:
18229 Generate code to avoid bugs in the multiply instructions for the MN10300
18230 processors. This is the default.
18232 @item -mno-mult-bug
18233 @opindex mno-mult-bug
18234 Do not generate code to avoid bugs in the multiply instructions for the
18235 MN10300 processors.
18239 Generate code using features specific to the AM33 processor.
18243 Do not generate code using features specific to the AM33 processor. This
18248 Generate code using features specific to the AM33/2.0 processor.
18252 Generate code using features specific to the AM34 processor.
18254 @item -mtune=@var{cpu-type}
18256 Use the timing characteristics of the indicated CPU type when
18257 scheduling instructions. This does not change the targeted processor
18258 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18259 @samp{am33-2} or @samp{am34}.
18261 @item -mreturn-pointer-on-d0
18262 @opindex mreturn-pointer-on-d0
18263 When generating a function that returns a pointer, return the pointer
18264 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18265 only in @code{a0}, and attempts to call such functions without a prototype
18266 result in errors. Note that this option is on by default; use
18267 @option{-mno-return-pointer-on-d0} to disable it.
18271 Do not link in the C run-time initialization object file.
18275 Indicate to the linker that it should perform a relaxation optimization pass
18276 to shorten branches, calls and absolute memory addresses. This option only
18277 has an effect when used on the command line for the final link step.
18279 This option makes symbolic debugging impossible.
18283 Allow the compiler to generate @emph{Long Instruction Word}
18284 instructions if the target is the @samp{AM33} or later. This is the
18285 default. This option defines the preprocessor macro @code{__LIW__}.
18289 Do not allow the compiler to generate @emph{Long Instruction Word}
18290 instructions. This option defines the preprocessor macro
18295 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18296 instructions if the target is the @samp{AM33} or later. This is the
18297 default. This option defines the preprocessor macro @code{__SETLB__}.
18301 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18302 instructions. This option defines the preprocessor macro
18303 @code{__NO_SETLB__}.
18307 @node Moxie Options
18308 @subsection Moxie Options
18309 @cindex Moxie Options
18315 Generate big-endian code. This is the default for @samp{moxie-*-*}
18320 Generate little-endian code.
18324 Generate mul.x and umul.x instructions. This is the default for
18325 @samp{moxiebox-*-*} configurations.
18329 Do not link in the C run-time initialization object file.
18333 @node MSP430 Options
18334 @subsection MSP430 Options
18335 @cindex MSP430 Options
18337 These options are defined for the MSP430:
18343 Force assembly output to always use hex constants. Normally such
18344 constants are signed decimals, but this option is available for
18345 testsuite and/or aesthetic purposes.
18349 Select the MCU to target. This is used to create a C preprocessor
18350 symbol based upon the MCU name, converted to upper case and pre- and
18351 post-fixed with @samp{__}. This in turn is used by the
18352 @file{msp430.h} header file to select an MCU-specific supplementary
18355 The option also sets the ISA to use. If the MCU name is one that is
18356 known to only support the 430 ISA then that is selected, otherwise the
18357 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18358 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18359 name selects the 430X ISA.
18361 In addition an MCU-specific linker script is added to the linker
18362 command line. The script's name is the name of the MCU with
18363 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18364 command line defines the C preprocessor symbol @code{__XXX__} and
18365 cause the linker to search for a script called @file{xxx.ld}.
18367 This option is also passed on to the assembler.
18371 Specifies the ISA to use. Accepted values are @samp{msp430},
18372 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18373 @option{-mmcu=} option should be used to select the ISA.
18377 Link to the simulator runtime libraries and linker script. Overrides
18378 any scripts that would be selected by the @option{-mmcu=} option.
18382 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18386 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18390 This option is passed to the assembler and linker, and allows the
18391 linker to perform certain optimizations that cannot be done until
18396 Describes the type of hardware multiply supported by the target.
18397 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18398 for the original 16-bit-only multiply supported by early MCUs.
18399 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18400 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18401 A value of @samp{auto} can also be given. This tells GCC to deduce
18402 the hardware multiply support based upon the MCU name provided by the
18403 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18404 @samp{32bit} hardware multiply support is assumed. If the MCU name is
18405 not recognised then no hardware multiply support is assumed.
18406 @code{auto} is the default setting.
18408 Hardware multiplies are normally performed by calling a library
18409 routine. This saves space in the generated code. When compiling at
18410 @option{-O3} or higher however the hardware multiplier is invoked
18411 inline. This makes for bigger, but faster code.
18413 The hardware multiply routines disable interrupts whilst running and
18414 restore the previous interrupt state when they finish. This makes
18415 them safe to use inside interrupt handlers as well as in normal code.
18419 Enable the use of a minimum runtime environment - no static
18420 initializers or constructors. This is intended for memory-constrained
18421 devices. The compiler includes special symbols in some objects
18422 that tell the linker and runtime which code fragments are required.
18424 @item -mcode-region=
18425 @itemx -mdata-region=
18426 @opindex mcode-region
18427 @opindex mdata-region
18428 These options tell the compiler where to place functions and data that
18429 do not have one of the @code{lower}, @code{upper}, @code{either} or
18430 @code{section} attributes. Possible values are @code{lower},
18431 @code{upper}, @code{either} or @code{any}. The first three behave
18432 like the corresponding attribute. The fourth possible value -
18433 @code{any} - is the default. It leaves placement entirely up to the
18434 linker script and how it assigns the standard sections (.text, .data
18435 etc) to the memory regions.
18439 @node NDS32 Options
18440 @subsection NDS32 Options
18441 @cindex NDS32 Options
18443 These options are defined for NDS32 implementations:
18448 @opindex mbig-endian
18449 Generate code in big-endian mode.
18451 @item -mlittle-endian
18452 @opindex mlittle-endian
18453 Generate code in little-endian mode.
18455 @item -mreduced-regs
18456 @opindex mreduced-regs
18457 Use reduced-set registers for register allocation.
18460 @opindex mfull-regs
18461 Use full-set registers for register allocation.
18465 Generate conditional move instructions.
18469 Do not generate conditional move instructions.
18473 Generate performance extension instructions.
18475 @item -mno-perf-ext
18476 @opindex mno-perf-ext
18477 Do not generate performance extension instructions.
18481 Generate v3 push25/pop25 instructions.
18484 @opindex mno-v3push
18485 Do not generate v3 push25/pop25 instructions.
18489 Generate 16-bit instructions.
18492 @opindex mno-16-bit
18493 Do not generate 16-bit instructions.
18495 @item -misr-vector-size=@var{num}
18496 @opindex misr-vector-size
18497 Specify the size of each interrupt vector, which must be 4 or 16.
18499 @item -mcache-block-size=@var{num}
18500 @opindex mcache-block-size
18501 Specify the size of each cache block,
18502 which must be a power of 2 between 4 and 512.
18504 @item -march=@var{arch}
18506 Specify the name of the target architecture.
18508 @item -mcmodel=@var{code-model}
18510 Set the code model to one of
18513 All the data and read-only data segments must be within 512KB addressing space.
18514 The text segment must be within 16MB addressing space.
18515 @item @samp{medium}
18516 The data segment must be within 512KB while the read-only data segment can be
18517 within 4GB addressing space. The text segment should be still within 16MB
18520 All the text and data segments can be within 4GB addressing space.
18524 @opindex mctor-dtor
18525 Enable constructor/destructor feature.
18529 Guide linker to relax instructions.
18533 @node Nios II Options
18534 @subsection Nios II Options
18535 @cindex Nios II options
18536 @cindex Altera Nios II options
18538 These are the options defined for the Altera Nios II processor.
18544 @cindex smaller data references
18545 Put global and static objects less than or equal to @var{num} bytes
18546 into the small data or BSS sections instead of the normal data or BSS
18547 sections. The default value of @var{num} is 8.
18549 @item -mgpopt=@var{option}
18554 Generate (do not generate) GP-relative accesses. The following
18555 @var{option} names are recognized:
18560 Do not generate GP-relative accesses.
18563 Generate GP-relative accesses for small data objects that are not
18564 external, weak, or uninitialized common symbols.
18565 Also use GP-relative addressing for objects that
18566 have been explicitly placed in a small data section via a @code{section}
18570 As for @samp{local}, but also generate GP-relative accesses for
18571 small data objects that are external, weak, or common. If you use this option,
18572 you must ensure that all parts of your program (including libraries) are
18573 compiled with the same @option{-G} setting.
18576 Generate GP-relative accesses for all data objects in the program. If you
18577 use this option, the entire data and BSS segments
18578 of your program must fit in 64K of memory and you must use an appropriate
18579 linker script to allocate them within the addressible range of the
18583 Generate GP-relative addresses for function pointers as well as data
18584 pointers. If you use this option, the entire text, data, and BSS segments
18585 of your program must fit in 64K of memory and you must use an appropriate
18586 linker script to allocate them within the addressible range of the
18591 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18592 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18594 The default is @option{-mgpopt} except when @option{-fpic} or
18595 @option{-fPIC} is specified to generate position-independent code.
18596 Note that the Nios II ABI does not permit GP-relative accesses from
18599 You may need to specify @option{-mno-gpopt} explicitly when building
18600 programs that include large amounts of small data, including large
18601 GOT data sections. In this case, the 16-bit offset for GP-relative
18602 addressing may not be large enough to allow access to the entire
18603 small data section.
18609 Generate little-endian (default) or big-endian (experimental) code,
18612 @item -march=@var{arch}
18614 This specifies the name of the target Nios II architecture. GCC uses this
18615 name to determine what kind of instructions it can emit when generating
18616 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
18618 The preprocessor macro @code{__nios2_arch__} is available to programs,
18619 with value 1 or 2, indicating the targeted ISA level.
18621 @item -mbypass-cache
18622 @itemx -mno-bypass-cache
18623 @opindex mno-bypass-cache
18624 @opindex mbypass-cache
18625 Force all load and store instructions to always bypass cache by
18626 using I/O variants of the instructions. The default is not to
18629 @item -mno-cache-volatile
18630 @itemx -mcache-volatile
18631 @opindex mcache-volatile
18632 @opindex mno-cache-volatile
18633 Volatile memory access bypass the cache using the I/O variants of
18634 the load and store instructions. The default is not to bypass the cache.
18636 @item -mno-fast-sw-div
18637 @itemx -mfast-sw-div
18638 @opindex mno-fast-sw-div
18639 @opindex mfast-sw-div
18640 Do not use table-based fast divide for small numbers. The default
18641 is to use the fast divide at @option{-O3} and above.
18645 @itemx -mno-hw-mulx
18649 @opindex mno-hw-mul
18651 @opindex mno-hw-mulx
18653 @opindex mno-hw-div
18655 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18656 instructions by the compiler. The default is to emit @code{mul}
18657 and not emit @code{div} and @code{mulx}.
18663 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
18664 CDX (code density) instructions. Enabling these instructions also
18665 requires @option{-march=r2}. Since these instructions are optional
18666 extensions to the R2 architecture, the default is not to emit them.
18668 @item -mcustom-@var{insn}=@var{N}
18669 @itemx -mno-custom-@var{insn}
18670 @opindex mcustom-@var{insn}
18671 @opindex mno-custom-@var{insn}
18672 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18673 custom instruction with encoding @var{N} when generating code that uses
18674 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18675 instruction 253 for single-precision floating-point add operations instead
18676 of the default behavior of using a library call.
18678 The following values of @var{insn} are supported. Except as otherwise
18679 noted, floating-point operations are expected to be implemented with
18680 normal IEEE 754 semantics and correspond directly to the C operators or the
18681 equivalent GCC built-in functions (@pxref{Other Builtins}).
18683 Single-precision floating point:
18686 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18687 Binary arithmetic operations.
18693 Unary absolute value.
18695 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18696 Comparison operations.
18698 @item @samp{fmins}, @samp{fmaxs}
18699 Floating-point minimum and maximum. These instructions are only
18700 generated if @option{-ffinite-math-only} is specified.
18702 @item @samp{fsqrts}
18703 Unary square root operation.
18705 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18706 Floating-point trigonometric and exponential functions. These instructions
18707 are only generated if @option{-funsafe-math-optimizations} is also specified.
18711 Double-precision floating point:
18714 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18715 Binary arithmetic operations.
18721 Unary absolute value.
18723 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18724 Comparison operations.
18726 @item @samp{fmind}, @samp{fmaxd}
18727 Double-precision minimum and maximum. These instructions are only
18728 generated if @option{-ffinite-math-only} is specified.
18730 @item @samp{fsqrtd}
18731 Unary square root operation.
18733 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18734 Double-precision trigonometric and exponential functions. These instructions
18735 are only generated if @option{-funsafe-math-optimizations} is also specified.
18741 @item @samp{fextsd}
18742 Conversion from single precision to double precision.
18744 @item @samp{ftruncds}
18745 Conversion from double precision to single precision.
18747 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18748 Conversion from floating point to signed or unsigned integer types, with
18749 truncation towards zero.
18752 Conversion from single-precision floating point to signed integer,
18753 rounding to the nearest integer and ties away from zero.
18754 This corresponds to the @code{__builtin_lroundf} function when
18755 @option{-fno-math-errno} is used.
18757 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18758 Conversion from signed or unsigned integer types to floating-point types.
18762 In addition, all of the following transfer instructions for internal
18763 registers X and Y must be provided to use any of the double-precision
18764 floating-point instructions. Custom instructions taking two
18765 double-precision source operands expect the first operand in the
18766 64-bit register X. The other operand (or only operand of a unary
18767 operation) is given to the custom arithmetic instruction with the
18768 least significant half in source register @var{src1} and the most
18769 significant half in @var{src2}. A custom instruction that returns a
18770 double-precision result returns the most significant 32 bits in the
18771 destination register and the other half in 32-bit register Y.
18772 GCC automatically generates the necessary code sequences to write
18773 register X and/or read register Y when double-precision floating-point
18774 instructions are used.
18779 Write @var{src1} into the least significant half of X and @var{src2} into
18780 the most significant half of X.
18783 Write @var{src1} into Y.
18785 @item @samp{frdxhi}, @samp{frdxlo}
18786 Read the most or least (respectively) significant half of X and store it in
18790 Read the value of Y and store it into @var{dest}.
18793 Note that you can gain more local control over generation of Nios II custom
18794 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18795 and @code{target("no-custom-@var{insn}")} function attributes
18796 (@pxref{Function Attributes})
18797 or pragmas (@pxref{Function Specific Option Pragmas}).
18799 @item -mcustom-fpu-cfg=@var{name}
18800 @opindex mcustom-fpu-cfg
18802 This option enables a predefined, named set of custom instruction encodings
18803 (see @option{-mcustom-@var{insn}} above).
18804 Currently, the following sets are defined:
18806 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18807 @gccoptlist{-mcustom-fmuls=252 @gol
18808 -mcustom-fadds=253 @gol
18809 -mcustom-fsubs=254 @gol
18810 -fsingle-precision-constant}
18812 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18813 @gccoptlist{-mcustom-fmuls=252 @gol
18814 -mcustom-fadds=253 @gol
18815 -mcustom-fsubs=254 @gol
18816 -mcustom-fdivs=255 @gol
18817 -fsingle-precision-constant}
18819 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18820 @gccoptlist{-mcustom-floatus=243 @gol
18821 -mcustom-fixsi=244 @gol
18822 -mcustom-floatis=245 @gol
18823 -mcustom-fcmpgts=246 @gol
18824 -mcustom-fcmples=249 @gol
18825 -mcustom-fcmpeqs=250 @gol
18826 -mcustom-fcmpnes=251 @gol
18827 -mcustom-fmuls=252 @gol
18828 -mcustom-fadds=253 @gol
18829 -mcustom-fsubs=254 @gol
18830 -mcustom-fdivs=255 @gol
18831 -fsingle-precision-constant}
18833 Custom instruction assignments given by individual
18834 @option{-mcustom-@var{insn}=} options override those given by
18835 @option{-mcustom-fpu-cfg=}, regardless of the
18836 order of the options on the command line.
18838 Note that you can gain more local control over selection of a FPU
18839 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18840 function attribute (@pxref{Function Attributes})
18841 or pragma (@pxref{Function Specific Option Pragmas}).
18845 These additional @samp{-m} options are available for the Altera Nios II
18846 ELF (bare-metal) target:
18852 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18853 startup and termination code, and is typically used in conjunction with
18854 @option{-msys-crt0=} to specify the location of the alternate startup code
18855 provided by the HAL BSP.
18859 Link with a limited version of the C library, @option{-lsmallc}, rather than
18862 @item -msys-crt0=@var{startfile}
18864 @var{startfile} is the file name of the startfile (crt0) to use
18865 when linking. This option is only useful in conjunction with @option{-mhal}.
18867 @item -msys-lib=@var{systemlib}
18869 @var{systemlib} is the library name of the library that provides
18870 low-level system calls required by the C library,
18871 e.g. @code{read} and @code{write}.
18872 This option is typically used to link with a library provided by a HAL BSP.
18876 @node Nvidia PTX Options
18877 @subsection Nvidia PTX Options
18878 @cindex Nvidia PTX options
18879 @cindex nvptx options
18881 These options are defined for Nvidia PTX:
18889 Generate code for 32-bit or 64-bit ABI.
18892 @opindex mmainkernel
18893 Link in code for a __main kernel. This is for stand-alone instead of
18894 offloading execution.
18898 @node PDP-11 Options
18899 @subsection PDP-11 Options
18900 @cindex PDP-11 Options
18902 These options are defined for the PDP-11:
18907 Use hardware FPP floating point. This is the default. (FIS floating
18908 point on the PDP-11/40 is not supported.)
18911 @opindex msoft-float
18912 Do not use hardware floating point.
18916 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18920 Return floating-point results in memory. This is the default.
18924 Generate code for a PDP-11/40.
18928 Generate code for a PDP-11/45. This is the default.
18932 Generate code for a PDP-11/10.
18934 @item -mbcopy-builtin
18935 @opindex mbcopy-builtin
18936 Use inline @code{movmemhi} patterns for copying memory. This is the
18941 Do not use inline @code{movmemhi} patterns for copying memory.
18947 Use 16-bit @code{int}. This is the default.
18953 Use 32-bit @code{int}.
18956 @itemx -mno-float32
18958 @opindex mno-float32
18959 Use 64-bit @code{float}. This is the default.
18962 @itemx -mno-float64
18964 @opindex mno-float64
18965 Use 32-bit @code{float}.
18969 Use @code{abshi2} pattern. This is the default.
18973 Do not use @code{abshi2} pattern.
18975 @item -mbranch-expensive
18976 @opindex mbranch-expensive
18977 Pretend that branches are expensive. This is for experimenting with
18978 code generation only.
18980 @item -mbranch-cheap
18981 @opindex mbranch-cheap
18982 Do not pretend that branches are expensive. This is the default.
18986 Use Unix assembler syntax. This is the default when configured for
18987 @samp{pdp11-*-bsd}.
18991 Use DEC assembler syntax. This is the default when configured for any
18992 PDP-11 target other than @samp{pdp11-*-bsd}.
18995 @node picoChip Options
18996 @subsection picoChip Options
18997 @cindex picoChip options
18999 These @samp{-m} options are defined for picoChip implementations:
19003 @item -mae=@var{ae_type}
19005 Set the instruction set, register set, and instruction scheduling
19006 parameters for array element type @var{ae_type}. Supported values
19007 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19009 @option{-mae=ANY} selects a completely generic AE type. Code
19010 generated with this option runs on any of the other AE types. The
19011 code is not as efficient as it would be if compiled for a specific
19012 AE type, and some types of operation (e.g., multiplication) do not
19013 work properly on all types of AE.
19015 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19016 for compiled code, and is the default.
19018 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19019 option may suffer from poor performance of byte (char) manipulation,
19020 since the DSP AE does not provide hardware support for byte load/stores.
19022 @item -msymbol-as-address
19023 Enable the compiler to directly use a symbol name as an address in a
19024 load/store instruction, without first loading it into a
19025 register. Typically, the use of this option generates larger
19026 programs, which run faster than when the option isn't used. However, the
19027 results vary from program to program, so it is left as a user option,
19028 rather than being permanently enabled.
19030 @item -mno-inefficient-warnings
19031 Disables warnings about the generation of inefficient code. These
19032 warnings can be generated, for example, when compiling code that
19033 performs byte-level memory operations on the MAC AE type. The MAC AE has
19034 no hardware support for byte-level memory operations, so all byte
19035 load/stores must be synthesized from word load/store operations. This is
19036 inefficient and a warning is generated to indicate
19037 that you should rewrite the code to avoid byte operations, or to target
19038 an AE type that has the necessary hardware support. This option disables
19043 @node PowerPC Options
19044 @subsection PowerPC Options
19045 @cindex PowerPC options
19047 These are listed under @xref{RS/6000 and PowerPC Options}.
19050 @subsection RL78 Options
19051 @cindex RL78 Options
19057 Links in additional target libraries to support operation within a
19066 Specifies the type of hardware multiplication and division support to
19067 be used. The simplest is @code{none}, which uses software for both
19068 multiplication and division. This is the default. The @code{g13}
19069 value is for the hardware multiply/divide peripheral found on the
19070 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19071 the multiplication and division instructions supported by the RL78/G14
19072 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19073 the value @code{mg10} is an alias for @code{none}.
19075 In addition a C preprocessor macro is defined, based upon the setting
19076 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19077 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19084 Specifies the RL78 core to target. The default is the G14 core, also
19085 known as an S3 core or just RL78. The G13 or S2 core does not have
19086 multiply or divide instructions, instead it uses a hardware peripheral
19087 for these operations. The G10 or S1 core does not have register
19088 banks, so it uses a different calling convention.
19090 If this option is set it also selects the type of hardware multiply
19091 support to use, unless this is overridden by an explicit
19092 @option{-mmul=none} option on the command line. Thus specifying
19093 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19094 peripheral and specifying @option{-mcpu=g10} disables the use of
19095 hardware multipications altogether.
19097 Note, although the RL78/G14 core is the default target, specifying
19098 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19099 change the behaviour of the toolchain since it also enables G14
19100 hardware multiply support. If these options are not specified on the
19101 command line then software multiplication routines will be used even
19102 though the code targets the RL78 core. This is for backwards
19103 compatibility with older toolchains which did not have hardware
19104 multiply and divide support.
19106 In addition a C preprocessor macro is defined, based upon the setting
19107 of this option. Possible values are: @code{__RL78_G10__},
19108 @code{__RL78_G13__} or @code{__RL78_G14__}.
19118 These are aliases for the corresponding @option{-mcpu=} option. They
19119 are provided for backwards compatibility.
19123 Allow the compiler to use all of the available registers. By default
19124 registers @code{r24..r31} are reserved for use in interrupt handlers.
19125 With this option enabled these registers can be used in ordinary
19128 @item -m64bit-doubles
19129 @itemx -m32bit-doubles
19130 @opindex m64bit-doubles
19131 @opindex m32bit-doubles
19132 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19133 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19134 @option{-m32bit-doubles}.
19138 @node RS/6000 and PowerPC Options
19139 @subsection IBM RS/6000 and PowerPC Options
19140 @cindex RS/6000 and PowerPC Options
19141 @cindex IBM RS/6000 and PowerPC Options
19143 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19145 @item -mpowerpc-gpopt
19146 @itemx -mno-powerpc-gpopt
19147 @itemx -mpowerpc-gfxopt
19148 @itemx -mno-powerpc-gfxopt
19151 @itemx -mno-powerpc64
19155 @itemx -mno-popcntb
19157 @itemx -mno-popcntd
19166 @itemx -mno-hard-dfp
19167 @opindex mpowerpc-gpopt
19168 @opindex mno-powerpc-gpopt
19169 @opindex mpowerpc-gfxopt
19170 @opindex mno-powerpc-gfxopt
19171 @opindex mpowerpc64
19172 @opindex mno-powerpc64
19176 @opindex mno-popcntb
19178 @opindex mno-popcntd
19184 @opindex mno-mfpgpr
19186 @opindex mno-hard-dfp
19187 You use these options to specify which instructions are available on the
19188 processor you are using. The default value of these options is
19189 determined when configuring GCC@. Specifying the
19190 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19191 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19192 rather than the options listed above.
19194 Specifying @option{-mpowerpc-gpopt} allows
19195 GCC to use the optional PowerPC architecture instructions in the
19196 General Purpose group, including floating-point square root. Specifying
19197 @option{-mpowerpc-gfxopt} allows GCC to
19198 use the optional PowerPC architecture instructions in the Graphics
19199 group, including floating-point select.
19201 The @option{-mmfcrf} option allows GCC to generate the move from
19202 condition register field instruction implemented on the POWER4
19203 processor and other processors that support the PowerPC V2.01
19205 The @option{-mpopcntb} option allows GCC to generate the popcount and
19206 double-precision FP reciprocal estimate instruction implemented on the
19207 POWER5 processor and other processors that support the PowerPC V2.02
19209 The @option{-mpopcntd} option allows GCC to generate the popcount
19210 instruction implemented on the POWER7 processor and other processors
19211 that support the PowerPC V2.06 architecture.
19212 The @option{-mfprnd} option allows GCC to generate the FP round to
19213 integer instructions implemented on the POWER5+ processor and other
19214 processors that support the PowerPC V2.03 architecture.
19215 The @option{-mcmpb} option allows GCC to generate the compare bytes
19216 instruction implemented on the POWER6 processor and other processors
19217 that support the PowerPC V2.05 architecture.
19218 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19219 general-purpose register instructions implemented on the POWER6X
19220 processor and other processors that support the extended PowerPC V2.05
19222 The @option{-mhard-dfp} option allows GCC to generate the decimal
19223 floating-point instructions implemented on some POWER processors.
19225 The @option{-mpowerpc64} option allows GCC to generate the additional
19226 64-bit instructions that are found in the full PowerPC64 architecture
19227 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19228 @option{-mno-powerpc64}.
19230 @item -mcpu=@var{cpu_type}
19232 Set architecture type, register usage, and
19233 instruction scheduling parameters for machine type @var{cpu_type}.
19234 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19235 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19236 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19237 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19238 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19239 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19240 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19241 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19242 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19243 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19244 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
19246 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19247 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19248 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19249 architecture machine types, with an appropriate, generic processor
19250 model assumed for scheduling purposes.
19252 The other options specify a specific processor. Code generated under
19253 those options runs best on that processor, and may not run at all on
19256 The @option{-mcpu} options automatically enable or disable the
19259 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19260 -mpopcntb -mpopcntd -mpowerpc64 @gol
19261 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19262 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19263 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19264 -mquad-memory -mquad-memory-atomic}
19266 The particular options set for any particular CPU varies between
19267 compiler versions, depending on what setting seems to produce optimal
19268 code for that CPU; it doesn't necessarily reflect the actual hardware's
19269 capabilities. If you wish to set an individual option to a particular
19270 value, you may specify it after the @option{-mcpu} option, like
19271 @option{-mcpu=970 -mno-altivec}.
19273 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19274 not enabled or disabled by the @option{-mcpu} option at present because
19275 AIX does not have full support for these options. You may still
19276 enable or disable them individually if you're sure it'll work in your
19279 @item -mtune=@var{cpu_type}
19281 Set the instruction scheduling parameters for machine type
19282 @var{cpu_type}, but do not set the architecture type or register usage,
19283 as @option{-mcpu=@var{cpu_type}} does. The same
19284 values for @var{cpu_type} are used for @option{-mtune} as for
19285 @option{-mcpu}. If both are specified, the code generated uses the
19286 architecture and registers set by @option{-mcpu}, but the
19287 scheduling parameters set by @option{-mtune}.
19289 @item -mcmodel=small
19290 @opindex mcmodel=small
19291 Generate PowerPC64 code for the small model: The TOC is limited to
19294 @item -mcmodel=medium
19295 @opindex mcmodel=medium
19296 Generate PowerPC64 code for the medium model: The TOC and other static
19297 data may be up to a total of 4G in size.
19299 @item -mcmodel=large
19300 @opindex mcmodel=large
19301 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19302 in size. Other data and code is only limited by the 64-bit address
19306 @itemx -mno-altivec
19308 @opindex mno-altivec
19309 Generate code that uses (does not use) AltiVec instructions, and also
19310 enable the use of built-in functions that allow more direct access to
19311 the AltiVec instruction set. You may also need to set
19312 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19315 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19316 @option{-maltivec=be}, the element order for Altivec intrinsics such
19317 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19318 match array element order corresponding to the endianness of the
19319 target. That is, element zero identifies the leftmost element in a
19320 vector register when targeting a big-endian platform, and identifies
19321 the rightmost element in a vector register when targeting a
19322 little-endian platform.
19325 @opindex maltivec=be
19326 Generate Altivec instructions using big-endian element order,
19327 regardless of whether the target is big- or little-endian. This is
19328 the default when targeting a big-endian platform.
19330 The element order is used to interpret element numbers in Altivec
19331 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19332 @code{vec_insert}. By default, these match array element order
19333 corresponding to the endianness for the target.
19336 @opindex maltivec=le
19337 Generate Altivec instructions using little-endian element order,
19338 regardless of whether the target is big- or little-endian. This is
19339 the default when targeting a little-endian platform. This option is
19340 currently ignored when targeting a big-endian platform.
19342 The element order is used to interpret element numbers in Altivec
19343 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19344 @code{vec_insert}. By default, these match array element order
19345 corresponding to the endianness for the target.
19350 @opindex mno-vrsave
19351 Generate VRSAVE instructions when generating AltiVec code.
19353 @item -mgen-cell-microcode
19354 @opindex mgen-cell-microcode
19355 Generate Cell microcode instructions.
19357 @item -mwarn-cell-microcode
19358 @opindex mwarn-cell-microcode
19359 Warn when a Cell microcode instruction is emitted. An example
19360 of a Cell microcode instruction is a variable shift.
19363 @opindex msecure-plt
19364 Generate code that allows @command{ld} and @command{ld.so}
19365 to build executables and shared
19366 libraries with non-executable @code{.plt} and @code{.got} sections.
19368 32-bit SYSV ABI option.
19372 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19374 requires @code{.plt} and @code{.got}
19375 sections that are both writable and executable.
19376 This is a PowerPC 32-bit SYSV ABI option.
19382 This switch enables or disables the generation of ISEL instructions.
19384 @item -misel=@var{yes/no}
19385 This switch has been deprecated. Use @option{-misel} and
19386 @option{-mno-isel} instead.
19392 This switch enables or disables the generation of SPE simd
19398 @opindex mno-paired
19399 This switch enables or disables the generation of PAIRED simd
19402 @item -mspe=@var{yes/no}
19403 This option has been deprecated. Use @option{-mspe} and
19404 @option{-mno-spe} instead.
19410 Generate code that uses (does not use) vector/scalar (VSX)
19411 instructions, and also enable the use of built-in functions that allow
19412 more direct access to the VSX instruction set.
19417 @opindex mno-crypto
19418 Enable the use (disable) of the built-in functions that allow direct
19419 access to the cryptographic instructions that were added in version
19420 2.07 of the PowerPC ISA.
19422 @item -mdirect-move
19423 @itemx -mno-direct-move
19424 @opindex mdirect-move
19425 @opindex mno-direct-move
19426 Generate code that uses (does not use) the instructions to move data
19427 between the general purpose registers and the vector/scalar (VSX)
19428 registers that were added in version 2.07 of the PowerPC ISA.
19430 @item -mpower8-fusion
19431 @itemx -mno-power8-fusion
19432 @opindex mpower8-fusion
19433 @opindex mno-power8-fusion
19434 Generate code that keeps (does not keeps) some integer operations
19435 adjacent so that the instructions can be fused together on power8 and
19438 @item -mpower8-vector
19439 @itemx -mno-power8-vector
19440 @opindex mpower8-vector
19441 @opindex mno-power8-vector
19442 Generate code that uses (does not use) the vector and scalar
19443 instructions that were added in version 2.07 of the PowerPC ISA. Also
19444 enable the use of built-in functions that allow more direct access to
19445 the vector instructions.
19447 @item -mquad-memory
19448 @itemx -mno-quad-memory
19449 @opindex mquad-memory
19450 @opindex mno-quad-memory
19451 Generate code that uses (does not use) the non-atomic quad word memory
19452 instructions. The @option{-mquad-memory} option requires use of
19455 @item -mquad-memory-atomic
19456 @itemx -mno-quad-memory-atomic
19457 @opindex mquad-memory-atomic
19458 @opindex mno-quad-memory-atomic
19459 Generate code that uses (does not use) the atomic quad word memory
19460 instructions. The @option{-mquad-memory-atomic} option requires use of
19463 @item -mupper-regs-df
19464 @itemx -mno-upper-regs-df
19465 @opindex mupper-regs-df
19466 @opindex mno-upper-regs-df
19467 Generate code that uses (does not use) the scalar double precision
19468 instructions that target all 64 registers in the vector/scalar
19469 floating point register set that were added in version 2.06 of the
19470 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19471 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19472 @option{-mvsx} options.
19474 @item -mupper-regs-sf
19475 @itemx -mno-upper-regs-sf
19476 @opindex mupper-regs-sf
19477 @opindex mno-upper-regs-sf
19478 Generate code that uses (does not use) the scalar single precision
19479 instructions that target all 64 registers in the vector/scalar
19480 floating point register set that were added in version 2.07 of the
19481 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19482 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19486 @itemx -mno-upper-regs
19487 @opindex mupper-regs
19488 @opindex mno-upper-regs
19489 Generate code that uses (does not use) the scalar
19490 instructions that target all 64 registers in the vector/scalar
19491 floating point register set, depending on the model of the machine.
19493 If the @option{-mno-upper-regs} option is used, it turns off both
19494 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19496 @item -mfloat-gprs=@var{yes/single/double/no}
19497 @itemx -mfloat-gprs
19498 @opindex mfloat-gprs
19499 This switch enables or disables the generation of floating-point
19500 operations on the general-purpose registers for architectures that
19503 The argument @samp{yes} or @samp{single} enables the use of
19504 single-precision floating-point operations.
19506 The argument @samp{double} enables the use of single and
19507 double-precision floating-point operations.
19509 The argument @samp{no} disables floating-point operations on the
19510 general-purpose registers.
19512 This option is currently only available on the MPC854x.
19518 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19519 targets (including GNU/Linux). The 32-bit environment sets int, long
19520 and pointer to 32 bits and generates code that runs on any PowerPC
19521 variant. The 64-bit environment sets int to 32 bits and long and
19522 pointer to 64 bits, and generates code for PowerPC64, as for
19523 @option{-mpowerpc64}.
19526 @itemx -mno-fp-in-toc
19527 @itemx -mno-sum-in-toc
19528 @itemx -mminimal-toc
19530 @opindex mno-fp-in-toc
19531 @opindex mno-sum-in-toc
19532 @opindex mminimal-toc
19533 Modify generation of the TOC (Table Of Contents), which is created for
19534 every executable file. The @option{-mfull-toc} option is selected by
19535 default. In that case, GCC allocates at least one TOC entry for
19536 each unique non-automatic variable reference in your program. GCC
19537 also places floating-point constants in the TOC@. However, only
19538 16,384 entries are available in the TOC@.
19540 If you receive a linker error message that saying you have overflowed
19541 the available TOC space, you can reduce the amount of TOC space used
19542 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19543 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19544 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19545 generate code to calculate the sum of an address and a constant at
19546 run time instead of putting that sum into the TOC@. You may specify one
19547 or both of these options. Each causes GCC to produce very slightly
19548 slower and larger code at the expense of conserving TOC space.
19550 If you still run out of space in the TOC even when you specify both of
19551 these options, specify @option{-mminimal-toc} instead. This option causes
19552 GCC to make only one TOC entry for every file. When you specify this
19553 option, GCC produces code that is slower and larger but which
19554 uses extremely little TOC space. You may wish to use this option
19555 only on files that contain less frequently-executed code.
19561 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19562 @code{long} type, and the infrastructure needed to support them.
19563 Specifying @option{-maix64} implies @option{-mpowerpc64},
19564 while @option{-maix32} disables the 64-bit ABI and
19565 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19568 @itemx -mno-xl-compat
19569 @opindex mxl-compat
19570 @opindex mno-xl-compat
19571 Produce code that conforms more closely to IBM XL compiler semantics
19572 when using AIX-compatible ABI@. Pass floating-point arguments to
19573 prototyped functions beyond the register save area (RSA) on the stack
19574 in addition to argument FPRs. Do not assume that most significant
19575 double in 128-bit long double value is properly rounded when comparing
19576 values and converting to double. Use XL symbol names for long double
19579 The AIX calling convention was extended but not initially documented to
19580 handle an obscure K&R C case of calling a function that takes the
19581 address of its arguments with fewer arguments than declared. IBM XL
19582 compilers access floating-point arguments that do not fit in the
19583 RSA from the stack when a subroutine is compiled without
19584 optimization. Because always storing floating-point arguments on the
19585 stack is inefficient and rarely needed, this option is not enabled by
19586 default and only is necessary when calling subroutines compiled by IBM
19587 XL compilers without optimization.
19591 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19592 application written to use message passing with special startup code to
19593 enable the application to run. The system must have PE installed in the
19594 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19595 must be overridden with the @option{-specs=} option to specify the
19596 appropriate directory location. The Parallel Environment does not
19597 support threads, so the @option{-mpe} option and the @option{-pthread}
19598 option are incompatible.
19600 @item -malign-natural
19601 @itemx -malign-power
19602 @opindex malign-natural
19603 @opindex malign-power
19604 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19605 @option{-malign-natural} overrides the ABI-defined alignment of larger
19606 types, such as floating-point doubles, on their natural size-based boundary.
19607 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19608 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19610 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19614 @itemx -mhard-float
19615 @opindex msoft-float
19616 @opindex mhard-float
19617 Generate code that does not use (uses) the floating-point register set.
19618 Software floating-point emulation is provided if you use the
19619 @option{-msoft-float} option, and pass the option to GCC when linking.
19621 @item -msingle-float
19622 @itemx -mdouble-float
19623 @opindex msingle-float
19624 @opindex mdouble-float
19625 Generate code for single- or double-precision floating-point operations.
19626 @option{-mdouble-float} implies @option{-msingle-float}.
19629 @opindex msimple-fpu
19630 Do not generate @code{sqrt} and @code{div} instructions for hardware
19631 floating-point unit.
19633 @item -mfpu=@var{name}
19635 Specify type of floating-point unit. Valid values for @var{name} are
19636 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19637 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19638 @samp{sp_full} (equivalent to @option{-msingle-float}),
19639 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19642 @opindex mxilinx-fpu
19643 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19646 @itemx -mno-multiple
19648 @opindex mno-multiple
19649 Generate code that uses (does not use) the load multiple word
19650 instructions and the store multiple word instructions. These
19651 instructions are generated by default on POWER systems, and not
19652 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19653 PowerPC systems, since those instructions do not work when the
19654 processor is in little-endian mode. The exceptions are PPC740 and
19655 PPC750 which permit these instructions in little-endian mode.
19660 @opindex mno-string
19661 Generate code that uses (does not use) the load string instructions
19662 and the store string word instructions to save multiple registers and
19663 do small block moves. These instructions are generated by default on
19664 POWER systems, and not generated on PowerPC systems. Do not use
19665 @option{-mstring} on little-endian PowerPC systems, since those
19666 instructions do not work when the processor is in little-endian mode.
19667 The exceptions are PPC740 and PPC750 which permit these instructions
19668 in little-endian mode.
19673 @opindex mno-update
19674 Generate code that uses (does not use) the load or store instructions
19675 that update the base register to the address of the calculated memory
19676 location. These instructions are generated by default. If you use
19677 @option{-mno-update}, there is a small window between the time that the
19678 stack pointer is updated and the address of the previous frame is
19679 stored, which means code that walks the stack frame across interrupts or
19680 signals may get corrupted data.
19682 @item -mavoid-indexed-addresses
19683 @itemx -mno-avoid-indexed-addresses
19684 @opindex mavoid-indexed-addresses
19685 @opindex mno-avoid-indexed-addresses
19686 Generate code that tries to avoid (not avoid) the use of indexed load
19687 or store instructions. These instructions can incur a performance
19688 penalty on Power6 processors in certain situations, such as when
19689 stepping through large arrays that cross a 16M boundary. This option
19690 is enabled by default when targeting Power6 and disabled otherwise.
19693 @itemx -mno-fused-madd
19694 @opindex mfused-madd
19695 @opindex mno-fused-madd
19696 Generate code that uses (does not use) the floating-point multiply and
19697 accumulate instructions. These instructions are generated by default
19698 if hardware floating point is used. The machine-dependent
19699 @option{-mfused-madd} option is now mapped to the machine-independent
19700 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19701 mapped to @option{-ffp-contract=off}.
19707 Generate code that uses (does not use) the half-word multiply and
19708 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19709 These instructions are generated by default when targeting those
19716 Generate code that uses (does not use) the string-search @samp{dlmzb}
19717 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19718 generated by default when targeting those processors.
19720 @item -mno-bit-align
19722 @opindex mno-bit-align
19723 @opindex mbit-align
19724 On System V.4 and embedded PowerPC systems do not (do) force structures
19725 and unions that contain bit-fields to be aligned to the base type of the
19728 For example, by default a structure containing nothing but 8
19729 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19730 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19731 the structure is aligned to a 1-byte boundary and is 1 byte in
19734 @item -mno-strict-align
19735 @itemx -mstrict-align
19736 @opindex mno-strict-align
19737 @opindex mstrict-align
19738 On System V.4 and embedded PowerPC systems do not (do) assume that
19739 unaligned memory references are handled by the system.
19741 @item -mrelocatable
19742 @itemx -mno-relocatable
19743 @opindex mrelocatable
19744 @opindex mno-relocatable
19745 Generate code that allows (does not allow) a static executable to be
19746 relocated to a different address at run time. A simple embedded
19747 PowerPC system loader should relocate the entire contents of
19748 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19749 a table of 32-bit addresses generated by this option. For this to
19750 work, all objects linked together must be compiled with
19751 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19752 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19754 @item -mrelocatable-lib
19755 @itemx -mno-relocatable-lib
19756 @opindex mrelocatable-lib
19757 @opindex mno-relocatable-lib
19758 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19759 @code{.fixup} section to allow static executables to be relocated at
19760 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19761 alignment of @option{-mrelocatable}. Objects compiled with
19762 @option{-mrelocatable-lib} may be linked with objects compiled with
19763 any combination of the @option{-mrelocatable} options.
19769 On System V.4 and embedded PowerPC systems do not (do) assume that
19770 register 2 contains a pointer to a global area pointing to the addresses
19771 used in the program.
19774 @itemx -mlittle-endian
19776 @opindex mlittle-endian
19777 On System V.4 and embedded PowerPC systems compile code for the
19778 processor in little-endian mode. The @option{-mlittle-endian} option is
19779 the same as @option{-mlittle}.
19782 @itemx -mbig-endian
19784 @opindex mbig-endian
19785 On System V.4 and embedded PowerPC systems compile code for the
19786 processor in big-endian mode. The @option{-mbig-endian} option is
19787 the same as @option{-mbig}.
19789 @item -mdynamic-no-pic
19790 @opindex mdynamic-no-pic
19791 On Darwin and Mac OS X systems, compile code so that it is not
19792 relocatable, but that its external references are relocatable. The
19793 resulting code is suitable for applications, but not shared
19796 @item -msingle-pic-base
19797 @opindex msingle-pic-base
19798 Treat the register used for PIC addressing as read-only, rather than
19799 loading it in the prologue for each function. The runtime system is
19800 responsible for initializing this register with an appropriate value
19801 before execution begins.
19803 @item -mprioritize-restricted-insns=@var{priority}
19804 @opindex mprioritize-restricted-insns
19805 This option controls the priority that is assigned to
19806 dispatch-slot restricted instructions during the second scheduling
19807 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19808 or @samp{2} to assign no, highest, or second-highest (respectively)
19809 priority to dispatch-slot restricted
19812 @item -msched-costly-dep=@var{dependence_type}
19813 @opindex msched-costly-dep
19814 This option controls which dependences are considered costly
19815 by the target during instruction scheduling. The argument
19816 @var{dependence_type} takes one of the following values:
19820 No dependence is costly.
19823 All dependences are costly.
19825 @item @samp{true_store_to_load}
19826 A true dependence from store to load is costly.
19828 @item @samp{store_to_load}
19829 Any dependence from store to load is costly.
19832 Any dependence for which the latency is greater than or equal to
19833 @var{number} is costly.
19836 @item -minsert-sched-nops=@var{scheme}
19837 @opindex minsert-sched-nops
19838 This option controls which NOP insertion scheme is used during
19839 the second scheduling pass. The argument @var{scheme} takes one of the
19847 Pad with NOPs any dispatch group that has vacant issue slots,
19848 according to the scheduler's grouping.
19850 @item @samp{regroup_exact}
19851 Insert NOPs to force costly dependent insns into
19852 separate groups. Insert exactly as many NOPs as needed to force an insn
19853 to a new group, according to the estimated processor grouping.
19856 Insert NOPs to force costly dependent insns into
19857 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19861 @opindex mcall-sysv
19862 On System V.4 and embedded PowerPC systems compile code using calling
19863 conventions that adhere to the March 1995 draft of the System V
19864 Application Binary Interface, PowerPC processor supplement. This is the
19865 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19867 @item -mcall-sysv-eabi
19869 @opindex mcall-sysv-eabi
19870 @opindex mcall-eabi
19871 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19873 @item -mcall-sysv-noeabi
19874 @opindex mcall-sysv-noeabi
19875 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19877 @item -mcall-aixdesc
19879 On System V.4 and embedded PowerPC systems compile code for the AIX
19883 @opindex mcall-linux
19884 On System V.4 and embedded PowerPC systems compile code for the
19885 Linux-based GNU system.
19887 @item -mcall-freebsd
19888 @opindex mcall-freebsd
19889 On System V.4 and embedded PowerPC systems compile code for the
19890 FreeBSD operating system.
19892 @item -mcall-netbsd
19893 @opindex mcall-netbsd
19894 On System V.4 and embedded PowerPC systems compile code for the
19895 NetBSD operating system.
19897 @item -mcall-openbsd
19898 @opindex mcall-netbsd
19899 On System V.4 and embedded PowerPC systems compile code for the
19900 OpenBSD operating system.
19902 @item -maix-struct-return
19903 @opindex maix-struct-return
19904 Return all structures in memory (as specified by the AIX ABI)@.
19906 @item -msvr4-struct-return
19907 @opindex msvr4-struct-return
19908 Return structures smaller than 8 bytes in registers (as specified by the
19911 @item -mabi=@var{abi-type}
19913 Extend the current ABI with a particular extension, or remove such extension.
19914 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19915 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19916 @samp{elfv1}, @samp{elfv2}@.
19920 Extend the current ABI with SPE ABI extensions. This does not change
19921 the default ABI, instead it adds the SPE ABI extensions to the current
19925 @opindex mabi=no-spe
19926 Disable Book-E SPE ABI extensions for the current ABI@.
19928 @item -mabi=ibmlongdouble
19929 @opindex mabi=ibmlongdouble
19930 Change the current ABI to use IBM extended-precision long double.
19931 This is a PowerPC 32-bit SYSV ABI option.
19933 @item -mabi=ieeelongdouble
19934 @opindex mabi=ieeelongdouble
19935 Change the current ABI to use IEEE extended-precision long double.
19936 This is a PowerPC 32-bit Linux ABI option.
19939 @opindex mabi=elfv1
19940 Change the current ABI to use the ELFv1 ABI.
19941 This is the default ABI for big-endian PowerPC 64-bit Linux.
19942 Overriding the default ABI requires special system support and is
19943 likely to fail in spectacular ways.
19946 @opindex mabi=elfv2
19947 Change the current ABI to use the ELFv2 ABI.
19948 This is the default ABI for little-endian PowerPC 64-bit Linux.
19949 Overriding the default ABI requires special system support and is
19950 likely to fail in spectacular ways.
19953 @itemx -mno-prototype
19954 @opindex mprototype
19955 @opindex mno-prototype
19956 On System V.4 and embedded PowerPC systems assume that all calls to
19957 variable argument functions are properly prototyped. Otherwise, the
19958 compiler must insert an instruction before every non-prototyped call to
19959 set or clear bit 6 of the condition code register (@code{CR}) to
19960 indicate whether floating-point values are passed in the floating-point
19961 registers in case the function takes variable arguments. With
19962 @option{-mprototype}, only calls to prototyped variable argument functions
19963 set or clear the bit.
19967 On embedded PowerPC systems, assume that the startup module is called
19968 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19969 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19974 On embedded PowerPC systems, assume that the startup module is called
19975 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19980 On embedded PowerPC systems, assume that the startup module is called
19981 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19984 @item -myellowknife
19985 @opindex myellowknife
19986 On embedded PowerPC systems, assume that the startup module is called
19987 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19992 On System V.4 and embedded PowerPC systems, specify that you are
19993 compiling for a VxWorks system.
19997 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19998 header to indicate that @samp{eabi} extended relocations are used.
20004 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20005 Embedded Applications Binary Interface (EABI), which is a set of
20006 modifications to the System V.4 specifications. Selecting @option{-meabi}
20007 means that the stack is aligned to an 8-byte boundary, a function
20008 @code{__eabi} is called from @code{main} to set up the EABI
20009 environment, and the @option{-msdata} option can use both @code{r2} and
20010 @code{r13} to point to two separate small data areas. Selecting
20011 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20012 no EABI initialization function is called from @code{main}, and the
20013 @option{-msdata} option only uses @code{r13} to point to a single
20014 small data area. The @option{-meabi} option is on by default if you
20015 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20018 @opindex msdata=eabi
20019 On System V.4 and embedded PowerPC systems, put small initialized
20020 @code{const} global and static data in the @code{.sdata2} section, which
20021 is pointed to by register @code{r2}. Put small initialized
20022 non-@code{const} global and static data in the @code{.sdata} section,
20023 which is pointed to by register @code{r13}. Put small uninitialized
20024 global and static data in the @code{.sbss} section, which is adjacent to
20025 the @code{.sdata} section. The @option{-msdata=eabi} option is
20026 incompatible with the @option{-mrelocatable} option. The
20027 @option{-msdata=eabi} option also sets the @option{-memb} option.
20030 @opindex msdata=sysv
20031 On System V.4 and embedded PowerPC systems, put small global and static
20032 data in the @code{.sdata} section, which is pointed to by register
20033 @code{r13}. Put small uninitialized global and static data in the
20034 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20035 The @option{-msdata=sysv} option is incompatible with the
20036 @option{-mrelocatable} option.
20038 @item -msdata=default
20040 @opindex msdata=default
20042 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20043 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20044 same as @option{-msdata=sysv}.
20047 @opindex msdata=data
20048 On System V.4 and embedded PowerPC systems, put small global
20049 data in the @code{.sdata} section. Put small uninitialized global
20050 data in the @code{.sbss} section. Do not use register @code{r13}
20051 to address small data however. This is the default behavior unless
20052 other @option{-msdata} options are used.
20056 @opindex msdata=none
20058 On embedded PowerPC systems, put all initialized global and static data
20059 in the @code{.data} section, and all uninitialized data in the
20060 @code{.bss} section.
20062 @item -mblock-move-inline-limit=@var{num}
20063 @opindex mblock-move-inline-limit
20064 Inline all block moves (such as calls to @code{memcpy} or structure
20065 copies) less than or equal to @var{num} bytes. The minimum value for
20066 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20067 targets. The default value is target-specific.
20071 @cindex smaller data references (PowerPC)
20072 @cindex .sdata/.sdata2 references (PowerPC)
20073 On embedded PowerPC systems, put global and static items less than or
20074 equal to @var{num} bytes into the small data or BSS sections instead of
20075 the normal data or BSS section. By default, @var{num} is 8. The
20076 @option{-G @var{num}} switch is also passed to the linker.
20077 All modules should be compiled with the same @option{-G @var{num}} value.
20080 @itemx -mno-regnames
20082 @opindex mno-regnames
20083 On System V.4 and embedded PowerPC systems do (do not) emit register
20084 names in the assembly language output using symbolic forms.
20087 @itemx -mno-longcall
20089 @opindex mno-longcall
20090 By default assume that all calls are far away so that a longer and more
20091 expensive calling sequence is required. This is required for calls
20092 farther than 32 megabytes (33,554,432 bytes) from the current location.
20093 A short call is generated if the compiler knows
20094 the call cannot be that far away. This setting can be overridden by
20095 the @code{shortcall} function attribute, or by @code{#pragma
20098 Some linkers are capable of detecting out-of-range calls and generating
20099 glue code on the fly. On these systems, long calls are unnecessary and
20100 generate slower code. As of this writing, the AIX linker can do this,
20101 as can the GNU linker for PowerPC/64. It is planned to add this feature
20102 to the GNU linker for 32-bit PowerPC systems as well.
20104 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20105 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20106 addresses represent the callee and the branch island. The
20107 Darwin/PPC linker prefers the first address and generates a @code{bl
20108 callee} if the PPC @code{bl} instruction reaches the callee directly;
20109 otherwise, the linker generates @code{bl L42} to call the branch
20110 island. The branch island is appended to the body of the
20111 calling function; it computes the full 32-bit address of the callee
20114 On Mach-O (Darwin) systems, this option directs the compiler emit to
20115 the glue for every direct call, and the Darwin linker decides whether
20116 to use or discard it.
20118 In the future, GCC may ignore all longcall specifications
20119 when the linker is known to generate glue.
20121 @item -mtls-markers
20122 @itemx -mno-tls-markers
20123 @opindex mtls-markers
20124 @opindex mno-tls-markers
20125 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20126 specifying the function argument. The relocation allows the linker to
20127 reliably associate function call with argument setup instructions for
20128 TLS optimization, which in turn allows GCC to better schedule the
20133 Adds support for multithreading with the @dfn{pthreads} library.
20134 This option sets flags for both the preprocessor and linker.
20139 This option enables use of the reciprocal estimate and
20140 reciprocal square root estimate instructions with additional
20141 Newton-Raphson steps to increase precision instead of doing a divide or
20142 square root and divide for floating-point arguments. You should use
20143 the @option{-ffast-math} option when using @option{-mrecip} (or at
20144 least @option{-funsafe-math-optimizations},
20145 @option{-finite-math-only}, @option{-freciprocal-math} and
20146 @option{-fno-trapping-math}). Note that while the throughput of the
20147 sequence is generally higher than the throughput of the non-reciprocal
20148 instruction, the precision of the sequence can be decreased by up to 2
20149 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20152 @item -mrecip=@var{opt}
20153 @opindex mrecip=opt
20154 This option controls which reciprocal estimate instructions
20155 may be used. @var{opt} is a comma-separated list of options, which may
20156 be preceded by a @code{!} to invert the option:
20161 Enable all estimate instructions.
20164 Enable the default instructions, equivalent to @option{-mrecip}.
20167 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20170 Enable the reciprocal approximation instructions for both
20171 single and double precision.
20174 Enable the single-precision reciprocal approximation instructions.
20177 Enable the double-precision reciprocal approximation instructions.
20180 Enable the reciprocal square root approximation instructions for both
20181 single and double precision.
20184 Enable the single-precision reciprocal square root approximation instructions.
20187 Enable the double-precision reciprocal square root approximation instructions.
20191 So, for example, @option{-mrecip=all,!rsqrtd} enables
20192 all of the reciprocal estimate instructions, except for the
20193 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20194 which handle the double-precision reciprocal square root calculations.
20196 @item -mrecip-precision
20197 @itemx -mno-recip-precision
20198 @opindex mrecip-precision
20199 Assume (do not assume) that the reciprocal estimate instructions
20200 provide higher-precision estimates than is mandated by the PowerPC
20201 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20202 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20203 The double-precision square root estimate instructions are not generated by
20204 default on low-precision machines, since they do not provide an
20205 estimate that converges after three steps.
20207 @item -mveclibabi=@var{type}
20208 @opindex mveclibabi
20209 Specifies the ABI type to use for vectorizing intrinsics using an
20210 external library. The only type supported at present is @samp{mass},
20211 which specifies to use IBM's Mathematical Acceleration Subsystem
20212 (MASS) libraries for vectorizing intrinsics using external libraries.
20213 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20214 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20215 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20216 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20217 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20218 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20219 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20220 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20221 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20222 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20223 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20224 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20225 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20226 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20227 for power7. Both @option{-ftree-vectorize} and
20228 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20229 libraries must be specified at link time.
20234 Generate (do not generate) the @code{friz} instruction when the
20235 @option{-funsafe-math-optimizations} option is used to optimize
20236 rounding of floating-point values to 64-bit integer and back to floating
20237 point. The @code{friz} instruction does not return the same value if
20238 the floating-point number is too large to fit in an integer.
20240 @item -mpointers-to-nested-functions
20241 @itemx -mno-pointers-to-nested-functions
20242 @opindex mpointers-to-nested-functions
20243 Generate (do not generate) code to load up the static chain register
20244 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20245 systems where a function pointer points to a 3-word descriptor giving
20246 the function address, TOC value to be loaded in register @code{r2}, and
20247 static chain value to be loaded in register @code{r11}. The
20248 @option{-mpointers-to-nested-functions} is on by default. You cannot
20249 call through pointers to nested functions or pointers
20250 to functions compiled in other languages that use the static chain if
20251 you use @option{-mno-pointers-to-nested-functions}.
20253 @item -msave-toc-indirect
20254 @itemx -mno-save-toc-indirect
20255 @opindex msave-toc-indirect
20256 Generate (do not generate) code to save the TOC value in the reserved
20257 stack location in the function prologue if the function calls through
20258 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20259 saved in the prologue, it is saved just before the call through the
20260 pointer. The @option{-mno-save-toc-indirect} option is the default.
20262 @item -mcompat-align-parm
20263 @itemx -mno-compat-align-parm
20264 @opindex mcompat-align-parm
20265 Generate (do not generate) code to pass structure parameters with a
20266 maximum alignment of 64 bits, for compatibility with older versions
20269 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20270 structure parameter on a 128-bit boundary when that structure contained
20271 a member requiring 128-bit alignment. This is corrected in more
20272 recent versions of GCC. This option may be used to generate code
20273 that is compatible with functions compiled with older versions of
20276 The @option{-mno-compat-align-parm} option is the default.
20280 @subsection RX Options
20283 These command-line options are defined for RX targets:
20286 @item -m64bit-doubles
20287 @itemx -m32bit-doubles
20288 @opindex m64bit-doubles
20289 @opindex m32bit-doubles
20290 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20291 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20292 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20293 works on 32-bit values, which is why the default is
20294 @option{-m32bit-doubles}.
20300 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20301 floating-point hardware. The default is enabled for the RX600
20302 series and disabled for the RX200 series.
20304 Floating-point instructions are only generated for 32-bit floating-point
20305 values, however, so the FPU hardware is not used for doubles if the
20306 @option{-m64bit-doubles} option is used.
20308 @emph{Note} If the @option{-fpu} option is enabled then
20309 @option{-funsafe-math-optimizations} is also enabled automatically.
20310 This is because the RX FPU instructions are themselves unsafe.
20312 @item -mcpu=@var{name}
20314 Selects the type of RX CPU to be targeted. Currently three types are
20315 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20316 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20318 The only difference between @samp{RX600} and @samp{RX610} is that the
20319 @samp{RX610} does not support the @code{MVTIPL} instruction.
20321 The @samp{RX200} series does not have a hardware floating-point unit
20322 and so @option{-nofpu} is enabled by default when this type is
20325 @item -mbig-endian-data
20326 @itemx -mlittle-endian-data
20327 @opindex mbig-endian-data
20328 @opindex mlittle-endian-data
20329 Store data (but not code) in the big-endian format. The default is
20330 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20333 @item -msmall-data-limit=@var{N}
20334 @opindex msmall-data-limit
20335 Specifies the maximum size in bytes of global and static variables
20336 which can be placed into the small data area. Using the small data
20337 area can lead to smaller and faster code, but the size of area is
20338 limited and it is up to the programmer to ensure that the area does
20339 not overflow. Also when the small data area is used one of the RX's
20340 registers (usually @code{r13}) is reserved for use pointing to this
20341 area, so it is no longer available for use by the compiler. This
20342 could result in slower and/or larger code if variables are pushed onto
20343 the stack instead of being held in this register.
20345 Note, common variables (variables that have not been initialized) and
20346 constants are not placed into the small data area as they are assigned
20347 to other sections in the output executable.
20349 The default value is zero, which disables this feature. Note, this
20350 feature is not enabled by default with higher optimization levels
20351 (@option{-O2} etc) because of the potentially detrimental effects of
20352 reserving a register. It is up to the programmer to experiment and
20353 discover whether this feature is of benefit to their program. See the
20354 description of the @option{-mpid} option for a description of how the
20355 actual register to hold the small data area pointer is chosen.
20361 Use the simulator runtime. The default is to use the libgloss
20362 board-specific runtime.
20364 @item -mas100-syntax
20365 @itemx -mno-as100-syntax
20366 @opindex mas100-syntax
20367 @opindex mno-as100-syntax
20368 When generating assembler output use a syntax that is compatible with
20369 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20370 assembler, but it has some restrictions so it is not generated by default.
20372 @item -mmax-constant-size=@var{N}
20373 @opindex mmax-constant-size
20374 Specifies the maximum size, in bytes, of a constant that can be used as
20375 an operand in a RX instruction. Although the RX instruction set does
20376 allow constants of up to 4 bytes in length to be used in instructions,
20377 a longer value equates to a longer instruction. Thus in some
20378 circumstances it can be beneficial to restrict the size of constants
20379 that are used in instructions. Constants that are too big are instead
20380 placed into a constant pool and referenced via register indirection.
20382 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20383 or 4 means that constants of any size are allowed.
20387 Enable linker relaxation. Linker relaxation is a process whereby the
20388 linker attempts to reduce the size of a program by finding shorter
20389 versions of various instructions. Disabled by default.
20391 @item -mint-register=@var{N}
20392 @opindex mint-register
20393 Specify the number of registers to reserve for fast interrupt handler
20394 functions. The value @var{N} can be between 0 and 4. A value of 1
20395 means that register @code{r13} is reserved for the exclusive use
20396 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20397 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20398 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20399 A value of 0, the default, does not reserve any registers.
20401 @item -msave-acc-in-interrupts
20402 @opindex msave-acc-in-interrupts
20403 Specifies that interrupt handler functions should preserve the
20404 accumulator register. This is only necessary if normal code might use
20405 the accumulator register, for example because it performs 64-bit
20406 multiplications. The default is to ignore the accumulator as this
20407 makes the interrupt handlers faster.
20413 Enables the generation of position independent data. When enabled any
20414 access to constant data is done via an offset from a base address
20415 held in a register. This allows the location of constant data to be
20416 determined at run time without requiring the executable to be
20417 relocated, which is a benefit to embedded applications with tight
20418 memory constraints. Data that can be modified is not affected by this
20421 Note, using this feature reserves a register, usually @code{r13}, for
20422 the constant data base address. This can result in slower and/or
20423 larger code, especially in complicated functions.
20425 The actual register chosen to hold the constant data base address
20426 depends upon whether the @option{-msmall-data-limit} and/or the
20427 @option{-mint-register} command-line options are enabled. Starting
20428 with register @code{r13} and proceeding downwards, registers are
20429 allocated first to satisfy the requirements of @option{-mint-register},
20430 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20431 is possible for the small data area register to be @code{r8} if both
20432 @option{-mint-register=4} and @option{-mpid} are specified on the
20435 By default this feature is not enabled. The default can be restored
20436 via the @option{-mno-pid} command-line option.
20438 @item -mno-warn-multiple-fast-interrupts
20439 @itemx -mwarn-multiple-fast-interrupts
20440 @opindex mno-warn-multiple-fast-interrupts
20441 @opindex mwarn-multiple-fast-interrupts
20442 Prevents GCC from issuing a warning message if it finds more than one
20443 fast interrupt handler when it is compiling a file. The default is to
20444 issue a warning for each extra fast interrupt handler found, as the RX
20445 only supports one such interrupt.
20447 @item -mallow-string-insns
20448 @itemx -mno-allow-string-insns
20449 @opindex mallow-string-insns
20450 @opindex mno-allow-string-insns
20451 Enables or disables the use of the string manipulation instructions
20452 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20453 @code{SWHILE} and also the @code{RMPA} instruction. These
20454 instructions may prefetch data, which is not safe to do if accessing
20455 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20456 for more information).
20458 The default is to allow these instructions, but it is not possible for
20459 GCC to reliably detect all circumstances where a string instruction
20460 might be used to access an I/O register, so their use cannot be
20461 disabled automatically. Instead it is reliant upon the programmer to
20462 use the @option{-mno-allow-string-insns} option if their program
20463 accesses I/O space.
20465 When the instructions are enabled GCC defines the C preprocessor
20466 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20467 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20470 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20471 has special significance to the RX port when used with the
20472 @code{interrupt} function attribute. This attribute indicates a
20473 function intended to process fast interrupts. GCC ensures
20474 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20475 and/or @code{r13} and only provided that the normal use of the
20476 corresponding registers have been restricted via the
20477 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20480 @node S/390 and zSeries Options
20481 @subsection S/390 and zSeries Options
20482 @cindex S/390 and zSeries Options
20484 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20488 @itemx -msoft-float
20489 @opindex mhard-float
20490 @opindex msoft-float
20491 Use (do not use) the hardware floating-point instructions and registers
20492 for floating-point operations. When @option{-msoft-float} is specified,
20493 functions in @file{libgcc.a} are used to perform floating-point
20494 operations. When @option{-mhard-float} is specified, the compiler
20495 generates IEEE floating-point instructions. This is the default.
20498 @itemx -mno-hard-dfp
20500 @opindex mno-hard-dfp
20501 Use (do not use) the hardware decimal-floating-point instructions for
20502 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20503 specified, functions in @file{libgcc.a} are used to perform
20504 decimal-floating-point operations. When @option{-mhard-dfp} is
20505 specified, the compiler generates decimal-floating-point hardware
20506 instructions. This is the default for @option{-march=z9-ec} or higher.
20508 @item -mlong-double-64
20509 @itemx -mlong-double-128
20510 @opindex mlong-double-64
20511 @opindex mlong-double-128
20512 These switches control the size of @code{long double} type. A size
20513 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20514 type. This is the default.
20517 @itemx -mno-backchain
20518 @opindex mbackchain
20519 @opindex mno-backchain
20520 Store (do not store) the address of the caller's frame as backchain pointer
20521 into the callee's stack frame.
20522 A backchain may be needed to allow debugging using tools that do not understand
20523 DWARF 2 call frame information.
20524 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20525 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20526 the backchain is placed into the topmost word of the 96/160 byte register
20529 In general, code compiled with @option{-mbackchain} is call-compatible with
20530 code compiled with @option{-mmo-backchain}; however, use of the backchain
20531 for debugging purposes usually requires that the whole binary is built with
20532 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20533 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20534 to build a linux kernel use @option{-msoft-float}.
20536 The default is to not maintain the backchain.
20538 @item -mpacked-stack
20539 @itemx -mno-packed-stack
20540 @opindex mpacked-stack
20541 @opindex mno-packed-stack
20542 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20543 specified, the compiler uses the all fields of the 96/160 byte register save
20544 area only for their default purpose; unused fields still take up stack space.
20545 When @option{-mpacked-stack} is specified, register save slots are densely
20546 packed at the top of the register save area; unused space is reused for other
20547 purposes, allowing for more efficient use of the available stack space.
20548 However, when @option{-mbackchain} is also in effect, the topmost word of
20549 the save area is always used to store the backchain, and the return address
20550 register is always saved two words below the backchain.
20552 As long as the stack frame backchain is not used, code generated with
20553 @option{-mpacked-stack} is call-compatible with code generated with
20554 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20555 S/390 or zSeries generated code that uses the stack frame backchain at run
20556 time, not just for debugging purposes. Such code is not call-compatible
20557 with code compiled with @option{-mpacked-stack}. Also, note that the
20558 combination of @option{-mbackchain},
20559 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20560 to build a linux kernel use @option{-msoft-float}.
20562 The default is to not use the packed stack layout.
20565 @itemx -mno-small-exec
20566 @opindex msmall-exec
20567 @opindex mno-small-exec
20568 Generate (or do not generate) code using the @code{bras} instruction
20569 to do subroutine calls.
20570 This only works reliably if the total executable size does not
20571 exceed 64k. The default is to use the @code{basr} instruction instead,
20572 which does not have this limitation.
20578 When @option{-m31} is specified, generate code compliant to the
20579 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20580 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20581 particular to generate 64-bit instructions. For the @samp{s390}
20582 targets, the default is @option{-m31}, while the @samp{s390x}
20583 targets default to @option{-m64}.
20589 When @option{-mzarch} is specified, generate code using the
20590 instructions available on z/Architecture.
20591 When @option{-mesa} is specified, generate code using the
20592 instructions available on ESA/390. Note that @option{-mesa} is
20593 not possible with @option{-m64}.
20594 When generating code compliant to the GNU/Linux for S/390 ABI,
20595 the default is @option{-mesa}. When generating code compliant
20596 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20602 The @option{-mhtm} option enables a set of builtins making use of
20603 instructions available with the transactional execution facility
20604 introduced with the IBM zEnterprise EC12 machine generation
20605 @ref{S/390 System z Built-in Functions}.
20606 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
20612 When @option{-mvx} is specified, generate code using the instructions
20613 available with the vector extension facility introduced with the IBM
20614 z13 machine generation.
20615 This option changes the ABI for some vector type values with regard to
20616 alignment and calling conventions. In case vector type values are
20617 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
20618 command will be added to mark the resulting binary with the ABI used.
20619 @option{-mvx} is enabled by default when using @option{-march=z13}.
20622 @itemx -mno-zvector
20624 @opindex mno-zvector
20625 The @option{-mzvector} option enables vector language extensions and
20626 builtins using instructions available with the vector extension
20627 facility introduced with the IBM z13 machine generation.
20628 This option adds support for @samp{vector} to be used as a keyword to
20629 define vector type variables and arguments. @samp{vector} is only
20630 available when GNU extensions are enabled. It will not be expanded
20631 when requesting strict standard compliance e.g. with @option{-std=c99}.
20632 In addition to the GCC low-level builtins @option{-mzvector} enables
20633 a set of builtins added for compatibility with Altivec-style
20634 implementations like Power and Cell. In order to make use of these
20635 builtins the header file @file{vecintrin.h} needs to be included.
20636 @option{-mzvector} is disabled by default.
20642 Generate (or do not generate) code using the @code{mvcle} instruction
20643 to perform block moves. When @option{-mno-mvcle} is specified,
20644 use a @code{mvc} loop instead. This is the default unless optimizing for
20651 Print (or do not print) additional debug information when compiling.
20652 The default is to not print debug information.
20654 @item -march=@var{cpu-type}
20656 Generate code that runs on @var{cpu-type}, which is the name of a system
20657 representing a certain processor type. Possible values for
20658 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20659 @samp{z9-109}, @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12},
20661 When generating code using the instructions available on z/Architecture,
20662 the default is @option{-march=z900}. Otherwise, the default is
20663 @option{-march=g5}.
20665 @item -mtune=@var{cpu-type}
20667 Tune to @var{cpu-type} everything applicable about the generated code,
20668 except for the ABI and the set of available instructions.
20669 The list of @var{cpu-type} values is the same as for @option{-march}.
20670 The default is the value used for @option{-march}.
20673 @itemx -mno-tpf-trace
20674 @opindex mtpf-trace
20675 @opindex mno-tpf-trace
20676 Generate code that adds (does not add) in TPF OS specific branches to trace
20677 routines in the operating system. This option is off by default, even
20678 when compiling for the TPF OS@.
20681 @itemx -mno-fused-madd
20682 @opindex mfused-madd
20683 @opindex mno-fused-madd
20684 Generate code that uses (does not use) the floating-point multiply and
20685 accumulate instructions. These instructions are generated by default if
20686 hardware floating point is used.
20688 @item -mwarn-framesize=@var{framesize}
20689 @opindex mwarn-framesize
20690 Emit a warning if the current function exceeds the given frame size. Because
20691 this is a compile-time check it doesn't need to be a real problem when the program
20692 runs. It is intended to identify functions that most probably cause
20693 a stack overflow. It is useful to be used in an environment with limited stack
20694 size e.g.@: the linux kernel.
20696 @item -mwarn-dynamicstack
20697 @opindex mwarn-dynamicstack
20698 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20699 arrays. This is generally a bad idea with a limited stack size.
20701 @item -mstack-guard=@var{stack-guard}
20702 @itemx -mstack-size=@var{stack-size}
20703 @opindex mstack-guard
20704 @opindex mstack-size
20705 If these options are provided the S/390 back end emits additional instructions in
20706 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20707 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20708 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20709 the frame size of the compiled function is chosen.
20710 These options are intended to be used to help debugging stack overflow problems.
20711 The additionally emitted code causes only little overhead and hence can also be
20712 used in production-like systems without greater performance degradation. The given
20713 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20714 @var{stack-guard} without exceeding 64k.
20715 In order to be efficient the extra code makes the assumption that the stack starts
20716 at an address aligned to the value given by @var{stack-size}.
20717 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20719 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20721 If the hotpatch option is enabled, a ``hot-patching'' function
20722 prologue is generated for all functions in the compilation unit.
20723 The funtion label is prepended with the given number of two-byte
20724 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20725 the label, 2 * @var{post-halfwords} bytes are appended, using the
20726 largest NOP like instructions the architecture allows (maximum
20729 If both arguments are zero, hotpatching is disabled.
20731 This option can be overridden for individual functions with the
20732 @code{hotpatch} attribute.
20735 @node Score Options
20736 @subsection Score Options
20737 @cindex Score Options
20739 These options are defined for Score implementations:
20744 Compile code for big-endian mode. This is the default.
20748 Compile code for little-endian mode.
20752 Disable generation of @code{bcnz} instructions.
20756 Enable generation of unaligned load and store instructions.
20760 Enable the use of multiply-accumulate instructions. Disabled by default.
20764 Specify the SCORE5 as the target architecture.
20768 Specify the SCORE5U of the target architecture.
20772 Specify the SCORE7 as the target architecture. This is the default.
20776 Specify the SCORE7D as the target architecture.
20780 @subsection SH Options
20782 These @samp{-m} options are defined for the SH implementations:
20787 Generate code for the SH1.
20791 Generate code for the SH2.
20794 Generate code for the SH2e.
20798 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20799 that the floating-point unit is not used.
20801 @item -m2a-single-only
20802 @opindex m2a-single-only
20803 Generate code for the SH2a-FPU, in such a way that no double-precision
20804 floating-point operations are used.
20807 @opindex m2a-single
20808 Generate code for the SH2a-FPU assuming the floating-point unit is in
20809 single-precision mode by default.
20813 Generate code for the SH2a-FPU assuming the floating-point unit is in
20814 double-precision mode by default.
20818 Generate code for the SH3.
20822 Generate code for the SH3e.
20826 Generate code for the SH4 without a floating-point unit.
20828 @item -m4-single-only
20829 @opindex m4-single-only
20830 Generate code for the SH4 with a floating-point unit that only
20831 supports single-precision arithmetic.
20835 Generate code for the SH4 assuming the floating-point unit is in
20836 single-precision mode by default.
20840 Generate code for the SH4.
20844 Generate code for SH4-100.
20846 @item -m4-100-nofpu
20847 @opindex m4-100-nofpu
20848 Generate code for SH4-100 in such a way that the
20849 floating-point unit is not used.
20851 @item -m4-100-single
20852 @opindex m4-100-single
20853 Generate code for SH4-100 assuming the floating-point unit is in
20854 single-precision mode by default.
20856 @item -m4-100-single-only
20857 @opindex m4-100-single-only
20858 Generate code for SH4-100 in such a way that no double-precision
20859 floating-point operations are used.
20863 Generate code for SH4-200.
20865 @item -m4-200-nofpu
20866 @opindex m4-200-nofpu
20867 Generate code for SH4-200 without in such a way that the
20868 floating-point unit is not used.
20870 @item -m4-200-single
20871 @opindex m4-200-single
20872 Generate code for SH4-200 assuming the floating-point unit is in
20873 single-precision mode by default.
20875 @item -m4-200-single-only
20876 @opindex m4-200-single-only
20877 Generate code for SH4-200 in such a way that no double-precision
20878 floating-point operations are used.
20882 Generate code for SH4-300.
20884 @item -m4-300-nofpu
20885 @opindex m4-300-nofpu
20886 Generate code for SH4-300 without in such a way that the
20887 floating-point unit is not used.
20889 @item -m4-300-single
20890 @opindex m4-300-single
20891 Generate code for SH4-300 in such a way that no double-precision
20892 floating-point operations are used.
20894 @item -m4-300-single-only
20895 @opindex m4-300-single-only
20896 Generate code for SH4-300 in such a way that no double-precision
20897 floating-point operations are used.
20901 Generate code for SH4-340 (no MMU, no FPU).
20905 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20910 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20911 floating-point unit is not used.
20913 @item -m4a-single-only
20914 @opindex m4a-single-only
20915 Generate code for the SH4a, in such a way that no double-precision
20916 floating-point operations are used.
20919 @opindex m4a-single
20920 Generate code for the SH4a assuming the floating-point unit is in
20921 single-precision mode by default.
20925 Generate code for the SH4a.
20929 Same as @option{-m4a-nofpu}, except that it implicitly passes
20930 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20931 instructions at the moment.
20935 Compile code for the processor in big-endian mode.
20939 Compile code for the processor in little-endian mode.
20943 Align doubles at 64-bit boundaries. Note that this changes the calling
20944 conventions, and thus some functions from the standard C library do
20945 not work unless you recompile it first with @option{-mdalign}.
20949 Shorten some address references at link time, when possible; uses the
20950 linker option @option{-relax}.
20954 Use 32-bit offsets in @code{switch} tables. The default is to use
20959 Enable the use of bit manipulation instructions on SH2A.
20963 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20964 alignment constraints.
20968 Comply with the calling conventions defined by Renesas.
20971 @opindex mno-renesas
20972 Comply with the calling conventions defined for GCC before the Renesas
20973 conventions were available. This option is the default for all
20974 targets of the SH toolchain.
20977 @opindex mnomacsave
20978 Mark the @code{MAC} register as call-clobbered, even if
20979 @option{-mrenesas} is given.
20985 Control the IEEE compliance of floating-point comparisons, which affects the
20986 handling of cases where the result of a comparison is unordered. By default
20987 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20988 enabled @option{-mno-ieee} is implicitly set, which results in faster
20989 floating-point greater-equal and less-equal comparisons. The implcit settings
20990 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20992 @item -minline-ic_invalidate
20993 @opindex minline-ic_invalidate
20994 Inline code to invalidate instruction cache entries after setting up
20995 nested function trampolines.
20996 This option has no effect if @option{-musermode} is in effect and the selected
20997 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20999 If the selected code generation option does not allow the use of the @code{icbi}
21000 instruction, and @option{-musermode} is not in effect, the inlined code
21001 manipulates the instruction cache address array directly with an associative
21002 write. This not only requires privileged mode at run time, but it also
21003 fails if the cache line had been mapped via the TLB and has become unmapped.
21007 Dump instruction size and location in the assembly code.
21010 @opindex mpadstruct
21011 This option is deprecated. It pads structures to multiple of 4 bytes,
21012 which is incompatible with the SH ABI@.
21014 @item -matomic-model=@var{model}
21015 @opindex matomic-model=@var{model}
21016 Sets the model of atomic operations and additional parameters as a comma
21017 separated list. For details on the atomic built-in functions see
21018 @ref{__atomic Builtins}. The following models and parameters are supported:
21023 Disable compiler generated atomic sequences and emit library calls for atomic
21024 operations. This is the default if the target is not @code{sh*-*-linux*}.
21027 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21028 built-in functions. The generated atomic sequences require additional support
21029 from the interrupt/exception handling code of the system and are only suitable
21030 for SH3* and SH4* single-core systems. This option is enabled by default when
21031 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21032 this option also partially utilizes the hardware atomic instructions
21033 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21034 @samp{strict} is specified.
21037 Generate software atomic sequences that use a variable in the thread control
21038 block. This is a variation of the gUSA sequences which can also be used on
21039 SH1* and SH2* targets. The generated atomic sequences require additional
21040 support from the interrupt/exception handling code of the system and are only
21041 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21042 parameter has to be specified as well.
21045 Generate software atomic sequences that temporarily disable interrupts by
21046 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21047 in privileged mode and is only suitable for single-core systems. Additional
21048 support from the interrupt/exception handling code of the system is not
21049 required. This model is enabled by default when the target is
21050 @code{sh*-*-linux*} and SH1* or SH2*.
21053 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21054 instructions only. This is only available on SH4A and is suitable for
21055 multi-core systems. Since the hardware instructions support only 32 bit atomic
21056 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21057 Code compiled with this option is also compatible with other software
21058 atomic model interrupt/exception handling systems if executed on an SH4A
21059 system. Additional support from the interrupt/exception handling code of the
21060 system is not required for this model.
21063 This parameter specifies the offset in bytes of the variable in the thread
21064 control block structure that should be used by the generated atomic sequences
21065 when the @samp{soft-tcb} model has been selected. For other models this
21066 parameter is ignored. The specified value must be an integer multiple of four
21067 and in the range 0-1020.
21070 This parameter prevents mixed usage of multiple atomic models, even if they
21071 are compatible, and makes the compiler generate atomic sequences of the
21072 specified model only.
21078 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21079 Notice that depending on the particular hardware and software configuration
21080 this can degrade overall performance due to the operand cache line flushes
21081 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21082 processors the @code{tas.b} instruction must be used with caution since it
21083 can result in data corruption for certain cache configurations.
21086 @opindex mprefergot
21087 When generating position-independent code, emit function calls using
21088 the Global Offset Table instead of the Procedure Linkage Table.
21091 @itemx -mno-usermode
21093 @opindex mno-usermode
21094 Don't allow (allow) the compiler generating privileged mode code. Specifying
21095 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21096 inlined code would not work in user mode. @option{-musermode} is the default
21097 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21098 @option{-musermode} has no effect, since there is no user mode.
21100 @item -multcost=@var{number}
21101 @opindex multcost=@var{number}
21102 Set the cost to assume for a multiply insn.
21104 @item -mdiv=@var{strategy}
21105 @opindex mdiv=@var{strategy}
21106 Set the division strategy to be used for integer division operations.
21107 @var{strategy} can be one of:
21112 Calls a library function that uses the single-step division instruction
21113 @code{div1} to perform the operation. Division by zero calculates an
21114 unspecified result and does not trap. This is the default except for SH4,
21115 SH2A and SHcompact.
21118 Calls a library function that performs the operation in double precision
21119 floating point. Division by zero causes a floating-point exception. This is
21120 the default for SHcompact with FPU. Specifying this for targets that do not
21121 have a double precision FPU defaults to @code{call-div1}.
21124 Calls a library function that uses a lookup table for small divisors and
21125 the @code{div1} instruction with case distinction for larger divisors. Division
21126 by zero calculates an unspecified result and does not trap. This is the default
21127 for SH4. Specifying this for targets that do not have dynamic shift
21128 instructions defaults to @code{call-div1}.
21132 When a division strategy has not been specified the default strategy is
21133 selected based on the current target. For SH2A the default strategy is to
21134 use the @code{divs} and @code{divu} instructions instead of library function
21137 @item -maccumulate-outgoing-args
21138 @opindex maccumulate-outgoing-args
21139 Reserve space once for outgoing arguments in the function prologue rather
21140 than around each call. Generally beneficial for performance and size. Also
21141 needed for unwinding to avoid changing the stack frame around conditional code.
21143 @item -mdivsi3_libfunc=@var{name}
21144 @opindex mdivsi3_libfunc=@var{name}
21145 Set the name of the library function used for 32-bit signed division to
21147 This only affects the name used in the @samp{call} division strategies, and
21148 the compiler still expects the same sets of input/output/clobbered registers as
21149 if this option were not present.
21151 @item -mfixed-range=@var{register-range}
21152 @opindex mfixed-range
21153 Generate code treating the given register range as fixed registers.
21154 A fixed register is one that the register allocator can not use. This is
21155 useful when compiling kernel code. A register range is specified as
21156 two registers separated by a dash. Multiple register ranges can be
21157 specified separated by a comma.
21159 @item -mbranch-cost=@var{num}
21160 @opindex mbranch-cost=@var{num}
21161 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21162 make the compiler try to generate more branch-free code if possible.
21163 If not specified the value is selected depending on the processor type that
21164 is being compiled for.
21167 @itemx -mno-zdcbranch
21168 @opindex mzdcbranch
21169 @opindex mno-zdcbranch
21170 Assume (do not assume) that zero displacement conditional branch instructions
21171 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21172 compiler prefers zero displacement branch code sequences. This is
21173 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21174 disabled by specifying @option{-mno-zdcbranch}.
21176 @item -mcbranch-force-delay-slot
21177 @opindex mcbranch-force-delay-slot
21178 Force the usage of delay slots for conditional branches, which stuffs the delay
21179 slot with a @code{nop} if a suitable instruction can't be found. By default
21180 this option is disabled. It can be enabled to work around hardware bugs as
21181 found in the original SH7055.
21184 @itemx -mno-fused-madd
21185 @opindex mfused-madd
21186 @opindex mno-fused-madd
21187 Generate code that uses (does not use) the floating-point multiply and
21188 accumulate instructions. These instructions are generated by default
21189 if hardware floating point is used. The machine-dependent
21190 @option{-mfused-madd} option is now mapped to the machine-independent
21191 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21192 mapped to @option{-ffp-contract=off}.
21198 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21199 and cosine approximations. The option @option{-mfsca} must be used in
21200 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21201 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21202 approximations even if @option{-funsafe-math-optimizations} is in effect.
21208 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21209 reciprocal square root approximations. The option @option{-mfsrra} must be used
21210 in combination with @option{-funsafe-math-optimizations} and
21211 @option{-ffinite-math-only}. It is enabled by default when generating code for
21212 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21213 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21216 @item -mpretend-cmove
21217 @opindex mpretend-cmove
21218 Prefer zero-displacement conditional branches for conditional move instruction
21219 patterns. This can result in faster code on the SH4 processor.
21223 @node Solaris 2 Options
21224 @subsection Solaris 2 Options
21225 @cindex Solaris 2 options
21227 These @samp{-m} options are supported on Solaris 2:
21230 @item -mclear-hwcap
21231 @opindex mclear-hwcap
21232 @option{-mclear-hwcap} tells the compiler to remove the hardware
21233 capabilities generated by the Solaris assembler. This is only necessary
21234 when object files use ISA extensions not supported by the current
21235 machine, but check at runtime whether or not to use them.
21237 @item -mimpure-text
21238 @opindex mimpure-text
21239 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21240 the compiler to not pass @option{-z text} to the linker when linking a
21241 shared object. Using this option, you can link position-dependent
21242 code into a shared object.
21244 @option{-mimpure-text} suppresses the ``relocations remain against
21245 allocatable but non-writable sections'' linker error message.
21246 However, the necessary relocations trigger copy-on-write, and the
21247 shared object is not actually shared across processes. Instead of
21248 using @option{-mimpure-text}, you should compile all source code with
21249 @option{-fpic} or @option{-fPIC}.
21253 These switches are supported in addition to the above on Solaris 2:
21258 Add support for multithreading using the POSIX threads library. This
21259 option sets flags for both the preprocessor and linker. This option does
21260 not affect the thread safety of object code produced by the compiler or
21261 that of libraries supplied with it.
21265 This is a synonym for @option{-pthreads}.
21268 @node SPARC Options
21269 @subsection SPARC Options
21270 @cindex SPARC options
21272 These @samp{-m} options are supported on the SPARC:
21275 @item -mno-app-regs
21277 @opindex mno-app-regs
21279 Specify @option{-mapp-regs} to generate output using the global registers
21280 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21281 global register 1, each global register 2 through 4 is then treated as an
21282 allocable register that is clobbered by function calls. This is the default.
21284 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21285 specify @option{-mno-app-regs}. You should compile libraries and system
21286 software with this option.
21292 With @option{-mflat}, the compiler does not generate save/restore instructions
21293 and uses a ``flat'' or single register window model. This model is compatible
21294 with the regular register window model. The local registers and the input
21295 registers (0--5) are still treated as ``call-saved'' registers and are
21296 saved on the stack as needed.
21298 With @option{-mno-flat} (the default), the compiler generates save/restore
21299 instructions (except for leaf functions). This is the normal operating mode.
21302 @itemx -mhard-float
21304 @opindex mhard-float
21305 Generate output containing floating-point instructions. This is the
21309 @itemx -msoft-float
21311 @opindex msoft-float
21312 Generate output containing library calls for floating point.
21313 @strong{Warning:} the requisite libraries are not available for all SPARC
21314 targets. Normally the facilities of the machine's usual C compiler are
21315 used, but this cannot be done directly in cross-compilation. You must make
21316 your own arrangements to provide suitable library functions for
21317 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21318 @samp{sparclite-*-*} do provide software floating-point support.
21320 @option{-msoft-float} changes the calling convention in the output file;
21321 therefore, it is only useful if you compile @emph{all} of a program with
21322 this option. In particular, you need to compile @file{libgcc.a}, the
21323 library that comes with GCC, with @option{-msoft-float} in order for
21326 @item -mhard-quad-float
21327 @opindex mhard-quad-float
21328 Generate output containing quad-word (long double) floating-point
21331 @item -msoft-quad-float
21332 @opindex msoft-quad-float
21333 Generate output containing library calls for quad-word (long double)
21334 floating-point instructions. The functions called are those specified
21335 in the SPARC ABI@. This is the default.
21337 As of this writing, there are no SPARC implementations that have hardware
21338 support for the quad-word floating-point instructions. They all invoke
21339 a trap handler for one of these instructions, and then the trap handler
21340 emulates the effect of the instruction. Because of the trap handler overhead,
21341 this is much slower than calling the ABI library routines. Thus the
21342 @option{-msoft-quad-float} option is the default.
21344 @item -mno-unaligned-doubles
21345 @itemx -munaligned-doubles
21346 @opindex mno-unaligned-doubles
21347 @opindex munaligned-doubles
21348 Assume that doubles have 8-byte alignment. This is the default.
21350 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21351 alignment only if they are contained in another type, or if they have an
21352 absolute address. Otherwise, it assumes they have 4-byte alignment.
21353 Specifying this option avoids some rare compatibility problems with code
21354 generated by other compilers. It is not the default because it results
21355 in a performance loss, especially for floating-point code.
21358 @itemx -mno-user-mode
21359 @opindex muser-mode
21360 @opindex mno-user-mode
21361 Do not generate code that can only run in supervisor mode. This is relevant
21362 only for the @code{casa} instruction emitted for the LEON3 processor. This
21365 @item -mno-faster-structs
21366 @itemx -mfaster-structs
21367 @opindex mno-faster-structs
21368 @opindex mfaster-structs
21369 With @option{-mfaster-structs}, the compiler assumes that structures
21370 should have 8-byte alignment. This enables the use of pairs of
21371 @code{ldd} and @code{std} instructions for copies in structure
21372 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21373 However, the use of this changed alignment directly violates the SPARC
21374 ABI@. Thus, it's intended only for use on targets where the developer
21375 acknowledges that their resulting code is not directly in line with
21376 the rules of the ABI@.
21378 @item -mcpu=@var{cpu_type}
21380 Set the instruction set, register set, and instruction scheduling parameters
21381 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21382 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21383 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21384 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21385 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21386 @samp{niagara3} and @samp{niagara4}.
21388 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21389 which selects the best architecture option for the host processor.
21390 @option{-mcpu=native} has no effect if GCC does not recognize
21393 Default instruction scheduling parameters are used for values that select
21394 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21395 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21397 Here is a list of each supported architecture and their supported
21405 supersparc, hypersparc, leon, leon3
21408 f930, f934, sparclite86x
21414 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21417 By default (unless configured otherwise), GCC generates code for the V7
21418 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21419 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21420 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21421 SPARCStation 1, 2, IPX etc.
21423 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21424 architecture. The only difference from V7 code is that the compiler emits
21425 the integer multiply and integer divide instructions which exist in SPARC-V8
21426 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21427 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21430 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21431 the SPARC architecture. This adds the integer multiply, integer divide step
21432 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21433 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21434 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21435 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21436 MB86934 chip, which is the more recent SPARClite with FPU@.
21438 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21439 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21440 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21441 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21442 optimizes it for the TEMIC SPARClet chip.
21444 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21445 architecture. This adds 64-bit integer and floating-point move instructions,
21446 3 additional floating-point condition code registers and conditional move
21447 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21448 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21449 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21450 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21451 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21452 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21453 additionally optimizes it for Sun UltraSPARC T2 chips. With
21454 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21455 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21456 additionally optimizes it for Sun UltraSPARC T4 chips.
21458 @item -mtune=@var{cpu_type}
21460 Set the instruction scheduling parameters for machine type
21461 @var{cpu_type}, but do not set the instruction set or register set that the
21462 option @option{-mcpu=@var{cpu_type}} does.
21464 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21465 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21466 that select a particular CPU implementation. Those are @samp{cypress},
21467 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21468 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21469 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21470 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21471 toolchains, @samp{native} can also be used.
21476 @opindex mno-v8plus
21477 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21478 difference from the V8 ABI is that the global and out registers are
21479 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21480 mode for all SPARC-V9 processors.
21486 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21487 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21493 With @option{-mvis2}, GCC generates code that takes advantage of
21494 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21495 default is @option{-mvis2} when targeting a cpu that supports such
21496 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21497 also sets @option{-mvis}.
21503 With @option{-mvis3}, GCC generates code that takes advantage of
21504 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21505 default is @option{-mvis3} when targeting a cpu that supports such
21506 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21507 also sets @option{-mvis2} and @option{-mvis}.
21512 @opindex mno-cbcond
21513 With @option{-mcbcond}, GCC generates code that takes advantage of
21514 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21515 The default is @option{-mcbcond} when targeting a cpu that supports such
21516 instructions, such as niagara-4 and later.
21522 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21523 population count instruction. The default is @option{-mpopc}
21524 when targeting a cpu that supports such instructions, such as Niagara-2 and
21531 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21532 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21533 when targeting a cpu that supports such instructions, such as Niagara-3 and
21537 @opindex mfix-at697f
21538 Enable the documented workaround for the single erratum of the Atmel AT697F
21539 processor (which corresponds to erratum #13 of the AT697E processor).
21542 @opindex mfix-ut699
21543 Enable the documented workarounds for the floating-point errata and the data
21544 cache nullify errata of the UT699 processor.
21547 These @samp{-m} options are supported in addition to the above
21548 on SPARC-V9 processors in 64-bit environments:
21555 Generate code for a 32-bit or 64-bit environment.
21556 The 32-bit environment sets int, long and pointer to 32 bits.
21557 The 64-bit environment sets int to 32 bits and long and pointer
21560 @item -mcmodel=@var{which}
21562 Set the code model to one of
21566 The Medium/Low code model: 64-bit addresses, programs
21567 must be linked in the low 32 bits of memory. Programs can be statically
21568 or dynamically linked.
21571 The Medium/Middle code model: 64-bit addresses, programs
21572 must be linked in the low 44 bits of memory, the text and data segments must
21573 be less than 2GB in size and the data segment must be located within 2GB of
21577 The Medium/Anywhere code model: 64-bit addresses, programs
21578 may be linked anywhere in memory, the text and data segments must be less
21579 than 2GB in size and the data segment must be located within 2GB of the
21583 The Medium/Anywhere code model for embedded systems:
21584 64-bit addresses, the text and data segments must be less than 2GB in
21585 size, both starting anywhere in memory (determined at link time). The
21586 global register %g4 points to the base of the data segment. Programs
21587 are statically linked and PIC is not supported.
21590 @item -mmemory-model=@var{mem-model}
21591 @opindex mmemory-model
21592 Set the memory model in force on the processor to one of
21596 The default memory model for the processor and operating system.
21599 Relaxed Memory Order
21602 Partial Store Order
21608 Sequential Consistency
21611 These memory models are formally defined in Appendix D of the Sparc V9
21612 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21615 @itemx -mno-stack-bias
21616 @opindex mstack-bias
21617 @opindex mno-stack-bias
21618 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21619 frame pointer if present, are offset by @minus{}2047 which must be added back
21620 when making stack frame references. This is the default in 64-bit mode.
21621 Otherwise, assume no such offset is present.
21625 @subsection SPU Options
21626 @cindex SPU options
21628 These @samp{-m} options are supported on the SPU:
21632 @itemx -merror-reloc
21633 @opindex mwarn-reloc
21634 @opindex merror-reloc
21636 The loader for SPU does not handle dynamic relocations. By default, GCC
21637 gives an error when it generates code that requires a dynamic
21638 relocation. @option{-mno-error-reloc} disables the error,
21639 @option{-mwarn-reloc} generates a warning instead.
21642 @itemx -munsafe-dma
21644 @opindex munsafe-dma
21646 Instructions that initiate or test completion of DMA must not be
21647 reordered with respect to loads and stores of the memory that is being
21649 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21650 memory accesses, but that can lead to inefficient code in places where the
21651 memory is known to not change. Rather than mark the memory as volatile,
21652 you can use @option{-msafe-dma} to tell the compiler to treat
21653 the DMA instructions as potentially affecting all memory.
21655 @item -mbranch-hints
21656 @opindex mbranch-hints
21658 By default, GCC generates a branch hint instruction to avoid
21659 pipeline stalls for always-taken or probably-taken branches. A hint
21660 is not generated closer than 8 instructions away from its branch.
21661 There is little reason to disable them, except for debugging purposes,
21662 or to make an object a little bit smaller.
21666 @opindex msmall-mem
21667 @opindex mlarge-mem
21669 By default, GCC generates code assuming that addresses are never larger
21670 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21671 a full 32-bit address.
21676 By default, GCC links against startup code that assumes the SPU-style
21677 main function interface (which has an unconventional parameter list).
21678 With @option{-mstdmain}, GCC links your program against startup
21679 code that assumes a C99-style interface to @code{main}, including a
21680 local copy of @code{argv} strings.
21682 @item -mfixed-range=@var{register-range}
21683 @opindex mfixed-range
21684 Generate code treating the given register range as fixed registers.
21685 A fixed register is one that the register allocator cannot use. This is
21686 useful when compiling kernel code. A register range is specified as
21687 two registers separated by a dash. Multiple register ranges can be
21688 specified separated by a comma.
21694 Compile code assuming that pointers to the PPU address space accessed
21695 via the @code{__ea} named address space qualifier are either 32 or 64
21696 bits wide. The default is 32 bits. As this is an ABI-changing option,
21697 all object code in an executable must be compiled with the same setting.
21699 @item -maddress-space-conversion
21700 @itemx -mno-address-space-conversion
21701 @opindex maddress-space-conversion
21702 @opindex mno-address-space-conversion
21703 Allow/disallow treating the @code{__ea} address space as superset
21704 of the generic address space. This enables explicit type casts
21705 between @code{__ea} and generic pointer as well as implicit
21706 conversions of generic pointers to @code{__ea} pointers. The
21707 default is to allow address space pointer conversions.
21709 @item -mcache-size=@var{cache-size}
21710 @opindex mcache-size
21711 This option controls the version of libgcc that the compiler links to an
21712 executable and selects a software-managed cache for accessing variables
21713 in the @code{__ea} address space with a particular cache size. Possible
21714 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21715 and @samp{128}. The default cache size is 64KB.
21717 @item -matomic-updates
21718 @itemx -mno-atomic-updates
21719 @opindex matomic-updates
21720 @opindex mno-atomic-updates
21721 This option controls the version of libgcc that the compiler links to an
21722 executable and selects whether atomic updates to the software-managed
21723 cache of PPU-side variables are used. If you use atomic updates, changes
21724 to a PPU variable from SPU code using the @code{__ea} named address space
21725 qualifier do not interfere with changes to other PPU variables residing
21726 in the same cache line from PPU code. If you do not use atomic updates,
21727 such interference may occur; however, writing back cache lines is
21728 more efficient. The default behavior is to use atomic updates.
21731 @itemx -mdual-nops=@var{n}
21732 @opindex mdual-nops
21733 By default, GCC inserts nops to increase dual issue when it expects
21734 it to increase performance. @var{n} can be a value from 0 to 10. A
21735 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21736 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21738 @item -mhint-max-nops=@var{n}
21739 @opindex mhint-max-nops
21740 Maximum number of nops to insert for a branch hint. A branch hint must
21741 be at least 8 instructions away from the branch it is affecting. GCC
21742 inserts up to @var{n} nops to enforce this, otherwise it does not
21743 generate the branch hint.
21745 @item -mhint-max-distance=@var{n}
21746 @opindex mhint-max-distance
21747 The encoding of the branch hint instruction limits the hint to be within
21748 256 instructions of the branch it is affecting. By default, GCC makes
21749 sure it is within 125.
21752 @opindex msafe-hints
21753 Work around a hardware bug that causes the SPU to stall indefinitely.
21754 By default, GCC inserts the @code{hbrp} instruction to make sure
21755 this stall won't happen.
21759 @node System V Options
21760 @subsection Options for System V
21762 These additional options are available on System V Release 4 for
21763 compatibility with other compilers on those systems:
21768 Create a shared object.
21769 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21773 Identify the versions of each tool used by the compiler, in a
21774 @code{.ident} assembler directive in the output.
21778 Refrain from adding @code{.ident} directives to the output file (this is
21781 @item -YP,@var{dirs}
21783 Search the directories @var{dirs}, and no others, for libraries
21784 specified with @option{-l}.
21786 @item -Ym,@var{dir}
21788 Look in the directory @var{dir} to find the M4 preprocessor.
21789 The assembler uses this option.
21790 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21791 @c the generic assembler that comes with Solaris takes just -Ym.
21794 @node TILE-Gx Options
21795 @subsection TILE-Gx Options
21796 @cindex TILE-Gx options
21798 These @samp{-m} options are supported on the TILE-Gx:
21801 @item -mcmodel=small
21802 @opindex mcmodel=small
21803 Generate code for the small model. The distance for direct calls is
21804 limited to 500M in either direction. PC-relative addresses are 32
21805 bits. Absolute addresses support the full address range.
21807 @item -mcmodel=large
21808 @opindex mcmodel=large
21809 Generate code for the large model. There is no limitation on call
21810 distance, pc-relative addresses, or absolute addresses.
21812 @item -mcpu=@var{name}
21814 Selects the type of CPU to be targeted. Currently the only supported
21815 type is @samp{tilegx}.
21821 Generate code for a 32-bit or 64-bit environment. The 32-bit
21822 environment sets int, long, and pointer to 32 bits. The 64-bit
21823 environment sets int to 32 bits and long and pointer to 64 bits.
21826 @itemx -mlittle-endian
21827 @opindex mbig-endian
21828 @opindex mlittle-endian
21829 Generate code in big/little endian mode, respectively.
21832 @node TILEPro Options
21833 @subsection TILEPro Options
21834 @cindex TILEPro options
21836 These @samp{-m} options are supported on the TILEPro:
21839 @item -mcpu=@var{name}
21841 Selects the type of CPU to be targeted. Currently the only supported
21842 type is @samp{tilepro}.
21846 Generate code for a 32-bit environment, which sets int, long, and
21847 pointer to 32 bits. This is the only supported behavior so the flag
21848 is essentially ignored.
21852 @subsection V850 Options
21853 @cindex V850 Options
21855 These @samp{-m} options are defined for V850 implementations:
21859 @itemx -mno-long-calls
21860 @opindex mlong-calls
21861 @opindex mno-long-calls
21862 Treat all calls as being far away (near). If calls are assumed to be
21863 far away, the compiler always loads the function's address into a
21864 register, and calls indirect through the pointer.
21870 Do not optimize (do optimize) basic blocks that use the same index
21871 pointer 4 or more times to copy pointer into the @code{ep} register, and
21872 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21873 option is on by default if you optimize.
21875 @item -mno-prolog-function
21876 @itemx -mprolog-function
21877 @opindex mno-prolog-function
21878 @opindex mprolog-function
21879 Do not use (do use) external functions to save and restore registers
21880 at the prologue and epilogue of a function. The external functions
21881 are slower, but use less code space if more than one function saves
21882 the same number of registers. The @option{-mprolog-function} option
21883 is on by default if you optimize.
21887 Try to make the code as small as possible. At present, this just turns
21888 on the @option{-mep} and @option{-mprolog-function} options.
21890 @item -mtda=@var{n}
21892 Put static or global variables whose size is @var{n} bytes or less into
21893 the tiny data area that register @code{ep} points to. The tiny data
21894 area can hold up to 256 bytes in total (128 bytes for byte references).
21896 @item -msda=@var{n}
21898 Put static or global variables whose size is @var{n} bytes or less into
21899 the small data area that register @code{gp} points to. The small data
21900 area can hold up to 64 kilobytes.
21902 @item -mzda=@var{n}
21904 Put static or global variables whose size is @var{n} bytes or less into
21905 the first 32 kilobytes of memory.
21909 Specify that the target processor is the V850.
21913 Specify that the target processor is the V850E3V5. The preprocessor
21914 constant @code{__v850e3v5__} is defined if this option is used.
21918 Specify that the target processor is the V850E3V5. This is an alias for
21919 the @option{-mv850e3v5} option.
21923 Specify that the target processor is the V850E2V3. The preprocessor
21924 constant @code{__v850e2v3__} is defined if this option is used.
21928 Specify that the target processor is the V850E2. The preprocessor
21929 constant @code{__v850e2__} is defined if this option is used.
21933 Specify that the target processor is the V850E1. The preprocessor
21934 constants @code{__v850e1__} and @code{__v850e__} are defined if
21935 this option is used.
21939 Specify that the target processor is the V850ES. This is an alias for
21940 the @option{-mv850e1} option.
21944 Specify that the target processor is the V850E@. The preprocessor
21945 constant @code{__v850e__} is defined if this option is used.
21947 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21948 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21949 are defined then a default target processor is chosen and the
21950 relevant @samp{__v850*__} preprocessor constant is defined.
21952 The preprocessor constants @code{__v850} and @code{__v851__} are always
21953 defined, regardless of which processor variant is the target.
21955 @item -mdisable-callt
21956 @itemx -mno-disable-callt
21957 @opindex mdisable-callt
21958 @opindex mno-disable-callt
21959 This option suppresses generation of the @code{CALLT} instruction for the
21960 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21963 This option is enabled by default when the RH850 ABI is
21964 in use (see @option{-mrh850-abi}), and disabled by default when the
21965 GCC ABI is in use. If @code{CALLT} instructions are being generated
21966 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21972 Pass on (or do not pass on) the @option{-mrelax} command-line option
21976 @itemx -mno-long-jumps
21977 @opindex mlong-jumps
21978 @opindex mno-long-jumps
21979 Disable (or re-enable) the generation of PC-relative jump instructions.
21982 @itemx -mhard-float
21983 @opindex msoft-float
21984 @opindex mhard-float
21985 Disable (or re-enable) the generation of hardware floating point
21986 instructions. This option is only significant when the target
21987 architecture is @samp{V850E2V3} or higher. If hardware floating point
21988 instructions are being generated then the C preprocessor symbol
21989 @code{__FPU_OK__} is defined, otherwise the symbol
21990 @code{__NO_FPU__} is defined.
21994 Enables the use of the e3v5 LOOP instruction. The use of this
21995 instruction is not enabled by default when the e3v5 architecture is
21996 selected because its use is still experimental.
22000 @opindex mrh850-abi
22002 Enables support for the RH850 version of the V850 ABI. This is the
22003 default. With this version of the ABI the following rules apply:
22007 Integer sized structures and unions are returned via a memory pointer
22008 rather than a register.
22011 Large structures and unions (more than 8 bytes in size) are passed by
22015 Functions are aligned to 16-bit boundaries.
22018 The @option{-m8byte-align} command-line option is supported.
22021 The @option{-mdisable-callt} command-line option is enabled by
22022 default. The @option{-mno-disable-callt} command-line option is not
22026 When this version of the ABI is enabled the C preprocessor symbol
22027 @code{__V850_RH850_ABI__} is defined.
22031 Enables support for the old GCC version of the V850 ABI. With this
22032 version of the ABI the following rules apply:
22036 Integer sized structures and unions are returned in register @code{r10}.
22039 Large structures and unions (more than 8 bytes in size) are passed by
22043 Functions are aligned to 32-bit boundaries, unless optimizing for
22047 The @option{-m8byte-align} command-line option is not supported.
22050 The @option{-mdisable-callt} command-line option is supported but not
22051 enabled by default.
22054 When this version of the ABI is enabled the C preprocessor symbol
22055 @code{__V850_GCC_ABI__} is defined.
22057 @item -m8byte-align
22058 @itemx -mno-8byte-align
22059 @opindex m8byte-align
22060 @opindex mno-8byte-align
22061 Enables support for @code{double} and @code{long long} types to be
22062 aligned on 8-byte boundaries. The default is to restrict the
22063 alignment of all objects to at most 4-bytes. When
22064 @option{-m8byte-align} is in effect the C preprocessor symbol
22065 @code{__V850_8BYTE_ALIGN__} is defined.
22068 @opindex mbig-switch
22069 Generate code suitable for big switch tables. Use this option only if
22070 the assembler/linker complain about out of range branches within a switch
22075 This option causes r2 and r5 to be used in the code generated by
22076 the compiler. This setting is the default.
22078 @item -mno-app-regs
22079 @opindex mno-app-regs
22080 This option causes r2 and r5 to be treated as fixed registers.
22085 @subsection VAX Options
22086 @cindex VAX options
22088 These @samp{-m} options are defined for the VAX:
22093 Do not output certain jump instructions (@code{aobleq} and so on)
22094 that the Unix assembler for the VAX cannot handle across long
22099 Do output those jump instructions, on the assumption that the
22100 GNU assembler is being used.
22104 Output code for G-format floating-point numbers instead of D-format.
22107 @node Visium Options
22108 @subsection Visium Options
22109 @cindex Visium options
22115 A program which performs file I/O and is destined to run on an MCM target
22116 should be linked with this option. It causes the libraries libc.a and
22117 libdebug.a to be linked. The program should be run on the target under
22118 the control of the GDB remote debugging stub.
22122 A program which performs file I/O and is destined to run on the simulator
22123 should be linked with option. This causes libraries libc.a and libsim.a to
22127 @itemx -mhard-float
22129 @opindex mhard-float
22130 Generate code containing floating-point instructions. This is the
22134 @itemx -msoft-float
22136 @opindex msoft-float
22137 Generate code containing library calls for floating-point.
22139 @option{-msoft-float} changes the calling convention in the output file;
22140 therefore, it is only useful if you compile @emph{all} of a program with
22141 this option. In particular, you need to compile @file{libgcc.a}, the
22142 library that comes with GCC, with @option{-msoft-float} in order for
22145 @item -mcpu=@var{cpu_type}
22147 Set the instruction set, register set, and instruction scheduling parameters
22148 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22149 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22151 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22153 By default (unless configured otherwise), GCC generates code for the GR5
22154 variant of the Visium architecture.
22156 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22157 architecture. The only difference from GR5 code is that the compiler will
22158 generate block move instructions.
22160 @item -mtune=@var{cpu_type}
22162 Set the instruction scheduling parameters for machine type @var{cpu_type},
22163 but do not set the instruction set or register set that the option
22164 @option{-mcpu=@var{cpu_type}} would.
22168 Generate code for the supervisor mode, where there are no restrictions on
22169 the access to general registers. This is the default.
22172 @opindex muser-mode
22173 Generate code for the user mode, where the access to some general registers
22174 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22175 mode; on the GR6, only registers r29 to r31 are affected.
22179 @subsection VMS Options
22181 These @samp{-m} options are defined for the VMS implementations:
22184 @item -mvms-return-codes
22185 @opindex mvms-return-codes
22186 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22187 condition (e.g.@ error) codes.
22189 @item -mdebug-main=@var{prefix}
22190 @opindex mdebug-main=@var{prefix}
22191 Flag the first routine whose name starts with @var{prefix} as the main
22192 routine for the debugger.
22196 Default to 64-bit memory allocation routines.
22198 @item -mpointer-size=@var{size}
22199 @opindex mpointer-size=@var{size}
22200 Set the default size of pointers. Possible options for @var{size} are
22201 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22202 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22203 The later option disables @code{pragma pointer_size}.
22206 @node VxWorks Options
22207 @subsection VxWorks Options
22208 @cindex VxWorks Options
22210 The options in this section are defined for all VxWorks targets.
22211 Options specific to the target hardware are listed with the other
22212 options for that target.
22217 GCC can generate code for both VxWorks kernels and real time processes
22218 (RTPs). This option switches from the former to the latter. It also
22219 defines the preprocessor macro @code{__RTP__}.
22222 @opindex non-static
22223 Link an RTP executable against shared libraries rather than static
22224 libraries. The options @option{-static} and @option{-shared} can
22225 also be used for RTPs (@pxref{Link Options}); @option{-static}
22232 These options are passed down to the linker. They are defined for
22233 compatibility with Diab.
22236 @opindex Xbind-lazy
22237 Enable lazy binding of function calls. This option is equivalent to
22238 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22242 Disable lazy binding of function calls. This option is the default and
22243 is defined for compatibility with Diab.
22247 @subsection x86 Options
22248 @cindex x86 Options
22250 These @samp{-m} options are defined for the x86 family of computers.
22254 @item -march=@var{cpu-type}
22256 Generate instructions for the machine type @var{cpu-type}. In contrast to
22257 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22258 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22259 to generate code that may not run at all on processors other than the one
22260 indicated. Specifying @option{-march=@var{cpu-type}} implies
22261 @option{-mtune=@var{cpu-type}}.
22263 The choices for @var{cpu-type} are:
22267 This selects the CPU to generate code for at compilation time by determining
22268 the processor type of the compiling machine. Using @option{-march=native}
22269 enables all instruction subsets supported by the local machine (hence
22270 the result might not run on different machines). Using @option{-mtune=native}
22271 produces code optimized for the local machine under the constraints
22272 of the selected instruction set.
22275 Original Intel i386 CPU@.
22278 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22282 Intel Pentium CPU with no MMX support.
22285 Intel Lakemont MCU, based on Intel Pentium CPU.
22288 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22291 Intel Pentium Pro CPU@.
22294 When used with @option{-march}, the Pentium Pro
22295 instruction set is used, so the code runs on all i686 family chips.
22296 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22299 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22304 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22308 Intel Pentium M; low-power version of Intel Pentium III CPU
22309 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22313 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22316 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22320 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22321 SSE2 and SSE3 instruction set support.
22324 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22325 instruction set support.
22328 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22329 SSE4.1, SSE4.2 and POPCNT instruction set support.
22332 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22333 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22336 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22337 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22340 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22341 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22342 instruction set support.
22345 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22346 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22347 BMI, BMI2 and F16C instruction set support.
22350 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22351 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22352 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22355 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22356 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22357 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
22358 XSAVES instruction set support.
22361 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22362 instruction set support.
22365 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22366 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22369 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22370 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22371 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
22372 AVX512CD instruction set support.
22374 @item skylake-avx512
22375 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22376 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22377 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
22378 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
22381 AMD K6 CPU with MMX instruction set support.
22385 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
22388 @itemx athlon-tbird
22389 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22395 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22396 instruction set support.
22402 Processors based on the AMD K8 core with x86-64 instruction set support,
22403 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22404 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22405 instruction set extensions.)
22408 @itemx opteron-sse3
22409 @itemx athlon64-sse3
22410 Improved versions of AMD K8 cores with SSE3 instruction set support.
22414 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22415 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22416 instruction set extensions.)
22419 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22420 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22421 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22423 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22424 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22425 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22428 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22429 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22430 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22431 64-bit instruction set extensions.
22433 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22434 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22435 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22436 SSE4.2, ABM and 64-bit instruction set extensions.
22439 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
22440 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
22441 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
22442 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
22443 instruction set extensions.
22446 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22447 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22448 instruction set extensions.)
22451 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22452 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22453 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22456 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22460 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22461 instruction set support.
22464 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22465 implemented for this chip.)
22468 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22470 implemented for this chip.)
22473 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22476 @item -mtune=@var{cpu-type}
22478 Tune to @var{cpu-type} everything applicable about the generated code, except
22479 for the ABI and the set of available instructions.
22480 While picking a specific @var{cpu-type} schedules things appropriately
22481 for that particular chip, the compiler does not generate any code that
22482 cannot run on the default machine type unless you use a
22483 @option{-march=@var{cpu-type}} option.
22484 For example, if GCC is configured for i686-pc-linux-gnu
22485 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22486 but still runs on i686 machines.
22488 The choices for @var{cpu-type} are the same as for @option{-march}.
22489 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22493 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22494 If you know the CPU on which your code will run, then you should use
22495 the corresponding @option{-mtune} or @option{-march} option instead of
22496 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22497 of your application will have, then you should use this option.
22499 As new processors are deployed in the marketplace, the behavior of this
22500 option will change. Therefore, if you upgrade to a newer version of
22501 GCC, code generation controlled by this option will change to reflect
22503 that are most common at the time that version of GCC is released.
22505 There is no @option{-march=generic} option because @option{-march}
22506 indicates the instruction set the compiler can use, and there is no
22507 generic instruction set applicable to all processors. In contrast,
22508 @option{-mtune} indicates the processor (or, in this case, collection of
22509 processors) for which the code is optimized.
22512 Produce code optimized for the most current Intel processors, which are
22513 Haswell and Silvermont for this version of GCC. If you know the CPU
22514 on which your code will run, then you should use the corresponding
22515 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22516 But, if you want your application performs better on both Haswell and
22517 Silvermont, then you should use this option.
22519 As new Intel processors are deployed in the marketplace, the behavior of
22520 this option will change. Therefore, if you upgrade to a newer version of
22521 GCC, code generation controlled by this option will change to reflect
22522 the most current Intel processors at the time that version of GCC is
22525 There is no @option{-march=intel} option because @option{-march} indicates
22526 the instruction set the compiler can use, and there is no common
22527 instruction set applicable to all processors. In contrast,
22528 @option{-mtune} indicates the processor (or, in this case, collection of
22529 processors) for which the code is optimized.
22532 @item -mcpu=@var{cpu-type}
22534 A deprecated synonym for @option{-mtune}.
22536 @item -mfpmath=@var{unit}
22538 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22539 for @var{unit} are:
22543 Use the standard 387 floating-point coprocessor present on the majority of chips and
22544 emulated otherwise. Code compiled with this option runs almost everywhere.
22545 The temporary results are computed in 80-bit precision instead of the precision
22546 specified by the type, resulting in slightly different results compared to most
22547 of other chips. See @option{-ffloat-store} for more detailed description.
22549 This is the default choice for x86-32 targets.
22552 Use scalar floating-point instructions present in the SSE instruction set.
22553 This instruction set is supported by Pentium III and newer chips,
22554 and in the AMD line
22555 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22556 instruction set supports only single-precision arithmetic, thus the double and
22557 extended-precision arithmetic are still done using 387. A later version, present
22558 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22561 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22562 or @option{-msse2} switches to enable SSE extensions and make this option
22563 effective. For the x86-64 compiler, these extensions are enabled by default.
22565 The resulting code should be considerably faster in the majority of cases and avoid
22566 the numerical instability problems of 387 code, but may break some existing
22567 code that expects temporaries to be 80 bits.
22569 This is the default choice for the x86-64 compiler.
22574 Attempt to utilize both instruction sets at once. This effectively doubles the
22575 amount of available registers, and on chips with separate execution units for
22576 387 and SSE the execution resources too. Use this option with care, as it is
22577 still experimental, because the GCC register allocator does not model separate
22578 functional units well, resulting in unstable performance.
22581 @item -masm=@var{dialect}
22582 @opindex masm=@var{dialect}
22583 Output assembly instructions using selected @var{dialect}. Also affects
22584 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22585 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22586 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22587 not support @samp{intel}.
22590 @itemx -mno-ieee-fp
22592 @opindex mno-ieee-fp
22593 Control whether or not the compiler uses IEEE floating-point
22594 comparisons. These correctly handle the case where the result of a
22595 comparison is unordered.
22598 @opindex msoft-float
22599 Generate output containing library calls for floating point.
22601 @strong{Warning:} the requisite libraries are not part of GCC@.
22602 Normally the facilities of the machine's usual C compiler are used, but
22603 this can't be done directly in cross-compilation. You must make your
22604 own arrangements to provide suitable library functions for
22607 On machines where a function returns floating-point results in the 80387
22608 register stack, some floating-point opcodes may be emitted even if
22609 @option{-msoft-float} is used.
22611 @item -mno-fp-ret-in-387
22612 @opindex mno-fp-ret-in-387
22613 Do not use the FPU registers for return values of functions.
22615 The usual calling convention has functions return values of types
22616 @code{float} and @code{double} in an FPU register, even if there
22617 is no FPU@. The idea is that the operating system should emulate
22620 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22621 in ordinary CPU registers instead.
22623 @item -mno-fancy-math-387
22624 @opindex mno-fancy-math-387
22625 Some 387 emulators do not support the @code{sin}, @code{cos} and
22626 @code{sqrt} instructions for the 387. Specify this option to avoid
22627 generating those instructions. This option is the default on
22628 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22629 indicates that the target CPU always has an FPU and so the
22630 instruction does not need emulation. These
22631 instructions are not generated unless you also use the
22632 @option{-funsafe-math-optimizations} switch.
22634 @item -malign-double
22635 @itemx -mno-align-double
22636 @opindex malign-double
22637 @opindex mno-align-double
22638 Control whether GCC aligns @code{double}, @code{long double}, and
22639 @code{long long} variables on a two-word boundary or a one-word
22640 boundary. Aligning @code{double} variables on a two-word boundary
22641 produces code that runs somewhat faster on a Pentium at the
22642 expense of more memory.
22644 On x86-64, @option{-malign-double} is enabled by default.
22646 @strong{Warning:} if you use the @option{-malign-double} switch,
22647 structures containing the above types are aligned differently than
22648 the published application binary interface specifications for the x86-32
22649 and are not binary compatible with structures in code compiled
22650 without that switch.
22652 @item -m96bit-long-double
22653 @itemx -m128bit-long-double
22654 @opindex m96bit-long-double
22655 @opindex m128bit-long-double
22656 These switches control the size of @code{long double} type. The x86-32
22657 application binary interface specifies the size to be 96 bits,
22658 so @option{-m96bit-long-double} is the default in 32-bit mode.
22660 Modern architectures (Pentium and newer) prefer @code{long double}
22661 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22662 conforming to the ABI, this is not possible. So specifying
22663 @option{-m128bit-long-double} aligns @code{long double}
22664 to a 16-byte boundary by padding the @code{long double} with an additional
22667 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22668 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22670 Notice that neither of these options enable any extra precision over the x87
22671 standard of 80 bits for a @code{long double}.
22673 @strong{Warning:} if you override the default value for your target ABI, this
22674 changes the size of
22675 structures and arrays containing @code{long double} variables,
22676 as well as modifying the function calling convention for functions taking
22677 @code{long double}. Hence they are not binary-compatible
22678 with code compiled without that switch.
22680 @item -mlong-double-64
22681 @itemx -mlong-double-80
22682 @itemx -mlong-double-128
22683 @opindex mlong-double-64
22684 @opindex mlong-double-80
22685 @opindex mlong-double-128
22686 These switches control the size of @code{long double} type. A size
22687 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22688 type. This is the default for 32-bit Bionic C library. A size
22689 of 128 bits makes the @code{long double} type equivalent to the
22690 @code{__float128} type. This is the default for 64-bit Bionic C library.
22692 @strong{Warning:} if you override the default value for your target ABI, this
22693 changes the size of
22694 structures and arrays containing @code{long double} variables,
22695 as well as modifying the function calling convention for functions taking
22696 @code{long double}. Hence they are not binary-compatible
22697 with code compiled without that switch.
22699 @item -malign-data=@var{type}
22700 @opindex malign-data
22701 Control how GCC aligns variables. Supported values for @var{type} are
22702 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22703 and earlier, @samp{abi} uses alignment value as specified by the
22704 psABI, and @samp{cacheline} uses increased alignment value to match
22705 the cache line size. @samp{compat} is the default.
22707 @item -mlarge-data-threshold=@var{threshold}
22708 @opindex mlarge-data-threshold
22709 When @option{-mcmodel=medium} is specified, data objects larger than
22710 @var{threshold} are placed in the large data section. This value must be the
22711 same across all objects linked into the binary, and defaults to 65535.
22715 Use a different function-calling convention, in which functions that
22716 take a fixed number of arguments return with the @code{ret @var{num}}
22717 instruction, which pops their arguments while returning. This saves one
22718 instruction in the caller since there is no need to pop the arguments
22721 You can specify that an individual function is called with this calling
22722 sequence with the function attribute @code{stdcall}. You can also
22723 override the @option{-mrtd} option by using the function attribute
22724 @code{cdecl}. @xref{Function Attributes}.
22726 @strong{Warning:} this calling convention is incompatible with the one
22727 normally used on Unix, so you cannot use it if you need to call
22728 libraries compiled with the Unix compiler.
22730 Also, you must provide function prototypes for all functions that
22731 take variable numbers of arguments (including @code{printf});
22732 otherwise incorrect code is generated for calls to those
22735 In addition, seriously incorrect code results if you call a
22736 function with too many arguments. (Normally, extra arguments are
22737 harmlessly ignored.)
22739 @item -mregparm=@var{num}
22741 Control how many registers are used to pass integer arguments. By
22742 default, no registers are used to pass arguments, and at most 3
22743 registers can be used. You can control this behavior for a specific
22744 function by using the function attribute @code{regparm}.
22745 @xref{Function Attributes}.
22747 @strong{Warning:} if you use this switch, and
22748 @var{num} is nonzero, then you must build all modules with the same
22749 value, including any libraries. This includes the system libraries and
22753 @opindex msseregparm
22754 Use SSE register passing conventions for float and double arguments
22755 and return values. You can control this behavior for a specific
22756 function by using the function attribute @code{sseregparm}.
22757 @xref{Function Attributes}.
22759 @strong{Warning:} if you use this switch then you must build all
22760 modules with the same value, including any libraries. This includes
22761 the system libraries and startup modules.
22763 @item -mvect8-ret-in-mem
22764 @opindex mvect8-ret-in-mem
22765 Return 8-byte vectors in memory instead of MMX registers. This is the
22766 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22767 Studio compilers until version 12. Later compiler versions (starting
22768 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22769 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22770 you need to remain compatible with existing code produced by those
22771 previous compiler versions or older versions of GCC@.
22780 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22781 is specified, the significands of results of floating-point operations are
22782 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22783 significands of results of floating-point operations to 53 bits (double
22784 precision) and @option{-mpc80} rounds the significands of results of
22785 floating-point operations to 64 bits (extended double precision), which is
22786 the default. When this option is used, floating-point operations in higher
22787 precisions are not available to the programmer without setting the FPU
22788 control word explicitly.
22790 Setting the rounding of floating-point operations to less than the default
22791 80 bits can speed some programs by 2% or more. Note that some mathematical
22792 libraries assume that extended-precision (80-bit) floating-point operations
22793 are enabled by default; routines in such libraries could suffer significant
22794 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22795 when this option is used to set the precision to less than extended precision.
22797 @item -mstackrealign
22798 @opindex mstackrealign
22799 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22800 option generates an alternate prologue and epilogue that realigns the
22801 run-time stack if necessary. This supports mixing legacy codes that keep
22802 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22803 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22804 applicable to individual functions.
22806 @item -mpreferred-stack-boundary=@var{num}
22807 @opindex mpreferred-stack-boundary
22808 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22809 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22810 the default is 4 (16 bytes or 128 bits).
22812 @strong{Warning:} When generating code for the x86-64 architecture with
22813 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22814 used to keep the stack boundary aligned to 8 byte boundary. Since
22815 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22816 intended to be used in controlled environment where stack space is
22817 important limitation. This option leads to wrong code when functions
22818 compiled with 16 byte stack alignment (such as functions from a standard
22819 library) are called with misaligned stack. In this case, SSE
22820 instructions may lead to misaligned memory access traps. In addition,
22821 variable arguments are handled incorrectly for 16 byte aligned
22822 objects (including x87 long double and __int128), leading to wrong
22823 results. You must build all modules with
22824 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22825 includes the system libraries and startup modules.
22827 @item -mincoming-stack-boundary=@var{num}
22828 @opindex mincoming-stack-boundary
22829 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22830 boundary. If @option{-mincoming-stack-boundary} is not specified,
22831 the one specified by @option{-mpreferred-stack-boundary} is used.
22833 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22834 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22835 suffer significant run time performance penalties. On Pentium III, the
22836 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22837 properly if it is not 16-byte aligned.
22839 To ensure proper alignment of this values on the stack, the stack boundary
22840 must be as aligned as that required by any value stored on the stack.
22841 Further, every function must be generated such that it keeps the stack
22842 aligned. Thus calling a function compiled with a higher preferred
22843 stack boundary from a function compiled with a lower preferred stack
22844 boundary most likely misaligns the stack. It is recommended that
22845 libraries that use callbacks always use the default setting.
22847 This extra alignment does consume extra stack space, and generally
22848 increases code size. Code that is sensitive to stack space usage, such
22849 as embedded systems and operating system kernels, may want to reduce the
22850 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22907 @itemx -mavx512ifma
22908 @opindex mavx512ifma
22910 @itemx -mavx512vbmi
22911 @opindex mavx512vbmi
22923 @opindex mclfushopt
22943 @itemx -mprefetchwt1
22944 @opindex mprefetchwt1
22995 These switches enable the use of instructions in the MMX, SSE,
22996 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22997 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22998 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
22999 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX or 3DNow!@:
23000 extended instruction sets. Each has a corresponding @option{-mno-} option
23001 to disable use of these instructions.
23003 These extensions are also available as built-in functions: see
23004 @ref{x86 Built-in Functions}, for details of the functions enabled and
23005 disabled by these switches.
23007 To generate SSE/SSE2 instructions automatically from floating-point
23008 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
23010 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
23011 generates new AVX instructions or AVX equivalence for all SSEx instructions
23014 These options enable GCC to use these extended instructions in
23015 generated code, even without @option{-mfpmath=sse}. Applications that
23016 perform run-time CPU detection must compile separate files for each
23017 supported architecture, using the appropriate flags. In particular,
23018 the file containing the CPU detection code should be compiled without
23021 @item -mdump-tune-features
23022 @opindex mdump-tune-features
23023 This option instructs GCC to dump the names of the x86 performance
23024 tuning features and default settings. The names can be used in
23025 @option{-mtune-ctrl=@var{feature-list}}.
23027 @item -mtune-ctrl=@var{feature-list}
23028 @opindex mtune-ctrl=@var{feature-list}
23029 This option is used to do fine grain control of x86 code generation features.
23030 @var{feature-list} is a comma separated list of @var{feature} names. See also
23031 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
23032 on if it is not preceded with @samp{^}, otherwise, it is turned off.
23033 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
23034 developers. Using it may lead to code paths not covered by testing and can
23035 potentially result in compiler ICEs or runtime errors.
23038 @opindex mno-default
23039 This option instructs GCC to turn off all tunable features. See also
23040 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23044 This option instructs GCC to emit a @code{cld} instruction in the prologue
23045 of functions that use string instructions. String instructions depend on
23046 the DF flag to select between autoincrement or autodecrement mode. While the
23047 ABI specifies the DF flag to be cleared on function entry, some operating
23048 systems violate this specification by not clearing the DF flag in their
23049 exception dispatchers. The exception handler can be invoked with the DF flag
23050 set, which leads to wrong direction mode when string instructions are used.
23051 This option can be enabled by default on 32-bit x86 targets by configuring
23052 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23053 instructions can be suppressed with the @option{-mno-cld} compiler option
23057 @opindex mvzeroupper
23058 This option instructs GCC to emit a @code{vzeroupper} instruction
23059 before a transfer of control flow out of the function to minimize
23060 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23063 @item -mprefer-avx128
23064 @opindex mprefer-avx128
23065 This option instructs GCC to use 128-bit AVX instructions instead of
23066 256-bit AVX instructions in the auto-vectorizer.
23070 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23071 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23072 (or oword) data types.
23073 This is useful for high-resolution counters that can be updated
23074 by multiple processors (or cores). This instruction is generated as part of
23075 atomic built-in functions: see @ref{__sync Builtins} or
23076 @ref{__atomic Builtins} for details.
23080 This option enables generation of @code{SAHF} instructions in 64-bit code.
23081 Early Intel Pentium 4 CPUs with Intel 64 support,
23082 prior to the introduction of Pentium 4 G1 step in December 2005,
23083 lacked the @code{LAHF} and @code{SAHF} instructions
23084 which are supported by AMD64.
23085 These are load and store instructions, respectively, for certain status flags.
23086 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23087 @code{drem}, and @code{remainder} built-in functions;
23088 see @ref{Other Builtins} for details.
23092 This option enables use of the @code{movbe} instruction to implement
23093 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23097 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23098 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23099 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23103 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23104 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23105 with an additional Newton-Raphson step
23106 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23107 (and their vectorized
23108 variants) for single-precision floating-point arguments. These instructions
23109 are generated only when @option{-funsafe-math-optimizations} is enabled
23110 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
23111 Note that while the throughput of the sequence is higher than the throughput
23112 of the non-reciprocal instruction, the precision of the sequence can be
23113 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23115 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23116 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23117 combination), and doesn't need @option{-mrecip}.
23119 Also note that GCC emits the above sequence with additional Newton-Raphson step
23120 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23121 already with @option{-ffast-math} (or the above option combination), and
23122 doesn't need @option{-mrecip}.
23124 @item -mrecip=@var{opt}
23125 @opindex mrecip=opt
23126 This option controls which reciprocal estimate instructions
23127 may be used. @var{opt} is a comma-separated list of options, which may
23128 be preceded by a @samp{!} to invert the option:
23132 Enable all estimate instructions.
23135 Enable the default instructions, equivalent to @option{-mrecip}.
23138 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23141 Enable the approximation for scalar division.
23144 Enable the approximation for vectorized division.
23147 Enable the approximation for scalar square root.
23150 Enable the approximation for vectorized square root.
23153 So, for example, @option{-mrecip=all,!sqrt} enables
23154 all of the reciprocal approximations, except for square root.
23156 @item -mveclibabi=@var{type}
23157 @opindex mveclibabi
23158 Specifies the ABI type to use for vectorizing intrinsics using an
23159 external library. Supported values for @var{type} are @samp{svml}
23160 for the Intel short
23161 vector math library and @samp{acml} for the AMD math core library.
23162 To use this option, both @option{-ftree-vectorize} and
23163 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23164 ABI-compatible library must be specified at link time.
23166 GCC currently emits calls to @code{vmldExp2},
23167 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23168 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23169 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23170 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23171 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23172 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23173 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23174 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23175 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23176 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23177 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23178 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23179 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23180 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23181 when @option{-mveclibabi=acml} is used.
23183 @item -mabi=@var{name}
23185 Generate code for the specified calling convention. Permissible values
23186 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23187 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23188 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23189 You can control this behavior for specific functions by
23190 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23191 @xref{Function Attributes}.
23193 @item -mtls-dialect=@var{type}
23194 @opindex mtls-dialect
23195 Generate code to access thread-local storage using the @samp{gnu} or
23196 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23197 @samp{gnu2} is more efficient, but it may add compile- and run-time
23198 requirements that cannot be satisfied on all systems.
23201 @itemx -mno-push-args
23202 @opindex mpush-args
23203 @opindex mno-push-args
23204 Use PUSH operations to store outgoing parameters. This method is shorter
23205 and usually equally fast as method using SUB/MOV operations and is enabled
23206 by default. In some cases disabling it may improve performance because of
23207 improved scheduling and reduced dependencies.
23209 @item -maccumulate-outgoing-args
23210 @opindex maccumulate-outgoing-args
23211 If enabled, the maximum amount of space required for outgoing arguments is
23212 computed in the function prologue. This is faster on most modern CPUs
23213 because of reduced dependencies, improved scheduling and reduced stack usage
23214 when the preferred stack boundary is not equal to 2. The drawback is a notable
23215 increase in code size. This switch implies @option{-mno-push-args}.
23219 Support thread-safe exception handling on MinGW. Programs that rely
23220 on thread-safe exception handling must compile and link all code with the
23221 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23222 @option{-D_MT}; when linking, it links in a special thread helper library
23223 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23225 @item -mno-align-stringops
23226 @opindex mno-align-stringops
23227 Do not align the destination of inlined string operations. This switch reduces
23228 code size and improves performance in case the destination is already aligned,
23229 but GCC doesn't know about it.
23231 @item -minline-all-stringops
23232 @opindex minline-all-stringops
23233 By default GCC inlines string operations only when the destination is
23234 known to be aligned to least a 4-byte boundary.
23235 This enables more inlining and increases code
23236 size, but may improve performance of code that depends on fast
23237 @code{memcpy}, @code{strlen},
23238 and @code{memset} for short lengths.
23240 @item -minline-stringops-dynamically
23241 @opindex minline-stringops-dynamically
23242 For string operations of unknown size, use run-time checks with
23243 inline code for small blocks and a library call for large blocks.
23245 @item -mstringop-strategy=@var{alg}
23246 @opindex mstringop-strategy=@var{alg}
23247 Override the internal decision heuristic for the particular algorithm to use
23248 for inlining string operations. The allowed values for @var{alg} are:
23254 Expand using i386 @code{rep} prefix of the specified size.
23258 @itemx unrolled_loop
23259 Expand into an inline loop.
23262 Always use a library call.
23265 @item -mmemcpy-strategy=@var{strategy}
23266 @opindex mmemcpy-strategy=@var{strategy}
23267 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23268 should be inlined and what inline algorithm to use when the expected size
23269 of the copy operation is known. @var{strategy}
23270 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23271 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23272 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23273 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23274 in the list must be specified in increasing order. The minimal byte size for
23275 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23278 @item -mmemset-strategy=@var{strategy}
23279 @opindex mmemset-strategy=@var{strategy}
23280 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23281 @code{__builtin_memset} expansion.
23283 @item -momit-leaf-frame-pointer
23284 @opindex momit-leaf-frame-pointer
23285 Don't keep the frame pointer in a register for leaf functions. This
23286 avoids the instructions to save, set up, and restore frame pointers and
23287 makes an extra register available in leaf functions. The option
23288 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23289 which might make debugging harder.
23291 @item -mtls-direct-seg-refs
23292 @itemx -mno-tls-direct-seg-refs
23293 @opindex mtls-direct-seg-refs
23294 Controls whether TLS variables may be accessed with offsets from the
23295 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23296 or whether the thread base pointer must be added. Whether or not this
23297 is valid depends on the operating system, and whether it maps the
23298 segment to cover the entire TLS area.
23300 For systems that use the GNU C Library, the default is on.
23303 @itemx -mno-sse2avx
23305 Specify that the assembler should encode SSE instructions with VEX
23306 prefix. The option @option{-mavx} turns this on by default.
23311 If profiling is active (@option{-pg}), put the profiling
23312 counter call before the prologue.
23313 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23314 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23316 @item -mrecord-mcount
23317 @itemx -mno-record-mcount
23318 @opindex mrecord-mcount
23319 If profiling is active (@option{-pg}), generate a __mcount_loc section
23320 that contains pointers to each profiling call. This is useful for
23321 automatically patching and out calls.
23324 @itemx -mno-nop-mcount
23325 @opindex mnop-mcount
23326 If profiling is active (@option{-pg}), generate the calls to
23327 the profiling functions as nops. This is useful when they
23328 should be patched in later dynamically. This is likely only
23329 useful together with @option{-mrecord-mcount}.
23331 @item -mskip-rax-setup
23332 @itemx -mno-skip-rax-setup
23333 @opindex mskip-rax-setup
23334 When generating code for the x86-64 architecture with SSE extensions
23335 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23336 register when there are no variable arguments passed in vector registers.
23338 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23339 saving vector registers on stack when passing variable arguments, the
23340 impacts of this option are callees may waste some stack space,
23341 misbehave or jump to a random location. GCC 4.4 or newer don't have
23342 those issues, regardless the RAX register value.
23345 @itemx -mno-8bit-idiv
23346 @opindex m8bit-idiv
23347 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23348 much faster than 32-bit/64-bit integer divide. This option generates a
23349 run-time check. If both dividend and divisor are within range of 0
23350 to 255, 8-bit unsigned integer divide is used instead of
23351 32-bit/64-bit integer divide.
23353 @item -mavx256-split-unaligned-load
23354 @itemx -mavx256-split-unaligned-store
23355 @opindex mavx256-split-unaligned-load
23356 @opindex mavx256-split-unaligned-store
23357 Split 32-byte AVX unaligned load and store.
23359 @item -mstack-protector-guard=@var{guard}
23360 @opindex mstack-protector-guard=@var{guard}
23361 Generate stack protection code using canary at @var{guard}. Supported
23362 locations are @samp{global} for global canary or @samp{tls} for per-thread
23363 canary in the TLS block (the default). This option has effect only when
23364 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23368 These @samp{-m} switches are supported in addition to the above
23369 on x86-64 processors in 64-bit environments.
23382 Generate code for a 16-bit, 32-bit or 64-bit environment.
23383 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23385 generates code that runs on any i386 system.
23387 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23388 types to 64 bits, and generates code for the x86-64 architecture.
23389 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23390 and @option{-mdynamic-no-pic} options.
23392 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23394 generates code for the x86-64 architecture.
23396 The @option{-m16} option is the same as @option{-m32}, except for that
23397 it outputs the @code{.code16gcc} assembly directive at the beginning of
23398 the assembly output so that the binary can run in 16-bit mode.
23400 The @option{-miamcu} option generates code which conforms to Intel MCU
23401 psABI. It requires the @option{-m32} option to be turned on.
23403 @item -mno-red-zone
23404 @opindex mno-red-zone
23405 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23406 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23407 stack pointer that is not modified by signal or interrupt handlers
23408 and therefore can be used for temporary data without adjusting the stack
23409 pointer. The flag @option{-mno-red-zone} disables this red zone.
23411 @item -mcmodel=small
23412 @opindex mcmodel=small
23413 Generate code for the small code model: the program and its symbols must
23414 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23415 Programs can be statically or dynamically linked. This is the default
23418 @item -mcmodel=kernel
23419 @opindex mcmodel=kernel
23420 Generate code for the kernel code model. The kernel runs in the
23421 negative 2 GB of the address space.
23422 This model has to be used for Linux kernel code.
23424 @item -mcmodel=medium
23425 @opindex mcmodel=medium
23426 Generate code for the medium model: the program is linked in the lower 2
23427 GB of the address space. Small symbols are also placed there. Symbols
23428 with sizes larger than @option{-mlarge-data-threshold} are put into
23429 large data or BSS sections and can be located above 2GB. Programs can
23430 be statically or dynamically linked.
23432 @item -mcmodel=large
23433 @opindex mcmodel=large
23434 Generate code for the large model. This model makes no assumptions
23435 about addresses and sizes of sections.
23437 @item -maddress-mode=long
23438 @opindex maddress-mode=long
23439 Generate code for long address mode. This is only supported for 64-bit
23440 and x32 environments. It is the default address mode for 64-bit
23443 @item -maddress-mode=short
23444 @opindex maddress-mode=short
23445 Generate code for short address mode. This is only supported for 32-bit
23446 and x32 environments. It is the default address mode for 32-bit and
23450 @node x86 Windows Options
23451 @subsection x86 Windows Options
23452 @cindex x86 Windows Options
23453 @cindex Windows Options for x86
23455 These additional options are available for Microsoft Windows targets:
23461 specifies that a console application is to be generated, by
23462 instructing the linker to set the PE header subsystem type
23463 required for console applications.
23464 This option is available for Cygwin and MinGW targets and is
23465 enabled by default on those targets.
23469 This option is available for Cygwin and MinGW targets. It
23470 specifies that a DLL---a dynamic link library---is to be
23471 generated, enabling the selection of the required runtime
23472 startup object and entry point.
23474 @item -mnop-fun-dllimport
23475 @opindex mnop-fun-dllimport
23476 This option is available for Cygwin and MinGW targets. It
23477 specifies that the @code{dllimport} attribute should be ignored.
23481 This option is available for MinGW targets. It specifies
23482 that MinGW-specific thread support is to be used.
23486 This option is available for MinGW-w64 targets. It causes
23487 the @code{UNICODE} preprocessor macro to be predefined, and
23488 chooses Unicode-capable runtime startup code.
23492 This option is available for Cygwin and MinGW targets. It
23493 specifies that the typical Microsoft Windows predefined macros are to
23494 be set in the pre-processor, but does not influence the choice
23495 of runtime library/startup code.
23499 This option is available for Cygwin and MinGW targets. It
23500 specifies that a GUI application is to be generated by
23501 instructing the linker to set the PE header subsystem type
23504 @item -fno-set-stack-executable
23505 @opindex fno-set-stack-executable
23506 This option is available for MinGW targets. It specifies that
23507 the executable flag for the stack used by nested functions isn't
23508 set. This is necessary for binaries running in kernel mode of
23509 Microsoft Windows, as there the User32 API, which is used to set executable
23510 privileges, isn't available.
23512 @item -fwritable-relocated-rdata
23513 @opindex fno-writable-relocated-rdata
23514 This option is available for MinGW and Cygwin targets. It specifies
23515 that relocated-data in read-only section is put into .data
23516 section. This is a necessary for older runtimes not supporting
23517 modification of .rdata sections for pseudo-relocation.
23519 @item -mpe-aligned-commons
23520 @opindex mpe-aligned-commons
23521 This option is available for Cygwin and MinGW targets. It
23522 specifies that the GNU extension to the PE file format that
23523 permits the correct alignment of COMMON variables should be
23524 used when generating code. It is enabled by default if
23525 GCC detects that the target assembler found during configuration
23526 supports the feature.
23529 See also under @ref{x86 Options} for standard options.
23531 @node Xstormy16 Options
23532 @subsection Xstormy16 Options
23533 @cindex Xstormy16 Options
23535 These options are defined for Xstormy16:
23540 Choose startup files and linker script suitable for the simulator.
23543 @node Xtensa Options
23544 @subsection Xtensa Options
23545 @cindex Xtensa Options
23547 These options are supported for Xtensa targets:
23551 @itemx -mno-const16
23553 @opindex mno-const16
23554 Enable or disable use of @code{CONST16} instructions for loading
23555 constant values. The @code{CONST16} instruction is currently not a
23556 standard option from Tensilica. When enabled, @code{CONST16}
23557 instructions are always used in place of the standard @code{L32R}
23558 instructions. The use of @code{CONST16} is enabled by default only if
23559 the @code{L32R} instruction is not available.
23562 @itemx -mno-fused-madd
23563 @opindex mfused-madd
23564 @opindex mno-fused-madd
23565 Enable or disable use of fused multiply/add and multiply/subtract
23566 instructions in the floating-point option. This has no effect if the
23567 floating-point option is not also enabled. Disabling fused multiply/add
23568 and multiply/subtract instructions forces the compiler to use separate
23569 instructions for the multiply and add/subtract operations. This may be
23570 desirable in some cases where strict IEEE 754-compliant results are
23571 required: the fused multiply add/subtract instructions do not round the
23572 intermediate result, thereby producing results with @emph{more} bits of
23573 precision than specified by the IEEE standard. Disabling fused multiply
23574 add/subtract instructions also ensures that the program output is not
23575 sensitive to the compiler's ability to combine multiply and add/subtract
23578 @item -mserialize-volatile
23579 @itemx -mno-serialize-volatile
23580 @opindex mserialize-volatile
23581 @opindex mno-serialize-volatile
23582 When this option is enabled, GCC inserts @code{MEMW} instructions before
23583 @code{volatile} memory references to guarantee sequential consistency.
23584 The default is @option{-mserialize-volatile}. Use
23585 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23587 @item -mforce-no-pic
23588 @opindex mforce-no-pic
23589 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23590 position-independent code (PIC), this option disables PIC for compiling
23593 @item -mtext-section-literals
23594 @itemx -mno-text-section-literals
23595 @opindex mtext-section-literals
23596 @opindex mno-text-section-literals
23597 These options control the treatment of literal pools. The default is
23598 @option{-mno-text-section-literals}, which places literals in a separate
23599 section in the output file. This allows the literal pool to be placed
23600 in a data RAM/ROM, and it also allows the linker to combine literal
23601 pools from separate object files to remove redundant literals and
23602 improve code size. With @option{-mtext-section-literals}, the literals
23603 are interspersed in the text section in order to keep them as close as
23604 possible to their references. This may be necessary for large assembly
23605 files. Literals for each function are placed right before that function.
23607 @item -mauto-litpools
23608 @itemx -mno-auto-litpools
23609 @opindex mauto-litpools
23610 @opindex mno-auto-litpools
23611 These options control the treatment of literal pools. The default is
23612 @option{-mno-auto-litpools}, which places literals in a separate
23613 section in the output file unless @option{-mtext-section-literals} is
23614 used. With @option{-mauto-litpools} the literals are interspersed in
23615 the text section by the assembler. Compiler does not produce explicit
23616 @code{.literal} directives and loads literals into registers with
23617 @code{MOVI} instructions instead of @code{L32R} to let the assembler
23618 do relaxation and place literals as necessary. This option allows
23619 assembler to create several literal pools per function and assemble
23620 very big functions, which may not be possible with
23621 @option{-mtext-section-literals}.
23623 @item -mtarget-align
23624 @itemx -mno-target-align
23625 @opindex mtarget-align
23626 @opindex mno-target-align
23627 When this option is enabled, GCC instructs the assembler to
23628 automatically align instructions to reduce branch penalties at the
23629 expense of some code density. The assembler attempts to widen density
23630 instructions to align branch targets and the instructions following call
23631 instructions. If there are not enough preceding safe density
23632 instructions to align a target, no widening is performed. The
23633 default is @option{-mtarget-align}. These options do not affect the
23634 treatment of auto-aligned instructions like @code{LOOP}, which the
23635 assembler always aligns, either by widening density instructions or
23636 by inserting NOP instructions.
23639 @itemx -mno-longcalls
23640 @opindex mlongcalls
23641 @opindex mno-longcalls
23642 When this option is enabled, GCC instructs the assembler to translate
23643 direct calls to indirect calls unless it can determine that the target
23644 of a direct call is in the range allowed by the call instruction. This
23645 translation typically occurs for calls to functions in other source
23646 files. Specifically, the assembler translates a direct @code{CALL}
23647 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23648 The default is @option{-mno-longcalls}. This option should be used in
23649 programs where the call target can potentially be out of range. This
23650 option is implemented in the assembler, not the compiler, so the
23651 assembly code generated by GCC still shows direct call
23652 instructions---look at the disassembled object code to see the actual
23653 instructions. Note that the assembler uses an indirect call for
23654 every cross-file call, not just those that really are out of range.
23657 @node zSeries Options
23658 @subsection zSeries Options
23659 @cindex zSeries options
23661 These are listed under @xref{S/390 and zSeries Options}.
23663 @node Code Gen Options
23664 @section Options for Code Generation Conventions
23665 @cindex code generation conventions
23666 @cindex options, code generation
23667 @cindex run-time options
23669 These machine-independent options control the interface conventions
23670 used in code generation.
23672 Most of them have both positive and negative forms; the negative form
23673 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23674 one of the forms is listed---the one that is not the default. You
23675 can figure out the other form by either removing @samp{no-} or adding
23679 @item -fbounds-check
23680 @opindex fbounds-check
23681 For front ends that support it, generate additional code to check that
23682 indices used to access arrays are within the declared range. This is
23683 currently only supported by the Java and Fortran front ends, where
23684 this option defaults to true and false respectively.
23686 @item -fstack-reuse=@var{reuse-level}
23687 @opindex fstack_reuse
23688 This option controls stack space reuse for user declared local/auto variables
23689 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23690 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23691 local variables and temporaries, @samp{named_vars} enables the reuse only for
23692 user defined local variables with names, and @samp{none} disables stack reuse
23693 completely. The default value is @samp{all}. The option is needed when the
23694 program extends the lifetime of a scoped local variable or a compiler generated
23695 temporary beyond the end point defined by the language. When a lifetime of
23696 a variable ends, and if the variable lives in memory, the optimizing compiler
23697 has the freedom to reuse its stack space with other temporaries or scoped
23698 local variables whose live range does not overlap with it. Legacy code extending
23699 local lifetime is likely to break with the stack reuse optimization.
23718 if (*p == 10) // out of scope use of local1
23729 A(int k) : i(k), j(k) @{ @}
23736 void foo(const A& ar)
23743 foo(A(10)); // temp object's lifetime ends when foo returns
23749 ap->i+= 10; // ap references out of scope temp whose space
23750 // is reused with a. What is the value of ap->i?
23755 The lifetime of a compiler generated temporary is well defined by the C++
23756 standard. When a lifetime of a temporary ends, and if the temporary lives
23757 in memory, the optimizing compiler has the freedom to reuse its stack
23758 space with other temporaries or scoped local variables whose live range
23759 does not overlap with it. However some of the legacy code relies on
23760 the behavior of older compilers in which temporaries' stack space is
23761 not reused, the aggressive stack reuse can lead to runtime errors. This
23762 option is used to control the temporary stack reuse optimization.
23766 This option generates traps for signed overflow on addition, subtraction,
23767 multiplication operations.
23768 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
23769 @option{-ftrapv} @option{-fwrapv} on the command-line results in
23770 @option{-fwrapv} being effective. Note that only active options override, so
23771 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
23772 results in @option{-ftrapv} being effective.
23776 This option instructs the compiler to assume that signed arithmetic
23777 overflow of addition, subtraction and multiplication wraps around
23778 using twos-complement representation. This flag enables some optimizations
23779 and disables others. This option is enabled by default for the Java
23780 front end, as required by the Java language specification.
23781 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
23782 @option{-ftrapv} @option{-fwrapv} on the command-line results in
23783 @option{-fwrapv} being effective. Note that only active options override, so
23784 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
23785 results in @option{-ftrapv} being effective.
23788 @opindex fexceptions
23789 Enable exception handling. Generates extra code needed to propagate
23790 exceptions. For some targets, this implies GCC generates frame
23791 unwind information for all functions, which can produce significant data
23792 size overhead, although it does not affect execution. If you do not
23793 specify this option, GCC enables it by default for languages like
23794 C++ that normally require exception handling, and disables it for
23795 languages like C that do not normally require it. However, you may need
23796 to enable this option when compiling C code that needs to interoperate
23797 properly with exception handlers written in C++. You may also wish to
23798 disable this option if you are compiling older C++ programs that don't
23799 use exception handling.
23801 @item -fnon-call-exceptions
23802 @opindex fnon-call-exceptions
23803 Generate code that allows trapping instructions to throw exceptions.
23804 Note that this requires platform-specific runtime support that does
23805 not exist everywhere. Moreover, it only allows @emph{trapping}
23806 instructions to throw exceptions, i.e.@: memory references or floating-point
23807 instructions. It does not allow exceptions to be thrown from
23808 arbitrary signal handlers such as @code{SIGALRM}.
23810 @item -fdelete-dead-exceptions
23811 @opindex fdelete-dead-exceptions
23812 Consider that instructions that may throw exceptions but don't otherwise
23813 contribute to the execution of the program can be optimized away.
23814 This option is enabled by default for the Ada front end, as permitted by
23815 the Ada language specification.
23816 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23818 @item -funwind-tables
23819 @opindex funwind-tables
23820 Similar to @option{-fexceptions}, except that it just generates any needed
23821 static data, but does not affect the generated code in any other way.
23822 You normally do not need to enable this option; instead, a language processor
23823 that needs this handling enables it on your behalf.
23825 @item -fasynchronous-unwind-tables
23826 @opindex fasynchronous-unwind-tables
23827 Generate unwind table in DWARF 2 format, if supported by target machine. The
23828 table is exact at each instruction boundary, so it can be used for stack
23829 unwinding from asynchronous events (such as debugger or garbage collector).
23831 @item -fno-gnu-unique
23832 @opindex fno-gnu-unique
23833 On systems with recent GNU assembler and C library, the C++ compiler
23834 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23835 of template static data members and static local variables in inline
23836 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23837 is necessary to avoid problems with a library used by two different
23838 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23839 therefore disagreeing with the other one about the binding of the
23840 symbol. But this causes @code{dlclose} to be ignored for affected
23841 DSOs; if your program relies on reinitialization of a DSO via
23842 @code{dlclose} and @code{dlopen}, you can use
23843 @option{-fno-gnu-unique}.
23845 @item -fpcc-struct-return
23846 @opindex fpcc-struct-return
23847 Return ``short'' @code{struct} and @code{union} values in memory like
23848 longer ones, rather than in registers. This convention is less
23849 efficient, but it has the advantage of allowing intercallability between
23850 GCC-compiled files and files compiled with other compilers, particularly
23851 the Portable C Compiler (pcc).
23853 The precise convention for returning structures in memory depends
23854 on the target configuration macros.
23856 Short structures and unions are those whose size and alignment match
23857 that of some integer type.
23859 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23860 switch is not binary compatible with code compiled with the
23861 @option{-freg-struct-return} switch.
23862 Use it to conform to a non-default application binary interface.
23864 @item -freg-struct-return
23865 @opindex freg-struct-return
23866 Return @code{struct} and @code{union} values in registers when possible.
23867 This is more efficient for small structures than
23868 @option{-fpcc-struct-return}.
23870 If you specify neither @option{-fpcc-struct-return} nor
23871 @option{-freg-struct-return}, GCC defaults to whichever convention is
23872 standard for the target. If there is no standard convention, GCC
23873 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23874 the principal compiler. In those cases, we can choose the standard, and
23875 we chose the more efficient register return alternative.
23877 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23878 switch is not binary compatible with code compiled with the
23879 @option{-fpcc-struct-return} switch.
23880 Use it to conform to a non-default application binary interface.
23882 @item -fshort-enums
23883 @opindex fshort-enums
23884 Allocate to an @code{enum} type only as many bytes as it needs for the
23885 declared range of possible values. Specifically, the @code{enum} type
23886 is equivalent to the smallest integer type that has enough room.
23888 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23889 code that is not binary compatible with code generated without that switch.
23890 Use it to conform to a non-default application binary interface.
23892 @item -fshort-double
23893 @opindex fshort-double
23894 Use the same size for @code{double} as for @code{float}.
23896 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23897 code that is not binary compatible with code generated without that switch.
23898 Use it to conform to a non-default application binary interface.
23900 @item -fshort-wchar
23901 @opindex fshort-wchar
23902 Override the underlying type for @code{wchar_t} to be @code{short
23903 unsigned int} instead of the default for the target. This option is
23904 useful for building programs to run under WINE@.
23906 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23907 code that is not binary compatible with code generated without that switch.
23908 Use it to conform to a non-default application binary interface.
23911 @opindex fno-common
23912 In C code, controls the placement of uninitialized global variables.
23913 Unix C compilers have traditionally permitted multiple definitions of
23914 such variables in different compilation units by placing the variables
23916 This is the behavior specified by @option{-fcommon}, and is the default
23917 for GCC on most targets.
23918 On the other hand, this behavior is not required by ISO C, and on some
23919 targets may carry a speed or code size penalty on variable references.
23920 The @option{-fno-common} option specifies that the compiler should place
23921 uninitialized global variables in the data section of the object file,
23922 rather than generating them as common blocks.
23923 This has the effect that if the same variable is declared
23924 (without @code{extern}) in two different compilations,
23925 you get a multiple-definition error when you link them.
23926 In this case, you must compile with @option{-fcommon} instead.
23927 Compiling with @option{-fno-common} is useful on targets for which
23928 it provides better performance, or if you wish to verify that the
23929 program will work on other systems that always treat uninitialized
23930 variable declarations this way.
23934 Ignore the @code{#ident} directive.
23936 @item -finhibit-size-directive
23937 @opindex finhibit-size-directive
23938 Don't output a @code{.size} assembler directive, or anything else that
23939 would cause trouble if the function is split in the middle, and the
23940 two halves are placed at locations far apart in memory. This option is
23941 used when compiling @file{crtstuff.c}; you should not need to use it
23944 @item -fverbose-asm
23945 @opindex fverbose-asm
23946 Put extra commentary information in the generated assembly code to
23947 make it more readable. This option is generally only of use to those
23948 who actually need to read the generated assembly code (perhaps while
23949 debugging the compiler itself).
23951 @option{-fno-verbose-asm}, the default, causes the
23952 extra information to be omitted and is useful when comparing two assembler
23955 @item -frecord-gcc-switches
23956 @opindex frecord-gcc-switches
23957 This switch causes the command line used to invoke the
23958 compiler to be recorded into the object file that is being created.
23959 This switch is only implemented on some targets and the exact format
23960 of the recording is target and binary file format dependent, but it
23961 usually takes the form of a section containing ASCII text. This
23962 switch is related to the @option{-fverbose-asm} switch, but that
23963 switch only records information in the assembler output file as
23964 comments, so it never reaches the object file.
23965 See also @option{-grecord-gcc-switches} for another
23966 way of storing compiler options into the object file.
23970 @cindex global offset table
23972 Generate position-independent code (PIC) suitable for use in a shared
23973 library, if supported for the target machine. Such code accesses all
23974 constant addresses through a global offset table (GOT)@. The dynamic
23975 loader resolves the GOT entries when the program starts (the dynamic
23976 loader is not part of GCC; it is part of the operating system). If
23977 the GOT size for the linked executable exceeds a machine-specific
23978 maximum size, you get an error message from the linker indicating that
23979 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23980 instead. (These maximums are 8k on the SPARC and 32k
23981 on the m68k and RS/6000. The x86 has no such limit.)
23983 Position-independent code requires special support, and therefore works
23984 only on certain machines. For the x86, GCC supports PIC for System V
23985 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23986 position-independent.
23988 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23993 If supported for the target machine, emit position-independent code,
23994 suitable for dynamic linking and avoiding any limit on the size of the
23995 global offset table. This option makes a difference on the m68k,
23996 PowerPC and SPARC@.
23998 Position-independent code requires special support, and therefore works
23999 only on certain machines.
24001 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
24008 These options are similar to @option{-fpic} and @option{-fPIC}, but
24009 generated position independent code can be only linked into executables.
24010 Usually these options are used when @option{-pie} GCC option is
24011 used during linking.
24013 @option{-fpie} and @option{-fPIE} both define the macros
24014 @code{__pie__} and @code{__PIE__}. The macros have the value 1
24015 for @option{-fpie} and 2 for @option{-fPIE}.
24019 Do not use PLT for external function calls in position-independent code.
24020 Instead, load callee address at call site from GOT and branch to it.
24021 This leads to more efficient code by eliminating PLT stubs and exposing
24022 GOT load to optimizations. On architectures such as 32-bit x86 where
24023 PLT stubs expect GOT pointer in a specific register, this gives more
24024 register allocation freedom to the compiler. Lazy binding requires PLT:
24025 with @option{-fno-plt} all external symbols are resolved at load time.
24027 Alternatively, function attribute @code{noplt} can be used to avoid PLT
24028 for calls to specific external functions by marking those functions with
24031 Additionally, a few targets also convert calls to those functions that are
24032 marked to not use the PLT to use the GOT instead for non-position independent
24035 @item -fno-jump-tables
24036 @opindex fno-jump-tables
24037 Do not use jump tables for switch statements even where it would be
24038 more efficient than other code generation strategies. This option is
24039 of use in conjunction with @option{-fpic} or @option{-fPIC} for
24040 building code that forms part of a dynamic linker and cannot
24041 reference the address of a jump table. On some targets, jump tables
24042 do not require a GOT and this option is not needed.
24044 @item -ffixed-@var{reg}
24046 Treat the register named @var{reg} as a fixed register; generated code
24047 should never refer to it (except perhaps as a stack pointer, frame
24048 pointer or in some other fixed role).
24050 @var{reg} must be the name of a register. The register names accepted
24051 are machine-specific and are defined in the @code{REGISTER_NAMES}
24052 macro in the machine description macro file.
24054 This flag does not have a negative form, because it specifies a
24057 @item -fcall-used-@var{reg}
24058 @opindex fcall-used
24059 Treat the register named @var{reg} as an allocable register that is
24060 clobbered by function calls. It may be allocated for temporaries or
24061 variables that do not live across a call. Functions compiled this way
24062 do not save and restore the register @var{reg}.
24064 It is an error to use this flag with the frame pointer or stack pointer.
24065 Use of this flag for other registers that have fixed pervasive roles in
24066 the machine's execution model produces disastrous results.
24068 This flag does not have a negative form, because it specifies a
24071 @item -fcall-saved-@var{reg}
24072 @opindex fcall-saved
24073 Treat the register named @var{reg} as an allocable register saved by
24074 functions. It may be allocated even for temporaries or variables that
24075 live across a call. Functions compiled this way save and restore
24076 the register @var{reg} if they use it.
24078 It is an error to use this flag with the frame pointer or stack pointer.
24079 Use of this flag for other registers that have fixed pervasive roles in
24080 the machine's execution model produces disastrous results.
24082 A different sort of disaster results from the use of this flag for
24083 a register in which function values may be returned.
24085 This flag does not have a negative form, because it specifies a
24088 @item -fpack-struct[=@var{n}]
24089 @opindex fpack-struct
24090 Without a value specified, pack all structure members together without
24091 holes. When a value is specified (which must be a small power of two), pack
24092 structure members according to this value, representing the maximum
24093 alignment (that is, objects with default alignment requirements larger than
24094 this are output potentially unaligned at the next fitting location.
24096 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
24097 code that is not binary compatible with code generated without that switch.
24098 Additionally, it makes the code suboptimal.
24099 Use it to conform to a non-default application binary interface.
24101 @item -finstrument-functions
24102 @opindex finstrument-functions
24103 Generate instrumentation calls for entry and exit to functions. Just
24104 after function entry and just before function exit, the following
24105 profiling functions are called with the address of the current
24106 function and its call site. (On some platforms,
24107 @code{__builtin_return_address} does not work beyond the current
24108 function, so the call site information may not be available to the
24109 profiling functions otherwise.)
24112 void __cyg_profile_func_enter (void *this_fn,
24114 void __cyg_profile_func_exit (void *this_fn,
24118 The first argument is the address of the start of the current function,
24119 which may be looked up exactly in the symbol table.
24121 This instrumentation is also done for functions expanded inline in other
24122 functions. The profiling calls indicate where, conceptually, the
24123 inline function is entered and exited. This means that addressable
24124 versions of such functions must be available. If all your uses of a
24125 function are expanded inline, this may mean an additional expansion of
24126 code size. If you use @code{extern inline} in your C code, an
24127 addressable version of such functions must be provided. (This is
24128 normally the case anyway, but if you get lucky and the optimizer always
24129 expands the functions inline, you might have gotten away without
24130 providing static copies.)
24132 A function may be given the attribute @code{no_instrument_function}, in
24133 which case this instrumentation is not done. This can be used, for
24134 example, for the profiling functions listed above, high-priority
24135 interrupt routines, and any functions from which the profiling functions
24136 cannot safely be called (perhaps signal handlers, if the profiling
24137 routines generate output or allocate memory).
24139 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
24140 @opindex finstrument-functions-exclude-file-list
24142 Set the list of functions that are excluded from instrumentation (see
24143 the description of @option{-finstrument-functions}). If the file that
24144 contains a function definition matches with one of @var{file}, then
24145 that function is not instrumented. The match is done on substrings:
24146 if the @var{file} parameter is a substring of the file name, it is
24147 considered to be a match.
24152 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
24156 excludes any inline function defined in files whose pathnames
24157 contain @file{/bits/stl} or @file{include/sys}.
24159 If, for some reason, you want to include letter @samp{,} in one of
24160 @var{sym}, write @samp{\,}. For example,
24161 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
24162 (note the single quote surrounding the option).
24164 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
24165 @opindex finstrument-functions-exclude-function-list
24167 This is similar to @option{-finstrument-functions-exclude-file-list},
24168 but this option sets the list of function names to be excluded from
24169 instrumentation. The function name to be matched is its user-visible
24170 name, such as @code{vector<int> blah(const vector<int> &)}, not the
24171 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
24172 match is done on substrings: if the @var{sym} parameter is a substring
24173 of the function name, it is considered to be a match. For C99 and C++
24174 extended identifiers, the function name must be given in UTF-8, not
24175 using universal character names.
24177 @item -fstack-check
24178 @opindex fstack-check
24179 Generate code to verify that you do not go beyond the boundary of the
24180 stack. You should specify this flag if you are running in an
24181 environment with multiple threads, but you only rarely need to specify it in
24182 a single-threaded environment since stack overflow is automatically
24183 detected on nearly all systems if there is only one stack.
24185 Note that this switch does not actually cause checking to be done; the
24186 operating system or the language runtime must do that. The switch causes
24187 generation of code to ensure that they see the stack being extended.
24189 You can additionally specify a string parameter: @samp{no} means no
24190 checking, @samp{generic} means force the use of old-style checking,
24191 @samp{specific} means use the best checking method and is equivalent
24192 to bare @option{-fstack-check}.
24194 Old-style checking is a generic mechanism that requires no specific
24195 target support in the compiler but comes with the following drawbacks:
24199 Modified allocation strategy for large objects: they are always
24200 allocated dynamically if their size exceeds a fixed threshold.
24203 Fixed limit on the size of the static frame of functions: when it is
24204 topped by a particular function, stack checking is not reliable and
24205 a warning is issued by the compiler.
24208 Inefficiency: because of both the modified allocation strategy and the
24209 generic implementation, code performance is hampered.
24212 Note that old-style stack checking is also the fallback method for
24213 @samp{specific} if no target support has been added in the compiler.
24215 @item -fstack-limit-register=@var{reg}
24216 @itemx -fstack-limit-symbol=@var{sym}
24217 @itemx -fno-stack-limit
24218 @opindex fstack-limit-register
24219 @opindex fstack-limit-symbol
24220 @opindex fno-stack-limit
24221 Generate code to ensure that the stack does not grow beyond a certain value,
24222 either the value of a register or the address of a symbol. If a larger
24223 stack is required, a signal is raised at run time. For most targets,
24224 the signal is raised before the stack overruns the boundary, so
24225 it is possible to catch the signal without taking special precautions.
24227 For instance, if the stack starts at absolute address @samp{0x80000000}
24228 and grows downwards, you can use the flags
24229 @option{-fstack-limit-symbol=__stack_limit} and
24230 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
24231 of 128KB@. Note that this may only work with the GNU linker.
24233 @item -fsplit-stack
24234 @opindex fsplit-stack
24235 Generate code to automatically split the stack before it overflows.
24236 The resulting program has a discontiguous stack which can only
24237 overflow if the program is unable to allocate any more memory. This
24238 is most useful when running threaded programs, as it is no longer
24239 necessary to calculate a good stack size to use for each thread. This
24240 is currently only implemented for the x86 targets running
24243 When code compiled with @option{-fsplit-stack} calls code compiled
24244 without @option{-fsplit-stack}, there may not be much stack space
24245 available for the latter code to run. If compiling all code,
24246 including library code, with @option{-fsplit-stack} is not an option,
24247 then the linker can fix up these calls so that the code compiled
24248 without @option{-fsplit-stack} always has a large stack. Support for
24249 this is implemented in the gold linker in GNU binutils release 2.21
24252 @item -fleading-underscore
24253 @opindex fleading-underscore
24254 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
24255 change the way C symbols are represented in the object file. One use
24256 is to help link with legacy assembly code.
24258 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
24259 generate code that is not binary compatible with code generated without that
24260 switch. Use it to conform to a non-default application binary interface.
24261 Not all targets provide complete support for this switch.
24263 @item -ftls-model=@var{model}
24264 @opindex ftls-model
24265 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
24266 The @var{model} argument should be one of @samp{global-dynamic},
24267 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
24268 Note that the choice is subject to optimization: the compiler may use
24269 a more efficient model for symbols not visible outside of the translation
24270 unit, or if @option{-fpic} is not given on the command line.
24272 The default without @option{-fpic} is @samp{initial-exec}; with
24273 @option{-fpic} the default is @samp{global-dynamic}.
24275 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
24276 @opindex fvisibility
24277 Set the default ELF image symbol visibility to the specified option---all
24278 symbols are marked with this unless overridden within the code.
24279 Using this feature can very substantially improve linking and
24280 load times of shared object libraries, produce more optimized
24281 code, provide near-perfect API export and prevent symbol clashes.
24282 It is @strong{strongly} recommended that you use this in any shared objects
24285 Despite the nomenclature, @samp{default} always means public; i.e.,
24286 available to be linked against from outside the shared object.
24287 @samp{protected} and @samp{internal} are pretty useless in real-world
24288 usage so the only other commonly used option is @samp{hidden}.
24289 The default if @option{-fvisibility} isn't specified is
24290 @samp{default}, i.e., make every symbol public.
24292 A good explanation of the benefits offered by ensuring ELF
24293 symbols have the correct visibility is given by ``How To Write
24294 Shared Libraries'' by Ulrich Drepper (which can be found at
24295 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
24296 solution made possible by this option to marking things hidden when
24297 the default is public is to make the default hidden and mark things
24298 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
24299 and @code{__attribute__ ((visibility("default")))} instead of
24300 @code{__declspec(dllexport)} you get almost identical semantics with
24301 identical syntax. This is a great boon to those working with
24302 cross-platform projects.
24304 For those adding visibility support to existing code, you may find
24305 @code{#pragma GCC visibility} of use. This works by you enclosing
24306 the declarations you wish to set visibility for with (for example)
24307 @code{#pragma GCC visibility push(hidden)} and
24308 @code{#pragma GCC visibility pop}.
24309 Bear in mind that symbol visibility should be viewed @strong{as
24310 part of the API interface contract} and thus all new code should
24311 always specify visibility when it is not the default; i.e., declarations
24312 only for use within the local DSO should @strong{always} be marked explicitly
24313 as hidden as so to avoid PLT indirection overheads---making this
24314 abundantly clear also aids readability and self-documentation of the code.
24315 Note that due to ISO C++ specification requirements, @code{operator new} and
24316 @code{operator delete} must always be of default visibility.
24318 Be aware that headers from outside your project, in particular system
24319 headers and headers from any other library you use, may not be
24320 expecting to be compiled with visibility other than the default. You
24321 may need to explicitly say @code{#pragma GCC visibility push(default)}
24322 before including any such headers.
24324 @code{extern} declarations are not affected by @option{-fvisibility}, so
24325 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24326 no modifications. However, this means that calls to @code{extern}
24327 functions with no explicit visibility use the PLT, so it is more
24328 effective to use @code{__attribute ((visibility))} and/or
24329 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24330 declarations should be treated as hidden.
24332 Note that @option{-fvisibility} does affect C++ vague linkage
24333 entities. This means that, for instance, an exception class that is
24334 be thrown between DSOs must be explicitly marked with default
24335 visibility so that the @samp{type_info} nodes are unified between
24338 An overview of these techniques, their benefits and how to use them
24339 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24341 @item -fstrict-volatile-bitfields
24342 @opindex fstrict-volatile-bitfields
24343 This option should be used if accesses to volatile bit-fields (or other
24344 structure fields, although the compiler usually honors those types
24345 anyway) should use a single access of the width of the
24346 field's type, aligned to a natural alignment if possible. For
24347 example, targets with memory-mapped peripheral registers might require
24348 all such accesses to be 16 bits wide; with this flag you can
24349 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24350 is 16 bits on these targets) to force GCC to use 16-bit accesses
24351 instead of, perhaps, a more efficient 32-bit access.
24353 If this option is disabled, the compiler uses the most efficient
24354 instruction. In the previous example, that might be a 32-bit load
24355 instruction, even though that accesses bytes that do not contain
24356 any portion of the bit-field, or memory-mapped registers unrelated to
24357 the one being updated.
24359 In some cases, such as when the @code{packed} attribute is applied to a
24360 structure field, it may not be possible to access the field with a single
24361 read or write that is correctly aligned for the target machine. In this
24362 case GCC falls back to generating multiple accesses rather than code that
24363 will fault or truncate the result at run time.
24365 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24366 not allowed to touch non bit-field members. It is therefore recommended
24367 to define all bits of the field's type as bit-field members.
24369 The default value of this option is determined by the application binary
24370 interface for the target processor.
24372 @item -fsync-libcalls
24373 @opindex fsync-libcalls
24374 This option controls whether any out-of-line instance of the @code{__sync}
24375 family of functions may be used to implement the C++11 @code{__atomic}
24376 family of functions.
24378 The default value of this option is enabled, thus the only useful form
24379 of the option is @option{-fno-sync-libcalls}. This option is used in
24380 the implementation of the @file{libatomic} runtime library.
24386 @node Environment Variables
24387 @section Environment Variables Affecting GCC
24388 @cindex environment variables
24390 @c man begin ENVIRONMENT
24391 This section describes several environment variables that affect how GCC
24392 operates. Some of them work by specifying directories or prefixes to use
24393 when searching for various kinds of files. Some are used to specify other
24394 aspects of the compilation environment.
24396 Note that you can also specify places to search using options such as
24397 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24398 take precedence over places specified using environment variables, which
24399 in turn take precedence over those specified by the configuration of GCC@.
24400 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24401 GNU Compiler Collection (GCC) Internals}.
24406 @c @itemx LC_COLLATE
24408 @c @itemx LC_MONETARY
24409 @c @itemx LC_NUMERIC
24414 @c @findex LC_COLLATE
24415 @findex LC_MESSAGES
24416 @c @findex LC_MONETARY
24417 @c @findex LC_NUMERIC
24421 These environment variables control the way that GCC uses
24422 localization information which allows GCC to work with different
24423 national conventions. GCC inspects the locale categories
24424 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24425 so. These locale categories can be set to any value supported by your
24426 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24427 Kingdom encoded in UTF-8.
24429 The @env{LC_CTYPE} environment variable specifies character
24430 classification. GCC uses it to determine the character boundaries in
24431 a string; this is needed for some multibyte encodings that contain quote
24432 and escape characters that are otherwise interpreted as a string
24435 The @env{LC_MESSAGES} environment variable specifies the language to
24436 use in diagnostic messages.
24438 If the @env{LC_ALL} environment variable is set, it overrides the value
24439 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24440 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24441 environment variable. If none of these variables are set, GCC
24442 defaults to traditional C English behavior.
24446 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24447 files. GCC uses temporary files to hold the output of one stage of
24448 compilation which is to be used as input to the next stage: for example,
24449 the output of the preprocessor, which is the input to the compiler
24452 @item GCC_COMPARE_DEBUG
24453 @findex GCC_COMPARE_DEBUG
24454 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24455 @option{-fcompare-debug} to the compiler driver. See the documentation
24456 of this option for more details.
24458 @item GCC_EXEC_PREFIX
24459 @findex GCC_EXEC_PREFIX
24460 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24461 names of the subprograms executed by the compiler. No slash is added
24462 when this prefix is combined with the name of a subprogram, but you can
24463 specify a prefix that ends with a slash if you wish.
24465 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24466 an appropriate prefix to use based on the pathname it is invoked with.
24468 If GCC cannot find the subprogram using the specified prefix, it
24469 tries looking in the usual places for the subprogram.
24471 The default value of @env{GCC_EXEC_PREFIX} is
24472 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24473 the installed compiler. In many cases @var{prefix} is the value
24474 of @code{prefix} when you ran the @file{configure} script.
24476 Other prefixes specified with @option{-B} take precedence over this prefix.
24478 This prefix is also used for finding files such as @file{crt0.o} that are
24481 In addition, the prefix is used in an unusual way in finding the
24482 directories to search for header files. For each of the standard
24483 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24484 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24485 replacing that beginning with the specified prefix to produce an
24486 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24487 @file{foo/bar} just before it searches the standard directory
24488 @file{/usr/local/lib/bar}.
24489 If a standard directory begins with the configured
24490 @var{prefix} then the value of @var{prefix} is replaced by
24491 @env{GCC_EXEC_PREFIX} when looking for header files.
24493 @item COMPILER_PATH
24494 @findex COMPILER_PATH
24495 The value of @env{COMPILER_PATH} is a colon-separated list of
24496 directories, much like @env{PATH}. GCC tries the directories thus
24497 specified when searching for subprograms, if it can't find the
24498 subprograms using @env{GCC_EXEC_PREFIX}.
24501 @findex LIBRARY_PATH
24502 The value of @env{LIBRARY_PATH} is a colon-separated list of
24503 directories, much like @env{PATH}. When configured as a native compiler,
24504 GCC tries the directories thus specified when searching for special
24505 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24506 using GCC also uses these directories when searching for ordinary
24507 libraries for the @option{-l} option (but directories specified with
24508 @option{-L} come first).
24512 @cindex locale definition
24513 This variable is used to pass locale information to the compiler. One way in
24514 which this information is used is to determine the character set to be used
24515 when character literals, string literals and comments are parsed in C and C++.
24516 When the compiler is configured to allow multibyte characters,
24517 the following values for @env{LANG} are recognized:
24521 Recognize JIS characters.
24523 Recognize SJIS characters.
24525 Recognize EUCJP characters.
24528 If @env{LANG} is not defined, or if it has some other value, then the
24529 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24530 recognize and translate multibyte characters.
24534 Some additional environment variables affect the behavior of the
24537 @include cppenv.texi
24541 @node Precompiled Headers
24542 @section Using Precompiled Headers
24543 @cindex precompiled headers
24544 @cindex speed of compilation
24546 Often large projects have many header files that are included in every
24547 source file. The time the compiler takes to process these header files
24548 over and over again can account for nearly all of the time required to
24549 build the project. To make builds faster, GCC allows you to
24550 @dfn{precompile} a header file.
24552 To create a precompiled header file, simply compile it as you would any
24553 other file, if necessary using the @option{-x} option to make the driver
24554 treat it as a C or C++ header file. You may want to use a
24555 tool like @command{make} to keep the precompiled header up-to-date when
24556 the headers it contains change.
24558 A precompiled header file is searched for when @code{#include} is
24559 seen in the compilation. As it searches for the included file
24560 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24561 compiler looks for a precompiled header in each directory just before it
24562 looks for the include file in that directory. The name searched for is
24563 the name specified in the @code{#include} with @samp{.gch} appended. If
24564 the precompiled header file can't be used, it is ignored.
24566 For instance, if you have @code{#include "all.h"}, and you have
24567 @file{all.h.gch} in the same directory as @file{all.h}, then the
24568 precompiled header file is used if possible, and the original
24569 header is used otherwise.
24571 Alternatively, you might decide to put the precompiled header file in a
24572 directory and use @option{-I} to ensure that directory is searched
24573 before (or instead of) the directory containing the original header.
24574 Then, if you want to check that the precompiled header file is always
24575 used, you can put a file of the same name as the original header in this
24576 directory containing an @code{#error} command.
24578 This also works with @option{-include}. So yet another way to use
24579 precompiled headers, good for projects not designed with precompiled
24580 header files in mind, is to simply take most of the header files used by
24581 a project, include them from another header file, precompile that header
24582 file, and @option{-include} the precompiled header. If the header files
24583 have guards against multiple inclusion, they are skipped because
24584 they've already been included (in the precompiled header).
24586 If you need to precompile the same header file for different
24587 languages, targets, or compiler options, you can instead make a
24588 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24589 header in the directory, perhaps using @option{-o}. It doesn't matter
24590 what you call the files in the directory; every precompiled header in
24591 the directory is considered. The first precompiled header
24592 encountered in the directory that is valid for this compilation is
24593 used; they're searched in no particular order.
24595 There are many other possibilities, limited only by your imagination,
24596 good sense, and the constraints of your build system.
24598 A precompiled header file can be used only when these conditions apply:
24602 Only one precompiled header can be used in a particular compilation.
24605 A precompiled header can't be used once the first C token is seen. You
24606 can have preprocessor directives before a precompiled header; you cannot
24607 include a precompiled header from inside another header.
24610 The precompiled header file must be produced for the same language as
24611 the current compilation. You can't use a C precompiled header for a C++
24615 The precompiled header file must have been produced by the same compiler
24616 binary as the current compilation is using.
24619 Any macros defined before the precompiled header is included must
24620 either be defined in the same way as when the precompiled header was
24621 generated, or must not affect the precompiled header, which usually
24622 means that they don't appear in the precompiled header at all.
24624 The @option{-D} option is one way to define a macro before a
24625 precompiled header is included; using a @code{#define} can also do it.
24626 There are also some options that define macros implicitly, like
24627 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
24630 @item If debugging information is output when using the precompiled
24631 header, using @option{-g} or similar, the same kind of debugging information
24632 must have been output when building the precompiled header. However,
24633 a precompiled header built using @option{-g} can be used in a compilation
24634 when no debugging information is being output.
24636 @item The same @option{-m} options must generally be used when building
24637 and using the precompiled header. @xref{Submodel Options},
24638 for any cases where this rule is relaxed.
24640 @item Each of the following options must be the same when building and using
24641 the precompiled header:
24643 @gccoptlist{-fexceptions}
24646 Some other command-line options starting with @option{-f},
24647 @option{-p}, or @option{-O} must be defined in the same way as when
24648 the precompiled header was generated. At present, it's not clear
24649 which options are safe to change and which are not; the safest choice
24650 is to use exactly the same options when generating and using the
24651 precompiled header. The following are known to be safe:
24653 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24654 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24655 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24660 For all of these except the last, the compiler automatically
24661 ignores the precompiled header if the conditions aren't met. If you
24662 find an option combination that doesn't work and doesn't cause the
24663 precompiled header to be ignored, please consider filing a bug report,
24666 If you do use differing options when generating and using the
24667 precompiled header, the actual behavior is a mixture of the
24668 behavior for the options. For instance, if you use @option{-g} to
24669 generate the precompiled header but not when using it, you may or may
24670 not get debugging information for routines in the precompiled header.