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 -Wnarrowing @gol
202 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
203 -Weffc++ -Wstrict-null-sentinel @gol
204 -Wno-non-template-friend -Wold-style-cast @gol
205 -Woverloaded-virtual -Wno-pmf-conversions @gol
208 @item Objective-C and Objective-C++ Language Options
209 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
210 Objective-C and Objective-C++ Dialects}.
211 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
212 -fgnu-runtime -fnext-runtime @gol
213 -fno-nil-receivers @gol
214 -fobjc-abi-version=@var{n} @gol
215 -fobjc-call-cxx-cdtors @gol
216 -fobjc-direct-dispatch @gol
217 -fobjc-exceptions @gol
220 -fobjc-std=objc1 @gol
221 -fno-local-ivars @gol
222 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
223 -freplace-objc-classes @gol
226 -Wassign-intercept @gol
227 -Wno-protocol -Wselector @gol
228 -Wstrict-selector-match @gol
229 -Wundeclared-selector}
231 @item Diagnostic Message Formatting Options
232 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
233 @gccoptlist{-fmessage-length=@var{n} @gol
234 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
235 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
236 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
238 @item Warning Options
239 @xref{Warning Options,,Options to Request or Suppress Warnings}.
240 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
241 -pedantic-errors @gol
242 -w -Wextra -Wall -Waddress -Waggregate-return @gol
243 -Waggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
244 -Wbool-compare -Wframe-address @gol
245 -Wno-attributes -Wno-builtin-macro-redefined @gol
246 -Wc90-c99-compat -Wc99-c11-compat @gol
247 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
248 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
249 -Wconversion -Wcoverage-mismatch -Wdate-time -Wdelete-incomplete -Wno-cpp @gol
250 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
251 -Wdisabled-optimization @gol
252 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
253 -Wno-div-by-zero -Wdouble-promotion -Wempty-body -Wenum-compare @gol
254 -Wno-endif-labels -Werror -Werror=* @gol
255 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
256 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
257 -Wformat-security -Wformat-signedness -Wformat-y2k @gol
258 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
259 -Wignored-qualifiers -Wincompatible-pointer-types @gol
260 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
261 -Winit-self -Winline -Wno-int-conversion @gol
262 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
263 -Wnull-dereference @gol
264 -Winvalid-pch -Wlarger-than=@var{len} -Wunsafe-loop-optimizations @gol
265 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
266 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args @gol
267 -Wmisleading-indentation -Wmissing-braces @gol
268 -Wmissing-field-initializers -Wmissing-include-dirs @gol
269 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
270 -Wodr -Wno-overflow -Wopenmp-simd @gol
271 -Woverride-init-side-effects @gol
272 -Woverlength-strings -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
273 -Wparentheses -Wpedantic-ms-format -Wno-pedantic-ms-format @gol
274 -Wpointer-arith -Wno-pointer-to-int-cast @gol
275 -Wredundant-decls -Wno-return-local-addr @gol
276 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
277 -Wshift-overflow -Wshift-overflow=@var{n} @gol
278 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
279 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
280 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
281 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
282 -Wstrict-aliasing=n @gol -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
283 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
284 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
285 -Wmissing-format-attribute @gol
286 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
287 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
288 -Wtype-limits -Wundef @gol
289 -Wuninitialized -Wunknown-pragmas -Wno-pragmas @gol
290 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
291 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
292 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
293 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
294 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
295 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
296 -Wzero-as-null-pointer-constant}
298 @item C and Objective-C-only Warning Options
299 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
300 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
301 -Wold-style-declaration -Wold-style-definition @gol
302 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
303 -Wdeclaration-after-statement -Wpointer-sign}
305 @item Debugging Options
306 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
307 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
308 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
309 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
310 -fsanitize-undefined-trap-on-error @gol
311 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
312 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
313 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
314 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
315 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
316 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
317 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
318 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
319 -fchkp-use-wrappers @gol
320 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
321 -fdisable-ipa-@var{pass_name} @gol
322 -fdisable-rtl-@var{pass_name} @gol
323 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
324 -fdisable-tree-@var{pass_name} @gol
325 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
326 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
327 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
328 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
329 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
331 -fdump-statistics @gol
333 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
334 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
335 -fdump-tree-cfg -fdump-tree-alias @gol
337 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
338 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
339 -fdump-tree-gimple@r{[}-raw@r{]} @gol
340 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
341 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
342 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
343 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
344 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-nrv -fdump-tree-vect @gol
347 -fdump-tree-sink @gol
348 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
349 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
350 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
351 -fdump-tree-vtable-verify @gol
352 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
353 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
354 -fdump-final-insns=@var{file} @gol
355 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
356 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
357 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
358 -fenable-@var{kind}-@var{pass} @gol
359 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
360 -fdebug-types-section -fmem-report-wpa @gol
361 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
363 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
364 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
365 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
366 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
367 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
368 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
369 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
370 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
371 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
372 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
373 -fdebug-prefix-map=@var{old}=@var{new} @gol
374 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
375 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
376 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
377 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
378 -print-prog-name=@var{program} -print-search-dirs -Q @gol
379 -print-sysroot -print-sysroot-headers-suffix @gol
380 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
382 @item Optimization Options
383 @xref{Optimize Options,,Options that Control Optimization}.
384 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
385 -falign-jumps[=@var{n}] @gol
386 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
387 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
388 -fauto-inc-dec -fbranch-probabilities @gol
389 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
390 -fbtr-bb-exclusive -fcaller-saves @gol
391 -fcombine-stack-adjustments -fconserve-stack @gol
392 -fcompare-elim -fcprop-registers -fcrossjumping @gol
393 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
394 -fcx-limited-range @gol
395 -fdata-sections -fdce -fdelayed-branch @gol
396 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
397 -fdevirtualize-at-ltrans -fdse @gol
398 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
399 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
400 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
401 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
402 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
403 -fif-conversion2 -findirect-inlining @gol
404 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
405 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
406 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
407 -fira-algorithm=@var{algorithm} @gol
408 -fira-region=@var{region} -fira-hoist-pressure @gol
409 -fira-loop-pressure -fno-ira-share-save-slots @gol
410 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
411 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
412 -fivopts -fkeep-inline-functions -fkeep-static-consts @gol
413 -flive-range-shrinkage @gol
414 -floop-block -floop-interchange -floop-strip-mine @gol
415 -floop-unroll-and-jam -floop-nest-optimize @gol
416 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
417 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
418 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
419 -fmove-loop-invariants -fno-branch-count-reg @gol
420 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
421 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
422 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
423 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
424 -fomit-frame-pointer -foptimize-sibling-calls @gol
425 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
426 -fprefetch-loop-arrays -fprofile-report @gol
427 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
428 -fprofile-generate=@var{path} @gol
429 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
430 -fprofile-reorder-functions @gol
431 -freciprocal-math -free -frename-registers -freorder-blocks @gol
432 -freorder-blocks-and-partition -freorder-functions @gol
433 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
434 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
435 -fsched-spec-load -fsched-spec-load-dangerous @gol
436 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
437 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
438 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
439 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
440 -fschedule-fusion @gol
441 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
442 -fselective-scheduling -fselective-scheduling2 @gol
443 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
444 -fsemantic-interposition @gol
445 -fshrink-wrap -fsignaling-nans -fsingle-precision-constant @gol
446 -fsplit-ivs-in-unroller -fsplit-wide-types -fssa-phiopt @gol
447 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
448 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
449 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
450 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
451 -ftree-coalesce-inline-vars -ftree-coalesce-vars -ftree-copy-prop @gol
452 -ftree-copyrename -ftree-dce -ftree-dominator-opts -ftree-dse @gol
453 -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
454 -ftree-loop-if-convert-stores -ftree-loop-im @gol
455 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
456 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
457 -ftree-loop-vectorize @gol
458 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
459 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
460 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
461 -ftree-vectorize -ftree-vrp @gol
462 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
463 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
464 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
465 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
466 --param @var{name}=@var{value}
467 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
469 @item Preprocessor Options
470 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
471 @gccoptlist{-A@var{question}=@var{answer} @gol
472 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
473 -C -dD -dI -dM -dN @gol
474 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
475 -idirafter @var{dir} @gol
476 -include @var{file} -imacros @var{file} @gol
477 -iprefix @var{file} -iwithprefix @var{dir} @gol
478 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
479 -imultilib @var{dir} -isysroot @var{dir} @gol
480 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
481 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
482 -remap -trigraphs -undef -U@var{macro} @gol
483 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
485 @item Assembler Option
486 @xref{Assembler Options,,Passing Options to the Assembler}.
487 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
490 @xref{Link Options,,Options for Linking}.
491 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
492 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
493 -s -static -static-libgcc -static-libstdc++ @gol
494 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
495 -static-libmpx -static-libmpxwrappers @gol
496 -shared -shared-libgcc -symbolic @gol
497 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
498 -u @var{symbol} -z @var{keyword}}
500 @item Directory Options
501 @xref{Directory Options,,Options for Directory Search}.
502 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
503 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
504 --sysroot=@var{dir} --no-sysroot-suffix}
506 @item Machine Dependent Options
507 @xref{Submodel Options,,Hardware Models and Configurations}.
508 @c This list is ordered alphanumerically by subsection name.
509 @c Try and put the significant identifier (CPU or system) first,
510 @c so users have a clue at guessing where the ones they want will be.
512 @emph{AArch64 Options}
513 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
514 -mgeneral-regs-only @gol
515 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
517 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
518 -mtls-dialect=desc -mtls-dialect=traditional @gol
519 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
520 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
521 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
523 @emph{Adapteva Epiphany Options}
524 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
525 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
526 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
527 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
528 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
529 -msplit-vecmove-early -m1reg-@var{reg}}
532 @gccoptlist{-mbarrel-shifter @gol
533 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
534 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
535 -mea -mno-mpy -mmul32x16 -mmul64 @gol
536 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
537 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
538 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
539 -mepilogue-cfi -mlong-calls -mmedium-calls -msdata @gol
540 -mucb-mcount -mvolatile-cache @gol
541 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
542 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
543 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
544 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
545 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
546 -mtune=@var{cpu} -mmultcost=@var{num} -munalign-prob-threshold=@var{probability}}
549 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
550 -mabi=@var{name} @gol
551 -mapcs-stack-check -mno-apcs-stack-check @gol
552 -mapcs-float -mno-apcs-float @gol
553 -mapcs-reentrant -mno-apcs-reentrant @gol
554 -msched-prolog -mno-sched-prolog @gol
555 -mlittle-endian -mbig-endian @gol
556 -mfloat-abi=@var{name} @gol
557 -mfp16-format=@var{name}
558 -mthumb-interwork -mno-thumb-interwork @gol
559 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
560 -mtune=@var{name} -mprint-tune-info @gol
561 -mstructure-size-boundary=@var{n} @gol
562 -mabort-on-noreturn @gol
563 -mlong-calls -mno-long-calls @gol
564 -msingle-pic-base -mno-single-pic-base @gol
565 -mpic-register=@var{reg} @gol
566 -mnop-fun-dllimport @gol
567 -mpoke-function-name @gol
569 -mtpcs-frame -mtpcs-leaf-frame @gol
570 -mcaller-super-interworking -mcallee-super-interworking @gol
571 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
572 -mword-relocations @gol
573 -mfix-cortex-m3-ldrd @gol
574 -munaligned-access @gol
575 -mneon-for-64bits @gol
576 -mslow-flash-data @gol
577 -masm-syntax-unified @gol
581 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
582 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
583 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
585 @emph{Blackfin Options}
586 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
587 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
588 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
589 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
590 -mno-id-shared-library -mshared-library-id=@var{n} @gol
591 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
592 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
593 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
597 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
598 -msim -msdata=@var{sdata-type}}
601 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
602 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
603 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
604 -mstack-align -mdata-align -mconst-align @gol
605 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
606 -melf -maout -melinux -mlinux -sim -sim2 @gol
607 -mmul-bug-workaround -mno-mul-bug-workaround}
610 @gccoptlist{-mmac @gol
611 -mcr16cplus -mcr16c @gol
612 -msim -mint32 -mbit-ops
613 -mdata-model=@var{model}}
615 @emph{Darwin Options}
616 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
617 -arch_only -bind_at_load -bundle -bundle_loader @gol
618 -client_name -compatibility_version -current_version @gol
620 -dependency-file -dylib_file -dylinker_install_name @gol
621 -dynamic -dynamiclib -exported_symbols_list @gol
622 -filelist -flat_namespace -force_cpusubtype_ALL @gol
623 -force_flat_namespace -headerpad_max_install_names @gol
625 -image_base -init -install_name -keep_private_externs @gol
626 -multi_module -multiply_defined -multiply_defined_unused @gol
627 -noall_load -no_dead_strip_inits_and_terms @gol
628 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
629 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
630 -private_bundle -read_only_relocs -sectalign @gol
631 -sectobjectsymbols -whyload -seg1addr @gol
632 -sectcreate -sectobjectsymbols -sectorder @gol
633 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
634 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
635 -segprot -segs_read_only_addr -segs_read_write_addr @gol
636 -single_module -static -sub_library -sub_umbrella @gol
637 -twolevel_namespace -umbrella -undefined @gol
638 -unexported_symbols_list -weak_reference_mismatches @gol
639 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
640 -mkernel -mone-byte-bool}
642 @emph{DEC Alpha Options}
643 @gccoptlist{-mno-fp-regs -msoft-float @gol
644 -mieee -mieee-with-inexact -mieee-conformant @gol
645 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
646 -mtrap-precision=@var{mode} -mbuild-constants @gol
647 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
648 -mbwx -mmax -mfix -mcix @gol
649 -mfloat-vax -mfloat-ieee @gol
650 -mexplicit-relocs -msmall-data -mlarge-data @gol
651 -msmall-text -mlarge-text @gol
652 -mmemory-latency=@var{time}}
655 @gccoptlist{-msmall-model -mno-lsim}
658 @gccoptlist{-msim -mlra}
661 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
662 -mhard-float -msoft-float @gol
663 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
664 -mdouble -mno-double @gol
665 -mmedia -mno-media -mmuladd -mno-muladd @gol
666 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
667 -mlinked-fp -mlong-calls -malign-labels @gol
668 -mlibrary-pic -macc-4 -macc-8 @gol
669 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
670 -moptimize-membar -mno-optimize-membar @gol
671 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
672 -mvliw-branch -mno-vliw-branch @gol
673 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
674 -mno-nested-cond-exec -mtomcat-stats @gol
678 @emph{GNU/Linux Options}
679 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
680 -tno-android-cc -tno-android-ld}
682 @emph{H8/300 Options}
683 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
686 @gccoptlist{-march=@var{architecture-type} @gol
687 -mdisable-fpregs -mdisable-indexing @gol
688 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
689 -mfixed-range=@var{register-range} @gol
690 -mjump-in-delay -mlinker-opt -mlong-calls @gol
691 -mlong-load-store -mno-disable-fpregs @gol
692 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
693 -mno-jump-in-delay -mno-long-load-store @gol
694 -mno-portable-runtime -mno-soft-float @gol
695 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
696 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
697 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
698 -munix=@var{unix-std} -nolibdld -static -threads}
701 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
702 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
703 -mconstant-gp -mauto-pic -mfused-madd @gol
704 -minline-float-divide-min-latency @gol
705 -minline-float-divide-max-throughput @gol
706 -mno-inline-float-divide @gol
707 -minline-int-divide-min-latency @gol
708 -minline-int-divide-max-throughput @gol
709 -mno-inline-int-divide @gol
710 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
711 -mno-inline-sqrt @gol
712 -mdwarf2-asm -mearly-stop-bits @gol
713 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
714 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
715 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
716 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
717 -msched-spec-ldc -msched-spec-control-ldc @gol
718 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
719 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
720 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
721 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
724 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
725 -msign-extend-enabled -muser-enabled}
727 @emph{M32R/D Options}
728 @gccoptlist{-m32r2 -m32rx -m32r @gol
730 -malign-loops -mno-align-loops @gol
731 -missue-rate=@var{number} @gol
732 -mbranch-cost=@var{number} @gol
733 -mmodel=@var{code-size-model-type} @gol
734 -msdata=@var{sdata-type} @gol
735 -mno-flush-func -mflush-func=@var{name} @gol
736 -mno-flush-trap -mflush-trap=@var{number} @gol
740 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
742 @emph{M680x0 Options}
743 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
744 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
745 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
746 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
747 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
748 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
749 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
750 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
754 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
755 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
756 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
757 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
758 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
761 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
762 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
763 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
764 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
767 @emph{MicroBlaze Options}
768 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
769 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
770 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
771 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
772 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
775 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
776 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
777 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
778 -mips16 -mno-mips16 -mflip-mips16 @gol
779 -minterlink-compressed -mno-interlink-compressed @gol
780 -minterlink-mips16 -mno-interlink-mips16 @gol
781 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
782 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
783 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
784 -mno-float -msingle-float -mdouble-float @gol
785 -modd-spreg -mno-odd-spreg @gol
786 -mabs=@var{mode} -mnan=@var{encoding} @gol
787 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
790 -mvirt -mno-virt @gol
792 -mmicromips -mno-micromips @gol
793 -mfpu=@var{fpu-type} @gol
794 -msmartmips -mno-smartmips @gol
795 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
796 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
797 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
798 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
799 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
800 -membedded-data -mno-embedded-data @gol
801 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
802 -mcode-readable=@var{setting} @gol
803 -msplit-addresses -mno-split-addresses @gol
804 -mexplicit-relocs -mno-explicit-relocs @gol
805 -mcheck-zero-division -mno-check-zero-division @gol
806 -mdivide-traps -mdivide-breaks @gol
807 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
808 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
809 -mfix-24k -mno-fix-24k @gol
810 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
811 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
812 -mfix-vr4120 -mno-fix-vr4120 @gol
813 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
814 -mflush-func=@var{func} -mno-flush-func @gol
815 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
816 -mfp-exceptions -mno-fp-exceptions @gol
817 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
818 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address}
821 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
822 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
823 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
824 -mno-base-addresses -msingle-exit -mno-single-exit}
826 @emph{MN10300 Options}
827 @gccoptlist{-mmult-bug -mno-mult-bug @gol
828 -mno-am33 -mam33 -mam33-2 -mam34 @gol
829 -mtune=@var{cpu-type} @gol
830 -mreturn-pointer-on-d0 @gol
831 -mno-crt0 -mrelax -mliw -msetlb}
834 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
836 @emph{MSP430 Options}
837 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
838 -mcode-region= -mdata-region= @gol
842 @gccoptlist{-mbig-endian -mlittle-endian @gol
843 -mreduced-regs -mfull-regs @gol
844 -mcmov -mno-cmov @gol
845 -mperf-ext -mno-perf-ext @gol
846 -mv3push -mno-v3push @gol
847 -m16bit -mno-16bit @gol
848 -misr-vector-size=@var{num} @gol
849 -mcache-block-size=@var{num} @gol
850 -march=@var{arch} @gol
851 -mcmodel=@var{code-model} @gol
854 @emph{Nios II Options}
855 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
857 -mno-bypass-cache -mbypass-cache @gol
858 -mno-cache-volatile -mcache-volatile @gol
859 -mno-fast-sw-div -mfast-sw-div @gol
860 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
861 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
862 -mcustom-fpu-cfg=@var{name} @gol
863 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
864 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
866 @emph{Nvidia PTX Options}
867 @gccoptlist{-m32 -m64 -mmainkernel}
869 @emph{PDP-11 Options}
870 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
871 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
872 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
873 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
874 -mbranch-expensive -mbranch-cheap @gol
875 -munix-asm -mdec-asm}
877 @emph{picoChip Options}
878 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
879 -msymbol-as-address -mno-inefficient-warnings}
881 @emph{PowerPC Options}
882 See RS/6000 and PowerPC Options.
885 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
886 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
887 -m64bit-doubles -m32bit-doubles}
889 @emph{RS/6000 and PowerPC Options}
890 @gccoptlist{-mcpu=@var{cpu-type} @gol
891 -mtune=@var{cpu-type} @gol
892 -mcmodel=@var{code-model} @gol
894 -maltivec -mno-altivec @gol
895 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
896 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
897 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
898 -mfprnd -mno-fprnd @gol
899 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
900 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
901 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
902 -malign-power -malign-natural @gol
903 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
904 -msingle-float -mdouble-float -msimple-fpu @gol
905 -mstring -mno-string -mupdate -mno-update @gol
906 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
907 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
908 -mstrict-align -mno-strict-align -mrelocatable @gol
909 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
910 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
911 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
912 -mprioritize-restricted-insns=@var{priority} @gol
913 -msched-costly-dep=@var{dependence_type} @gol
914 -minsert-sched-nops=@var{scheme} @gol
915 -mcall-sysv -mcall-netbsd @gol
916 -maix-struct-return -msvr4-struct-return @gol
917 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
918 -mblock-move-inline-limit=@var{num} @gol
919 -misel -mno-isel @gol
920 -misel=yes -misel=no @gol
922 -mspe=yes -mspe=no @gol
924 -mgen-cell-microcode -mwarn-cell-microcode @gol
925 -mvrsave -mno-vrsave @gol
926 -mmulhw -mno-mulhw @gol
927 -mdlmzb -mno-dlmzb @gol
928 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
929 -mprototype -mno-prototype @gol
930 -msim -mmvme -mads -myellowknife -memb -msdata @gol
931 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
932 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
933 -mno-recip-precision @gol
934 -mveclibabi=@var{type} -mfriz -mno-friz @gol
935 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
936 -msave-toc-indirect -mno-save-toc-indirect @gol
937 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
938 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
939 -mquad-memory -mno-quad-memory @gol
940 -mquad-memory-atomic -mno-quad-memory-atomic @gol
941 -mcompat-align-parm -mno-compat-align-parm @gol
942 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
943 -mupper-regs -mno-upper-regs}
946 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
948 -mbig-endian-data -mlittle-endian-data @gol
951 -mas100-syntax -mno-as100-syntax@gol
953 -mmax-constant-size=@gol
956 -mallow-string-insns -mno-allow-string-insns@gol
957 -mno-warn-multiple-fast-interrupts@gol
958 -msave-acc-in-interrupts}
960 @emph{S/390 and zSeries Options}
961 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
962 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
963 -mlong-double-64 -mlong-double-128 @gol
964 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
965 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
966 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
967 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
968 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
969 -mhotpatch=@var{halfwords},@var{halfwords}}
972 @gccoptlist{-meb -mel @gol
976 -mscore5 -mscore5u -mscore7 -mscore7d}
979 @gccoptlist{-m1 -m2 -m2e @gol
980 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
982 -m4-nofpu -m4-single-only -m4-single -m4 @gol
983 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
984 -m5-64media -m5-64media-nofpu @gol
985 -m5-32media -m5-32media-nofpu @gol
986 -m5-compact -m5-compact-nofpu @gol
987 -mb -ml -mdalign -mrelax @gol
988 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
989 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
990 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
991 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
992 -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
993 -maccumulate-outgoing-args -minvalid-symbols @gol
994 -matomic-model=@var{atomic-model} @gol
995 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
996 -mcbranch-force-delay-slot @gol
997 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
998 -mpretend-cmove -mtas}
1000 @emph{Solaris 2 Options}
1001 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1004 @emph{SPARC Options}
1005 @gccoptlist{-mcpu=@var{cpu-type} @gol
1006 -mtune=@var{cpu-type} @gol
1007 -mcmodel=@var{code-model} @gol
1008 -mmemory-model=@var{mem-model} @gol
1009 -m32 -m64 -mapp-regs -mno-app-regs @gol
1010 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1011 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1012 -mhard-quad-float -msoft-quad-float @gol
1013 -mstack-bias -mno-stack-bias @gol
1014 -munaligned-doubles -mno-unaligned-doubles @gol
1015 -muser-mode -mno-user-mode @gol
1016 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1017 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1018 -mcbcond -mno-cbcond @gol
1019 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1020 -mfix-at697f -mfix-ut699}
1023 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1024 -msafe-dma -munsafe-dma @gol
1026 -msmall-mem -mlarge-mem -mstdmain @gol
1027 -mfixed-range=@var{register-range} @gol
1029 -maddress-space-conversion -mno-address-space-conversion @gol
1030 -mcache-size=@var{cache-size} @gol
1031 -matomic-updates -mno-atomic-updates}
1033 @emph{System V Options}
1034 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1036 @emph{TILE-Gx Options}
1037 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1038 -mcmodel=@var{code-model}}
1040 @emph{TILEPro Options}
1041 @gccoptlist{-mcpu=@var{cpu} -m32}
1044 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1045 -mprolog-function -mno-prolog-function -mspace @gol
1046 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1047 -mapp-regs -mno-app-regs @gol
1048 -mdisable-callt -mno-disable-callt @gol
1049 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1050 -mv850e -mv850 -mv850e3v5 @gol
1061 @gccoptlist{-mg -mgnu -munix}
1063 @emph{Visium Options}
1064 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1065 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1068 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1069 -mpointer-size=@var{size}}
1071 @emph{VxWorks Options}
1072 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1073 -Xbind-lazy -Xbind-now}
1076 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1077 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1078 -mfpmath=@var{unit} @gol
1079 -masm=@var{dialect} -mno-fancy-math-387 @gol
1080 -mno-fp-ret-in-387 -msoft-float @gol
1081 -mno-wide-multiply -mrtd -malign-double @gol
1082 -mpreferred-stack-boundary=@var{num} @gol
1083 -mincoming-stack-boundary=@var{num} @gol
1084 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1085 -mrecip -mrecip=@var{opt} @gol
1086 -mvzeroupper -mprefer-avx128 @gol
1087 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1088 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -msha @gol
1089 -maes -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mprefetchwt1 @gol
1090 -mclflushopt -mxsavec -mxsaves @gol
1091 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1092 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mthreads @gol
1093 -mno-align-stringops -minline-all-stringops @gol
1094 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1095 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1096 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1097 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1098 -mregparm=@var{num} -msseregparm @gol
1099 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1100 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1101 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1102 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1103 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1104 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1105 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1106 -malign-data=@var{type} -mstack-protector-guard=@var{guard}}
1108 @emph{x86 Windows Options}
1109 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1110 -mnop-fun-dllimport -mthread @gol
1111 -municode -mwin32 -mwindows -fno-set-stack-executable}
1113 @emph{Xstormy16 Options}
1116 @emph{Xtensa Options}
1117 @gccoptlist{-mconst16 -mno-const16 @gol
1118 -mfused-madd -mno-fused-madd @gol
1120 -mserialize-volatile -mno-serialize-volatile @gol
1121 -mtext-section-literals -mno-text-section-literals @gol
1122 -mtarget-align -mno-target-align @gol
1123 -mlongcalls -mno-longcalls}
1125 @emph{zSeries Options}
1126 See S/390 and zSeries Options.
1128 @item Code Generation Options
1129 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1130 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1131 -ffixed-@var{reg} -fexceptions @gol
1132 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1133 -fasynchronous-unwind-tables @gol
1134 -fno-gnu-unique @gol
1135 -finhibit-size-directive -finstrument-functions @gol
1136 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1137 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1138 -fno-common -fno-ident @gol
1139 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
1140 -fno-jump-tables @gol
1141 -frecord-gcc-switches @gol
1142 -freg-struct-return -fshort-enums @gol
1143 -fshort-double -fshort-wchar @gol
1144 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1145 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1146 -fno-stack-limit -fsplit-stack @gol
1147 -fleading-underscore -ftls-model=@var{model} @gol
1148 -fstack-reuse=@var{reuse_level} @gol
1149 -ftrapv -fwrapv -fbounds-check @gol
1150 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1151 -fstrict-volatile-bitfields -fsync-libcalls}
1155 @node Overall Options
1156 @section Options Controlling the Kind of Output
1158 Compilation can involve up to four stages: preprocessing, compilation
1159 proper, assembly and linking, always in that order. GCC is capable of
1160 preprocessing and compiling several files either into several
1161 assembler input files, or into one assembler input file; then each
1162 assembler input file produces an object file, and linking combines all
1163 the object files (those newly compiled, and those specified as input)
1164 into an executable file.
1166 @cindex file name suffix
1167 For any given input file, the file name suffix determines what kind of
1168 compilation is done:
1172 C source code that must be preprocessed.
1175 C source code that should not be preprocessed.
1178 C++ source code that should not be preprocessed.
1181 Objective-C source code. Note that you must link with the @file{libobjc}
1182 library to make an Objective-C program work.
1185 Objective-C source code that should not be preprocessed.
1189 Objective-C++ source code. Note that you must link with the @file{libobjc}
1190 library to make an Objective-C++ program work. Note that @samp{.M} refers
1191 to a literal capital M@.
1193 @item @var{file}.mii
1194 Objective-C++ source code that should not be preprocessed.
1197 C, C++, Objective-C or Objective-C++ header file to be turned into a
1198 precompiled header (default), or C, C++ header file to be turned into an
1199 Ada spec (via the @option{-fdump-ada-spec} switch).
1202 @itemx @var{file}.cp
1203 @itemx @var{file}.cxx
1204 @itemx @var{file}.cpp
1205 @itemx @var{file}.CPP
1206 @itemx @var{file}.c++
1208 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1209 the last two letters must both be literally @samp{x}. Likewise,
1210 @samp{.C} refers to a literal capital C@.
1214 Objective-C++ source code that must be preprocessed.
1216 @item @var{file}.mii
1217 Objective-C++ source code that should not be preprocessed.
1221 @itemx @var{file}.hp
1222 @itemx @var{file}.hxx
1223 @itemx @var{file}.hpp
1224 @itemx @var{file}.HPP
1225 @itemx @var{file}.h++
1226 @itemx @var{file}.tcc
1227 C++ header file to be turned into a precompiled header or Ada spec.
1230 @itemx @var{file}.for
1231 @itemx @var{file}.ftn
1232 Fixed form Fortran source code that should not be preprocessed.
1235 @itemx @var{file}.FOR
1236 @itemx @var{file}.fpp
1237 @itemx @var{file}.FPP
1238 @itemx @var{file}.FTN
1239 Fixed form Fortran source code that must be preprocessed (with the traditional
1242 @item @var{file}.f90
1243 @itemx @var{file}.f95
1244 @itemx @var{file}.f03
1245 @itemx @var{file}.f08
1246 Free form Fortran source code that should not be preprocessed.
1248 @item @var{file}.F90
1249 @itemx @var{file}.F95
1250 @itemx @var{file}.F03
1251 @itemx @var{file}.F08
1252 Free form Fortran source code that must be preprocessed (with the
1253 traditional preprocessor).
1258 @c FIXME: Descriptions of Java file types.
1264 @item @var{file}.ads
1265 Ada source code file that contains a library unit declaration (a
1266 declaration of a package, subprogram, or generic, or a generic
1267 instantiation), or a library unit renaming declaration (a package,
1268 generic, or subprogram renaming declaration). Such files are also
1271 @item @var{file}.adb
1272 Ada source code file containing a library unit body (a subprogram or
1273 package body). Such files are also called @dfn{bodies}.
1275 @c GCC also knows about some suffixes for languages not yet included:
1286 @itemx @var{file}.sx
1287 Assembler code that must be preprocessed.
1290 An object file to be fed straight into linking.
1291 Any file name with no recognized suffix is treated this way.
1295 You can specify the input language explicitly with the @option{-x} option:
1298 @item -x @var{language}
1299 Specify explicitly the @var{language} for the following input files
1300 (rather than letting the compiler choose a default based on the file
1301 name suffix). This option applies to all following input files until
1302 the next @option{-x} option. Possible values for @var{language} are:
1304 c c-header cpp-output
1305 c++ c++-header c++-cpp-output
1306 objective-c objective-c-header objective-c-cpp-output
1307 objective-c++ objective-c++-header objective-c++-cpp-output
1308 assembler assembler-with-cpp
1310 f77 f77-cpp-input f95 f95-cpp-input
1316 Turn off any specification of a language, so that subsequent files are
1317 handled according to their file name suffixes (as they are if @option{-x}
1318 has not been used at all).
1320 @item -pass-exit-codes
1321 @opindex pass-exit-codes
1322 Normally the @command{gcc} program exits with the code of 1 if any
1323 phase of the compiler returns a non-success return code. If you specify
1324 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1325 the numerically highest error produced by any phase returning an error
1326 indication. The C, C++, and Fortran front ends return 4 if an internal
1327 compiler error is encountered.
1330 If you only want some of the stages of compilation, you can use
1331 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1332 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1333 @command{gcc} is to stop. Note that some combinations (for example,
1334 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1339 Compile or assemble the source files, but do not link. The linking
1340 stage simply is not done. The ultimate output is in the form of an
1341 object file for each source file.
1343 By default, the object file name for a source file is made by replacing
1344 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1346 Unrecognized input files, not requiring compilation or assembly, are
1351 Stop after the stage of compilation proper; do not assemble. The output
1352 is in the form of an assembler code file for each non-assembler input
1355 By default, the assembler file name for a source file is made by
1356 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1358 Input files that don't require compilation are ignored.
1362 Stop after the preprocessing stage; do not run the compiler proper. The
1363 output is in the form of preprocessed source code, which is sent to the
1366 Input files that don't require preprocessing are ignored.
1368 @cindex output file option
1371 Place output in file @var{file}. This applies to whatever
1372 sort of output is being produced, whether it be an executable file,
1373 an object file, an assembler file or preprocessed C code.
1375 If @option{-o} is not specified, the default is to put an executable
1376 file in @file{a.out}, the object file for
1377 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1378 assembler file in @file{@var{source}.s}, a precompiled header file in
1379 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1384 Print (on standard error output) the commands executed to run the stages
1385 of compilation. Also print the version number of the compiler driver
1386 program and of the preprocessor and the compiler proper.
1390 Like @option{-v} except the commands are not executed and arguments
1391 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1392 This is useful for shell scripts to capture the driver-generated command lines.
1396 Use pipes rather than temporary files for communication between the
1397 various stages of compilation. This fails to work on some systems where
1398 the assembler is unable to read from a pipe; but the GNU assembler has
1403 Print (on the standard output) a description of the command-line options
1404 understood by @command{gcc}. If the @option{-v} option is also specified
1405 then @option{--help} is also passed on to the various processes
1406 invoked by @command{gcc}, so that they can display the command-line options
1407 they accept. If the @option{-Wextra} option has also been specified
1408 (prior to the @option{--help} option), then command-line options that
1409 have no documentation associated with them are also displayed.
1412 @opindex target-help
1413 Print (on the standard output) a description of target-specific command-line
1414 options for each tool. For some targets extra target-specific
1415 information may also be printed.
1417 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1418 Print (on the standard output) a description of the command-line
1419 options understood by the compiler that fit into all specified classes
1420 and qualifiers. These are the supported classes:
1423 @item @samp{optimizers}
1424 Display all of the optimization options supported by the
1427 @item @samp{warnings}
1428 Display all of the options controlling warning messages
1429 produced by the compiler.
1432 Display target-specific options. Unlike the
1433 @option{--target-help} option however, target-specific options of the
1434 linker and assembler are not displayed. This is because those
1435 tools do not currently support the extended @option{--help=} syntax.
1438 Display the values recognized by the @option{--param}
1441 @item @var{language}
1442 Display the options supported for @var{language}, where
1443 @var{language} is the name of one of the languages supported in this
1447 Display the options that are common to all languages.
1450 These are the supported qualifiers:
1453 @item @samp{undocumented}
1454 Display only those options that are undocumented.
1457 Display options taking an argument that appears after an equal
1458 sign in the same continuous piece of text, such as:
1459 @samp{--help=target}.
1461 @item @samp{separate}
1462 Display options taking an argument that appears as a separate word
1463 following the original option, such as: @samp{-o output-file}.
1466 Thus for example to display all the undocumented target-specific
1467 switches supported by the compiler, use:
1470 --help=target,undocumented
1473 The sense of a qualifier can be inverted by prefixing it with the
1474 @samp{^} character, so for example to display all binary warning
1475 options (i.e., ones that are either on or off and that do not take an
1476 argument) that have a description, use:
1479 --help=warnings,^joined,^undocumented
1482 The argument to @option{--help=} should not consist solely of inverted
1485 Combining several classes is possible, although this usually
1486 restricts the output so much that there is nothing to display. One
1487 case where it does work, however, is when one of the classes is
1488 @var{target}. For example, to display all the target-specific
1489 optimization options, use:
1492 --help=target,optimizers
1495 The @option{--help=} option can be repeated on the command line. Each
1496 successive use displays its requested class of options, skipping
1497 those that have already been displayed.
1499 If the @option{-Q} option appears on the command line before the
1500 @option{--help=} option, then the descriptive text displayed by
1501 @option{--help=} is changed. Instead of describing the displayed
1502 options, an indication is given as to whether the option is enabled,
1503 disabled or set to a specific value (assuming that the compiler
1504 knows this at the point where the @option{--help=} option is used).
1506 Here is a truncated example from the ARM port of @command{gcc}:
1509 % gcc -Q -mabi=2 --help=target -c
1510 The following options are target specific:
1512 -mabort-on-noreturn [disabled]
1516 The output is sensitive to the effects of previous command-line
1517 options, so for example it is possible to find out which optimizations
1518 are enabled at @option{-O2} by using:
1521 -Q -O2 --help=optimizers
1524 Alternatively you can discover which binary optimizations are enabled
1525 by @option{-O3} by using:
1528 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1529 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1530 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1533 @item -no-canonical-prefixes
1534 @opindex no-canonical-prefixes
1535 Do not expand any symbolic links, resolve references to @samp{/../}
1536 or @samp{/./}, or make the path absolute when generating a relative
1541 Display the version number and copyrights of the invoked GCC@.
1545 Invoke all subcommands under a wrapper program. The name of the
1546 wrapper program and its parameters are passed as a comma separated
1550 gcc -c t.c -wrapper gdb,--args
1554 This invokes all subprograms of @command{gcc} under
1555 @samp{gdb --args}, thus the invocation of @command{cc1} is
1556 @samp{gdb --args cc1 @dots{}}.
1558 @item -fplugin=@var{name}.so
1560 Load the plugin code in file @var{name}.so, assumed to be a
1561 shared object to be dlopen'd by the compiler. The base name of
1562 the shared object file is used to identify the plugin for the
1563 purposes of argument parsing (See
1564 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1565 Each plugin should define the callback functions specified in the
1568 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1569 @opindex fplugin-arg
1570 Define an argument called @var{key} with a value of @var{value}
1571 for the plugin called @var{name}.
1573 @item -fdump-ada-spec@r{[}-slim@r{]}
1574 @opindex fdump-ada-spec
1575 For C and C++ source and include files, generate corresponding Ada specs.
1576 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1577 GNAT User's Guide}, which provides detailed documentation on this feature.
1579 @item -fada-spec-parent=@var{unit}
1580 @opindex fada-spec-parent
1581 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1582 Ada specs as child units of parent @var{unit}.
1584 @item -fdump-go-spec=@var{file}
1585 @opindex fdump-go-spec
1586 For input files in any language, generate corresponding Go
1587 declarations in @var{file}. This generates Go @code{const},
1588 @code{type}, @code{var}, and @code{func} declarations which may be a
1589 useful way to start writing a Go interface to code written in some
1592 @include @value{srcdir}/../libiberty/at-file.texi
1596 @section Compiling C++ Programs
1598 @cindex suffixes for C++ source
1599 @cindex C++ source file suffixes
1600 C++ source files conventionally use one of the suffixes @samp{.C},
1601 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1602 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1603 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1604 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1605 files with these names and compiles them as C++ programs even if you
1606 call the compiler the same way as for compiling C programs (usually
1607 with the name @command{gcc}).
1611 However, the use of @command{gcc} does not add the C++ library.
1612 @command{g++} is a program that calls GCC and automatically specifies linking
1613 against the C++ library. It treats @samp{.c},
1614 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1615 files unless @option{-x} is used. This program is also useful when
1616 precompiling a C header file with a @samp{.h} extension for use in C++
1617 compilations. On many systems, @command{g++} is also installed with
1618 the name @command{c++}.
1620 @cindex invoking @command{g++}
1621 When you compile C++ programs, you may specify many of the same
1622 command-line options that you use for compiling programs in any
1623 language; or command-line options meaningful for C and related
1624 languages; or options that are meaningful only for C++ programs.
1625 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1626 explanations of options for languages related to C@.
1627 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1628 explanations of options that are meaningful only for C++ programs.
1630 @node C Dialect Options
1631 @section Options Controlling C Dialect
1632 @cindex dialect options
1633 @cindex language dialect options
1634 @cindex options, dialect
1636 The following options control the dialect of C (or languages derived
1637 from C, such as C++, Objective-C and Objective-C++) that the compiler
1641 @cindex ANSI support
1645 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1646 equivalent to @option{-std=c++98}.
1648 This turns off certain features of GCC that are incompatible with ISO
1649 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1650 such as the @code{asm} and @code{typeof} keywords, and
1651 predefined macros such as @code{unix} and @code{vax} that identify the
1652 type of system you are using. It also enables the undesirable and
1653 rarely used ISO trigraph feature. For the C compiler,
1654 it disables recognition of C++ style @samp{//} comments as well as
1655 the @code{inline} keyword.
1657 The alternate keywords @code{__asm__}, @code{__extension__},
1658 @code{__inline__} and @code{__typeof__} continue to work despite
1659 @option{-ansi}. You would not want to use them in an ISO C program, of
1660 course, but it is useful to put them in header files that might be included
1661 in compilations done with @option{-ansi}. Alternate predefined macros
1662 such as @code{__unix__} and @code{__vax__} are also available, with or
1663 without @option{-ansi}.
1665 The @option{-ansi} option does not cause non-ISO programs to be
1666 rejected gratuitously. For that, @option{-Wpedantic} is required in
1667 addition to @option{-ansi}. @xref{Warning Options}.
1669 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1670 option is used. Some header files may notice this macro and refrain
1671 from declaring certain functions or defining certain macros that the
1672 ISO standard doesn't call for; this is to avoid interfering with any
1673 programs that might use these names for other things.
1675 Functions that are normally built in but do not have semantics
1676 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1677 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1678 built-in functions provided by GCC}, for details of the functions
1683 Determine the language standard. @xref{Standards,,Language Standards
1684 Supported by GCC}, for details of these standard versions. This option
1685 is currently only supported when compiling C or C++.
1687 The compiler can accept several base standards, such as @samp{c90} or
1688 @samp{c++98}, and GNU dialects of those standards, such as
1689 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1690 compiler accepts all programs following that standard plus those
1691 using GNU extensions that do not contradict it. For example,
1692 @option{-std=c90} turns off certain features of GCC that are
1693 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1694 keywords, but not other GNU extensions that do not have a meaning in
1695 ISO C90, such as omitting the middle term of a @code{?:}
1696 expression. On the other hand, when a GNU dialect of a standard is
1697 specified, all features supported by the compiler are enabled, even when
1698 those features change the meaning of the base standard. As a result, some
1699 strict-conforming programs may be rejected. The particular standard
1700 is used by @option{-Wpedantic} to identify which features are GNU
1701 extensions given that version of the standard. For example
1702 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1703 comments, while @option{-std=gnu99 -Wpedantic} does not.
1705 A value for this option must be provided; possible values are
1711 Support all ISO C90 programs (certain GNU extensions that conflict
1712 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1714 @item iso9899:199409
1715 ISO C90 as modified in amendment 1.
1721 ISO C99. This standard is substantially completely supported, modulo
1722 bugs and floating-point issues
1723 (mainly but not entirely relating to optional C99 features from
1724 Annexes F and G). See
1725 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1726 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1731 ISO C11, the 2011 revision of the ISO C standard. This standard is
1732 substantially completely supported, modulo bugs, floating-point issues
1733 (mainly but not entirely relating to optional C11 features from
1734 Annexes F and G) and the optional Annexes K (Bounds-checking
1735 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1739 GNU dialect of ISO C90 (including some C99 features).
1743 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1747 GNU dialect of ISO C11. This is the default for C code.
1748 The name @samp{gnu1x} is deprecated.
1752 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1753 additional defect reports. Same as @option{-ansi} for C++ code.
1757 GNU dialect of @option{-std=c++98}. This is the default for
1762 The 2011 ISO C++ standard plus amendments.
1763 The name @samp{c++0x} is deprecated.
1767 GNU dialect of @option{-std=c++11}.
1768 The name @samp{gnu++0x} is deprecated.
1772 The 2014 ISO C++ standard plus amendments.
1773 The name @samp{c++1y} is deprecated.
1777 GNU dialect of @option{-std=c++14}.
1778 The name @samp{gnu++1y} is deprecated.
1781 The next revision of the ISO C++ standard, tentatively planned for
1782 2017. Support is highly experimental, and will almost certainly
1783 change in incompatible ways in future releases.
1786 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1787 and will almost certainly change in incompatible ways in future
1791 @item -fgnu89-inline
1792 @opindex fgnu89-inline
1793 The option @option{-fgnu89-inline} tells GCC to use the traditional
1794 GNU semantics for @code{inline} functions when in C99 mode.
1795 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1796 Using this option is roughly equivalent to adding the
1797 @code{gnu_inline} function attribute to all inline functions
1798 (@pxref{Function Attributes}).
1800 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1801 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1802 specifies the default behavior).
1803 This option is not supported in @option{-std=c90} or
1804 @option{-std=gnu90} mode.
1806 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1807 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1808 in effect for @code{inline} functions. @xref{Common Predefined
1809 Macros,,,cpp,The C Preprocessor}.
1811 @item -aux-info @var{filename}
1813 Output to the given filename prototyped declarations for all functions
1814 declared and/or defined in a translation unit, including those in header
1815 files. This option is silently ignored in any language other than C@.
1817 Besides declarations, the file indicates, in comments, the origin of
1818 each declaration (source file and line), whether the declaration was
1819 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1820 @samp{O} for old, respectively, in the first character after the line
1821 number and the colon), and whether it came from a declaration or a
1822 definition (@samp{C} or @samp{F}, respectively, in the following
1823 character). In the case of function definitions, a K&R-style list of
1824 arguments followed by their declarations is also provided, inside
1825 comments, after the declaration.
1827 @item -fallow-parameterless-variadic-functions
1828 @opindex fallow-parameterless-variadic-functions
1829 Accept variadic functions without named parameters.
1831 Although it is possible to define such a function, this is not very
1832 useful as it is not possible to read the arguments. This is only
1833 supported for C as this construct is allowed by C++.
1837 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1838 keyword, so that code can use these words as identifiers. You can use
1839 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1840 instead. @option{-ansi} implies @option{-fno-asm}.
1842 In C++, this switch only affects the @code{typeof} keyword, since
1843 @code{asm} and @code{inline} are standard keywords. You may want to
1844 use the @option{-fno-gnu-keywords} flag instead, which has the same
1845 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1846 switch only affects the @code{asm} and @code{typeof} keywords, since
1847 @code{inline} is a standard keyword in ISO C99.
1850 @itemx -fno-builtin-@var{function}
1851 @opindex fno-builtin
1852 @cindex built-in functions
1853 Don't recognize built-in functions that do not begin with
1854 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1855 functions provided by GCC}, for details of the functions affected,
1856 including those which are not built-in functions when @option{-ansi} or
1857 @option{-std} options for strict ISO C conformance are used because they
1858 do not have an ISO standard meaning.
1860 GCC normally generates special code to handle certain built-in functions
1861 more efficiently; for instance, calls to @code{alloca} may become single
1862 instructions which adjust the stack directly, and calls to @code{memcpy}
1863 may become inline copy loops. The resulting code is often both smaller
1864 and faster, but since the function calls no longer appear as such, you
1865 cannot set a breakpoint on those calls, nor can you change the behavior
1866 of the functions by linking with a different library. In addition,
1867 when a function is recognized as a built-in function, GCC may use
1868 information about that function to warn about problems with calls to
1869 that function, or to generate more efficient code, even if the
1870 resulting code still contains calls to that function. For example,
1871 warnings are given with @option{-Wformat} for bad calls to
1872 @code{printf} when @code{printf} is built in and @code{strlen} is
1873 known not to modify global memory.
1875 With the @option{-fno-builtin-@var{function}} option
1876 only the built-in function @var{function} is
1877 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1878 function is named that is not built-in in this version of GCC, this
1879 option is ignored. There is no corresponding
1880 @option{-fbuiltin-@var{function}} option; if you wish to enable
1881 built-in functions selectively when using @option{-fno-builtin} or
1882 @option{-ffreestanding}, you may define macros such as:
1885 #define abs(n) __builtin_abs ((n))
1886 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1891 @cindex hosted environment
1893 Assert that compilation targets a hosted environment. This implies
1894 @option{-fbuiltin}. A hosted environment is one in which the
1895 entire standard library is available, and in which @code{main} has a return
1896 type of @code{int}. Examples are nearly everything except a kernel.
1897 This is equivalent to @option{-fno-freestanding}.
1899 @item -ffreestanding
1900 @opindex ffreestanding
1901 @cindex hosted environment
1903 Assert that compilation targets a freestanding environment. This
1904 implies @option{-fno-builtin}. A freestanding environment
1905 is one in which the standard library may not exist, and program startup may
1906 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1907 This is equivalent to @option{-fno-hosted}.
1909 @xref{Standards,,Language Standards Supported by GCC}, for details of
1910 freestanding and hosted environments.
1914 @cindex OpenACC accelerator programming
1915 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1916 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1917 compiler generates accelerated code according to the OpenACC Application
1918 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1919 implies @option{-pthread}, and thus is only supported on targets that
1920 have support for @option{-pthread}.
1922 Note that this is an experimental feature, incomplete, and subject to
1923 change in future versions of GCC. See
1924 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1928 @cindex OpenMP parallel
1929 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1930 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1931 compiler generates parallel code according to the OpenMP Application
1932 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1933 implies @option{-pthread}, and thus is only supported on targets that
1934 have support for @option{-pthread}. @option{-fopenmp} implies
1935 @option{-fopenmp-simd}.
1938 @opindex fopenmp-simd
1941 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1942 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1947 @cindex Enable Cilk Plus
1948 Enable the usage of Cilk Plus language extension features for C/C++.
1949 When the option @option{-fcilkplus} is specified, enable the usage of
1950 the Cilk Plus Language extension features for C/C++. The present
1951 implementation follows ABI version 1.2. This is an experimental
1952 feature that is only partially complete, and whose interface may
1953 change in future versions of GCC as the official specification
1954 changes. Currently, all features but @code{_Cilk_for} have been
1959 When the option @option{-fgnu-tm} is specified, the compiler
1960 generates code for the Linux variant of Intel's current Transactional
1961 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1962 an experimental feature whose interface may change in future versions
1963 of GCC, as the official specification changes. Please note that not
1964 all architectures are supported for this feature.
1966 For more information on GCC's support for transactional memory,
1967 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1968 Transactional Memory Library}.
1970 Note that the transactional memory feature is not supported with
1971 non-call exceptions (@option{-fnon-call-exceptions}).
1973 @item -fms-extensions
1974 @opindex fms-extensions
1975 Accept some non-standard constructs used in Microsoft header files.
1977 In C++ code, this allows member names in structures to be similar
1978 to previous types declarations.
1987 Some cases of unnamed fields in structures and unions are only
1988 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1989 fields within structs/unions}, for details.
1991 Note that this option is off for all targets but x86
1992 targets using ms-abi.
1994 @item -fplan9-extensions
1995 @opindex fplan9-extensions
1996 Accept some non-standard constructs used in Plan 9 code.
1998 This enables @option{-fms-extensions}, permits passing pointers to
1999 structures with anonymous fields to functions that expect pointers to
2000 elements of the type of the field, and permits referring to anonymous
2001 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2002 struct/union fields within structs/unions}, for details. This is only
2003 supported for C, not C++.
2007 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2008 options for strict ISO C conformance) implies @option{-trigraphs}.
2010 @cindex traditional C language
2011 @cindex C language, traditional
2013 @itemx -traditional-cpp
2014 @opindex traditional-cpp
2015 @opindex traditional
2016 Formerly, these options caused GCC to attempt to emulate a pre-standard
2017 C compiler. They are now only supported with the @option{-E} switch.
2018 The preprocessor continues to support a pre-standard mode. See the GNU
2019 CPP manual for details.
2021 @item -fcond-mismatch
2022 @opindex fcond-mismatch
2023 Allow conditional expressions with mismatched types in the second and
2024 third arguments. The value of such an expression is void. This option
2025 is not supported for C++.
2027 @item -flax-vector-conversions
2028 @opindex flax-vector-conversions
2029 Allow implicit conversions between vectors with differing numbers of
2030 elements and/or incompatible element types. This option should not be
2033 @item -funsigned-char
2034 @opindex funsigned-char
2035 Let the type @code{char} be unsigned, like @code{unsigned char}.
2037 Each kind of machine has a default for what @code{char} should
2038 be. It is either like @code{unsigned char} by default or like
2039 @code{signed char} by default.
2041 Ideally, a portable program should always use @code{signed char} or
2042 @code{unsigned char} when it depends on the signedness of an object.
2043 But many programs have been written to use plain @code{char} and
2044 expect it to be signed, or expect it to be unsigned, depending on the
2045 machines they were written for. This option, and its inverse, let you
2046 make such a program work with the opposite default.
2048 The type @code{char} is always a distinct type from each of
2049 @code{signed char} or @code{unsigned char}, even though its behavior
2050 is always just like one of those two.
2053 @opindex fsigned-char
2054 Let the type @code{char} be signed, like @code{signed char}.
2056 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2057 the negative form of @option{-funsigned-char}. Likewise, the option
2058 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2060 @item -fsigned-bitfields
2061 @itemx -funsigned-bitfields
2062 @itemx -fno-signed-bitfields
2063 @itemx -fno-unsigned-bitfields
2064 @opindex fsigned-bitfields
2065 @opindex funsigned-bitfields
2066 @opindex fno-signed-bitfields
2067 @opindex fno-unsigned-bitfields
2068 These options control whether a bit-field is signed or unsigned, when the
2069 declaration does not use either @code{signed} or @code{unsigned}. By
2070 default, such a bit-field is signed, because this is consistent: the
2071 basic integer types such as @code{int} are signed types.
2074 @node C++ Dialect Options
2075 @section Options Controlling C++ Dialect
2077 @cindex compiler options, C++
2078 @cindex C++ options, command-line
2079 @cindex options, C++
2080 This section describes the command-line options that are only meaningful
2081 for C++ programs. You can also use most of the GNU compiler options
2082 regardless of what language your program is in. For example, you
2083 might compile a file @file{firstClass.C} like this:
2086 g++ -g -frepo -O -c firstClass.C
2090 In this example, only @option{-frepo} is an option meant
2091 only for C++ programs; you can use the other options with any
2092 language supported by GCC@.
2094 Here is a list of options that are @emph{only} for compiling C++ programs:
2098 @item -fabi-version=@var{n}
2099 @opindex fabi-version
2100 Use version @var{n} of the C++ ABI@. The default is version 0.
2102 Version 0 refers to the version conforming most closely to
2103 the C++ ABI specification. Therefore, the ABI obtained using version 0
2104 will change in different versions of G++ as ABI bugs are fixed.
2106 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2108 Version 2 is the version of the C++ ABI that first appeared in G++
2109 3.4, and was the default through G++ 4.9.
2111 Version 3 corrects an error in mangling a constant address as a
2114 Version 4, which first appeared in G++ 4.5, implements a standard
2115 mangling for vector types.
2117 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2118 attribute const/volatile on function pointer types, decltype of a
2119 plain decl, and use of a function parameter in the declaration of
2122 Version 6, which first appeared in G++ 4.7, corrects the promotion
2123 behavior of C++11 scoped enums and the mangling of template argument
2124 packs, const/static_cast, prefix ++ and --, and a class scope function
2125 used as a template argument.
2127 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2128 builtin type and corrects the mangling of lambdas in default argument
2131 Version 8, which first appeared in G++ 4.9, corrects the substitution
2132 behavior of function types with function-cv-qualifiers.
2134 See also @option{-Wabi}.
2136 @item -fabi-compat-version=@var{n}
2137 @opindex fabi-compat-version
2138 On targets that support strong aliases, G++
2139 works around mangling changes by creating an alias with the correct
2140 mangled name when defining a symbol with an incorrect mangled name.
2141 This switch specifies which ABI version to use for the alias.
2143 With @option{-fabi-version=0} (the default), this defaults to 2. If
2144 another ABI version is explicitly selected, this defaults to 0.
2146 The compatibility version is also set by @option{-Wabi=@var{n}}.
2148 @item -fno-access-control
2149 @opindex fno-access-control
2150 Turn off all access checking. This switch is mainly useful for working
2151 around bugs in the access control code.
2155 Check that the pointer returned by @code{operator new} is non-null
2156 before attempting to modify the storage allocated. This check is
2157 normally unnecessary because the C++ standard specifies that
2158 @code{operator new} only returns @code{0} if it is declared
2159 @code{throw()}, in which case the compiler always checks the
2160 return value even without this option. In all other cases, when
2161 @code{operator new} has a non-empty exception specification, memory
2162 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2163 @samp{new (nothrow)}.
2165 @item -fconstexpr-depth=@var{n}
2166 @opindex fconstexpr-depth
2167 Set the maximum nested evaluation depth for C++11 constexpr functions
2168 to @var{n}. A limit is needed to detect endless recursion during
2169 constant expression evaluation. The minimum specified by the standard
2172 @item -fdeduce-init-list
2173 @opindex fdeduce-init-list
2174 Enable deduction of a template type parameter as
2175 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2178 template <class T> auto forward(T t) -> decltype (realfn (t))
2185 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2189 This deduction was implemented as a possible extension to the
2190 originally proposed semantics for the C++11 standard, but was not part
2191 of the final standard, so it is disabled by default. This option is
2192 deprecated, and may be removed in a future version of G++.
2194 @item -ffriend-injection
2195 @opindex ffriend-injection
2196 Inject friend functions into the enclosing namespace, so that they are
2197 visible outside the scope of the class in which they are declared.
2198 Friend functions were documented to work this way in the old Annotated
2199 C++ Reference Manual.
2200 However, in ISO C++ a friend function that is not declared
2201 in an enclosing scope can only be found using argument dependent
2202 lookup. GCC defaults to the standard behavior.
2204 This option is for compatibility, and may be removed in a future
2207 @item -fno-elide-constructors
2208 @opindex fno-elide-constructors
2209 The C++ standard allows an implementation to omit creating a temporary
2210 that is only used to initialize another object of the same type.
2211 Specifying this option disables that optimization, and forces G++ to
2212 call the copy constructor in all cases.
2214 @item -fno-enforce-eh-specs
2215 @opindex fno-enforce-eh-specs
2216 Don't generate code to check for violation of exception specifications
2217 at run time. This option violates the C++ standard, but may be useful
2218 for reducing code size in production builds, much like defining
2219 @code{NDEBUG}. This does not give user code permission to throw
2220 exceptions in violation of the exception specifications; the compiler
2221 still optimizes based on the specifications, so throwing an
2222 unexpected exception results in undefined behavior at run time.
2224 @item -fextern-tls-init
2225 @itemx -fno-extern-tls-init
2226 @opindex fextern-tls-init
2227 @opindex fno-extern-tls-init
2228 The C++11 and OpenMP standards allow @code{thread_local} and
2229 @code{threadprivate} variables to have dynamic (runtime)
2230 initialization. To support this, any use of such a variable goes
2231 through a wrapper function that performs any necessary initialization.
2232 When the use and definition of the variable are in the same
2233 translation unit, this overhead can be optimized away, but when the
2234 use is in a different translation unit there is significant overhead
2235 even if the variable doesn't actually need dynamic initialization. If
2236 the programmer can be sure that no use of the variable in a
2237 non-defining TU needs to trigger dynamic initialization (either
2238 because the variable is statically initialized, or a use of the
2239 variable in the defining TU will be executed before any uses in
2240 another TU), they can avoid this overhead with the
2241 @option{-fno-extern-tls-init} option.
2243 On targets that support symbol aliases, the default is
2244 @option{-fextern-tls-init}. On targets that do not support symbol
2245 aliases, the default is @option{-fno-extern-tls-init}.
2248 @itemx -fno-for-scope
2250 @opindex fno-for-scope
2251 If @option{-ffor-scope} is specified, the scope of variables declared in
2252 a @i{for-init-statement} is limited to the @code{for} loop itself,
2253 as specified by the C++ standard.
2254 If @option{-fno-for-scope} is specified, the scope of variables declared in
2255 a @i{for-init-statement} extends to the end of the enclosing scope,
2256 as was the case in old versions of G++, and other (traditional)
2257 implementations of C++.
2259 If neither flag is given, the default is to follow the standard,
2260 but to allow and give a warning for old-style code that would
2261 otherwise be invalid, or have different behavior.
2263 @item -fno-gnu-keywords
2264 @opindex fno-gnu-keywords
2265 Do not recognize @code{typeof} as a keyword, so that code can use this
2266 word as an identifier. You can use the keyword @code{__typeof__} instead.
2267 @option{-ansi} implies @option{-fno-gnu-keywords}.
2269 @item -fno-implicit-templates
2270 @opindex fno-implicit-templates
2271 Never emit code for non-inline templates that are instantiated
2272 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2273 @xref{Template Instantiation}, for more information.
2275 @item -fno-implicit-inline-templates
2276 @opindex fno-implicit-inline-templates
2277 Don't emit code for implicit instantiations of inline templates, either.
2278 The default is to handle inlines differently so that compiles with and
2279 without optimization need the same set of explicit instantiations.
2281 @item -fno-implement-inlines
2282 @opindex fno-implement-inlines
2283 To save space, do not emit out-of-line copies of inline functions
2284 controlled by @code{#pragma implementation}. This causes linker
2285 errors if these functions are not inlined everywhere they are called.
2287 @item -fms-extensions
2288 @opindex fms-extensions
2289 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2290 int and getting a pointer to member function via non-standard syntax.
2292 @item -fno-nonansi-builtins
2293 @opindex fno-nonansi-builtins
2294 Disable built-in declarations of functions that are not mandated by
2295 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2296 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2299 @opindex fnothrow-opt
2300 Treat a @code{throw()} exception specification as if it were a
2301 @code{noexcept} specification to reduce or eliminate the text size
2302 overhead relative to a function with no exception specification. If
2303 the function has local variables of types with non-trivial
2304 destructors, the exception specification actually makes the
2305 function smaller because the EH cleanups for those variables can be
2306 optimized away. The semantic effect is that an exception thrown out of
2307 a function with such an exception specification results in a call
2308 to @code{terminate} rather than @code{unexpected}.
2310 @item -fno-operator-names
2311 @opindex fno-operator-names
2312 Do not treat the operator name keywords @code{and}, @code{bitand},
2313 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2314 synonyms as keywords.
2316 @item -fno-optional-diags
2317 @opindex fno-optional-diags
2318 Disable diagnostics that the standard says a compiler does not need to
2319 issue. Currently, the only such diagnostic issued by G++ is the one for
2320 a name having multiple meanings within a class.
2323 @opindex fpermissive
2324 Downgrade some diagnostics about nonconformant code from errors to
2325 warnings. Thus, using @option{-fpermissive} allows some
2326 nonconforming code to compile.
2328 @item -fno-pretty-templates
2329 @opindex fno-pretty-templates
2330 When an error message refers to a specialization of a function
2331 template, the compiler normally prints the signature of the
2332 template followed by the template arguments and any typedefs or
2333 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2334 rather than @code{void f(int)}) so that it's clear which template is
2335 involved. When an error message refers to a specialization of a class
2336 template, the compiler omits any template arguments that match
2337 the default template arguments for that template. If either of these
2338 behaviors make it harder to understand the error message rather than
2339 easier, you can use @option{-fno-pretty-templates} to disable them.
2343 Enable automatic template instantiation at link time. This option also
2344 implies @option{-fno-implicit-templates}. @xref{Template
2345 Instantiation}, for more information.
2349 Disable generation of information about every class with virtual
2350 functions for use by the C++ run-time type identification features
2351 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2352 of the language, you can save some space by using this flag. Note that
2353 exception handling uses the same information, but G++ generates it as
2354 needed. The @code{dynamic_cast} operator can still be used for casts that
2355 do not require run-time type information, i.e.@: casts to @code{void *} or to
2356 unambiguous base classes.
2358 @item -fsized-deallocation
2359 @opindex fsized-deallocation
2360 Enable the built-in global declarations
2362 void operator delete (void *, std::size_t) noexcept;
2363 void operator delete[] (void *, std::size_t) noexcept;
2365 as introduced in C++14. This is useful for user-defined replacement
2366 deallocation functions that, for example, use the size of the object
2367 to make deallocation faster. Enabled by default under
2368 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2369 warns about places that might want to add a definition.
2373 Emit statistics about front-end processing at the end of the compilation.
2374 This information is generally only useful to the G++ development team.
2376 @item -fstrict-enums
2377 @opindex fstrict-enums
2378 Allow the compiler to optimize using the assumption that a value of
2379 enumerated type can only be one of the values of the enumeration (as
2380 defined in the C++ standard; basically, a value that can be
2381 represented in the minimum number of bits needed to represent all the
2382 enumerators). This assumption may not be valid if the program uses a
2383 cast to convert an arbitrary integer value to the enumerated type.
2385 @item -ftemplate-backtrace-limit=@var{n}
2386 @opindex ftemplate-backtrace-limit
2387 Set the maximum number of template instantiation notes for a single
2388 warning or error to @var{n}. The default value is 10.
2390 @item -ftemplate-depth=@var{n}
2391 @opindex ftemplate-depth
2392 Set the maximum instantiation depth for template classes to @var{n}.
2393 A limit on the template instantiation depth is needed to detect
2394 endless recursions during template class instantiation. ANSI/ISO C++
2395 conforming programs must not rely on a maximum depth greater than 17
2396 (changed to 1024 in C++11). The default value is 900, as the compiler
2397 can run out of stack space before hitting 1024 in some situations.
2399 @item -fno-threadsafe-statics
2400 @opindex fno-threadsafe-statics
2401 Do not emit the extra code to use the routines specified in the C++
2402 ABI for thread-safe initialization of local statics. You can use this
2403 option to reduce code size slightly in code that doesn't need to be
2406 @item -fuse-cxa-atexit
2407 @opindex fuse-cxa-atexit
2408 Register destructors for objects with static storage duration with the
2409 @code{__cxa_atexit} function rather than the @code{atexit} function.
2410 This option is required for fully standards-compliant handling of static
2411 destructors, but only works if your C library supports
2412 @code{__cxa_atexit}.
2414 @item -fno-use-cxa-get-exception-ptr
2415 @opindex fno-use-cxa-get-exception-ptr
2416 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2417 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2418 if the runtime routine is not available.
2420 @item -fvisibility-inlines-hidden
2421 @opindex fvisibility-inlines-hidden
2422 This switch declares that the user does not attempt to compare
2423 pointers to inline functions or methods where the addresses of the two functions
2424 are taken in different shared objects.
2426 The effect of this is that GCC may, effectively, mark inline methods with
2427 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2428 appear in the export table of a DSO and do not require a PLT indirection
2429 when used within the DSO@. Enabling this option can have a dramatic effect
2430 on load and link times of a DSO as it massively reduces the size of the
2431 dynamic export table when the library makes heavy use of templates.
2433 The behavior of this switch is not quite the same as marking the
2434 methods as hidden directly, because it does not affect static variables
2435 local to the function or cause the compiler to deduce that
2436 the function is defined in only one shared object.
2438 You may mark a method as having a visibility explicitly to negate the
2439 effect of the switch for that method. For example, if you do want to
2440 compare pointers to a particular inline method, you might mark it as
2441 having default visibility. Marking the enclosing class with explicit
2442 visibility has no effect.
2444 Explicitly instantiated inline methods are unaffected by this option
2445 as their linkage might otherwise cross a shared library boundary.
2446 @xref{Template Instantiation}.
2448 @item -fvisibility-ms-compat
2449 @opindex fvisibility-ms-compat
2450 This flag attempts to use visibility settings to make GCC's C++
2451 linkage model compatible with that of Microsoft Visual Studio.
2453 The flag makes these changes to GCC's linkage model:
2457 It sets the default visibility to @code{hidden}, like
2458 @option{-fvisibility=hidden}.
2461 Types, but not their members, are not hidden by default.
2464 The One Definition Rule is relaxed for types without explicit
2465 visibility specifications that are defined in more than one
2466 shared object: those declarations are permitted if they are
2467 permitted when this option is not used.
2470 In new code it is better to use @option{-fvisibility=hidden} and
2471 export those classes that are intended to be externally visible.
2472 Unfortunately it is possible for code to rely, perhaps accidentally,
2473 on the Visual Studio behavior.
2475 Among the consequences of these changes are that static data members
2476 of the same type with the same name but defined in different shared
2477 objects are different, so changing one does not change the other;
2478 and that pointers to function members defined in different shared
2479 objects may not compare equal. When this flag is given, it is a
2480 violation of the ODR to define types with the same name differently.
2482 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2483 @opindex fvtable-verify
2484 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2485 feature that verifies at run time, for every virtual call, that
2486 the vtable pointer through which the call is made is valid for the type of
2487 the object, and has not been corrupted or overwritten. If an invalid vtable
2488 pointer is detected at run time, an error is reported and execution of the
2489 program is immediately halted.
2491 This option causes run-time data structures to be built at program startup,
2492 which are used for verifying the vtable pointers.
2493 The options @samp{std} and @samp{preinit}
2494 control the timing of when these data structures are built. In both cases the
2495 data structures are built before execution reaches @code{main}. Using
2496 @option{-fvtable-verify=std} causes the data structures to be built after
2497 shared libraries have been loaded and initialized.
2498 @option{-fvtable-verify=preinit} causes them to be built before shared
2499 libraries have been loaded and initialized.
2501 If this option appears multiple times in the command line with different
2502 values specified, @samp{none} takes highest priority over both @samp{std} and
2503 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2507 When used in conjunction with @option{-fvtable-verify=std} or
2508 @option{-fvtable-verify=preinit}, causes debug versions of the
2509 runtime functions for the vtable verification feature to be called.
2510 This flag also causes the compiler to log information about which
2511 vtable pointers it finds for each class.
2512 This information is written to a file named @file{vtv_set_ptr_data.log}
2513 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2514 if that is defined or the current working directory otherwise.
2516 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2517 file, be sure to delete any existing one.
2520 @opindex fvtv-counts
2521 This is a debugging flag. When used in conjunction with
2522 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2523 causes the compiler to keep track of the total number of virtual calls
2524 it encounters and the number of verifications it inserts. It also
2525 counts the number of calls to certain run-time library functions
2526 that it inserts and logs this information for each compilation unit.
2527 The compiler writes this information to a file named
2528 @file{vtv_count_data.log} in the directory named by the environment
2529 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2530 directory otherwise. It also counts the size of the vtable pointer sets
2531 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2532 in the same directory.
2534 Note: This feature @emph{appends} data to the log files. To get fresh log
2535 files, be sure to delete any existing ones.
2539 Do not use weak symbol support, even if it is provided by the linker.
2540 By default, G++ uses weak symbols if they are available. This
2541 option exists only for testing, and should not be used by end-users;
2542 it results in inferior code and has no benefits. This option may
2543 be removed in a future release of G++.
2547 Do not search for header files in the standard directories specific to
2548 C++, but do still search the other standard directories. (This option
2549 is used when building the C++ library.)
2552 In addition, these optimization, warning, and code generation options
2553 have meanings only for C++ programs:
2556 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2559 When an explicit @option{-fabi-version=@var{n}} option is used, causes
2560 G++ to warn when it generates code that is probably not compatible with the
2561 vendor-neutral C++ ABI@. Since G++ now defaults to
2562 @option{-fabi-version=0}, @option{-Wabi} has no effect unless either
2563 an older ABI version is selected (with @option{-fabi-version=@var{n}})
2564 or an older compatibility version is selected (with
2565 @option{-Wabi=@var{n}} or @option{-fabi-compat-version=@var{n}}).
2567 Although an effort has been made to warn about
2568 all such cases, there are probably some cases that are not warned about,
2569 even though G++ is generating incompatible code. There may also be
2570 cases where warnings are emitted even though the code that is generated
2573 You should rewrite your code to avoid these warnings if you are
2574 concerned about the fact that code generated by G++ may not be binary
2575 compatible with code generated by other compilers.
2577 @option{-Wabi} can also be used with an explicit version number to
2578 warn about compatibility with a particular @option{-fabi-version}
2579 level, e.g. @option{-Wabi=2} to warn about changes relative to
2580 @option{-fabi-version=2}. Specifying a version number also sets
2581 @option{-fabi-compat-version=@var{n}}.
2583 The known incompatibilities in @option{-fabi-version=2} (which was the
2584 default from GCC 3.4 to 4.9) include:
2589 A template with a non-type template parameter of reference type was
2590 mangled incorrectly:
2593 template <int &> struct S @{@};
2597 This was fixed in @option{-fabi-version=3}.
2600 SIMD vector types declared using @code{__attribute ((vector_size))} were
2601 mangled in a non-standard way that does not allow for overloading of
2602 functions taking vectors of different sizes.
2604 The mangling was changed in @option{-fabi-version=4}.
2607 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2608 qualifiers, and @code{decltype} of a plain declaration was folded away.
2610 These mangling issues were fixed in @option{-fabi-version=5}.
2613 Scoped enumerators passed as arguments to a variadic function are
2614 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2615 On most targets this does not actually affect the parameter passing
2616 ABI, as there is no way to pass an argument smaller than @code{int}.
2618 Also, the ABI changed the mangling of template argument packs,
2619 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2620 a class scope function used as a template argument.
2622 These issues were corrected in @option{-fabi-version=6}.
2625 Lambdas in default argument scope were mangled incorrectly, and the
2626 ABI changed the mangling of @code{nullptr_t}.
2628 These issues were corrected in @option{-fabi-version=7}.
2631 When mangling a function type with function-cv-qualifiers, the
2632 un-qualified function type was incorrectly treated as a substitution
2635 This was fixed in @option{-fabi-version=8}.
2638 It also warns about psABI-related changes. The known psABI changes at this
2644 For SysV/x86-64, unions with @code{long double} members are
2645 passed in memory as specified in psABI. For example:
2655 @code{union U} is always passed in memory.
2659 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2662 Warn when a type with an ABI tag is used in a context that does not
2663 have that ABI tag. See @ref{C++ Attributes} for more information
2666 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2667 @opindex Wctor-dtor-privacy
2668 @opindex Wno-ctor-dtor-privacy
2669 Warn when a class seems unusable because all the constructors or
2670 destructors in that class are private, and it has neither friends nor
2671 public static member functions. Also warn if there are no non-private
2672 methods, and there's at least one private member function that isn't
2673 a constructor or destructor.
2675 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2676 @opindex Wdelete-non-virtual-dtor
2677 @opindex Wno-delete-non-virtual-dtor
2678 Warn when @code{delete} is used to destroy an instance of a class that
2679 has virtual functions and non-virtual destructor. It is unsafe to delete
2680 an instance of a derived class through a pointer to a base class if the
2681 base class does not have a virtual destructor. This warning is enabled
2684 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2685 @opindex Wliteral-suffix
2686 @opindex Wno-literal-suffix
2687 Warn when a string or character literal is followed by a ud-suffix which does
2688 not begin with an underscore. As a conforming extension, GCC treats such
2689 suffixes as separate preprocessing tokens in order to maintain backwards
2690 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2694 #define __STDC_FORMAT_MACROS
2695 #include <inttypes.h>
2700 printf("My int64: %" PRId64"\n", i64);
2704 In this case, @code{PRId64} is treated as a separate preprocessing token.
2706 This warning is enabled by default.
2708 @item -Wlto-type-mismatch
2709 @opindex Wlto-type-mismatch
2710 @opindex Wno-lto-type-mistmach
2712 During the link-time optimization warn about type mismatches in between
2713 global declarations from different compilation units.
2714 Requires @option{-flto} to be enabled. Enabled by default.
2716 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2718 @opindex Wno-narrowing
2719 Warn when a narrowing conversion prohibited by C++11 occurs within
2723 int i = @{ 2.2 @}; // error: narrowing from double to int
2726 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2728 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2729 required by the standard. Note that this does not affect the meaning
2730 of well-formed code; narrowing conversions are still considered
2731 ill-formed in SFINAE context.
2733 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2735 @opindex Wno-noexcept
2736 Warn when a noexcept-expression evaluates to false because of a call
2737 to a function that does not have a non-throwing exception
2738 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2739 the compiler to never throw an exception.
2741 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2742 @opindex Wnon-virtual-dtor
2743 @opindex Wno-non-virtual-dtor
2744 Warn when a class has virtual functions and an accessible non-virtual
2745 destructor itself or in an accessible polymorphic base class, in which
2746 case it is possible but unsafe to delete an instance of a derived
2747 class through a pointer to the class itself or base class. This
2748 warning is automatically enabled if @option{-Weffc++} is specified.
2750 @item -Wreorder @r{(C++ and Objective-C++ only)}
2752 @opindex Wno-reorder
2753 @cindex reordering, warning
2754 @cindex warning for reordering of member initializers
2755 Warn when the order of member initializers given in the code does not
2756 match the order in which they must be executed. For instance:
2762 A(): j (0), i (1) @{ @}
2767 The compiler rearranges the member initializers for @code{i}
2768 and @code{j} to match the declaration order of the members, emitting
2769 a warning to that effect. This warning is enabled by @option{-Wall}.
2771 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2772 @opindex fext-numeric-literals
2773 @opindex fno-ext-numeric-literals
2774 Accept imaginary, fixed-point, or machine-defined
2775 literal number suffixes as GNU extensions.
2776 When this option is turned off these suffixes are treated
2777 as C++11 user-defined literal numeric suffixes.
2778 This is on by default for all pre-C++11 dialects and all GNU dialects:
2779 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2780 @option{-std=gnu++14}.
2781 This option is off by default
2782 for ISO C++11 onwards (@option{-std=c++11}, ...).
2785 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2788 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2791 Warn about violations of the following style guidelines from Scott Meyers'
2792 @cite{Effective C++} series of books:
2796 Define a copy constructor and an assignment operator for classes
2797 with dynamically-allocated memory.
2800 Prefer initialization to assignment in constructors.
2803 Have @code{operator=} return a reference to @code{*this}.
2806 Don't try to return a reference when you must return an object.
2809 Distinguish between prefix and postfix forms of increment and
2810 decrement operators.
2813 Never overload @code{&&}, @code{||}, or @code{,}.
2817 This option also enables @option{-Wnon-virtual-dtor}, which is also
2818 one of the effective C++ recommendations. However, the check is
2819 extended to warn about the lack of virtual destructor in accessible
2820 non-polymorphic bases classes too.
2822 When selecting this option, be aware that the standard library
2823 headers do not obey all of these guidelines; use @samp{grep -v}
2824 to filter out those warnings.
2826 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2827 @opindex Wstrict-null-sentinel
2828 @opindex Wno-strict-null-sentinel
2829 Warn about the use of an uncasted @code{NULL} as sentinel. When
2830 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2831 to @code{__null}. Although it is a null pointer constant rather than a
2832 null pointer, it is guaranteed to be of the same size as a pointer.
2833 But this use is not portable across different compilers.
2835 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2836 @opindex Wno-non-template-friend
2837 @opindex Wnon-template-friend
2838 Disable warnings when non-templatized friend functions are declared
2839 within a template. Since the advent of explicit template specification
2840 support in G++, if the name of the friend is an unqualified-id (i.e.,
2841 @samp{friend foo(int)}), the C++ language specification demands that the
2842 friend declare or define an ordinary, nontemplate function. (Section
2843 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2844 could be interpreted as a particular specialization of a templatized
2845 function. Because this non-conforming behavior is no longer the default
2846 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2847 check existing code for potential trouble spots and is on by default.
2848 This new compiler behavior can be turned off with
2849 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2850 but disables the helpful warning.
2852 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2853 @opindex Wold-style-cast
2854 @opindex Wno-old-style-cast
2855 Warn if an old-style (C-style) cast to a non-void type is used within
2856 a C++ program. The new-style casts (@code{dynamic_cast},
2857 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2858 less vulnerable to unintended effects and much easier to search for.
2860 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2861 @opindex Woverloaded-virtual
2862 @opindex Wno-overloaded-virtual
2863 @cindex overloaded virtual function, warning
2864 @cindex warning for overloaded virtual function
2865 Warn when a function declaration hides virtual functions from a
2866 base class. For example, in:
2873 struct B: public A @{
2878 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2889 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2890 @opindex Wno-pmf-conversions
2891 @opindex Wpmf-conversions
2892 Disable the diagnostic for converting a bound pointer to member function
2895 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2896 @opindex Wsign-promo
2897 @opindex Wno-sign-promo
2898 Warn when overload resolution chooses a promotion from unsigned or
2899 enumerated type to a signed type, over a conversion to an unsigned type of
2900 the same size. Previous versions of G++ tried to preserve
2901 unsignedness, but the standard mandates the current behavior.
2903 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2905 @opindex Wno-terminate
2906 Disable the warning about a throw-expression that will immediately
2907 result in a call to @code{terminate}.
2910 @node Objective-C and Objective-C++ Dialect Options
2911 @section Options Controlling Objective-C and Objective-C++ Dialects
2913 @cindex compiler options, Objective-C and Objective-C++
2914 @cindex Objective-C and Objective-C++ options, command-line
2915 @cindex options, Objective-C and Objective-C++
2916 (NOTE: This manual does not describe the Objective-C and Objective-C++
2917 languages themselves. @xref{Standards,,Language Standards
2918 Supported by GCC}, for references.)
2920 This section describes the command-line options that are only meaningful
2921 for Objective-C and Objective-C++ programs. You can also use most of
2922 the language-independent GNU compiler options.
2923 For example, you might compile a file @file{some_class.m} like this:
2926 gcc -g -fgnu-runtime -O -c some_class.m
2930 In this example, @option{-fgnu-runtime} is an option meant only for
2931 Objective-C and Objective-C++ programs; you can use the other options with
2932 any language supported by GCC@.
2934 Note that since Objective-C is an extension of the C language, Objective-C
2935 compilations may also use options specific to the C front-end (e.g.,
2936 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
2937 C++-specific options (e.g., @option{-Wabi}).
2939 Here is a list of options that are @emph{only} for compiling Objective-C
2940 and Objective-C++ programs:
2943 @item -fconstant-string-class=@var{class-name}
2944 @opindex fconstant-string-class
2945 Use @var{class-name} as the name of the class to instantiate for each
2946 literal string specified with the syntax @code{@@"@dots{}"}. The default
2947 class name is @code{NXConstantString} if the GNU runtime is being used, and
2948 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2949 @option{-fconstant-cfstrings} option, if also present, overrides the
2950 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2951 to be laid out as constant CoreFoundation strings.
2954 @opindex fgnu-runtime
2955 Generate object code compatible with the standard GNU Objective-C
2956 runtime. This is the default for most types of systems.
2958 @item -fnext-runtime
2959 @opindex fnext-runtime
2960 Generate output compatible with the NeXT runtime. This is the default
2961 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2962 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2965 @item -fno-nil-receivers
2966 @opindex fno-nil-receivers
2967 Assume that all Objective-C message dispatches (@code{[receiver
2968 message:arg]}) in this translation unit ensure that the receiver is
2969 not @code{nil}. This allows for more efficient entry points in the
2970 runtime to be used. This option is only available in conjunction with
2971 the NeXT runtime and ABI version 0 or 1.
2973 @item -fobjc-abi-version=@var{n}
2974 @opindex fobjc-abi-version
2975 Use version @var{n} of the Objective-C ABI for the selected runtime.
2976 This option is currently supported only for the NeXT runtime. In that
2977 case, Version 0 is the traditional (32-bit) ABI without support for
2978 properties and other Objective-C 2.0 additions. Version 1 is the
2979 traditional (32-bit) ABI with support for properties and other
2980 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
2981 nothing is specified, the default is Version 0 on 32-bit target
2982 machines, and Version 2 on 64-bit target machines.
2984 @item -fobjc-call-cxx-cdtors
2985 @opindex fobjc-call-cxx-cdtors
2986 For each Objective-C class, check if any of its instance variables is a
2987 C++ object with a non-trivial default constructor. If so, synthesize a
2988 special @code{- (id) .cxx_construct} instance method which runs
2989 non-trivial default constructors on any such instance variables, in order,
2990 and then return @code{self}. Similarly, check if any instance variable
2991 is a C++ object with a non-trivial destructor, and if so, synthesize a
2992 special @code{- (void) .cxx_destruct} method which runs
2993 all such default destructors, in reverse order.
2995 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
2996 methods thusly generated only operate on instance variables
2997 declared in the current Objective-C class, and not those inherited
2998 from superclasses. It is the responsibility of the Objective-C
2999 runtime to invoke all such methods in an object's inheritance
3000 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3001 by the runtime immediately after a new object instance is allocated;
3002 the @code{- (void) .cxx_destruct} methods are invoked immediately
3003 before the runtime deallocates an object instance.
3005 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3006 support for invoking the @code{- (id) .cxx_construct} and
3007 @code{- (void) .cxx_destruct} methods.
3009 @item -fobjc-direct-dispatch
3010 @opindex fobjc-direct-dispatch
3011 Allow fast jumps to the message dispatcher. On Darwin this is
3012 accomplished via the comm page.
3014 @item -fobjc-exceptions
3015 @opindex fobjc-exceptions
3016 Enable syntactic support for structured exception handling in
3017 Objective-C, similar to what is offered by C++ and Java. This option
3018 is required to use the Objective-C keywords @code{@@try},
3019 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3020 @code{@@synchronized}. This option is available with both the GNU
3021 runtime and the NeXT runtime (but not available in conjunction with
3022 the NeXT runtime on Mac OS X 10.2 and earlier).
3026 Enable garbage collection (GC) in Objective-C and Objective-C++
3027 programs. This option is only available with the NeXT runtime; the
3028 GNU runtime has a different garbage collection implementation that
3029 does not require special compiler flags.
3031 @item -fobjc-nilcheck
3032 @opindex fobjc-nilcheck
3033 For the NeXT runtime with version 2 of the ABI, check for a nil
3034 receiver in method invocations before doing the actual method call.
3035 This is the default and can be disabled using
3036 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3037 checked for nil in this way no matter what this flag is set to.
3038 Currently this flag does nothing when the GNU runtime, or an older
3039 version of the NeXT runtime ABI, is used.
3041 @item -fobjc-std=objc1
3043 Conform to the language syntax of Objective-C 1.0, the language
3044 recognized by GCC 4.0. This only affects the Objective-C additions to
3045 the C/C++ language; it does not affect conformance to C/C++ standards,
3046 which is controlled by the separate C/C++ dialect option flags. When
3047 this option is used with the Objective-C or Objective-C++ compiler,
3048 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3049 This is useful if you need to make sure that your Objective-C code can
3050 be compiled with older versions of GCC@.
3052 @item -freplace-objc-classes
3053 @opindex freplace-objc-classes
3054 Emit a special marker instructing @command{ld(1)} not to statically link in
3055 the resulting object file, and allow @command{dyld(1)} to load it in at
3056 run time instead. This is used in conjunction with the Fix-and-Continue
3057 debugging mode, where the object file in question may be recompiled and
3058 dynamically reloaded in the course of program execution, without the need
3059 to restart the program itself. Currently, Fix-and-Continue functionality
3060 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3065 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3066 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3067 compile time) with static class references that get initialized at load time,
3068 which improves run-time performance. Specifying the @option{-fzero-link} flag
3069 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3070 to be retained. This is useful in Zero-Link debugging mode, since it allows
3071 for individual class implementations to be modified during program execution.
3072 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3073 regardless of command-line options.
3075 @item -fno-local-ivars
3076 @opindex fno-local-ivars
3077 @opindex flocal-ivars
3078 By default instance variables in Objective-C can be accessed as if
3079 they were local variables from within the methods of the class they're
3080 declared in. This can lead to shadowing between instance variables
3081 and other variables declared either locally inside a class method or
3082 globally with the same name. Specifying the @option{-fno-local-ivars}
3083 flag disables this behavior thus avoiding variable shadowing issues.
3085 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3086 @opindex fivar-visibility
3087 Set the default instance variable visibility to the specified option
3088 so that instance variables declared outside the scope of any access
3089 modifier directives default to the specified visibility.
3093 Dump interface declarations for all classes seen in the source file to a
3094 file named @file{@var{sourcename}.decl}.
3096 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3097 @opindex Wassign-intercept
3098 @opindex Wno-assign-intercept
3099 Warn whenever an Objective-C assignment is being intercepted by the
3102 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3103 @opindex Wno-protocol
3105 If a class is declared to implement a protocol, a warning is issued for
3106 every method in the protocol that is not implemented by the class. The
3107 default behavior is to issue a warning for every method not explicitly
3108 implemented in the class, even if a method implementation is inherited
3109 from the superclass. If you use the @option{-Wno-protocol} option, then
3110 methods inherited from the superclass are considered to be implemented,
3111 and no warning is issued for them.
3113 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3115 @opindex Wno-selector
3116 Warn if multiple methods of different types for the same selector are
3117 found during compilation. The check is performed on the list of methods
3118 in the final stage of compilation. Additionally, a check is performed
3119 for each selector appearing in a @code{@@selector(@dots{})}
3120 expression, and a corresponding method for that selector has been found
3121 during compilation. Because these checks scan the method table only at
3122 the end of compilation, these warnings are not produced if the final
3123 stage of compilation is not reached, for example because an error is
3124 found during compilation, or because the @option{-fsyntax-only} option is
3127 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3128 @opindex Wstrict-selector-match
3129 @opindex Wno-strict-selector-match
3130 Warn if multiple methods with differing argument and/or return types are
3131 found for a given selector when attempting to send a message using this
3132 selector to a receiver of type @code{id} or @code{Class}. When this flag
3133 is off (which is the default behavior), the compiler omits such warnings
3134 if any differences found are confined to types that share the same size
3137 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3138 @opindex Wundeclared-selector
3139 @opindex Wno-undeclared-selector
3140 Warn if a @code{@@selector(@dots{})} expression referring to an
3141 undeclared selector is found. A selector is considered undeclared if no
3142 method with that name has been declared before the
3143 @code{@@selector(@dots{})} expression, either explicitly in an
3144 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3145 an @code{@@implementation} section. This option always performs its
3146 checks as soon as a @code{@@selector(@dots{})} expression is found,
3147 while @option{-Wselector} only performs its checks in the final stage of
3148 compilation. This also enforces the coding style convention
3149 that methods and selectors must be declared before being used.
3151 @item -print-objc-runtime-info
3152 @opindex print-objc-runtime-info
3153 Generate C header describing the largest structure that is passed by
3158 @node Diagnostic Message Formatting Options
3159 @section Options to Control Diagnostic Messages Formatting
3160 @cindex options to control diagnostics formatting
3161 @cindex diagnostic messages
3162 @cindex message formatting
3164 Traditionally, diagnostic messages have been formatted irrespective of
3165 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3166 options described below
3167 to control the formatting algorithm for diagnostic messages,
3168 e.g.@: how many characters per line, how often source location
3169 information should be reported. Note that some language front ends may not
3170 honor these options.
3173 @item -fmessage-length=@var{n}
3174 @opindex fmessage-length
3175 Try to format error messages so that they fit on lines of about
3176 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3177 done; each error message appears on a single line. This is the
3178 default for all front ends.
3180 @item -fdiagnostics-show-location=once
3181 @opindex fdiagnostics-show-location
3182 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3183 reporter to emit source location information @emph{once}; that is, in
3184 case the message is too long to fit on a single physical line and has to
3185 be wrapped, the source location won't be emitted (as prefix) again,
3186 over and over, in subsequent continuation lines. This is the default
3189 @item -fdiagnostics-show-location=every-line
3190 Only meaningful in line-wrapping mode. Instructs the diagnostic
3191 messages reporter to emit the same source location information (as
3192 prefix) for physical lines that result from the process of breaking
3193 a message which is too long to fit on a single line.
3195 @item -fdiagnostics-color[=@var{WHEN}]
3196 @itemx -fno-diagnostics-color
3197 @opindex fdiagnostics-color
3198 @cindex highlight, color, colour
3199 @vindex GCC_COLORS @r{environment variable}
3200 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3201 or @samp{auto}. The default depends on how the compiler has been configured,
3202 it can be any of the above @var{WHEN} options or also @samp{never}
3203 if @env{GCC_COLORS} environment variable isn't present in the environment,
3204 and @samp{auto} otherwise.
3205 @samp{auto} means to use color only when the standard error is a terminal.
3206 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3207 aliases for @option{-fdiagnostics-color=always} and
3208 @option{-fdiagnostics-color=never}, respectively.
3210 The colors are defined by the environment variable @env{GCC_COLORS}.
3211 Its value is a colon-separated list of capabilities and Select Graphic
3212 Rendition (SGR) substrings. SGR commands are interpreted by the
3213 terminal or terminal emulator. (See the section in the documentation
3214 of your text terminal for permitted values and their meanings as
3215 character attributes.) These substring values are integers in decimal
3216 representation and can be concatenated with semicolons.
3217 Common values to concatenate include
3219 @samp{4} for underline,
3221 @samp{7} for inverse,
3222 @samp{39} for default foreground color,
3223 @samp{30} to @samp{37} for foreground colors,
3224 @samp{90} to @samp{97} for 16-color mode foreground colors,
3225 @samp{38;5;0} to @samp{38;5;255}
3226 for 88-color and 256-color modes foreground colors,
3227 @samp{49} for default background color,
3228 @samp{40} to @samp{47} for background colors,
3229 @samp{100} to @samp{107} for 16-color mode background colors,
3230 and @samp{48;5;0} to @samp{48;5;255}
3231 for 88-color and 256-color modes background colors.
3233 The default @env{GCC_COLORS} is
3235 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3238 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3239 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3240 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3241 string disables colors.
3242 Supported capabilities are as follows.
3246 @vindex error GCC_COLORS @r{capability}
3247 SGR substring for error: markers.
3250 @vindex warning GCC_COLORS @r{capability}
3251 SGR substring for warning: markers.
3254 @vindex note GCC_COLORS @r{capability}
3255 SGR substring for note: markers.
3258 @vindex caret GCC_COLORS @r{capability}
3259 SGR substring for caret line.
3262 @vindex locus GCC_COLORS @r{capability}
3263 SGR substring for location information, @samp{file:line} or
3264 @samp{file:line:column} etc.
3267 @vindex quote GCC_COLORS @r{capability}
3268 SGR substring for information printed within quotes.
3271 @item -fno-diagnostics-show-option
3272 @opindex fno-diagnostics-show-option
3273 @opindex fdiagnostics-show-option
3274 By default, each diagnostic emitted includes text indicating the
3275 command-line option that directly controls the diagnostic (if such an
3276 option is known to the diagnostic machinery). Specifying the
3277 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3279 @item -fno-diagnostics-show-caret
3280 @opindex fno-diagnostics-show-caret
3281 @opindex fdiagnostics-show-caret
3282 By default, each diagnostic emitted includes the original source line
3283 and a caret '^' indicating the column. This option suppresses this
3284 information. The source line is truncated to @var{n} characters, if
3285 the @option{-fmessage-length=n} option is given. When the output is done
3286 to the terminal, the width is limited to the width given by the
3287 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3291 @node Warning Options
3292 @section Options to Request or Suppress Warnings
3293 @cindex options to control warnings
3294 @cindex warning messages
3295 @cindex messages, warning
3296 @cindex suppressing warnings
3298 Warnings are diagnostic messages that report constructions that
3299 are not inherently erroneous but that are risky or suggest there
3300 may have been an error.
3302 The following language-independent options do not enable specific
3303 warnings but control the kinds of diagnostics produced by GCC@.
3306 @cindex syntax checking
3308 @opindex fsyntax-only
3309 Check the code for syntax errors, but don't do anything beyond that.
3311 @item -fmax-errors=@var{n}
3312 @opindex fmax-errors
3313 Limits the maximum number of error messages to @var{n}, at which point
3314 GCC bails out rather than attempting to continue processing the source
3315 code. If @var{n} is 0 (the default), there is no limit on the number
3316 of error messages produced. If @option{-Wfatal-errors} is also
3317 specified, then @option{-Wfatal-errors} takes precedence over this
3322 Inhibit all warning messages.
3327 Make all warnings into errors.
3332 Make the specified warning into an error. The specifier for a warning
3333 is appended; for example @option{-Werror=switch} turns the warnings
3334 controlled by @option{-Wswitch} into errors. This switch takes a
3335 negative form, to be used to negate @option{-Werror} for specific
3336 warnings; for example @option{-Wno-error=switch} makes
3337 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3340 The warning message for each controllable warning includes the
3341 option that controls the warning. That option can then be used with
3342 @option{-Werror=} and @option{-Wno-error=} as described above.
3343 (Printing of the option in the warning message can be disabled using the
3344 @option{-fno-diagnostics-show-option} flag.)
3346 Note that specifying @option{-Werror=}@var{foo} automatically implies
3347 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3350 @item -Wfatal-errors
3351 @opindex Wfatal-errors
3352 @opindex Wno-fatal-errors
3353 This option causes the compiler to abort compilation on the first error
3354 occurred rather than trying to keep going and printing further error
3359 You can request many specific warnings with options beginning with
3360 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3361 implicit declarations. Each of these specific warning options also
3362 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3363 example, @option{-Wno-implicit}. This manual lists only one of the
3364 two forms, whichever is not the default. For further
3365 language-specific options also refer to @ref{C++ Dialect Options} and
3366 @ref{Objective-C and Objective-C++ Dialect Options}.
3368 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3369 options, such as @option{-Wunused}, which may turn on further options,
3370 such as @option{-Wunused-value}. The combined effect of positive and
3371 negative forms is that more specific options have priority over less
3372 specific ones, independently of their position in the command-line. For
3373 options of the same specificity, the last one takes effect. Options
3374 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3375 as if they appeared at the end of the command-line.
3377 When an unrecognized warning option is requested (e.g.,
3378 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3379 that the option is not recognized. However, if the @option{-Wno-} form
3380 is used, the behavior is slightly different: no diagnostic is
3381 produced for @option{-Wno-unknown-warning} unless other diagnostics
3382 are being produced. This allows the use of new @option{-Wno-} options
3383 with old compilers, but if something goes wrong, the compiler
3384 warns that an unrecognized option is present.
3391 Issue all the warnings demanded by strict ISO C and ISO C++;
3392 reject all programs that use forbidden extensions, and some other
3393 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3394 version of the ISO C standard specified by any @option{-std} option used.
3396 Valid ISO C and ISO C++ programs should compile properly with or without
3397 this option (though a rare few require @option{-ansi} or a
3398 @option{-std} option specifying the required version of ISO C)@. However,
3399 without this option, certain GNU extensions and traditional C and C++
3400 features are supported as well. With this option, they are rejected.
3402 @option{-Wpedantic} does not cause warning messages for use of the
3403 alternate keywords whose names begin and end with @samp{__}. Pedantic
3404 warnings are also disabled in the expression that follows
3405 @code{__extension__}. However, only system header files should use
3406 these escape routes; application programs should avoid them.
3407 @xref{Alternate Keywords}.
3409 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3410 C conformance. They soon find that it does not do quite what they want:
3411 it finds some non-ISO practices, but not all---only those for which
3412 ISO C @emph{requires} a diagnostic, and some others for which
3413 diagnostics have been added.
3415 A feature to report any failure to conform to ISO C might be useful in
3416 some instances, but would require considerable additional work and would
3417 be quite different from @option{-Wpedantic}. We don't have plans to
3418 support such a feature in the near future.
3420 Where the standard specified with @option{-std} represents a GNU
3421 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3422 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3423 extended dialect is based. Warnings from @option{-Wpedantic} are given
3424 where they are required by the base standard. (It does not make sense
3425 for such warnings to be given only for features not in the specified GNU
3426 C dialect, since by definition the GNU dialects of C include all
3427 features the compiler supports with the given option, and there would be
3428 nothing to warn about.)
3430 @item -pedantic-errors
3431 @opindex pedantic-errors
3432 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3433 requires a diagnostic, in some cases where there is undefined behavior
3434 at compile-time and in some other cases that do not prevent compilation
3435 of programs that are valid according to the standard. This is not
3436 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3437 by this option and not enabled by the latter and vice versa.
3442 This enables all the warnings about constructions that some users
3443 consider questionable, and that are easy to avoid (or modify to
3444 prevent the warning), even in conjunction with macros. This also
3445 enables some language-specific warnings described in @ref{C++ Dialect
3446 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3448 @option{-Wall} turns on the following warning flags:
3450 @gccoptlist{-Waddress @gol
3451 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3452 -Wc++11-compat -Wc++14-compat@gol
3453 -Wchar-subscripts @gol
3454 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3455 -Wimplicit-int @r{(C and Objective-C only)} @gol
3456 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3460 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3461 -Wmaybe-uninitialized @gol
3462 -Wmissing-braces @r{(only for C/ObjC)} @gol
3469 -Wsequence-point @gol
3470 -Wsign-compare @r{(only in C++)} @gol
3471 -Wstrict-aliasing @gol
3472 -Wstrict-overflow=1 @gol
3474 -Wtautological-compare @gol
3476 -Wuninitialized @gol
3477 -Wunknown-pragmas @gol
3478 -Wunused-function @gol
3481 -Wunused-variable @gol
3482 -Wvolatile-register-var @gol
3485 Note that some warning flags are not implied by @option{-Wall}. Some of
3486 them warn about constructions that users generally do not consider
3487 questionable, but which occasionally you might wish to check for;
3488 others warn about constructions that are necessary or hard to avoid in
3489 some cases, and there is no simple way to modify the code to suppress
3490 the warning. Some of them are enabled by @option{-Wextra} but many of
3491 them must be enabled individually.
3497 This enables some extra warning flags that are not enabled by
3498 @option{-Wall}. (This option used to be called @option{-W}. The older
3499 name is still supported, but the newer name is more descriptive.)
3501 @gccoptlist{-Wclobbered @gol
3503 -Wignored-qualifiers @gol
3504 -Wmissing-field-initializers @gol
3505 -Wmissing-parameter-type @r{(C only)} @gol
3506 -Wold-style-declaration @r{(C only)} @gol
3507 -Woverride-init @gol
3510 -Wuninitialized @gol
3511 -Wshift-negative-value @gol
3512 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3513 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3516 The option @option{-Wextra} also prints warning messages for the
3522 A pointer is compared against integer zero with @code{<}, @code{<=},
3523 @code{>}, or @code{>=}.
3526 (C++ only) An enumerator and a non-enumerator both appear in a
3527 conditional expression.
3530 (C++ only) Ambiguous virtual bases.
3533 (C++ only) Subscripting an array that has been declared @code{register}.
3536 (C++ only) Taking the address of a variable that has been declared
3540 (C++ only) A base class is not initialized in a derived class's copy
3545 @item -Wchar-subscripts
3546 @opindex Wchar-subscripts
3547 @opindex Wno-char-subscripts
3548 Warn if an array subscript has type @code{char}. This is a common cause
3549 of error, as programmers often forget that this type is signed on some
3551 This warning is enabled by @option{-Wall}.
3555 @opindex Wno-comment
3556 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3557 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3558 This warning is enabled by @option{-Wall}.
3560 @item -Wno-coverage-mismatch
3561 @opindex Wno-coverage-mismatch
3562 Warn if feedback profiles do not match when using the
3563 @option{-fprofile-use} option.
3564 If a source file is changed between compiling with @option{-fprofile-gen} and
3565 with @option{-fprofile-use}, the files with the profile feedback can fail
3566 to match the source file and GCC cannot use the profile feedback
3567 information. By default, this warning is enabled and is treated as an
3568 error. @option{-Wno-coverage-mismatch} can be used to disable the
3569 warning or @option{-Wno-error=coverage-mismatch} can be used to
3570 disable the error. Disabling the error for this warning can result in
3571 poorly optimized code and is useful only in the
3572 case of very minor changes such as bug fixes to an existing code-base.
3573 Completely disabling the warning is not recommended.
3576 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3578 Suppress warning messages emitted by @code{#warning} directives.
3580 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3581 @opindex Wdouble-promotion
3582 @opindex Wno-double-promotion
3583 Give a warning when a value of type @code{float} is implicitly
3584 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3585 floating-point unit implement @code{float} in hardware, but emulate
3586 @code{double} in software. On such a machine, doing computations
3587 using @code{double} values is much more expensive because of the
3588 overhead required for software emulation.
3590 It is easy to accidentally do computations with @code{double} because
3591 floating-point literals are implicitly of type @code{double}. For
3595 float area(float radius)
3597 return 3.14159 * radius * radius;
3601 the compiler performs the entire computation with @code{double}
3602 because the floating-point literal is a @code{double}.
3605 @itemx -Wformat=@var{n}
3608 @opindex ffreestanding
3609 @opindex fno-builtin
3611 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3612 the arguments supplied have types appropriate to the format string
3613 specified, and that the conversions specified in the format string make
3614 sense. This includes standard functions, and others specified by format
3615 attributes (@pxref{Function Attributes}), in the @code{printf},
3616 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3617 not in the C standard) families (or other target-specific families).
3618 Which functions are checked without format attributes having been
3619 specified depends on the standard version selected, and such checks of
3620 functions without the attribute specified are disabled by
3621 @option{-ffreestanding} or @option{-fno-builtin}.
3623 The formats are checked against the format features supported by GNU
3624 libc version 2.2. These include all ISO C90 and C99 features, as well
3625 as features from the Single Unix Specification and some BSD and GNU
3626 extensions. Other library implementations may not support all these
3627 features; GCC does not support warning about features that go beyond a
3628 particular library's limitations. However, if @option{-Wpedantic} is used
3629 with @option{-Wformat}, warnings are given about format features not
3630 in the selected standard version (but not for @code{strfmon} formats,
3631 since those are not in any version of the C standard). @xref{C Dialect
3632 Options,,Options Controlling C Dialect}.
3639 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3640 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3641 @option{-Wformat} also checks for null format arguments for several
3642 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3643 aspects of this level of format checking can be disabled by the
3644 options: @option{-Wno-format-contains-nul},
3645 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3646 @option{-Wformat} is enabled by @option{-Wall}.
3648 @item -Wno-format-contains-nul
3649 @opindex Wno-format-contains-nul
3650 @opindex Wformat-contains-nul
3651 If @option{-Wformat} is specified, do not warn about format strings that
3654 @item -Wno-format-extra-args
3655 @opindex Wno-format-extra-args
3656 @opindex Wformat-extra-args
3657 If @option{-Wformat} is specified, do not warn about excess arguments to a
3658 @code{printf} or @code{scanf} format function. The C standard specifies
3659 that such arguments are ignored.
3661 Where the unused arguments lie between used arguments that are
3662 specified with @samp{$} operand number specifications, normally
3663 warnings are still given, since the implementation could not know what
3664 type to pass to @code{va_arg} to skip the unused arguments. However,
3665 in the case of @code{scanf} formats, this option suppresses the
3666 warning if the unused arguments are all pointers, since the Single
3667 Unix Specification says that such unused arguments are allowed.
3669 @item -Wno-format-zero-length
3670 @opindex Wno-format-zero-length
3671 @opindex Wformat-zero-length
3672 If @option{-Wformat} is specified, do not warn about zero-length formats.
3673 The C standard specifies that zero-length formats are allowed.
3678 Enable @option{-Wformat} plus additional format checks. Currently
3679 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3682 @item -Wformat-nonliteral
3683 @opindex Wformat-nonliteral
3684 @opindex Wno-format-nonliteral
3685 If @option{-Wformat} is specified, also warn if the format string is not a
3686 string literal and so cannot be checked, unless the format function
3687 takes its format arguments as a @code{va_list}.
3689 @item -Wformat-security
3690 @opindex Wformat-security
3691 @opindex Wno-format-security
3692 If @option{-Wformat} is specified, also warn about uses of format
3693 functions that represent possible security problems. At present, this
3694 warns about calls to @code{printf} and @code{scanf} functions where the
3695 format string is not a string literal and there are no format arguments,
3696 as in @code{printf (foo);}. This may be a security hole if the format
3697 string came from untrusted input and contains @samp{%n}. (This is
3698 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3699 in future warnings may be added to @option{-Wformat-security} that are not
3700 included in @option{-Wformat-nonliteral}.)
3702 @item -Wformat-signedness
3703 @opindex Wformat-signedness
3704 @opindex Wno-format-signedness
3705 If @option{-Wformat} is specified, also warn if the format string
3706 requires an unsigned argument and the argument is signed and vice versa.
3709 @opindex Wformat-y2k
3710 @opindex Wno-format-y2k
3711 If @option{-Wformat} is specified, also warn about @code{strftime}
3712 formats that may yield only a two-digit year.
3717 @opindex Wno-nonnull
3718 Warn about passing a null pointer for arguments marked as
3719 requiring a non-null value by the @code{nonnull} function attribute.
3721 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3722 can be disabled with the @option{-Wno-nonnull} option.
3724 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3726 @opindex Wno-init-self
3727 Warn about uninitialized variables that are initialized with themselves.
3728 Note this option can only be used with the @option{-Wuninitialized} option.
3730 For example, GCC warns about @code{i} being uninitialized in the
3731 following snippet only when @option{-Winit-self} has been specified:
3742 This warning is enabled by @option{-Wall} in C++.
3744 @item -Wimplicit-int @r{(C and Objective-C only)}
3745 @opindex Wimplicit-int
3746 @opindex Wno-implicit-int
3747 Warn when a declaration does not specify a type.
3748 This warning is enabled by @option{-Wall}.
3750 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3751 @opindex Wimplicit-function-declaration
3752 @opindex Wno-implicit-function-declaration
3753 Give a warning whenever a function is used before being declared. In
3754 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3755 enabled by default and it is made into an error by
3756 @option{-pedantic-errors}. This warning is also enabled by
3759 @item -Wimplicit @r{(C and Objective-C only)}
3761 @opindex Wno-implicit
3762 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3763 This warning is enabled by @option{-Wall}.
3765 @item -Wignored-qualifiers @r{(C and C++ only)}
3766 @opindex Wignored-qualifiers
3767 @opindex Wno-ignored-qualifiers
3768 Warn if the return type of a function has a type qualifier
3769 such as @code{const}. For ISO C such a type qualifier has no effect,
3770 since the value returned by a function is not an lvalue.
3771 For C++, the warning is only emitted for scalar types or @code{void}.
3772 ISO C prohibits qualified @code{void} return types on function
3773 definitions, so such return types always receive a warning
3774 even without this option.
3776 This warning is also enabled by @option{-Wextra}.
3781 Warn if the type of @code{main} is suspicious. @code{main} should be
3782 a function with external linkage, returning int, taking either zero
3783 arguments, two, or three arguments of appropriate types. This warning
3784 is enabled by default in C++ and is enabled by either @option{-Wall}
3785 or @option{-Wpedantic}.
3787 @item -Wmisleading-indentation @r{(C and C++ only)}
3788 @opindex Wmisleading-indentation
3789 @opindex Wno-misleading-indentation
3790 Warn when the indentation of the code does not reflect the block structure.
3791 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3792 @code{for} clauses with a guarded statement that does not use braces,
3793 followed by an unguarded statement with the same indentation.
3795 This warning is disabled by default.
3797 In the following example, the call to ``bar'' is misleadingly indented as
3798 if it were guarded by the ``if'' conditional.
3801 if (some_condition ())
3803 bar (); /* Gotcha: this is not guarded by the "if". */
3806 In the case of mixed tabs and spaces, the warning uses the
3807 @option{-ftabstop=} option to determine if the statements line up
3810 The warning is not issued for code involving multiline preprocessor logic
3811 such as the following example.
3816 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3822 The warning is not issued after a @code{#line} directive, since this
3823 typically indicates autogenerated code, and no assumptions can be made
3824 about the layout of the file that the directive references.
3826 @item -Wmissing-braces
3827 @opindex Wmissing-braces
3828 @opindex Wno-missing-braces
3829 Warn if an aggregate or union initializer is not fully bracketed. In
3830 the following example, the initializer for @code{a} is not fully
3831 bracketed, but that for @code{b} is fully bracketed. This warning is
3832 enabled by @option{-Wall} in C.
3835 int a[2][2] = @{ 0, 1, 2, 3 @};
3836 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3839 This warning is enabled by @option{-Wall}.
3841 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3842 @opindex Wmissing-include-dirs
3843 @opindex Wno-missing-include-dirs
3844 Warn if a user-supplied include directory does not exist.
3847 @opindex Wparentheses
3848 @opindex Wno-parentheses
3849 Warn if parentheses are omitted in certain contexts, such
3850 as when there is an assignment in a context where a truth value
3851 is expected, or when operators are nested whose precedence people
3852 often get confused about.
3854 Also warn if a comparison like @code{x<=y<=z} appears; this is
3855 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3856 interpretation from that of ordinary mathematical notation.
3858 Also warn about constructions where there may be confusion to which
3859 @code{if} statement an @code{else} branch belongs. Here is an example of
3874 In C/C++, every @code{else} branch belongs to the innermost possible
3875 @code{if} statement, which in this example is @code{if (b)}. This is
3876 often not what the programmer expected, as illustrated in the above
3877 example by indentation the programmer chose. When there is the
3878 potential for this confusion, GCC issues a warning when this flag
3879 is specified. To eliminate the warning, add explicit braces around
3880 the innermost @code{if} statement so there is no way the @code{else}
3881 can belong to the enclosing @code{if}. The resulting code
3898 Also warn for dangerous uses of the GNU extension to
3899 @code{?:} with omitted middle operand. When the condition
3900 in the @code{?}: operator is a boolean expression, the omitted value is
3901 always 1. Often programmers expect it to be a value computed
3902 inside the conditional expression instead.
3904 This warning is enabled by @option{-Wall}.
3906 @item -Wsequence-point
3907 @opindex Wsequence-point
3908 @opindex Wno-sequence-point
3909 Warn about code that may have undefined semantics because of violations
3910 of sequence point rules in the C and C++ standards.
3912 The C and C++ standards define the order in which expressions in a C/C++
3913 program are evaluated in terms of @dfn{sequence points}, which represent
3914 a partial ordering between the execution of parts of the program: those
3915 executed before the sequence point, and those executed after it. These
3916 occur after the evaluation of a full expression (one which is not part
3917 of a larger expression), after the evaluation of the first operand of a
3918 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
3919 function is called (but after the evaluation of its arguments and the
3920 expression denoting the called function), and in certain other places.
3921 Other than as expressed by the sequence point rules, the order of
3922 evaluation of subexpressions of an expression is not specified. All
3923 these rules describe only a partial order rather than a total order,
3924 since, for example, if two functions are called within one expression
3925 with no sequence point between them, the order in which the functions
3926 are called is not specified. However, the standards committee have
3927 ruled that function calls do not overlap.
3929 It is not specified when between sequence points modifications to the
3930 values of objects take effect. Programs whose behavior depends on this
3931 have undefined behavior; the C and C++ standards specify that ``Between
3932 the previous and next sequence point an object shall have its stored
3933 value modified at most once by the evaluation of an expression.
3934 Furthermore, the prior value shall be read only to determine the value
3935 to be stored.''. If a program breaks these rules, the results on any
3936 particular implementation are entirely unpredictable.
3938 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
3939 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
3940 diagnosed by this option, and it may give an occasional false positive
3941 result, but in general it has been found fairly effective at detecting
3942 this sort of problem in programs.
3944 The standard is worded confusingly, therefore there is some debate
3945 over the precise meaning of the sequence point rules in subtle cases.
3946 Links to discussions of the problem, including proposed formal
3947 definitions, may be found on the GCC readings page, at
3948 @uref{http://gcc.gnu.org/@/readings.html}.
3950 This warning is enabled by @option{-Wall} for C and C++.
3952 @item -Wno-return-local-addr
3953 @opindex Wno-return-local-addr
3954 @opindex Wreturn-local-addr
3955 Do not warn about returning a pointer (or in C++, a reference) to a
3956 variable that goes out of scope after the function returns.
3959 @opindex Wreturn-type
3960 @opindex Wno-return-type
3961 Warn whenever a function is defined with a return type that defaults
3962 to @code{int}. Also warn about any @code{return} statement with no
3963 return value in a function whose return type is not @code{void}
3964 (falling off the end of the function body is considered returning
3965 without a value), and about a @code{return} statement with an
3966 expression in a function whose return type is @code{void}.
3968 For C++, a function without return type always produces a diagnostic
3969 message, even when @option{-Wno-return-type} is specified. The only
3970 exceptions are @code{main} and functions defined in system headers.
3972 This warning is enabled by @option{-Wall}.
3974 @item -Wshift-count-negative
3975 @opindex Wshift-count-negative
3976 @opindex Wno-shift-count-negative
3977 Warn if shift count is negative. This warning is enabled by default.
3979 @item -Wshift-count-overflow
3980 @opindex Wshift-count-overflow
3981 @opindex Wno-shift-count-overflow
3982 Warn if shift count >= width of type. This warning is enabled by default.
3984 @item -Wshift-negative-value
3985 @opindex Wshift-negative-value
3986 @opindex Wno-shift-negative-value
3987 Warn if left shifting a negative value. This warning is enabled by
3988 @option{-Wextra} in C99 and C++11 modes (and newer).
3990 @item -Wshift-overflow
3991 @itemx -Wshift-overflow=@var{n}
3992 @opindex Wshift-overflow
3993 @opindex Wno-shift-overflow
3994 Warn about left shift overflows. This warning is enabled by
3995 default in C99 and C++11 modes (and newer).
3998 @item -Wshift-overflow=1
3999 This is the warning level of @option{-Wshift-overflow} and is enabled
4000 by default in C99 and C++11 modes (and newer). This warning level does
4001 not warn about left-shifting 1 into the sign bit. (However, in C, such
4002 an overflow is still rejected in contexts where an integer constant expression
4005 @item -Wshift-overflow=2
4006 This warning level also warns about left-shifting 1 into the sign bit,
4007 unless C++14 mode is active.
4013 Warn whenever a @code{switch} statement has an index of enumerated type
4014 and lacks a @code{case} for one or more of the named codes of that
4015 enumeration. (The presence of a @code{default} label prevents this
4016 warning.) @code{case} labels outside the enumeration range also
4017 provoke warnings when this option is used (even if there is a
4018 @code{default} label).
4019 This warning is enabled by @option{-Wall}.
4021 @item -Wswitch-default
4022 @opindex Wswitch-default
4023 @opindex Wno-switch-default
4024 Warn whenever a @code{switch} statement does not have a @code{default}
4028 @opindex Wswitch-enum
4029 @opindex Wno-switch-enum
4030 Warn whenever a @code{switch} statement has an index of enumerated type
4031 and lacks a @code{case} for one or more of the named codes of that
4032 enumeration. @code{case} labels outside the enumeration range also
4033 provoke warnings when this option is used. The only difference
4034 between @option{-Wswitch} and this option is that this option gives a
4035 warning about an omitted enumeration code even if there is a
4036 @code{default} label.
4039 @opindex Wswitch-bool
4040 @opindex Wno-switch-bool
4041 Warn whenever a @code{switch} statement has an index of boolean type
4042 and the case values are outside the range of a boolean type.
4043 It is possible to suppress this warning by casting the controlling
4044 expression to a type other than @code{bool}. For example:
4047 switch ((int) (a == 4))
4053 This warning is enabled by default for C and C++ programs.
4055 @item -Wsync-nand @r{(C and C++ only)}
4057 @opindex Wno-sync-nand
4058 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4059 built-in functions are used. These functions changed semantics in GCC 4.4.
4063 @opindex Wno-trigraphs
4064 Warn if any trigraphs are encountered that might change the meaning of
4065 the program (trigraphs within comments are not warned about).
4066 This warning is enabled by @option{-Wall}.
4068 @item -Wunused-but-set-parameter
4069 @opindex Wunused-but-set-parameter
4070 @opindex Wno-unused-but-set-parameter
4071 Warn whenever a function parameter is assigned to, but otherwise unused
4072 (aside from its declaration).
4074 To suppress this warning use the @code{unused} attribute
4075 (@pxref{Variable Attributes}).
4077 This warning is also enabled by @option{-Wunused} together with
4080 @item -Wunused-but-set-variable
4081 @opindex Wunused-but-set-variable
4082 @opindex Wno-unused-but-set-variable
4083 Warn whenever a local variable is assigned to, but otherwise unused
4084 (aside from its declaration).
4085 This warning is enabled by @option{-Wall}.
4087 To suppress this warning use the @code{unused} attribute
4088 (@pxref{Variable Attributes}).
4090 This warning is also enabled by @option{-Wunused}, which is enabled
4093 @item -Wunused-function
4094 @opindex Wunused-function
4095 @opindex Wno-unused-function
4096 Warn whenever a static function is declared but not defined or a
4097 non-inline static function is unused.
4098 This warning is enabled by @option{-Wall}.
4100 @item -Wunused-label
4101 @opindex Wunused-label
4102 @opindex Wno-unused-label
4103 Warn whenever a label is declared but not used.
4104 This warning is enabled by @option{-Wall}.
4106 To suppress this warning use the @code{unused} attribute
4107 (@pxref{Variable Attributes}).
4109 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4110 @opindex Wunused-local-typedefs
4111 Warn when a typedef locally defined in a function is not used.
4112 This warning is enabled by @option{-Wall}.
4114 @item -Wunused-parameter
4115 @opindex Wunused-parameter
4116 @opindex Wno-unused-parameter
4117 Warn whenever a function parameter is unused aside from its declaration.
4119 To suppress this warning use the @code{unused} attribute
4120 (@pxref{Variable Attributes}).
4122 @item -Wno-unused-result
4123 @opindex Wunused-result
4124 @opindex Wno-unused-result
4125 Do not warn if a caller of a function marked with attribute
4126 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4127 its return value. The default is @option{-Wunused-result}.
4129 @item -Wunused-variable
4130 @opindex Wunused-variable
4131 @opindex Wno-unused-variable
4132 Warn whenever a local variable or non-constant static variable is unused
4133 aside from its declaration.
4134 This warning is enabled by @option{-Wall}.
4136 To suppress this warning use the @code{unused} attribute
4137 (@pxref{Variable Attributes}).
4139 @item -Wunused-value
4140 @opindex Wunused-value
4141 @opindex Wno-unused-value
4142 Warn whenever a statement computes a result that is explicitly not
4143 used. To suppress this warning cast the unused expression to
4144 @code{void}. This includes an expression-statement or the left-hand
4145 side of a comma expression that contains no side effects. For example,
4146 an expression such as @code{x[i,j]} causes a warning, while
4147 @code{x[(void)i,j]} does not.
4149 This warning is enabled by @option{-Wall}.
4154 All the above @option{-Wunused} options combined.
4156 In order to get a warning about an unused function parameter, you must
4157 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4158 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4160 @item -Wnull-dereference
4161 @opindex Wnull-dereference
4162 @opindex Wno-null-dereference
4163 Warn if the compiler detects paths that trigger erroneous or
4164 undefined behavior due to dereferencing a null pointer. This option
4165 is only active when @option{-fdelete-null-pointer-checks} is active,
4166 which is enabled by optimizations in most targets. The precision of
4167 the warnings depends on the optimization options used. This option is
4168 enabled by @option{-Wall}.
4170 @item -Wuninitialized
4171 @opindex Wuninitialized
4172 @opindex Wno-uninitialized
4173 Warn if an automatic variable is used without first being initialized
4174 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4175 warn if a non-static reference or non-static @code{const} member
4176 appears in a class without constructors.
4178 If you want to warn about code that uses the uninitialized value of the
4179 variable in its own initializer, use the @option{-Winit-self} option.
4181 These warnings occur for individual uninitialized or clobbered
4182 elements of structure, union or array variables as well as for
4183 variables that are uninitialized or clobbered as a whole. They do
4184 not occur for variables or elements declared @code{volatile}. Because
4185 these warnings depend on optimization, the exact variables or elements
4186 for which there are warnings depends on the precise optimization
4187 options and version of GCC used.
4189 Note that there may be no warning about a variable that is used only
4190 to compute a value that itself is never used, because such
4191 computations may be deleted by data flow analysis before the warnings
4194 @item -Wmaybe-uninitialized
4195 @opindex Wmaybe-uninitialized
4196 @opindex Wno-maybe-uninitialized
4197 For an automatic variable, if there exists a path from the function
4198 entry to a use of the variable that is initialized, but there exist
4199 some other paths for which the variable is not initialized, the compiler
4200 emits a warning if it cannot prove the uninitialized paths are not
4201 executed at run time. These warnings are made optional because GCC is
4202 not smart enough to see all the reasons why the code might be correct
4203 in spite of appearing to have an error. Here is one example of how
4224 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4225 always initialized, but GCC doesn't know this. To suppress the
4226 warning, you need to provide a default case with assert(0) or
4229 @cindex @code{longjmp} warnings
4230 This option also warns when a non-volatile automatic variable might be
4231 changed by a call to @code{longjmp}. These warnings as well are possible
4232 only in optimizing compilation.
4234 The compiler sees only the calls to @code{setjmp}. It cannot know
4235 where @code{longjmp} will be called; in fact, a signal handler could
4236 call it at any point in the code. As a result, you may get a warning
4237 even when there is in fact no problem because @code{longjmp} cannot
4238 in fact be called at the place that would cause a problem.
4240 Some spurious warnings can be avoided if you declare all the functions
4241 you use that never return as @code{noreturn}. @xref{Function
4244 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4246 @item -Wunknown-pragmas
4247 @opindex Wunknown-pragmas
4248 @opindex Wno-unknown-pragmas
4249 @cindex warning for unknown pragmas
4250 @cindex unknown pragmas, warning
4251 @cindex pragmas, warning of unknown
4252 Warn when a @code{#pragma} directive is encountered that is not understood by
4253 GCC@. If this command-line option is used, warnings are even issued
4254 for unknown pragmas in system header files. This is not the case if
4255 the warnings are only enabled by the @option{-Wall} command-line option.
4258 @opindex Wno-pragmas
4260 Do not warn about misuses of pragmas, such as incorrect parameters,
4261 invalid syntax, or conflicts between pragmas. See also
4262 @option{-Wunknown-pragmas}.
4264 @item -Wstrict-aliasing
4265 @opindex Wstrict-aliasing
4266 @opindex Wno-strict-aliasing
4267 This option is only active when @option{-fstrict-aliasing} is active.
4268 It warns about code that might break the strict aliasing rules that the
4269 compiler is using for optimization. The warning does not catch all
4270 cases, but does attempt to catch the more common pitfalls. It is
4271 included in @option{-Wall}.
4272 It is equivalent to @option{-Wstrict-aliasing=3}
4274 @item -Wstrict-aliasing=n
4275 @opindex Wstrict-aliasing=n
4276 This option is only active when @option{-fstrict-aliasing} is active.
4277 It warns about code that might break the strict aliasing rules that the
4278 compiler is using for optimization.
4279 Higher levels correspond to higher accuracy (fewer false positives).
4280 Higher levels also correspond to more effort, similar to the way @option{-O}
4282 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4284 Level 1: Most aggressive, quick, least accurate.
4285 Possibly useful when higher levels
4286 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4287 false negatives. However, it has many false positives.
4288 Warns for all pointer conversions between possibly incompatible types,
4289 even if never dereferenced. Runs in the front end only.
4291 Level 2: Aggressive, quick, not too precise.
4292 May still have many false positives (not as many as level 1 though),
4293 and few false negatives (but possibly more than level 1).
4294 Unlike level 1, it only warns when an address is taken. Warns about
4295 incomplete types. Runs in the front end only.
4297 Level 3 (default for @option{-Wstrict-aliasing}):
4298 Should have very few false positives and few false
4299 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4300 Takes care of the common pun+dereference pattern in the front end:
4301 @code{*(int*)&some_float}.
4302 If optimization is enabled, it also runs in the back end, where it deals
4303 with multiple statement cases using flow-sensitive points-to information.
4304 Only warns when the converted pointer is dereferenced.
4305 Does not warn about incomplete types.
4307 @item -Wstrict-overflow
4308 @itemx -Wstrict-overflow=@var{n}
4309 @opindex Wstrict-overflow
4310 @opindex Wno-strict-overflow
4311 This option is only active when @option{-fstrict-overflow} is active.
4312 It warns about cases where the compiler optimizes based on the
4313 assumption that signed overflow does not occur. Note that it does not
4314 warn about all cases where the code might overflow: it only warns
4315 about cases where the compiler implements some optimization. Thus
4316 this warning depends on the optimization level.
4318 An optimization that assumes that signed overflow does not occur is
4319 perfectly safe if the values of the variables involved are such that
4320 overflow never does, in fact, occur. Therefore this warning can
4321 easily give a false positive: a warning about code that is not
4322 actually a problem. To help focus on important issues, several
4323 warning levels are defined. No warnings are issued for the use of
4324 undefined signed overflow when estimating how many iterations a loop
4325 requires, in particular when determining whether a loop will be
4329 @item -Wstrict-overflow=1
4330 Warn about cases that are both questionable and easy to avoid. For
4331 example, with @option{-fstrict-overflow}, the compiler simplifies
4332 @code{x + 1 > x} to @code{1}. This level of
4333 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4334 are not, and must be explicitly requested.
4336 @item -Wstrict-overflow=2
4337 Also warn about other cases where a comparison is simplified to a
4338 constant. For example: @code{abs (x) >= 0}. This can only be
4339 simplified when @option{-fstrict-overflow} is in effect, because
4340 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4341 zero. @option{-Wstrict-overflow} (with no level) is the same as
4342 @option{-Wstrict-overflow=2}.
4344 @item -Wstrict-overflow=3
4345 Also warn about other cases where a comparison is simplified. For
4346 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4348 @item -Wstrict-overflow=4
4349 Also warn about other simplifications not covered by the above cases.
4350 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4352 @item -Wstrict-overflow=5
4353 Also warn about cases where the compiler reduces the magnitude of a
4354 constant involved in a comparison. For example: @code{x + 2 > y} is
4355 simplified to @code{x + 1 >= y}. This is reported only at the
4356 highest warning level because this simplification applies to many
4357 comparisons, so this warning level gives a very large number of
4361 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4362 @opindex Wsuggest-attribute=
4363 @opindex Wno-suggest-attribute=
4364 Warn for cases where adding an attribute may be beneficial. The
4365 attributes currently supported are listed below.
4368 @item -Wsuggest-attribute=pure
4369 @itemx -Wsuggest-attribute=const
4370 @itemx -Wsuggest-attribute=noreturn
4371 @opindex Wsuggest-attribute=pure
4372 @opindex Wno-suggest-attribute=pure
4373 @opindex Wsuggest-attribute=const
4374 @opindex Wno-suggest-attribute=const
4375 @opindex Wsuggest-attribute=noreturn
4376 @opindex Wno-suggest-attribute=noreturn
4378 Warn about functions that might be candidates for attributes
4379 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4380 functions visible in other compilation units or (in the case of @code{pure} and
4381 @code{const}) if it cannot prove that the function returns normally. A function
4382 returns normally if it doesn't contain an infinite loop or return abnormally
4383 by throwing, calling @code{abort} or trapping. This analysis requires option
4384 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4385 higher. Higher optimization levels improve the accuracy of the analysis.
4387 @item -Wsuggest-attribute=format
4388 @itemx -Wmissing-format-attribute
4389 @opindex Wsuggest-attribute=format
4390 @opindex Wmissing-format-attribute
4391 @opindex Wno-suggest-attribute=format
4392 @opindex Wno-missing-format-attribute
4396 Warn about function pointers that might be candidates for @code{format}
4397 attributes. Note these are only possible candidates, not absolute ones.
4398 GCC guesses that function pointers with @code{format} attributes that
4399 are used in assignment, initialization, parameter passing or return
4400 statements should have a corresponding @code{format} attribute in the
4401 resulting type. I.e.@: the left-hand side of the assignment or
4402 initialization, the type of the parameter variable, or the return type
4403 of the containing function respectively should also have a @code{format}
4404 attribute to avoid the warning.
4406 GCC also warns about function definitions that might be
4407 candidates for @code{format} attributes. Again, these are only
4408 possible candidates. GCC guesses that @code{format} attributes
4409 might be appropriate for any function that calls a function like
4410 @code{vprintf} or @code{vscanf}, but this might not always be the
4411 case, and some functions for which @code{format} attributes are
4412 appropriate may not be detected.
4415 @item -Wsuggest-final-types
4416 @opindex Wno-suggest-final-types
4417 @opindex Wsuggest-final-types
4418 Warn about types with virtual methods where code quality would be improved
4419 if the type were declared with the C++11 @code{final} specifier,
4421 declared in an anonymous namespace. This allows GCC to more aggressively
4422 devirtualize the polymorphic calls. This warning is more effective with link
4423 time optimization, where the information about the class hierarchy graph is
4426 @item -Wsuggest-final-methods
4427 @opindex Wno-suggest-final-methods
4428 @opindex Wsuggest-final-methods
4429 Warn about virtual methods where code quality would be improved if the method
4430 were declared with the C++11 @code{final} specifier,
4431 or, if possible, its type were
4432 declared in an anonymous namespace or with the @code{final} specifier.
4434 more effective with link time optimization, where the information about the
4435 class hierarchy graph is more complete. It is recommended to first consider
4436 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4439 @item -Wsuggest-override
4440 Warn about overriding virtual functions that are not marked with the override
4443 @item -Warray-bounds
4444 @itemx -Warray-bounds=@var{n}
4445 @opindex Wno-array-bounds
4446 @opindex Warray-bounds
4447 This option is only active when @option{-ftree-vrp} is active
4448 (default for @option{-O2} and above). It warns about subscripts to arrays
4449 that are always out of bounds. This warning is enabled by @option{-Wall}.
4452 @item -Warray-bounds=1
4453 This is the warning level of @option{-Warray-bounds} and is enabled
4454 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4456 @item -Warray-bounds=2
4457 This warning level also warns about out of bounds access for
4458 arrays at the end of a struct and for arrays accessed through
4459 pointers. This warning level may give a larger number of
4460 false positives and is deactivated by default.
4463 @item -Wbool-compare
4464 @opindex Wno-bool-compare
4465 @opindex Wbool-compare
4466 Warn about boolean expression compared with an integer value different from
4467 @code{true}/@code{false}. For instance, the following comparison is
4472 if ((n > 1) == 2) @{ @dots{} @}
4474 This warning is enabled by @option{-Wall}.
4476 @item -Wframe-address
4477 @opindex Wno-frame-address
4478 @opindex Wframe-address
4479 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4480 is called with an argument greater than 0. Such calls may return indeterminate
4481 values or crash the program. The warning is included in @option{-Wall}.
4483 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4484 @opindex Wno-discarded-qualifiers
4485 @opindex Wdiscarded-qualifiers
4486 Do not warn if type qualifiers on pointers are being discarded.
4487 Typically, the compiler warns if a @code{const char *} variable is
4488 passed to a function that takes a @code{char *} parameter. This option
4489 can be used to suppress such a warning.
4491 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4492 @opindex Wno-discarded-array-qualifiers
4493 @opindex Wdiscarded-array-qualifiers
4494 Do not warn if type qualifiers on arrays which are pointer targets
4495 are being discarded. Typically, the compiler warns if a
4496 @code{const int (*)[]} variable is passed to a function that
4497 takes a @code{int (*)[]} parameter. This option can be used to
4498 suppress such a warning.
4500 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4501 @opindex Wno-incompatible-pointer-types
4502 @opindex Wincompatible-pointer-types
4503 Do not warn when there is a conversion between pointers that have incompatible
4504 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4505 which warns for pointer argument passing or assignment with different
4508 @item -Wno-int-conversion @r{(C and Objective-C only)}
4509 @opindex Wno-int-conversion
4510 @opindex Wint-conversion
4511 Do not warn about incompatible integer to pointer and pointer to integer
4512 conversions. This warning is about implicit conversions; for explicit
4513 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4514 @option{-Wno-pointer-to-int-cast} may be used.
4516 @item -Wno-div-by-zero
4517 @opindex Wno-div-by-zero
4518 @opindex Wdiv-by-zero
4519 Do not warn about compile-time integer division by zero. Floating-point
4520 division by zero is not warned about, as it can be a legitimate way of
4521 obtaining infinities and NaNs.
4523 @item -Wsystem-headers
4524 @opindex Wsystem-headers
4525 @opindex Wno-system-headers
4526 @cindex warnings from system headers
4527 @cindex system headers, warnings from
4528 Print warning messages for constructs found in system header files.
4529 Warnings from system headers are normally suppressed, on the assumption
4530 that they usually do not indicate real problems and would only make the
4531 compiler output harder to read. Using this command-line option tells
4532 GCC to emit warnings from system headers as if they occurred in user
4533 code. However, note that using @option{-Wall} in conjunction with this
4534 option does @emph{not} warn about unknown pragmas in system
4535 headers---for that, @option{-Wunknown-pragmas} must also be used.
4537 @item -Wtautological-compare
4538 @opindex Wtautological-compare
4539 @opindex Wno-tautological-compare
4540 Warn if a self-comparison always evaluates to true or false. This
4541 warning detects various mistakes such as:
4545 if (i > i) @{ @dots{} @}
4547 This warning is enabled by @option{-Wall}.
4550 @opindex Wtrampolines
4551 @opindex Wno-trampolines
4552 Warn about trampolines generated for pointers to nested functions.
4553 A trampoline is a small piece of data or code that is created at run
4554 time on the stack when the address of a nested function is taken, and is
4555 used to call the nested function indirectly. For some targets, it is
4556 made up of data only and thus requires no special treatment. But, for
4557 most targets, it is made up of code and thus requires the stack to be
4558 made executable in order for the program to work properly.
4561 @opindex Wfloat-equal
4562 @opindex Wno-float-equal
4563 Warn if floating-point values are used in equality comparisons.
4565 The idea behind this is that sometimes it is convenient (for the
4566 programmer) to consider floating-point values as approximations to
4567 infinitely precise real numbers. If you are doing this, then you need
4568 to compute (by analyzing the code, or in some other way) the maximum or
4569 likely maximum error that the computation introduces, and allow for it
4570 when performing comparisons (and when producing output, but that's a
4571 different problem). In particular, instead of testing for equality, you
4572 should check to see whether the two values have ranges that overlap; and
4573 this is done with the relational operators, so equality comparisons are
4576 @item -Wtraditional @r{(C and Objective-C only)}
4577 @opindex Wtraditional
4578 @opindex Wno-traditional
4579 Warn about certain constructs that behave differently in traditional and
4580 ISO C@. Also warn about ISO C constructs that have no traditional C
4581 equivalent, and/or problematic constructs that should be avoided.
4585 Macro parameters that appear within string literals in the macro body.
4586 In traditional C macro replacement takes place within string literals,
4587 but in ISO C it does not.
4590 In traditional C, some preprocessor directives did not exist.
4591 Traditional preprocessors only considered a line to be a directive
4592 if the @samp{#} appeared in column 1 on the line. Therefore
4593 @option{-Wtraditional} warns about directives that traditional C
4594 understands but ignores because the @samp{#} does not appear as the
4595 first character on the line. It also suggests you hide directives like
4596 @code{#pragma} not understood by traditional C by indenting them. Some
4597 traditional implementations do not recognize @code{#elif}, so this option
4598 suggests avoiding it altogether.
4601 A function-like macro that appears without arguments.
4604 The unary plus operator.
4607 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4608 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4609 constants.) Note, these suffixes appear in macros defined in the system
4610 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4611 Use of these macros in user code might normally lead to spurious
4612 warnings, however GCC's integrated preprocessor has enough context to
4613 avoid warning in these cases.
4616 A function declared external in one block and then used after the end of
4620 A @code{switch} statement has an operand of type @code{long}.
4623 A non-@code{static} function declaration follows a @code{static} one.
4624 This construct is not accepted by some traditional C compilers.
4627 The ISO type of an integer constant has a different width or
4628 signedness from its traditional type. This warning is only issued if
4629 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4630 typically represent bit patterns, are not warned about.
4633 Usage of ISO string concatenation is detected.
4636 Initialization of automatic aggregates.
4639 Identifier conflicts with labels. Traditional C lacks a separate
4640 namespace for labels.
4643 Initialization of unions. If the initializer is zero, the warning is
4644 omitted. This is done under the assumption that the zero initializer in
4645 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4646 initializer warnings and relies on default initialization to zero in the
4650 Conversions by prototypes between fixed/floating-point values and vice
4651 versa. The absence of these prototypes when compiling with traditional
4652 C causes serious problems. This is a subset of the possible
4653 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4656 Use of ISO C style function definitions. This warning intentionally is
4657 @emph{not} issued for prototype declarations or variadic functions
4658 because these ISO C features appear in your code when using
4659 libiberty's traditional C compatibility macros, @code{PARAMS} and
4660 @code{VPARAMS}. This warning is also bypassed for nested functions
4661 because that feature is already a GCC extension and thus not relevant to
4662 traditional C compatibility.
4665 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4666 @opindex Wtraditional-conversion
4667 @opindex Wno-traditional-conversion
4668 Warn if a prototype causes a type conversion that is different from what
4669 would happen to the same argument in the absence of a prototype. This
4670 includes conversions of fixed point to floating and vice versa, and
4671 conversions changing the width or signedness of a fixed-point argument
4672 except when the same as the default promotion.
4674 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4675 @opindex Wdeclaration-after-statement
4676 @opindex Wno-declaration-after-statement
4677 Warn when a declaration is found after a statement in a block. This
4678 construct, known from C++, was introduced with ISO C99 and is by default
4679 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4684 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4686 @item -Wno-endif-labels
4687 @opindex Wno-endif-labels
4688 @opindex Wendif-labels
4689 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4694 Warn whenever a local variable or type declaration shadows another
4695 variable, parameter, type, class member (in C++), or instance variable
4696 (in Objective-C) or whenever a built-in function is shadowed. Note
4697 that in C++, the compiler warns if a local variable shadows an
4698 explicit typedef, but not if it shadows a struct/class/enum.
4700 @item -Wno-shadow-ivar @r{(Objective-C only)}
4701 @opindex Wno-shadow-ivar
4702 @opindex Wshadow-ivar
4703 Do not warn whenever a local variable shadows an instance variable in an
4706 @item -Wlarger-than=@var{len}
4707 @opindex Wlarger-than=@var{len}
4708 @opindex Wlarger-than-@var{len}
4709 Warn whenever an object of larger than @var{len} bytes is defined.
4711 @item -Wframe-larger-than=@var{len}
4712 @opindex Wframe-larger-than
4713 Warn if the size of a function frame is larger than @var{len} bytes.
4714 The computation done to determine the stack frame size is approximate
4715 and not conservative.
4716 The actual requirements may be somewhat greater than @var{len}
4717 even if you do not get a warning. In addition, any space allocated
4718 via @code{alloca}, variable-length arrays, or related constructs
4719 is not included by the compiler when determining
4720 whether or not to issue a warning.
4722 @item -Wno-free-nonheap-object
4723 @opindex Wno-free-nonheap-object
4724 @opindex Wfree-nonheap-object
4725 Do not warn when attempting to free an object that was not allocated
4728 @item -Wstack-usage=@var{len}
4729 @opindex Wstack-usage
4730 Warn if the stack usage of a function might be larger than @var{len} bytes.
4731 The computation done to determine the stack usage is conservative.
4732 Any space allocated via @code{alloca}, variable-length arrays, or related
4733 constructs is included by the compiler when determining whether or not to
4736 The message is in keeping with the output of @option{-fstack-usage}.
4740 If the stack usage is fully static but exceeds the specified amount, it's:
4743 warning: stack usage is 1120 bytes
4746 If the stack usage is (partly) dynamic but bounded, it's:
4749 warning: stack usage might be 1648 bytes
4752 If the stack usage is (partly) dynamic and not bounded, it's:
4755 warning: stack usage might be unbounded
4759 @item -Wunsafe-loop-optimizations
4760 @opindex Wunsafe-loop-optimizations
4761 @opindex Wno-unsafe-loop-optimizations
4762 Warn if the loop cannot be optimized because the compiler cannot
4763 assume anything on the bounds of the loop indices. With
4764 @option{-funsafe-loop-optimizations} warn if the compiler makes
4767 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4768 @opindex Wno-pedantic-ms-format
4769 @opindex Wpedantic-ms-format
4770 When used in combination with @option{-Wformat}
4771 and @option{-pedantic} without GNU extensions, this option
4772 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4773 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4774 which depend on the MS runtime.
4776 @item -Wpointer-arith
4777 @opindex Wpointer-arith
4778 @opindex Wno-pointer-arith
4779 Warn about anything that depends on the ``size of'' a function type or
4780 of @code{void}. GNU C assigns these types a size of 1, for
4781 convenience in calculations with @code{void *} pointers and pointers
4782 to functions. In C++, warn also when an arithmetic operation involves
4783 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4786 @opindex Wtype-limits
4787 @opindex Wno-type-limits
4788 Warn if a comparison is always true or always false due to the limited
4789 range of the data type, but do not warn for constant expressions. For
4790 example, warn if an unsigned variable is compared against zero with
4791 @code{<} or @code{>=}. This warning is also enabled by
4794 @item -Wbad-function-cast @r{(C and Objective-C only)}
4795 @opindex Wbad-function-cast
4796 @opindex Wno-bad-function-cast
4797 Warn when a function call is cast to a non-matching type.
4798 For example, warn if a call to a function returning an integer type
4799 is cast to a pointer type.
4801 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4802 @opindex Wc90-c99-compat
4803 @opindex Wno-c90-c99-compat
4804 Warn about features not present in ISO C90, but present in ISO C99.
4805 For instance, warn about use of variable length arrays, @code{long long}
4806 type, @code{bool} type, compound literals, designated initializers, and so
4807 on. This option is independent of the standards mode. Warnings are disabled
4808 in the expression that follows @code{__extension__}.
4810 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4811 @opindex Wc99-c11-compat
4812 @opindex Wno-c99-c11-compat
4813 Warn about features not present in ISO C99, but present in ISO C11.
4814 For instance, warn about use of anonymous structures and unions,
4815 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4816 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4817 and so on. This option is independent of the standards mode. Warnings are
4818 disabled in the expression that follows @code{__extension__}.
4820 @item -Wc++-compat @r{(C and Objective-C only)}
4821 @opindex Wc++-compat
4822 Warn about ISO C constructs that are outside of the common subset of
4823 ISO C and ISO C++, e.g.@: request for implicit conversion from
4824 @code{void *} to a pointer to non-@code{void} type.
4826 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4827 @opindex Wc++11-compat
4828 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4829 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4830 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4831 enabled by @option{-Wall}.
4833 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4834 @opindex Wc++14-compat
4835 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4836 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4840 @opindex Wno-cast-qual
4841 Warn whenever a pointer is cast so as to remove a type qualifier from
4842 the target type. For example, warn if a @code{const char *} is cast
4843 to an ordinary @code{char *}.
4845 Also warn when making a cast that introduces a type qualifier in an
4846 unsafe way. For example, casting @code{char **} to @code{const char **}
4847 is unsafe, as in this example:
4850 /* p is char ** value. */
4851 const char **q = (const char **) p;
4852 /* Assignment of readonly string to const char * is OK. */
4854 /* Now char** pointer points to read-only memory. */
4859 @opindex Wcast-align
4860 @opindex Wno-cast-align
4861 Warn whenever a pointer is cast such that the required alignment of the
4862 target is increased. For example, warn if a @code{char *} is cast to
4863 an @code{int *} on machines where integers can only be accessed at
4864 two- or four-byte boundaries.
4866 @item -Wwrite-strings
4867 @opindex Wwrite-strings
4868 @opindex Wno-write-strings
4869 When compiling C, give string constants the type @code{const
4870 char[@var{length}]} so that copying the address of one into a
4871 non-@code{const} @code{char *} pointer produces a warning. These
4872 warnings help you find at compile time code that can try to write
4873 into a string constant, but only if you have been very careful about
4874 using @code{const} in declarations and prototypes. Otherwise, it is
4875 just a nuisance. This is why we did not make @option{-Wall} request
4878 When compiling C++, warn about the deprecated conversion from string
4879 literals to @code{char *}. This warning is enabled by default for C++
4884 @opindex Wno-clobbered
4885 Warn for variables that might be changed by @code{longjmp} or
4886 @code{vfork}. This warning is also enabled by @option{-Wextra}.
4888 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
4889 @opindex Wconditionally-supported
4890 @opindex Wno-conditionally-supported
4891 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
4894 @opindex Wconversion
4895 @opindex Wno-conversion
4896 Warn for implicit conversions that may alter a value. This includes
4897 conversions between real and integer, like @code{abs (x)} when
4898 @code{x} is @code{double}; conversions between signed and unsigned,
4899 like @code{unsigned ui = -1}; and conversions to smaller types, like
4900 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
4901 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
4902 changed by the conversion like in @code{abs (2.0)}. Warnings about
4903 conversions between signed and unsigned integers can be disabled by
4904 using @option{-Wno-sign-conversion}.
4906 For C++, also warn for confusing overload resolution for user-defined
4907 conversions; and conversions that never use a type conversion
4908 operator: conversions to @code{void}, the same type, a base class or a
4909 reference to them. Warnings about conversions between signed and
4910 unsigned integers are disabled by default in C++ unless
4911 @option{-Wsign-conversion} is explicitly enabled.
4913 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
4914 @opindex Wconversion-null
4915 @opindex Wno-conversion-null
4916 Do not warn for conversions between @code{NULL} and non-pointer
4917 types. @option{-Wconversion-null} is enabled by default.
4919 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
4920 @opindex Wzero-as-null-pointer-constant
4921 @opindex Wno-zero-as-null-pointer-constant
4922 Warn when a literal '0' is used as null pointer constant. This can
4923 be useful to facilitate the conversion to @code{nullptr} in C++11.
4927 @opindex Wno-date-time
4928 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
4929 are encountered as they might prevent bit-wise-identical reproducible
4932 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
4933 @opindex Wdelete-incomplete
4934 @opindex Wno-delete-incomplete
4935 Warn when deleting a pointer to incomplete type, which may cause
4936 undefined behavior at runtime. This warning is enabled by default.
4938 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
4939 @opindex Wuseless-cast
4940 @opindex Wno-useless-cast
4941 Warn when an expression is casted to its own type.
4944 @opindex Wempty-body
4945 @opindex Wno-empty-body
4946 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
4947 while} statement. This warning is also enabled by @option{-Wextra}.
4949 @item -Wenum-compare
4950 @opindex Wenum-compare
4951 @opindex Wno-enum-compare
4952 Warn about a comparison between values of different enumerated types.
4953 In C++ enumeral mismatches in conditional expressions are also
4954 diagnosed and the warning is enabled by default. In C this warning is
4955 enabled by @option{-Wall}.
4957 @item -Wjump-misses-init @r{(C, Objective-C only)}
4958 @opindex Wjump-misses-init
4959 @opindex Wno-jump-misses-init
4960 Warn if a @code{goto} statement or a @code{switch} statement jumps
4961 forward across the initialization of a variable, or jumps backward to a
4962 label after the variable has been initialized. This only warns about
4963 variables that are initialized when they are declared. This warning is
4964 only supported for C and Objective-C; in C++ this sort of branch is an
4967 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
4968 can be disabled with the @option{-Wno-jump-misses-init} option.
4970 @item -Wsign-compare
4971 @opindex Wsign-compare
4972 @opindex Wno-sign-compare
4973 @cindex warning for comparison of signed and unsigned values
4974 @cindex comparison of signed and unsigned values, warning
4975 @cindex signed and unsigned values, comparison warning
4976 Warn when a comparison between signed and unsigned values could produce
4977 an incorrect result when the signed value is converted to unsigned.
4978 This warning is also enabled by @option{-Wextra}; to get the other warnings
4979 of @option{-Wextra} without this warning, use @option{-Wextra -Wno-sign-compare}.
4981 @item -Wsign-conversion
4982 @opindex Wsign-conversion
4983 @opindex Wno-sign-conversion
4984 Warn for implicit conversions that may change the sign of an integer
4985 value, like assigning a signed integer expression to an unsigned
4986 integer variable. An explicit cast silences the warning. In C, this
4987 option is enabled also by @option{-Wconversion}.
4989 @item -Wfloat-conversion
4990 @opindex Wfloat-conversion
4991 @opindex Wno-float-conversion
4992 Warn for implicit conversions that reduce the precision of a real value.
4993 This includes conversions from real to integer, and from higher precision
4994 real to lower precision real values. This option is also enabled by
4995 @option{-Wconversion}.
4997 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
4998 @opindex Wsized-deallocation
4999 @opindex Wno-sized-deallocation
5000 Warn about a definition of an unsized deallocation function
5002 void operator delete (void *) noexcept;
5003 void operator delete[] (void *) noexcept;
5005 without a definition of the corresponding sized deallocation function
5007 void operator delete (void *, std::size_t) noexcept;
5008 void operator delete[] (void *, std::size_t) noexcept;
5010 or vice versa. Enabled by @option{-Wextra} along with
5011 @option{-fsized-deallocation}.
5013 @item -Wsizeof-pointer-memaccess
5014 @opindex Wsizeof-pointer-memaccess
5015 @opindex Wno-sizeof-pointer-memaccess
5016 Warn for suspicious length parameters to certain string and memory built-in
5017 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5018 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5019 but a pointer, and suggests a possible fix, or about
5020 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5023 @item -Wsizeof-array-argument
5024 @opindex Wsizeof-array-argument
5025 @opindex Wno-sizeof-array-argument
5026 Warn when the @code{sizeof} operator is applied to a parameter that is
5027 declared as an array in a function definition. This warning is enabled by
5028 default for C and C++ programs.
5030 @item -Wmemset-transposed-args
5031 @opindex Wmemset-transposed-args
5032 @opindex Wno-memset-transposed-args
5033 Warn for suspicious calls to the @code{memset} built-in function, if the
5034 second argument is not zero and the third argument is zero. This warns e.g.@
5035 about @code{memset (buf, sizeof buf, 0)} where most probably
5036 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5037 is only emitted if the third argument is literal zero. If it is some
5038 expression that is folded to zero, a cast of zero to some type, etc.,
5039 it is far less likely that the user has mistakenly exchanged the arguments
5040 and no warning is emitted. This warning is enabled by @option{-Wall}.
5044 @opindex Wno-address
5045 Warn about suspicious uses of memory addresses. These include using
5046 the address of a function in a conditional expression, such as
5047 @code{void func(void); if (func)}, and comparisons against the memory
5048 address of a string literal, such as @code{if (x == "abc")}. Such
5049 uses typically indicate a programmer error: the address of a function
5050 always evaluates to true, so their use in a conditional usually
5051 indicate that the programmer forgot the parentheses in a function
5052 call; and comparisons against string literals result in unspecified
5053 behavior and are not portable in C, so they usually indicate that the
5054 programmer intended to use @code{strcmp}. This warning is enabled by
5058 @opindex Wlogical-op
5059 @opindex Wno-logical-op
5060 Warn about suspicious uses of logical operators in expressions.
5061 This includes using logical operators in contexts where a
5062 bit-wise operator is likely to be expected. Also warns when
5063 the operands of a logical operator are the same:
5066 if (a < 0 && a < 0) @{ @dots{} @}
5069 @item -Wlogical-not-parentheses
5070 @opindex Wlogical-not-parentheses
5071 @opindex Wno-logical-not-parentheses
5072 Warn about logical not used on the left hand side operand of a comparison.
5073 This option does not warn if the RHS operand is of a boolean type. Its
5074 purpose is to detect suspicious code like the following:
5078 if (!a > 1) @{ @dots{} @}
5081 It is possible to suppress the warning by wrapping the LHS into
5084 if ((!a) > 1) @{ @dots{} @}
5087 This warning is enabled by @option{-Wall}.
5089 @item -Waggregate-return
5090 @opindex Waggregate-return
5091 @opindex Wno-aggregate-return
5092 Warn if any functions that return structures or unions are defined or
5093 called. (In languages where you can return an array, this also elicits
5096 @item -Wno-aggressive-loop-optimizations
5097 @opindex Wno-aggressive-loop-optimizations
5098 @opindex Waggressive-loop-optimizations
5099 Warn if in a loop with constant number of iterations the compiler detects
5100 undefined behavior in some statement during one or more of the iterations.
5102 @item -Wno-attributes
5103 @opindex Wno-attributes
5104 @opindex Wattributes
5105 Do not warn if an unexpected @code{__attribute__} is used, such as
5106 unrecognized attributes, function attributes applied to variables,
5107 etc. This does not stop errors for incorrect use of supported
5110 @item -Wno-builtin-macro-redefined
5111 @opindex Wno-builtin-macro-redefined
5112 @opindex Wbuiltin-macro-redefined
5113 Do not warn if certain built-in macros are redefined. This suppresses
5114 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5115 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5117 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5118 @opindex Wstrict-prototypes
5119 @opindex Wno-strict-prototypes
5120 Warn if a function is declared or defined without specifying the
5121 argument types. (An old-style function definition is permitted without
5122 a warning if preceded by a declaration that specifies the argument
5125 @item -Wold-style-declaration @r{(C and Objective-C only)}
5126 @opindex Wold-style-declaration
5127 @opindex Wno-old-style-declaration
5128 Warn for obsolescent usages, according to the C Standard, in a
5129 declaration. For example, warn if storage-class specifiers like
5130 @code{static} are not the first things in a declaration. This warning
5131 is also enabled by @option{-Wextra}.
5133 @item -Wold-style-definition @r{(C and Objective-C only)}
5134 @opindex Wold-style-definition
5135 @opindex Wno-old-style-definition
5136 Warn if an old-style function definition is used. A warning is given
5137 even if there is a previous prototype.
5139 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5140 @opindex Wmissing-parameter-type
5141 @opindex Wno-missing-parameter-type
5142 A function parameter is declared without a type specifier in K&R-style
5149 This warning is also enabled by @option{-Wextra}.
5151 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5152 @opindex Wmissing-prototypes
5153 @opindex Wno-missing-prototypes
5154 Warn if a global function is defined without a previous prototype
5155 declaration. This warning is issued even if the definition itself
5156 provides a prototype. Use this option to detect global functions
5157 that do not have a matching prototype declaration in a header file.
5158 This option is not valid for C++ because all function declarations
5159 provide prototypes and a non-matching declaration declares an
5160 overload rather than conflict with an earlier declaration.
5161 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5163 @item -Wmissing-declarations
5164 @opindex Wmissing-declarations
5165 @opindex Wno-missing-declarations
5166 Warn if a global function is defined without a previous declaration.
5167 Do so even if the definition itself provides a prototype.
5168 Use this option to detect global functions that are not declared in
5169 header files. In C, no warnings are issued for functions with previous
5170 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5171 missing prototypes. In C++, no warnings are issued for function templates,
5172 or for inline functions, or for functions in anonymous namespaces.
5174 @item -Wmissing-field-initializers
5175 @opindex Wmissing-field-initializers
5176 @opindex Wno-missing-field-initializers
5180 Warn if a structure's initializer has some fields missing. For
5181 example, the following code causes such a warning, because
5182 @code{x.h} is implicitly zero:
5185 struct s @{ int f, g, h; @};
5186 struct s x = @{ 3, 4 @};
5189 This option does not warn about designated initializers, so the following
5190 modification does not trigger a warning:
5193 struct s @{ int f, g, h; @};
5194 struct s x = @{ .f = 3, .g = 4 @};
5197 In C++ this option does not warn either about the empty @{ @}
5198 initializer, for example:
5201 struct s @{ int f, g, h; @};
5205 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5206 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5208 @item -Wno-multichar
5209 @opindex Wno-multichar
5211 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5212 Usually they indicate a typo in the user's code, as they have
5213 implementation-defined values, and should not be used in portable code.
5215 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5216 @opindex Wnormalized=
5217 @opindex Wnormalized
5218 @opindex Wno-normalized
5221 @cindex character set, input normalization
5222 In ISO C and ISO C++, two identifiers are different if they are
5223 different sequences of characters. However, sometimes when characters
5224 outside the basic ASCII character set are used, you can have two
5225 different character sequences that look the same. To avoid confusion,
5226 the ISO 10646 standard sets out some @dfn{normalization rules} which
5227 when applied ensure that two sequences that look the same are turned into
5228 the same sequence. GCC can warn you if you are using identifiers that
5229 have not been normalized; this option controls that warning.
5231 There are four levels of warning supported by GCC@. The default is
5232 @option{-Wnormalized=nfc}, which warns about any identifier that is
5233 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5234 recommended form for most uses. It is equivalent to
5235 @option{-Wnormalized}.
5237 Unfortunately, there are some characters allowed in identifiers by
5238 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5239 identifiers. That is, there's no way to use these symbols in portable
5240 ISO C or C++ and have all your identifiers in NFC@.
5241 @option{-Wnormalized=id} suppresses the warning for these characters.
5242 It is hoped that future versions of the standards involved will correct
5243 this, which is why this option is not the default.
5245 You can switch the warning off for all characters by writing
5246 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5247 only do this if you are using some other normalization scheme (like
5248 ``D''), because otherwise you can easily create bugs that are
5249 literally impossible to see.
5251 Some characters in ISO 10646 have distinct meanings but look identical
5252 in some fonts or display methodologies, especially once formatting has
5253 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5254 LETTER N'', displays just like a regular @code{n} that has been
5255 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5256 normalization scheme to convert all these into a standard form as
5257 well, and GCC warns if your code is not in NFKC if you use
5258 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5259 about every identifier that contains the letter O because it might be
5260 confused with the digit 0, and so is not the default, but may be
5261 useful as a local coding convention if the programming environment
5262 cannot be fixed to display these characters distinctly.
5264 @item -Wno-deprecated
5265 @opindex Wno-deprecated
5266 @opindex Wdeprecated
5267 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5269 @item -Wno-deprecated-declarations
5270 @opindex Wno-deprecated-declarations
5271 @opindex Wdeprecated-declarations
5272 Do not warn about uses of functions (@pxref{Function Attributes}),
5273 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5274 Attributes}) marked as deprecated by using the @code{deprecated}
5278 @opindex Wno-overflow
5280 Do not warn about compile-time overflow in constant expressions.
5285 Warn about One Definition Rule violations during link-time optimization.
5286 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5289 @opindex Wopenm-simd
5290 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5291 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5292 option can be used to relax the cost model.
5294 @item -Woverride-init @r{(C and Objective-C only)}
5295 @opindex Woverride-init
5296 @opindex Wno-override-init
5300 Warn if an initialized field without side effects is overridden when
5301 using designated initializers (@pxref{Designated Inits, , Designated
5304 This warning is included in @option{-Wextra}. To get other
5305 @option{-Wextra} warnings without this one, use @option{-Wextra
5306 -Wno-override-init}.
5308 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5309 @opindex Woverride-init-side-effects
5310 @opindex Wno-override-init-side-effects
5311 Warn if an initialized field with side effects is overridden when
5312 using designated initializers (@pxref{Designated Inits, , Designated
5313 Initializers}). This warning is enabled by default.
5318 Warn if a structure is given the packed attribute, but the packed
5319 attribute has no effect on the layout or size of the structure.
5320 Such structures may be mis-aligned for little benefit. For
5321 instance, in this code, the variable @code{f.x} in @code{struct bar}
5322 is misaligned even though @code{struct bar} does not itself
5323 have the packed attribute:
5330 @} __attribute__((packed));
5338 @item -Wpacked-bitfield-compat
5339 @opindex Wpacked-bitfield-compat
5340 @opindex Wno-packed-bitfield-compat
5341 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5342 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5343 the change can lead to differences in the structure layout. GCC
5344 informs you when the offset of such a field has changed in GCC 4.4.
5345 For example there is no longer a 4-bit padding between field @code{a}
5346 and @code{b} in this structure:
5353 @} __attribute__ ((packed));
5356 This warning is enabled by default. Use
5357 @option{-Wno-packed-bitfield-compat} to disable this warning.
5362 Warn if padding is included in a structure, either to align an element
5363 of the structure or to align the whole structure. Sometimes when this
5364 happens it is possible to rearrange the fields of the structure to
5365 reduce the padding and so make the structure smaller.
5367 @item -Wredundant-decls
5368 @opindex Wredundant-decls
5369 @opindex Wno-redundant-decls
5370 Warn if anything is declared more than once in the same scope, even in
5371 cases where multiple declaration is valid and changes nothing.
5373 @item -Wnested-externs @r{(C and Objective-C only)}
5374 @opindex Wnested-externs
5375 @opindex Wno-nested-externs
5376 Warn if an @code{extern} declaration is encountered within a function.
5378 @item -Wno-inherited-variadic-ctor
5379 @opindex Winherited-variadic-ctor
5380 @opindex Wno-inherited-variadic-ctor
5381 Suppress warnings about use of C++11 inheriting constructors when the
5382 base class inherited from has a C variadic constructor; the warning is
5383 on by default because the ellipsis is not inherited.
5388 Warn if a function that is declared as inline cannot be inlined.
5389 Even with this option, the compiler does not warn about failures to
5390 inline functions declared in system headers.
5392 The compiler uses a variety of heuristics to determine whether or not
5393 to inline a function. For example, the compiler takes into account
5394 the size of the function being inlined and the amount of inlining
5395 that has already been done in the current function. Therefore,
5396 seemingly insignificant changes in the source program can cause the
5397 warnings produced by @option{-Winline} to appear or disappear.
5399 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5400 @opindex Wno-invalid-offsetof
5401 @opindex Winvalid-offsetof
5402 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5403 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5404 to a non-standard-layout type is undefined. In existing C++ implementations,
5405 however, @code{offsetof} typically gives meaningful results.
5406 This flag is for users who are aware that they are
5407 writing nonportable code and who have deliberately chosen to ignore the
5410 The restrictions on @code{offsetof} may be relaxed in a future version
5411 of the C++ standard.
5413 @item -Wno-int-to-pointer-cast
5414 @opindex Wno-int-to-pointer-cast
5415 @opindex Wint-to-pointer-cast
5416 Suppress warnings from casts to pointer type of an integer of a
5417 different size. In C++, casting to a pointer type of smaller size is
5418 an error. @option{Wint-to-pointer-cast} is enabled by default.
5421 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5422 @opindex Wno-pointer-to-int-cast
5423 @opindex Wpointer-to-int-cast
5424 Suppress warnings from casts from a pointer to an integer type of a
5428 @opindex Winvalid-pch
5429 @opindex Wno-invalid-pch
5430 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5431 the search path but can't be used.
5435 @opindex Wno-long-long
5436 Warn if @code{long long} type is used. This is enabled by either
5437 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5438 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5440 @item -Wvariadic-macros
5441 @opindex Wvariadic-macros
5442 @opindex Wno-variadic-macros
5443 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5444 alternate syntax is used in ISO C99 mode. This is enabled by either
5445 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5446 messages, use @option{-Wno-variadic-macros}.
5450 @opindex Wno-varargs
5451 Warn upon questionable usage of the macros used to handle variable
5452 arguments like @code{va_start}. This is default. To inhibit the
5453 warning messages, use @option{-Wno-varargs}.
5455 @item -Wvector-operation-performance
5456 @opindex Wvector-operation-performance
5457 @opindex Wno-vector-operation-performance
5458 Warn if vector operation is not implemented via SIMD capabilities of the
5459 architecture. Mainly useful for the performance tuning.
5460 Vector operation can be implemented @code{piecewise}, which means that the
5461 scalar operation is performed on every vector element;
5462 @code{in parallel}, which means that the vector operation is implemented
5463 using scalars of wider type, which normally is more performance efficient;
5464 and @code{as a single scalar}, which means that vector fits into a
5467 @item -Wno-virtual-move-assign
5468 @opindex Wvirtual-move-assign
5469 @opindex Wno-virtual-move-assign
5470 Suppress warnings about inheriting from a virtual base with a
5471 non-trivial C++11 move assignment operator. This is dangerous because
5472 if the virtual base is reachable along more than one path, it is
5473 moved multiple times, which can mean both objects end up in the
5474 moved-from state. If the move assignment operator is written to avoid
5475 moving from a moved-from object, this warning can be disabled.
5480 Warn if variable length array is used in the code.
5481 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5482 the variable length array.
5484 @item -Wvolatile-register-var
5485 @opindex Wvolatile-register-var
5486 @opindex Wno-volatile-register-var
5487 Warn if a register variable is declared volatile. The volatile
5488 modifier does not inhibit all optimizations that may eliminate reads
5489 and/or writes to register variables. This warning is enabled by
5492 @item -Wdisabled-optimization
5493 @opindex Wdisabled-optimization
5494 @opindex Wno-disabled-optimization
5495 Warn if a requested optimization pass is disabled. This warning does
5496 not generally indicate that there is anything wrong with your code; it
5497 merely indicates that GCC's optimizers are unable to handle the code
5498 effectively. Often, the problem is that your code is too big or too
5499 complex; GCC refuses to optimize programs when the optimization
5500 itself is likely to take inordinate amounts of time.
5502 @item -Wpointer-sign @r{(C and Objective-C only)}
5503 @opindex Wpointer-sign
5504 @opindex Wno-pointer-sign
5505 Warn for pointer argument passing or assignment with different signedness.
5506 This option is only supported for C and Objective-C@. It is implied by
5507 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5508 @option{-Wno-pointer-sign}.
5510 @item -Wstack-protector
5511 @opindex Wstack-protector
5512 @opindex Wno-stack-protector
5513 This option is only active when @option{-fstack-protector} is active. It
5514 warns about functions that are not protected against stack smashing.
5516 @item -Woverlength-strings
5517 @opindex Woverlength-strings
5518 @opindex Wno-overlength-strings
5519 Warn about string constants that are longer than the ``minimum
5520 maximum'' length specified in the C standard. Modern compilers
5521 generally allow string constants that are much longer than the
5522 standard's minimum limit, but very portable programs should avoid
5523 using longer strings.
5525 The limit applies @emph{after} string constant concatenation, and does
5526 not count the trailing NUL@. In C90, the limit was 509 characters; in
5527 C99, it was raised to 4095. C++98 does not specify a normative
5528 minimum maximum, so we do not diagnose overlength strings in C++@.
5530 This option is implied by @option{-Wpedantic}, and can be disabled with
5531 @option{-Wno-overlength-strings}.
5533 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5534 @opindex Wunsuffixed-float-constants
5536 Issue a warning for any floating constant that does not have
5537 a suffix. When used together with @option{-Wsystem-headers} it
5538 warns about such constants in system header files. This can be useful
5539 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5540 from the decimal floating-point extension to C99.
5542 @item -Wno-designated-init @r{(C and Objective-C only)}
5543 Suppress warnings when a positional initializer is used to initialize
5544 a structure that has been marked with the @code{designated_init}
5549 @node Debugging Options
5550 @section Options for Debugging Your Program or GCC
5551 @cindex options, debugging
5552 @cindex debugging information options
5554 GCC has various special options that are used for debugging
5555 either your program or GCC:
5560 Produce debugging information in the operating system's native format
5561 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5564 On most systems that use stabs format, @option{-g} enables use of extra
5565 debugging information that only GDB can use; this extra information
5566 makes debugging work better in GDB but probably makes other debuggers
5568 refuse to read the program. If you want to control for certain whether
5569 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5570 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5572 GCC allows you to use @option{-g} with
5573 @option{-O}. The shortcuts taken by optimized code may occasionally
5574 produce surprising results: some variables you declared may not exist
5575 at all; flow of control may briefly move where you did not expect it;
5576 some statements may not be executed because they compute constant
5577 results or their values are already at hand; some statements may
5578 execute in different places because they have been moved out of loops.
5580 Nevertheless it proves possible to debug optimized output. This makes
5581 it reasonable to use the optimizer for programs that might have bugs.
5583 The following options are useful when GCC is generated with the
5584 capability for more than one debugging format.
5587 @opindex gsplit-dwarf
5588 Separate as much dwarf debugging information as possible into a
5589 separate output file with the extension .dwo. This option allows
5590 the build system to avoid linking files with debug information. To
5591 be useful, this option requires a debugger capable of reading .dwo
5596 Produce debugging information for use by GDB@. This means to use the
5597 most expressive format available (DWARF 2, stabs, or the native format
5598 if neither of those are supported), including GDB extensions if at all
5603 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5605 @item -ggnu-pubnames
5606 @opindex ggnu-pubnames
5607 Generate .debug_pubnames and .debug_pubtypes sections in a format
5608 suitable for conversion into a GDB@ index. This option is only useful
5609 with a linker that can produce GDB@ index version 7.
5613 Produce debugging information in stabs format (if that is supported),
5614 without GDB extensions. This is the format used by DBX on most BSD
5615 systems. On MIPS, Alpha and System V Release 4 systems this option
5616 produces stabs debugging output that is not understood by DBX or SDB@.
5617 On System V Release 4 systems this option requires the GNU assembler.
5619 @item -feliminate-unused-debug-symbols
5620 @opindex feliminate-unused-debug-symbols
5621 Produce debugging information in stabs format (if that is supported),
5622 for only symbols that are actually used.
5624 @item -femit-class-debug-always
5625 @opindex femit-class-debug-always
5626 Instead of emitting debugging information for a C++ class in only one
5627 object file, emit it in all object files using the class. This option
5628 should be used only with debuggers that are unable to handle the way GCC
5629 normally emits debugging information for classes because using this
5630 option increases the size of debugging information by as much as a
5633 @item -fdebug-types-section
5634 @opindex fdebug-types-section
5635 @opindex fno-debug-types-section
5636 When using DWARF Version 4 or higher, type DIEs can be put into
5637 their own @code{.debug_types} section instead of making them part of the
5638 @code{.debug_info} section. It is more efficient to put them in a separate
5639 comdat sections since the linker can then remove duplicates.
5640 But not all DWARF consumers support @code{.debug_types} sections yet
5641 and on some objects @code{.debug_types} produces larger instead of smaller
5642 debugging information.
5646 Produce debugging information in stabs format (if that is supported),
5647 using GNU extensions understood only by the GNU debugger (GDB)@. The
5648 use of these extensions is likely to make other debuggers crash or
5649 refuse to read the program.
5653 Produce debugging information in COFF format (if that is supported).
5654 This is the format used by SDB on most System V systems prior to
5659 Produce debugging information in XCOFF format (if that is supported).
5660 This is the format used by the DBX debugger on IBM RS/6000 systems.
5664 Produce debugging information in XCOFF format (if that is supported),
5665 using GNU extensions understood only by the GNU debugger (GDB)@. The
5666 use of these extensions is likely to make other debuggers crash or
5667 refuse to read the program, and may cause assemblers other than the GNU
5668 assembler (GAS) to fail with an error.
5670 @item -gdwarf-@var{version}
5671 @opindex gdwarf-@var{version}
5672 Produce debugging information in DWARF format (if that is supported).
5673 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5674 for most targets is 4. DWARF Version 5 is only experimental.
5676 Note that with DWARF Version 2, some ports require and always
5677 use some non-conflicting DWARF 3 extensions in the unwind tables.
5679 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5680 for maximum benefit.
5682 @item -grecord-gcc-switches
5683 @opindex grecord-gcc-switches
5684 This switch causes the command-line options used to invoke the
5685 compiler that may affect code generation to be appended to the
5686 DW_AT_producer attribute in DWARF debugging information. The options
5687 are concatenated with spaces separating them from each other and from
5688 the compiler version. See also @option{-frecord-gcc-switches} for another
5689 way of storing compiler options into the object file. This is the default.
5691 @item -gno-record-gcc-switches
5692 @opindex gno-record-gcc-switches
5693 Disallow appending command-line options to the DW_AT_producer attribute
5694 in DWARF debugging information.
5696 @item -gstrict-dwarf
5697 @opindex gstrict-dwarf
5698 Disallow using extensions of later DWARF standard version than selected
5699 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5700 DWARF extensions from later standard versions is allowed.
5702 @item -gno-strict-dwarf
5703 @opindex gno-strict-dwarf
5704 Allow using extensions of later DWARF standard version than selected with
5705 @option{-gdwarf-@var{version}}.
5707 @item -gz@r{[}=@var{type}@r{]}
5709 Produce compressed debug sections in DWARF format, if that is supported.
5710 If @var{type} is not given, the default type depends on the capabilities
5711 of the assembler and linker used. @var{type} may be one of
5712 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5713 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5714 compression in traditional GNU format). If the linker doesn't support
5715 writing compressed debug sections, the option is rejected. Otherwise,
5716 if the assembler does not support them, @option{-gz} is silently ignored
5717 when producing object files.
5721 Produce debugging information in Alpha/VMS debug format (if that is
5722 supported). This is the format used by DEBUG on Alpha/VMS systems.
5725 @itemx -ggdb@var{level}
5726 @itemx -gstabs@var{level}
5727 @itemx -gcoff@var{level}
5728 @itemx -gxcoff@var{level}
5729 @itemx -gvms@var{level}
5730 Request debugging information and also use @var{level} to specify how
5731 much information. The default level is 2.
5733 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5736 Level 1 produces minimal information, enough for making backtraces in
5737 parts of the program that you don't plan to debug. This includes
5738 descriptions of functions and external variables, and line number
5739 tables, but no information about local variables.
5741 Level 3 includes extra information, such as all the macro definitions
5742 present in the program. Some debuggers support macro expansion when
5743 you use @option{-g3}.
5745 @option{-gdwarf-2} does not accept a concatenated debug level, because
5746 GCC used to support an option @option{-gdwarf} that meant to generate
5747 debug information in version 1 of the DWARF format (which is very
5748 different from version 2), and it would have been too confusing. That
5749 debug format is long obsolete, but the option cannot be changed now.
5750 Instead use an additional @option{-g@var{level}} option to change the
5751 debug level for DWARF.
5755 Turn off generation of debug info, if leaving out this option
5756 generates it, or turn it on at level 2 otherwise. The position of this
5757 argument in the command line does not matter; it takes effect after all
5758 other options are processed, and it does so only once, no matter how
5759 many times it is given. This is mainly intended to be used with
5760 @option{-fcompare-debug}.
5762 @item -fsanitize=address
5763 @opindex fsanitize=address
5764 Enable AddressSanitizer, a fast memory error detector.
5765 Memory access instructions are instrumented to detect
5766 out-of-bounds and use-after-free bugs.
5767 See @uref{http://code.google.com/p/address-sanitizer/} for
5768 more details. The run-time behavior can be influenced using the
5769 @env{ASAN_OPTIONS} environment variable; see
5770 @url{https://code.google.com/p/address-sanitizer/wiki/Flags#Run-time_flags} for
5771 a list of supported options.
5773 @item -fsanitize=kernel-address
5774 @opindex fsanitize=kernel-address
5775 Enable AddressSanitizer for Linux kernel.
5776 See @uref{http://code.google.com/p/address-sanitizer/wiki/AddressSanitizerForKernel} for more details.
5778 @item -fsanitize=thread
5779 @opindex fsanitize=thread
5780 Enable ThreadSanitizer, a fast data race detector.
5781 Memory access instructions are instrumented to detect
5782 data race bugs. See @uref{http://code.google.com/p/thread-sanitizer/} for more
5783 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5784 environment variable; see
5785 @url{https://code.google.com/p/thread-sanitizer/wiki/Flags} for a list of
5788 @item -fsanitize=leak
5789 @opindex fsanitize=leak
5790 Enable LeakSanitizer, a memory leak detector.
5791 This option only matters for linking of executables and if neither
5792 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5793 case the executable is linked against a library that overrides @code{malloc}
5794 and other allocator functions. See
5795 @uref{https://code.google.com/p/address-sanitizer/wiki/LeakSanitizer} for more
5796 details. The run-time behavior can be influenced using the
5797 @env{LSAN_OPTIONS} environment variable.
5799 @item -fsanitize=undefined
5800 @opindex fsanitize=undefined
5801 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5802 Various computations are instrumented to detect undefined behavior
5803 at runtime. Current suboptions are:
5807 @item -fsanitize=shift
5808 @opindex fsanitize=shift
5809 This option enables checking that the result of a shift operation is
5810 not undefined. Note that what exactly is considered undefined differs
5811 slightly between C and C++, as well as between ISO C90 and C99, etc.
5813 @item -fsanitize=integer-divide-by-zero
5814 @opindex fsanitize=integer-divide-by-zero
5815 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5817 @item -fsanitize=unreachable
5818 @opindex fsanitize=unreachable
5819 With this option, the compiler turns the @code{__builtin_unreachable}
5820 call into a diagnostics message call instead. When reaching the
5821 @code{__builtin_unreachable} call, the behavior is undefined.
5823 @item -fsanitize=vla-bound
5824 @opindex fsanitize=vla-bound
5825 This option instructs the compiler to check that the size of a variable
5826 length array is positive.
5828 @item -fsanitize=null
5829 @opindex fsanitize=null
5830 This option enables pointer checking. Particularly, the application
5831 built with this option turned on will issue an error message when it
5832 tries to dereference a NULL pointer, or if a reference (possibly an
5833 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5834 on an object pointed by a NULL pointer.
5836 @item -fsanitize=return
5837 @opindex fsanitize=return
5838 This option enables return statement checking. Programs
5839 built with this option turned on will issue an error message
5840 when the end of a non-void function is reached without actually
5841 returning a value. This option works in C++ only.
5843 @item -fsanitize=signed-integer-overflow
5844 @opindex fsanitize=signed-integer-overflow
5845 This option enables signed integer overflow checking. We check that
5846 the result of @code{+}, @code{*}, and both unary and binary @code{-}
5847 does not overflow in the signed arithmetics. Note, integer promotion
5848 rules must be taken into account. That is, the following is not an
5851 signed char a = SCHAR_MAX;
5855 @item -fsanitize=bounds
5856 @opindex fsanitize=bounds
5857 This option enables instrumentation of array bounds. Various out of bounds
5858 accesses are detected. Flexible array members, flexible array member-like
5859 arrays, and initializers of variables with static storage are not instrumented.
5861 @item -fsanitize=bounds-strict
5862 @opindex fsanitize=bounds-strict
5863 This option enables strict instrumentation of array bounds. Most out of bounds
5864 accesses are detected, including flexible array members and flexible array
5865 member-like arrays. Initializers of variables with static storage are not
5868 @item -fsanitize=alignment
5869 @opindex fsanitize=alignment
5871 This option enables checking of alignment of pointers when they are
5872 dereferenced, or when a reference is bound to insufficiently aligned target,
5873 or when a method or constructor is invoked on insufficiently aligned object.
5875 @item -fsanitize=object-size
5876 @opindex fsanitize=object-size
5877 This option enables instrumentation of memory references using the
5878 @code{__builtin_object_size} function. Various out of bounds pointer
5879 accesses are detected.
5881 @item -fsanitize=float-divide-by-zero
5882 @opindex fsanitize=float-divide-by-zero
5883 Detect floating-point division by zero. Unlike other similar options,
5884 @option{-fsanitize=float-divide-by-zero} is not enabled by
5885 @option{-fsanitize=undefined}, since floating-point division by zero can
5886 be a legitimate way of obtaining infinities and NaNs.
5888 @item -fsanitize=float-cast-overflow
5889 @opindex fsanitize=float-cast-overflow
5890 This option enables floating-point type to integer conversion checking.
5891 We check that the result of the conversion does not overflow.
5892 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
5893 not enabled by @option{-fsanitize=undefined}.
5894 This option does not work well with @code{FE_INVALID} exceptions enabled.
5896 @item -fsanitize=nonnull-attribute
5897 @opindex fsanitize=nonnull-attribute
5899 This option enables instrumentation of calls, checking whether null values
5900 are not passed to arguments marked as requiring a non-null value by the
5901 @code{nonnull} function attribute.
5903 @item -fsanitize=returns-nonnull-attribute
5904 @opindex fsanitize=returns-nonnull-attribute
5906 This option enables instrumentation of return statements in functions
5907 marked with @code{returns_nonnull} function attribute, to detect returning
5908 of null values from such functions.
5910 @item -fsanitize=bool
5911 @opindex fsanitize=bool
5913 This option enables instrumentation of loads from bool. If a value other
5914 than 0/1 is loaded, a run-time error is issued.
5916 @item -fsanitize=enum
5917 @opindex fsanitize=enum
5919 This option enables instrumentation of loads from an enum type. If
5920 a value outside the range of values for the enum type is loaded,
5921 a run-time error is issued.
5923 @item -fsanitize=vptr
5924 @opindex fsanitize=vptr
5926 This option enables instrumentation of C++ member function calls, member
5927 accesses and some conversions between pointers to base and derived classes,
5928 to verify the referenced object has the correct dynamic type.
5932 While @option{-ftrapv} causes traps for signed overflows to be emitted,
5933 @option{-fsanitize=undefined} gives a diagnostic message.
5934 This currently works only for the C family of languages.
5936 @item -fno-sanitize=all
5937 @opindex fno-sanitize=all
5939 This option disables all previously enabled sanitizers.
5940 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
5943 @item -fasan-shadow-offset=@var{number}
5944 @opindex fasan-shadow-offset
5945 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
5946 It is useful for experimenting with different shadow memory layouts in
5947 Kernel AddressSanitizer.
5949 @item -fsanitize-sections=@var{s1},@var{s2},...
5950 @opindex fsanitize-sections
5951 Sanitize global variables in selected user-defined sections. @var{si} may
5954 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
5955 @opindex fsanitize-recover
5956 @opindex fno-sanitize-recover
5957 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
5958 mentioned in comma-separated list of @var{opts}. Enabling this option
5959 for a sanitizer component causes it to attempt to continue
5960 running the program as if no error happened. This means multiple
5961 runtime errors can be reported in a single program run, and the exit
5962 code of the program may indicate success even when errors
5963 have been reported. The @option{-fno-sanitize-recover=} option
5964 can be used to alter
5965 this behavior: only the first detected error is reported
5966 and program then exits with a non-zero exit code.
5968 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
5969 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
5970 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero} and
5971 @option{-fsanitize=kernel-address}. For these sanitizers error recovery is turned on by default.
5972 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
5973 accepted, the former enables recovery for all sanitizers that support it,
5974 the latter disables recovery for all sanitizers that support it.
5976 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
5978 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5981 Similarly @option{-fno-sanitize-recover} is equivalent to
5983 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
5986 @item -fsanitize-undefined-trap-on-error
5987 @opindex fsanitize-undefined-trap-on-error
5988 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
5989 report undefined behavior using @code{__builtin_trap} rather than
5990 a @code{libubsan} library routine. The advantage of this is that the
5991 @code{libubsan} library is not needed and is not linked in, so this
5992 is usable even in freestanding environments.
5994 @item -fcheck-pointer-bounds
5995 @opindex fcheck-pointer-bounds
5996 @opindex fno-check-pointer-bounds
5997 @cindex Pointer Bounds Checker options
5998 Enable Pointer Bounds Checker instrumentation. Each memory reference
5999 is instrumented with checks of the pointer used for memory access against
6000 bounds associated with that pointer.
6003 is only an implementation for Intel MPX available, thus x86 target
6004 and @option{-mmpx} are required to enable this feature.
6005 MPX-based instrumentation requires
6006 a runtime library to enable MPX in hardware and handle bounds
6007 violation signals. By default when @option{-fcheck-pointer-bounds}
6008 and @option{-mmpx} options are used to link a program, the GCC driver
6009 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
6010 library. It also passes '-z bndplt' to a linker in case it supports this
6011 option (which is checked on libmpx configuration). Note that old versions
6012 of linker may ignore option. Gold linker doesn't support '-z bndplt'
6013 option. With no '-z bndplt' support in linker all calls to dynamic libraries
6014 lose passed bounds reducing overall protection level. It's highly
6015 recommended to use linker with '-z bndplt' support. In case such linker
6016 is not available it is adviced to always use @option{-static-libmpxwrappers}
6017 for better protection level or use @option{-static} to completely avoid
6018 external calls to dynamic libraries. MPX-based instrumentation
6019 may be used for debugging and also may be included in production code
6020 to increase program security. Depending on usage, you may
6021 have different requirements for the runtime library. The current version
6022 of the MPX runtime library is more oriented for use as a debugging
6023 tool. MPX runtime library usage implies @option{-lpthread}. See
6024 also @option{-static-libmpx}. The runtime library behavior can be
6025 influenced using various @env{CHKP_RT_*} environment variables. See
6026 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
6029 Generated instrumentation may be controlled by various
6030 @option{-fchkp-*} options and by the @code{bnd_variable_size}
6031 structure field attribute (@pxref{Type Attributes}) and
6032 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
6033 (@pxref{Function Attributes}). GCC also provides a number of built-in
6034 functions for controlling the Pointer Bounds Checker. @xref{Pointer
6035 Bounds Checker builtins}, for more information.
6037 @item -fchkp-check-incomplete-type
6038 @opindex fchkp-check-incomplete-type
6039 @opindex fno-chkp-check-incomplete-type
6040 Generate pointer bounds checks for variables with incomplete type.
6043 @item -fchkp-narrow-bounds
6044 @opindex fchkp-narrow-bounds
6045 @opindex fno-chkp-narrow-bounds
6046 Controls bounds used by Pointer Bounds Checker for pointers to object
6047 fields. If narrowing is enabled then field bounds are used. Otherwise
6048 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
6049 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
6051 @item -fchkp-first-field-has-own-bounds
6052 @opindex fchkp-first-field-has-own-bounds
6053 @opindex fno-chkp-first-field-has-own-bounds
6054 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
6055 first field in the structure. By default a pointer to the first field has
6056 the same bounds as a pointer to the whole structure.
6058 @item -fchkp-narrow-to-innermost-array
6059 @opindex fchkp-narrow-to-innermost-array
6060 @opindex fno-chkp-narrow-to-innermost-array
6061 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
6062 case of nested static array access. By default this option is disabled and
6063 bounds of the outermost array are used.
6065 @item -fchkp-optimize
6066 @opindex fchkp-optimize
6067 @opindex fno-chkp-optimize
6068 Enables Pointer Bounds Checker optimizations. Enabled by default at
6069 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
6071 @item -fchkp-use-fast-string-functions
6072 @opindex fchkp-use-fast-string-functions
6073 @opindex fno-chkp-use-fast-string-functions
6074 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
6075 by Pointer Bounds Checker. Disabled by default.
6077 @item -fchkp-use-nochk-string-functions
6078 @opindex fchkp-use-nochk-string-functions
6079 @opindex fno-chkp-use-nochk-string-functions
6080 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
6081 by Pointer Bounds Checker. Disabled by default.
6083 @item -fchkp-use-static-bounds
6084 @opindex fchkp-use-static-bounds
6085 @opindex fno-chkp-use-static-bounds
6086 Allow Pointer Bounds Checker to generate static bounds holding
6087 bounds of static variables. Enabled by default.
6089 @item -fchkp-use-static-const-bounds
6090 @opindex fchkp-use-static-const-bounds
6091 @opindex fno-chkp-use-static-const-bounds
6092 Use statically-initialized bounds for constant bounds instead of
6093 generating them each time they are required. By default enabled when
6094 @option{-fchkp-use-static-bounds} is enabled.
6096 @item -fchkp-treat-zero-dynamic-size-as-infinite
6097 @opindex fchkp-treat-zero-dynamic-size-as-infinite
6098 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
6099 With this option, objects with incomplete type whose
6100 dynamically-obtained size is zero are treated as having infinite size
6101 instead by Pointer Bounds
6102 Checker. This option may be helpful if a program is linked with a library
6103 missing size information for some symbols. Disabled by default.
6105 @item -fchkp-check-read
6106 @opindex fchkp-check-read
6107 @opindex fno-chkp-check-read
6108 Instructs Pointer Bounds Checker to generate checks for all read
6109 accesses to memory. Enabled by default.
6111 @item -fchkp-check-write
6112 @opindex fchkp-check-write
6113 @opindex fno-chkp-check-write
6114 Instructs Pointer Bounds Checker to generate checks for all write
6115 accesses to memory. Enabled by default.
6117 @item -fchkp-store-bounds
6118 @opindex fchkp-store-bounds
6119 @opindex fno-chkp-store-bounds
6120 Instructs Pointer Bounds Checker to generate bounds stores for
6121 pointer writes. Enabled by default.
6123 @item -fchkp-instrument-calls
6124 @opindex fchkp-instrument-calls
6125 @opindex fno-chkp-instrument-calls
6126 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
6129 @item -fchkp-instrument-marked-only
6130 @opindex fchkp-instrument-marked-only
6131 @opindex fno-chkp-instrument-marked-only
6132 Instructs Pointer Bounds Checker to instrument only functions
6133 marked with the @code{bnd_instrument} attribute
6134 (@pxref{Function Attributes}). Disabled by default.
6136 @item -fchkp-use-wrappers
6137 @opindex fchkp-use-wrappers
6138 @opindex fno-chkp-use-wrappers
6139 Allows Pointer Bounds Checker to replace calls to built-in functions
6140 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6141 is used to link a program, the GCC driver automatically links
6142 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6145 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6146 @opindex fdump-final-insns
6147 Dump the final internal representation (RTL) to @var{file}. If the
6148 optional argument is omitted (or if @var{file} is @code{.}), the name
6149 of the dump file is determined by appending @code{.gkd} to the
6150 compilation output file name.
6152 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6153 @opindex fcompare-debug
6154 @opindex fno-compare-debug
6155 If no error occurs during compilation, run the compiler a second time,
6156 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6157 passed to the second compilation. Dump the final internal
6158 representation in both compilations, and print an error if they differ.
6160 If the equal sign is omitted, the default @option{-gtoggle} is used.
6162 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6163 and nonzero, implicitly enables @option{-fcompare-debug}. If
6164 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6165 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6168 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6169 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6170 of the final representation and the second compilation, preventing even
6171 @env{GCC_COMPARE_DEBUG} from taking effect.
6173 To verify full coverage during @option{-fcompare-debug} testing, set
6174 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6175 which GCC rejects as an invalid option in any actual compilation
6176 (rather than preprocessing, assembly or linking). To get just a
6177 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6178 not overridden} will do.
6180 @item -fcompare-debug-second
6181 @opindex fcompare-debug-second
6182 This option is implicitly passed to the compiler for the second
6183 compilation requested by @option{-fcompare-debug}, along with options to
6184 silence warnings, and omitting other options that would cause
6185 side-effect compiler outputs to files or to the standard output. Dump
6186 files and preserved temporary files are renamed so as to contain the
6187 @code{.gk} additional extension during the second compilation, to avoid
6188 overwriting those generated by the first.
6190 When this option is passed to the compiler driver, it causes the
6191 @emph{first} compilation to be skipped, which makes it useful for little
6192 other than debugging the compiler proper.
6194 @item -feliminate-dwarf2-dups
6195 @opindex feliminate-dwarf2-dups
6196 Compress DWARF 2 debugging information by eliminating duplicated
6197 information about each symbol. This option only makes sense when
6198 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6200 @item -femit-struct-debug-baseonly
6201 @opindex femit-struct-debug-baseonly
6202 Emit debug information for struct-like types
6203 only when the base name of the compilation source file
6204 matches the base name of file in which the struct is defined.
6206 This option substantially reduces the size of debugging information,
6207 but at significant potential loss in type information to the debugger.
6208 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6209 See @option{-femit-struct-debug-detailed} for more detailed control.
6211 This option works only with DWARF 2.
6213 @item -femit-struct-debug-reduced
6214 @opindex femit-struct-debug-reduced
6215 Emit debug information for struct-like types
6216 only when the base name of the compilation source file
6217 matches the base name of file in which the type is defined,
6218 unless the struct is a template or defined in a system header.
6220 This option significantly reduces the size of debugging information,
6221 with some potential loss in type information to the debugger.
6222 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6223 See @option{-femit-struct-debug-detailed} for more detailed control.
6225 This option works only with DWARF 2.
6227 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6228 @opindex femit-struct-debug-detailed
6229 Specify the struct-like types
6230 for which the compiler generates debug information.
6231 The intent is to reduce duplicate struct debug information
6232 between different object files within the same program.
6234 This option is a detailed version of
6235 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6236 which serves for most needs.
6238 A specification has the syntax@*
6239 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6241 The optional first word limits the specification to
6242 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6243 A struct type is used directly when it is the type of a variable, member.
6244 Indirect uses arise through pointers to structs.
6245 That is, when use of an incomplete struct is valid, the use is indirect.
6247 @samp{struct one direct; struct two * indirect;}.
6249 The optional second word limits the specification to
6250 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6251 Generic structs are a bit complicated to explain.
6252 For C++, these are non-explicit specializations of template classes,
6253 or non-template classes within the above.
6254 Other programming languages have generics,
6255 but @option{-femit-struct-debug-detailed} does not yet implement them.
6257 The third word specifies the source files for those
6258 structs for which the compiler should emit debug information.
6259 The values @samp{none} and @samp{any} have the normal meaning.
6260 The value @samp{base} means that
6261 the base of name of the file in which the type declaration appears
6262 must match the base of the name of the main compilation file.
6263 In practice, this means that when compiling @file{foo.c}, debug information
6264 is generated for types declared in that file and @file{foo.h},
6265 but not other header files.
6266 The value @samp{sys} means those types satisfying @samp{base}
6267 or declared in system or compiler headers.
6269 You may need to experiment to determine the best settings for your application.
6271 The default is @option{-femit-struct-debug-detailed=all}.
6273 This option works only with DWARF 2.
6275 @item -fno-merge-debug-strings
6276 @opindex fmerge-debug-strings
6277 @opindex fno-merge-debug-strings
6278 Direct the linker to not merge together strings in the debugging
6279 information that are identical in different object files. Merging is
6280 not supported by all assemblers or linkers. Merging decreases the size
6281 of the debug information in the output file at the cost of increasing
6282 link processing time. Merging is enabled by default.
6284 @item -fdebug-prefix-map=@var{old}=@var{new}
6285 @opindex fdebug-prefix-map
6286 When compiling files in directory @file{@var{old}}, record debugging
6287 information describing them as in @file{@var{new}} instead.
6289 @item -fno-dwarf2-cfi-asm
6290 @opindex fdwarf2-cfi-asm
6291 @opindex fno-dwarf2-cfi-asm
6292 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6293 instead of using GAS @code{.cfi_*} directives.
6295 @cindex @command{prof}
6298 Generate extra code to write profile information suitable for the
6299 analysis program @command{prof}. You must use this option when compiling
6300 the source files you want data about, and you must also use it when
6303 @cindex @command{gprof}
6306 Generate extra code to write profile information suitable for the
6307 analysis program @command{gprof}. You must use this option when compiling
6308 the source files you want data about, and you must also use it when
6313 Makes the compiler print out each function name as it is compiled, and
6314 print some statistics about each pass when it finishes.
6317 @opindex ftime-report
6318 Makes the compiler print some statistics about the time consumed by each
6319 pass when it finishes.
6322 @opindex fmem-report
6323 Makes the compiler print some statistics about permanent memory
6324 allocation when it finishes.
6326 @item -fmem-report-wpa
6327 @opindex fmem-report-wpa
6328 Makes the compiler print some statistics about permanent memory
6329 allocation for the WPA phase only.
6331 @item -fpre-ipa-mem-report
6332 @opindex fpre-ipa-mem-report
6333 @item -fpost-ipa-mem-report
6334 @opindex fpost-ipa-mem-report
6335 Makes the compiler print some statistics about permanent memory
6336 allocation before or after interprocedural optimization.
6338 @item -fprofile-report
6339 @opindex fprofile-report
6340 Makes the compiler print some statistics about consistency of the
6341 (estimated) profile and effect of individual passes.
6344 @opindex fstack-usage
6345 Makes the compiler output stack usage information for the program, on a
6346 per-function basis. The filename for the dump is made by appending
6347 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6348 the output file, if explicitly specified and it is not an executable,
6349 otherwise it is the basename of the source file. An entry is made up
6354 The name of the function.
6358 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6361 The qualifier @code{static} means that the function manipulates the stack
6362 statically: a fixed number of bytes are allocated for the frame on function
6363 entry and released on function exit; no stack adjustments are otherwise made
6364 in the function. The second field is this fixed number of bytes.
6366 The qualifier @code{dynamic} means that the function manipulates the stack
6367 dynamically: in addition to the static allocation described above, stack
6368 adjustments are made in the body of the function, for example to push/pop
6369 arguments around function calls. If the qualifier @code{bounded} is also
6370 present, the amount of these adjustments is bounded at compile time and
6371 the second field is an upper bound of the total amount of stack used by
6372 the function. If it is not present, the amount of these adjustments is
6373 not bounded at compile time and the second field only represents the
6376 @item -fprofile-arcs
6377 @opindex fprofile-arcs
6378 Add code so that program flow @dfn{arcs} are instrumented. During
6379 execution the program records how many times each branch and call is
6380 executed and how many times it is taken or returns. When the compiled
6381 program exits it saves this data to a file called
6382 @file{@var{auxname}.gcda} for each source file. The data may be used for
6383 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6384 test coverage analysis (@option{-ftest-coverage}). Each object file's
6385 @var{auxname} is generated from the name of the output file, if
6386 explicitly specified and it is not the final executable, otherwise it is
6387 the basename of the source file. In both cases any suffix is removed
6388 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6389 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6390 @xref{Cross-profiling}.
6392 @cindex @command{gcov}
6396 This option is used to compile and link code instrumented for coverage
6397 analysis. The option is a synonym for @option{-fprofile-arcs}
6398 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6399 linking). See the documentation for those options for more details.
6404 Compile the source files with @option{-fprofile-arcs} plus optimization
6405 and code generation options. For test coverage analysis, use the
6406 additional @option{-ftest-coverage} option. You do not need to profile
6407 every source file in a program.
6410 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6411 (the latter implies the former).
6414 Run the program on a representative workload to generate the arc profile
6415 information. This may be repeated any number of times. You can run
6416 concurrent instances of your program, and provided that the file system
6417 supports locking, the data files will be correctly updated. Also
6418 @code{fork} calls are detected and correctly handled (double counting
6422 For profile-directed optimizations, compile the source files again with
6423 the same optimization and code generation options plus
6424 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6425 Control Optimization}).
6428 For test coverage analysis, use @command{gcov} to produce human readable
6429 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6430 @command{gcov} documentation for further information.
6434 With @option{-fprofile-arcs}, for each function of your program GCC
6435 creates a program flow graph, then finds a spanning tree for the graph.
6436 Only arcs that are not on the spanning tree have to be instrumented: the
6437 compiler adds code to count the number of times that these arcs are
6438 executed. When an arc is the only exit or only entrance to a block, the
6439 instrumentation code can be added to the block; otherwise, a new basic
6440 block must be created to hold the instrumentation code.
6443 @item -ftest-coverage
6444 @opindex ftest-coverage
6445 Produce a notes file that the @command{gcov} code-coverage utility
6446 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6447 show program coverage. Each source file's note file is called
6448 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6449 above for a description of @var{auxname} and instructions on how to
6450 generate test coverage data. Coverage data matches the source files
6451 more closely if you do not optimize.
6453 @item -fdbg-cnt-list
6454 @opindex fdbg-cnt-list
6455 Print the name and the counter upper bound for all debug counters.
6458 @item -fdbg-cnt=@var{counter-value-list}
6460 Set the internal debug counter upper bound. @var{counter-value-list}
6461 is a comma-separated list of @var{name}:@var{value} pairs
6462 which sets the upper bound of each debug counter @var{name} to @var{value}.
6463 All debug counters have the initial upper bound of @code{UINT_MAX};
6464 thus @code{dbg_cnt} returns true always unless the upper bound
6465 is set by this option.
6466 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6467 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6469 @item -fenable-@var{kind}-@var{pass}
6470 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6474 This is a set of options that are used to explicitly disable/enable
6475 optimization passes. These options are intended for use for debugging GCC.
6476 Compiler users should use regular options for enabling/disabling
6481 @item -fdisable-ipa-@var{pass}
6482 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6483 statically invoked in the compiler multiple times, the pass name should be
6484 appended with a sequential number starting from 1.
6486 @item -fdisable-rtl-@var{pass}
6487 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6488 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6489 statically invoked in the compiler multiple times, the pass name should be
6490 appended with a sequential number starting from 1. @var{range-list} is a
6491 comma-separated list of function ranges or assembler names. Each range is a number
6492 pair separated by a colon. The range is inclusive in both ends. If the range
6493 is trivial, the number pair can be simplified as a single number. If the
6494 function's call graph node's @var{uid} falls within one of the specified ranges,
6495 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6496 function header of a dump file, and the pass names can be dumped by using
6497 option @option{-fdump-passes}.
6499 @item -fdisable-tree-@var{pass}
6500 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6501 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6504 @item -fenable-ipa-@var{pass}
6505 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6506 statically invoked in the compiler multiple times, the pass name should be
6507 appended with a sequential number starting from 1.
6509 @item -fenable-rtl-@var{pass}
6510 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6511 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6512 description and examples.
6514 @item -fenable-tree-@var{pass}
6515 @itemx -fenable-tree-@var{pass}=@var{range-list}
6516 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6517 of option arguments.
6521 Here are some examples showing uses of these options.
6525 # disable ccp1 for all functions
6527 # disable complete unroll for function whose cgraph node uid is 1
6528 -fenable-tree-cunroll=1
6529 # disable gcse2 for functions at the following ranges [1,1],
6530 # [300,400], and [400,1000]
6531 # disable gcse2 for functions foo and foo2
6532 -fdisable-rtl-gcse2=foo,foo2
6533 # disable early inlining
6534 -fdisable-tree-einline
6535 # disable ipa inlining
6536 -fdisable-ipa-inline
6537 # enable tree full unroll
6538 -fenable-tree-unroll
6542 @item -d@var{letters}
6543 @itemx -fdump-rtl-@var{pass}
6544 @itemx -fdump-rtl-@var{pass}=@var{filename}
6546 @opindex fdump-rtl-@var{pass}
6547 Says to make debugging dumps during compilation at times specified by
6548 @var{letters}. This is used for debugging the RTL-based passes of the
6549 compiler. The file names for most of the dumps are made by appending
6550 a pass number and a word to the @var{dumpname}, and the files are
6551 created in the directory of the output file. In case of
6552 @option{=@var{filename}} option, the dump is output on the given file
6553 instead of the pass numbered dump files. Note that the pass number is
6554 computed statically as passes get registered into the pass manager.
6555 Thus the numbering is not related to the dynamic order of execution of
6556 passes. In particular, a pass installed by a plugin could have a
6557 number over 200 even if it executed quite early. @var{dumpname} is
6558 generated from the name of the output file, if explicitly specified
6559 and it is not an executable, otherwise it is the basename of the
6560 source file. These switches may have different effects when
6561 @option{-E} is used for preprocessing.
6563 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6564 @option{-d} option @var{letters}. Here are the possible
6565 letters for use in @var{pass} and @var{letters}, and their meanings:
6569 @item -fdump-rtl-alignments
6570 @opindex fdump-rtl-alignments
6571 Dump after branch alignments have been computed.
6573 @item -fdump-rtl-asmcons
6574 @opindex fdump-rtl-asmcons
6575 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6577 @item -fdump-rtl-auto_inc_dec
6578 @opindex fdump-rtl-auto_inc_dec
6579 Dump after auto-inc-dec discovery. This pass is only run on
6580 architectures that have auto inc or auto dec instructions.
6582 @item -fdump-rtl-barriers
6583 @opindex fdump-rtl-barriers
6584 Dump after cleaning up the barrier instructions.
6586 @item -fdump-rtl-bbpart
6587 @opindex fdump-rtl-bbpart
6588 Dump after partitioning hot and cold basic blocks.
6590 @item -fdump-rtl-bbro
6591 @opindex fdump-rtl-bbro
6592 Dump after block reordering.
6594 @item -fdump-rtl-btl1
6595 @itemx -fdump-rtl-btl2
6596 @opindex fdump-rtl-btl2
6597 @opindex fdump-rtl-btl2
6598 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6599 after the two branch
6600 target load optimization passes.
6602 @item -fdump-rtl-bypass
6603 @opindex fdump-rtl-bypass
6604 Dump after jump bypassing and control flow optimizations.
6606 @item -fdump-rtl-combine
6607 @opindex fdump-rtl-combine
6608 Dump after the RTL instruction combination pass.
6610 @item -fdump-rtl-compgotos
6611 @opindex fdump-rtl-compgotos
6612 Dump after duplicating the computed gotos.
6614 @item -fdump-rtl-ce1
6615 @itemx -fdump-rtl-ce2
6616 @itemx -fdump-rtl-ce3
6617 @opindex fdump-rtl-ce1
6618 @opindex fdump-rtl-ce2
6619 @opindex fdump-rtl-ce3
6620 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6621 @option{-fdump-rtl-ce3} enable dumping after the three
6622 if conversion passes.
6624 @item -fdump-rtl-cprop_hardreg
6625 @opindex fdump-rtl-cprop_hardreg
6626 Dump after hard register copy propagation.
6628 @item -fdump-rtl-csa
6629 @opindex fdump-rtl-csa
6630 Dump after combining stack adjustments.
6632 @item -fdump-rtl-cse1
6633 @itemx -fdump-rtl-cse2
6634 @opindex fdump-rtl-cse1
6635 @opindex fdump-rtl-cse2
6636 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6637 the two common subexpression elimination passes.
6639 @item -fdump-rtl-dce
6640 @opindex fdump-rtl-dce
6641 Dump after the standalone dead code elimination passes.
6643 @item -fdump-rtl-dbr
6644 @opindex fdump-rtl-dbr
6645 Dump after delayed branch scheduling.
6647 @item -fdump-rtl-dce1
6648 @itemx -fdump-rtl-dce2
6649 @opindex fdump-rtl-dce1
6650 @opindex fdump-rtl-dce2
6651 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6652 the two dead store elimination passes.
6655 @opindex fdump-rtl-eh
6656 Dump after finalization of EH handling code.
6658 @item -fdump-rtl-eh_ranges
6659 @opindex fdump-rtl-eh_ranges
6660 Dump after conversion of EH handling range regions.
6662 @item -fdump-rtl-expand
6663 @opindex fdump-rtl-expand
6664 Dump after RTL generation.
6666 @item -fdump-rtl-fwprop1
6667 @itemx -fdump-rtl-fwprop2
6668 @opindex fdump-rtl-fwprop1
6669 @opindex fdump-rtl-fwprop2
6670 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6671 dumping after the two forward propagation passes.
6673 @item -fdump-rtl-gcse1
6674 @itemx -fdump-rtl-gcse2
6675 @opindex fdump-rtl-gcse1
6676 @opindex fdump-rtl-gcse2
6677 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6678 after global common subexpression elimination.
6680 @item -fdump-rtl-init-regs
6681 @opindex fdump-rtl-init-regs
6682 Dump after the initialization of the registers.
6684 @item -fdump-rtl-initvals
6685 @opindex fdump-rtl-initvals
6686 Dump after the computation of the initial value sets.
6688 @item -fdump-rtl-into_cfglayout
6689 @opindex fdump-rtl-into_cfglayout
6690 Dump after converting to cfglayout mode.
6692 @item -fdump-rtl-ira
6693 @opindex fdump-rtl-ira
6694 Dump after iterated register allocation.
6696 @item -fdump-rtl-jump
6697 @opindex fdump-rtl-jump
6698 Dump after the second jump optimization.
6700 @item -fdump-rtl-loop2
6701 @opindex fdump-rtl-loop2
6702 @option{-fdump-rtl-loop2} enables dumping after the rtl
6703 loop optimization passes.
6705 @item -fdump-rtl-mach
6706 @opindex fdump-rtl-mach
6707 Dump after performing the machine dependent reorganization pass, if that
6710 @item -fdump-rtl-mode_sw
6711 @opindex fdump-rtl-mode_sw
6712 Dump after removing redundant mode switches.
6714 @item -fdump-rtl-rnreg
6715 @opindex fdump-rtl-rnreg
6716 Dump after register renumbering.
6718 @item -fdump-rtl-outof_cfglayout
6719 @opindex fdump-rtl-outof_cfglayout
6720 Dump after converting from cfglayout mode.
6722 @item -fdump-rtl-peephole2
6723 @opindex fdump-rtl-peephole2
6724 Dump after the peephole pass.
6726 @item -fdump-rtl-postreload
6727 @opindex fdump-rtl-postreload
6728 Dump after post-reload optimizations.
6730 @item -fdump-rtl-pro_and_epilogue
6731 @opindex fdump-rtl-pro_and_epilogue
6732 Dump after generating the function prologues and epilogues.
6734 @item -fdump-rtl-sched1
6735 @itemx -fdump-rtl-sched2
6736 @opindex fdump-rtl-sched1
6737 @opindex fdump-rtl-sched2
6738 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6739 after the basic block scheduling passes.
6741 @item -fdump-rtl-ree
6742 @opindex fdump-rtl-ree
6743 Dump after sign/zero extension elimination.
6745 @item -fdump-rtl-seqabstr
6746 @opindex fdump-rtl-seqabstr
6747 Dump after common sequence discovery.
6749 @item -fdump-rtl-shorten
6750 @opindex fdump-rtl-shorten
6751 Dump after shortening branches.
6753 @item -fdump-rtl-sibling
6754 @opindex fdump-rtl-sibling
6755 Dump after sibling call optimizations.
6757 @item -fdump-rtl-split1
6758 @itemx -fdump-rtl-split2
6759 @itemx -fdump-rtl-split3
6760 @itemx -fdump-rtl-split4
6761 @itemx -fdump-rtl-split5
6762 @opindex fdump-rtl-split1
6763 @opindex fdump-rtl-split2
6764 @opindex fdump-rtl-split3
6765 @opindex fdump-rtl-split4
6766 @opindex fdump-rtl-split5
6767 These options enable dumping after five rounds of
6768 instruction splitting.
6770 @item -fdump-rtl-sms
6771 @opindex fdump-rtl-sms
6772 Dump after modulo scheduling. This pass is only run on some
6775 @item -fdump-rtl-stack
6776 @opindex fdump-rtl-stack
6777 Dump after conversion from GCC's ``flat register file'' registers to the
6778 x87's stack-like registers. This pass is only run on x86 variants.
6780 @item -fdump-rtl-subreg1
6781 @itemx -fdump-rtl-subreg2
6782 @opindex fdump-rtl-subreg1
6783 @opindex fdump-rtl-subreg2
6784 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6785 the two subreg expansion passes.
6787 @item -fdump-rtl-unshare
6788 @opindex fdump-rtl-unshare
6789 Dump after all rtl has been unshared.
6791 @item -fdump-rtl-vartrack
6792 @opindex fdump-rtl-vartrack
6793 Dump after variable tracking.
6795 @item -fdump-rtl-vregs
6796 @opindex fdump-rtl-vregs
6797 Dump after converting virtual registers to hard registers.
6799 @item -fdump-rtl-web
6800 @opindex fdump-rtl-web
6801 Dump after live range splitting.
6803 @item -fdump-rtl-regclass
6804 @itemx -fdump-rtl-subregs_of_mode_init
6805 @itemx -fdump-rtl-subregs_of_mode_finish
6806 @itemx -fdump-rtl-dfinit
6807 @itemx -fdump-rtl-dfinish
6808 @opindex fdump-rtl-regclass
6809 @opindex fdump-rtl-subregs_of_mode_init
6810 @opindex fdump-rtl-subregs_of_mode_finish
6811 @opindex fdump-rtl-dfinit
6812 @opindex fdump-rtl-dfinish
6813 These dumps are defined but always produce empty files.
6816 @itemx -fdump-rtl-all
6818 @opindex fdump-rtl-all
6819 Produce all the dumps listed above.
6823 Annotate the assembler output with miscellaneous debugging information.
6827 Dump all macro definitions, at the end of preprocessing, in addition to
6832 Produce a core dump whenever an error occurs.
6836 Annotate the assembler output with a comment indicating which
6837 pattern and alternative is used. The length of each instruction is
6842 Dump the RTL in the assembler output as a comment before each instruction.
6843 Also turns on @option{-dp} annotation.
6847 Just generate RTL for a function instead of compiling it. Usually used
6848 with @option{-fdump-rtl-expand}.
6852 @opindex fdump-noaddr
6853 When doing debugging dumps, suppress address output. This makes it more
6854 feasible to use diff on debugging dumps for compiler invocations with
6855 different compiler binaries and/or different
6856 text / bss / data / heap / stack / dso start locations.
6859 @opindex freport-bug
6860 Collect and dump debug information into temporary file if ICE in C/C++
6863 @item -fdump-unnumbered
6864 @opindex fdump-unnumbered
6865 When doing debugging dumps, suppress instruction numbers and address output.
6866 This makes it more feasible to use diff on debugging dumps for compiler
6867 invocations with different options, in particular with and without
6870 @item -fdump-unnumbered-links
6871 @opindex fdump-unnumbered-links
6872 When doing debugging dumps (see @option{-d} option above), suppress
6873 instruction numbers for the links to the previous and next instructions
6876 @item -fdump-translation-unit @r{(C++ only)}
6877 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
6878 @opindex fdump-translation-unit
6879 Dump a representation of the tree structure for the entire translation
6880 unit to a file. The file name is made by appending @file{.tu} to the
6881 source file name, and the file is created in the same directory as the
6882 output file. If the @samp{-@var{options}} form is used, @var{options}
6883 controls the details of the dump as described for the
6884 @option{-fdump-tree} options.
6886 @item -fdump-class-hierarchy @r{(C++ only)}
6887 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
6888 @opindex fdump-class-hierarchy
6889 Dump a representation of each class's hierarchy and virtual function
6890 table layout to a file. The file name is made by appending
6891 @file{.class} to the source file name, and the file is created in the
6892 same directory as the output file. If the @samp{-@var{options}} form
6893 is used, @var{options} controls the details of the dump as described
6894 for the @option{-fdump-tree} options.
6896 @item -fdump-ipa-@var{switch}
6898 Control the dumping at various stages of inter-procedural analysis
6899 language tree to a file. The file name is generated by appending a
6900 switch specific suffix to the source file name, and the file is created
6901 in the same directory as the output file. The following dumps are
6906 Enables all inter-procedural analysis dumps.
6909 Dumps information about call-graph optimization, unused function removal,
6910 and inlining decisions.
6913 Dump after function inlining.
6918 @opindex fdump-passes
6919 Dump the list of optimization passes that are turned on and off by
6920 the current command-line options.
6922 @item -fdump-statistics-@var{option}
6923 @opindex fdump-statistics
6924 Enable and control dumping of pass statistics in a separate file. The
6925 file name is generated by appending a suffix ending in
6926 @samp{.statistics} to the source file name, and the file is created in
6927 the same directory as the output file. If the @samp{-@var{option}}
6928 form is used, @samp{-stats} causes counters to be summed over the
6929 whole compilation unit while @samp{-details} dumps every event as
6930 the passes generate them. The default with no option is to sum
6931 counters for each function compiled.
6933 @item -fdump-tree-@var{switch}
6934 @itemx -fdump-tree-@var{switch}-@var{options}
6935 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
6937 Control the dumping at various stages of processing the intermediate
6938 language tree to a file. The file name is generated by appending a
6939 switch-specific suffix to the source file name, and the file is
6940 created in the same directory as the output file. In case of
6941 @option{=@var{filename}} option, the dump is output on the given file
6942 instead of the auto named dump files. If the @samp{-@var{options}}
6943 form is used, @var{options} is a list of @samp{-} separated options
6944 which control the details of the dump. Not all options are applicable
6945 to all dumps; those that are not meaningful are ignored. The
6946 following options are available
6950 Print the address of each node. Usually this is not meaningful as it
6951 changes according to the environment and source file. Its primary use
6952 is for tying up a dump file with a debug environment.
6954 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
6955 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
6956 use working backward from mangled names in the assembly file.
6958 When dumping front-end intermediate representations, inhibit dumping
6959 of members of a scope or body of a function merely because that scope
6960 has been reached. Only dump such items when they are directly reachable
6963 When dumping pretty-printed trees, this option inhibits dumping the
6964 bodies of control structures.
6966 When dumping RTL, print the RTL in slim (condensed) form instead of
6967 the default LISP-like representation.
6969 Print a raw representation of the tree. By default, trees are
6970 pretty-printed into a C-like representation.
6972 Enable more detailed dumps (not honored by every dump option). Also
6973 include information from the optimization passes.
6975 Enable dumping various statistics about the pass (not honored by every dump
6978 Enable showing basic block boundaries (disabled in raw dumps).
6980 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
6981 dump a representation of the control flow graph suitable for viewing with
6982 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
6983 the file is pretty-printed as a subgraph, so that GraphViz can render them
6984 all in a single plot.
6986 This option currently only works for RTL dumps, and the RTL is always
6987 dumped in slim form.
6989 Enable showing virtual operands for every statement.
6991 Enable showing line numbers for statements.
6993 Enable showing the unique ID (@code{DECL_UID}) for each variable.
6995 Enable showing the tree dump for each statement.
6997 Enable showing the EH region number holding each statement.
6999 Enable showing scalar evolution analysis details.
7001 Enable showing optimization information (only available in certain
7004 Enable showing missed optimization information (only available in certain
7007 Enable other detailed optimization information (only available in
7009 @item =@var{filename}
7010 Instead of an auto named dump file, output into the given file
7011 name. The file names @file{stdout} and @file{stderr} are treated
7012 specially and are considered already open standard streams. For
7016 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
7017 -fdump-tree-pre=stderr file.c
7020 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
7021 output on to @file{stderr}. If two conflicting dump filenames are
7022 given for the same pass, then the latter option overrides the earlier
7026 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
7027 and @option{lineno}.
7030 Turn on all optimization options, i.e., @option{optimized},
7031 @option{missed}, and @option{note}.
7034 The following tree dumps are possible:
7038 @opindex fdump-tree-original
7039 Dump before any tree based optimization, to @file{@var{file}.original}.
7042 @opindex fdump-tree-optimized
7043 Dump after all tree based optimization, to @file{@var{file}.optimized}.
7046 @opindex fdump-tree-gimple
7047 Dump each function before and after the gimplification pass to a file. The
7048 file name is made by appending @file{.gimple} to the source file name.
7051 @opindex fdump-tree-cfg
7052 Dump the control flow graph of each function to a file. The file name is
7053 made by appending @file{.cfg} to the source file name.
7056 @opindex fdump-tree-ch
7057 Dump each function after copying loop headers. The file name is made by
7058 appending @file{.ch} to the source file name.
7061 @opindex fdump-tree-ssa
7062 Dump SSA related information to a file. The file name is made by appending
7063 @file{.ssa} to the source file name.
7066 @opindex fdump-tree-alias
7067 Dump aliasing information for each function. The file name is made by
7068 appending @file{.alias} to the source file name.
7071 @opindex fdump-tree-ccp
7072 Dump each function after CCP@. The file name is made by appending
7073 @file{.ccp} to the source file name.
7076 @opindex fdump-tree-storeccp
7077 Dump each function after STORE-CCP@. The file name is made by appending
7078 @file{.storeccp} to the source file name.
7081 @opindex fdump-tree-pre
7082 Dump trees after partial redundancy elimination. The file name is made
7083 by appending @file{.pre} to the source file name.
7086 @opindex fdump-tree-fre
7087 Dump trees after full redundancy elimination. The file name is made
7088 by appending @file{.fre} to the source file name.
7091 @opindex fdump-tree-copyprop
7092 Dump trees after copy propagation. The file name is made
7093 by appending @file{.copyprop} to the source file name.
7095 @item store_copyprop
7096 @opindex fdump-tree-store_copyprop
7097 Dump trees after store copy-propagation. The file name is made
7098 by appending @file{.store_copyprop} to the source file name.
7101 @opindex fdump-tree-dce
7102 Dump each function after dead code elimination. The file name is made by
7103 appending @file{.dce} to the source file name.
7106 @opindex fdump-tree-sra
7107 Dump each function after performing scalar replacement of aggregates. The
7108 file name is made by appending @file{.sra} to the source file name.
7111 @opindex fdump-tree-sink
7112 Dump each function after performing code sinking. The file name is made
7113 by appending @file{.sink} to the source file name.
7116 @opindex fdump-tree-dom
7117 Dump each function after applying dominator tree optimizations. The file
7118 name is made by appending @file{.dom} to the source file name.
7121 @opindex fdump-tree-dse
7122 Dump each function after applying dead store elimination. The file
7123 name is made by appending @file{.dse} to the source file name.
7126 @opindex fdump-tree-phiopt
7127 Dump each function after optimizing PHI nodes into straightline code. The file
7128 name is made by appending @file{.phiopt} to the source file name.
7131 @opindex fdump-tree-forwprop
7132 Dump each function after forward propagating single use variables. The file
7133 name is made by appending @file{.forwprop} to the source file name.
7136 @opindex fdump-tree-copyrename
7137 Dump each function after applying the copy rename optimization. The file
7138 name is made by appending @file{.copyrename} to the source file name.
7141 @opindex fdump-tree-nrv
7142 Dump each function after applying the named return value optimization on
7143 generic trees. The file name is made by appending @file{.nrv} to the source
7147 @opindex fdump-tree-vect
7148 Dump each function after applying vectorization of loops. The file name is
7149 made by appending @file{.vect} to the source file name.
7152 @opindex fdump-tree-slp
7153 Dump each function after applying vectorization of basic blocks. The file name
7154 is made by appending @file{.slp} to the source file name.
7157 @opindex fdump-tree-vrp
7158 Dump each function after Value Range Propagation (VRP). The file name
7159 is made by appending @file{.vrp} to the source file name.
7162 @opindex fdump-tree-all
7163 Enable all the available tree dumps with the flags provided in this option.
7167 @itemx -fopt-info-@var{options}
7168 @itemx -fopt-info-@var{options}=@var{filename}
7170 Controls optimization dumps from various optimization passes. If the
7171 @samp{-@var{options}} form is used, @var{options} is a list of
7172 @samp{-} separated option keywords to select the dump details and
7175 The @var{options} can be divided into two groups: options describing the
7176 verbosity of the dump, and options describing which optimizations
7177 should be included. The options from both the groups can be freely
7178 mixed as they are non-overlapping. However, in case of any conflicts,
7179 the later options override the earlier options on the command
7182 The following options control the dump verbosity:
7186 Print information when an optimization is successfully applied. It is
7187 up to a pass to decide which information is relevant. For example, the
7188 vectorizer passes print the source location of loops which are
7189 successfully vectorized.
7191 Print information about missed optimizations. Individual passes
7192 control which information to include in the output.
7194 Print verbose information about optimizations, such as certain
7195 transformations, more detailed messages about decisions etc.
7197 Print detailed optimization information. This includes
7198 @samp{optimized}, @samp{missed}, and @samp{note}.
7201 One or more of the following option keywords can be used to describe a
7202 group of optimizations:
7206 Enable dumps from all interprocedural optimizations.
7208 Enable dumps from all loop optimizations.
7210 Enable dumps from all inlining optimizations.
7212 Enable dumps from all vectorization optimizations.
7214 Enable dumps from all optimizations. This is a superset of
7215 the optimization groups listed above.
7219 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7220 info about successful optimizations from all the passes.
7222 If the @var{filename} is provided, then the dumps from all the
7223 applicable optimizations are concatenated into the @var{filename}.
7224 Otherwise the dump is output onto @file{stderr}. Though multiple
7225 @option{-fopt-info} options are accepted, only one of them can include
7226 a @var{filename}. If other filenames are provided then all but the
7227 first such option are ignored.
7229 Note that the output @var{filename} is overwritten
7230 in case of multiple translation units. If a combined output from
7231 multiple translation units is desired, @file{stderr} should be used
7234 In the following example, the optimization info is output to
7243 gcc -O3 -fopt-info-missed=missed.all
7247 outputs missed optimization report from all the passes into
7248 @file{missed.all}, and this one:
7251 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7255 prints information about missed optimization opportunities from
7256 vectorization passes on @file{stderr}.
7257 Note that @option{-fopt-info-vec-missed} is equivalent to
7258 @option{-fopt-info-missed-vec}.
7262 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7266 outputs information about missed optimizations as well as
7267 optimized locations from all the inlining passes into
7273 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7277 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7278 in conflict since only one output file is allowed. In this case, only
7279 the first option takes effect and the subsequent options are
7280 ignored. Thus only @file{vec.miss} is produced which contains
7281 dumps from the vectorizer about missed opportunities.
7283 @item -frandom-seed=@var{number}
7284 @opindex frandom-seed
7285 This option provides a seed that GCC uses in place of
7286 random numbers in generating certain symbol names
7287 that have to be different in every compiled file. It is also used to
7288 place unique stamps in coverage data files and the object files that
7289 produce them. You can use the @option{-frandom-seed} option to produce
7290 reproducibly identical object files.
7292 The @var{number} should be different for every file you compile.
7294 @item -fsched-verbose=@var{n}
7295 @opindex fsched-verbose
7296 On targets that use instruction scheduling, this option controls the
7297 amount of debugging output the scheduler prints. This information is
7298 written to standard error, unless @option{-fdump-rtl-sched1} or
7299 @option{-fdump-rtl-sched2} is specified, in which case it is output
7300 to the usual dump listing file, @file{.sched1} or @file{.sched2}
7301 respectively. However for @var{n} greater than nine, the output is
7302 always printed to standard error.
7304 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7305 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7306 For @var{n} greater than one, it also output basic block probabilities,
7307 detailed ready list information and unit/insn info. For @var{n} greater
7308 than two, it includes RTL at abort point, control-flow and regions info.
7309 And for @var{n} over four, @option{-fsched-verbose} also includes
7313 @itemx -save-temps=cwd
7315 Store the usual ``temporary'' intermediate files permanently; place them
7316 in the current directory and name them based on the source file. Thus,
7317 compiling @file{foo.c} with @option{-c -save-temps} produces files
7318 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7319 preprocessed @file{foo.i} output file even though the compiler now
7320 normally uses an integrated preprocessor.
7322 When used in combination with the @option{-x} command-line option,
7323 @option{-save-temps} is sensible enough to avoid over writing an
7324 input source file with the same extension as an intermediate file.
7325 The corresponding intermediate file may be obtained by renaming the
7326 source file before using @option{-save-temps}.
7328 If you invoke GCC in parallel, compiling several different source
7329 files that share a common base name in different subdirectories or the
7330 same source file compiled for multiple output destinations, it is
7331 likely that the different parallel compilers will interfere with each
7332 other, and overwrite the temporary files. For instance:
7335 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7336 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7339 may result in @file{foo.i} and @file{foo.o} being written to
7340 simultaneously by both compilers.
7342 @item -save-temps=obj
7343 @opindex save-temps=obj
7344 Store the usual ``temporary'' intermediate files permanently. If the
7345 @option{-o} option is used, the temporary files are based on the
7346 object file. If the @option{-o} option is not used, the
7347 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7352 gcc -save-temps=obj -c foo.c
7353 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7354 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7358 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7359 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7360 @file{dir2/yfoobar.o}.
7362 @item -time@r{[}=@var{file}@r{]}
7364 Report the CPU time taken by each subprocess in the compilation
7365 sequence. For C source files, this is the compiler proper and assembler
7366 (plus the linker if linking is done).
7368 Without the specification of an output file, the output looks like this:
7375 The first number on each line is the ``user time'', that is time spent
7376 executing the program itself. The second number is ``system time'',
7377 time spent executing operating system routines on behalf of the program.
7378 Both numbers are in seconds.
7380 With the specification of an output file, the output is appended to the
7381 named file, and it looks like this:
7384 0.12 0.01 cc1 @var{options}
7385 0.00 0.01 as @var{options}
7388 The ``user time'' and the ``system time'' are moved before the program
7389 name, and the options passed to the program are displayed, so that one
7390 can later tell what file was being compiled, and with which options.
7392 @item -fvar-tracking
7393 @opindex fvar-tracking
7394 Run variable tracking pass. It computes where variables are stored at each
7395 position in code. Better debugging information is then generated
7396 (if the debugging information format supports this information).
7398 It is enabled by default when compiling with optimization (@option{-Os},
7399 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7400 the debug info format supports it.
7402 @item -fvar-tracking-assignments
7403 @opindex fvar-tracking-assignments
7404 @opindex fno-var-tracking-assignments
7405 Annotate assignments to user variables early in the compilation and
7406 attempt to carry the annotations over throughout the compilation all the
7407 way to the end, in an attempt to improve debug information while
7408 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7410 It can be enabled even if var-tracking is disabled, in which case
7411 annotations are created and maintained, but discarded at the end.
7412 By default, this flag is enabled together with @option{-fvar-tracking},
7413 except when selective scheduling is enabled.
7415 @item -fvar-tracking-assignments-toggle
7416 @opindex fvar-tracking-assignments-toggle
7417 @opindex fno-var-tracking-assignments-toggle
7418 Toggle @option{-fvar-tracking-assignments}, in the same way that
7419 @option{-gtoggle} toggles @option{-g}.
7421 @item -print-file-name=@var{library}
7422 @opindex print-file-name
7423 Print the full absolute name of the library file @var{library} that
7424 would be used when linking---and don't do anything else. With this
7425 option, GCC does not compile or link anything; it just prints the
7428 @item -print-multi-directory
7429 @opindex print-multi-directory
7430 Print the directory name corresponding to the multilib selected by any
7431 other switches present in the command line. This directory is supposed
7432 to exist in @env{GCC_EXEC_PREFIX}.
7434 @item -print-multi-lib
7435 @opindex print-multi-lib
7436 Print the mapping from multilib directory names to compiler switches
7437 that enable them. The directory name is separated from the switches by
7438 @samp{;}, and each switch starts with an @samp{@@} instead of the
7439 @samp{-}, without spaces between multiple switches. This is supposed to
7440 ease shell processing.
7442 @item -print-multi-os-directory
7443 @opindex print-multi-os-directory
7444 Print the path to OS libraries for the selected
7445 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7446 present in the @file{lib} subdirectory and no multilibs are used, this is
7447 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7448 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7449 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7450 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7452 @item -print-multiarch
7453 @opindex print-multiarch
7454 Print the path to OS libraries for the selected multiarch,
7455 relative to some @file{lib} subdirectory.
7457 @item -print-prog-name=@var{program}
7458 @opindex print-prog-name
7459 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7461 @item -print-libgcc-file-name
7462 @opindex print-libgcc-file-name
7463 Same as @option{-print-file-name=libgcc.a}.
7465 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7466 but you do want to link with @file{libgcc.a}. You can do:
7469 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7472 @item -print-search-dirs
7473 @opindex print-search-dirs
7474 Print the name of the configured installation directory and a list of
7475 program and library directories @command{gcc} searches---and don't do anything else.
7477 This is useful when @command{gcc} prints the error message
7478 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7479 To resolve this you either need to put @file{cpp0} and the other compiler
7480 components where @command{gcc} expects to find them, or you can set the environment
7481 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7482 Don't forget the trailing @samp{/}.
7483 @xref{Environment Variables}.
7485 @item -print-sysroot
7486 @opindex print-sysroot
7487 Print the target sysroot directory that is used during
7488 compilation. This is the target sysroot specified either at configure
7489 time or using the @option{--sysroot} option, possibly with an extra
7490 suffix that depends on compilation options. If no target sysroot is
7491 specified, the option prints nothing.
7493 @item -print-sysroot-headers-suffix
7494 @opindex print-sysroot-headers-suffix
7495 Print the suffix added to the target sysroot when searching for
7496 headers, or give an error if the compiler is not configured with such
7497 a suffix---and don't do anything else.
7500 @opindex dumpmachine
7501 Print the compiler's target machine (for example,
7502 @samp{i686-pc-linux-gnu})---and don't do anything else.
7505 @opindex dumpversion
7506 Print the compiler version (for example, @code{3.0})---and don't do
7511 Print the compiler's built-in specs---and don't do anything else. (This
7512 is used when GCC itself is being built.) @xref{Spec Files}.
7514 @item -fno-eliminate-unused-debug-types
7515 @opindex feliminate-unused-debug-types
7516 @opindex fno-eliminate-unused-debug-types
7517 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7518 output for types that are nowhere used in the source file being compiled.
7519 Sometimes it is useful to have GCC emit debugging
7520 information for all types declared in a compilation
7521 unit, regardless of whether or not they are actually used
7522 in that compilation unit, for example
7523 if, in the debugger, you want to cast a value to a type that is
7524 not actually used in your program (but is declared). More often,
7525 however, this results in a significant amount of wasted space.
7528 @node Optimize Options
7529 @section Options That Control Optimization
7530 @cindex optimize options
7531 @cindex options, optimization
7533 These options control various sorts of optimizations.
7535 Without any optimization option, the compiler's goal is to reduce the
7536 cost of compilation and to make debugging produce the expected
7537 results. Statements are independent: if you stop the program with a
7538 breakpoint between statements, you can then assign a new value to any
7539 variable or change the program counter to any other statement in the
7540 function and get exactly the results you expect from the source
7543 Turning on optimization flags makes the compiler attempt to improve
7544 the performance and/or code size at the expense of compilation time
7545 and possibly the ability to debug the program.
7547 The compiler performs optimization based on the knowledge it has of the
7548 program. Compiling multiple files at once to a single output file mode allows
7549 the compiler to use information gained from all of the files when compiling
7552 Not all optimizations are controlled directly by a flag. Only
7553 optimizations that have a flag are listed in this section.
7555 Most optimizations are only enabled if an @option{-O} level is set on
7556 the command line. Otherwise they are disabled, even if individual
7557 optimization flags are specified.
7559 Depending on the target and how GCC was configured, a slightly different
7560 set of optimizations may be enabled at each @option{-O} level than
7561 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7562 to find out the exact set of optimizations that are enabled at each level.
7563 @xref{Overall Options}, for examples.
7570 Optimize. Optimizing compilation takes somewhat more time, and a lot
7571 more memory for a large function.
7573 With @option{-O}, the compiler tries to reduce code size and execution
7574 time, without performing any optimizations that take a great deal of
7577 @option{-O} turns on the following optimization flags:
7580 -fbranch-count-reg @gol
7581 -fcombine-stack-adjustments @gol
7583 -fcprop-registers @gol
7586 -fdelayed-branch @gol
7588 -fforward-propagate @gol
7589 -fguess-branch-probability @gol
7590 -fif-conversion2 @gol
7591 -fif-conversion @gol
7592 -finline-functions-called-once @gol
7593 -fipa-pure-const @gol
7595 -fipa-reference @gol
7596 -fmerge-constants @gol
7597 -fmove-loop-invariants @gol
7599 -fsplit-wide-types @gol
7604 -ftree-copy-prop @gol
7605 -ftree-copyrename @gol
7607 -ftree-dominator-opts @gol
7609 -ftree-forwprop @gol
7619 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7620 where doing so does not interfere with debugging.
7624 Optimize even more. GCC performs nearly all supported optimizations
7625 that do not involve a space-speed tradeoff.
7626 As compared to @option{-O}, this option increases both compilation time
7627 and the performance of the generated code.
7629 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7630 also turns on the following optimization flags:
7631 @gccoptlist{-fthread-jumps @gol
7632 -falign-functions -falign-jumps @gol
7633 -falign-loops -falign-labels @gol
7636 -fcse-follow-jumps -fcse-skip-blocks @gol
7637 -fdelete-null-pointer-checks @gol
7638 -fdevirtualize -fdevirtualize-speculatively @gol
7639 -fexpensive-optimizations @gol
7640 -fgcse -fgcse-lm @gol
7641 -fhoist-adjacent-loads @gol
7642 -finline-small-functions @gol
7643 -findirect-inlining @gol
7645 -fipa-cp-alignment @gol
7648 -fisolate-erroneous-paths-dereference @gol
7650 -foptimize-sibling-calls @gol
7651 -foptimize-strlen @gol
7652 -fpartial-inlining @gol
7654 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
7655 -frerun-cse-after-loop @gol
7656 -fsched-interblock -fsched-spec @gol
7657 -fschedule-insns -fschedule-insns2 @gol
7658 -fstrict-aliasing -fstrict-overflow @gol
7659 -ftree-builtin-call-dce @gol
7660 -ftree-switch-conversion -ftree-tail-merge @gol
7665 Please note the warning under @option{-fgcse} about
7666 invoking @option{-O2} on programs that use computed gotos.
7670 Optimize yet more. @option{-O3} turns on all optimizations specified
7671 by @option{-O2} and also turns on the @option{-finline-functions},
7672 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7673 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7674 @option{-ftree-loop-distribute-patterns},
7675 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7676 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7680 Reduce compilation time and make debugging produce the expected
7681 results. This is the default.
7685 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7686 do not typically increase code size. It also performs further
7687 optimizations designed to reduce code size.
7689 @option{-Os} disables the following optimization flags:
7690 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7691 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
7692 -fprefetch-loop-arrays}
7696 Disregard strict standards compliance. @option{-Ofast} enables all
7697 @option{-O3} optimizations. It also enables optimizations that are not
7698 valid for all standard-compliant programs.
7699 It turns on @option{-ffast-math} and the Fortran-specific
7700 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7704 Optimize debugging experience. @option{-Og} enables optimizations
7705 that do not interfere with debugging. It should be the optimization
7706 level of choice for the standard edit-compile-debug cycle, offering
7707 a reasonable level of optimization while maintaining fast compilation
7708 and a good debugging experience.
7710 If you use multiple @option{-O} options, with or without level numbers,
7711 the last such option is the one that is effective.
7714 Options of the form @option{-f@var{flag}} specify machine-independent
7715 flags. Most flags have both positive and negative forms; the negative
7716 form of @option{-ffoo} is @option{-fno-foo}. In the table
7717 below, only one of the forms is listed---the one you typically
7718 use. You can figure out the other form by either removing @samp{no-}
7721 The following options control specific optimizations. They are either
7722 activated by @option{-O} options or are related to ones that are. You
7723 can use the following flags in the rare cases when ``fine-tuning'' of
7724 optimizations to be performed is desired.
7727 @item -fno-defer-pop
7728 @opindex fno-defer-pop
7729 Always pop the arguments to each function call as soon as that function
7730 returns. For machines that must pop arguments after a function call,
7731 the compiler normally lets arguments accumulate on the stack for several
7732 function calls and pops them all at once.
7734 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7736 @item -fforward-propagate
7737 @opindex fforward-propagate
7738 Perform a forward propagation pass on RTL@. The pass tries to combine two
7739 instructions and checks if the result can be simplified. If loop unrolling
7740 is active, two passes are performed and the second is scheduled after
7743 This option is enabled by default at optimization levels @option{-O},
7744 @option{-O2}, @option{-O3}, @option{-Os}.
7746 @item -ffp-contract=@var{style}
7747 @opindex ffp-contract
7748 @option{-ffp-contract=off} disables floating-point expression contraction.
7749 @option{-ffp-contract=fast} enables floating-point expression contraction
7750 such as forming of fused multiply-add operations if the target has
7751 native support for them.
7752 @option{-ffp-contract=on} enables floating-point expression contraction
7753 if allowed by the language standard. This is currently not implemented
7754 and treated equal to @option{-ffp-contract=off}.
7756 The default is @option{-ffp-contract=fast}.
7758 @item -fomit-frame-pointer
7759 @opindex fomit-frame-pointer
7760 Don't keep the frame pointer in a register for functions that
7761 don't need one. This avoids the instructions to save, set up and
7762 restore frame pointers; it also makes an extra register available
7763 in many functions. @strong{It also makes debugging impossible on
7766 On some machines, such as the VAX, this flag has no effect, because
7767 the standard calling sequence automatically handles the frame pointer
7768 and nothing is saved by pretending it doesn't exist. The
7769 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7770 whether a target machine supports this flag. @xref{Registers,,Register
7771 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7773 The default setting (when not optimizing for
7774 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7775 @option{-fomit-frame-pointer}. You can configure GCC with the
7776 @option{--enable-frame-pointer} configure option to change the default.
7778 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7780 @item -foptimize-sibling-calls
7781 @opindex foptimize-sibling-calls
7782 Optimize sibling and tail recursive calls.
7784 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7786 @item -foptimize-strlen
7787 @opindex foptimize-strlen
7788 Optimize various standard C string functions (e.g. @code{strlen},
7789 @code{strchr} or @code{strcpy}) and
7790 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7792 Enabled at levels @option{-O2}, @option{-O3}.
7796 Do not expand any functions inline apart from those marked with
7797 the @code{always_inline} attribute. This is the default when not
7800 Single functions can be exempted from inlining by marking them
7801 with the @code{noinline} attribute.
7803 @item -finline-small-functions
7804 @opindex finline-small-functions
7805 Integrate functions into their callers when their body is smaller than expected
7806 function call code (so overall size of program gets smaller). The compiler
7807 heuristically decides which functions are simple enough to be worth integrating
7808 in this way. This inlining applies to all functions, even those not declared
7811 Enabled at level @option{-O2}.
7813 @item -findirect-inlining
7814 @opindex findirect-inlining
7815 Inline also indirect calls that are discovered to be known at compile
7816 time thanks to previous inlining. This option has any effect only
7817 when inlining itself is turned on by the @option{-finline-functions}
7818 or @option{-finline-small-functions} options.
7820 Enabled at level @option{-O2}.
7822 @item -finline-functions
7823 @opindex finline-functions
7824 Consider all functions for inlining, even if they are not declared inline.
7825 The compiler heuristically decides which functions are worth integrating
7828 If all calls to a given function are integrated, and the function is
7829 declared @code{static}, then the function is normally not output as
7830 assembler code in its own right.
7832 Enabled at level @option{-O3}.
7834 @item -finline-functions-called-once
7835 @opindex finline-functions-called-once
7836 Consider all @code{static} functions called once for inlining into their
7837 caller even if they are not marked @code{inline}. If a call to a given
7838 function is integrated, then the function is not output as assembler code
7841 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
7843 @item -fearly-inlining
7844 @opindex fearly-inlining
7845 Inline functions marked by @code{always_inline} and functions whose body seems
7846 smaller than the function call overhead early before doing
7847 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
7848 makes profiling significantly cheaper and usually inlining faster on programs
7849 having large chains of nested wrapper functions.
7855 Perform interprocedural scalar replacement of aggregates, removal of
7856 unused parameters and replacement of parameters passed by reference
7857 by parameters passed by value.
7859 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
7861 @item -finline-limit=@var{n}
7862 @opindex finline-limit
7863 By default, GCC limits the size of functions that can be inlined. This flag
7864 allows coarse control of this limit. @var{n} is the size of functions that
7865 can be inlined in number of pseudo instructions.
7867 Inlining is actually controlled by a number of parameters, which may be
7868 specified individually by using @option{--param @var{name}=@var{value}}.
7869 The @option{-finline-limit=@var{n}} option sets some of these parameters
7873 @item max-inline-insns-single
7874 is set to @var{n}/2.
7875 @item max-inline-insns-auto
7876 is set to @var{n}/2.
7879 See below for a documentation of the individual
7880 parameters controlling inlining and for the defaults of these parameters.
7882 @emph{Note:} there may be no value to @option{-finline-limit} that results
7883 in default behavior.
7885 @emph{Note:} pseudo instruction represents, in this particular context, an
7886 abstract measurement of function's size. In no way does it represent a count
7887 of assembly instructions and as such its exact meaning might change from one
7888 release to an another.
7890 @item -fno-keep-inline-dllexport
7891 @opindex fno-keep-inline-dllexport
7892 This is a more fine-grained version of @option{-fkeep-inline-functions},
7893 which applies only to functions that are declared using the @code{dllexport}
7894 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
7897 @item -fkeep-inline-functions
7898 @opindex fkeep-inline-functions
7899 In C, emit @code{static} functions that are declared @code{inline}
7900 into the object file, even if the function has been inlined into all
7901 of its callers. This switch does not affect functions using the
7902 @code{extern inline} extension in GNU C90@. In C++, emit any and all
7903 inline functions into the object file.
7905 @item -fkeep-static-consts
7906 @opindex fkeep-static-consts
7907 Emit variables declared @code{static const} when optimization isn't turned
7908 on, even if the variables aren't referenced.
7910 GCC enables this option by default. If you want to force the compiler to
7911 check if a variable is referenced, regardless of whether or not
7912 optimization is turned on, use the @option{-fno-keep-static-consts} option.
7914 @item -fmerge-constants
7915 @opindex fmerge-constants
7916 Attempt to merge identical constants (string constants and floating-point
7917 constants) across compilation units.
7919 This option is the default for optimized compilation if the assembler and
7920 linker support it. Use @option{-fno-merge-constants} to inhibit this
7923 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7925 @item -fmerge-all-constants
7926 @opindex fmerge-all-constants
7927 Attempt to merge identical constants and identical variables.
7929 This option implies @option{-fmerge-constants}. In addition to
7930 @option{-fmerge-constants} this considers e.g.@: even constant initialized
7931 arrays or initialized constant variables with integral or floating-point
7932 types. Languages like C or C++ require each variable, including multiple
7933 instances of the same variable in recursive calls, to have distinct locations,
7934 so using this option results in non-conforming
7937 @item -fmodulo-sched
7938 @opindex fmodulo-sched
7939 Perform swing modulo scheduling immediately before the first scheduling
7940 pass. This pass looks at innermost loops and reorders their
7941 instructions by overlapping different iterations.
7943 @item -fmodulo-sched-allow-regmoves
7944 @opindex fmodulo-sched-allow-regmoves
7945 Perform more aggressive SMS-based modulo scheduling with register moves
7946 allowed. By setting this flag certain anti-dependences edges are
7947 deleted, which triggers the generation of reg-moves based on the
7948 life-range analysis. This option is effective only with
7949 @option{-fmodulo-sched} enabled.
7951 @item -fno-branch-count-reg
7952 @opindex fno-branch-count-reg
7953 Do not use ``decrement and branch'' instructions on a count register,
7954 but instead generate a sequence of instructions that decrement a
7955 register, compare it against zero, then branch based upon the result.
7956 This option is only meaningful on architectures that support such
7957 instructions, which include x86, PowerPC, IA-64 and S/390.
7959 Enabled by default at @option{-O1} and higher.
7961 The default is @option{-fbranch-count-reg}.
7963 @item -fno-function-cse
7964 @opindex fno-function-cse
7965 Do not put function addresses in registers; make each instruction that
7966 calls a constant function contain the function's address explicitly.
7968 This option results in less efficient code, but some strange hacks
7969 that alter the assembler output may be confused by the optimizations
7970 performed when this option is not used.
7972 The default is @option{-ffunction-cse}
7974 @item -fno-zero-initialized-in-bss
7975 @opindex fno-zero-initialized-in-bss
7976 If the target supports a BSS section, GCC by default puts variables that
7977 are initialized to zero into BSS@. This can save space in the resulting
7980 This option turns off this behavior because some programs explicitly
7981 rely on variables going to the data section---e.g., so that the
7982 resulting executable can find the beginning of that section and/or make
7983 assumptions based on that.
7985 The default is @option{-fzero-initialized-in-bss}.
7987 @item -fthread-jumps
7988 @opindex fthread-jumps
7989 Perform optimizations that check to see if a jump branches to a
7990 location where another comparison subsumed by the first is found. If
7991 so, the first branch is redirected to either the destination of the
7992 second branch or a point immediately following it, depending on whether
7993 the condition is known to be true or false.
7995 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7997 @item -fsplit-wide-types
7998 @opindex fsplit-wide-types
7999 When using a type that occupies multiple registers, such as @code{long
8000 long} on a 32-bit system, split the registers apart and allocate them
8001 independently. This normally generates better code for those types,
8002 but may make debugging more difficult.
8004 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8007 @item -fcse-follow-jumps
8008 @opindex fcse-follow-jumps
8009 In common subexpression elimination (CSE), scan through jump instructions
8010 when the target of the jump is not reached by any other path. For
8011 example, when CSE encounters an @code{if} statement with an
8012 @code{else} clause, CSE follows the jump when the condition
8015 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8017 @item -fcse-skip-blocks
8018 @opindex fcse-skip-blocks
8019 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8020 follow jumps that conditionally skip over blocks. When CSE
8021 encounters a simple @code{if} statement with no else clause,
8022 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8023 body of the @code{if}.
8025 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8027 @item -frerun-cse-after-loop
8028 @opindex frerun-cse-after-loop
8029 Re-run common subexpression elimination after loop optimizations are
8032 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8036 Perform a global common subexpression elimination pass.
8037 This pass also performs global constant and copy propagation.
8039 @emph{Note:} When compiling a program using computed gotos, a GCC
8040 extension, you may get better run-time performance if you disable
8041 the global common subexpression elimination pass by adding
8042 @option{-fno-gcse} to the command line.
8044 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8048 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8049 attempts to move loads that are only killed by stores into themselves. This
8050 allows a loop containing a load/store sequence to be changed to a load outside
8051 the loop, and a copy/store within the loop.
8053 Enabled by default when @option{-fgcse} is enabled.
8057 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8058 global common subexpression elimination. This pass attempts to move
8059 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8060 loops containing a load/store sequence can be changed to a load before
8061 the loop and a store after the loop.
8063 Not enabled at any optimization level.
8067 When @option{-fgcse-las} is enabled, the global common subexpression
8068 elimination pass eliminates redundant loads that come after stores to the
8069 same memory location (both partial and full redundancies).
8071 Not enabled at any optimization level.
8073 @item -fgcse-after-reload
8074 @opindex fgcse-after-reload
8075 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8076 pass is performed after reload. The purpose of this pass is to clean up
8079 @item -faggressive-loop-optimizations
8080 @opindex faggressive-loop-optimizations
8081 This option tells the loop optimizer to use language constraints to
8082 derive bounds for the number of iterations of a loop. This assumes that
8083 loop code does not invoke undefined behavior by for example causing signed
8084 integer overflows or out-of-bound array accesses. The bounds for the
8085 number of iterations of a loop are used to guide loop unrolling and peeling
8086 and loop exit test optimizations.
8087 This option is enabled by default.
8089 @item -funsafe-loop-optimizations
8090 @opindex funsafe-loop-optimizations
8091 This option tells the loop optimizer to assume that loop indices do not
8092 overflow, and that loops with nontrivial exit condition are not
8093 infinite. This enables a wider range of loop optimizations even if
8094 the loop optimizer itself cannot prove that these assumptions are valid.
8095 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
8096 if it finds this kind of loop.
8098 @item -fcrossjumping
8099 @opindex fcrossjumping
8100 Perform cross-jumping transformation.
8101 This transformation unifies equivalent code and saves code size. The
8102 resulting code may or may not perform better than without cross-jumping.
8104 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8106 @item -fauto-inc-dec
8107 @opindex fauto-inc-dec
8108 Combine increments or decrements of addresses with memory accesses.
8109 This pass is always skipped on architectures that do not have
8110 instructions to support this. Enabled by default at @option{-O} and
8111 higher on architectures that support this.
8115 Perform dead code elimination (DCE) on RTL@.
8116 Enabled by default at @option{-O} and higher.
8120 Perform dead store elimination (DSE) on RTL@.
8121 Enabled by default at @option{-O} and higher.
8123 @item -fif-conversion
8124 @opindex fif-conversion
8125 Attempt to transform conditional jumps into branch-less equivalents. This
8126 includes use of conditional moves, min, max, set flags and abs instructions, and
8127 some tricks doable by standard arithmetics. The use of conditional execution
8128 on chips where it is available is controlled by @option{-fif-conversion2}.
8130 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8132 @item -fif-conversion2
8133 @opindex fif-conversion2
8134 Use conditional execution (where available) to transform conditional jumps into
8135 branch-less equivalents.
8137 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8139 @item -fdeclone-ctor-dtor
8140 @opindex fdeclone-ctor-dtor
8141 The C++ ABI requires multiple entry points for constructors and
8142 destructors: one for a base subobject, one for a complete object, and
8143 one for a virtual destructor that calls operator delete afterwards.
8144 For a hierarchy with virtual bases, the base and complete variants are
8145 clones, which means two copies of the function. With this option, the
8146 base and complete variants are changed to be thunks that call a common
8149 Enabled by @option{-Os}.
8151 @item -fdelete-null-pointer-checks
8152 @opindex fdelete-null-pointer-checks
8153 Assume that programs cannot safely dereference null pointers, and that
8154 no code or data element resides at address zero.
8155 This option enables simple constant
8156 folding optimizations at all optimization levels. In addition, other
8157 optimization passes in GCC use this flag to control global dataflow
8158 analyses that eliminate useless checks for null pointers; these assume
8159 that a memory access to address zero always results in a trap, so
8160 that if a pointer is checked after it has already been dereferenced,
8163 Note however that in some environments this assumption is not true.
8164 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8165 for programs that depend on that behavior.
8167 This option is enabled by default on most targets. On Nios II ELF, it
8168 defaults to off. On AVR and CR16, this option is completely disabled.
8170 Passes that use the dataflow information
8171 are enabled independently at different optimization levels.
8173 @item -fdevirtualize
8174 @opindex fdevirtualize
8175 Attempt to convert calls to virtual functions to direct calls. This
8176 is done both within a procedure and interprocedurally as part of
8177 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8178 propagation (@option{-fipa-cp}).
8179 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8181 @item -fdevirtualize-speculatively
8182 @opindex fdevirtualize-speculatively
8183 Attempt to convert calls to virtual functions to speculative direct calls.
8184 Based on the analysis of the type inheritance graph, determine for a given call
8185 the set of likely targets. If the set is small, preferably of size 1, change
8186 the call into a conditional deciding between direct and indirect calls. The
8187 speculative calls enable more optimizations, such as inlining. When they seem
8188 useless after further optimization, they are converted back into original form.
8190 @item -fdevirtualize-at-ltrans
8191 @opindex fdevirtualize-at-ltrans
8192 Stream extra information needed for aggressive devirtualization when running
8193 the link-time optimizer in local transformation mode.
8194 This option enables more devirtualization but
8195 significantly increases the size of streamed data. For this reason it is
8196 disabled by default.
8198 @item -fexpensive-optimizations
8199 @opindex fexpensive-optimizations
8200 Perform a number of minor optimizations that are relatively expensive.
8202 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8206 Attempt to remove redundant extension instructions. This is especially
8207 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8208 registers after writing to their lower 32-bit half.
8210 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8211 @option{-O3}, @option{-Os}.
8213 @item -fno-lifetime-dse
8214 @opindex fno-lifetime-dse
8215 In C++ the value of an object is only affected by changes within its
8216 lifetime: when the constructor begins, the object has an indeterminate
8217 value, and any changes during the lifetime of the object are dead when
8218 the object is destroyed. Normally dead store elimination will take
8219 advantage of this; if your code relies on the value of the object
8220 storage persisting beyond the lifetime of the object, you can use this
8221 flag to disable this optimization.
8223 @item -flive-range-shrinkage
8224 @opindex flive-range-shrinkage
8225 Attempt to decrease register pressure through register live range
8226 shrinkage. This is helpful for fast processors with small or moderate
8229 @item -fira-algorithm=@var{algorithm}
8230 @opindex fira-algorithm
8231 Use the specified coloring algorithm for the integrated register
8232 allocator. The @var{algorithm} argument can be @samp{priority}, which
8233 specifies Chow's priority coloring, or @samp{CB}, which specifies
8234 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8235 for all architectures, but for those targets that do support it, it is
8236 the default because it generates better code.
8238 @item -fira-region=@var{region}
8239 @opindex fira-region
8240 Use specified regions for the integrated register allocator. The
8241 @var{region} argument should be one of the following:
8246 Use all loops as register allocation regions.
8247 This can give the best results for machines with a small and/or
8248 irregular register set.
8251 Use all loops except for loops with small register pressure
8252 as the regions. This value usually gives
8253 the best results in most cases and for most architectures,
8254 and is enabled by default when compiling with optimization for speed
8255 (@option{-O}, @option{-O2}, @dots{}).
8258 Use all functions as a single region.
8259 This typically results in the smallest code size, and is enabled by default for
8260 @option{-Os} or @option{-O0}.
8264 @item -fira-hoist-pressure
8265 @opindex fira-hoist-pressure
8266 Use IRA to evaluate register pressure in the code hoisting pass for
8267 decisions to hoist expressions. This option usually results in smaller
8268 code, but it can slow the compiler down.
8270 This option is enabled at level @option{-Os} for all targets.
8272 @item -fira-loop-pressure
8273 @opindex fira-loop-pressure
8274 Use IRA to evaluate register pressure in loops for decisions to move
8275 loop invariants. This option usually results in generation
8276 of faster and smaller code on machines with large register files (>= 32
8277 registers), but it can slow the compiler down.
8279 This option is enabled at level @option{-O3} for some targets.
8281 @item -fno-ira-share-save-slots
8282 @opindex fno-ira-share-save-slots
8283 Disable sharing of stack slots used for saving call-used hard
8284 registers living through a call. Each hard register gets a
8285 separate stack slot, and as a result function stack frames are
8288 @item -fno-ira-share-spill-slots
8289 @opindex fno-ira-share-spill-slots
8290 Disable sharing of stack slots allocated for pseudo-registers. Each
8291 pseudo-register that does not get a hard register gets a separate
8292 stack slot, and as a result function stack frames are larger.
8294 @item -fira-verbose=@var{n}
8295 @opindex fira-verbose
8296 Control the verbosity of the dump file for the integrated register allocator.
8297 The default value is 5. If the value @var{n} is greater or equal to 10,
8298 the dump output is sent to stderr using the same format as @var{n} minus 10.
8302 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8303 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8304 values if it is profitable.
8306 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8308 @item -fdelayed-branch
8309 @opindex fdelayed-branch
8310 If supported for the target machine, attempt to reorder instructions
8311 to exploit instruction slots available after delayed branch
8314 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8316 @item -fschedule-insns
8317 @opindex fschedule-insns
8318 If supported for the target machine, attempt to reorder instructions to
8319 eliminate execution stalls due to required data being unavailable. This
8320 helps machines that have slow floating point or memory load instructions
8321 by allowing other instructions to be issued until the result of the load
8322 or floating-point instruction is required.
8324 Enabled at levels @option{-O2}, @option{-O3}.
8326 @item -fschedule-insns2
8327 @opindex fschedule-insns2
8328 Similar to @option{-fschedule-insns}, but requests an additional pass of
8329 instruction scheduling after register allocation has been done. This is
8330 especially useful on machines with a relatively small number of
8331 registers and where memory load instructions take more than one cycle.
8333 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8335 @item -fno-sched-interblock
8336 @opindex fno-sched-interblock
8337 Don't schedule instructions across basic blocks. This is normally
8338 enabled by default when scheduling before register allocation, i.e.@:
8339 with @option{-fschedule-insns} or at @option{-O2} or higher.
8341 @item -fno-sched-spec
8342 @opindex fno-sched-spec
8343 Don't allow speculative motion of non-load instructions. This is normally
8344 enabled by default when scheduling before register allocation, i.e.@:
8345 with @option{-fschedule-insns} or at @option{-O2} or higher.
8347 @item -fsched-pressure
8348 @opindex fsched-pressure
8349 Enable register pressure sensitive insn scheduling before register
8350 allocation. This only makes sense when scheduling before register
8351 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8352 @option{-O2} or higher. Usage of this option can improve the
8353 generated code and decrease its size by preventing register pressure
8354 increase above the number of available hard registers and subsequent
8355 spills in register allocation.
8357 @item -fsched-spec-load
8358 @opindex fsched-spec-load
8359 Allow speculative motion of some load instructions. This only makes
8360 sense when scheduling before register allocation, i.e.@: with
8361 @option{-fschedule-insns} or at @option{-O2} or higher.
8363 @item -fsched-spec-load-dangerous
8364 @opindex fsched-spec-load-dangerous
8365 Allow speculative motion of more load instructions. This only makes
8366 sense when scheduling before register allocation, i.e.@: with
8367 @option{-fschedule-insns} or at @option{-O2} or higher.
8369 @item -fsched-stalled-insns
8370 @itemx -fsched-stalled-insns=@var{n}
8371 @opindex fsched-stalled-insns
8372 Define how many insns (if any) can be moved prematurely from the queue
8373 of stalled insns into the ready list during the second scheduling pass.
8374 @option{-fno-sched-stalled-insns} means that no insns are moved
8375 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8376 on how many queued insns can be moved prematurely.
8377 @option{-fsched-stalled-insns} without a value is equivalent to
8378 @option{-fsched-stalled-insns=1}.
8380 @item -fsched-stalled-insns-dep
8381 @itemx -fsched-stalled-insns-dep=@var{n}
8382 @opindex fsched-stalled-insns-dep
8383 Define how many insn groups (cycles) are examined for a dependency
8384 on a stalled insn that is a candidate for premature removal from the queue
8385 of stalled insns. This has an effect only during the second scheduling pass,
8386 and only if @option{-fsched-stalled-insns} is used.
8387 @option{-fno-sched-stalled-insns-dep} is equivalent to
8388 @option{-fsched-stalled-insns-dep=0}.
8389 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8390 @option{-fsched-stalled-insns-dep=1}.
8392 @item -fsched2-use-superblocks
8393 @opindex fsched2-use-superblocks
8394 When scheduling after register allocation, use superblock scheduling.
8395 This allows motion across basic block boundaries,
8396 resulting in faster schedules. This option is experimental, as not all machine
8397 descriptions used by GCC model the CPU closely enough to avoid unreliable
8398 results from the algorithm.
8400 This only makes sense when scheduling after register allocation, i.e.@: with
8401 @option{-fschedule-insns2} or at @option{-O2} or higher.
8403 @item -fsched-group-heuristic
8404 @opindex fsched-group-heuristic
8405 Enable the group heuristic in the scheduler. This heuristic favors
8406 the instruction that belongs to a schedule group. This is enabled
8407 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8408 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8410 @item -fsched-critical-path-heuristic
8411 @opindex fsched-critical-path-heuristic
8412 Enable the critical-path heuristic in the scheduler. This heuristic favors
8413 instructions on the critical path. This is enabled by default when
8414 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8415 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8417 @item -fsched-spec-insn-heuristic
8418 @opindex fsched-spec-insn-heuristic
8419 Enable the speculative instruction heuristic in the scheduler. This
8420 heuristic favors speculative instructions with greater dependency weakness.
8421 This is enabled by default when scheduling is enabled, i.e.@:
8422 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8423 or at @option{-O2} or higher.
8425 @item -fsched-rank-heuristic
8426 @opindex fsched-rank-heuristic
8427 Enable the rank heuristic in the scheduler. This heuristic favors
8428 the instruction belonging to a basic block with greater size or frequency.
8429 This is enabled by default when scheduling is enabled, i.e.@:
8430 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8431 at @option{-O2} or higher.
8433 @item -fsched-last-insn-heuristic
8434 @opindex fsched-last-insn-heuristic
8435 Enable the last-instruction heuristic in the scheduler. This heuristic
8436 favors the instruction that is less dependent on the last instruction
8437 scheduled. This is enabled by default when scheduling is enabled,
8438 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8439 at @option{-O2} or higher.
8441 @item -fsched-dep-count-heuristic
8442 @opindex fsched-dep-count-heuristic
8443 Enable the dependent-count heuristic in the scheduler. This heuristic
8444 favors the instruction that has more instructions depending on it.
8445 This is enabled by default when scheduling is enabled, i.e.@:
8446 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8447 at @option{-O2} or higher.
8449 @item -freschedule-modulo-scheduled-loops
8450 @opindex freschedule-modulo-scheduled-loops
8451 Modulo scheduling is performed before traditional scheduling. If a loop
8452 is modulo scheduled, later scheduling passes may change its schedule.
8453 Use this option to control that behavior.
8455 @item -fselective-scheduling
8456 @opindex fselective-scheduling
8457 Schedule instructions using selective scheduling algorithm. Selective
8458 scheduling runs instead of the first scheduler pass.
8460 @item -fselective-scheduling2
8461 @opindex fselective-scheduling2
8462 Schedule instructions using selective scheduling algorithm. Selective
8463 scheduling runs instead of the second scheduler pass.
8465 @item -fsel-sched-pipelining
8466 @opindex fsel-sched-pipelining
8467 Enable software pipelining of innermost loops during selective scheduling.
8468 This option has no effect unless one of @option{-fselective-scheduling} or
8469 @option{-fselective-scheduling2} is turned on.
8471 @item -fsel-sched-pipelining-outer-loops
8472 @opindex fsel-sched-pipelining-outer-loops
8473 When pipelining loops during selective scheduling, also pipeline outer loops.
8474 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8476 @item -fsemantic-interposition
8477 @opindex fsemantic-interposition
8478 Some object formats, like ELF, allow interposing of symbols by the
8480 This means that for symbols exported from the DSO, the compiler cannot perform
8481 interprocedural propagation, inlining and other optimizations in anticipation
8482 that the function or variable in question may change. While this feature is
8483 useful, for example, to rewrite memory allocation functions by a debugging
8484 implementation, it is expensive in the terms of code quality.
8485 With @option{-fno-semantic-interposition} the compiler assumes that
8486 if interposition happens for functions the overwriting function will have
8487 precisely the same semantics (and side effects).
8488 Similarly if interposition happens
8489 for variables, the constructor of the variable will be the same. The flag
8490 has no effect for functions explicitly declared inline
8491 (where it is never allowed for interposition to change semantics)
8492 and for symbols explicitly declared weak.
8495 @opindex fshrink-wrap
8496 Emit function prologues only before parts of the function that need it,
8497 rather than at the top of the function. This flag is enabled by default at
8498 @option{-O} and higher.
8500 @item -fcaller-saves
8501 @opindex fcaller-saves
8502 Enable allocation of values to registers that are clobbered by
8503 function calls, by emitting extra instructions to save and restore the
8504 registers around such calls. Such allocation is done only when it
8505 seems to result in better code.
8507 This option is always enabled by default on certain machines, usually
8508 those which have no call-preserved registers to use instead.
8510 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8512 @item -fcombine-stack-adjustments
8513 @opindex fcombine-stack-adjustments
8514 Tracks stack adjustments (pushes and pops) and stack memory references
8515 and then tries to find ways to combine them.
8517 Enabled by default at @option{-O1} and higher.
8521 Use caller save registers for allocation if those registers are not used by
8522 any called function. In that case it is not necessary to save and restore
8523 them around calls. This is only possible if called functions are part of
8524 same compilation unit as current function and they are compiled before it.
8526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8528 @item -fconserve-stack
8529 @opindex fconserve-stack
8530 Attempt to minimize stack usage. The compiler attempts to use less
8531 stack space, even if that makes the program slower. This option
8532 implies setting the @option{large-stack-frame} parameter to 100
8533 and the @option{large-stack-frame-growth} parameter to 400.
8535 @item -ftree-reassoc
8536 @opindex ftree-reassoc
8537 Perform reassociation on trees. This flag is enabled by default
8538 at @option{-O} and higher.
8542 Perform partial redundancy elimination (PRE) on trees. This flag is
8543 enabled by default at @option{-O2} and @option{-O3}.
8545 @item -ftree-partial-pre
8546 @opindex ftree-partial-pre
8547 Make partial redundancy elimination (PRE) more aggressive. This flag is
8548 enabled by default at @option{-O3}.
8550 @item -ftree-forwprop
8551 @opindex ftree-forwprop
8552 Perform forward propagation on trees. This flag is enabled by default
8553 at @option{-O} and higher.
8557 Perform full redundancy elimination (FRE) on trees. The difference
8558 between FRE and PRE is that FRE only considers expressions
8559 that are computed on all paths leading to the redundant computation.
8560 This analysis is faster than PRE, though it exposes fewer redundancies.
8561 This flag is enabled by default at @option{-O} and higher.
8563 @item -ftree-phiprop
8564 @opindex ftree-phiprop
8565 Perform hoisting of loads from conditional pointers on trees. This
8566 pass is enabled by default at @option{-O} and higher.
8568 @item -fhoist-adjacent-loads
8569 @opindex fhoist-adjacent-loads
8570 Speculatively hoist loads from both branches of an if-then-else if the
8571 loads are from adjacent locations in the same structure and the target
8572 architecture has a conditional move instruction. This flag is enabled
8573 by default at @option{-O2} and higher.
8575 @item -ftree-copy-prop
8576 @opindex ftree-copy-prop
8577 Perform copy propagation on trees. This pass eliminates unnecessary
8578 copy operations. This flag is enabled by default at @option{-O} and
8581 @item -fipa-pure-const
8582 @opindex fipa-pure-const
8583 Discover which functions are pure or constant.
8584 Enabled by default at @option{-O} and higher.
8586 @item -fipa-reference
8587 @opindex fipa-reference
8588 Discover which static variables do not escape the
8590 Enabled by default at @option{-O} and higher.
8594 Perform interprocedural pointer analysis and interprocedural modification
8595 and reference analysis. This option can cause excessive memory and
8596 compile-time usage on large compilation units. It is not enabled by
8597 default at any optimization level.
8600 @opindex fipa-profile
8601 Perform interprocedural profile propagation. The functions called only from
8602 cold functions are marked as cold. Also functions executed once (such as
8603 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8604 functions and loop less parts of functions executed once are then optimized for
8606 Enabled by default at @option{-O} and higher.
8610 Perform interprocedural constant propagation.
8611 This optimization analyzes the program to determine when values passed
8612 to functions are constants and then optimizes accordingly.
8613 This optimization can substantially increase performance
8614 if the application has constants passed to functions.
8615 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8617 @item -fipa-cp-clone
8618 @opindex fipa-cp-clone
8619 Perform function cloning to make interprocedural constant propagation stronger.
8620 When enabled, interprocedural constant propagation performs function cloning
8621 when externally visible function can be called with constant arguments.
8622 Because this optimization can create multiple copies of functions,
8623 it may significantly increase code size
8624 (see @option{--param ipcp-unit-growth=@var{value}}).
8625 This flag is enabled by default at @option{-O3}.
8627 @item -fipa-cp-alignment
8628 @opindex -fipa-cp-alignment
8629 When enabled, this optimization propagates alignment of function
8630 parameters to support better vectorization and string operations.
8632 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8633 requires that @option{-fipa-cp} is enabled.
8637 Perform Identical Code Folding for functions and read-only variables.
8638 The optimization reduces code size and may disturb unwind stacks by replacing
8639 a function by equivalent one with a different name. The optimization works
8640 more effectively with link time optimization enabled.
8642 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8643 works on different levels and thus the optimizations are not same - there are
8644 equivalences that are found only by GCC and equivalences found only by Gold.
8646 This flag is enabled by default at @option{-O2} and @option{-Os}.
8648 @item -fisolate-erroneous-paths-dereference
8649 @opindex fisolate-erroneous-paths-dereference
8650 Detect paths that trigger erroneous or undefined behavior due to
8651 dereferencing a null pointer. Isolate those paths from the main control
8652 flow and turn the statement with erroneous or undefined behavior into a trap.
8653 This flag is enabled by default at @option{-O2} and higher and depends on
8654 @option{-fdelete-null-pointer-checks} also being enabled.
8656 @item -fisolate-erroneous-paths-attribute
8657 @opindex fisolate-erroneous-paths-attribute
8658 Detect paths that trigger erroneous or undefined behavior due a null value
8659 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8660 attribute. Isolate those paths from the main control flow and turn the
8661 statement with erroneous or undefined behavior into a trap. This is not
8662 currently enabled, but may be enabled by @option{-O2} in the future.
8666 Perform forward store motion on trees. This flag is
8667 enabled by default at @option{-O} and higher.
8669 @item -ftree-bit-ccp
8670 @opindex ftree-bit-ccp
8671 Perform sparse conditional bit constant propagation on trees and propagate
8672 pointer alignment information.
8673 This pass only operates on local scalar variables and is enabled by default
8674 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8678 Perform sparse conditional constant propagation (CCP) on trees. This
8679 pass only operates on local scalar variables and is enabled by default
8680 at @option{-O} and higher.
8683 @opindex fssa-phiopt
8684 Perform pattern matching on SSA PHI nodes to optimize conditional
8685 code. This pass is enabled by default at @option{-O} and higher.
8687 @item -ftree-switch-conversion
8688 @opindex ftree-switch-conversion
8689 Perform conversion of simple initializations in a switch to
8690 initializations from a scalar array. This flag is enabled by default
8691 at @option{-O2} and higher.
8693 @item -ftree-tail-merge
8694 @opindex ftree-tail-merge
8695 Look for identical code sequences. When found, replace one with a jump to the
8696 other. This optimization is known as tail merging or cross jumping. This flag
8697 is enabled by default at @option{-O2} and higher. The compilation time
8699 be limited using @option{max-tail-merge-comparisons} parameter and
8700 @option{max-tail-merge-iterations} parameter.
8704 Perform dead code elimination (DCE) on trees. This flag is enabled by
8705 default at @option{-O} and higher.
8707 @item -ftree-builtin-call-dce
8708 @opindex ftree-builtin-call-dce
8709 Perform conditional dead code elimination (DCE) for calls to built-in functions
8710 that may set @code{errno} but are otherwise side-effect free. This flag is
8711 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8714 @item -ftree-dominator-opts
8715 @opindex ftree-dominator-opts
8716 Perform a variety of simple scalar cleanups (constant/copy
8717 propagation, redundancy elimination, range propagation and expression
8718 simplification) based on a dominator tree traversal. This also
8719 performs jump threading (to reduce jumps to jumps). This flag is
8720 enabled by default at @option{-O} and higher.
8724 Perform dead store elimination (DSE) on trees. A dead store is a store into
8725 a memory location that is later overwritten by another store without
8726 any intervening loads. In this case the earlier store can be deleted. This
8727 flag is enabled by default at @option{-O} and higher.
8731 Perform loop header copying on trees. This is beneficial since it increases
8732 effectiveness of code motion optimizations. It also saves one jump. This flag
8733 is enabled by default at @option{-O} and higher. It is not enabled
8734 for @option{-Os}, since it usually increases code size.
8736 @item -ftree-loop-optimize
8737 @opindex ftree-loop-optimize
8738 Perform loop optimizations on trees. This flag is enabled by default
8739 at @option{-O} and higher.
8741 @item -ftree-loop-linear
8742 @opindex ftree-loop-linear
8743 Perform loop interchange transformations on tree. Same as
8744 @option{-floop-interchange}. To use this code transformation, GCC has
8745 to be configured with @option{--with-isl} to enable the Graphite loop
8746 transformation infrastructure.
8748 @item -floop-interchange
8749 @opindex floop-interchange
8750 Perform loop interchange transformations on loops. Interchanging two
8751 nested loops switches the inner and outer loops. For example, given a
8756 A(J, I) = A(J, I) * C
8761 loop interchange transforms the loop as if it were written:
8765 A(J, I) = A(J, I) * C
8769 which can be beneficial when @code{N} is larger than the caches,
8770 because in Fortran, the elements of an array are stored in memory
8771 contiguously by column, and the original loop iterates over rows,
8772 potentially creating at each access a cache miss. This optimization
8773 applies to all the languages supported by GCC and is not limited to
8774 Fortran. To use this code transformation, GCC has to be configured
8775 with @option{--with-isl} to enable the Graphite loop transformation
8778 @item -floop-strip-mine
8779 @opindex floop-strip-mine
8780 Perform loop strip mining transformations on loops. Strip mining
8781 splits a loop into two nested loops. The outer loop has strides
8782 equal to the strip size and the inner loop has strides of the
8783 original loop within a strip. The strip length can be changed
8784 using the @option{loop-block-tile-size} parameter. For example,
8792 loop strip mining transforms the loop as if it were written:
8795 DO I = II, min (II + 50, N)
8800 This optimization applies to all the languages supported by GCC and is
8801 not limited to Fortran. To use this code transformation, GCC has to
8802 be configured with @option{--with-isl} to enable the Graphite loop
8803 transformation infrastructure.
8806 @opindex floop-block
8807 Perform loop blocking transformations on loops. Blocking strip mines
8808 each loop in the loop nest such that the memory accesses of the
8809 element loops fit inside caches. The strip length can be changed
8810 using the @option{loop-block-tile-size} parameter. For example, given
8815 A(J, I) = B(I) + C(J)
8820 loop blocking transforms the loop as if it were written:
8824 DO I = II, min (II + 50, N)
8825 DO J = JJ, min (JJ + 50, M)
8826 A(J, I) = B(I) + C(J)
8832 which can be beneficial when @code{M} is larger than the caches,
8833 because the innermost loop iterates over a smaller amount of data
8834 which can be kept in the caches. This optimization applies to all the
8835 languages supported by GCC and is not limited to Fortran. To use this
8836 code transformation, GCC has to be configured with @option{--with-isl}
8837 to enable the Graphite loop transformation infrastructure.
8839 @item -fgraphite-identity
8840 @opindex fgraphite-identity
8841 Enable the identity transformation for graphite. For every SCoP we generate
8842 the polyhedral representation and transform it back to gimple. Using
8843 @option{-fgraphite-identity} we can check the costs or benefits of the
8844 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8845 are also performed by the code generator ISL, like index splitting and
8846 dead code elimination in loops.
8848 @item -floop-nest-optimize
8849 @opindex floop-nest-optimize
8850 Enable the ISL based loop nest optimizer. This is a generic loop nest
8851 optimizer based on the Pluto optimization algorithms. It calculates a loop
8852 structure optimized for data-locality and parallelism. This option
8855 @item -floop-unroll-and-jam
8856 @opindex floop-unroll-and-jam
8857 Enable unroll and jam for the ISL based loop nest optimizer. The unroll
8858 factor can be changed using the @option{loop-unroll-jam-size} parameter.
8859 The unrolled dimension (counting from the most inner one) can be changed
8860 using the @option{loop-unroll-jam-depth} parameter. .
8862 @item -floop-parallelize-all
8863 @opindex floop-parallelize-all
8864 Use the Graphite data dependence analysis to identify loops that can
8865 be parallelized. Parallelize all the loops that can be analyzed to
8866 not contain loop carried dependences without checking that it is
8867 profitable to parallelize the loops.
8869 @item -ftree-loop-if-convert
8870 @opindex ftree-loop-if-convert
8871 Attempt to transform conditional jumps in the innermost loops to
8872 branch-less equivalents. The intent is to remove control-flow from
8873 the innermost loops in order to improve the ability of the
8874 vectorization pass to handle these loops. This is enabled by default
8875 if vectorization is enabled.
8877 @item -ftree-loop-if-convert-stores
8878 @opindex ftree-loop-if-convert-stores
8879 Attempt to also if-convert conditional jumps containing memory writes.
8880 This transformation can be unsafe for multi-threaded programs as it
8881 transforms conditional memory writes into unconditional memory writes.
8884 for (i = 0; i < N; i++)
8890 for (i = 0; i < N; i++)
8891 A[i] = cond ? expr : A[i];
8893 potentially producing data races.
8895 @item -ftree-loop-distribution
8896 @opindex ftree-loop-distribution
8897 Perform loop distribution. This flag can improve cache performance on
8898 big loop bodies and allow further loop optimizations, like
8899 parallelization or vectorization, to take place. For example, the loop
8916 @item -ftree-loop-distribute-patterns
8917 @opindex ftree-loop-distribute-patterns
8918 Perform loop distribution of patterns that can be code generated with
8919 calls to a library. This flag is enabled by default at @option{-O3}.
8921 This pass distributes the initialization loops and generates a call to
8922 memset zero. For example, the loop
8938 and the initialization loop is transformed into a call to memset zero.
8940 @item -ftree-loop-im
8941 @opindex ftree-loop-im
8942 Perform loop invariant motion on trees. This pass moves only invariants that
8943 are hard to handle at RTL level (function calls, operations that expand to
8944 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
8945 operands of conditions that are invariant out of the loop, so that we can use
8946 just trivial invariantness analysis in loop unswitching. The pass also includes
8949 @item -ftree-loop-ivcanon
8950 @opindex ftree-loop-ivcanon
8951 Create a canonical counter for number of iterations in loops for which
8952 determining number of iterations requires complicated analysis. Later
8953 optimizations then may determine the number easily. Useful especially
8954 in connection with unrolling.
8958 Perform induction variable optimizations (strength reduction, induction
8959 variable merging and induction variable elimination) on trees.
8961 @item -ftree-parallelize-loops=n
8962 @opindex ftree-parallelize-loops
8963 Parallelize loops, i.e., split their iteration space to run in n threads.
8964 This is only possible for loops whose iterations are independent
8965 and can be arbitrarily reordered. The optimization is only
8966 profitable on multiprocessor machines, for loops that are CPU-intensive,
8967 rather than constrained e.g.@: by memory bandwidth. This option
8968 implies @option{-pthread}, and thus is only supported on targets
8969 that have support for @option{-pthread}.
8973 Perform function-local points-to analysis on trees. This flag is
8974 enabled by default at @option{-O} and higher.
8978 Perform scalar replacement of aggregates. This pass replaces structure
8979 references with scalars to prevent committing structures to memory too
8980 early. This flag is enabled by default at @option{-O} and higher.
8982 @item -ftree-copyrename
8983 @opindex ftree-copyrename
8984 Perform copy renaming on trees. This pass attempts to rename compiler
8985 temporaries to other variables at copy locations, usually resulting in
8986 variable names which more closely resemble the original variables. This flag
8987 is enabled by default at @option{-O} and higher.
8989 @item -ftree-coalesce-inlined-vars
8990 @opindex ftree-coalesce-inlined-vars
8991 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
8992 combine small user-defined variables too, but only if they are inlined
8993 from other functions. It is a more limited form of
8994 @option{-ftree-coalesce-vars}. This may harm debug information of such
8995 inlined variables, but it keeps variables of the inlined-into
8996 function apart from each other, such that they are more likely to
8997 contain the expected values in a debugging session.
8999 @item -ftree-coalesce-vars
9000 @opindex ftree-coalesce-vars
9001 Tell the copyrename pass (see @option{-ftree-copyrename}) to attempt to
9002 combine small user-defined variables too, instead of just compiler
9003 temporaries. This may severely limit the ability to debug an optimized
9004 program compiled with @option{-fno-var-tracking-assignments}. In the
9005 negated form, this flag prevents SSA coalescing of user variables,
9006 including inlined ones. This option is enabled by default.
9010 Perform temporary expression replacement during the SSA->normal phase. Single
9011 use/single def temporaries are replaced at their use location with their
9012 defining expression. This results in non-GIMPLE code, but gives the expanders
9013 much more complex trees to work on resulting in better RTL generation. This is
9014 enabled by default at @option{-O} and higher.
9018 Perform straight-line strength reduction on trees. This recognizes related
9019 expressions involving multiplications and replaces them by less expensive
9020 calculations when possible. This is enabled by default at @option{-O} and
9023 @item -ftree-vectorize
9024 @opindex ftree-vectorize
9025 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9026 and @option{-ftree-slp-vectorize} if not explicitly specified.
9028 @item -ftree-loop-vectorize
9029 @opindex ftree-loop-vectorize
9030 Perform loop vectorization on trees. This flag is enabled by default at
9031 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9033 @item -ftree-slp-vectorize
9034 @opindex ftree-slp-vectorize
9035 Perform basic block vectorization on trees. This flag is enabled by default at
9036 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9038 @item -fvect-cost-model=@var{model}
9039 @opindex fvect-cost-model
9040 Alter the cost model used for vectorization. The @var{model} argument
9041 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9042 With the @samp{unlimited} model the vectorized code-path is assumed
9043 to be profitable while with the @samp{dynamic} model a runtime check
9044 guards the vectorized code-path to enable it only for iteration
9045 counts that will likely execute faster than when executing the original
9046 scalar loop. The @samp{cheap} model disables vectorization of
9047 loops where doing so would be cost prohibitive for example due to
9048 required runtime checks for data dependence or alignment but otherwise
9049 is equal to the @samp{dynamic} model.
9050 The default cost model depends on other optimization flags and is
9051 either @samp{dynamic} or @samp{cheap}.
9053 @item -fsimd-cost-model=@var{model}
9054 @opindex fsimd-cost-model
9055 Alter the cost model used for vectorization of loops marked with the OpenMP
9056 or Cilk Plus simd directive. The @var{model} argument should be one of
9057 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9058 have the same meaning as described in @option{-fvect-cost-model} and by
9059 default a cost model defined with @option{-fvect-cost-model} is used.
9063 Perform Value Range Propagation on trees. This is similar to the
9064 constant propagation pass, but instead of values, ranges of values are
9065 propagated. This allows the optimizers to remove unnecessary range
9066 checks like array bound checks and null pointer checks. This is
9067 enabled by default at @option{-O2} and higher. Null pointer check
9068 elimination is only done if @option{-fdelete-null-pointer-checks} is
9071 @item -fsplit-ivs-in-unroller
9072 @opindex fsplit-ivs-in-unroller
9073 Enables expression of values of induction variables in later iterations
9074 of the unrolled loop using the value in the first iteration. This breaks
9075 long dependency chains, thus improving efficiency of the scheduling passes.
9077 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9078 same effect. However, that is not reliable in cases where the loop body
9079 is more complicated than a single basic block. It also does not work at all
9080 on some architectures due to restrictions in the CSE pass.
9082 This optimization is enabled by default.
9084 @item -fvariable-expansion-in-unroller
9085 @opindex fvariable-expansion-in-unroller
9086 With this option, the compiler creates multiple copies of some
9087 local variables when unrolling a loop, which can result in superior code.
9089 @item -fpartial-inlining
9090 @opindex fpartial-inlining
9091 Inline parts of functions. This option has any effect only
9092 when inlining itself is turned on by the @option{-finline-functions}
9093 or @option{-finline-small-functions} options.
9095 Enabled at level @option{-O2}.
9097 @item -fpredictive-commoning
9098 @opindex fpredictive-commoning
9099 Perform predictive commoning optimization, i.e., reusing computations
9100 (especially memory loads and stores) performed in previous
9101 iterations of loops.
9103 This option is enabled at level @option{-O3}.
9105 @item -fprefetch-loop-arrays
9106 @opindex fprefetch-loop-arrays
9107 If supported by the target machine, generate instructions to prefetch
9108 memory to improve the performance of loops that access large arrays.
9110 This option may generate better or worse code; results are highly
9111 dependent on the structure of loops within the source code.
9113 Disabled at level @option{-Os}.
9116 @itemx -fno-peephole2
9117 @opindex fno-peephole
9118 @opindex fno-peephole2
9119 Disable any machine-specific peephole optimizations. The difference
9120 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9121 are implemented in the compiler; some targets use one, some use the
9122 other, a few use both.
9124 @option{-fpeephole} is enabled by default.
9125 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9127 @item -fno-guess-branch-probability
9128 @opindex fno-guess-branch-probability
9129 Do not guess branch probabilities using heuristics.
9131 GCC uses heuristics to guess branch probabilities if they are
9132 not provided by profiling feedback (@option{-fprofile-arcs}). These
9133 heuristics are based on the control flow graph. If some branch probabilities
9134 are specified by @code{__builtin_expect}, then the heuristics are
9135 used to guess branch probabilities for the rest of the control flow graph,
9136 taking the @code{__builtin_expect} info into account. The interactions
9137 between the heuristics and @code{__builtin_expect} can be complex, and in
9138 some cases, it may be useful to disable the heuristics so that the effects
9139 of @code{__builtin_expect} are easier to understand.
9141 The default is @option{-fguess-branch-probability} at levels
9142 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9144 @item -freorder-blocks
9145 @opindex freorder-blocks
9146 Reorder basic blocks in the compiled function in order to reduce number of
9147 taken branches and improve code locality.
9149 Enabled at levels @option{-O2}, @option{-O3}.
9151 @item -freorder-blocks-and-partition
9152 @opindex freorder-blocks-and-partition
9153 In addition to reordering basic blocks in the compiled function, in order
9154 to reduce number of taken branches, partitions hot and cold basic blocks
9155 into separate sections of the assembly and .o files, to improve
9156 paging and cache locality performance.
9158 This optimization is automatically turned off in the presence of
9159 exception handling, for linkonce sections, for functions with a user-defined
9160 section attribute and on any architecture that does not support named
9163 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9165 @item -freorder-functions
9166 @opindex freorder-functions
9167 Reorder functions in the object file in order to
9168 improve code locality. This is implemented by using special
9169 subsections @code{.text.hot} for most frequently executed functions and
9170 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9171 the linker so object file format must support named sections and linker must
9172 place them in a reasonable way.
9174 Also profile feedback must be available to make this option effective. See
9175 @option{-fprofile-arcs} for details.
9177 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9179 @item -fstrict-aliasing
9180 @opindex fstrict-aliasing
9181 Allow the compiler to assume the strictest aliasing rules applicable to
9182 the language being compiled. For C (and C++), this activates
9183 optimizations based on the type of expressions. In particular, an
9184 object of one type is assumed never to reside at the same address as an
9185 object of a different type, unless the types are almost the same. For
9186 example, an @code{unsigned int} can alias an @code{int}, but not a
9187 @code{void*} or a @code{double}. A character type may alias any other
9190 @anchor{Type-punning}Pay special attention to code like this:
9203 The practice of reading from a different union member than the one most
9204 recently written to (called ``type-punning'') is common. Even with
9205 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9206 is accessed through the union type. So, the code above works as
9207 expected. @xref{Structures unions enumerations and bit-fields
9208 implementation}. However, this code might not:
9219 Similarly, access by taking the address, casting the resulting pointer
9220 and dereferencing the result has undefined behavior, even if the cast
9221 uses a union type, e.g.:
9225 return ((union a_union *) &d)->i;
9229 The @option{-fstrict-aliasing} option is enabled at levels
9230 @option{-O2}, @option{-O3}, @option{-Os}.
9232 @item -fstrict-overflow
9233 @opindex fstrict-overflow
9234 Allow the compiler to assume strict signed overflow rules, depending
9235 on the language being compiled. For C (and C++) this means that
9236 overflow when doing arithmetic with signed numbers is undefined, which
9237 means that the compiler may assume that it does not happen. This
9238 permits various optimizations. For example, the compiler assumes
9239 that an expression like @code{i + 10 > i} is always true for
9240 signed @code{i}. This assumption is only valid if signed overflow is
9241 undefined, as the expression is false if @code{i + 10} overflows when
9242 using twos complement arithmetic. When this option is in effect any
9243 attempt to determine whether an operation on signed numbers
9244 overflows must be written carefully to not actually involve overflow.
9246 This option also allows the compiler to assume strict pointer
9247 semantics: given a pointer to an object, if adding an offset to that
9248 pointer does not produce a pointer to the same object, the addition is
9249 undefined. This permits the compiler to conclude that @code{p + u >
9250 p} is always true for a pointer @code{p} and unsigned integer
9251 @code{u}. This assumption is only valid because pointer wraparound is
9252 undefined, as the expression is false if @code{p + u} overflows using
9253 twos complement arithmetic.
9255 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9256 that integer signed overflow is fully defined: it wraps. When
9257 @option{-fwrapv} is used, there is no difference between
9258 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9259 integers. With @option{-fwrapv} certain types of overflow are
9260 permitted. For example, if the compiler gets an overflow when doing
9261 arithmetic on constants, the overflowed value can still be used with
9262 @option{-fwrapv}, but not otherwise.
9264 The @option{-fstrict-overflow} option is enabled at levels
9265 @option{-O2}, @option{-O3}, @option{-Os}.
9267 @item -falign-functions
9268 @itemx -falign-functions=@var{n}
9269 @opindex falign-functions
9270 Align the start of functions to the next power-of-two greater than
9271 @var{n}, skipping up to @var{n} bytes. For instance,
9272 @option{-falign-functions=32} aligns functions to the next 32-byte
9273 boundary, but @option{-falign-functions=24} aligns to the next
9274 32-byte boundary only if this can be done by skipping 23 bytes or less.
9276 @option{-fno-align-functions} and @option{-falign-functions=1} are
9277 equivalent and mean that functions are not aligned.
9279 Some assemblers only support this flag when @var{n} is a power of two;
9280 in that case, it is rounded up.
9282 If @var{n} is not specified or is zero, use a machine-dependent default.
9284 Enabled at levels @option{-O2}, @option{-O3}.
9286 @item -falign-labels
9287 @itemx -falign-labels=@var{n}
9288 @opindex falign-labels
9289 Align all branch targets to a power-of-two boundary, skipping up to
9290 @var{n} bytes like @option{-falign-functions}. This option can easily
9291 make code slower, because it must insert dummy operations for when the
9292 branch target is reached in the usual flow of the code.
9294 @option{-fno-align-labels} and @option{-falign-labels=1} are
9295 equivalent and mean that labels are not aligned.
9297 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9298 are greater than this value, then their values are used instead.
9300 If @var{n} is not specified or is zero, use a machine-dependent default
9301 which is very likely to be @samp{1}, meaning no alignment.
9303 Enabled at levels @option{-O2}, @option{-O3}.
9306 @itemx -falign-loops=@var{n}
9307 @opindex falign-loops
9308 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9309 like @option{-falign-functions}. If the loops are
9310 executed many times, this makes up for any execution of the dummy
9313 @option{-fno-align-loops} and @option{-falign-loops=1} are
9314 equivalent and mean that loops are not aligned.
9316 If @var{n} is not specified or is zero, use a machine-dependent default.
9318 Enabled at levels @option{-O2}, @option{-O3}.
9321 @itemx -falign-jumps=@var{n}
9322 @opindex falign-jumps
9323 Align branch targets to a power-of-two boundary, for branch targets
9324 where the targets can only be reached by jumping, skipping up to @var{n}
9325 bytes like @option{-falign-functions}. In this case, no dummy operations
9328 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9329 equivalent and mean that loops are not aligned.
9331 If @var{n} is not specified or is zero, use a machine-dependent default.
9333 Enabled at levels @option{-O2}, @option{-O3}.
9335 @item -funit-at-a-time
9336 @opindex funit-at-a-time
9337 This option is left for compatibility reasons. @option{-funit-at-a-time}
9338 has no effect, while @option{-fno-unit-at-a-time} implies
9339 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9343 @item -fno-toplevel-reorder
9344 @opindex fno-toplevel-reorder
9345 Do not reorder top-level functions, variables, and @code{asm}
9346 statements. Output them in the same order that they appear in the
9347 input file. When this option is used, unreferenced static variables
9348 are not removed. This option is intended to support existing code
9349 that relies on a particular ordering. For new code, it is better to
9350 use attributes when possible.
9352 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9353 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9358 Constructs webs as commonly used for register allocation purposes and assign
9359 each web individual pseudo register. This allows the register allocation pass
9360 to operate on pseudos directly, but also strengthens several other optimization
9361 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9362 however, make debugging impossible, since variables no longer stay in a
9365 Enabled by default with @option{-funroll-loops}.
9367 @item -fwhole-program
9368 @opindex fwhole-program
9369 Assume that the current compilation unit represents the whole program being
9370 compiled. All public functions and variables with the exception of @code{main}
9371 and those merged by attribute @code{externally_visible} become static functions
9372 and in effect are optimized more aggressively by interprocedural optimizers.
9374 This option should not be used in combination with @option{-flto}.
9375 Instead relying on a linker plugin should provide safer and more precise
9378 @item -flto[=@var{n}]
9380 This option runs the standard link-time optimizer. When invoked
9381 with source code, it generates GIMPLE (one of GCC's internal
9382 representations) and writes it to special ELF sections in the object
9383 file. When the object files are linked together, all the function
9384 bodies are read from these ELF sections and instantiated as if they
9385 had been part of the same translation unit.
9387 To use the link-time optimizer, @option{-flto} and optimization
9388 options should be specified at compile time and during the final link.
9392 gcc -c -O2 -flto foo.c
9393 gcc -c -O2 -flto bar.c
9394 gcc -o myprog -flto -O2 foo.o bar.o
9397 The first two invocations to GCC save a bytecode representation
9398 of GIMPLE into special ELF sections inside @file{foo.o} and
9399 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9400 @file{foo.o} and @file{bar.o}, merges the two files into a single
9401 internal image, and compiles the result as usual. Since both
9402 @file{foo.o} and @file{bar.o} are merged into a single image, this
9403 causes all the interprocedural analyses and optimizations in GCC to
9404 work across the two files as if they were a single one. This means,
9405 for example, that the inliner is able to inline functions in
9406 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9408 Another (simpler) way to enable link-time optimization is:
9411 gcc -o myprog -flto -O2 foo.c bar.c
9414 The above generates bytecode for @file{foo.c} and @file{bar.c},
9415 merges them together into a single GIMPLE representation and optimizes
9416 them as usual to produce @file{myprog}.
9418 The only important thing to keep in mind is that to enable link-time
9419 optimizations you need to use the GCC driver to perform the link-step.
9420 GCC then automatically performs link-time optimization if any of the
9421 objects involved were compiled with the @option{-flto} command-line option.
9423 should specify the optimization options to be used for link-time
9424 optimization though GCC tries to be clever at guessing an
9425 optimization level to use from the options used at compile-time
9426 if you fail to specify one at link-time. You can always override
9427 the automatic decision to do link-time optimization at link-time
9428 by passing @option{-fno-lto} to the link command.
9430 To make whole program optimization effective, it is necessary to make
9431 certain whole program assumptions. The compiler needs to know
9432 what functions and variables can be accessed by libraries and runtime
9433 outside of the link-time optimized unit. When supported by the linker,
9434 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9435 to the compiler about used and externally visible symbols. When
9436 the linker plugin is not available, @option{-fwhole-program} should be
9437 used to allow the compiler to make these assumptions, which leads
9438 to more aggressive optimization decisions.
9440 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9441 compiled with @option{-flto}, the generated object file is larger than
9442 a regular object file because it contains GIMPLE bytecodes and the usual
9443 final code (see @option{-ffat-lto-objects}. This means that
9444 object files with LTO information can be linked as normal object
9445 files; if @option{-fno-lto} is passed to the linker, no
9446 interprocedural optimizations are applied. Note that when
9447 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9448 but you cannot perform a regular, non-LTO link on them.
9450 Additionally, the optimization flags used to compile individual files
9451 are not necessarily related to those used at link time. For instance,
9454 gcc -c -O0 -ffat-lto-objects -flto foo.c
9455 gcc -c -O0 -ffat-lto-objects -flto bar.c
9456 gcc -o myprog -O3 foo.o bar.o
9459 This produces individual object files with unoptimized assembler
9460 code, but the resulting binary @file{myprog} is optimized at
9461 @option{-O3}. If, instead, the final binary is generated with
9462 @option{-fno-lto}, then @file{myprog} is not optimized.
9464 When producing the final binary, GCC only
9465 applies link-time optimizations to those files that contain bytecode.
9466 Therefore, you can mix and match object files and libraries with
9467 GIMPLE bytecodes and final object code. GCC automatically selects
9468 which files to optimize in LTO mode and which files to link without
9471 There are some code generation flags preserved by GCC when
9472 generating bytecodes, as they need to be used during the final link
9473 stage. Generally options specified at link-time override those
9474 specified at compile-time.
9476 If you do not specify an optimization level option @option{-O} at
9477 link-time then GCC computes one based on the optimization levels
9478 used when compiling the object files. The highest optimization
9481 Currently, the following options and their setting are take from
9482 the first object file that explicitely specified it:
9483 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9484 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9485 and all the @option{-m} target flags.
9487 Certain ABI changing flags are required to match in all compilation-units
9488 and trying to override this at link-time with a conflicting value
9489 is ignored. This includes options such as @option{-freg-struct-return}
9490 and @option{-fpcc-struct-return}.
9492 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9493 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9494 are passed through to the link stage and merged conservatively for
9495 conflicting translation units. Specifically
9496 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9497 precedence and for example @option{-ffp-contract=off} takes precedence
9498 over @option{-ffp-contract=fast}. You can override them at linke-time.
9500 It is recommended that you compile all the files participating in the
9501 same link with the same options and also specify those options at
9504 If LTO encounters objects with C linkage declared with incompatible
9505 types in separate translation units to be linked together (undefined
9506 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9507 issued. The behavior is still undefined at run time. Similar
9508 diagnostics may be raised for other languages.
9510 Another feature of LTO is that it is possible to apply interprocedural
9511 optimizations on files written in different languages:
9516 gfortran -c -flto baz.f90
9517 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9520 Notice that the final link is done with @command{g++} to get the C++
9521 runtime libraries and @option{-lgfortran} is added to get the Fortran
9522 runtime libraries. In general, when mixing languages in LTO mode, you
9523 should use the same link command options as when mixing languages in a
9524 regular (non-LTO) compilation.
9526 If object files containing GIMPLE bytecode are stored in a library archive, say
9527 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9528 are using a linker with plugin support. To create static libraries suitable
9529 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9530 and @command{ranlib};
9531 to show the symbols of object files with GIMPLE bytecode, use
9532 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9533 and @command{nm} have been compiled with plugin support. At link time, use the the
9534 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9535 the LTO optimization process:
9538 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9541 With the linker plugin enabled, the linker extracts the needed
9542 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9543 to make them part of the aggregated GIMPLE image to be optimized.
9545 If you are not using a linker with plugin support and/or do not
9546 enable the linker plugin, then the objects inside @file{libfoo.a}
9547 are extracted and linked as usual, but they do not participate
9548 in the LTO optimization process. In order to make a static library suitable
9549 for both LTO optimization and usual linkage, compile its object files with
9550 @option{-flto} @option{-ffat-lto-objects}.
9552 Link-time optimizations do not require the presence of the whole program to
9553 operate. If the program does not require any symbols to be exported, it is
9554 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9555 the interprocedural optimizers to use more aggressive assumptions which may
9556 lead to improved optimization opportunities.
9557 Use of @option{-fwhole-program} is not needed when linker plugin is
9558 active (see @option{-fuse-linker-plugin}).
9560 The current implementation of LTO makes no
9561 attempt to generate bytecode that is portable between different
9562 types of hosts. The bytecode files are versioned and there is a
9563 strict version check, so bytecode files generated in one version of
9564 GCC do not work with an older or newer version of GCC.
9566 Link-time optimization does not work well with generation of debugging
9567 information. Combining @option{-flto} with
9568 @option{-g} is currently experimental and expected to produce unexpected
9571 If you specify the optional @var{n}, the optimization and code
9572 generation done at link time is executed in parallel using @var{n}
9573 parallel jobs by utilizing an installed @command{make} program. The
9574 environment variable @env{MAKE} may be used to override the program
9575 used. The default value for @var{n} is 1.
9577 You can also specify @option{-flto=jobserver} to use GNU make's
9578 job server mode to determine the number of parallel jobs. This
9579 is useful when the Makefile calling GCC is already executing in parallel.
9580 You must prepend a @samp{+} to the command recipe in the parent Makefile
9581 for this to work. This option likely only works if @env{MAKE} is
9584 @item -flto-partition=@var{alg}
9585 @opindex flto-partition
9586 Specify the partitioning algorithm used by the link-time optimizer.
9587 The value is either @samp{1to1} to specify a partitioning mirroring
9588 the original source files or @samp{balanced} to specify partitioning
9589 into equally sized chunks (whenever possible) or @samp{max} to create
9590 new partition for every symbol where possible. Specifying @samp{none}
9591 as an algorithm disables partitioning and streaming completely.
9592 The default value is @samp{balanced}. While @samp{1to1} can be used
9593 as an workaround for various code ordering issues, the @samp{max}
9594 partitioning is intended for internal testing only.
9595 The value @samp{one} specifies that exactly one partition should be
9596 used while the value @samp{none} bypasses partitioning and executes
9597 the link-time optimization step directly from the WPA phase.
9599 @item -flto-odr-type-merging
9600 @opindex flto-odr-type-merging
9601 Enable streaming of mangled types names of C++ types and their unification
9602 at linktime. This increases size of LTO object files, but enable
9603 diagnostics about One Definition Rule violations.
9605 @item -flto-compression-level=@var{n}
9606 @opindex flto-compression-level
9607 This option specifies the level of compression used for intermediate
9608 language written to LTO object files, and is only meaningful in
9609 conjunction with LTO mode (@option{-flto}). Valid
9610 values are 0 (no compression) to 9 (maximum compression). Values
9611 outside this range are clamped to either 0 or 9. If the option is not
9612 given, a default balanced compression setting is used.
9615 @opindex flto-report
9616 Prints a report with internal details on the workings of the link-time
9617 optimizer. The contents of this report vary from version to version.
9618 It is meant to be useful to GCC developers when processing object
9619 files in LTO mode (via @option{-flto}).
9621 Disabled by default.
9623 @item -flto-report-wpa
9624 @opindex flto-report-wpa
9625 Like @option{-flto-report}, but only print for the WPA phase of Link
9628 @item -fuse-linker-plugin
9629 @opindex fuse-linker-plugin
9630 Enables the use of a linker plugin during link-time optimization. This
9631 option relies on plugin support in the linker, which is available in gold
9632 or in GNU ld 2.21 or newer.
9634 This option enables the extraction of object files with GIMPLE bytecode out
9635 of library archives. This improves the quality of optimization by exposing
9636 more code to the link-time optimizer. This information specifies what
9637 symbols can be accessed externally (by non-LTO object or during dynamic
9638 linking). Resulting code quality improvements on binaries (and shared
9639 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9640 See @option{-flto} for a description of the effect of this flag and how to
9643 This option is enabled by default when LTO support in GCC is enabled
9644 and GCC was configured for use with
9645 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9647 @item -ffat-lto-objects
9648 @opindex ffat-lto-objects
9649 Fat LTO objects are object files that contain both the intermediate language
9650 and the object code. This makes them usable for both LTO linking and normal
9651 linking. This option is effective only when compiling with @option{-flto}
9652 and is ignored at link time.
9654 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9655 requires the complete toolchain to be aware of LTO. It requires a linker with
9656 linker plugin support for basic functionality. Additionally,
9657 @command{nm}, @command{ar} and @command{ranlib}
9658 need to support linker plugins to allow a full-featured build environment
9659 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9660 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9661 to these tools. With non fat LTO makefiles need to be modified to use them.
9663 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9666 @item -fcompare-elim
9667 @opindex fcompare-elim
9668 After register allocation and post-register allocation instruction splitting,
9669 identify arithmetic instructions that compute processor flags similar to a
9670 comparison operation based on that arithmetic. If possible, eliminate the
9671 explicit comparison operation.
9673 This pass only applies to certain targets that cannot explicitly represent
9674 the comparison operation before register allocation is complete.
9676 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9678 @item -fcprop-registers
9679 @opindex fcprop-registers
9680 After register allocation and post-register allocation instruction splitting,
9681 perform a copy-propagation pass to try to reduce scheduling dependencies
9682 and occasionally eliminate the copy.
9684 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9686 @item -fprofile-correction
9687 @opindex fprofile-correction
9688 Profiles collected using an instrumented binary for multi-threaded programs may
9689 be inconsistent due to missed counter updates. When this option is specified,
9690 GCC uses heuristics to correct or smooth out such inconsistencies. By
9691 default, GCC emits an error message when an inconsistent profile is detected.
9693 @item -fprofile-dir=@var{path}
9694 @opindex fprofile-dir
9696 Set the directory to search for the profile data files in to @var{path}.
9697 This option affects only the profile data generated by
9698 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9699 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9700 and its related options. Both absolute and relative paths can be used.
9701 By default, GCC uses the current directory as @var{path}, thus the
9702 profile data file appears in the same directory as the object file.
9704 @item -fprofile-generate
9705 @itemx -fprofile-generate=@var{path}
9706 @opindex fprofile-generate
9708 Enable options usually used for instrumenting application to produce
9709 profile useful for later recompilation with profile feedback based
9710 optimization. You must use @option{-fprofile-generate} both when
9711 compiling and when linking your program.
9713 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9715 If @var{path} is specified, GCC looks at the @var{path} to find
9716 the profile feedback data files. See @option{-fprofile-dir}.
9719 @itemx -fprofile-use=@var{path}
9720 @opindex fprofile-use
9721 Enable profile feedback-directed optimizations,
9722 and the following optimizations
9723 which are generally profitable only with profile feedback available:
9724 @option{-fbranch-probabilities}, @option{-fvpt},
9725 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9726 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9728 By default, GCC emits an error message if the feedback profiles do not
9729 match the source code. This error can be turned into a warning by using
9730 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9733 If @var{path} is specified, GCC looks at the @var{path} to find
9734 the profile feedback data files. See @option{-fprofile-dir}.
9736 @item -fauto-profile
9737 @itemx -fauto-profile=@var{path}
9738 @opindex fauto-profile
9739 Enable sampling-based feedback-directed optimizations,
9740 and the following optimizations
9741 which are generally profitable only with profile feedback available:
9742 @option{-fbranch-probabilities}, @option{-fvpt},
9743 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9744 @option{-ftree-vectorize},
9745 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9746 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9747 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9749 @var{path} is the name of a file containing AutoFDO profile information.
9750 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9752 Producing an AutoFDO profile data file requires running your program
9753 with the @command{perf} utility on a supported GNU/Linux target system.
9754 For more information, see @uref{https://perf.wiki.kernel.org/}.
9758 perf record -e br_inst_retired:near_taken -b -o perf.data \
9762 Then use the @command{create_gcov} tool to convert the raw profile data
9763 to a format that can be used by GCC.@ You must also supply the
9764 unstripped binary for your program to this tool.
9765 See @uref{https://github.com/google/autofdo}.
9769 create_gcov --binary=your_program.unstripped --profile=perf.data \
9774 The following options control compiler behavior regarding floating-point
9775 arithmetic. These options trade off between speed and
9776 correctness. All must be specifically enabled.
9780 @opindex ffloat-store
9781 Do not store floating-point variables in registers, and inhibit other
9782 options that might change whether a floating-point value is taken from a
9785 @cindex floating-point precision
9786 This option prevents undesirable excess precision on machines such as
9787 the 68000 where the floating registers (of the 68881) keep more
9788 precision than a @code{double} is supposed to have. Similarly for the
9789 x86 architecture. For most programs, the excess precision does only
9790 good, but a few programs rely on the precise definition of IEEE floating
9791 point. Use @option{-ffloat-store} for such programs, after modifying
9792 them to store all pertinent intermediate computations into variables.
9794 @item -fexcess-precision=@var{style}
9795 @opindex fexcess-precision
9796 This option allows further control over excess precision on machines
9797 where floating-point registers have more precision than the IEEE
9798 @code{float} and @code{double} types and the processor does not
9799 support operations rounding to those types. By default,
9800 @option{-fexcess-precision=fast} is in effect; this means that
9801 operations are carried out in the precision of the registers and that
9802 it is unpredictable when rounding to the types specified in the source
9803 code takes place. When compiling C, if
9804 @option{-fexcess-precision=standard} is specified then excess
9805 precision follows the rules specified in ISO C99; in particular,
9806 both casts and assignments cause values to be rounded to their
9807 semantic types (whereas @option{-ffloat-store} only affects
9808 assignments). This option is enabled by default for C if a strict
9809 conformance option such as @option{-std=c99} is used.
9812 @option{-fexcess-precision=standard} is not implemented for languages
9813 other than C, and has no effect if
9814 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9815 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9816 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9817 semantics apply without excess precision, and in the latter, rounding
9822 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9823 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9824 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9826 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9828 This option is not turned on by any @option{-O} option besides
9829 @option{-Ofast} since it can result in incorrect output for programs
9830 that depend on an exact implementation of IEEE or ISO rules/specifications
9831 for math functions. It may, however, yield faster code for programs
9832 that do not require the guarantees of these specifications.
9834 @item -fno-math-errno
9835 @opindex fno-math-errno
9836 Do not set @code{errno} after calling math functions that are executed
9837 with a single instruction, e.g., @code{sqrt}. A program that relies on
9838 IEEE exceptions for math error handling may want to use this flag
9839 for speed while maintaining IEEE arithmetic compatibility.
9841 This option is not turned on by any @option{-O} option since
9842 it can result in incorrect output for programs that depend on
9843 an exact implementation of IEEE or ISO rules/specifications for
9844 math functions. It may, however, yield faster code for programs
9845 that do not require the guarantees of these specifications.
9847 The default is @option{-fmath-errno}.
9849 On Darwin systems, the math library never sets @code{errno}. There is
9850 therefore no reason for the compiler to consider the possibility that
9851 it might, and @option{-fno-math-errno} is the default.
9853 @item -funsafe-math-optimizations
9854 @opindex funsafe-math-optimizations
9856 Allow optimizations for floating-point arithmetic that (a) assume
9857 that arguments and results are valid and (b) may violate IEEE or
9858 ANSI standards. When used at link-time, it may include libraries
9859 or startup files that change the default FPU control word or other
9860 similar optimizations.
9862 This option is not turned on by any @option{-O} option since
9863 it can result in incorrect output for programs that depend on
9864 an exact implementation of IEEE or ISO rules/specifications for
9865 math functions. It may, however, yield faster code for programs
9866 that do not require the guarantees of these specifications.
9867 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9868 @option{-fassociative-math} and @option{-freciprocal-math}.
9870 The default is @option{-fno-unsafe-math-optimizations}.
9872 @item -fassociative-math
9873 @opindex fassociative-math
9875 Allow re-association of operands in series of floating-point operations.
9876 This violates the ISO C and C++ language standard by possibly changing
9877 computation result. NOTE: re-ordering may change the sign of zero as
9878 well as ignore NaNs and inhibit or create underflow or overflow (and
9879 thus cannot be used on code that relies on rounding behavior like
9880 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9881 and thus may not be used when ordered comparisons are required.
9882 This option requires that both @option{-fno-signed-zeros} and
9883 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9884 much sense with @option{-frounding-math}. For Fortran the option
9885 is automatically enabled when both @option{-fno-signed-zeros} and
9886 @option{-fno-trapping-math} are in effect.
9888 The default is @option{-fno-associative-math}.
9890 @item -freciprocal-math
9891 @opindex freciprocal-math
9893 Allow the reciprocal of a value to be used instead of dividing by
9894 the value if this enables optimizations. For example @code{x / y}
9895 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
9896 is subject to common subexpression elimination. Note that this loses
9897 precision and increases the number of flops operating on the value.
9899 The default is @option{-fno-reciprocal-math}.
9901 @item -ffinite-math-only
9902 @opindex ffinite-math-only
9903 Allow optimizations for floating-point arithmetic that assume
9904 that arguments and results are not NaNs or +-Infs.
9906 This option is not turned on by any @option{-O} option since
9907 it can result in incorrect output for programs that depend on
9908 an exact implementation of IEEE or ISO rules/specifications for
9909 math functions. It may, however, yield faster code for programs
9910 that do not require the guarantees of these specifications.
9912 The default is @option{-fno-finite-math-only}.
9914 @item -fno-signed-zeros
9915 @opindex fno-signed-zeros
9916 Allow optimizations for floating-point arithmetic that ignore the
9917 signedness of zero. IEEE arithmetic specifies the behavior of
9918 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
9919 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
9920 This option implies that the sign of a zero result isn't significant.
9922 The default is @option{-fsigned-zeros}.
9924 @item -fno-trapping-math
9925 @opindex fno-trapping-math
9926 Compile code assuming that floating-point operations cannot generate
9927 user-visible traps. These traps include division by zero, overflow,
9928 underflow, inexact result and invalid operation. This option requires
9929 that @option{-fno-signaling-nans} be in effect. Setting this option may
9930 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
9932 This option should never be turned on by any @option{-O} option since
9933 it can result in incorrect output for programs that depend on
9934 an exact implementation of IEEE or ISO rules/specifications for
9937 The default is @option{-ftrapping-math}.
9939 @item -frounding-math
9940 @opindex frounding-math
9941 Disable transformations and optimizations that assume default floating-point
9942 rounding behavior. This is round-to-zero for all floating point
9943 to integer conversions, and round-to-nearest for all other arithmetic
9944 truncations. This option should be specified for programs that change
9945 the FP rounding mode dynamically, or that may be executed with a
9946 non-default rounding mode. This option disables constant folding of
9947 floating-point expressions at compile time (which may be affected by
9948 rounding mode) and arithmetic transformations that are unsafe in the
9949 presence of sign-dependent rounding modes.
9951 The default is @option{-fno-rounding-math}.
9953 This option is experimental and does not currently guarantee to
9954 disable all GCC optimizations that are affected by rounding mode.
9955 Future versions of GCC may provide finer control of this setting
9956 using C99's @code{FENV_ACCESS} pragma. This command-line option
9957 will be used to specify the default state for @code{FENV_ACCESS}.
9959 @item -fsignaling-nans
9960 @opindex fsignaling-nans
9961 Compile code assuming that IEEE signaling NaNs may generate user-visible
9962 traps during floating-point operations. Setting this option disables
9963 optimizations that may change the number of exceptions visible with
9964 signaling NaNs. This option implies @option{-ftrapping-math}.
9966 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
9969 The default is @option{-fno-signaling-nans}.
9971 This option is experimental and does not currently guarantee to
9972 disable all GCC optimizations that affect signaling NaN behavior.
9974 @item -fsingle-precision-constant
9975 @opindex fsingle-precision-constant
9976 Treat floating-point constants as single precision instead of
9977 implicitly converting them to double-precision constants.
9979 @item -fcx-limited-range
9980 @opindex fcx-limited-range
9981 When enabled, this option states that a range reduction step is not
9982 needed when performing complex division. Also, there is no checking
9983 whether the result of a complex multiplication or division is @code{NaN
9984 + I*NaN}, with an attempt to rescue the situation in that case. The
9985 default is @option{-fno-cx-limited-range}, but is enabled by
9986 @option{-ffast-math}.
9988 This option controls the default setting of the ISO C99
9989 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
9992 @item -fcx-fortran-rules
9993 @opindex fcx-fortran-rules
9994 Complex multiplication and division follow Fortran rules. Range
9995 reduction is done as part of complex division, but there is no checking
9996 whether the result of a complex multiplication or division is @code{NaN
9997 + I*NaN}, with an attempt to rescue the situation in that case.
9999 The default is @option{-fno-cx-fortran-rules}.
10003 The following options control optimizations that may improve
10004 performance, but are not enabled by any @option{-O} options. This
10005 section includes experimental options that may produce broken code.
10008 @item -fbranch-probabilities
10009 @opindex fbranch-probabilities
10010 After running a program compiled with @option{-fprofile-arcs}
10011 (@pxref{Debugging Options,, Options for Debugging Your Program or
10012 @command{gcc}}), you can compile it a second time using
10013 @option{-fbranch-probabilities}, to improve optimizations based on
10014 the number of times each branch was taken. When a program
10015 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10016 counts to a file called @file{@var{sourcename}.gcda} for each source
10017 file. The information in this data file is very dependent on the
10018 structure of the generated code, so you must use the same source code
10019 and the same optimization options for both compilations.
10021 With @option{-fbranch-probabilities}, GCC puts a
10022 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10023 These can be used to improve optimization. Currently, they are only
10024 used in one place: in @file{reorg.c}, instead of guessing which path a
10025 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10026 exactly determine which path is taken more often.
10028 @item -fprofile-values
10029 @opindex fprofile-values
10030 If combined with @option{-fprofile-arcs}, it adds code so that some
10031 data about values of expressions in the program is gathered.
10033 With @option{-fbranch-probabilities}, it reads back the data gathered
10034 from profiling values of expressions for usage in optimizations.
10036 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
10038 @item -fprofile-reorder-functions
10039 @opindex fprofile-reorder-functions
10040 Function reordering based on profile instrumentation collects
10041 first time of execution of a function and orders these functions
10042 in ascending order.
10044 Enabled with @option{-fprofile-use}.
10048 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10049 to add code to gather information about values of expressions.
10051 With @option{-fbranch-probabilities}, it reads back the data gathered
10052 and actually performs the optimizations based on them.
10053 Currently the optimizations include specialization of division operations
10054 using the knowledge about the value of the denominator.
10056 @item -frename-registers
10057 @opindex frename-registers
10058 Attempt to avoid false dependencies in scheduled code by making use
10059 of registers left over after register allocation. This optimization
10060 most benefits processors with lots of registers. Depending on the
10061 debug information format adopted by the target, however, it can
10062 make debugging impossible, since variables no longer stay in
10063 a ``home register''.
10065 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
10067 @item -fschedule-fusion
10068 @opindex fschedule-fusion
10069 Performs a target dependent pass over the instruction stream to schedule
10070 instructions of same type together because target machine can execute them
10071 more efficiently if they are adjacent to each other in the instruction flow.
10073 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10077 Perform tail duplication to enlarge superblock size. This transformation
10078 simplifies the control flow of the function allowing other optimizations to do
10081 Enabled with @option{-fprofile-use}.
10083 @item -funroll-loops
10084 @opindex funroll-loops
10085 Unroll loops whose number of iterations can be determined at compile time or
10086 upon entry to the loop. @option{-funroll-loops} implies
10087 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10088 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10089 a small constant number of iterations). This option makes code larger, and may
10090 or may not make it run faster.
10092 Enabled with @option{-fprofile-use}.
10094 @item -funroll-all-loops
10095 @opindex funroll-all-loops
10096 Unroll all loops, even if their number of iterations is uncertain when
10097 the loop is entered. This usually makes programs run more slowly.
10098 @option{-funroll-all-loops} implies the same options as
10099 @option{-funroll-loops}.
10102 @opindex fpeel-loops
10103 Peels loops for which there is enough information that they do not
10104 roll much (from profile feedback). It also turns on complete loop peeling
10105 (i.e.@: complete removal of loops with small constant number of iterations).
10107 Enabled with @option{-fprofile-use}.
10109 @item -fmove-loop-invariants
10110 @opindex fmove-loop-invariants
10111 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10112 at level @option{-O1}
10114 @item -funswitch-loops
10115 @opindex funswitch-loops
10116 Move branches with loop invariant conditions out of the loop, with duplicates
10117 of the loop on both branches (modified according to result of the condition).
10119 @item -ffunction-sections
10120 @itemx -fdata-sections
10121 @opindex ffunction-sections
10122 @opindex fdata-sections
10123 Place each function or data item into its own section in the output
10124 file if the target supports arbitrary sections. The name of the
10125 function or the name of the data item determines the section's name
10126 in the output file.
10128 Use these options on systems where the linker can perform optimizations
10129 to improve locality of reference in the instruction space. Most systems
10130 using the ELF object format and SPARC processors running Solaris 2 have
10131 linkers with such optimizations. AIX may have these optimizations in
10134 Only use these options when there are significant benefits from doing
10135 so. When you specify these options, the assembler and linker
10136 create larger object and executable files and are also slower.
10137 You cannot use @command{gprof} on all systems if you
10138 specify this option, and you may have problems with debugging if
10139 you specify both this option and @option{-g}.
10141 @item -fbranch-target-load-optimize
10142 @opindex fbranch-target-load-optimize
10143 Perform branch target register load optimization before prologue / epilogue
10145 The use of target registers can typically be exposed only during reload,
10146 thus hoisting loads out of loops and doing inter-block scheduling needs
10147 a separate optimization pass.
10149 @item -fbranch-target-load-optimize2
10150 @opindex fbranch-target-load-optimize2
10151 Perform branch target register load optimization after prologue / epilogue
10154 @item -fbtr-bb-exclusive
10155 @opindex fbtr-bb-exclusive
10156 When performing branch target register load optimization, don't reuse
10157 branch target registers within any basic block.
10159 @item -fstack-protector
10160 @opindex fstack-protector
10161 Emit extra code to check for buffer overflows, such as stack smashing
10162 attacks. This is done by adding a guard variable to functions with
10163 vulnerable objects. This includes functions that call @code{alloca}, and
10164 functions with buffers larger than 8 bytes. The guards are initialized
10165 when a function is entered and then checked when the function exits.
10166 If a guard check fails, an error message is printed and the program exits.
10168 @item -fstack-protector-all
10169 @opindex fstack-protector-all
10170 Like @option{-fstack-protector} except that all functions are protected.
10172 @item -fstack-protector-strong
10173 @opindex fstack-protector-strong
10174 Like @option{-fstack-protector} but includes additional functions to
10175 be protected --- those that have local array definitions, or have
10176 references to local frame addresses.
10178 @item -fstack-protector-explicit
10179 @opindex fstack-protector-explicit
10180 Like @option{-fstack-protector} but only protects those functions which
10181 have the @code{stack_protect} attribute
10184 @opindex fstdarg-opt
10185 Optimize the prologue of variadic argument functions with respect to usage of
10188 @item -fsection-anchors
10189 @opindex fsection-anchors
10190 Try to reduce the number of symbolic address calculations by using
10191 shared ``anchor'' symbols to address nearby objects. This transformation
10192 can help to reduce the number of GOT entries and GOT accesses on some
10195 For example, the implementation of the following function @code{foo}:
10198 static int a, b, c;
10199 int foo (void) @{ return a + b + c; @}
10203 usually calculates the addresses of all three variables, but if you
10204 compile it with @option{-fsection-anchors}, it accesses the variables
10205 from a common anchor point instead. The effect is similar to the
10206 following pseudocode (which isn't valid C):
10211 register int *xr = &x;
10212 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10216 Not all targets support this option.
10218 @item --param @var{name}=@var{value}
10220 In some places, GCC uses various constants to control the amount of
10221 optimization that is done. For example, GCC does not inline functions
10222 that contain more than a certain number of instructions. You can
10223 control some of these constants on the command line using the
10224 @option{--param} option.
10226 The names of specific parameters, and the meaning of the values, are
10227 tied to the internals of the compiler, and are subject to change
10228 without notice in future releases.
10230 In each case, the @var{value} is an integer. The allowable choices for
10234 @item predictable-branch-outcome
10235 When branch is predicted to be taken with probability lower than this threshold
10236 (in percent), then it is considered well predictable. The default is 10.
10238 @item max-crossjump-edges
10239 The maximum number of incoming edges to consider for cross-jumping.
10240 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10241 the number of edges incoming to each block. Increasing values mean
10242 more aggressive optimization, making the compilation time increase with
10243 probably small improvement in executable size.
10245 @item min-crossjump-insns
10246 The minimum number of instructions that must be matched at the end
10247 of two blocks before cross-jumping is performed on them. This
10248 value is ignored in the case where all instructions in the block being
10249 cross-jumped from are matched. The default value is 5.
10251 @item max-grow-copy-bb-insns
10252 The maximum code size expansion factor when copying basic blocks
10253 instead of jumping. The expansion is relative to a jump instruction.
10254 The default value is 8.
10256 @item max-goto-duplication-insns
10257 The maximum number of instructions to duplicate to a block that jumps
10258 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10259 passes, GCC factors computed gotos early in the compilation process,
10260 and unfactors them as late as possible. Only computed jumps at the
10261 end of a basic blocks with no more than max-goto-duplication-insns are
10262 unfactored. The default value is 8.
10264 @item max-delay-slot-insn-search
10265 The maximum number of instructions to consider when looking for an
10266 instruction to fill a delay slot. If more than this arbitrary number of
10267 instructions are searched, the time savings from filling the delay slot
10268 are minimal, so stop searching. Increasing values mean more
10269 aggressive optimization, making the compilation time increase with probably
10270 small improvement in execution time.
10272 @item max-delay-slot-live-search
10273 When trying to fill delay slots, the maximum number of instructions to
10274 consider when searching for a block with valid live register
10275 information. Increasing this arbitrarily chosen value means more
10276 aggressive optimization, increasing the compilation time. This parameter
10277 should be removed when the delay slot code is rewritten to maintain the
10278 control-flow graph.
10280 @item max-gcse-memory
10281 The approximate maximum amount of memory that can be allocated in
10282 order to perform the global common subexpression elimination
10283 optimization. If more memory than specified is required, the
10284 optimization is not done.
10286 @item max-gcse-insertion-ratio
10287 If the ratio of expression insertions to deletions is larger than this value
10288 for any expression, then RTL PRE inserts or removes the expression and thus
10289 leaves partially redundant computations in the instruction stream. The default value is 20.
10291 @item max-pending-list-length
10292 The maximum number of pending dependencies scheduling allows
10293 before flushing the current state and starting over. Large functions
10294 with few branches or calls can create excessively large lists which
10295 needlessly consume memory and resources.
10297 @item max-modulo-backtrack-attempts
10298 The maximum number of backtrack attempts the scheduler should make
10299 when modulo scheduling a loop. Larger values can exponentially increase
10302 @item max-inline-insns-single
10303 Several parameters control the tree inliner used in GCC@.
10304 This number sets the maximum number of instructions (counted in GCC's
10305 internal representation) in a single function that the tree inliner
10306 considers for inlining. This only affects functions declared
10307 inline and methods implemented in a class declaration (C++).
10308 The default value is 400.
10310 @item max-inline-insns-auto
10311 When you use @option{-finline-functions} (included in @option{-O3}),
10312 a lot of functions that would otherwise not be considered for inlining
10313 by the compiler are investigated. To those functions, a different
10314 (more restrictive) limit compared to functions declared inline can
10316 The default value is 40.
10318 @item inline-min-speedup
10319 When estimated performance improvement of caller + callee runtime exceeds this
10320 threshold (in precent), the function can be inlined regardless the limit on
10321 @option{--param max-inline-insns-single} and @option{--param
10322 max-inline-insns-auto}.
10324 @item large-function-insns
10325 The limit specifying really large functions. For functions larger than this
10326 limit after inlining, inlining is constrained by
10327 @option{--param large-function-growth}. This parameter is useful primarily
10328 to avoid extreme compilation time caused by non-linear algorithms used by the
10330 The default value is 2700.
10332 @item large-function-growth
10333 Specifies maximal growth of large function caused by inlining in percents.
10334 The default value is 100 which limits large function growth to 2.0 times
10337 @item large-unit-insns
10338 The limit specifying large translation unit. Growth caused by inlining of
10339 units larger than this limit is limited by @option{--param inline-unit-growth}.
10340 For small units this might be too tight.
10341 For example, consider a unit consisting of function A
10342 that is inline and B that just calls A three times. If B is small relative to
10343 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10344 large units consisting of small inlineable functions, however, the overall unit
10345 growth limit is needed to avoid exponential explosion of code size. Thus for
10346 smaller units, the size is increased to @option{--param large-unit-insns}
10347 before applying @option{--param inline-unit-growth}. The default is 10000.
10349 @item inline-unit-growth
10350 Specifies maximal overall growth of the compilation unit caused by inlining.
10351 The default value is 20 which limits unit growth to 1.2 times the original
10352 size. Cold functions (either marked cold via an attribute or by profile
10353 feedback) are not accounted into the unit size.
10355 @item ipcp-unit-growth
10356 Specifies maximal overall growth of the compilation unit caused by
10357 interprocedural constant propagation. The default value is 10 which limits
10358 unit growth to 1.1 times the original size.
10360 @item large-stack-frame
10361 The limit specifying large stack frames. While inlining the algorithm is trying
10362 to not grow past this limit too much. The default value is 256 bytes.
10364 @item large-stack-frame-growth
10365 Specifies maximal growth of large stack frames caused by inlining in percents.
10366 The default value is 1000 which limits large stack frame growth to 11 times
10369 @item max-inline-insns-recursive
10370 @itemx max-inline-insns-recursive-auto
10371 Specifies the maximum number of instructions an out-of-line copy of a
10372 self-recursive inline
10373 function can grow into by performing recursive inlining.
10375 @option{--param max-inline-insns-recursive} applies to functions
10377 For functions not declared inline, recursive inlining
10378 happens only when @option{-finline-functions} (included in @option{-O3}) is
10379 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10380 default value is 450.
10382 @item max-inline-recursive-depth
10383 @itemx max-inline-recursive-depth-auto
10384 Specifies the maximum recursion depth used for recursive inlining.
10386 @option{--param max-inline-recursive-depth} applies to functions
10387 declared inline. For functions not declared inline, recursive inlining
10388 happens only when @option{-finline-functions} (included in @option{-O3}) is
10389 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10390 default value is 8.
10392 @item min-inline-recursive-probability
10393 Recursive inlining is profitable only for function having deep recursion
10394 in average and can hurt for function having little recursion depth by
10395 increasing the prologue size or complexity of function body to other
10398 When profile feedback is available (see @option{-fprofile-generate}) the actual
10399 recursion depth can be guessed from probability that function recurses via a
10400 given call expression. This parameter limits inlining only to call expressions
10401 whose probability exceeds the given threshold (in percents).
10402 The default value is 10.
10404 @item early-inlining-insns
10405 Specify growth that the early inliner can make. In effect it increases
10406 the amount of inlining for code having a large abstraction penalty.
10407 The default value is 14.
10409 @item max-early-inliner-iterations
10410 Limit of iterations of the early inliner. This basically bounds
10411 the number of nested indirect calls the early inliner can resolve.
10412 Deeper chains are still handled by late inlining.
10414 @item comdat-sharing-probability
10415 Probability (in percent) that C++ inline function with comdat visibility
10416 are shared across multiple compilation units. The default value is 20.
10418 @item profile-func-internal-id
10419 A parameter to control whether to use function internal id in profile
10420 database lookup. If the value is 0, the compiler uses an id that
10421 is based on function assembler name and filename, which makes old profile
10422 data more tolerant to source changes such as function reordering etc.
10423 The default value is 0.
10425 @item min-vect-loop-bound
10426 The minimum number of iterations under which loops are not vectorized
10427 when @option{-ftree-vectorize} is used. The number of iterations after
10428 vectorization needs to be greater than the value specified by this option
10429 to allow vectorization. The default value is 0.
10431 @item gcse-cost-distance-ratio
10432 Scaling factor in calculation of maximum distance an expression
10433 can be moved by GCSE optimizations. This is currently supported only in the
10434 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10435 is with simple expressions, i.e., the expressions that have cost
10436 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10437 hoisting of simple expressions. The default value is 10.
10439 @item gcse-unrestricted-cost
10440 Cost, roughly measured as the cost of a single typical machine
10441 instruction, at which GCSE optimizations do not constrain
10442 the distance an expression can travel. This is currently
10443 supported only in the code hoisting pass. The lesser the cost,
10444 the more aggressive code hoisting is. Specifying 0
10445 allows all expressions to travel unrestricted distances.
10446 The default value is 3.
10448 @item max-hoist-depth
10449 The depth of search in the dominator tree for expressions to hoist.
10450 This is used to avoid quadratic behavior in hoisting algorithm.
10451 The value of 0 does not limit on the search, but may slow down compilation
10452 of huge functions. The default value is 30.
10454 @item max-tail-merge-comparisons
10455 The maximum amount of similar bbs to compare a bb with. This is used to
10456 avoid quadratic behavior in tree tail merging. The default value is 10.
10458 @item max-tail-merge-iterations
10459 The maximum amount of iterations of the pass over the function. This is used to
10460 limit compilation time in tree tail merging. The default value is 2.
10462 @item max-unrolled-insns
10463 The maximum number of instructions that a loop may have to be unrolled.
10464 If a loop is unrolled, this parameter also determines how many times
10465 the loop code is unrolled.
10467 @item max-average-unrolled-insns
10468 The maximum number of instructions biased by probabilities of their execution
10469 that a loop may have to be unrolled. If a loop is unrolled,
10470 this parameter also determines how many times the loop code is unrolled.
10472 @item max-unroll-times
10473 The maximum number of unrollings of a single loop.
10475 @item max-peeled-insns
10476 The maximum number of instructions that a loop may have to be peeled.
10477 If a loop is peeled, this parameter also determines how many times
10478 the loop code is peeled.
10480 @item max-peel-times
10481 The maximum number of peelings of a single loop.
10483 @item max-peel-branches
10484 The maximum number of branches on the hot path through the peeled sequence.
10486 @item max-completely-peeled-insns
10487 The maximum number of insns of a completely peeled loop.
10489 @item max-completely-peel-times
10490 The maximum number of iterations of a loop to be suitable for complete peeling.
10492 @item max-completely-peel-loop-nest-depth
10493 The maximum depth of a loop nest suitable for complete peeling.
10495 @item max-unswitch-insns
10496 The maximum number of insns of an unswitched loop.
10498 @item max-unswitch-level
10499 The maximum number of branches unswitched in a single loop.
10501 @item lim-expensive
10502 The minimum cost of an expensive expression in the loop invariant motion.
10504 @item iv-consider-all-candidates-bound
10505 Bound on number of candidates for induction variables, below which
10506 all candidates are considered for each use in induction variable
10507 optimizations. If there are more candidates than this,
10508 only the most relevant ones are considered to avoid quadratic time complexity.
10510 @item iv-max-considered-uses
10511 The induction variable optimizations give up on loops that contain more
10512 induction variable uses.
10514 @item iv-always-prune-cand-set-bound
10515 If the number of candidates in the set is smaller than this value,
10516 always try to remove unnecessary ivs from the set
10517 when adding a new one.
10519 @item scev-max-expr-size
10520 Bound on size of expressions used in the scalar evolutions analyzer.
10521 Large expressions slow the analyzer.
10523 @item scev-max-expr-complexity
10524 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10525 Complex expressions slow the analyzer.
10527 @item vect-max-version-for-alignment-checks
10528 The maximum number of run-time checks that can be performed when
10529 doing loop versioning for alignment in the vectorizer.
10531 @item vect-max-version-for-alias-checks
10532 The maximum number of run-time checks that can be performed when
10533 doing loop versioning for alias in the vectorizer.
10535 @item vect-max-peeling-for-alignment
10536 The maximum number of loop peels to enhance access alignment
10537 for vectorizer. Value -1 means 'no limit'.
10539 @item max-iterations-to-track
10540 The maximum number of iterations of a loop the brute-force algorithm
10541 for analysis of the number of iterations of the loop tries to evaluate.
10543 @item hot-bb-count-ws-permille
10544 A basic block profile count is considered hot if it contributes to
10545 the given permillage (i.e. 0...1000) of the entire profiled execution.
10547 @item hot-bb-frequency-fraction
10548 Select fraction of the entry block frequency of executions of basic block in
10549 function given basic block needs to have to be considered hot.
10551 @item max-predicted-iterations
10552 The maximum number of loop iterations we predict statically. This is useful
10553 in cases where a function contains a single loop with known bound and
10554 another loop with unknown bound.
10555 The known number of iterations is predicted correctly, while
10556 the unknown number of iterations average to roughly 10. This means that the
10557 loop without bounds appears artificially cold relative to the other one.
10559 @item builtin-expect-probability
10560 Control the probability of the expression having the specified value. This
10561 parameter takes a percentage (i.e. 0 ... 100) as input.
10562 The default probability of 90 is obtained empirically.
10564 @item align-threshold
10566 Select fraction of the maximal frequency of executions of a basic block in
10567 a function to align the basic block.
10569 @item align-loop-iterations
10571 A loop expected to iterate at least the selected number of iterations is
10574 @item tracer-dynamic-coverage
10575 @itemx tracer-dynamic-coverage-feedback
10577 This value is used to limit superblock formation once the given percentage of
10578 executed instructions is covered. This limits unnecessary code size
10581 The @option{tracer-dynamic-coverage-feedback} parameter
10582 is used only when profile
10583 feedback is available. The real profiles (as opposed to statically estimated
10584 ones) are much less balanced allowing the threshold to be larger value.
10586 @item tracer-max-code-growth
10587 Stop tail duplication once code growth has reached given percentage. This is
10588 a rather artificial limit, as most of the duplicates are eliminated later in
10589 cross jumping, so it may be set to much higher values than is the desired code
10592 @item tracer-min-branch-ratio
10594 Stop reverse growth when the reverse probability of best edge is less than this
10595 threshold (in percent).
10597 @item tracer-min-branch-ratio
10598 @itemx tracer-min-branch-ratio-feedback
10600 Stop forward growth if the best edge has probability lower than this
10603 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10604 compilation for profile feedback and one for compilation without. The value
10605 for compilation with profile feedback needs to be more conservative (higher) in
10606 order to make tracer effective.
10608 @item max-cse-path-length
10610 The maximum number of basic blocks on path that CSE considers.
10613 @item max-cse-insns
10614 The maximum number of instructions CSE processes before flushing.
10615 The default is 1000.
10617 @item ggc-min-expand
10619 GCC uses a garbage collector to manage its own memory allocation. This
10620 parameter specifies the minimum percentage by which the garbage
10621 collector's heap should be allowed to expand between collections.
10622 Tuning this may improve compilation speed; it has no effect on code
10625 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10626 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10627 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10628 GCC is not able to calculate RAM on a particular platform, the lower
10629 bound of 30% is used. Setting this parameter and
10630 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10631 every opportunity. This is extremely slow, but can be useful for
10634 @item ggc-min-heapsize
10636 Minimum size of the garbage collector's heap before it begins bothering
10637 to collect garbage. The first collection occurs after the heap expands
10638 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10639 tuning this may improve compilation speed, and has no effect on code
10642 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10643 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10644 with a lower bound of 4096 (four megabytes) and an upper bound of
10645 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10646 particular platform, the lower bound is used. Setting this parameter
10647 very large effectively disables garbage collection. Setting this
10648 parameter and @option{ggc-min-expand} to zero causes a full collection
10649 to occur at every opportunity.
10651 @item max-reload-search-insns
10652 The maximum number of instruction reload should look backward for equivalent
10653 register. Increasing values mean more aggressive optimization, making the
10654 compilation time increase with probably slightly better performance.
10655 The default value is 100.
10657 @item max-cselib-memory-locations
10658 The maximum number of memory locations cselib should take into account.
10659 Increasing values mean more aggressive optimization, making the compilation time
10660 increase with probably slightly better performance. The default value is 500.
10662 @item reorder-blocks-duplicate
10663 @itemx reorder-blocks-duplicate-feedback
10665 Used by the basic block reordering pass to decide whether to use unconditional
10666 branch or duplicate the code on its destination. Code is duplicated when its
10667 estimated size is smaller than this value multiplied by the estimated size of
10668 unconditional jump in the hot spots of the program.
10670 The @option{reorder-block-duplicate-feedback} parameter
10671 is used only when profile
10672 feedback is available. It may be set to higher values than
10673 @option{reorder-block-duplicate} since information about the hot spots is more
10676 @item max-sched-ready-insns
10677 The maximum number of instructions ready to be issued the scheduler should
10678 consider at any given time during the first scheduling pass. Increasing
10679 values mean more thorough searches, making the compilation time increase
10680 with probably little benefit. The default value is 100.
10682 @item max-sched-region-blocks
10683 The maximum number of blocks in a region to be considered for
10684 interblock scheduling. The default value is 10.
10686 @item max-pipeline-region-blocks
10687 The maximum number of blocks in a region to be considered for
10688 pipelining in the selective scheduler. The default value is 15.
10690 @item max-sched-region-insns
10691 The maximum number of insns in a region to be considered for
10692 interblock scheduling. The default value is 100.
10694 @item max-pipeline-region-insns
10695 The maximum number of insns in a region to be considered for
10696 pipelining in the selective scheduler. The default value is 200.
10698 @item min-spec-prob
10699 The minimum probability (in percents) of reaching a source block
10700 for interblock speculative scheduling. The default value is 40.
10702 @item max-sched-extend-regions-iters
10703 The maximum number of iterations through CFG to extend regions.
10704 A value of 0 (the default) disables region extensions.
10706 @item max-sched-insn-conflict-delay
10707 The maximum conflict delay for an insn to be considered for speculative motion.
10708 The default value is 3.
10710 @item sched-spec-prob-cutoff
10711 The minimal probability of speculation success (in percents), so that
10712 speculative insns are scheduled.
10713 The default value is 40.
10715 @item sched-spec-state-edge-prob-cutoff
10716 The minimum probability an edge must have for the scheduler to save its
10718 The default value is 10.
10720 @item sched-mem-true-dep-cost
10721 Minimal distance (in CPU cycles) between store and load targeting same
10722 memory locations. The default value is 1.
10724 @item selsched-max-lookahead
10725 The maximum size of the lookahead window of selective scheduling. It is a
10726 depth of search for available instructions.
10727 The default value is 50.
10729 @item selsched-max-sched-times
10730 The maximum number of times that an instruction is scheduled during
10731 selective scheduling. This is the limit on the number of iterations
10732 through which the instruction may be pipelined. The default value is 2.
10734 @item selsched-max-insns-to-rename
10735 The maximum number of best instructions in the ready list that are considered
10736 for renaming in the selective scheduler. The default value is 2.
10739 The minimum value of stage count that swing modulo scheduler
10740 generates. The default value is 2.
10742 @item max-last-value-rtl
10743 The maximum size measured as number of RTLs that can be recorded in an expression
10744 in combiner for a pseudo register as last known value of that register. The default
10747 @item max-combine-insns
10748 The maximum number of instructions the RTL combiner tries to combine.
10749 The default value is 2 at @option{-Og} and 4 otherwise.
10751 @item integer-share-limit
10752 Small integer constants can use a shared data structure, reducing the
10753 compiler's memory usage and increasing its speed. This sets the maximum
10754 value of a shared integer constant. The default value is 256.
10756 @item ssp-buffer-size
10757 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10758 protection when @option{-fstack-protection} is used.
10760 @item min-size-for-stack-sharing
10761 The minimum size of variables taking part in stack slot sharing when not
10762 optimizing. The default value is 32.
10764 @item max-jump-thread-duplication-stmts
10765 Maximum number of statements allowed in a block that needs to be
10766 duplicated when threading jumps.
10768 @item max-fields-for-field-sensitive
10769 Maximum number of fields in a structure treated in
10770 a field sensitive manner during pointer analysis. The default is zero
10771 for @option{-O0} and @option{-O1},
10772 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10774 @item prefetch-latency
10775 Estimate on average number of instructions that are executed before
10776 prefetch finishes. The distance prefetched ahead is proportional
10777 to this constant. Increasing this number may also lead to less
10778 streams being prefetched (see @option{simultaneous-prefetches}).
10780 @item simultaneous-prefetches
10781 Maximum number of prefetches that can run at the same time.
10783 @item l1-cache-line-size
10784 The size of cache line in L1 cache, in bytes.
10786 @item l1-cache-size
10787 The size of L1 cache, in kilobytes.
10789 @item l2-cache-size
10790 The size of L2 cache, in kilobytes.
10792 @item min-insn-to-prefetch-ratio
10793 The minimum ratio between the number of instructions and the
10794 number of prefetches to enable prefetching in a loop.
10796 @item prefetch-min-insn-to-mem-ratio
10797 The minimum ratio between the number of instructions and the
10798 number of memory references to enable prefetching in a loop.
10800 @item use-canonical-types
10801 Whether the compiler should use the ``canonical'' type system. By
10802 default, this should always be 1, which uses a more efficient internal
10803 mechanism for comparing types in C++ and Objective-C++. However, if
10804 bugs in the canonical type system are causing compilation failures,
10805 set this value to 0 to disable canonical types.
10807 @item switch-conversion-max-branch-ratio
10808 Switch initialization conversion refuses to create arrays that are
10809 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10810 branches in the switch.
10812 @item max-partial-antic-length
10813 Maximum length of the partial antic set computed during the tree
10814 partial redundancy elimination optimization (@option{-ftree-pre}) when
10815 optimizing at @option{-O3} and above. For some sorts of source code
10816 the enhanced partial redundancy elimination optimization can run away,
10817 consuming all of the memory available on the host machine. This
10818 parameter sets a limit on the length of the sets that are computed,
10819 which prevents the runaway behavior. Setting a value of 0 for
10820 this parameter allows an unlimited set length.
10822 @item sccvn-max-scc-size
10823 Maximum size of a strongly connected component (SCC) during SCCVN
10824 processing. If this limit is hit, SCCVN processing for the whole
10825 function is not done and optimizations depending on it are
10826 disabled. The default maximum SCC size is 10000.
10828 @item sccvn-max-alias-queries-per-access
10829 Maximum number of alias-oracle queries we perform when looking for
10830 redundancies for loads and stores. If this limit is hit the search
10831 is aborted and the load or store is not considered redundant. The
10832 number of queries is algorithmically limited to the number of
10833 stores on all paths from the load to the function entry.
10834 The default maxmimum number of queries is 1000.
10836 @item ira-max-loops-num
10837 IRA uses regional register allocation by default. If a function
10838 contains more loops than the number given by this parameter, only at most
10839 the given number of the most frequently-executed loops form regions
10840 for regional register allocation. The default value of the
10843 @item ira-max-conflict-table-size
10844 Although IRA uses a sophisticated algorithm to compress the conflict
10845 table, the table can still require excessive amounts of memory for
10846 huge functions. If the conflict table for a function could be more
10847 than the size in MB given by this parameter, the register allocator
10848 instead uses a faster, simpler, and lower-quality
10849 algorithm that does not require building a pseudo-register conflict table.
10850 The default value of the parameter is 2000.
10852 @item ira-loop-reserved-regs
10853 IRA can be used to evaluate more accurate register pressure in loops
10854 for decisions to move loop invariants (see @option{-O3}). The number
10855 of available registers reserved for some other purposes is given
10856 by this parameter. The default value of the parameter is 2, which is
10857 the minimal number of registers needed by typical instructions.
10858 This value is the best found from numerous experiments.
10860 @item lra-inheritance-ebb-probability-cutoff
10861 LRA tries to reuse values reloaded in registers in subsequent insns.
10862 This optimization is called inheritance. EBB is used as a region to
10863 do this optimization. The parameter defines a minimal fall-through
10864 edge probability in percentage used to add BB to inheritance EBB in
10865 LRA. The default value of the parameter is 40. The value was chosen
10866 from numerous runs of SPEC2000 on x86-64.
10868 @item loop-invariant-max-bbs-in-loop
10869 Loop invariant motion can be very expensive, both in compilation time and
10870 in amount of needed compile-time memory, with very large loops. Loops
10871 with more basic blocks than this parameter won't have loop invariant
10872 motion optimization performed on them. The default value of the
10873 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10875 @item loop-max-datarefs-for-datadeps
10876 Building data dapendencies is expensive for very large loops. This
10877 parameter limits the number of data references in loops that are
10878 considered for data dependence analysis. These large loops are no
10879 handled by the optimizations using loop data dependencies.
10880 The default value is 1000.
10882 @item max-vartrack-size
10883 Sets a maximum number of hash table slots to use during variable
10884 tracking dataflow analysis of any function. If this limit is exceeded
10885 with variable tracking at assignments enabled, analysis for that
10886 function is retried without it, after removing all debug insns from
10887 the function. If the limit is exceeded even without debug insns, var
10888 tracking analysis is completely disabled for the function. Setting
10889 the parameter to zero makes it unlimited.
10891 @item max-vartrack-expr-depth
10892 Sets a maximum number of recursion levels when attempting to map
10893 variable names or debug temporaries to value expressions. This trades
10894 compilation time for more complete debug information. If this is set too
10895 low, value expressions that are available and could be represented in
10896 debug information may end up not being used; setting this higher may
10897 enable the compiler to find more complex debug expressions, but compile
10898 time and memory use may grow. The default is 12.
10900 @item min-nondebug-insn-uid
10901 Use uids starting at this parameter for nondebug insns. The range below
10902 the parameter is reserved exclusively for debug insns created by
10903 @option{-fvar-tracking-assignments}, but debug insns may get
10904 (non-overlapping) uids above it if the reserved range is exhausted.
10906 @item ipa-sra-ptr-growth-factor
10907 IPA-SRA replaces a pointer to an aggregate with one or more new
10908 parameters only when their cumulative size is less or equal to
10909 @option{ipa-sra-ptr-growth-factor} times the size of the original
10912 @item sra-max-scalarization-size-Ospeed
10913 @item sra-max-scalarization-size-Osize
10914 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
10915 replace scalar parts of aggregates with uses of independent scalar
10916 variables. These parameters control the maximum size, in storage units,
10917 of aggregate which is considered for replacement when compiling for
10919 (@option{sra-max-scalarization-size-Ospeed}) or size
10920 (@option{sra-max-scalarization-size-Osize}) respectively.
10922 @item tm-max-aggregate-size
10923 When making copies of thread-local variables in a transaction, this
10924 parameter specifies the size in bytes after which variables are
10925 saved with the logging functions as opposed to save/restore code
10926 sequence pairs. This option only applies when using
10929 @item graphite-max-nb-scop-params
10930 To avoid exponential effects in the Graphite loop transforms, the
10931 number of parameters in a Static Control Part (SCoP) is bounded. The
10932 default value is 10 parameters. A variable whose value is unknown at
10933 compilation time and defined outside a SCoP is a parameter of the SCoP.
10935 @item graphite-max-bbs-per-function
10936 To avoid exponential effects in the detection of SCoPs, the size of
10937 the functions analyzed by Graphite is bounded. The default value is
10940 @item loop-block-tile-size
10941 Loop blocking or strip mining transforms, enabled with
10942 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
10943 loop in the loop nest by a given number of iterations. The strip
10944 length can be changed using the @option{loop-block-tile-size}
10945 parameter. The default value is 51 iterations.
10947 @item loop-unroll-jam-size
10948 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
10949 default value is 4.
10951 @item loop-unroll-jam-depth
10952 Specify the dimension to be unrolled (counting from the most inner loop)
10953 for the @option{-floop-unroll-and-jam}. The default value is 2.
10955 @item ipa-cp-value-list-size
10956 IPA-CP attempts to track all possible values and types passed to a function's
10957 parameter in order to propagate them and perform devirtualization.
10958 @option{ipa-cp-value-list-size} is the maximum number of values and types it
10959 stores per one formal parameter of a function.
10961 @item ipa-cp-eval-threshold
10962 IPA-CP calculates its own score of cloning profitability heuristics
10963 and performs those cloning opportunities with scores that exceed
10964 @option{ipa-cp-eval-threshold}.
10966 @item ipa-cp-recursion-penalty
10967 Percentage penalty the recursive functions will receive when they
10968 are evaluated for cloning.
10970 @item ipa-cp-single-call-penalty
10971 Percentage penalty functions containg a single call to another
10972 function will receive when they are evaluated for cloning.
10975 @item ipa-max-agg-items
10976 IPA-CP is also capable to propagate a number of scalar values passed
10977 in an aggregate. @option{ipa-max-agg-items} controls the maximum
10978 number of such values per one parameter.
10980 @item ipa-cp-loop-hint-bonus
10981 When IPA-CP determines that a cloning candidate would make the number
10982 of iterations of a loop known, it adds a bonus of
10983 @option{ipa-cp-loop-hint-bonus} to the profitability score of
10986 @item ipa-cp-array-index-hint-bonus
10987 When IPA-CP determines that a cloning candidate would make the index of
10988 an array access known, it adds a bonus of
10989 @option{ipa-cp-array-index-hint-bonus} to the profitability
10990 score of the candidate.
10992 @item ipa-max-aa-steps
10993 During its analysis of function bodies, IPA-CP employs alias analysis
10994 in order to track values pointed to by function parameters. In order
10995 not spend too much time analyzing huge functions, it gives up and
10996 consider all memory clobbered after examining
10997 @option{ipa-max-aa-steps} statements modifying memory.
10999 @item lto-partitions
11000 Specify desired number of partitions produced during WHOPR compilation.
11001 The number of partitions should exceed the number of CPUs used for compilation.
11002 The default value is 32.
11004 @item lto-minpartition
11005 Size of minimal partition for WHOPR (in estimated instructions).
11006 This prevents expenses of splitting very small programs into too many
11009 @item cxx-max-namespaces-for-diagnostic-help
11010 The maximum number of namespaces to consult for suggestions when C++
11011 name lookup fails for an identifier. The default is 1000.
11013 @item sink-frequency-threshold
11014 The maximum relative execution frequency (in percents) of the target block
11015 relative to a statement's original block to allow statement sinking of a
11016 statement. Larger numbers result in more aggressive statement sinking.
11017 The default value is 75. A small positive adjustment is applied for
11018 statements with memory operands as those are even more profitable so sink.
11020 @item max-stores-to-sink
11021 The maximum number of conditional stores paires that can be sunk. Set to 0
11022 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11023 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
11025 @item allow-store-data-races
11026 Allow optimizers to introduce new data races on stores.
11027 Set to 1 to allow, otherwise to 0. This option is enabled by default
11028 at optimization level @option{-Ofast}.
11030 @item case-values-threshold
11031 The smallest number of different values for which it is best to use a
11032 jump-table instead of a tree of conditional branches. If the value is
11033 0, use the default for the machine. The default is 0.
11035 @item tree-reassoc-width
11036 Set the maximum number of instructions executed in parallel in
11037 reassociated tree. This parameter overrides target dependent
11038 heuristics used by default if has non zero value.
11040 @item sched-pressure-algorithm
11041 Choose between the two available implementations of
11042 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11043 and is the more likely to prevent instructions from being reordered.
11044 Algorithm 2 was designed to be a compromise between the relatively
11045 conservative approach taken by algorithm 1 and the rather aggressive
11046 approach taken by the default scheduler. It relies more heavily on
11047 having a regular register file and accurate register pressure classes.
11048 See @file{haifa-sched.c} in the GCC sources for more details.
11050 The default choice depends on the target.
11052 @item max-slsr-cand-scan
11053 Set the maximum number of existing candidates that are considered when
11054 seeking a basis for a new straight-line strength reduction candidate.
11057 Enable buffer overflow detection for global objects. This kind
11058 of protection is enabled by default if you are using
11059 @option{-fsanitize=address} option.
11060 To disable global objects protection use @option{--param asan-globals=0}.
11063 Enable buffer overflow detection for stack objects. This kind of
11064 protection is enabled by default when using@option{-fsanitize=address}.
11065 To disable stack protection use @option{--param asan-stack=0} option.
11067 @item asan-instrument-reads
11068 Enable buffer overflow detection for memory reads. This kind of
11069 protection is enabled by default when using @option{-fsanitize=address}.
11070 To disable memory reads protection use
11071 @option{--param asan-instrument-reads=0}.
11073 @item asan-instrument-writes
11074 Enable buffer overflow detection for memory writes. This kind of
11075 protection is enabled by default when using @option{-fsanitize=address}.
11076 To disable memory writes protection use
11077 @option{--param asan-instrument-writes=0} option.
11079 @item asan-memintrin
11080 Enable detection for built-in functions. This kind of protection
11081 is enabled by default when using @option{-fsanitize=address}.
11082 To disable built-in functions protection use
11083 @option{--param asan-memintrin=0}.
11085 @item asan-use-after-return
11086 Enable detection of use-after-return. This kind of protection
11087 is enabled by default when using @option{-fsanitize=address} option.
11088 To disable use-after-return detection use
11089 @option{--param asan-use-after-return=0}.
11091 @item asan-instrumentation-with-call-threshold
11092 If number of memory accesses in function being instrumented
11093 is greater or equal to this number, use callbacks instead of inline checks.
11094 E.g. to disable inline code use
11095 @option{--param asan-instrumentation-with-call-threshold=0}.
11097 @item chkp-max-ctor-size
11098 Static constructors generated by Pointer Bounds Checker may become very
11099 large and significantly increase compile time at optimization level
11100 @option{-O1} and higher. This parameter is a maximum nubmer of statements
11101 in a single generated constructor. Default value is 5000.
11103 @item max-fsm-thread-path-insns
11104 Maximum number of instructions to copy when duplicating blocks on a
11105 finite state automaton jump thread path. The default is 100.
11107 @item max-fsm-thread-length
11108 Maximum number of basic blocks on a finite state automaton jump thread
11109 path. The default is 10.
11111 @item max-fsm-thread-paths
11112 Maximum number of new jump thread paths to create for a finite state
11113 automaton. The default is 50.
11118 @node Preprocessor Options
11119 @section Options Controlling the Preprocessor
11120 @cindex preprocessor options
11121 @cindex options, preprocessor
11123 These options control the C preprocessor, which is run on each C source
11124 file before actual compilation.
11126 If you use the @option{-E} option, nothing is done except preprocessing.
11127 Some of these options make sense only together with @option{-E} because
11128 they cause the preprocessor output to be unsuitable for actual
11132 @item -Wp,@var{option}
11134 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11135 and pass @var{option} directly through to the preprocessor. If
11136 @var{option} contains commas, it is split into multiple options at the
11137 commas. However, many options are modified, translated or interpreted
11138 by the compiler driver before being passed to the preprocessor, and
11139 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11140 interface is undocumented and subject to change, so whenever possible
11141 you should avoid using @option{-Wp} and let the driver handle the
11144 @item -Xpreprocessor @var{option}
11145 @opindex Xpreprocessor
11146 Pass @var{option} as an option to the preprocessor. You can use this to
11147 supply system-specific preprocessor options that GCC does not
11150 If you want to pass an option that takes an argument, you must use
11151 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11153 @item -no-integrated-cpp
11154 @opindex no-integrated-cpp
11155 Perform preprocessing as a separate pass before compilation.
11156 By default, GCC performs preprocessing as an integrated part of
11157 input tokenization and parsing.
11158 If this option is provided, the appropriate language front end
11159 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11160 and Objective-C, respectively) is instead invoked twice,
11161 once for preprocessing only and once for actual compilation
11162 of the preprocessed input.
11163 This option may be useful in conjunction with the @option{-B} or
11164 @option{-wrapper} options to specify an alternate preprocessor or
11165 perform additional processing of the program source between
11166 normal preprocessing and compilation.
11169 @include cppopts.texi
11171 @node Assembler Options
11172 @section Passing Options to the Assembler
11174 @c prevent bad page break with this line
11175 You can pass options to the assembler.
11178 @item -Wa,@var{option}
11180 Pass @var{option} as an option to the assembler. If @var{option}
11181 contains commas, it is split into multiple options at the commas.
11183 @item -Xassembler @var{option}
11184 @opindex Xassembler
11185 Pass @var{option} as an option to the assembler. You can use this to
11186 supply system-specific assembler options that GCC does not
11189 If you want to pass an option that takes an argument, you must use
11190 @option{-Xassembler} twice, once for the option and once for the argument.
11195 @section Options for Linking
11196 @cindex link options
11197 @cindex options, linking
11199 These options come into play when the compiler links object files into
11200 an executable output file. They are meaningless if the compiler is
11201 not doing a link step.
11205 @item @var{object-file-name}
11206 A file name that does not end in a special recognized suffix is
11207 considered to name an object file or library. (Object files are
11208 distinguished from libraries by the linker according to the file
11209 contents.) If linking is done, these object files are used as input
11218 If any of these options is used, then the linker is not run, and
11219 object file names should not be used as arguments. @xref{Overall
11223 @opindex fuse-ld=bfd
11224 Use the @command{bfd} linker instead of the default linker.
11226 @item -fuse-ld=gold
11227 @opindex fuse-ld=gold
11228 Use the @command{gold} linker instead of the default linker.
11231 @item -l@var{library}
11232 @itemx -l @var{library}
11234 Search the library named @var{library} when linking. (The second
11235 alternative with the library as a separate argument is only for
11236 POSIX compliance and is not recommended.)
11238 It makes a difference where in the command you write this option; the
11239 linker searches and processes libraries and object files in the order they
11240 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11241 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11242 to functions in @samp{z}, those functions may not be loaded.
11244 The linker searches a standard list of directories for the library,
11245 which is actually a file named @file{lib@var{library}.a}. The linker
11246 then uses this file as if it had been specified precisely by name.
11248 The directories searched include several standard system directories
11249 plus any that you specify with @option{-L}.
11251 Normally the files found this way are library files---archive files
11252 whose members are object files. The linker handles an archive file by
11253 scanning through it for members which define symbols that have so far
11254 been referenced but not defined. But if the file that is found is an
11255 ordinary object file, it is linked in the usual fashion. The only
11256 difference between using an @option{-l} option and specifying a file name
11257 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11258 and searches several directories.
11262 You need this special case of the @option{-l} option in order to
11263 link an Objective-C or Objective-C++ program.
11265 @item -nostartfiles
11266 @opindex nostartfiles
11267 Do not use the standard system startup files when linking.
11268 The standard system libraries are used normally, unless @option{-nostdlib}
11269 or @option{-nodefaultlibs} is used.
11271 @item -nodefaultlibs
11272 @opindex nodefaultlibs
11273 Do not use the standard system libraries when linking.
11274 Only the libraries you specify are passed to the linker, and options
11275 specifying linkage of the system libraries, such as @option{-static-libgcc}
11276 or @option{-shared-libgcc}, are ignored.
11277 The standard startup files are used normally, unless @option{-nostartfiles}
11280 The compiler may generate calls to @code{memcmp},
11281 @code{memset}, @code{memcpy} and @code{memmove}.
11282 These entries are usually resolved by entries in
11283 libc. These entry points should be supplied through some other
11284 mechanism when this option is specified.
11288 Do not use the standard system startup files or libraries when linking.
11289 No startup files and only the libraries you specify are passed to
11290 the linker, and options specifying linkage of the system libraries, such as
11291 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11293 The compiler may generate calls to @code{memcmp}, @code{memset},
11294 @code{memcpy} and @code{memmove}.
11295 These entries are usually resolved by entries in
11296 libc. These entry points should be supplied through some other
11297 mechanism when this option is specified.
11299 @cindex @option{-lgcc}, use with @option{-nostdlib}
11300 @cindex @option{-nostdlib} and unresolved references
11301 @cindex unresolved references and @option{-nostdlib}
11302 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11303 @cindex @option{-nodefaultlibs} and unresolved references
11304 @cindex unresolved references and @option{-nodefaultlibs}
11305 One of the standard libraries bypassed by @option{-nostdlib} and
11306 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11307 which GCC uses to overcome shortcomings of particular machines, or special
11308 needs for some languages.
11309 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11310 Collection (GCC) Internals},
11311 for more discussion of @file{libgcc.a}.)
11312 In most cases, you need @file{libgcc.a} even when you want to avoid
11313 other standard libraries. In other words, when you specify @option{-nostdlib}
11314 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11315 This ensures that you have no unresolved references to internal GCC
11316 library subroutines.
11317 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11318 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11319 GNU Compiler Collection (GCC) Internals}.)
11323 Produce a position independent executable on targets that support it.
11324 For predictable results, you must also specify the same set of options
11325 used for compilation (@option{-fpie}, @option{-fPIE},
11326 or model suboptions) when you specify this linker option.
11330 Don't produce a position independent executable.
11334 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11335 that support it. This instructs the linker to add all symbols, not
11336 only used ones, to the dynamic symbol table. This option is needed
11337 for some uses of @code{dlopen} or to allow obtaining backtraces
11338 from within a program.
11342 Remove all symbol table and relocation information from the executable.
11346 On systems that support dynamic linking, this prevents linking with the shared
11347 libraries. On other systems, this option has no effect.
11351 Produce a shared object which can then be linked with other objects to
11352 form an executable. Not all systems support this option. For predictable
11353 results, you must also specify the same set of options used for compilation
11354 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11355 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11356 needs to build supplementary stub code for constructors to work. On
11357 multi-libbed systems, @samp{gcc -shared} must select the correct support
11358 libraries to link against. Failing to supply the correct flags may lead
11359 to subtle defects. Supplying them in cases where they are not necessary
11362 @item -shared-libgcc
11363 @itemx -static-libgcc
11364 @opindex shared-libgcc
11365 @opindex static-libgcc
11366 On systems that provide @file{libgcc} as a shared library, these options
11367 force the use of either the shared or static version, respectively.
11368 If no shared version of @file{libgcc} was built when the compiler was
11369 configured, these options have no effect.
11371 There are several situations in which an application should use the
11372 shared @file{libgcc} instead of the static version. The most common
11373 of these is when the application wishes to throw and catch exceptions
11374 across different shared libraries. In that case, each of the libraries
11375 as well as the application itself should use the shared @file{libgcc}.
11377 Therefore, the G++ and GCJ drivers automatically add
11378 @option{-shared-libgcc} whenever you build a shared library or a main
11379 executable, because C++ and Java programs typically use exceptions, so
11380 this is the right thing to do.
11382 If, instead, you use the GCC driver to create shared libraries, you may
11383 find that they are not always linked with the shared @file{libgcc}.
11384 If GCC finds, at its configuration time, that you have a non-GNU linker
11385 or a GNU linker that does not support option @option{--eh-frame-hdr},
11386 it links the shared version of @file{libgcc} into shared libraries
11387 by default. Otherwise, it takes advantage of the linker and optimizes
11388 away the linking with the shared version of @file{libgcc}, linking with
11389 the static version of libgcc by default. This allows exceptions to
11390 propagate through such shared libraries, without incurring relocation
11391 costs at library load time.
11393 However, if a library or main executable is supposed to throw or catch
11394 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11395 for the languages used in the program, or using the option
11396 @option{-shared-libgcc}, such that it is linked with the shared
11399 @item -static-libasan
11400 @opindex static-libasan
11401 When the @option{-fsanitize=address} option is used to link a program,
11402 the GCC driver automatically links against @option{libasan}. If
11403 @file{libasan} is available as a shared library, and the @option{-static}
11404 option is not used, then this links against the shared version of
11405 @file{libasan}. The @option{-static-libasan} option directs the GCC
11406 driver to link @file{libasan} statically, without necessarily linking
11407 other libraries statically.
11409 @item -static-libtsan
11410 @opindex static-libtsan
11411 When the @option{-fsanitize=thread} option is used to link a program,
11412 the GCC driver automatically links against @option{libtsan}. If
11413 @file{libtsan} is available as a shared library, and the @option{-static}
11414 option is not used, then this links against the shared version of
11415 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11416 driver to link @file{libtsan} statically, without necessarily linking
11417 other libraries statically.
11419 @item -static-liblsan
11420 @opindex static-liblsan
11421 When the @option{-fsanitize=leak} option is used to link a program,
11422 the GCC driver automatically links against @option{liblsan}. If
11423 @file{liblsan} is available as a shared library, and the @option{-static}
11424 option is not used, then this links against the shared version of
11425 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11426 driver to link @file{liblsan} statically, without necessarily linking
11427 other libraries statically.
11429 @item -static-libubsan
11430 @opindex static-libubsan
11431 When the @option{-fsanitize=undefined} option is used to link a program,
11432 the GCC driver automatically links against @option{libubsan}. If
11433 @file{libubsan} is available as a shared library, and the @option{-static}
11434 option is not used, then this links against the shared version of
11435 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11436 driver to link @file{libubsan} statically, without necessarily linking
11437 other libraries statically.
11439 @item -static-libmpx
11440 @opindex static-libmpx
11441 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11442 used to link a program, the GCC driver automatically links against
11443 @file{libmpx}. If @file{libmpx} is available as a shared library,
11444 and the @option{-static} option is not used, then this links against
11445 the shared version of @file{libmpx}. The @option{-static-libmpx}
11446 option directs the GCC driver to link @file{libmpx} statically,
11447 without necessarily linking other libraries statically.
11449 @item -static-libmpxwrappers
11450 @opindex static-libmpxwrappers
11451 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11452 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11453 GCC driver automatically links against @file{libmpxwrappers}. If
11454 @file{libmpxwrappers} is available as a shared library, and the
11455 @option{-static} option is not used, then this links against the shared
11456 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11457 option directs the GCC driver to link @file{libmpxwrappers} statically,
11458 without necessarily linking other libraries statically.
11460 @item -static-libstdc++
11461 @opindex static-libstdc++
11462 When the @command{g++} program is used to link a C++ program, it
11463 normally automatically links against @option{libstdc++}. If
11464 @file{libstdc++} is available as a shared library, and the
11465 @option{-static} option is not used, then this links against the
11466 shared version of @file{libstdc++}. That is normally fine. However, it
11467 is sometimes useful to freeze the version of @file{libstdc++} used by
11468 the program without going all the way to a fully static link. The
11469 @option{-static-libstdc++} option directs the @command{g++} driver to
11470 link @file{libstdc++} statically, without necessarily linking other
11471 libraries statically.
11475 Bind references to global symbols when building a shared object. Warn
11476 about any unresolved references (unless overridden by the link editor
11477 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11480 @item -T @var{script}
11482 @cindex linker script
11483 Use @var{script} as the linker script. This option is supported by most
11484 systems using the GNU linker. On some targets, such as bare-board
11485 targets without an operating system, the @option{-T} option may be required
11486 when linking to avoid references to undefined symbols.
11488 @item -Xlinker @var{option}
11490 Pass @var{option} as an option to the linker. You can use this to
11491 supply system-specific linker options that GCC does not recognize.
11493 If you want to pass an option that takes a separate argument, you must use
11494 @option{-Xlinker} twice, once for the option and once for the argument.
11495 For example, to pass @option{-assert definitions}, you must write
11496 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11497 @option{-Xlinker "-assert definitions"}, because this passes the entire
11498 string as a single argument, which is not what the linker expects.
11500 When using the GNU linker, it is usually more convenient to pass
11501 arguments to linker options using the @option{@var{option}=@var{value}}
11502 syntax than as separate arguments. For example, you can specify
11503 @option{-Xlinker -Map=output.map} rather than
11504 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11505 this syntax for command-line options.
11507 @item -Wl,@var{option}
11509 Pass @var{option} as an option to the linker. If @var{option} contains
11510 commas, it is split into multiple options at the commas. You can use this
11511 syntax to pass an argument to the option.
11512 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11513 linker. When using the GNU linker, you can also get the same effect with
11514 @option{-Wl,-Map=output.map}.
11516 @item -u @var{symbol}
11518 Pretend the symbol @var{symbol} is undefined, to force linking of
11519 library modules to define it. You can use @option{-u} multiple times with
11520 different symbols to force loading of additional library modules.
11522 @item -z @var{keyword}
11524 @option{-z} is passed directly on to the linker along with the keyword
11525 @var{keyword}. See the section in the documentation of your linker for
11526 permitted values and their meanings.
11529 @node Directory Options
11530 @section Options for Directory Search
11531 @cindex directory options
11532 @cindex options, directory search
11533 @cindex search path
11535 These options specify directories to search for header files, for
11536 libraries and for parts of the compiler:
11541 Add the directory @var{dir} to the head of the list of directories to be
11542 searched for header files. This can be used to override a system header
11543 file, substituting your own version, since these directories are
11544 searched before the system header file directories. However, you should
11545 not use this option to add directories that contain vendor-supplied
11546 system header files (use @option{-isystem} for that). If you use more than
11547 one @option{-I} option, the directories are scanned in left-to-right
11548 order; the standard system directories come after.
11550 If a standard system include directory, or a directory specified with
11551 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11552 option is ignored. The directory is still searched but as a
11553 system directory at its normal position in the system include chain.
11554 This is to ensure that GCC's procedure to fix buggy system headers and
11555 the ordering for the @code{include_next} directive are not inadvertently changed.
11556 If you really need to change the search order for system directories,
11557 use the @option{-nostdinc} and/or @option{-isystem} options.
11559 @item -iplugindir=@var{dir}
11560 @opindex iplugindir=
11561 Set the directory to search for plugins that are passed
11562 by @option{-fplugin=@var{name}} instead of
11563 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11564 to be used by the user, but only passed by the driver.
11566 @item -iquote@var{dir}
11568 Add the directory @var{dir} to the head of the list of directories to
11569 be searched for header files only for the case of @code{#include
11570 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11571 otherwise just like @option{-I}.
11575 Add directory @var{dir} to the list of directories to be searched
11578 @item -B@var{prefix}
11580 This option specifies where to find the executables, libraries,
11581 include files, and data files of the compiler itself.
11583 The compiler driver program runs one or more of the subprograms
11584 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11585 @var{prefix} as a prefix for each program it tries to run, both with and
11586 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11588 For each subprogram to be run, the compiler driver first tries the
11589 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11590 is not specified, the driver tries two standard prefixes,
11591 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11592 those results in a file name that is found, the unmodified program
11593 name is searched for using the directories specified in your
11594 @env{PATH} environment variable.
11596 The compiler checks to see if the path provided by @option{-B}
11597 refers to a directory, and if necessary it adds a directory
11598 separator character at the end of the path.
11600 @option{-B} prefixes that effectively specify directory names also apply
11601 to libraries in the linker, because the compiler translates these
11602 options into @option{-L} options for the linker. They also apply to
11603 include files in the preprocessor, because the compiler translates these
11604 options into @option{-isystem} options for the preprocessor. In this case,
11605 the compiler appends @samp{include} to the prefix.
11607 The runtime support file @file{libgcc.a} can also be searched for using
11608 the @option{-B} prefix, if needed. If it is not found there, the two
11609 standard prefixes above are tried, and that is all. The file is left
11610 out of the link if it is not found by those means.
11612 Another way to specify a prefix much like the @option{-B} prefix is to use
11613 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11616 As a special kludge, if the path provided by @option{-B} is
11617 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11618 9, then it is replaced by @file{[dir/]include}. This is to help
11619 with boot-strapping the compiler.
11621 @item -specs=@var{file}
11623 Process @var{file} after the compiler reads in the standard @file{specs}
11624 file, in order to override the defaults which the @command{gcc} driver
11625 program uses when determining what switches to pass to @command{cc1},
11626 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11627 @option{-specs=@var{file}} can be specified on the command line, and they
11628 are processed in order, from left to right.
11630 @item --sysroot=@var{dir}
11632 Use @var{dir} as the logical root directory for headers and libraries.
11633 For example, if the compiler normally searches for headers in
11634 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11635 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11637 If you use both this option and the @option{-isysroot} option, then
11638 the @option{--sysroot} option applies to libraries, but the
11639 @option{-isysroot} option applies to header files.
11641 The GNU linker (beginning with version 2.16) has the necessary support
11642 for this option. If your linker does not support this option, the
11643 header file aspect of @option{--sysroot} still works, but the
11644 library aspect does not.
11646 @item --no-sysroot-suffix
11647 @opindex no-sysroot-suffix
11648 For some targets, a suffix is added to the root directory specified
11649 with @option{--sysroot}, depending on the other options used, so that
11650 headers may for example be found in
11651 @file{@var{dir}/@var{suffix}/usr/include} instead of
11652 @file{@var{dir}/usr/include}. This option disables the addition of
11657 This option has been deprecated. Please use @option{-iquote} instead for
11658 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11660 Any directories you specify with @option{-I} options before the @option{-I-}
11661 option are searched only for the case of @code{#include "@var{file}"};
11662 they are not searched for @code{#include <@var{file}>}.
11664 If additional directories are specified with @option{-I} options after
11665 the @option{-I-} option, these directories are searched for all @code{#include}
11666 directives. (Ordinarily @emph{all} @option{-I} directories are used
11669 In addition, the @option{-I-} option inhibits the use of the current
11670 directory (where the current input file came from) as the first search
11671 directory for @code{#include "@var{file}"}. There is no way to
11672 override this effect of @option{-I-}. With @option{-I.} you can specify
11673 searching the directory that is current when the compiler is
11674 invoked. That is not exactly the same as what the preprocessor does
11675 by default, but it is often satisfactory.
11677 @option{-I-} does not inhibit the use of the standard system directories
11678 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11685 @section Specifying Subprocesses and the Switches to Pass to Them
11688 @command{gcc} is a driver program. It performs its job by invoking a
11689 sequence of other programs to do the work of compiling, assembling and
11690 linking. GCC interprets its command-line parameters and uses these to
11691 deduce which programs it should invoke, and which command-line options
11692 it ought to place on their command lines. This behavior is controlled
11693 by @dfn{spec strings}. In most cases there is one spec string for each
11694 program that GCC can invoke, but a few programs have multiple spec
11695 strings to control their behavior. The spec strings built into GCC can
11696 be overridden by using the @option{-specs=} command-line switch to specify
11699 @dfn{Spec files} are plaintext files that are used to construct spec
11700 strings. They consist of a sequence of directives separated by blank
11701 lines. The type of directive is determined by the first non-whitespace
11702 character on the line, which can be one of the following:
11705 @item %@var{command}
11706 Issues a @var{command} to the spec file processor. The commands that can
11710 @item %include <@var{file}>
11711 @cindex @code{%include}
11712 Search for @var{file} and insert its text at the current point in the
11715 @item %include_noerr <@var{file}>
11716 @cindex @code{%include_noerr}
11717 Just like @samp{%include}, but do not generate an error message if the include
11718 file cannot be found.
11720 @item %rename @var{old_name} @var{new_name}
11721 @cindex @code{%rename}
11722 Rename the spec string @var{old_name} to @var{new_name}.
11726 @item *[@var{spec_name}]:
11727 This tells the compiler to create, override or delete the named spec
11728 string. All lines after this directive up to the next directive or
11729 blank line are considered to be the text for the spec string. If this
11730 results in an empty string then the spec is deleted. (Or, if the
11731 spec did not exist, then nothing happens.) Otherwise, if the spec
11732 does not currently exist a new spec is created. If the spec does
11733 exist then its contents are overridden by the text of this
11734 directive, unless the first character of that text is the @samp{+}
11735 character, in which case the text is appended to the spec.
11737 @item [@var{suffix}]:
11738 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11739 and up to the next directive or blank line are considered to make up the
11740 spec string for the indicated suffix. When the compiler encounters an
11741 input file with the named suffix, it processes the spec string in
11742 order to work out how to compile that file. For example:
11746 z-compile -input %i
11749 This says that any input file whose name ends in @samp{.ZZ} should be
11750 passed to the program @samp{z-compile}, which should be invoked with the
11751 command-line switch @option{-input} and with the result of performing the
11752 @samp{%i} substitution. (See below.)
11754 As an alternative to providing a spec string, the text following a
11755 suffix directive can be one of the following:
11758 @item @@@var{language}
11759 This says that the suffix is an alias for a known @var{language}. This is
11760 similar to using the @option{-x} command-line switch to GCC to specify a
11761 language explicitly. For example:
11768 Says that .ZZ files are, in fact, C++ source files.
11771 This causes an error messages saying:
11774 @var{name} compiler not installed on this system.
11778 GCC already has an extensive list of suffixes built into it.
11779 This directive adds an entry to the end of the list of suffixes, but
11780 since the list is searched from the end backwards, it is effectively
11781 possible to override earlier entries using this technique.
11785 GCC has the following spec strings built into it. Spec files can
11786 override these strings or create their own. Note that individual
11787 targets can also add their own spec strings to this list.
11790 asm Options to pass to the assembler
11791 asm_final Options to pass to the assembler post-processor
11792 cpp Options to pass to the C preprocessor
11793 cc1 Options to pass to the C compiler
11794 cc1plus Options to pass to the C++ compiler
11795 endfile Object files to include at the end of the link
11796 link Options to pass to the linker
11797 lib Libraries to include on the command line to the linker
11798 libgcc Decides which GCC support library to pass to the linker
11799 linker Sets the name of the linker
11800 predefines Defines to be passed to the C preprocessor
11801 signed_char Defines to pass to CPP to say whether @code{char} is signed
11803 startfile Object files to include at the start of the link
11806 Here is a small example of a spec file:
11809 %rename lib old_lib
11812 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11815 This example renames the spec called @samp{lib} to @samp{old_lib} and
11816 then overrides the previous definition of @samp{lib} with a new one.
11817 The new definition adds in some extra command-line options before
11818 including the text of the old definition.
11820 @dfn{Spec strings} are a list of command-line options to be passed to their
11821 corresponding program. In addition, the spec strings can contain
11822 @samp{%}-prefixed sequences to substitute variable text or to
11823 conditionally insert text into the command line. Using these constructs
11824 it is possible to generate quite complex command lines.
11826 Here is a table of all defined @samp{%}-sequences for spec
11827 strings. Note that spaces are not generated automatically around the
11828 results of expanding these sequences. Therefore you can concatenate them
11829 together or combine them with constant text in a single argument.
11833 Substitute one @samp{%} into the program name or argument.
11836 Substitute the name of the input file being processed.
11839 Substitute the basename of the input file being processed.
11840 This is the substring up to (and not including) the last period
11841 and not including the directory.
11844 This is the same as @samp{%b}, but include the file suffix (text after
11848 Marks the argument containing or following the @samp{%d} as a
11849 temporary file name, so that that file is deleted if GCC exits
11850 successfully. Unlike @samp{%g}, this contributes no text to the
11853 @item %g@var{suffix}
11854 Substitute a file name that has suffix @var{suffix} and is chosen
11855 once per compilation, and mark the argument in the same way as
11856 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11857 name is now chosen in a way that is hard to predict even when previously
11858 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11859 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11860 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11861 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11862 was simply substituted with a file name chosen once per compilation,
11863 without regard to any appended suffix (which was therefore treated
11864 just like ordinary text), making such attacks more likely to succeed.
11866 @item %u@var{suffix}
11867 Like @samp{%g}, but generates a new temporary file name
11868 each time it appears instead of once per compilation.
11870 @item %U@var{suffix}
11871 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11872 new one if there is no such last file name. In the absence of any
11873 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11874 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11875 involves the generation of two distinct file names, one
11876 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11877 simply substituted with a file name chosen for the previous @samp{%u},
11878 without regard to any appended suffix.
11880 @item %j@var{suffix}
11881 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
11882 writable, and if @option{-save-temps} is not used;
11883 otherwise, substitute the name
11884 of a temporary file, just like @samp{%u}. This temporary file is not
11885 meant for communication between processes, but rather as a junk
11886 disposal mechanism.
11888 @item %|@var{suffix}
11889 @itemx %m@var{suffix}
11890 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
11891 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
11892 all. These are the two most common ways to instruct a program that it
11893 should read from standard input or write to standard output. If you
11894 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
11895 construct: see for example @file{f/lang-specs.h}.
11897 @item %.@var{SUFFIX}
11898 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
11899 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
11900 terminated by the next space or %.
11903 Marks the argument containing or following the @samp{%w} as the
11904 designated output file of this compilation. This puts the argument
11905 into the sequence of arguments that @samp{%o} substitutes.
11908 Substitutes the names of all the output files, with spaces
11909 automatically placed around them. You should write spaces
11910 around the @samp{%o} as well or the results are undefined.
11911 @samp{%o} is for use in the specs for running the linker.
11912 Input files whose names have no recognized suffix are not compiled
11913 at all, but they are included among the output files, so they are
11917 Substitutes the suffix for object files. Note that this is
11918 handled specially when it immediately follows @samp{%g, %u, or %U},
11919 because of the need for those to form complete file names. The
11920 handling is such that @samp{%O} is treated exactly as if it had already
11921 been substituted, except that @samp{%g, %u, and %U} do not currently
11922 support additional @var{suffix} characters following @samp{%O} as they do
11923 following, for example, @samp{.o}.
11926 Substitutes the standard macro predefinitions for the
11927 current target machine. Use this when running @command{cpp}.
11930 Like @samp{%p}, but puts @samp{__} before and after the name of each
11931 predefined macro, except for macros that start with @samp{__} or with
11932 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
11936 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
11937 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
11938 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
11939 and @option{-imultilib} as necessary.
11942 Current argument is the name of a library or startup file of some sort.
11943 Search for that file in a standard list of directories and substitute
11944 the full name found. The current working directory is included in the
11945 list of directories scanned.
11948 Current argument is the name of a linker script. Search for that file
11949 in the current list of directories to scan for libraries. If the file
11950 is located insert a @option{--script} option into the command line
11951 followed by the full path name found. If the file is not found then
11952 generate an error message. Note: the current working directory is not
11956 Print @var{str} as an error message. @var{str} is terminated by a newline.
11957 Use this when inconsistent options are detected.
11959 @item %(@var{name})
11960 Substitute the contents of spec string @var{name} at this point.
11962 @item %x@{@var{option}@}
11963 Accumulate an option for @samp{%X}.
11966 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
11970 Output the accumulated assembler options specified by @option{-Wa}.
11973 Output the accumulated preprocessor options specified by @option{-Wp}.
11976 Process the @code{asm} spec. This is used to compute the
11977 switches to be passed to the assembler.
11980 Process the @code{asm_final} spec. This is a spec string for
11981 passing switches to an assembler post-processor, if such a program is
11985 Process the @code{link} spec. This is the spec for computing the
11986 command line passed to the linker. Typically it makes use of the
11987 @samp{%L %G %S %D and %E} sequences.
11990 Dump out a @option{-L} option for each directory that GCC believes might
11991 contain startup files. If the target supports multilibs then the
11992 current multilib directory is prepended to each of these paths.
11995 Process the @code{lib} spec. This is a spec string for deciding which
11996 libraries are included on the command line to the linker.
11999 Process the @code{libgcc} spec. This is a spec string for deciding
12000 which GCC support library is included on the command line to the linker.
12003 Process the @code{startfile} spec. This is a spec for deciding which
12004 object files are the first ones passed to the linker. Typically
12005 this might be a file named @file{crt0.o}.
12008 Process the @code{endfile} spec. This is a spec string that specifies
12009 the last object files that are passed to the linker.
12012 Process the @code{cpp} spec. This is used to construct the arguments
12013 to be passed to the C preprocessor.
12016 Process the @code{cc1} spec. This is used to construct the options to be
12017 passed to the actual C compiler (@command{cc1}).
12020 Process the @code{cc1plus} spec. This is used to construct the options to be
12021 passed to the actual C++ compiler (@command{cc1plus}).
12024 Substitute the variable part of a matched option. See below.
12025 Note that each comma in the substituted string is replaced by
12029 Remove all occurrences of @code{-S} from the command line. Note---this
12030 command is position dependent. @samp{%} commands in the spec string
12031 before this one see @code{-S}, @samp{%} commands in the spec string
12032 after this one do not.
12034 @item %:@var{function}(@var{args})
12035 Call the named function @var{function}, passing it @var{args}.
12036 @var{args} is first processed as a nested spec string, then split
12037 into an argument vector in the usual fashion. The function returns
12038 a string which is processed as if it had appeared literally as part
12039 of the current spec.
12041 The following built-in spec functions are provided:
12044 @item @code{getenv}
12045 The @code{getenv} spec function takes two arguments: an environment
12046 variable name and a string. If the environment variable is not
12047 defined, a fatal error is issued. Otherwise, the return value is the
12048 value of the environment variable concatenated with the string. For
12049 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
12052 %:getenv(TOPDIR /include)
12055 expands to @file{/path/to/top/include}.
12057 @item @code{if-exists}
12058 The @code{if-exists} spec function takes one argument, an absolute
12059 pathname to a file. If the file exists, @code{if-exists} returns the
12060 pathname. Here is a small example of its usage:
12064 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
12067 @item @code{if-exists-else}
12068 The @code{if-exists-else} spec function is similar to the @code{if-exists}
12069 spec function, except that it takes two arguments. The first argument is
12070 an absolute pathname to a file. If the file exists, @code{if-exists-else}
12071 returns the pathname. If it does not exist, it returns the second argument.
12072 This way, @code{if-exists-else} can be used to select one file or another,
12073 based on the existence of the first. Here is a small example of its usage:
12077 crt0%O%s %:if-exists(crti%O%s) \
12078 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
12081 @item @code{replace-outfile}
12082 The @code{replace-outfile} spec function takes two arguments. It looks for the
12083 first argument in the outfiles array and replaces it with the second argument. Here
12084 is a small example of its usage:
12087 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
12090 @item @code{remove-outfile}
12091 The @code{remove-outfile} spec function takes one argument. It looks for the
12092 first argument in the outfiles array and removes it. Here is a small example
12096 %:remove-outfile(-lm)
12099 @item @code{pass-through-libs}
12100 The @code{pass-through-libs} spec function takes any number of arguments. It
12101 finds any @option{-l} options and any non-options ending in @file{.a} (which it
12102 assumes are the names of linker input library archive files) and returns a
12103 result containing all the found arguments each prepended by
12104 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
12105 intended to be passed to the LTO linker plugin.
12108 %:pass-through-libs(%G %L %G)
12111 @item @code{print-asm-header}
12112 The @code{print-asm-header} function takes no arguments and simply
12113 prints a banner like:
12119 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12122 It is used to separate compiler options from assembler options
12123 in the @option{--target-help} output.
12126 @item %@{@code{S}@}
12127 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12128 If that switch is not specified, this substitutes nothing. Note that
12129 the leading dash is omitted when specifying this option, and it is
12130 automatically inserted if the substitution is performed. Thus the spec
12131 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12132 and outputs the command-line option @option{-foo}.
12134 @item %W@{@code{S}@}
12135 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12136 deleted on failure.
12138 @item %@{@code{S}*@}
12139 Substitutes all the switches specified to GCC whose names start
12140 with @code{-S}, but which also take an argument. This is used for
12141 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12142 GCC considers @option{-o foo} as being
12143 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12144 text, including the space. Thus two arguments are generated.
12146 @item %@{@code{S}*&@code{T}*@}
12147 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12148 (the order of @code{S} and @code{T} in the spec is not significant).
12149 There can be any number of ampersand-separated variables; for each the
12150 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12152 @item %@{@code{S}:@code{X}@}
12153 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12155 @item %@{!@code{S}:@code{X}@}
12156 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12158 @item %@{@code{S}*:@code{X}@}
12159 Substitutes @code{X} if one or more switches whose names start with
12160 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12161 once, no matter how many such switches appeared. However, if @code{%*}
12162 appears somewhere in @code{X}, then @code{X} is substituted once
12163 for each matching switch, with the @code{%*} replaced by the part of
12164 that switch matching the @code{*}.
12166 If @code{%*} appears as the last part of a spec sequence then a space
12167 is added after the end of the last substitution. If there is more
12168 text in the sequence, however, then a space is not generated. This
12169 allows the @code{%*} substitution to be used as part of a larger
12170 string. For example, a spec string like this:
12173 %@{mcu=*:--script=%*/memory.ld@}
12177 when matching an option like @option{-mcu=newchip} produces:
12180 --script=newchip/memory.ld
12183 @item %@{.@code{S}:@code{X}@}
12184 Substitutes @code{X}, if processing a file with suffix @code{S}.
12186 @item %@{!.@code{S}:@code{X}@}
12187 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12189 @item %@{,@code{S}:@code{X}@}
12190 Substitutes @code{X}, if processing a file for language @code{S}.
12192 @item %@{!,@code{S}:@code{X}@}
12193 Substitutes @code{X}, if not processing a file for language @code{S}.
12195 @item %@{@code{S}|@code{P}:@code{X}@}
12196 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12197 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12198 @code{*} sequences as well, although they have a stronger binding than
12199 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12200 alternatives must be starred, and only the first matching alternative
12203 For example, a spec string like this:
12206 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12210 outputs the following command-line options from the following input
12211 command-line options:
12216 -d fred.c -foo -baz -boggle
12217 -d jim.d -bar -baz -boggle
12220 @item %@{S:X; T:Y; :D@}
12222 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12223 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12224 be as many clauses as you need. This may be combined with @code{.},
12225 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12230 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12231 construct may contain other nested @samp{%} constructs or spaces, or
12232 even newlines. They are processed as usual, as described above.
12233 Trailing white space in @code{X} is ignored. White space may also
12234 appear anywhere on the left side of the colon in these constructs,
12235 except between @code{.} or @code{*} and the corresponding word.
12237 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12238 handled specifically in these constructs. If another value of
12239 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12240 @option{-W} switch is found later in the command line, the earlier
12241 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12242 just one letter, which passes all matching options.
12244 The character @samp{|} at the beginning of the predicate text is used to
12245 indicate that a command should be piped to the following command, but
12246 only if @option{-pipe} is specified.
12248 It is built into GCC which switches take arguments and which do not.
12249 (You might think it would be useful to generalize this to allow each
12250 compiler's spec to say which switches take arguments. But this cannot
12251 be done in a consistent fashion. GCC cannot even decide which input
12252 files have been specified without knowing which switches take arguments,
12253 and it must know which input files to compile in order to tell which
12256 GCC also knows implicitly that arguments starting in @option{-l} are to be
12257 treated as compiler output files, and passed to the linker in their
12258 proper position among the other output files.
12260 @c man begin OPTIONS
12262 @node Target Options
12263 @section Specifying Target Machine and Compiler Version
12264 @cindex target options
12265 @cindex cross compiling
12266 @cindex specifying machine version
12267 @cindex specifying compiler version and target machine
12268 @cindex compiler version, specifying
12269 @cindex target machine, specifying
12271 The usual way to run GCC is to run the executable called @command{gcc}, or
12272 @command{@var{machine}-gcc} when cross-compiling, or
12273 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12274 one that was installed last.
12276 @node Submodel Options
12277 @section Hardware Models and Configurations
12278 @cindex submodel options
12279 @cindex specifying hardware config
12280 @cindex hardware models and configurations, specifying
12281 @cindex machine dependent options
12283 Each target machine types can have its own
12284 special options, starting with @samp{-m}, to choose among various
12285 hardware models or configurations---for example, 68010 vs 68020,
12286 floating coprocessor or none. A single installed version of the
12287 compiler can compile for any model or configuration, according to the
12290 Some configurations of the compiler also support additional special
12291 options, usually for compatibility with other compilers on the same
12294 @c This list is ordered alphanumerically by subsection name.
12295 @c It should be the same order and spelling as these options are listed
12296 @c in Machine Dependent Options
12299 * AArch64 Options::
12300 * Adapteva Epiphany Options::
12304 * Blackfin Options::
12309 * DEC Alpha Options::
12313 * GNU/Linux Options::
12323 * MicroBlaze Options::
12326 * MN10300 Options::
12330 * Nios II Options::
12331 * Nvidia PTX Options::
12333 * picoChip Options::
12334 * PowerPC Options::
12336 * RS/6000 and PowerPC Options::
12338 * S/390 and zSeries Options::
12341 * Solaris 2 Options::
12344 * System V Options::
12345 * TILE-Gx Options::
12346 * TILEPro Options::
12351 * VxWorks Options::
12353 * x86 Windows Options::
12354 * Xstormy16 Options::
12356 * zSeries Options::
12359 @node AArch64 Options
12360 @subsection AArch64 Options
12361 @cindex AArch64 Options
12363 These options are defined for AArch64 implementations:
12367 @item -mabi=@var{name}
12369 Generate code for the specified data model. Permissible values
12370 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12371 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12372 but long int and pointer are 64-bit.
12374 The default depends on the specific target configuration. Note that
12375 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12376 entire program with the same ABI, and link with a compatible set of libraries.
12379 @opindex mbig-endian
12380 Generate big-endian code. This is the default when GCC is configured for an
12381 @samp{aarch64_be-*-*} target.
12383 @item -mgeneral-regs-only
12384 @opindex mgeneral-regs-only
12385 Generate code which uses only the general-purpose registers. This is equivalent
12386 to feature modifier @option{nofp} of @option{-march} or @option{-mcpu}, except
12387 that @option{-mgeneral-regs-only} takes precedence over any conflicting feature
12388 modifier regardless of sequence.
12390 @item -mlittle-endian
12391 @opindex mlittle-endian
12392 Generate little-endian code. This is the default when GCC is configured for an
12393 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12395 @item -mcmodel=tiny
12396 @opindex mcmodel=tiny
12397 Generate code for the tiny code model. The program and its statically defined
12398 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12399 be statically or dynamically linked. This model is not fully implemented and
12400 mostly treated as @samp{small}.
12402 @item -mcmodel=small
12403 @opindex mcmodel=small
12404 Generate code for the small code model. The program and its statically defined
12405 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12406 be statically or dynamically linked. This is the default code model.
12408 @item -mcmodel=large
12409 @opindex mcmodel=large
12410 Generate code for the large code model. This makes no assumptions about
12411 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12412 statically linked only.
12414 @item -mstrict-align
12415 @opindex mstrict-align
12416 Do not assume that unaligned memory references are handled by the system.
12418 @item -momit-leaf-frame-pointer
12419 @itemx -mno-omit-leaf-frame-pointer
12420 @opindex momit-leaf-frame-pointer
12421 @opindex mno-omit-leaf-frame-pointer
12422 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12425 @item -mtls-dialect=desc
12426 @opindex mtls-dialect=desc
12427 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12428 of TLS variables. This is the default.
12430 @item -mtls-dialect=traditional
12431 @opindex mtls-dialect=traditional
12432 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12435 @item -mfix-cortex-a53-835769
12436 @itemx -mno-fix-cortex-a53-835769
12437 @opindex mfix-cortex-a53-835769
12438 @opindex mno-fix-cortex-a53-835769
12439 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12440 This involves inserting a NOP instruction between memory instructions and
12441 64-bit integer multiply-accumulate instructions.
12443 @item -mfix-cortex-a53-843419
12444 @itemx -mno-fix-cortex-a53-843419
12445 @opindex mfix-cortex-a53-843419
12446 @opindex mno-fix-cortex-a53-843419
12447 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12448 This erratum workaround is made at link time and this will only pass the
12449 corresponding flag to the linker.
12451 @item -march=@var{name}
12453 Specify the name of the target architecture, optionally suffixed by one or
12454 more feature modifiers. This option has the form
12455 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12457 The permissible values for @var{arch} are @samp{armv8-a} or
12460 For the permissible values for @var{feature}, see the sub-section on
12461 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12462 Feature Modifiers}. Where conflicting feature modifiers are
12463 specified, the right-most feature is used.
12465 Additionally on native AArch64 GNU/Linux systems the value
12466 @samp{native} is available. This option causes the compiler to pick the
12467 architecture of the host system. If the compiler is unable to recognize the
12468 architecture of the host system this option has no effect.
12470 GCC uses @var{name} to determine what kind of instructions it can emit
12471 when generating assembly code. If @option{-march} is specified
12472 without either of @option{-mtune} or @option{-mcpu} also being
12473 specified, the code is tuned to perform well across a range of target
12474 processors implementing the target architecture.
12476 @item -mtune=@var{name}
12478 Specify the name of the target processor for which GCC should tune the
12479 performance of the code. Permissible values for this option are:
12480 @samp{generic}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
12481 @samp{exynos-m1}, @samp{thunderx}, @samp{xgene1}.
12483 Additionally, this option can specify that GCC should tune the performance
12484 of the code for a big.LITTLE system. Permissible values for this
12485 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12487 Additionally on native AArch64 GNU/Linux systems the value
12488 @samp{native} is available. This option causes the compiler to pick
12489 the architecture of and tune the performance of the code for the
12490 processor of the host system. If the compiler is unable to recognize
12491 the processor of the host system this option has no effect.
12493 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12494 are specified, the code is tuned to perform well across a range
12495 of target processors.
12497 This option cannot be suffixed by feature modifiers.
12499 @item -mcpu=@var{name}
12501 Specify the name of the target processor, optionally suffixed by one
12502 or more feature modifiers. This option has the form
12503 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
12504 the permissible values for @var{cpu} are the same as those available
12505 for @option{-mtune}. The permissible values for @var{feature} are
12506 documented in the sub-section on
12507 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12508 Feature Modifiers}. Where conflicting feature modifiers are
12509 specified, the right-most feature is used.
12511 Additionally on native AArch64 GNU/Linux systems the value
12512 @samp{native} is available. This option causes the compiler to tune
12513 the performance of the code for the processor of the host system. If
12514 the compiler is unable to recognize the processor of the host system
12515 this option has no effect.
12517 GCC uses @var{name} to determine what kind of instructions it can emit when
12518 generating assembly code (as if by @option{-march}) and to determine
12519 the target processor for which to tune for performance (as if
12520 by @option{-mtune}). Where this option is used in conjunction
12521 with @option{-march} or @option{-mtune}, those options take precedence
12522 over the appropriate part of this option.
12524 @item -moverride=@var{string}
12526 Override tuning decisions made by the back-end in response to a
12527 @option{-mtune=} switch. The syntax, semantics, and accepted values
12528 for @var{string} in this option are not guaranteed to be consistent
12531 This option is only intended to be useful when developing GCC.
12534 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12535 @anchor{aarch64-feature-modifiers}
12536 @cindex @option{-march} feature modifiers
12537 @cindex @option{-mcpu} feature modifiers
12538 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
12539 the following and their inverses @option{no@var{feature}}:
12543 Enable CRC extension.
12545 Enable Crypto extension. This also enables Advanced SIMD and floating-point
12548 Enable floating-point instructions. This is on by default for all possible
12549 values for options @option{-march} and @option{-mcpu}.
12551 Enable Advanced SIMD instructions. This also enables floating-point
12552 instructions. This is on by default for all possible values for options
12553 @option{-march} and @option{-mcpu}.
12555 Enable Large System Extension instructions.
12557 Enable Privileged Access Never support.
12559 Enable Limited Ordering Regions support.
12561 Enable ARMv8.1 Advanced SIMD instructions. This implies Advanced SIMD
12566 That is, @option{crypto} implies @option{simd} implies @option{fp}.
12567 Conversely, @option{nofp} (or equivalently, @option{-mgeneral-regs-only})
12568 implies @option{nosimd} implies @option{nocrypto}.
12570 @node Adapteva Epiphany Options
12571 @subsection Adapteva Epiphany Options
12573 These @samp{-m} options are defined for Adapteva Epiphany:
12576 @item -mhalf-reg-file
12577 @opindex mhalf-reg-file
12578 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12579 That allows code to run on hardware variants that lack these registers.
12581 @item -mprefer-short-insn-regs
12582 @opindex mprefer-short-insn-regs
12583 Preferrentially allocate registers that allow short instruction generation.
12584 This can result in increased instruction count, so this may either reduce or
12585 increase overall code size.
12587 @item -mbranch-cost=@var{num}
12588 @opindex mbranch-cost
12589 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12590 This cost is only a heuristic and is not guaranteed to produce
12591 consistent results across releases.
12595 Enable the generation of conditional moves.
12597 @item -mnops=@var{num}
12599 Emit @var{num} NOPs before every other generated instruction.
12601 @item -mno-soft-cmpsf
12602 @opindex mno-soft-cmpsf
12603 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12604 and test the flags. This is faster than a software comparison, but can
12605 get incorrect results in the presence of NaNs, or when two different small
12606 numbers are compared such that their difference is calculated as zero.
12607 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12608 software comparisons.
12610 @item -mstack-offset=@var{num}
12611 @opindex mstack-offset
12612 Set the offset between the top of the stack and the stack pointer.
12613 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12614 can be used by leaf functions without stack allocation.
12615 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12616 Note also that this option changes the ABI; compiling a program with a
12617 different stack offset than the libraries have been compiled with
12618 generally does not work.
12619 This option can be useful if you want to evaluate if a different stack
12620 offset would give you better code, but to actually use a different stack
12621 offset to build working programs, it is recommended to configure the
12622 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12624 @item -mno-round-nearest
12625 @opindex mno-round-nearest
12626 Make the scheduler assume that the rounding mode has been set to
12627 truncating. The default is @option{-mround-nearest}.
12630 @opindex mlong-calls
12631 If not otherwise specified by an attribute, assume all calls might be beyond
12632 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12633 function address into a register before performing a (otherwise direct) call.
12634 This is the default.
12636 @item -mshort-calls
12637 @opindex short-calls
12638 If not otherwise specified by an attribute, assume all direct calls are
12639 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12640 for direct calls. The default is @option{-mlong-calls}.
12644 Assume addresses can be loaded as 16-bit unsigned values. This does not
12645 apply to function addresses for which @option{-mlong-calls} semantics
12648 @item -mfp-mode=@var{mode}
12650 Set the prevailing mode of the floating-point unit.
12651 This determines the floating-point mode that is provided and expected
12652 at function call and return time. Making this mode match the mode you
12653 predominantly need at function start can make your programs smaller and
12654 faster by avoiding unnecessary mode switches.
12656 @var{mode} can be set to one the following values:
12660 Any mode at function entry is valid, and retained or restored when
12661 the function returns, and when it calls other functions.
12662 This mode is useful for compiling libraries or other compilation units
12663 you might want to incorporate into different programs with different
12664 prevailing FPU modes, and the convenience of being able to use a single
12665 object file outweighs the size and speed overhead for any extra
12666 mode switching that might be needed, compared with what would be needed
12667 with a more specific choice of prevailing FPU mode.
12670 This is the mode used for floating-point calculations with
12671 truncating (i.e.@: round towards zero) rounding mode. That includes
12672 conversion from floating point to integer.
12674 @item round-nearest
12675 This is the mode used for floating-point calculations with
12676 round-to-nearest-or-even rounding mode.
12679 This is the mode used to perform integer calculations in the FPU, e.g.@:
12680 integer multiply, or integer multiply-and-accumulate.
12683 The default is @option{-mfp-mode=caller}
12685 @item -mnosplit-lohi
12686 @itemx -mno-postinc
12687 @itemx -mno-postmodify
12688 @opindex mnosplit-lohi
12689 @opindex mno-postinc
12690 @opindex mno-postmodify
12691 Code generation tweaks that disable, respectively, splitting of 32-bit
12692 loads, generation of post-increment addresses, and generation of
12693 post-modify addresses. The defaults are @option{msplit-lohi},
12694 @option{-mpost-inc}, and @option{-mpost-modify}.
12696 @item -mnovect-double
12697 @opindex mno-vect-double
12698 Change the preferred SIMD mode to SImode. The default is
12699 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12701 @item -max-vect-align=@var{num}
12702 @opindex max-vect-align
12703 The maximum alignment for SIMD vector mode types.
12704 @var{num} may be 4 or 8. The default is 8.
12705 Note that this is an ABI change, even though many library function
12706 interfaces are unaffected if they don't use SIMD vector modes
12707 in places that affect size and/or alignment of relevant types.
12709 @item -msplit-vecmove-early
12710 @opindex msplit-vecmove-early
12711 Split vector moves into single word moves before reload. In theory this
12712 can give better register allocation, but so far the reverse seems to be
12713 generally the case.
12715 @item -m1reg-@var{reg}
12717 Specify a register to hold the constant @minus{}1, which makes loading small negative
12718 constants and certain bitmasks faster.
12719 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12720 which specify use of that register as a fixed register,
12721 and @samp{none}, which means that no register is used for this
12722 purpose. The default is @option{-m1reg-none}.
12727 @subsection ARC Options
12728 @cindex ARC options
12730 The following options control the architecture variant for which code
12733 @c architecture variants
12736 @item -mbarrel-shifter
12737 @opindex mbarrel-shifter
12738 Generate instructions supported by barrel shifter. This is the default
12739 unless @option{-mcpu=ARC601} is in effect.
12741 @item -mcpu=@var{cpu}
12743 Set architecture type, register usage, and instruction scheduling
12744 parameters for @var{cpu}. There are also shortcut alias options
12745 available for backward compatibility and convenience. Supported
12746 values for @var{cpu} are
12752 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12756 Compile for ARC601. Alias: @option{-mARC601}.
12761 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12762 This is the default when configured with @option{--with-cpu=arc700}@.
12767 @itemx -mdpfp-compact
12768 @opindex mdpfp-compact
12769 FPX: Generate Double Precision FPX instructions, tuned for the compact
12773 @opindex mdpfp-fast
12774 FPX: Generate Double Precision FPX instructions, tuned for the fast
12777 @item -mno-dpfp-lrsr
12778 @opindex mno-dpfp-lrsr
12779 Disable LR and SR instructions from using FPX extension aux registers.
12783 Generate Extended arithmetic instructions. Currently only
12784 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12785 supported. This is always enabled for @option{-mcpu=ARC700}.
12789 Do not generate mpy instructions for ARC700.
12793 Generate 32x16 bit multiply and mac instructions.
12797 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12801 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12806 @itemx -mspfp-compact
12807 @opindex mspfp-compact
12808 FPX: Generate Single Precision FPX instructions, tuned for the compact
12812 @opindex mspfp-fast
12813 FPX: Generate Single Precision FPX instructions, tuned for the fast
12818 Enable generation of ARC SIMD instructions via target-specific
12819 builtins. Only valid for @option{-mcpu=ARC700}.
12822 @opindex msoft-float
12823 This option ignored; it is provided for compatibility purposes only.
12824 Software floating point code is emitted by default, and this default
12825 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12826 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12827 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12831 Generate swap instructions.
12835 The following options are passed through to the assembler, and also
12836 define preprocessor macro symbols.
12838 @c Flags used by the assembler, but for which we define preprocessor
12839 @c macro symbols as well.
12842 @opindex mdsp-packa
12843 Passed down to the assembler to enable the DSP Pack A extensions.
12844 Also sets the preprocessor symbol @code{__Xdsp_packa}.
12848 Passed down to the assembler to enable the dual viterbi butterfly
12849 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
12851 @c ARC700 4.10 extension instruction
12854 Passed down to the assembler to enable the Locked Load/Store
12855 Conditional extension. Also sets the preprocessor symbol
12860 Passed down to the assembler. Also sets the preprocessor symbol
12861 @code{__Xxmac_d16}.
12865 Passed down to the assembler. Also sets the preprocessor symbol
12868 @c ARC700 4.10 extension instruction
12871 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
12872 extension instruction. Also sets the preprocessor symbol
12875 @c ARC700 4.10 extension instruction
12878 Passed down to the assembler to enable the swap byte ordering
12879 extension instruction. Also sets the preprocessor symbol
12883 @opindex mtelephony
12884 Passed down to the assembler to enable dual and single operand
12885 instructions for telephony. Also sets the preprocessor symbol
12886 @code{__Xtelephony}.
12890 Passed down to the assembler to enable the XY Memory extension. Also
12891 sets the preprocessor symbol @code{__Xxy}.
12895 The following options control how the assembly code is annotated:
12897 @c Assembly annotation options
12901 Annotate assembler instructions with estimated addresses.
12903 @item -mannotate-align
12904 @opindex mannotate-align
12905 Explain what alignment considerations lead to the decision to make an
12906 instruction short or long.
12910 The following options are passed through to the linker:
12912 @c options passed through to the linker
12916 Passed through to the linker, to specify use of the @code{arclinux} emulation.
12917 This option is enabled by default in tool chains built for
12918 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
12919 when profiling is not requested.
12921 @item -marclinux_prof
12922 @opindex marclinux_prof
12923 Passed through to the linker, to specify use of the
12924 @code{arclinux_prof} emulation. This option is enabled by default in
12925 tool chains built for @w{@code{arc-linux-uclibc}} and
12926 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
12930 The following options control the semantics of generated code:
12932 @c semantically relevant code generation options
12934 @item -mepilogue-cfi
12935 @opindex mepilogue-cfi
12936 Enable generation of call frame information for epilogues.
12938 @item -mno-epilogue-cfi
12939 @opindex mno-epilogue-cfi
12940 Disable generation of call frame information for epilogues.
12943 @opindex mlong-calls
12944 Generate call insns as register indirect calls, thus providing access
12945 to the full 32-bit address range.
12947 @item -mmedium-calls
12948 @opindex mmedium-calls
12949 Don't use less than 25 bit addressing range for calls, which is the
12950 offset available for an unconditional branch-and-link
12951 instruction. Conditional execution of function calls is suppressed, to
12952 allow use of the 25-bit range, rather than the 21-bit range with
12953 conditional branch-and-link. This is the default for tool chains built
12954 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
12958 Do not generate sdata references. This is the default for tool chains
12959 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
12963 @opindex mucb-mcount
12964 Instrument with mcount calls as used in UCB code. I.e. do the
12965 counting in the callee, not the caller. By default ARC instrumentation
12966 counts in the caller.
12968 @item -mvolatile-cache
12969 @opindex mvolatile-cache
12970 Use ordinarily cached memory accesses for volatile references. This is the
12973 @item -mno-volatile-cache
12974 @opindex mno-volatile-cache
12975 Enable cache bypass for volatile references.
12979 The following options fine tune code generation:
12980 @c code generation tuning options
12983 @opindex malign-call
12984 Do alignment optimizations for call instructions.
12986 @item -mauto-modify-reg
12987 @opindex mauto-modify-reg
12988 Enable the use of pre/post modify with register displacement.
12990 @item -mbbit-peephole
12991 @opindex mbbit-peephole
12992 Enable bbit peephole2.
12996 This option disables a target-specific pass in @file{arc_reorg} to
12997 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
12998 generation driven by the combiner pass.
13000 @item -mcase-vector-pcrel
13001 @opindex mcase-vector-pcrel
13002 Use pc-relative switch case tables - this enables case table shortening.
13003 This is the default for @option{-Os}.
13005 @item -mcompact-casesi
13006 @opindex mcompact-casesi
13007 Enable compact casesi pattern.
13008 This is the default for @option{-Os}.
13010 @item -mno-cond-exec
13011 @opindex mno-cond-exec
13012 Disable ARCompact specific pass to generate conditional execution instructions.
13013 Due to delay slot scheduling and interactions between operand numbers,
13014 literal sizes, instruction lengths, and the support for conditional execution,
13015 the target-independent pass to generate conditional execution is often lacking,
13016 so the ARC port has kept a special pass around that tries to find more
13017 conditional execution generating opportunities after register allocation,
13018 branch shortening, and delay slot scheduling have been done. This pass
13019 generally, but not always, improves performance and code size, at the cost of
13020 extra compilation time, which is why there is an option to switch it off.
13021 If you have a problem with call instructions exceeding their allowable
13022 offset range because they are conditionalized, you should consider using
13023 @option{-mmedium-calls} instead.
13025 @item -mearly-cbranchsi
13026 @opindex mearly-cbranchsi
13027 Enable pre-reload use of the cbranchsi pattern.
13029 @item -mexpand-adddi
13030 @opindex mexpand-adddi
13031 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13032 @code{add.f}, @code{adc} etc.
13034 @item -mindexed-loads
13035 @opindex mindexed-loads
13036 Enable the use of indexed loads. This can be problematic because some
13037 optimizers then assume that indexed stores exist, which is not
13042 Enable Local Register Allocation. This is still experimental for ARC,
13043 so by default the compiler uses standard reload
13044 (i.e. @option{-mno-lra}).
13046 @item -mlra-priority-none
13047 @opindex mlra-priority-none
13048 Don't indicate any priority for target registers.
13050 @item -mlra-priority-compact
13051 @opindex mlra-priority-compact
13052 Indicate target register priority for r0..r3 / r12..r15.
13054 @item -mlra-priority-noncompact
13055 @opindex mlra-priority-noncompact
13056 Reduce target regsiter priority for r0..r3 / r12..r15.
13058 @item -mno-millicode
13059 @opindex mno-millicode
13060 When optimizing for size (using @option{-Os}), prologues and epilogues
13061 that have to save or restore a large number of registers are often
13062 shortened by using call to a special function in libgcc; this is
13063 referred to as a @emph{millicode} call. As these calls can pose
13064 performance issues, and/or cause linking issues when linking in a
13065 nonstandard way, this option is provided to turn off millicode call
13069 @opindex mmixed-code
13070 Tweak register allocation to help 16-bit instruction generation.
13071 This generally has the effect of decreasing the average instruction size
13072 while increasing the instruction count.
13076 Enable 'q' instruction alternatives.
13077 This is the default for @option{-Os}.
13081 Enable Rcq constraint handling - most short code generation depends on this.
13082 This is the default.
13086 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13087 This is the default.
13089 @item -msize-level=@var{level}
13090 @opindex msize-level
13091 Fine-tune size optimization with regards to instruction lengths and alignment.
13092 The recognized values for @var{level} are:
13095 No size optimization. This level is deprecated and treated like @samp{1}.
13098 Short instructions are used opportunistically.
13101 In addition, alignment of loops and of code after barriers are dropped.
13104 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13108 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13109 the behavior when this is not set is equivalent to level @samp{1}.
13111 @item -mtune=@var{cpu}
13113 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13114 by @option{-mcpu=}.
13116 Supported values for @var{cpu} are
13120 Tune for ARC600 cpu.
13123 Tune for ARC601 cpu.
13126 Tune for ARC700 cpu with standard multiplier block.
13129 Tune for ARC700 cpu with XMAC block.
13132 Tune for ARC725D cpu.
13135 Tune for ARC750D cpu.
13139 @item -mmultcost=@var{num}
13141 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13142 normal instruction.
13144 @item -munalign-prob-threshold=@var{probability}
13145 @opindex munalign-prob-threshold
13146 Set probability threshold for unaligning branches.
13147 When tuning for @samp{ARC700} and optimizing for speed, branches without
13148 filled delay slot are preferably emitted unaligned and long, unless
13149 profiling indicates that the probability for the branch to be taken
13150 is below @var{probability}. @xref{Cross-profiling}.
13151 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13155 The following options are maintained for backward compatibility, but
13156 are now deprecated and will be removed in a future release:
13158 @c Deprecated options
13166 @opindex mbig-endian
13169 Compile code for big endian targets. Use of these options is now
13170 deprecated. Users wanting big-endian code, should use the
13171 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13172 building the tool chain, for which big-endian is the default.
13174 @item -mlittle-endian
13175 @opindex mlittle-endian
13178 Compile code for little endian targets. Use of these options is now
13179 deprecated. Users wanting little-endian code should use the
13180 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13181 building the tool chain, for which little-endian is the default.
13183 @item -mbarrel_shifter
13184 @opindex mbarrel_shifter
13185 Replaced by @option{-mbarrel-shifter}.
13187 @item -mdpfp_compact
13188 @opindex mdpfp_compact
13189 Replaced by @option{-mdpfp-compact}.
13192 @opindex mdpfp_fast
13193 Replaced by @option{-mdpfp-fast}.
13196 @opindex mdsp_packa
13197 Replaced by @option{-mdsp-packa}.
13201 Replaced by @option{-mea}.
13205 Replaced by @option{-mmac-24}.
13209 Replaced by @option{-mmac-d16}.
13211 @item -mspfp_compact
13212 @opindex mspfp_compact
13213 Replaced by @option{-mspfp-compact}.
13216 @opindex mspfp_fast
13217 Replaced by @option{-mspfp-fast}.
13219 @item -mtune=@var{cpu}
13221 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13222 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13223 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13225 @item -multcost=@var{num}
13227 Replaced by @option{-mmultcost}.
13232 @subsection ARM Options
13233 @cindex ARM options
13235 These @samp{-m} options are defined for the ARM port:
13238 @item -mabi=@var{name}
13240 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13241 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13244 @opindex mapcs-frame
13245 Generate a stack frame that is compliant with the ARM Procedure Call
13246 Standard for all functions, even if this is not strictly necessary for
13247 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13248 with this option causes the stack frames not to be generated for
13249 leaf functions. The default is @option{-mno-apcs-frame}.
13250 This option is deprecated.
13254 This is a synonym for @option{-mapcs-frame} and is deprecated.
13257 @c not currently implemented
13258 @item -mapcs-stack-check
13259 @opindex mapcs-stack-check
13260 Generate code to check the amount of stack space available upon entry to
13261 every function (that actually uses some stack space). If there is
13262 insufficient space available then either the function
13263 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13264 called, depending upon the amount of stack space required. The runtime
13265 system is required to provide these functions. The default is
13266 @option{-mno-apcs-stack-check}, since this produces smaller code.
13268 @c not currently implemented
13270 @opindex mapcs-float
13271 Pass floating-point arguments using the floating-point registers. This is
13272 one of the variants of the APCS@. This option is recommended if the
13273 target hardware has a floating-point unit or if a lot of floating-point
13274 arithmetic is going to be performed by the code. The default is
13275 @option{-mno-apcs-float}, since the size of integer-only code is
13276 slightly increased if @option{-mapcs-float} is used.
13278 @c not currently implemented
13279 @item -mapcs-reentrant
13280 @opindex mapcs-reentrant
13281 Generate reentrant, position-independent code. The default is
13282 @option{-mno-apcs-reentrant}.
13285 @item -mthumb-interwork
13286 @opindex mthumb-interwork
13287 Generate code that supports calling between the ARM and Thumb
13288 instruction sets. Without this option, on pre-v5 architectures, the
13289 two instruction sets cannot be reliably used inside one program. The
13290 default is @option{-mno-thumb-interwork}, since slightly larger code
13291 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13292 configurations this option is meaningless.
13294 @item -mno-sched-prolog
13295 @opindex mno-sched-prolog
13296 Prevent the reordering of instructions in the function prologue, or the
13297 merging of those instruction with the instructions in the function's
13298 body. This means that all functions start with a recognizable set
13299 of instructions (or in fact one of a choice from a small set of
13300 different function prologues), and this information can be used to
13301 locate the start of functions inside an executable piece of code. The
13302 default is @option{-msched-prolog}.
13304 @item -mfloat-abi=@var{name}
13305 @opindex mfloat-abi
13306 Specifies which floating-point ABI to use. Permissible values
13307 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13309 Specifying @samp{soft} causes GCC to generate output containing
13310 library calls for floating-point operations.
13311 @samp{softfp} allows the generation of code using hardware floating-point
13312 instructions, but still uses the soft-float calling conventions.
13313 @samp{hard} allows generation of floating-point instructions
13314 and uses FPU-specific calling conventions.
13316 The default depends on the specific target configuration. Note that
13317 the hard-float and soft-float ABIs are not link-compatible; you must
13318 compile your entire program with the same ABI, and link with a
13319 compatible set of libraries.
13321 @item -mlittle-endian
13322 @opindex mlittle-endian
13323 Generate code for a processor running in little-endian mode. This is
13324 the default for all standard configurations.
13327 @opindex mbig-endian
13328 Generate code for a processor running in big-endian mode; the default is
13329 to compile code for a little-endian processor.
13331 @item -march=@var{name}
13333 This specifies the name of the target ARM architecture. GCC uses this
13334 name to determine what kind of instructions it can emit when generating
13335 assembly code. This option can be used in conjunction with or instead
13336 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13337 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13338 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13339 @samp{armv6}, @samp{armv6j},
13340 @samp{armv6t2}, @samp{armv6z}, @samp{armv6kz}, @samp{armv6-m},
13341 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13342 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc},
13343 @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13345 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13348 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13349 architecture together with the optional CRC32 extensions.
13351 @option{-march=native} causes the compiler to auto-detect the architecture
13352 of the build computer. At present, this feature is only supported on
13353 GNU/Linux, and not all architectures are recognized. If the auto-detect
13354 is unsuccessful the option has no effect.
13356 @item -mtune=@var{name}
13358 This option specifies the name of the target ARM processor for
13359 which GCC should tune the performance of the code.
13360 For some ARM implementations better performance can be obtained by using
13362 Permissible names are: @samp{arm2}, @samp{arm250},
13363 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13364 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13365 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13366 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13368 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13369 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13370 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13371 @samp{strongarm1110},
13372 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13373 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13374 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13375 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13376 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13377 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13378 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13379 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
13380 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
13381 @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
13383 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13388 @samp{cortex-m0plus},
13389 @samp{cortex-m1.small-multiply},
13390 @samp{cortex-m0.small-multiply},
13391 @samp{cortex-m0plus.small-multiply},
13393 @samp{marvell-pj4},
13394 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13395 @samp{fa526}, @samp{fa626},
13396 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13399 Additionally, this option can specify that GCC should tune the performance
13400 of the code for a big.LITTLE system. Permissible names are:
13401 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
13402 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
13404 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13405 performance for a blend of processors within architecture @var{arch}.
13406 The aim is to generate code that run well on the current most popular
13407 processors, balancing between optimizations that benefit some CPUs in the
13408 range, and avoiding performance pitfalls of other CPUs. The effects of
13409 this option may change in future GCC versions as CPU models come and go.
13411 @option{-mtune=native} causes the compiler to auto-detect the CPU
13412 of the build computer. At present, this feature is only supported on
13413 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13414 unsuccessful the option has no effect.
13416 @item -mcpu=@var{name}
13418 This specifies the name of the target ARM processor. GCC uses this name
13419 to derive the name of the target ARM architecture (as if specified
13420 by @option{-march}) and the ARM processor type for which to tune for
13421 performance (as if specified by @option{-mtune}). Where this option
13422 is used in conjunction with @option{-march} or @option{-mtune},
13423 those options take precedence over the appropriate part of this option.
13425 Permissible names for this option are the same as those for
13428 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13429 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13430 See @option{-mtune} for more information.
13432 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13433 of the build computer. At present, this feature is only supported on
13434 GNU/Linux, and not all architectures are recognized. If the auto-detect
13435 is unsuccessful the option has no effect.
13437 @item -mfpu=@var{name}
13439 This specifies what floating-point hardware (or hardware emulation) is
13440 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13441 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13442 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13443 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13444 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13445 @samp{fp-armv8}, @samp{neon-fp-armv8}, and @samp{crypto-neon-fp-armv8}.
13447 If @option{-msoft-float} is specified this specifies the format of
13448 floating-point values.
13450 If the selected floating-point hardware includes the NEON extension
13451 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13452 operations are not generated by GCC's auto-vectorization pass unless
13453 @option{-funsafe-math-optimizations} is also specified. This is
13454 because NEON hardware does not fully implement the IEEE 754 standard for
13455 floating-point arithmetic (in particular denormal values are treated as
13456 zero), so the use of NEON instructions may lead to a loss of precision.
13458 @item -mfp16-format=@var{name}
13459 @opindex mfp16-format
13460 Specify the format of the @code{__fp16} half-precision floating-point type.
13461 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13462 the default is @samp{none}, in which case the @code{__fp16} type is not
13463 defined. @xref{Half-Precision}, for more information.
13465 @item -mstructure-size-boundary=@var{n}
13466 @opindex mstructure-size-boundary
13467 The sizes of all structures and unions are rounded up to a multiple
13468 of the number of bits set by this option. Permissible values are 8, 32
13469 and 64. The default value varies for different toolchains. For the COFF
13470 targeted toolchain the default value is 8. A value of 64 is only allowed
13471 if the underlying ABI supports it.
13473 Specifying a larger number can produce faster, more efficient code, but
13474 can also increase the size of the program. Different values are potentially
13475 incompatible. Code compiled with one value cannot necessarily expect to
13476 work with code or libraries compiled with another value, if they exchange
13477 information using structures or unions.
13479 @item -mabort-on-noreturn
13480 @opindex mabort-on-noreturn
13481 Generate a call to the function @code{abort} at the end of a
13482 @code{noreturn} function. It is executed if the function tries to
13486 @itemx -mno-long-calls
13487 @opindex mlong-calls
13488 @opindex mno-long-calls
13489 Tells the compiler to perform function calls by first loading the
13490 address of the function into a register and then performing a subroutine
13491 call on this register. This switch is needed if the target function
13492 lies outside of the 64-megabyte addressing range of the offset-based
13493 version of subroutine call instruction.
13495 Even if this switch is enabled, not all function calls are turned
13496 into long calls. The heuristic is that static functions, functions
13497 that have the @code{short_call} attribute, functions that are inside
13498 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13499 definitions have already been compiled within the current compilation
13500 unit are not turned into long calls. The exceptions to this rule are
13501 that weak function definitions, functions with the @code{long_call}
13502 attribute or the @code{section} attribute, and functions that are within
13503 the scope of a @code{#pragma long_calls} directive are always
13504 turned into long calls.
13506 This feature is not enabled by default. Specifying
13507 @option{-mno-long-calls} restores the default behavior, as does
13508 placing the function calls within the scope of a @code{#pragma
13509 long_calls_off} directive. Note these switches have no effect on how
13510 the compiler generates code to handle function calls via function
13513 @item -msingle-pic-base
13514 @opindex msingle-pic-base
13515 Treat the register used for PIC addressing as read-only, rather than
13516 loading it in the prologue for each function. The runtime system is
13517 responsible for initializing this register with an appropriate value
13518 before execution begins.
13520 @item -mpic-register=@var{reg}
13521 @opindex mpic-register
13522 Specify the register to be used for PIC addressing.
13523 For standard PIC base case, the default is any suitable register
13524 determined by compiler. For single PIC base case, the default is
13525 @samp{R9} if target is EABI based or stack-checking is enabled,
13526 otherwise the default is @samp{R10}.
13528 @item -mpic-data-is-text-relative
13529 @opindex mpic-data-is-text-relative
13530 Assume that each data segments are relative to text segment at load time.
13531 Therefore, it permits addressing data using PC-relative operations.
13532 This option is on by default for targets other than VxWorks RTP.
13534 @item -mpoke-function-name
13535 @opindex mpoke-function-name
13536 Write the name of each function into the text section, directly
13537 preceding the function prologue. The generated code is similar to this:
13541 .ascii "arm_poke_function_name", 0
13544 .word 0xff000000 + (t1 - t0)
13545 arm_poke_function_name
13547 stmfd sp!, @{fp, ip, lr, pc@}
13551 When performing a stack backtrace, code can inspect the value of
13552 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13553 location @code{pc - 12} and the top 8 bits are set, then we know that
13554 there is a function name embedded immediately preceding this location
13555 and has length @code{((pc[-3]) & 0xff000000)}.
13562 Select between generating code that executes in ARM and Thumb
13563 states. The default for most configurations is to generate code
13564 that executes in ARM state, but the default can be changed by
13565 configuring GCC with the @option{--with-mode=}@var{state}
13568 You can also override the ARM and Thumb mode for each function
13569 by using the @code{target("thumb")} and @code{target("arm")} function attributes
13570 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13573 @opindex mtpcs-frame
13574 Generate a stack frame that is compliant with the Thumb Procedure Call
13575 Standard for all non-leaf functions. (A leaf function is one that does
13576 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13578 @item -mtpcs-leaf-frame
13579 @opindex mtpcs-leaf-frame
13580 Generate a stack frame that is compliant with the Thumb Procedure Call
13581 Standard for all leaf functions. (A leaf function is one that does
13582 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13584 @item -mcallee-super-interworking
13585 @opindex mcallee-super-interworking
13586 Gives all externally visible functions in the file being compiled an ARM
13587 instruction set header which switches to Thumb mode before executing the
13588 rest of the function. This allows these functions to be called from
13589 non-interworking code. This option is not valid in AAPCS configurations
13590 because interworking is enabled by default.
13592 @item -mcaller-super-interworking
13593 @opindex mcaller-super-interworking
13594 Allows calls via function pointers (including virtual functions) to
13595 execute correctly regardless of whether the target code has been
13596 compiled for interworking or not. There is a small overhead in the cost
13597 of executing a function pointer if this option is enabled. This option
13598 is not valid in AAPCS configurations because interworking is enabled
13601 @item -mtp=@var{name}
13603 Specify the access model for the thread local storage pointer. The valid
13604 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13605 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13606 (supported in the arm6k architecture), and @samp{auto}, which uses the
13607 best available method for the selected processor. The default setting is
13610 @item -mtls-dialect=@var{dialect}
13611 @opindex mtls-dialect
13612 Specify the dialect to use for accessing thread local storage. Two
13613 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13614 @samp{gnu} dialect selects the original GNU scheme for supporting
13615 local and global dynamic TLS models. The @samp{gnu2} dialect
13616 selects the GNU descriptor scheme, which provides better performance
13617 for shared libraries. The GNU descriptor scheme is compatible with
13618 the original scheme, but does require new assembler, linker and
13619 library support. Initial and local exec TLS models are unaffected by
13620 this option and always use the original scheme.
13622 @item -mword-relocations
13623 @opindex mword-relocations
13624 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13625 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13626 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13629 @item -mfix-cortex-m3-ldrd
13630 @opindex mfix-cortex-m3-ldrd
13631 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13632 with overlapping destination and base registers are used. This option avoids
13633 generating these instructions. This option is enabled by default when
13634 @option{-mcpu=cortex-m3} is specified.
13636 @item -munaligned-access
13637 @itemx -mno-unaligned-access
13638 @opindex munaligned-access
13639 @opindex mno-unaligned-access
13640 Enables (or disables) reading and writing of 16- and 32- bit values
13641 from addresses that are not 16- or 32- bit aligned. By default
13642 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13643 architectures, and enabled for all other architectures. If unaligned
13644 access is not enabled then words in packed data structures are
13645 accessed a byte at a time.
13647 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13648 generated object file to either true or false, depending upon the
13649 setting of this option. If unaligned access is enabled then the
13650 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13653 @item -mneon-for-64bits
13654 @opindex mneon-for-64bits
13655 Enables using Neon to handle scalar 64-bits operations. This is
13656 disabled by default since the cost of moving data from core registers
13659 @item -mslow-flash-data
13660 @opindex mslow-flash-data
13661 Assume loading data from flash is slower than fetching instruction.
13662 Therefore literal load is minimized for better performance.
13663 This option is only supported when compiling for ARMv7 M-profile and
13666 @item -masm-syntax-unified
13667 @opindex masm-syntax-unified
13668 Assume inline assembler is using unified asm syntax. The default is
13669 currently off which implies divided syntax. Currently this option is
13670 available only for Thumb1 and has no effect on ARM state and Thumb2.
13671 However, this may change in future releases of GCC. Divided syntax
13672 should be considered deprecated.
13674 @item -mrestrict-it
13675 @opindex mrestrict-it
13676 Restricts generation of IT blocks to conform to the rules of ARMv8.
13677 IT blocks can only contain a single 16-bit instruction from a select
13678 set of instructions. This option is on by default for ARMv8 Thumb mode.
13680 @item -mprint-tune-info
13681 @opindex mprint-tune-info
13682 Print CPU tuning information as comment in assembler file. This is
13683 an option used only for regression testing of the compiler and not
13684 intended for ordinary use in compiling code. This option is disabled
13689 @subsection AVR Options
13690 @cindex AVR Options
13692 These options are defined for AVR implementations:
13695 @item -mmcu=@var{mcu}
13697 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13699 The default for this option is@tie{}@samp{avr2}.
13701 GCC supports the following AVR devices and ISAs:
13703 @include avr-mmcu.texi
13705 @item -maccumulate-args
13706 @opindex maccumulate-args
13707 Accumulate outgoing function arguments and acquire/release the needed
13708 stack space for outgoing function arguments once in function
13709 prologue/epilogue. Without this option, outgoing arguments are pushed
13710 before calling a function and popped afterwards.
13712 Popping the arguments after the function call can be expensive on
13713 AVR so that accumulating the stack space might lead to smaller
13714 executables because arguments need not to be removed from the
13715 stack after such a function call.
13717 This option can lead to reduced code size for functions that perform
13718 several calls to functions that get their arguments on the stack like
13719 calls to printf-like functions.
13721 @item -mbranch-cost=@var{cost}
13722 @opindex mbranch-cost
13723 Set the branch costs for conditional branch instructions to
13724 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13725 integers. The default branch cost is 0.
13727 @item -mcall-prologues
13728 @opindex mcall-prologues
13729 Functions prologues/epilogues are expanded as calls to appropriate
13730 subroutines. Code size is smaller.
13734 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13735 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13736 and @code{long long} is 4 bytes. Please note that this option does not
13737 conform to the C standards, but it results in smaller code
13740 @item -mn-flash=@var{num}
13742 Assume that the flash memory has a size of
13743 @var{num} times 64@tie{}KiB.
13745 @item -mno-interrupts
13746 @opindex mno-interrupts
13747 Generated code is not compatible with hardware interrupts.
13748 Code size is smaller.
13752 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13753 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13754 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13755 the assembler's command line and the @option{--relax} option to the
13756 linker's command line.
13758 Jump relaxing is performed by the linker because jump offsets are not
13759 known before code is located. Therefore, the assembler code generated by the
13760 compiler is the same, but the instructions in the executable may
13761 differ from instructions in the assembler code.
13763 Relaxing must be turned on if linker stubs are needed, see the
13764 section on @code{EIND} and linker stubs below.
13768 Assume that the device supports the Read-Modify-Write
13769 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13773 Treat the stack pointer register as an 8-bit register,
13774 i.e.@: assume the high byte of the stack pointer is zero.
13775 In general, you don't need to set this option by hand.
13777 This option is used internally by the compiler to select and
13778 build multilibs for architectures @code{avr2} and @code{avr25}.
13779 These architectures mix devices with and without @code{SPH}.
13780 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13781 the compiler driver adds or removes this option from the compiler
13782 proper's command line, because the compiler then knows if the device
13783 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13788 Use address register @code{X} in a way proposed by the hardware. This means
13789 that @code{X} is only used in indirect, post-increment or
13790 pre-decrement addressing.
13792 Without this option, the @code{X} register may be used in the same way
13793 as @code{Y} or @code{Z} which then is emulated by additional
13795 For example, loading a value with @code{X+const} addressing with a
13796 small non-negative @code{const < 64} to a register @var{Rn} is
13800 adiw r26, const ; X += const
13801 ld @var{Rn}, X ; @var{Rn} = *X
13802 sbiw r26, const ; X -= const
13806 @opindex mtiny-stack
13807 Only change the lower 8@tie{}bits of the stack pointer.
13810 @opindex nodevicelib
13811 Don't link against AVR-LibC's device specific library @code{libdev.a}.
13813 @item -Waddr-space-convert
13814 @opindex Waddr-space-convert
13815 Warn about conversions between address spaces in the case where the
13816 resulting address space is not contained in the incoming address space.
13819 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
13820 @cindex @code{EIND}
13821 Pointers in the implementation are 16@tie{}bits wide.
13822 The address of a function or label is represented as word address so
13823 that indirect jumps and calls can target any code address in the
13824 range of 64@tie{}Ki words.
13826 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
13827 bytes of program memory space, there is a special function register called
13828 @code{EIND} that serves as most significant part of the target address
13829 when @code{EICALL} or @code{EIJMP} instructions are used.
13831 Indirect jumps and calls on these devices are handled as follows by
13832 the compiler and are subject to some limitations:
13837 The compiler never sets @code{EIND}.
13840 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
13841 instructions or might read @code{EIND} directly in order to emulate an
13842 indirect call/jump by means of a @code{RET} instruction.
13845 The compiler assumes that @code{EIND} never changes during the startup
13846 code or during the application. In particular, @code{EIND} is not
13847 saved/restored in function or interrupt service routine
13851 For indirect calls to functions and computed goto, the linker
13852 generates @emph{stubs}. Stubs are jump pads sometimes also called
13853 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
13854 The stub contains a direct jump to the desired address.
13857 Linker relaxation must be turned on so that the linker generates
13858 the stubs correctly in all situations. See the compiler option
13859 @option{-mrelax} and the linker option @option{--relax}.
13860 There are corner cases where the linker is supposed to generate stubs
13861 but aborts without relaxation and without a helpful error message.
13864 The default linker script is arranged for code with @code{EIND = 0}.
13865 If code is supposed to work for a setup with @code{EIND != 0}, a custom
13866 linker script has to be used in order to place the sections whose
13867 name start with @code{.trampolines} into the segment where @code{EIND}
13871 The startup code from libgcc never sets @code{EIND}.
13872 Notice that startup code is a blend of code from libgcc and AVR-LibC.
13873 For the impact of AVR-LibC on @code{EIND}, see the
13874 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
13877 It is legitimate for user-specific startup code to set up @code{EIND}
13878 early, for example by means of initialization code located in
13879 section @code{.init3}. Such code runs prior to general startup code
13880 that initializes RAM and calls constructors, but after the bit
13881 of startup code from AVR-LibC that sets @code{EIND} to the segment
13882 where the vector table is located.
13884 #include <avr/io.h>
13887 __attribute__((section(".init3"),naked,used,no_instrument_function))
13888 init3_set_eind (void)
13890 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
13891 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
13896 The @code{__trampolines_start} symbol is defined in the linker script.
13899 Stubs are generated automatically by the linker if
13900 the following two conditions are met:
13903 @item The address of a label is taken by means of the @code{gs} modifier
13904 (short for @emph{generate stubs}) like so:
13906 LDI r24, lo8(gs(@var{func}))
13907 LDI r25, hi8(gs(@var{func}))
13909 @item The final location of that label is in a code segment
13910 @emph{outside} the segment where the stubs are located.
13914 The compiler emits such @code{gs} modifiers for code labels in the
13915 following situations:
13917 @item Taking address of a function or code label.
13918 @item Computed goto.
13919 @item If prologue-save function is used, see @option{-mcall-prologues}
13920 command-line option.
13921 @item Switch/case dispatch tables. If you do not want such dispatch
13922 tables you can specify the @option{-fno-jump-tables} command-line option.
13923 @item C and C++ constructors/destructors called during startup/shutdown.
13924 @item If the tools hit a @code{gs()} modifier explained above.
13928 Jumping to non-symbolic addresses like so is @emph{not} supported:
13933 /* Call function at word address 0x2 */
13934 return ((int(*)(void)) 0x2)();
13938 Instead, a stub has to be set up, i.e.@: the function has to be called
13939 through a symbol (@code{func_4} in the example):
13944 extern int func_4 (void);
13946 /* Call function at byte address 0x4 */
13951 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
13952 Alternatively, @code{func_4} can be defined in the linker script.
13955 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
13956 @cindex @code{RAMPD}
13957 @cindex @code{RAMPX}
13958 @cindex @code{RAMPY}
13959 @cindex @code{RAMPZ}
13960 Some AVR devices support memories larger than the 64@tie{}KiB range
13961 that can be accessed with 16-bit pointers. To access memory locations
13962 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
13963 register is used as high part of the address:
13964 The @code{X}, @code{Y}, @code{Z} address register is concatenated
13965 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
13966 register, respectively, to get a wide address. Similarly,
13967 @code{RAMPD} is used together with direct addressing.
13971 The startup code initializes the @code{RAMP} special function
13972 registers with zero.
13975 If a @ref{AVR Named Address Spaces,named address space} other than
13976 generic or @code{__flash} is used, then @code{RAMPZ} is set
13977 as needed before the operation.
13980 If the device supports RAM larger than 64@tie{}KiB and the compiler
13981 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
13982 is reset to zero after the operation.
13985 If the device comes with a specific @code{RAMP} register, the ISR
13986 prologue/epilogue saves/restores that SFR and initializes it with
13987 zero in case the ISR code might (implicitly) use it.
13990 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
13991 If you use inline assembler to read from locations outside the
13992 16-bit address range and change one of the @code{RAMP} registers,
13993 you must reset it to zero after the access.
13997 @subsubsection AVR Built-in Macros
13999 GCC defines several built-in macros so that the user code can test
14000 for the presence or absence of features. Almost any of the following
14001 built-in macros are deduced from device capabilities and thus
14002 triggered by the @option{-mmcu=} command-line option.
14004 For even more AVR-specific built-in macros see
14005 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14010 Build-in macro that resolves to a decimal number that identifies the
14011 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14012 Possible values are:
14014 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14015 @code{4}, @code{5}, @code{51}, @code{6}
14017 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14018 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14022 @code{100}, @code{102}, @code{104},
14023 @code{105}, @code{106}, @code{107}
14025 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14026 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14027 If @var{mcu} specifies a device, this built-in macro is set
14028 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14029 defined to @code{4}.
14031 @item __AVR_@var{Device}__
14032 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14033 the device's name. For example, @option{-mmcu=atmega8} defines the
14034 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14035 @code{__AVR_ATtiny261A__}, etc.
14037 The built-in macros' names follow
14038 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14039 the device name as from the AVR user manual. The difference between
14040 @var{Device} in the built-in macro and @var{device} in
14041 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14043 If @var{device} is not a device but only a core architecture like
14044 @samp{avr51}, this macro is not defined.
14046 @item __AVR_DEVICE_NAME__
14047 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14048 the device's name. For example, with @option{-mmcu=atmega8} the macro
14049 is defined to @code{atmega8}.
14051 If @var{device} is not a device but only a core architecture like
14052 @samp{avr51}, this macro is not defined.
14054 @item __AVR_XMEGA__
14055 The device / architecture belongs to the XMEGA family of devices.
14057 @item __AVR_HAVE_ELPM__
14058 The device has the @code{ELPM} instruction.
14060 @item __AVR_HAVE_ELPMX__
14061 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14062 R@var{n},Z+} instructions.
14064 @item __AVR_HAVE_MOVW__
14065 The device has the @code{MOVW} instruction to perform 16-bit
14066 register-register moves.
14068 @item __AVR_HAVE_LPMX__
14069 The device has the @code{LPM R@var{n},Z} and
14070 @code{LPM R@var{n},Z+} instructions.
14072 @item __AVR_HAVE_MUL__
14073 The device has a hardware multiplier.
14075 @item __AVR_HAVE_JMP_CALL__
14076 The device has the @code{JMP} and @code{CALL} instructions.
14077 This is the case for devices with at least 16@tie{}KiB of program
14080 @item __AVR_HAVE_EIJMP_EICALL__
14081 @itemx __AVR_3_BYTE_PC__
14082 The device has the @code{EIJMP} and @code{EICALL} instructions.
14083 This is the case for devices with more than 128@tie{}KiB of program memory.
14084 This also means that the program counter
14085 (PC) is 3@tie{}bytes wide.
14087 @item __AVR_2_BYTE_PC__
14088 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14089 with up to 128@tie{}KiB of program memory.
14091 @item __AVR_HAVE_8BIT_SP__
14092 @itemx __AVR_HAVE_16BIT_SP__
14093 The stack pointer (SP) register is treated as 8-bit respectively
14094 16-bit register by the compiler.
14095 The definition of these macros is affected by @option{-mtiny-stack}.
14097 @item __AVR_HAVE_SPH__
14099 The device has the SPH (high part of stack pointer) special function
14100 register or has an 8-bit stack pointer, respectively.
14101 The definition of these macros is affected by @option{-mmcu=} and
14102 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14105 @item __AVR_HAVE_RAMPD__
14106 @itemx __AVR_HAVE_RAMPX__
14107 @itemx __AVR_HAVE_RAMPY__
14108 @itemx __AVR_HAVE_RAMPZ__
14109 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14110 @code{RAMPZ} special function register, respectively.
14112 @item __NO_INTERRUPTS__
14113 This macro reflects the @option{-mno-interrupts} command-line option.
14115 @item __AVR_ERRATA_SKIP__
14116 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14117 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14118 instructions because of a hardware erratum. Skip instructions are
14119 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14120 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14123 @item __AVR_ISA_RMW__
14124 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14126 @item __AVR_SFR_OFFSET__=@var{offset}
14127 Instructions that can address I/O special function registers directly
14128 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14129 address as if addressed by an instruction to access RAM like @code{LD}
14130 or @code{STS}. This offset depends on the device architecture and has
14131 to be subtracted from the RAM address in order to get the
14132 respective I/O@tie{}address.
14134 @item __WITH_AVRLIBC__
14135 The compiler is configured to be used together with AVR-Libc.
14136 See the @option{--with-avrlibc} configure option.
14140 @node Blackfin Options
14141 @subsection Blackfin Options
14142 @cindex Blackfin Options
14145 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14147 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14148 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14149 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14150 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14151 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14152 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14153 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14154 @samp{bf561}, @samp{bf592}.
14156 The optional @var{sirevision} specifies the silicon revision of the target
14157 Blackfin processor. Any workarounds available for the targeted silicon revision
14158 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14159 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14160 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14161 hexadecimal digits representing the major and minor numbers in the silicon
14162 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14163 is not defined. If @var{sirevision} is @samp{any}, the
14164 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14165 If this optional @var{sirevision} is not used, GCC assumes the latest known
14166 silicon revision of the targeted Blackfin processor.
14168 GCC defines a preprocessor macro for the specified @var{cpu}.
14169 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14170 provided by libgloss to be linked in if @option{-msim} is not given.
14172 Without this option, @samp{bf532} is used as the processor by default.
14174 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14175 only the preprocessor macro is defined.
14179 Specifies that the program will be run on the simulator. This causes
14180 the simulator BSP provided by libgloss to be linked in. This option
14181 has effect only for @samp{bfin-elf} toolchain.
14182 Certain other options, such as @option{-mid-shared-library} and
14183 @option{-mfdpic}, imply @option{-msim}.
14185 @item -momit-leaf-frame-pointer
14186 @opindex momit-leaf-frame-pointer
14187 Don't keep the frame pointer in a register for leaf functions. This
14188 avoids the instructions to save, set up and restore frame pointers and
14189 makes an extra register available in leaf functions. The option
14190 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14191 which might make debugging harder.
14193 @item -mspecld-anomaly
14194 @opindex mspecld-anomaly
14195 When enabled, the compiler ensures that the generated code does not
14196 contain speculative loads after jump instructions. If this option is used,
14197 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14199 @item -mno-specld-anomaly
14200 @opindex mno-specld-anomaly
14201 Don't generate extra code to prevent speculative loads from occurring.
14203 @item -mcsync-anomaly
14204 @opindex mcsync-anomaly
14205 When enabled, the compiler ensures that the generated code does not
14206 contain CSYNC or SSYNC instructions too soon after conditional branches.
14207 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14209 @item -mno-csync-anomaly
14210 @opindex mno-csync-anomaly
14211 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14212 occurring too soon after a conditional branch.
14216 When enabled, the compiler is free to take advantage of the knowledge that
14217 the entire program fits into the low 64k of memory.
14220 @opindex mno-low-64k
14221 Assume that the program is arbitrarily large. This is the default.
14223 @item -mstack-check-l1
14224 @opindex mstack-check-l1
14225 Do stack checking using information placed into L1 scratchpad memory by the
14228 @item -mid-shared-library
14229 @opindex mid-shared-library
14230 Generate code that supports shared libraries via the library ID method.
14231 This allows for execute in place and shared libraries in an environment
14232 without virtual memory management. This option implies @option{-fPIC}.
14233 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14235 @item -mno-id-shared-library
14236 @opindex mno-id-shared-library
14237 Generate code that doesn't assume ID-based shared libraries are being used.
14238 This is the default.
14240 @item -mleaf-id-shared-library
14241 @opindex mleaf-id-shared-library
14242 Generate code that supports shared libraries via the library ID method,
14243 but assumes that this library or executable won't link against any other
14244 ID shared libraries. That allows the compiler to use faster code for jumps
14247 @item -mno-leaf-id-shared-library
14248 @opindex mno-leaf-id-shared-library
14249 Do not assume that the code being compiled won't link against any ID shared
14250 libraries. Slower code is generated for jump and call insns.
14252 @item -mshared-library-id=n
14253 @opindex mshared-library-id
14254 Specifies the identification number of the ID-based shared library being
14255 compiled. Specifying a value of 0 generates more compact code; specifying
14256 other values forces the allocation of that number to the current
14257 library but is no more space- or time-efficient than omitting this option.
14261 Generate code that allows the data segment to be located in a different
14262 area of memory from the text segment. This allows for execute in place in
14263 an environment without virtual memory management by eliminating relocations
14264 against the text section.
14266 @item -mno-sep-data
14267 @opindex mno-sep-data
14268 Generate code that assumes that the data segment follows the text segment.
14269 This is the default.
14272 @itemx -mno-long-calls
14273 @opindex mlong-calls
14274 @opindex mno-long-calls
14275 Tells the compiler to perform function calls by first loading the
14276 address of the function into a register and then performing a subroutine
14277 call on this register. This switch is needed if the target function
14278 lies outside of the 24-bit addressing range of the offset-based
14279 version of subroutine call instruction.
14281 This feature is not enabled by default. Specifying
14282 @option{-mno-long-calls} restores the default behavior. Note these
14283 switches have no effect on how the compiler generates code to handle
14284 function calls via function pointers.
14288 Link with the fast floating-point library. This library relaxes some of
14289 the IEEE floating-point standard's rules for checking inputs against
14290 Not-a-Number (NAN), in the interest of performance.
14293 @opindex minline-plt
14294 Enable inlining of PLT entries in function calls to functions that are
14295 not known to bind locally. It has no effect without @option{-mfdpic}.
14298 @opindex mmulticore
14299 Build a standalone application for multicore Blackfin processors.
14300 This option causes proper start files and link scripts supporting
14301 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14302 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14304 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14305 selects the one-application-per-core programming model. Without
14306 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14307 programming model is used. In this model, the main function of Core B
14308 should be named as @code{coreb_main}.
14310 If this option is not used, the single-core application programming
14315 Build a standalone application for Core A of BF561 when using
14316 the one-application-per-core programming model. Proper start files
14317 and link scripts are used to support Core A, and the macro
14318 @code{__BFIN_COREA} is defined.
14319 This option can only be used in conjunction with @option{-mmulticore}.
14323 Build a standalone application for Core B of BF561 when using
14324 the one-application-per-core programming model. Proper start files
14325 and link scripts are used to support Core B, and the macro
14326 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14327 should be used instead of @code{main}.
14328 This option can only be used in conjunction with @option{-mmulticore}.
14332 Build a standalone application for SDRAM. Proper start files and
14333 link scripts are used to put the application into SDRAM, and the macro
14334 @code{__BFIN_SDRAM} is defined.
14335 The loader should initialize SDRAM before loading the application.
14339 Assume that ICPLBs are enabled at run time. This has an effect on certain
14340 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14341 are enabled; for standalone applications the default is off.
14345 @subsection C6X Options
14346 @cindex C6X Options
14349 @item -march=@var{name}
14351 This specifies the name of the target architecture. GCC uses this
14352 name to determine what kind of instructions it can emit when generating
14353 assembly code. Permissible names are: @samp{c62x},
14354 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14357 @opindex mbig-endian
14358 Generate code for a big-endian target.
14360 @item -mlittle-endian
14361 @opindex mlittle-endian
14362 Generate code for a little-endian target. This is the default.
14366 Choose startup files and linker script suitable for the simulator.
14368 @item -msdata=default
14369 @opindex msdata=default
14370 Put small global and static data in the @code{.neardata} section,
14371 which is pointed to by register @code{B14}. Put small uninitialized
14372 global and static data in the @code{.bss} section, which is adjacent
14373 to the @code{.neardata} section. Put small read-only data into the
14374 @code{.rodata} section. The corresponding sections used for large
14375 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14378 @opindex msdata=all
14379 Put all data, not just small objects, into the sections reserved for
14380 small data, and use addressing relative to the @code{B14} register to
14384 @opindex msdata=none
14385 Make no use of the sections reserved for small data, and use absolute
14386 addresses to access all data. Put all initialized global and static
14387 data in the @code{.fardata} section, and all uninitialized data in the
14388 @code{.far} section. Put all constant data into the @code{.const}
14393 @subsection CRIS Options
14394 @cindex CRIS Options
14396 These options are defined specifically for the CRIS ports.
14399 @item -march=@var{architecture-type}
14400 @itemx -mcpu=@var{architecture-type}
14403 Generate code for the specified architecture. The choices for
14404 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14405 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14406 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14409 @item -mtune=@var{architecture-type}
14411 Tune to @var{architecture-type} everything applicable about the generated
14412 code, except for the ABI and the set of available instructions. The
14413 choices for @var{architecture-type} are the same as for
14414 @option{-march=@var{architecture-type}}.
14416 @item -mmax-stack-frame=@var{n}
14417 @opindex mmax-stack-frame
14418 Warn when the stack frame of a function exceeds @var{n} bytes.
14424 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14425 @option{-march=v3} and @option{-march=v8} respectively.
14427 @item -mmul-bug-workaround
14428 @itemx -mno-mul-bug-workaround
14429 @opindex mmul-bug-workaround
14430 @opindex mno-mul-bug-workaround
14431 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14432 models where it applies. This option is active by default.
14436 Enable CRIS-specific verbose debug-related information in the assembly
14437 code. This option also has the effect of turning off the @samp{#NO_APP}
14438 formatted-code indicator to the assembler at the beginning of the
14443 Do not use condition-code results from previous instruction; always emit
14444 compare and test instructions before use of condition codes.
14446 @item -mno-side-effects
14447 @opindex mno-side-effects
14448 Do not emit instructions with side effects in addressing modes other than
14451 @item -mstack-align
14452 @itemx -mno-stack-align
14453 @itemx -mdata-align
14454 @itemx -mno-data-align
14455 @itemx -mconst-align
14456 @itemx -mno-const-align
14457 @opindex mstack-align
14458 @opindex mno-stack-align
14459 @opindex mdata-align
14460 @opindex mno-data-align
14461 @opindex mconst-align
14462 @opindex mno-const-align
14463 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14464 stack frame, individual data and constants to be aligned for the maximum
14465 single data access size for the chosen CPU model. The default is to
14466 arrange for 32-bit alignment. ABI details such as structure layout are
14467 not affected by these options.
14475 Similar to the stack- data- and const-align options above, these options
14476 arrange for stack frame, writable data and constants to all be 32-bit,
14477 16-bit or 8-bit aligned. The default is 32-bit alignment.
14479 @item -mno-prologue-epilogue
14480 @itemx -mprologue-epilogue
14481 @opindex mno-prologue-epilogue
14482 @opindex mprologue-epilogue
14483 With @option{-mno-prologue-epilogue}, the normal function prologue and
14484 epilogue which set up the stack frame are omitted and no return
14485 instructions or return sequences are generated in the code. Use this
14486 option only together with visual inspection of the compiled code: no
14487 warnings or errors are generated when call-saved registers must be saved,
14488 or storage for local variables needs to be allocated.
14492 @opindex mno-gotplt
14494 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14495 instruction sequences that load addresses for functions from the PLT part
14496 of the GOT rather than (traditional on other architectures) calls to the
14497 PLT@. The default is @option{-mgotplt}.
14501 Legacy no-op option only recognized with the cris-axis-elf and
14502 cris-axis-linux-gnu targets.
14506 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14510 This option, recognized for the cris-axis-elf, arranges
14511 to link with input-output functions from a simulator library. Code,
14512 initialized data and zero-initialized data are allocated consecutively.
14516 Like @option{-sim}, but pass linker options to locate initialized data at
14517 0x40000000 and zero-initialized data at 0x80000000.
14521 @subsection CR16 Options
14522 @cindex CR16 Options
14524 These options are defined specifically for the CR16 ports.
14530 Enable the use of multiply-accumulate instructions. Disabled by default.
14534 @opindex mcr16cplus
14536 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14541 Links the library libsim.a which is in compatible with simulator. Applicable
14542 to ELF compiler only.
14546 Choose integer type as 32-bit wide.
14550 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14552 @item -mdata-model=@var{model}
14553 @opindex mdata-model
14554 Choose a data model. The choices for @var{model} are @samp{near},
14555 @samp{far} or @samp{medium}. @samp{medium} is default.
14556 However, @samp{far} is not valid with @option{-mcr16c}, as the
14557 CR16C architecture does not support the far data model.
14560 @node Darwin Options
14561 @subsection Darwin Options
14562 @cindex Darwin options
14564 These options are defined for all architectures running the Darwin operating
14567 FSF GCC on Darwin does not create ``fat'' object files; it creates
14568 an object file for the single architecture that GCC was built to
14569 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14570 @option{-arch} options are used; it does so by running the compiler or
14571 linker multiple times and joining the results together with
14574 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14575 @samp{i686}) is determined by the flags that specify the ISA
14576 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14577 @option{-force_cpusubtype_ALL} option can be used to override this.
14579 The Darwin tools vary in their behavior when presented with an ISA
14580 mismatch. The assembler, @file{as}, only permits instructions to
14581 be used that are valid for the subtype of the file it is generating,
14582 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14583 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14584 and prints an error if asked to create a shared library with a less
14585 restrictive subtype than its input files (for instance, trying to put
14586 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14587 for executables, @command{ld}, quietly gives the executable the most
14588 restrictive subtype of any of its input files.
14593 Add the framework directory @var{dir} to the head of the list of
14594 directories to be searched for header files. These directories are
14595 interleaved with those specified by @option{-I} options and are
14596 scanned in a left-to-right order.
14598 A framework directory is a directory with frameworks in it. A
14599 framework is a directory with a @file{Headers} and/or
14600 @file{PrivateHeaders} directory contained directly in it that ends
14601 in @file{.framework}. The name of a framework is the name of this
14602 directory excluding the @file{.framework}. Headers associated with
14603 the framework are found in one of those two directories, with
14604 @file{Headers} being searched first. A subframework is a framework
14605 directory that is in a framework's @file{Frameworks} directory.
14606 Includes of subframework headers can only appear in a header of a
14607 framework that contains the subframework, or in a sibling subframework
14608 header. Two subframeworks are siblings if they occur in the same
14609 framework. A subframework should not have the same name as a
14610 framework; a warning is issued if this is violated. Currently a
14611 subframework cannot have subframeworks; in the future, the mechanism
14612 may be extended to support this. The standard frameworks can be found
14613 in @file{/System/Library/Frameworks} and
14614 @file{/Library/Frameworks}. An example include looks like
14615 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14616 the name of the framework and @file{header.h} is found in the
14617 @file{PrivateHeaders} or @file{Headers} directory.
14619 @item -iframework@var{dir}
14620 @opindex iframework
14621 Like @option{-F} except the directory is a treated as a system
14622 directory. The main difference between this @option{-iframework} and
14623 @option{-F} is that with @option{-iframework} the compiler does not
14624 warn about constructs contained within header files found via
14625 @var{dir}. This option is valid only for the C family of languages.
14629 Emit debugging information for symbols that are used. For stabs
14630 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14631 This is by default ON@.
14635 Emit debugging information for all symbols and types.
14637 @item -mmacosx-version-min=@var{version}
14638 The earliest version of MacOS X that this executable will run on
14639 is @var{version}. Typical values of @var{version} include @code{10.1},
14640 @code{10.2}, and @code{10.3.9}.
14642 If the compiler was built to use the system's headers by default,
14643 then the default for this option is the system version on which the
14644 compiler is running, otherwise the default is to make choices that
14645 are compatible with as many systems and code bases as possible.
14649 Enable kernel development mode. The @option{-mkernel} option sets
14650 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14651 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14652 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14653 applicable. This mode also sets @option{-mno-altivec},
14654 @option{-msoft-float}, @option{-fno-builtin} and
14655 @option{-mlong-branch} for PowerPC targets.
14657 @item -mone-byte-bool
14658 @opindex mone-byte-bool
14659 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14660 By default @code{sizeof(bool)} is @code{4} when compiling for
14661 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14662 option has no effect on x86.
14664 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14665 to generate code that is not binary compatible with code generated
14666 without that switch. Using this switch may require recompiling all
14667 other modules in a program, including system libraries. Use this
14668 switch to conform to a non-default data model.
14670 @item -mfix-and-continue
14671 @itemx -ffix-and-continue
14672 @itemx -findirect-data
14673 @opindex mfix-and-continue
14674 @opindex ffix-and-continue
14675 @opindex findirect-data
14676 Generate code suitable for fast turnaround development, such as to
14677 allow GDB to dynamically load @file{.o} files into already-running
14678 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14679 are provided for backwards compatibility.
14683 Loads all members of static archive libraries.
14684 See man ld(1) for more information.
14686 @item -arch_errors_fatal
14687 @opindex arch_errors_fatal
14688 Cause the errors having to do with files that have the wrong architecture
14691 @item -bind_at_load
14692 @opindex bind_at_load
14693 Causes the output file to be marked such that the dynamic linker will
14694 bind all undefined references when the file is loaded or launched.
14698 Produce a Mach-o bundle format file.
14699 See man ld(1) for more information.
14701 @item -bundle_loader @var{executable}
14702 @opindex bundle_loader
14703 This option specifies the @var{executable} that will load the build
14704 output file being linked. See man ld(1) for more information.
14707 @opindex dynamiclib
14708 When passed this option, GCC produces a dynamic library instead of
14709 an executable when linking, using the Darwin @file{libtool} command.
14711 @item -force_cpusubtype_ALL
14712 @opindex force_cpusubtype_ALL
14713 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14714 one controlled by the @option{-mcpu} or @option{-march} option.
14716 @item -allowable_client @var{client_name}
14717 @itemx -client_name
14718 @itemx -compatibility_version
14719 @itemx -current_version
14721 @itemx -dependency-file
14723 @itemx -dylinker_install_name
14725 @itemx -exported_symbols_list
14728 @itemx -flat_namespace
14729 @itemx -force_flat_namespace
14730 @itemx -headerpad_max_install_names
14733 @itemx -install_name
14734 @itemx -keep_private_externs
14735 @itemx -multi_module
14736 @itemx -multiply_defined
14737 @itemx -multiply_defined_unused
14740 @itemx -no_dead_strip_inits_and_terms
14741 @itemx -nofixprebinding
14742 @itemx -nomultidefs
14744 @itemx -noseglinkedit
14745 @itemx -pagezero_size
14747 @itemx -prebind_all_twolevel_modules
14748 @itemx -private_bundle
14750 @itemx -read_only_relocs
14752 @itemx -sectobjectsymbols
14756 @itemx -sectobjectsymbols
14759 @itemx -segs_read_only_addr
14761 @itemx -segs_read_write_addr
14762 @itemx -seg_addr_table
14763 @itemx -seg_addr_table_filename
14764 @itemx -seglinkedit
14766 @itemx -segs_read_only_addr
14767 @itemx -segs_read_write_addr
14768 @itemx -single_module
14770 @itemx -sub_library
14772 @itemx -sub_umbrella
14773 @itemx -twolevel_namespace
14776 @itemx -unexported_symbols_list
14777 @itemx -weak_reference_mismatches
14778 @itemx -whatsloaded
14779 @opindex allowable_client
14780 @opindex client_name
14781 @opindex compatibility_version
14782 @opindex current_version
14783 @opindex dead_strip
14784 @opindex dependency-file
14785 @opindex dylib_file
14786 @opindex dylinker_install_name
14788 @opindex exported_symbols_list
14790 @opindex flat_namespace
14791 @opindex force_flat_namespace
14792 @opindex headerpad_max_install_names
14793 @opindex image_base
14795 @opindex install_name
14796 @opindex keep_private_externs
14797 @opindex multi_module
14798 @opindex multiply_defined
14799 @opindex multiply_defined_unused
14800 @opindex noall_load
14801 @opindex no_dead_strip_inits_and_terms
14802 @opindex nofixprebinding
14803 @opindex nomultidefs
14805 @opindex noseglinkedit
14806 @opindex pagezero_size
14808 @opindex prebind_all_twolevel_modules
14809 @opindex private_bundle
14810 @opindex read_only_relocs
14812 @opindex sectobjectsymbols
14815 @opindex sectcreate
14816 @opindex sectobjectsymbols
14819 @opindex segs_read_only_addr
14820 @opindex segs_read_write_addr
14821 @opindex seg_addr_table
14822 @opindex seg_addr_table_filename
14823 @opindex seglinkedit
14825 @opindex segs_read_only_addr
14826 @opindex segs_read_write_addr
14827 @opindex single_module
14829 @opindex sub_library
14830 @opindex sub_umbrella
14831 @opindex twolevel_namespace
14834 @opindex unexported_symbols_list
14835 @opindex weak_reference_mismatches
14836 @opindex whatsloaded
14837 These options are passed to the Darwin linker. The Darwin linker man page
14838 describes them in detail.
14841 @node DEC Alpha Options
14842 @subsection DEC Alpha Options
14844 These @samp{-m} options are defined for the DEC Alpha implementations:
14847 @item -mno-soft-float
14848 @itemx -msoft-float
14849 @opindex mno-soft-float
14850 @opindex msoft-float
14851 Use (do not use) the hardware floating-point instructions for
14852 floating-point operations. When @option{-msoft-float} is specified,
14853 functions in @file{libgcc.a} are used to perform floating-point
14854 operations. Unless they are replaced by routines that emulate the
14855 floating-point operations, or compiled in such a way as to call such
14856 emulations routines, these routines issue floating-point
14857 operations. If you are compiling for an Alpha without floating-point
14858 operations, you must ensure that the library is built so as not to call
14861 Note that Alpha implementations without floating-point operations are
14862 required to have floating-point registers.
14865 @itemx -mno-fp-regs
14867 @opindex mno-fp-regs
14868 Generate code that uses (does not use) the floating-point register set.
14869 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
14870 register set is not used, floating-point operands are passed in integer
14871 registers as if they were integers and floating-point results are passed
14872 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
14873 so any function with a floating-point argument or return value called by code
14874 compiled with @option{-mno-fp-regs} must also be compiled with that
14877 A typical use of this option is building a kernel that does not use,
14878 and hence need not save and restore, any floating-point registers.
14882 The Alpha architecture implements floating-point hardware optimized for
14883 maximum performance. It is mostly compliant with the IEEE floating-point
14884 standard. However, for full compliance, software assistance is
14885 required. This option generates code fully IEEE-compliant code
14886 @emph{except} that the @var{inexact-flag} is not maintained (see below).
14887 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
14888 defined during compilation. The resulting code is less efficient but is
14889 able to correctly support denormalized numbers and exceptional IEEE
14890 values such as not-a-number and plus/minus infinity. Other Alpha
14891 compilers call this option @option{-ieee_with_no_inexact}.
14893 @item -mieee-with-inexact
14894 @opindex mieee-with-inexact
14895 This is like @option{-mieee} except the generated code also maintains
14896 the IEEE @var{inexact-flag}. Turning on this option causes the
14897 generated code to implement fully-compliant IEEE math. In addition to
14898 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
14899 macro. On some Alpha implementations the resulting code may execute
14900 significantly slower than the code generated by default. Since there is
14901 very little code that depends on the @var{inexact-flag}, you should
14902 normally not specify this option. Other Alpha compilers call this
14903 option @option{-ieee_with_inexact}.
14905 @item -mfp-trap-mode=@var{trap-mode}
14906 @opindex mfp-trap-mode
14907 This option controls what floating-point related traps are enabled.
14908 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
14909 The trap mode can be set to one of four values:
14913 This is the default (normal) setting. The only traps that are enabled
14914 are the ones that cannot be disabled in software (e.g., division by zero
14918 In addition to the traps enabled by @samp{n}, underflow traps are enabled
14922 Like @samp{u}, but the instructions are marked to be safe for software
14923 completion (see Alpha architecture manual for details).
14926 Like @samp{su}, but inexact traps are enabled as well.
14929 @item -mfp-rounding-mode=@var{rounding-mode}
14930 @opindex mfp-rounding-mode
14931 Selects the IEEE rounding mode. Other Alpha compilers call this option
14932 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
14937 Normal IEEE rounding mode. Floating-point numbers are rounded towards
14938 the nearest machine number or towards the even machine number in case
14942 Round towards minus infinity.
14945 Chopped rounding mode. Floating-point numbers are rounded towards zero.
14948 Dynamic rounding mode. A field in the floating-point control register
14949 (@var{fpcr}, see Alpha architecture reference manual) controls the
14950 rounding mode in effect. The C library initializes this register for
14951 rounding towards plus infinity. Thus, unless your program modifies the
14952 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
14955 @item -mtrap-precision=@var{trap-precision}
14956 @opindex mtrap-precision
14957 In the Alpha architecture, floating-point traps are imprecise. This
14958 means without software assistance it is impossible to recover from a
14959 floating trap and program execution normally needs to be terminated.
14960 GCC can generate code that can assist operating system trap handlers
14961 in determining the exact location that caused a floating-point trap.
14962 Depending on the requirements of an application, different levels of
14963 precisions can be selected:
14967 Program precision. This option is the default and means a trap handler
14968 can only identify which program caused a floating-point exception.
14971 Function precision. The trap handler can determine the function that
14972 caused a floating-point exception.
14975 Instruction precision. The trap handler can determine the exact
14976 instruction that caused a floating-point exception.
14979 Other Alpha compilers provide the equivalent options called
14980 @option{-scope_safe} and @option{-resumption_safe}.
14982 @item -mieee-conformant
14983 @opindex mieee-conformant
14984 This option marks the generated code as IEEE conformant. You must not
14985 use this option unless you also specify @option{-mtrap-precision=i} and either
14986 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
14987 is to emit the line @samp{.eflag 48} in the function prologue of the
14988 generated assembly file.
14990 @item -mbuild-constants
14991 @opindex mbuild-constants
14992 Normally GCC examines a 32- or 64-bit integer constant to
14993 see if it can construct it from smaller constants in two or three
14994 instructions. If it cannot, it outputs the constant as a literal and
14995 generates code to load it from the data segment at run time.
14997 Use this option to require GCC to construct @emph{all} integer constants
14998 using code, even if it takes more instructions (the maximum is six).
15000 You typically use this option to build a shared library dynamic
15001 loader. Itself a shared library, it must relocate itself in memory
15002 before it can find the variables and constants in its own data segment.
15020 Indicate whether GCC should generate code to use the optional BWX,
15021 CIX, FIX and MAX instruction sets. The default is to use the instruction
15022 sets supported by the CPU type specified via @option{-mcpu=} option or that
15023 of the CPU on which GCC was built if none is specified.
15026 @itemx -mfloat-ieee
15027 @opindex mfloat-vax
15028 @opindex mfloat-ieee
15029 Generate code that uses (does not use) VAX F and G floating-point
15030 arithmetic instead of IEEE single and double precision.
15032 @item -mexplicit-relocs
15033 @itemx -mno-explicit-relocs
15034 @opindex mexplicit-relocs
15035 @opindex mno-explicit-relocs
15036 Older Alpha assemblers provided no way to generate symbol relocations
15037 except via assembler macros. Use of these macros does not allow
15038 optimal instruction scheduling. GNU binutils as of version 2.12
15039 supports a new syntax that allows the compiler to explicitly mark
15040 which relocations should apply to which instructions. This option
15041 is mostly useful for debugging, as GCC detects the capabilities of
15042 the assembler when it is built and sets the default accordingly.
15045 @itemx -mlarge-data
15046 @opindex msmall-data
15047 @opindex mlarge-data
15048 When @option{-mexplicit-relocs} is in effect, static data is
15049 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15050 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15051 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15052 16-bit relocations off of the @code{$gp} register. This limits the
15053 size of the small data area to 64KB, but allows the variables to be
15054 directly accessed via a single instruction.
15056 The default is @option{-mlarge-data}. With this option the data area
15057 is limited to just below 2GB@. Programs that require more than 2GB of
15058 data must use @code{malloc} or @code{mmap} to allocate the data in the
15059 heap instead of in the program's data segment.
15061 When generating code for shared libraries, @option{-fpic} implies
15062 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15065 @itemx -mlarge-text
15066 @opindex msmall-text
15067 @opindex mlarge-text
15068 When @option{-msmall-text} is used, the compiler assumes that the
15069 code of the entire program (or shared library) fits in 4MB, and is
15070 thus reachable with a branch instruction. When @option{-msmall-data}
15071 is used, the compiler can assume that all local symbols share the
15072 same @code{$gp} value, and thus reduce the number of instructions
15073 required for a function call from 4 to 1.
15075 The default is @option{-mlarge-text}.
15077 @item -mcpu=@var{cpu_type}
15079 Set the instruction set and instruction scheduling parameters for
15080 machine type @var{cpu_type}. You can specify either the @samp{EV}
15081 style name or the corresponding chip number. GCC supports scheduling
15082 parameters for the EV4, EV5 and EV6 family of processors and
15083 chooses the default values for the instruction set from the processor
15084 you specify. If you do not specify a processor type, GCC defaults
15085 to the processor on which the compiler was built.
15087 Supported values for @var{cpu_type} are
15093 Schedules as an EV4 and has no instruction set extensions.
15097 Schedules as an EV5 and has no instruction set extensions.
15101 Schedules as an EV5 and supports the BWX extension.
15106 Schedules as an EV5 and supports the BWX and MAX extensions.
15110 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15114 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15117 Native toolchains also support the value @samp{native},
15118 which selects the best architecture option for the host processor.
15119 @option{-mcpu=native} has no effect if GCC does not recognize
15122 @item -mtune=@var{cpu_type}
15124 Set only the instruction scheduling parameters for machine type
15125 @var{cpu_type}. The instruction set is not changed.
15127 Native toolchains also support the value @samp{native},
15128 which selects the best architecture option for the host processor.
15129 @option{-mtune=native} has no effect if GCC does not recognize
15132 @item -mmemory-latency=@var{time}
15133 @opindex mmemory-latency
15134 Sets the latency the scheduler should assume for typical memory
15135 references as seen by the application. This number is highly
15136 dependent on the memory access patterns used by the application
15137 and the size of the external cache on the machine.
15139 Valid options for @var{time} are
15143 A decimal number representing clock cycles.
15149 The compiler contains estimates of the number of clock cycles for
15150 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15151 (also called Dcache, Scache, and Bcache), as well as to main memory.
15152 Note that L3 is only valid for EV5.
15158 @subsection FR30 Options
15159 @cindex FR30 Options
15161 These options are defined specifically for the FR30 port.
15165 @item -msmall-model
15166 @opindex msmall-model
15167 Use the small address space model. This can produce smaller code, but
15168 it does assume that all symbolic values and addresses fit into a
15173 Assume that runtime support has been provided and so there is no need
15174 to include the simulator library (@file{libsim.a}) on the linker
15180 @subsection FT32 Options
15181 @cindex FT32 Options
15183 These options are defined specifically for the FT32 port.
15189 Specifies that the program will be run on the simulator. This causes
15190 an alternate runtime startup and library to be linked.
15191 You must not use this option when generating programs that will run on
15192 real hardware; you must provide your own runtime library for whatever
15193 I/O functions are needed.
15197 Enable Local Register Allocation. This is still experimental for FT32,
15198 so by default the compiler uses standard reload.
15203 @subsection FRV Options
15204 @cindex FRV Options
15210 Only use the first 32 general-purpose registers.
15215 Use all 64 general-purpose registers.
15220 Use only the first 32 floating-point registers.
15225 Use all 64 floating-point registers.
15228 @opindex mhard-float
15230 Use hardware instructions for floating-point operations.
15233 @opindex msoft-float
15235 Use library routines for floating-point operations.
15240 Dynamically allocate condition code registers.
15245 Do not try to dynamically allocate condition code registers, only
15246 use @code{icc0} and @code{fcc0}.
15251 Change ABI to use double word insns.
15256 Do not use double word instructions.
15261 Use floating-point double instructions.
15264 @opindex mno-double
15266 Do not use floating-point double instructions.
15271 Use media instructions.
15276 Do not use media instructions.
15281 Use multiply and add/subtract instructions.
15284 @opindex mno-muladd
15286 Do not use multiply and add/subtract instructions.
15291 Select the FDPIC ABI, which uses function descriptors to represent
15292 pointers to functions. Without any PIC/PIE-related options, it
15293 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15294 assumes GOT entries and small data are within a 12-bit range from the
15295 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15296 are computed with 32 bits.
15297 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15300 @opindex minline-plt
15302 Enable inlining of PLT entries in function calls to functions that are
15303 not known to bind locally. It has no effect without @option{-mfdpic}.
15304 It's enabled by default if optimizing for speed and compiling for
15305 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15306 optimization option such as @option{-O3} or above is present in the
15312 Assume a large TLS segment when generating thread-local code.
15317 Do not assume a large TLS segment when generating thread-local code.
15322 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15323 that is known to be in read-only sections. It's enabled by default,
15324 except for @option{-fpic} or @option{-fpie}: even though it may help
15325 make the global offset table smaller, it trades 1 instruction for 4.
15326 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15327 one of which may be shared by multiple symbols, and it avoids the need
15328 for a GOT entry for the referenced symbol, so it's more likely to be a
15329 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15331 @item -multilib-library-pic
15332 @opindex multilib-library-pic
15334 Link with the (library, not FD) pic libraries. It's implied by
15335 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15336 @option{-fpic} without @option{-mfdpic}. You should never have to use
15340 @opindex mlinked-fp
15342 Follow the EABI requirement of always creating a frame pointer whenever
15343 a stack frame is allocated. This option is enabled by default and can
15344 be disabled with @option{-mno-linked-fp}.
15347 @opindex mlong-calls
15349 Use indirect addressing to call functions outside the current
15350 compilation unit. This allows the functions to be placed anywhere
15351 within the 32-bit address space.
15353 @item -malign-labels
15354 @opindex malign-labels
15356 Try to align labels to an 8-byte boundary by inserting NOPs into the
15357 previous packet. This option only has an effect when VLIW packing
15358 is enabled. It doesn't create new packets; it merely adds NOPs to
15361 @item -mlibrary-pic
15362 @opindex mlibrary-pic
15364 Generate position-independent EABI code.
15369 Use only the first four media accumulator registers.
15374 Use all eight media accumulator registers.
15379 Pack VLIW instructions.
15384 Do not pack VLIW instructions.
15387 @opindex mno-eflags
15389 Do not mark ABI switches in e_flags.
15392 @opindex mcond-move
15394 Enable the use of conditional-move instructions (default).
15396 This switch is mainly for debugging the compiler and will likely be removed
15397 in a future version.
15399 @item -mno-cond-move
15400 @opindex mno-cond-move
15402 Disable the use of conditional-move instructions.
15404 This switch is mainly for debugging the compiler and will likely be removed
15405 in a future version.
15410 Enable the use of conditional set instructions (default).
15412 This switch is mainly for debugging the compiler and will likely be removed
15413 in a future version.
15418 Disable the use of conditional set instructions.
15420 This switch is mainly for debugging the compiler and will likely be removed
15421 in a future version.
15424 @opindex mcond-exec
15426 Enable the use of conditional execution (default).
15428 This switch is mainly for debugging the compiler and will likely be removed
15429 in a future version.
15431 @item -mno-cond-exec
15432 @opindex mno-cond-exec
15434 Disable the use of conditional execution.
15436 This switch is mainly for debugging the compiler and will likely be removed
15437 in a future version.
15439 @item -mvliw-branch
15440 @opindex mvliw-branch
15442 Run a pass to pack branches into VLIW instructions (default).
15444 This switch is mainly for debugging the compiler and will likely be removed
15445 in a future version.
15447 @item -mno-vliw-branch
15448 @opindex mno-vliw-branch
15450 Do not run a pass to pack branches into VLIW instructions.
15452 This switch is mainly for debugging the compiler and will likely be removed
15453 in a future version.
15455 @item -mmulti-cond-exec
15456 @opindex mmulti-cond-exec
15458 Enable optimization of @code{&&} and @code{||} in conditional execution
15461 This switch is mainly for debugging the compiler and will likely be removed
15462 in a future version.
15464 @item -mno-multi-cond-exec
15465 @opindex mno-multi-cond-exec
15467 Disable optimization of @code{&&} and @code{||} in conditional execution.
15469 This switch is mainly for debugging the compiler and will likely be removed
15470 in a future version.
15472 @item -mnested-cond-exec
15473 @opindex mnested-cond-exec
15475 Enable nested conditional execution optimizations (default).
15477 This switch is mainly for debugging the compiler and will likely be removed
15478 in a future version.
15480 @item -mno-nested-cond-exec
15481 @opindex mno-nested-cond-exec
15483 Disable nested conditional execution optimizations.
15485 This switch is mainly for debugging the compiler and will likely be removed
15486 in a future version.
15488 @item -moptimize-membar
15489 @opindex moptimize-membar
15491 This switch removes redundant @code{membar} instructions from the
15492 compiler-generated code. It is enabled by default.
15494 @item -mno-optimize-membar
15495 @opindex mno-optimize-membar
15497 This switch disables the automatic removal of redundant @code{membar}
15498 instructions from the generated code.
15500 @item -mtomcat-stats
15501 @opindex mtomcat-stats
15503 Cause gas to print out tomcat statistics.
15505 @item -mcpu=@var{cpu}
15508 Select the processor type for which to generate code. Possible values are
15509 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15510 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15514 @node GNU/Linux Options
15515 @subsection GNU/Linux Options
15517 These @samp{-m} options are defined for GNU/Linux targets:
15522 Use the GNU C library. This is the default except
15523 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
15524 @samp{*-*-linux-*android*} targets.
15528 Use uClibc C library. This is the default on
15529 @samp{*-*-linux-*uclibc*} targets.
15533 Use the musl C library. This is the default on
15534 @samp{*-*-linux-*musl*} targets.
15538 Use Bionic C library. This is the default on
15539 @samp{*-*-linux-*android*} targets.
15543 Compile code compatible with Android platform. This is the default on
15544 @samp{*-*-linux-*android*} targets.
15546 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15547 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15548 this option makes the GCC driver pass Android-specific options to the linker.
15549 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15552 @item -tno-android-cc
15553 @opindex tno-android-cc
15554 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15555 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15556 @option{-fno-rtti} by default.
15558 @item -tno-android-ld
15559 @opindex tno-android-ld
15560 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15561 linking options to the linker.
15565 @node H8/300 Options
15566 @subsection H8/300 Options
15568 These @samp{-m} options are defined for the H8/300 implementations:
15573 Shorten some address references at link time, when possible; uses the
15574 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15575 ld, Using ld}, for a fuller description.
15579 Generate code for the H8/300H@.
15583 Generate code for the H8S@.
15587 Generate code for the H8S and H8/300H in the normal mode. This switch
15588 must be used either with @option{-mh} or @option{-ms}.
15592 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15596 Extended registers are stored on stack before execution of function
15597 with monitor attribute. Default option is @option{-mexr}.
15598 This option is valid only for H8S targets.
15602 Extended registers are not stored on stack before execution of function
15603 with monitor attribute. Default option is @option{-mno-exr}.
15604 This option is valid only for H8S targets.
15608 Make @code{int} data 32 bits by default.
15611 @opindex malign-300
15612 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15613 The default for the H8/300H and H8S is to align longs and floats on
15615 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15616 This option has no effect on the H8/300.
15620 @subsection HPPA Options
15621 @cindex HPPA Options
15623 These @samp{-m} options are defined for the HPPA family of computers:
15626 @item -march=@var{architecture-type}
15628 Generate code for the specified architecture. The choices for
15629 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15630 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15631 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15632 architecture option for your machine. Code compiled for lower numbered
15633 architectures runs on higher numbered architectures, but not the
15636 @item -mpa-risc-1-0
15637 @itemx -mpa-risc-1-1
15638 @itemx -mpa-risc-2-0
15639 @opindex mpa-risc-1-0
15640 @opindex mpa-risc-1-1
15641 @opindex mpa-risc-2-0
15642 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15644 @item -mjump-in-delay
15645 @opindex mjump-in-delay
15646 This option is ignored and provided for compatibility purposes only.
15648 @item -mdisable-fpregs
15649 @opindex mdisable-fpregs
15650 Prevent floating-point registers from being used in any manner. This is
15651 necessary for compiling kernels that perform lazy context switching of
15652 floating-point registers. If you use this option and attempt to perform
15653 floating-point operations, the compiler aborts.
15655 @item -mdisable-indexing
15656 @opindex mdisable-indexing
15657 Prevent the compiler from using indexing address modes. This avoids some
15658 rather obscure problems when compiling MIG generated code under MACH@.
15660 @item -mno-space-regs
15661 @opindex mno-space-regs
15662 Generate code that assumes the target has no space registers. This allows
15663 GCC to generate faster indirect calls and use unscaled index address modes.
15665 Such code is suitable for level 0 PA systems and kernels.
15667 @item -mfast-indirect-calls
15668 @opindex mfast-indirect-calls
15669 Generate code that assumes calls never cross space boundaries. This
15670 allows GCC to emit code that performs faster indirect calls.
15672 This option does not work in the presence of shared libraries or nested
15675 @item -mfixed-range=@var{register-range}
15676 @opindex mfixed-range
15677 Generate code treating the given register range as fixed registers.
15678 A fixed register is one that the register allocator cannot use. This is
15679 useful when compiling kernel code. A register range is specified as
15680 two registers separated by a dash. Multiple register ranges can be
15681 specified separated by a comma.
15683 @item -mlong-load-store
15684 @opindex mlong-load-store
15685 Generate 3-instruction load and store sequences as sometimes required by
15686 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15689 @item -mportable-runtime
15690 @opindex mportable-runtime
15691 Use the portable calling conventions proposed by HP for ELF systems.
15695 Enable the use of assembler directives only GAS understands.
15697 @item -mschedule=@var{cpu-type}
15699 Schedule code according to the constraints for the machine type
15700 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15701 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15702 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15703 proper scheduling option for your machine. The default scheduling is
15707 @opindex mlinker-opt
15708 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15709 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15710 linkers in which they give bogus error messages when linking some programs.
15713 @opindex msoft-float
15714 Generate output containing library calls for floating point.
15715 @strong{Warning:} the requisite libraries are not available for all HPPA
15716 targets. Normally the facilities of the machine's usual C compiler are
15717 used, but this cannot be done directly in cross-compilation. You must make
15718 your own arrangements to provide suitable library functions for
15721 @option{-msoft-float} changes the calling convention in the output file;
15722 therefore, it is only useful if you compile @emph{all} of a program with
15723 this option. In particular, you need to compile @file{libgcc.a}, the
15724 library that comes with GCC, with @option{-msoft-float} in order for
15729 Generate the predefine, @code{_SIO}, for server IO@. The default is
15730 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15731 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15732 options are available under HP-UX and HI-UX@.
15736 Use options specific to GNU @command{ld}.
15737 This passes @option{-shared} to @command{ld} when
15738 building a shared library. It is the default when GCC is configured,
15739 explicitly or implicitly, with the GNU linker. This option does not
15740 affect which @command{ld} is called; it only changes what parameters
15741 are passed to that @command{ld}.
15742 The @command{ld} that is called is determined by the
15743 @option{--with-ld} configure option, GCC's program search path, and
15744 finally by the user's @env{PATH}. The linker used by GCC can be printed
15745 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15746 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15750 Use options specific to HP @command{ld}.
15751 This passes @option{-b} to @command{ld} when building
15752 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15753 links. It is the default when GCC is configured, explicitly or
15754 implicitly, with the HP linker. This option does not affect
15755 which @command{ld} is called; it only changes what parameters are passed to that
15757 The @command{ld} that is called is determined by the @option{--with-ld}
15758 configure option, GCC's program search path, and finally by the user's
15759 @env{PATH}. The linker used by GCC can be printed using @samp{which
15760 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15761 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15764 @opindex mno-long-calls
15765 Generate code that uses long call sequences. This ensures that a call
15766 is always able to reach linker generated stubs. The default is to generate
15767 long calls only when the distance from the call site to the beginning
15768 of the function or translation unit, as the case may be, exceeds a
15769 predefined limit set by the branch type being used. The limits for
15770 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15771 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15774 Distances are measured from the beginning of functions when using the
15775 @option{-ffunction-sections} option, or when using the @option{-mgas}
15776 and @option{-mno-portable-runtime} options together under HP-UX with
15779 It is normally not desirable to use this option as it degrades
15780 performance. However, it may be useful in large applications,
15781 particularly when partial linking is used to build the application.
15783 The types of long calls used depends on the capabilities of the
15784 assembler and linker, and the type of code being generated. The
15785 impact on systems that support long absolute calls, and long pic
15786 symbol-difference or pc-relative calls should be relatively small.
15787 However, an indirect call is used on 32-bit ELF systems in pic code
15788 and it is quite long.
15790 @item -munix=@var{unix-std}
15792 Generate compiler predefines and select a startfile for the specified
15793 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
15794 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
15795 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
15796 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
15797 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
15800 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
15801 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
15802 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
15803 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
15804 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
15805 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
15807 It is @emph{important} to note that this option changes the interfaces
15808 for various library routines. It also affects the operational behavior
15809 of the C library. Thus, @emph{extreme} care is needed in using this
15812 Library code that is intended to operate with more than one UNIX
15813 standard must test, set and restore the variable @code{__xpg4_extended_mask}
15814 as appropriate. Most GNU software doesn't provide this capability.
15818 Suppress the generation of link options to search libdld.sl when the
15819 @option{-static} option is specified on HP-UX 10 and later.
15823 The HP-UX implementation of setlocale in libc has a dependency on
15824 libdld.sl. There isn't an archive version of libdld.sl. Thus,
15825 when the @option{-static} option is specified, special link options
15826 are needed to resolve this dependency.
15828 On HP-UX 10 and later, the GCC driver adds the necessary options to
15829 link with libdld.sl when the @option{-static} option is specified.
15830 This causes the resulting binary to be dynamic. On the 64-bit port,
15831 the linkers generate dynamic binaries by default in any case. The
15832 @option{-nolibdld} option can be used to prevent the GCC driver from
15833 adding these link options.
15837 Add support for multithreading with the @dfn{dce thread} library
15838 under HP-UX@. This option sets flags for both the preprocessor and
15842 @node IA-64 Options
15843 @subsection IA-64 Options
15844 @cindex IA-64 Options
15846 These are the @samp{-m} options defined for the Intel IA-64 architecture.
15850 @opindex mbig-endian
15851 Generate code for a big-endian target. This is the default for HP-UX@.
15853 @item -mlittle-endian
15854 @opindex mlittle-endian
15855 Generate code for a little-endian target. This is the default for AIX5
15861 @opindex mno-gnu-as
15862 Generate (or don't) code for the GNU assembler. This is the default.
15863 @c Also, this is the default if the configure option @option{--with-gnu-as}
15869 @opindex mno-gnu-ld
15870 Generate (or don't) code for the GNU linker. This is the default.
15871 @c Also, this is the default if the configure option @option{--with-gnu-ld}
15876 Generate code that does not use a global pointer register. The result
15877 is not position independent code, and violates the IA-64 ABI@.
15879 @item -mvolatile-asm-stop
15880 @itemx -mno-volatile-asm-stop
15881 @opindex mvolatile-asm-stop
15882 @opindex mno-volatile-asm-stop
15883 Generate (or don't) a stop bit immediately before and after volatile asm
15886 @item -mregister-names
15887 @itemx -mno-register-names
15888 @opindex mregister-names
15889 @opindex mno-register-names
15890 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
15891 the stacked registers. This may make assembler output more readable.
15897 Disable (or enable) optimizations that use the small data section. This may
15898 be useful for working around optimizer bugs.
15900 @item -mconstant-gp
15901 @opindex mconstant-gp
15902 Generate code that uses a single constant global pointer value. This is
15903 useful when compiling kernel code.
15907 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
15908 This is useful when compiling firmware code.
15910 @item -minline-float-divide-min-latency
15911 @opindex minline-float-divide-min-latency
15912 Generate code for inline divides of floating-point values
15913 using the minimum latency algorithm.
15915 @item -minline-float-divide-max-throughput
15916 @opindex minline-float-divide-max-throughput
15917 Generate code for inline divides of floating-point values
15918 using the maximum throughput algorithm.
15920 @item -mno-inline-float-divide
15921 @opindex mno-inline-float-divide
15922 Do not generate inline code for divides of floating-point values.
15924 @item -minline-int-divide-min-latency
15925 @opindex minline-int-divide-min-latency
15926 Generate code for inline divides of integer values
15927 using the minimum latency algorithm.
15929 @item -minline-int-divide-max-throughput
15930 @opindex minline-int-divide-max-throughput
15931 Generate code for inline divides of integer values
15932 using the maximum throughput algorithm.
15934 @item -mno-inline-int-divide
15935 @opindex mno-inline-int-divide
15936 Do not generate inline code for divides of integer values.
15938 @item -minline-sqrt-min-latency
15939 @opindex minline-sqrt-min-latency
15940 Generate code for inline square roots
15941 using the minimum latency algorithm.
15943 @item -minline-sqrt-max-throughput
15944 @opindex minline-sqrt-max-throughput
15945 Generate code for inline square roots
15946 using the maximum throughput algorithm.
15948 @item -mno-inline-sqrt
15949 @opindex mno-inline-sqrt
15950 Do not generate inline code for @code{sqrt}.
15953 @itemx -mno-fused-madd
15954 @opindex mfused-madd
15955 @opindex mno-fused-madd
15956 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
15957 instructions. The default is to use these instructions.
15959 @item -mno-dwarf2-asm
15960 @itemx -mdwarf2-asm
15961 @opindex mno-dwarf2-asm
15962 @opindex mdwarf2-asm
15963 Don't (or do) generate assembler code for the DWARF 2 line number debugging
15964 info. This may be useful when not using the GNU assembler.
15966 @item -mearly-stop-bits
15967 @itemx -mno-early-stop-bits
15968 @opindex mearly-stop-bits
15969 @opindex mno-early-stop-bits
15970 Allow stop bits to be placed earlier than immediately preceding the
15971 instruction that triggered the stop bit. This can improve instruction
15972 scheduling, but does not always do so.
15974 @item -mfixed-range=@var{register-range}
15975 @opindex mfixed-range
15976 Generate code treating the given register range as fixed registers.
15977 A fixed register is one that the register allocator cannot use. This is
15978 useful when compiling kernel code. A register range is specified as
15979 two registers separated by a dash. Multiple register ranges can be
15980 specified separated by a comma.
15982 @item -mtls-size=@var{tls-size}
15984 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
15987 @item -mtune=@var{cpu-type}
15989 Tune the instruction scheduling for a particular CPU, Valid values are
15990 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
15991 and @samp{mckinley}.
15997 Generate code for a 32-bit or 64-bit environment.
15998 The 32-bit environment sets int, long and pointer to 32 bits.
15999 The 64-bit environment sets int to 32 bits and long and pointer
16000 to 64 bits. These are HP-UX specific flags.
16002 @item -mno-sched-br-data-spec
16003 @itemx -msched-br-data-spec
16004 @opindex mno-sched-br-data-spec
16005 @opindex msched-br-data-spec
16006 (Dis/En)able data speculative scheduling before reload.
16007 This results in generation of @code{ld.a} instructions and
16008 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16009 The default is 'disable'.
16011 @item -msched-ar-data-spec
16012 @itemx -mno-sched-ar-data-spec
16013 @opindex msched-ar-data-spec
16014 @opindex mno-sched-ar-data-spec
16015 (En/Dis)able data speculative scheduling after reload.
16016 This results in generation of @code{ld.a} instructions and
16017 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16018 The default is 'enable'.
16020 @item -mno-sched-control-spec
16021 @itemx -msched-control-spec
16022 @opindex mno-sched-control-spec
16023 @opindex msched-control-spec
16024 (Dis/En)able control speculative scheduling. This feature is
16025 available only during region scheduling (i.e.@: before reload).
16026 This results in generation of the @code{ld.s} instructions and
16027 the corresponding check instructions @code{chk.s}.
16028 The default is 'disable'.
16030 @item -msched-br-in-data-spec
16031 @itemx -mno-sched-br-in-data-spec
16032 @opindex msched-br-in-data-spec
16033 @opindex mno-sched-br-in-data-spec
16034 (En/Dis)able speculative scheduling of the instructions that
16035 are dependent on the data speculative loads before reload.
16036 This is effective only with @option{-msched-br-data-spec} enabled.
16037 The default is 'enable'.
16039 @item -msched-ar-in-data-spec
16040 @itemx -mno-sched-ar-in-data-spec
16041 @opindex msched-ar-in-data-spec
16042 @opindex mno-sched-ar-in-data-spec
16043 (En/Dis)able speculative scheduling of the instructions that
16044 are dependent on the data speculative loads after reload.
16045 This is effective only with @option{-msched-ar-data-spec} enabled.
16046 The default is 'enable'.
16048 @item -msched-in-control-spec
16049 @itemx -mno-sched-in-control-spec
16050 @opindex msched-in-control-spec
16051 @opindex mno-sched-in-control-spec
16052 (En/Dis)able speculative scheduling of the instructions that
16053 are dependent on the control speculative loads.
16054 This is effective only with @option{-msched-control-spec} enabled.
16055 The default is 'enable'.
16057 @item -mno-sched-prefer-non-data-spec-insns
16058 @itemx -msched-prefer-non-data-spec-insns
16059 @opindex mno-sched-prefer-non-data-spec-insns
16060 @opindex msched-prefer-non-data-spec-insns
16061 If enabled, data-speculative instructions are chosen for schedule
16062 only if there are no other choices at the moment. This makes
16063 the use of the data speculation much more conservative.
16064 The default is 'disable'.
16066 @item -mno-sched-prefer-non-control-spec-insns
16067 @itemx -msched-prefer-non-control-spec-insns
16068 @opindex mno-sched-prefer-non-control-spec-insns
16069 @opindex msched-prefer-non-control-spec-insns
16070 If enabled, control-speculative instructions are chosen for schedule
16071 only if there are no other choices at the moment. This makes
16072 the use of the control speculation much more conservative.
16073 The default is 'disable'.
16075 @item -mno-sched-count-spec-in-critical-path
16076 @itemx -msched-count-spec-in-critical-path
16077 @opindex mno-sched-count-spec-in-critical-path
16078 @opindex msched-count-spec-in-critical-path
16079 If enabled, speculative dependencies are considered during
16080 computation of the instructions priorities. This makes the use of the
16081 speculation a bit more conservative.
16082 The default is 'disable'.
16084 @item -msched-spec-ldc
16085 @opindex msched-spec-ldc
16086 Use a simple data speculation check. This option is on by default.
16088 @item -msched-control-spec-ldc
16089 @opindex msched-spec-ldc
16090 Use a simple check for control speculation. This option is on by default.
16092 @item -msched-stop-bits-after-every-cycle
16093 @opindex msched-stop-bits-after-every-cycle
16094 Place a stop bit after every cycle when scheduling. This option is on
16097 @item -msched-fp-mem-deps-zero-cost
16098 @opindex msched-fp-mem-deps-zero-cost
16099 Assume that floating-point stores and loads are not likely to cause a conflict
16100 when placed into the same instruction group. This option is disabled by
16103 @item -msel-sched-dont-check-control-spec
16104 @opindex msel-sched-dont-check-control-spec
16105 Generate checks for control speculation in selective scheduling.
16106 This flag is disabled by default.
16108 @item -msched-max-memory-insns=@var{max-insns}
16109 @opindex msched-max-memory-insns
16110 Limit on the number of memory insns per instruction group, giving lower
16111 priority to subsequent memory insns attempting to schedule in the same
16112 instruction group. Frequently useful to prevent cache bank conflicts.
16113 The default value is 1.
16115 @item -msched-max-memory-insns-hard-limit
16116 @opindex msched-max-memory-insns-hard-limit
16117 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16118 disallowing more than that number in an instruction group.
16119 Otherwise, the limit is ``soft'', meaning that non-memory operations
16120 are preferred when the limit is reached, but memory operations may still
16126 @subsection LM32 Options
16127 @cindex LM32 options
16129 These @option{-m} options are defined for the LatticeMico32 architecture:
16132 @item -mbarrel-shift-enabled
16133 @opindex mbarrel-shift-enabled
16134 Enable barrel-shift instructions.
16136 @item -mdivide-enabled
16137 @opindex mdivide-enabled
16138 Enable divide and modulus instructions.
16140 @item -mmultiply-enabled
16141 @opindex multiply-enabled
16142 Enable multiply instructions.
16144 @item -msign-extend-enabled
16145 @opindex msign-extend-enabled
16146 Enable sign extend instructions.
16148 @item -muser-enabled
16149 @opindex muser-enabled
16150 Enable user-defined instructions.
16155 @subsection M32C Options
16156 @cindex M32C options
16159 @item -mcpu=@var{name}
16161 Select the CPU for which code is generated. @var{name} may be one of
16162 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16163 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16164 the M32C/80 series.
16168 Specifies that the program will be run on the simulator. This causes
16169 an alternate runtime library to be linked in which supports, for
16170 example, file I/O@. You must not use this option when generating
16171 programs that will run on real hardware; you must provide your own
16172 runtime library for whatever I/O functions are needed.
16174 @item -memregs=@var{number}
16176 Specifies the number of memory-based pseudo-registers GCC uses
16177 during code generation. These pseudo-registers are used like real
16178 registers, so there is a tradeoff between GCC's ability to fit the
16179 code into available registers, and the performance penalty of using
16180 memory instead of registers. Note that all modules in a program must
16181 be compiled with the same value for this option. Because of that, you
16182 must not use this option with GCC's default runtime libraries.
16186 @node M32R/D Options
16187 @subsection M32R/D Options
16188 @cindex M32R/D options
16190 These @option{-m} options are defined for Renesas M32R/D architectures:
16195 Generate code for the M32R/2@.
16199 Generate code for the M32R/X@.
16203 Generate code for the M32R@. This is the default.
16205 @item -mmodel=small
16206 @opindex mmodel=small
16207 Assume all objects live in the lower 16MB of memory (so that their addresses
16208 can be loaded with the @code{ld24} instruction), and assume all subroutines
16209 are reachable with the @code{bl} instruction.
16210 This is the default.
16212 The addressability of a particular object can be set with the
16213 @code{model} attribute.
16215 @item -mmodel=medium
16216 @opindex mmodel=medium
16217 Assume objects may be anywhere in the 32-bit address space (the compiler
16218 generates @code{seth/add3} instructions to load their addresses), and
16219 assume all subroutines are reachable with the @code{bl} instruction.
16221 @item -mmodel=large
16222 @opindex mmodel=large
16223 Assume objects may be anywhere in the 32-bit address space (the compiler
16224 generates @code{seth/add3} instructions to load their addresses), and
16225 assume subroutines may not be reachable with the @code{bl} instruction
16226 (the compiler generates the much slower @code{seth/add3/jl}
16227 instruction sequence).
16230 @opindex msdata=none
16231 Disable use of the small data area. Variables are put into
16232 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16233 @code{section} attribute has been specified).
16234 This is the default.
16236 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16237 Objects may be explicitly put in the small data area with the
16238 @code{section} attribute using one of these sections.
16240 @item -msdata=sdata
16241 @opindex msdata=sdata
16242 Put small global and static data in the small data area, but do not
16243 generate special code to reference them.
16246 @opindex msdata=use
16247 Put small global and static data in the small data area, and generate
16248 special instructions to reference them.
16252 @cindex smaller data references
16253 Put global and static objects less than or equal to @var{num} bytes
16254 into the small data or BSS sections instead of the normal data or BSS
16255 sections. The default value of @var{num} is 8.
16256 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16257 for this option to have any effect.
16259 All modules should be compiled with the same @option{-G @var{num}} value.
16260 Compiling with different values of @var{num} may or may not work; if it
16261 doesn't the linker gives an error message---incorrect code is not
16266 Makes the M32R-specific code in the compiler display some statistics
16267 that might help in debugging programs.
16269 @item -malign-loops
16270 @opindex malign-loops
16271 Align all loops to a 32-byte boundary.
16273 @item -mno-align-loops
16274 @opindex mno-align-loops
16275 Do not enforce a 32-byte alignment for loops. This is the default.
16277 @item -missue-rate=@var{number}
16278 @opindex missue-rate=@var{number}
16279 Issue @var{number} instructions per cycle. @var{number} can only be 1
16282 @item -mbranch-cost=@var{number}
16283 @opindex mbranch-cost=@var{number}
16284 @var{number} can only be 1 or 2. If it is 1 then branches are
16285 preferred over conditional code, if it is 2, then the opposite applies.
16287 @item -mflush-trap=@var{number}
16288 @opindex mflush-trap=@var{number}
16289 Specifies the trap number to use to flush the cache. The default is
16290 12. Valid numbers are between 0 and 15 inclusive.
16292 @item -mno-flush-trap
16293 @opindex mno-flush-trap
16294 Specifies that the cache cannot be flushed by using a trap.
16296 @item -mflush-func=@var{name}
16297 @opindex mflush-func=@var{name}
16298 Specifies the name of the operating system function to call to flush
16299 the cache. The default is @samp{_flush_cache}, but a function call
16300 is only used if a trap is not available.
16302 @item -mno-flush-func
16303 @opindex mno-flush-func
16304 Indicates that there is no OS function for flushing the cache.
16308 @node M680x0 Options
16309 @subsection M680x0 Options
16310 @cindex M680x0 options
16312 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16313 The default settings depend on which architecture was selected when
16314 the compiler was configured; the defaults for the most common choices
16318 @item -march=@var{arch}
16320 Generate code for a specific M680x0 or ColdFire instruction set
16321 architecture. Permissible values of @var{arch} for M680x0
16322 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16323 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16324 architectures are selected according to Freescale's ISA classification
16325 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16326 @samp{isab} and @samp{isac}.
16328 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16329 code for a ColdFire target. The @var{arch} in this macro is one of the
16330 @option{-march} arguments given above.
16332 When used together, @option{-march} and @option{-mtune} select code
16333 that runs on a family of similar processors but that is optimized
16334 for a particular microarchitecture.
16336 @item -mcpu=@var{cpu}
16338 Generate code for a specific M680x0 or ColdFire processor.
16339 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16340 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16341 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16342 below, which also classifies the CPUs into families:
16344 @multitable @columnfractions 0.20 0.80
16345 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16346 @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}
16347 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16348 @item @samp{5206e} @tab @samp{5206e}
16349 @item @samp{5208} @tab @samp{5207} @samp{5208}
16350 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16351 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16352 @item @samp{5216} @tab @samp{5214} @samp{5216}
16353 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16354 @item @samp{5225} @tab @samp{5224} @samp{5225}
16355 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16356 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16357 @item @samp{5249} @tab @samp{5249}
16358 @item @samp{5250} @tab @samp{5250}
16359 @item @samp{5271} @tab @samp{5270} @samp{5271}
16360 @item @samp{5272} @tab @samp{5272}
16361 @item @samp{5275} @tab @samp{5274} @samp{5275}
16362 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16363 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16364 @item @samp{5307} @tab @samp{5307}
16365 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16366 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16367 @item @samp{5407} @tab @samp{5407}
16368 @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}
16371 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16372 @var{arch} is compatible with @var{cpu}. Other combinations of
16373 @option{-mcpu} and @option{-march} are rejected.
16375 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16376 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16377 where the value of @var{family} is given by the table above.
16379 @item -mtune=@var{tune}
16381 Tune the code for a particular microarchitecture within the
16382 constraints set by @option{-march} and @option{-mcpu}.
16383 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16384 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16385 and @samp{cpu32}. The ColdFire microarchitectures
16386 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16388 You can also use @option{-mtune=68020-40} for code that needs
16389 to run relatively well on 68020, 68030 and 68040 targets.
16390 @option{-mtune=68020-60} is similar but includes 68060 targets
16391 as well. These two options select the same tuning decisions as
16392 @option{-m68020-40} and @option{-m68020-60} respectively.
16394 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16395 when tuning for 680x0 architecture @var{arch}. It also defines
16396 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16397 option is used. If GCC is tuning for a range of architectures,
16398 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16399 it defines the macros for every architecture in the range.
16401 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16402 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16403 of the arguments given above.
16409 Generate output for a 68000. This is the default
16410 when the compiler is configured for 68000-based systems.
16411 It is equivalent to @option{-march=68000}.
16413 Use this option for microcontrollers with a 68000 or EC000 core,
16414 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16418 Generate output for a 68010. This is the default
16419 when the compiler is configured for 68010-based systems.
16420 It is equivalent to @option{-march=68010}.
16426 Generate output for a 68020. This is the default
16427 when the compiler is configured for 68020-based systems.
16428 It is equivalent to @option{-march=68020}.
16432 Generate output for a 68030. This is the default when the compiler is
16433 configured for 68030-based systems. It is equivalent to
16434 @option{-march=68030}.
16438 Generate output for a 68040. This is the default when the compiler is
16439 configured for 68040-based systems. It is equivalent to
16440 @option{-march=68040}.
16442 This option inhibits the use of 68881/68882 instructions that have to be
16443 emulated by software on the 68040. Use this option if your 68040 does not
16444 have code to emulate those instructions.
16448 Generate output for a 68060. This is the default when the compiler is
16449 configured for 68060-based systems. It is equivalent to
16450 @option{-march=68060}.
16452 This option inhibits the use of 68020 and 68881/68882 instructions that
16453 have to be emulated by software on the 68060. Use this option if your 68060
16454 does not have code to emulate those instructions.
16458 Generate output for a CPU32. This is the default
16459 when the compiler is configured for CPU32-based systems.
16460 It is equivalent to @option{-march=cpu32}.
16462 Use this option for microcontrollers with a
16463 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16464 68336, 68340, 68341, 68349 and 68360.
16468 Generate output for a 520X ColdFire CPU@. This is the default
16469 when the compiler is configured for 520X-based systems.
16470 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16471 in favor of that option.
16473 Use this option for microcontroller with a 5200 core, including
16474 the MCF5202, MCF5203, MCF5204 and MCF5206.
16478 Generate output for a 5206e ColdFire CPU@. The option is now
16479 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16483 Generate output for a member of the ColdFire 528X family.
16484 The option is now deprecated in favor of the equivalent
16485 @option{-mcpu=528x}.
16489 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16490 in favor of the equivalent @option{-mcpu=5307}.
16494 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16495 in favor of the equivalent @option{-mcpu=5407}.
16499 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16500 This includes use of hardware floating-point instructions.
16501 The option is equivalent to @option{-mcpu=547x}, and is now
16502 deprecated in favor of that option.
16506 Generate output for a 68040, without using any of the new instructions.
16507 This results in code that can run relatively efficiently on either a
16508 68020/68881 or a 68030 or a 68040. The generated code does use the
16509 68881 instructions that are emulated on the 68040.
16511 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16515 Generate output for a 68060, without using any of the new instructions.
16516 This results in code that can run relatively efficiently on either a
16517 68020/68881 or a 68030 or a 68040. The generated code does use the
16518 68881 instructions that are emulated on the 68060.
16520 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16524 @opindex mhard-float
16526 Generate floating-point instructions. This is the default for 68020
16527 and above, and for ColdFire devices that have an FPU@. It defines the
16528 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16529 on ColdFire targets.
16532 @opindex msoft-float
16533 Do not generate floating-point instructions; use library calls instead.
16534 This is the default for 68000, 68010, and 68832 targets. It is also
16535 the default for ColdFire devices that have no FPU.
16541 Generate (do not generate) ColdFire hardware divide and remainder
16542 instructions. If @option{-march} is used without @option{-mcpu},
16543 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16544 architectures. Otherwise, the default is taken from the target CPU
16545 (either the default CPU, or the one specified by @option{-mcpu}). For
16546 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16547 @option{-mcpu=5206e}.
16549 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16553 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16554 Additionally, parameters passed on the stack are also aligned to a
16555 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16559 Do not consider type @code{int} to be 16 bits wide. This is the default.
16562 @itemx -mno-bitfield
16563 @opindex mnobitfield
16564 @opindex mno-bitfield
16565 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16566 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16570 Do use the bit-field instructions. The @option{-m68020} option implies
16571 @option{-mbitfield}. This is the default if you use a configuration
16572 designed for a 68020.
16576 Use a different function-calling convention, in which functions
16577 that take a fixed number of arguments return with the @code{rtd}
16578 instruction, which pops their arguments while returning. This
16579 saves one instruction in the caller since there is no need to pop
16580 the arguments there.
16582 This calling convention is incompatible with the one normally
16583 used on Unix, so you cannot use it if you need to call libraries
16584 compiled with the Unix compiler.
16586 Also, you must provide function prototypes for all functions that
16587 take variable numbers of arguments (including @code{printf});
16588 otherwise incorrect code is generated for calls to those
16591 In addition, seriously incorrect code results if you call a
16592 function with too many arguments. (Normally, extra arguments are
16593 harmlessly ignored.)
16595 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16596 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16600 Do not use the calling conventions selected by @option{-mrtd}.
16601 This is the default.
16604 @itemx -mno-align-int
16605 @opindex malign-int
16606 @opindex mno-align-int
16607 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16608 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16609 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16610 Aligning variables on 32-bit boundaries produces code that runs somewhat
16611 faster on processors with 32-bit busses at the expense of more memory.
16613 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16614 aligns structures containing the above types differently than
16615 most published application binary interface specifications for the m68k.
16619 Use the pc-relative addressing mode of the 68000 directly, instead of
16620 using a global offset table. At present, this option implies @option{-fpic},
16621 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16622 not presently supported with @option{-mpcrel}, though this could be supported for
16623 68020 and higher processors.
16625 @item -mno-strict-align
16626 @itemx -mstrict-align
16627 @opindex mno-strict-align
16628 @opindex mstrict-align
16629 Do not (do) assume that unaligned memory references are handled by
16633 Generate code that allows the data segment to be located in a different
16634 area of memory from the text segment. This allows for execute-in-place in
16635 an environment without virtual memory management. This option implies
16638 @item -mno-sep-data
16639 Generate code that assumes that the data segment follows the text segment.
16640 This is the default.
16642 @item -mid-shared-library
16643 Generate code that supports shared libraries via the library ID method.
16644 This allows for execute-in-place and shared libraries in an environment
16645 without virtual memory management. This option implies @option{-fPIC}.
16647 @item -mno-id-shared-library
16648 Generate code that doesn't assume ID-based shared libraries are being used.
16649 This is the default.
16651 @item -mshared-library-id=n
16652 Specifies the identification number of the ID-based shared library being
16653 compiled. Specifying a value of 0 generates more compact code; specifying
16654 other values forces the allocation of that number to the current
16655 library, but is no more space- or time-efficient than omitting this option.
16661 When generating position-independent code for ColdFire, generate code
16662 that works if the GOT has more than 8192 entries. This code is
16663 larger and slower than code generated without this option. On M680x0
16664 processors, this option is not needed; @option{-fPIC} suffices.
16666 GCC normally uses a single instruction to load values from the GOT@.
16667 While this is relatively efficient, it only works if the GOT
16668 is smaller than about 64k. Anything larger causes the linker
16669 to report an error such as:
16671 @cindex relocation truncated to fit (ColdFire)
16673 relocation truncated to fit: R_68K_GOT16O foobar
16676 If this happens, you should recompile your code with @option{-mxgot}.
16677 It should then work with very large GOTs. However, code generated with
16678 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16679 the value of a global symbol.
16681 Note that some linkers, including newer versions of the GNU linker,
16682 can create multiple GOTs and sort GOT entries. If you have such a linker,
16683 you should only need to use @option{-mxgot} when compiling a single
16684 object file that accesses more than 8192 GOT entries. Very few do.
16686 These options have no effect unless GCC is generating
16687 position-independent code.
16691 @node MCore Options
16692 @subsection MCore Options
16693 @cindex MCore options
16695 These are the @samp{-m} options defined for the Motorola M*Core
16701 @itemx -mno-hardlit
16703 @opindex mno-hardlit
16704 Inline constants into the code stream if it can be done in two
16705 instructions or less.
16711 Use the divide instruction. (Enabled by default).
16713 @item -mrelax-immediate
16714 @itemx -mno-relax-immediate
16715 @opindex mrelax-immediate
16716 @opindex mno-relax-immediate
16717 Allow arbitrary-sized immediates in bit operations.
16719 @item -mwide-bitfields
16720 @itemx -mno-wide-bitfields
16721 @opindex mwide-bitfields
16722 @opindex mno-wide-bitfields
16723 Always treat bit-fields as @code{int}-sized.
16725 @item -m4byte-functions
16726 @itemx -mno-4byte-functions
16727 @opindex m4byte-functions
16728 @opindex mno-4byte-functions
16729 Force all functions to be aligned to a 4-byte boundary.
16731 @item -mcallgraph-data
16732 @itemx -mno-callgraph-data
16733 @opindex mcallgraph-data
16734 @opindex mno-callgraph-data
16735 Emit callgraph information.
16738 @itemx -mno-slow-bytes
16739 @opindex mslow-bytes
16740 @opindex mno-slow-bytes
16741 Prefer word access when reading byte quantities.
16743 @item -mlittle-endian
16744 @itemx -mbig-endian
16745 @opindex mlittle-endian
16746 @opindex mbig-endian
16747 Generate code for a little-endian target.
16753 Generate code for the 210 processor.
16757 Assume that runtime support has been provided and so omit the
16758 simulator library (@file{libsim.a)} from the linker command line.
16760 @item -mstack-increment=@var{size}
16761 @opindex mstack-increment
16762 Set the maximum amount for a single stack increment operation. Large
16763 values can increase the speed of programs that contain functions
16764 that need a large amount of stack space, but they can also trigger a
16765 segmentation fault if the stack is extended too much. The default
16771 @subsection MeP Options
16772 @cindex MeP options
16778 Enables the @code{abs} instruction, which is the absolute difference
16779 between two registers.
16783 Enables all the optional instructions---average, multiply, divide, bit
16784 operations, leading zero, absolute difference, min/max, clip, and
16790 Enables the @code{ave} instruction, which computes the average of two
16793 @item -mbased=@var{n}
16795 Variables of size @var{n} bytes or smaller are placed in the
16796 @code{.based} section by default. Based variables use the @code{$tp}
16797 register as a base register, and there is a 128-byte limit to the
16798 @code{.based} section.
16802 Enables the bit operation instructions---bit test (@code{btstm}), set
16803 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
16804 test-and-set (@code{tas}).
16806 @item -mc=@var{name}
16808 Selects which section constant data is placed in. @var{name} may
16809 be @samp{tiny}, @samp{near}, or @samp{far}.
16813 Enables the @code{clip} instruction. Note that @option{-mclip} is not
16814 useful unless you also provide @option{-mminmax}.
16816 @item -mconfig=@var{name}
16818 Selects one of the built-in core configurations. Each MeP chip has
16819 one or more modules in it; each module has a core CPU and a variety of
16820 coprocessors, optional instructions, and peripherals. The
16821 @code{MeP-Integrator} tool, not part of GCC, provides these
16822 configurations through this option; using this option is the same as
16823 using all the corresponding command-line options. The default
16824 configuration is @samp{default}.
16828 Enables the coprocessor instructions. By default, this is a 32-bit
16829 coprocessor. Note that the coprocessor is normally enabled via the
16830 @option{-mconfig=} option.
16834 Enables the 32-bit coprocessor's instructions.
16838 Enables the 64-bit coprocessor's instructions.
16842 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
16846 Causes constant variables to be placed in the @code{.near} section.
16850 Enables the @code{div} and @code{divu} instructions.
16854 Generate big-endian code.
16858 Generate little-endian code.
16860 @item -mio-volatile
16861 @opindex mio-volatile
16862 Tells the compiler that any variable marked with the @code{io}
16863 attribute is to be considered volatile.
16867 Causes variables to be assigned to the @code{.far} section by default.
16871 Enables the @code{leadz} (leading zero) instruction.
16875 Causes variables to be assigned to the @code{.near} section by default.
16879 Enables the @code{min} and @code{max} instructions.
16883 Enables the multiplication and multiply-accumulate instructions.
16887 Disables all the optional instructions enabled by @option{-mall-opts}.
16891 Enables the @code{repeat} and @code{erepeat} instructions, used for
16892 low-overhead looping.
16896 Causes all variables to default to the @code{.tiny} section. Note
16897 that there is a 65536-byte limit to this section. Accesses to these
16898 variables use the @code{%gp} base register.
16902 Enables the saturation instructions. Note that the compiler does not
16903 currently generate these itself, but this option is included for
16904 compatibility with other tools, like @code{as}.
16908 Link the SDRAM-based runtime instead of the default ROM-based runtime.
16912 Link the simulator run-time libraries.
16916 Link the simulator runtime libraries, excluding built-in support
16917 for reset and exception vectors and tables.
16921 Causes all functions to default to the @code{.far} section. Without
16922 this option, functions default to the @code{.near} section.
16924 @item -mtiny=@var{n}
16926 Variables that are @var{n} bytes or smaller are allocated to the
16927 @code{.tiny} section. These variables use the @code{$gp} base
16928 register. The default for this option is 4, but note that there's a
16929 65536-byte limit to the @code{.tiny} section.
16933 @node MicroBlaze Options
16934 @subsection MicroBlaze Options
16935 @cindex MicroBlaze Options
16940 @opindex msoft-float
16941 Use software emulation for floating point (default).
16944 @opindex mhard-float
16945 Use hardware floating-point instructions.
16949 Do not optimize block moves, use @code{memcpy}.
16951 @item -mno-clearbss
16952 @opindex mno-clearbss
16953 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
16955 @item -mcpu=@var{cpu-type}
16957 Use features of, and schedule code for, the given CPU.
16958 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
16959 where @var{X} is a major version, @var{YY} is the minor version, and
16960 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
16961 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
16963 @item -mxl-soft-mul
16964 @opindex mxl-soft-mul
16965 Use software multiply emulation (default).
16967 @item -mxl-soft-div
16968 @opindex mxl-soft-div
16969 Use software emulation for divides (default).
16971 @item -mxl-barrel-shift
16972 @opindex mxl-barrel-shift
16973 Use the hardware barrel shifter.
16975 @item -mxl-pattern-compare
16976 @opindex mxl-pattern-compare
16977 Use pattern compare instructions.
16979 @item -msmall-divides
16980 @opindex msmall-divides
16981 Use table lookup optimization for small signed integer divisions.
16983 @item -mxl-stack-check
16984 @opindex mxl-stack-check
16985 This option is deprecated. Use @option{-fstack-check} instead.
16988 @opindex mxl-gp-opt
16989 Use GP-relative @code{.sdata}/@code{.sbss} sections.
16991 @item -mxl-multiply-high
16992 @opindex mxl-multiply-high
16993 Use multiply high instructions for high part of 32x32 multiply.
16995 @item -mxl-float-convert
16996 @opindex mxl-float-convert
16997 Use hardware floating-point conversion instructions.
16999 @item -mxl-float-sqrt
17000 @opindex mxl-float-sqrt
17001 Use hardware floating-point square root instruction.
17004 @opindex mbig-endian
17005 Generate code for a big-endian target.
17007 @item -mlittle-endian
17008 @opindex mlittle-endian
17009 Generate code for a little-endian target.
17012 @opindex mxl-reorder
17013 Use reorder instructions (swap and byte reversed load/store).
17015 @item -mxl-mode-@var{app-model}
17016 Select application model @var{app-model}. Valid models are
17019 normal executable (default), uses startup code @file{crt0.o}.
17022 for use with Xilinx Microprocessor Debugger (XMD) based
17023 software intrusive debug agent called xmdstub. This uses startup file
17024 @file{crt1.o} and sets the start address of the program to 0x800.
17027 for applications that are loaded using a bootloader.
17028 This model uses startup file @file{crt2.o} which does not contain a processor
17029 reset vector handler. This is suitable for transferring control on a
17030 processor reset to the bootloader rather than the application.
17033 for applications that do not require any of the
17034 MicroBlaze vectors. This option may be useful for applications running
17035 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17038 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17039 @option{-mxl-mode-@var{app-model}}.
17044 @subsection MIPS Options
17045 @cindex MIPS options
17051 Generate big-endian code.
17055 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17058 @item -march=@var{arch}
17060 Generate code that runs on @var{arch}, which can be the name of a
17061 generic MIPS ISA, or the name of a particular processor.
17063 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17064 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17065 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17066 @samp{mips64r5} and @samp{mips64r6}.
17067 The processor names are:
17068 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17069 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17070 @samp{5kc}, @samp{5kf},
17072 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17073 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17074 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17075 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17076 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17079 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17081 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17082 @samp{m5100}, @samp{m5101},
17083 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17086 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17087 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17088 @samp{rm7000}, @samp{rm9000},
17089 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17092 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17093 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17094 @samp{xlr} and @samp{xlp}.
17095 The special value @samp{from-abi} selects the
17096 most compatible architecture for the selected ABI (that is,
17097 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17099 The native Linux/GNU toolchain also supports the value @samp{native},
17100 which selects the best architecture option for the host processor.
17101 @option{-march=native} has no effect if GCC does not recognize
17104 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17105 (for example, @option{-march=r2k}). Prefixes are optional, and
17106 @samp{vr} may be written @samp{r}.
17108 Names of the form @samp{@var{n}f2_1} refer to processors with
17109 FPUs clocked at half the rate of the core, names of the form
17110 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17111 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17112 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17113 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17114 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17115 accepted as synonyms for @samp{@var{n}f1_1}.
17117 GCC defines two macros based on the value of this option. The first
17118 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17119 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17120 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17121 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17122 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17124 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17125 above. In other words, it has the full prefix and does not
17126 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17127 the macro names the resolved architecture (either @code{"mips1"} or
17128 @code{"mips3"}). It names the default architecture when no
17129 @option{-march} option is given.
17131 @item -mtune=@var{arch}
17133 Optimize for @var{arch}. Among other things, this option controls
17134 the way instructions are scheduled, and the perceived cost of arithmetic
17135 operations. The list of @var{arch} values is the same as for
17138 When this option is not used, GCC optimizes for the processor
17139 specified by @option{-march}. By using @option{-march} and
17140 @option{-mtune} together, it is possible to generate code that
17141 runs on a family of processors, but optimize the code for one
17142 particular member of that family.
17144 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17145 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17146 @option{-march} ones described above.
17150 Equivalent to @option{-march=mips1}.
17154 Equivalent to @option{-march=mips2}.
17158 Equivalent to @option{-march=mips3}.
17162 Equivalent to @option{-march=mips4}.
17166 Equivalent to @option{-march=mips32}.
17170 Equivalent to @option{-march=mips32r3}.
17174 Equivalent to @option{-march=mips32r5}.
17178 Equivalent to @option{-march=mips32r6}.
17182 Equivalent to @option{-march=mips64}.
17186 Equivalent to @option{-march=mips64r2}.
17190 Equivalent to @option{-march=mips64r3}.
17194 Equivalent to @option{-march=mips64r5}.
17198 Equivalent to @option{-march=mips64r6}.
17203 @opindex mno-mips16
17204 Generate (do not generate) MIPS16 code. If GCC is targeting a
17205 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17207 MIPS16 code generation can also be controlled on a per-function basis
17208 by means of @code{mips16} and @code{nomips16} attributes.
17209 @xref{Function Attributes}, for more information.
17211 @item -mflip-mips16
17212 @opindex mflip-mips16
17213 Generate MIPS16 code on alternating functions. This option is provided
17214 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17215 not intended for ordinary use in compiling user code.
17217 @item -minterlink-compressed
17218 @item -mno-interlink-compressed
17219 @opindex minterlink-compressed
17220 @opindex mno-interlink-compressed
17221 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17222 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17224 For example, code using the standard ISA encoding cannot jump directly
17225 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17226 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17227 knows that the target of the jump is not compressed.
17229 @item -minterlink-mips16
17230 @itemx -mno-interlink-mips16
17231 @opindex minterlink-mips16
17232 @opindex mno-interlink-mips16
17233 Aliases of @option{-minterlink-compressed} and
17234 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17235 and are retained for backwards compatibility.
17247 Generate code for the given ABI@.
17249 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17250 generates 64-bit code when you select a 64-bit architecture, but you
17251 can use @option{-mgp32} to get 32-bit code instead.
17253 For information about the O64 ABI, see
17254 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17256 GCC supports a variant of the o32 ABI in which floating-point registers
17257 are 64 rather than 32 bits wide. You can select this combination with
17258 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17259 and @code{mfhc1} instructions and is therefore only supported for
17260 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17262 The register assignments for arguments and return values remain the
17263 same, but each scalar value is passed in a single 64-bit register
17264 rather than a pair of 32-bit registers. For example, scalar
17265 floating-point values are returned in @samp{$f0} only, not a
17266 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17267 remains the same in that the even-numbered double-precision registers
17270 Two additional variants of the o32 ABI are supported to enable
17271 a transition from 32-bit to 64-bit registers. These are FPXX
17272 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17273 The FPXX extension mandates that all code must execute correctly
17274 when run using 32-bit or 64-bit registers. The code can be interlinked
17275 with either FP32 or FP64, but not both.
17276 The FP64A extension is similar to the FP64 extension but forbids the
17277 use of odd-numbered single-precision registers. This can be used
17278 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17279 processors and allows both FP32 and FP64A code to interlink and
17280 run in the same process without changing FPU modes.
17283 @itemx -mno-abicalls
17285 @opindex mno-abicalls
17286 Generate (do not generate) code that is suitable for SVR4-style
17287 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17292 Generate (do not generate) code that is fully position-independent,
17293 and that can therefore be linked into shared libraries. This option
17294 only affects @option{-mabicalls}.
17296 All @option{-mabicalls} code has traditionally been position-independent,
17297 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17298 as an extension, the GNU toolchain allows executables to use absolute
17299 accesses for locally-binding symbols. It can also use shorter GP
17300 initialization sequences and generate direct calls to locally-defined
17301 functions. This mode is selected by @option{-mno-shared}.
17303 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17304 objects that can only be linked by the GNU linker. However, the option
17305 does not affect the ABI of the final executable; it only affects the ABI
17306 of relocatable objects. Using @option{-mno-shared} generally makes
17307 executables both smaller and quicker.
17309 @option{-mshared} is the default.
17315 Assume (do not assume) that the static and dynamic linkers
17316 support PLTs and copy relocations. This option only affects
17317 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17318 has no effect without @option{-msym32}.
17320 You can make @option{-mplt} the default by configuring
17321 GCC with @option{--with-mips-plt}. The default is
17322 @option{-mno-plt} otherwise.
17328 Lift (do not lift) the usual restrictions on the size of the global
17331 GCC normally uses a single instruction to load values from the GOT@.
17332 While this is relatively efficient, it only works if the GOT
17333 is smaller than about 64k. Anything larger causes the linker
17334 to report an error such as:
17336 @cindex relocation truncated to fit (MIPS)
17338 relocation truncated to fit: R_MIPS_GOT16 foobar
17341 If this happens, you should recompile your code with @option{-mxgot}.
17342 This works with very large GOTs, although the code is also
17343 less efficient, since it takes three instructions to fetch the
17344 value of a global symbol.
17346 Note that some linkers can create multiple GOTs. If you have such a
17347 linker, you should only need to use @option{-mxgot} when a single object
17348 file accesses more than 64k's worth of GOT entries. Very few do.
17350 These options have no effect unless GCC is generating position
17355 Assume that general-purpose registers are 32 bits wide.
17359 Assume that general-purpose registers are 64 bits wide.
17363 Assume that floating-point registers are 32 bits wide.
17367 Assume that floating-point registers are 64 bits wide.
17371 Do not assume the width of floating-point registers.
17374 @opindex mhard-float
17375 Use floating-point coprocessor instructions.
17378 @opindex msoft-float
17379 Do not use floating-point coprocessor instructions. Implement
17380 floating-point calculations using library calls instead.
17384 Equivalent to @option{-msoft-float}, but additionally asserts that the
17385 program being compiled does not perform any floating-point operations.
17386 This option is presently supported only by some bare-metal MIPS
17387 configurations, where it may select a special set of libraries
17388 that lack all floating-point support (including, for example, the
17389 floating-point @code{printf} formats).
17390 If code compiled with @option{-mno-float} accidentally contains
17391 floating-point operations, it is likely to suffer a link-time
17392 or run-time failure.
17394 @item -msingle-float
17395 @opindex msingle-float
17396 Assume that the floating-point coprocessor only supports single-precision
17399 @item -mdouble-float
17400 @opindex mdouble-float
17401 Assume that the floating-point coprocessor supports double-precision
17402 operations. This is the default.
17405 @itemx -mno-odd-spreg
17406 @opindex modd-spreg
17407 @opindex mno-odd-spreg
17408 Enable the use of odd-numbered single-precision floating-point registers
17409 for the o32 ABI. This is the default for processors that are known to
17410 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17414 @itemx -mabs=legacy
17416 @opindex mabs=legacy
17417 These options control the treatment of the special not-a-number (NaN)
17418 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17419 @code{neg.@i{fmt}} machine instructions.
17421 By default or when @option{-mabs=legacy} is used the legacy
17422 treatment is selected. In this case these instructions are considered
17423 arithmetic and avoided where correct operation is required and the
17424 input operand might be a NaN. A longer sequence of instructions that
17425 manipulate the sign bit of floating-point datum manually is used
17426 instead unless the @option{-ffinite-math-only} option has also been
17429 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17430 this case these instructions are considered non-arithmetic and therefore
17431 operating correctly in all cases, including in particular where the
17432 input operand is a NaN. These instructions are therefore always used
17433 for the respective operations.
17436 @itemx -mnan=legacy
17438 @opindex mnan=legacy
17439 These options control the encoding of the special not-a-number (NaN)
17440 IEEE 754 floating-point data.
17442 The @option{-mnan=legacy} option selects the legacy encoding. In this
17443 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17444 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17445 by the first bit of their trailing significand field being 1.
17447 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17448 this case qNaNs are denoted by the first bit of their trailing
17449 significand field being 1, whereas sNaNs are denoted by the first bit of
17450 their trailing significand field being 0.
17452 The default is @option{-mnan=legacy} unless GCC has been configured with
17453 @option{--with-nan=2008}.
17459 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17460 implement atomic memory built-in functions. When neither option is
17461 specified, GCC uses the instructions if the target architecture
17464 @option{-mllsc} is useful if the runtime environment can emulate the
17465 instructions and @option{-mno-llsc} can be useful when compiling for
17466 nonstandard ISAs. You can make either option the default by
17467 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17468 respectively. @option{--with-llsc} is the default for some
17469 configurations; see the installation documentation for details.
17475 Use (do not use) revision 1 of the MIPS DSP ASE@.
17476 @xref{MIPS DSP Built-in Functions}. This option defines the
17477 preprocessor macro @code{__mips_dsp}. It also defines
17478 @code{__mips_dsp_rev} to 1.
17484 Use (do not use) revision 2 of the MIPS DSP ASE@.
17485 @xref{MIPS DSP Built-in Functions}. This option defines the
17486 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17487 It also defines @code{__mips_dsp_rev} to 2.
17490 @itemx -mno-smartmips
17491 @opindex msmartmips
17492 @opindex mno-smartmips
17493 Use (do not use) the MIPS SmartMIPS ASE.
17495 @item -mpaired-single
17496 @itemx -mno-paired-single
17497 @opindex mpaired-single
17498 @opindex mno-paired-single
17499 Use (do not use) paired-single floating-point instructions.
17500 @xref{MIPS Paired-Single Support}. This option requires
17501 hardware floating-point support to be enabled.
17507 Use (do not use) MIPS Digital Media Extension instructions.
17508 This option can only be used when generating 64-bit code and requires
17509 hardware floating-point support to be enabled.
17514 @opindex mno-mips3d
17515 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17516 The option @option{-mips3d} implies @option{-mpaired-single}.
17519 @itemx -mno-micromips
17520 @opindex mmicromips
17521 @opindex mno-mmicromips
17522 Generate (do not generate) microMIPS code.
17524 MicroMIPS code generation can also be controlled on a per-function basis
17525 by means of @code{micromips} and @code{nomicromips} attributes.
17526 @xref{Function Attributes}, for more information.
17532 Use (do not use) MT Multithreading instructions.
17538 Use (do not use) the MIPS MCU ASE instructions.
17544 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17550 Use (do not use) the MIPS Virtualization Application Specific instructions.
17556 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17560 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17561 an explanation of the default and the way that the pointer size is
17566 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17568 The default size of @code{int}s, @code{long}s and pointers depends on
17569 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17570 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17571 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17572 or the same size as integer registers, whichever is smaller.
17578 Assume (do not assume) that all symbols have 32-bit values, regardless
17579 of the selected ABI@. This option is useful in combination with
17580 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17581 to generate shorter and faster references to symbolic addresses.
17585 Put definitions of externally-visible data in a small data section
17586 if that data is no bigger than @var{num} bytes. GCC can then generate
17587 more efficient accesses to the data; see @option{-mgpopt} for details.
17589 The default @option{-G} option depends on the configuration.
17591 @item -mlocal-sdata
17592 @itemx -mno-local-sdata
17593 @opindex mlocal-sdata
17594 @opindex mno-local-sdata
17595 Extend (do not extend) the @option{-G} behavior to local data too,
17596 such as to static variables in C@. @option{-mlocal-sdata} is the
17597 default for all configurations.
17599 If the linker complains that an application is using too much small data,
17600 you might want to try rebuilding the less performance-critical parts with
17601 @option{-mno-local-sdata}. You might also want to build large
17602 libraries with @option{-mno-local-sdata}, so that the libraries leave
17603 more room for the main program.
17605 @item -mextern-sdata
17606 @itemx -mno-extern-sdata
17607 @opindex mextern-sdata
17608 @opindex mno-extern-sdata
17609 Assume (do not assume) that externally-defined data is in
17610 a small data section if the size of that data is within the @option{-G} limit.
17611 @option{-mextern-sdata} is the default for all configurations.
17613 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17614 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17615 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17616 is placed in a small data section. If @var{Var} is defined by another
17617 module, you must either compile that module with a high-enough
17618 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17619 definition. If @var{Var} is common, you must link the application
17620 with a high-enough @option{-G} setting.
17622 The easiest way of satisfying these restrictions is to compile
17623 and link every module with the same @option{-G} option. However,
17624 you may wish to build a library that supports several different
17625 small data limits. You can do this by compiling the library with
17626 the highest supported @option{-G} setting and additionally using
17627 @option{-mno-extern-sdata} to stop the library from making assumptions
17628 about externally-defined data.
17634 Use (do not use) GP-relative accesses for symbols that are known to be
17635 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17636 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17639 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17640 might not hold the value of @code{_gp}. For example, if the code is
17641 part of a library that might be used in a boot monitor, programs that
17642 call boot monitor routines pass an unknown value in @code{$gp}.
17643 (In such situations, the boot monitor itself is usually compiled
17644 with @option{-G0}.)
17646 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17647 @option{-mno-extern-sdata}.
17649 @item -membedded-data
17650 @itemx -mno-embedded-data
17651 @opindex membedded-data
17652 @opindex mno-embedded-data
17653 Allocate variables to the read-only data section first if possible, then
17654 next in the small data section if possible, otherwise in data. This gives
17655 slightly slower code than the default, but reduces the amount of RAM required
17656 when executing, and thus may be preferred for some embedded systems.
17658 @item -muninit-const-in-rodata
17659 @itemx -mno-uninit-const-in-rodata
17660 @opindex muninit-const-in-rodata
17661 @opindex mno-uninit-const-in-rodata
17662 Put uninitialized @code{const} variables in the read-only data section.
17663 This option is only meaningful in conjunction with @option{-membedded-data}.
17665 @item -mcode-readable=@var{setting}
17666 @opindex mcode-readable
17667 Specify whether GCC may generate code that reads from executable sections.
17668 There are three possible settings:
17671 @item -mcode-readable=yes
17672 Instructions may freely access executable sections. This is the
17675 @item -mcode-readable=pcrel
17676 MIPS16 PC-relative load instructions can access executable sections,
17677 but other instructions must not do so. This option is useful on 4KSc
17678 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17679 It is also useful on processors that can be configured to have a dual
17680 instruction/data SRAM interface and that, like the M4K, automatically
17681 redirect PC-relative loads to the instruction RAM.
17683 @item -mcode-readable=no
17684 Instructions must not access executable sections. This option can be
17685 useful on targets that are configured to have a dual instruction/data
17686 SRAM interface but that (unlike the M4K) do not automatically redirect
17687 PC-relative loads to the instruction RAM.
17690 @item -msplit-addresses
17691 @itemx -mno-split-addresses
17692 @opindex msplit-addresses
17693 @opindex mno-split-addresses
17694 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17695 relocation operators. This option has been superseded by
17696 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17698 @item -mexplicit-relocs
17699 @itemx -mno-explicit-relocs
17700 @opindex mexplicit-relocs
17701 @opindex mno-explicit-relocs
17702 Use (do not use) assembler relocation operators when dealing with symbolic
17703 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17704 is to use assembler macros instead.
17706 @option{-mexplicit-relocs} is the default if GCC was configured
17707 to use an assembler that supports relocation operators.
17709 @item -mcheck-zero-division
17710 @itemx -mno-check-zero-division
17711 @opindex mcheck-zero-division
17712 @opindex mno-check-zero-division
17713 Trap (do not trap) on integer division by zero.
17715 The default is @option{-mcheck-zero-division}.
17717 @item -mdivide-traps
17718 @itemx -mdivide-breaks
17719 @opindex mdivide-traps
17720 @opindex mdivide-breaks
17721 MIPS systems check for division by zero by generating either a
17722 conditional trap or a break instruction. Using traps results in
17723 smaller code, but is only supported on MIPS II and later. Also, some
17724 versions of the Linux kernel have a bug that prevents trap from
17725 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17726 allow conditional traps on architectures that support them and
17727 @option{-mdivide-breaks} to force the use of breaks.
17729 The default is usually @option{-mdivide-traps}, but this can be
17730 overridden at configure time using @option{--with-divide=breaks}.
17731 Divide-by-zero checks can be completely disabled using
17732 @option{-mno-check-zero-division}.
17737 @opindex mno-memcpy
17738 Force (do not force) the use of @code{memcpy} for non-trivial block
17739 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17740 most constant-sized copies.
17743 @itemx -mno-long-calls
17744 @opindex mlong-calls
17745 @opindex mno-long-calls
17746 Disable (do not disable) use of the @code{jal} instruction. Calling
17747 functions using @code{jal} is more efficient but requires the caller
17748 and callee to be in the same 256 megabyte segment.
17750 This option has no effect on abicalls code. The default is
17751 @option{-mno-long-calls}.
17757 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17758 instructions, as provided by the R4650 ISA@.
17764 Enable (disable) use of the @code{madd} and @code{msub} integer
17765 instructions. The default is @option{-mimadd} on architectures
17766 that support @code{madd} and @code{msub} except for the 74k
17767 architecture where it was found to generate slower code.
17770 @itemx -mno-fused-madd
17771 @opindex mfused-madd
17772 @opindex mno-fused-madd
17773 Enable (disable) use of the floating-point multiply-accumulate
17774 instructions, when they are available. The default is
17775 @option{-mfused-madd}.
17777 On the R8000 CPU when multiply-accumulate instructions are used,
17778 the intermediate product is calculated to infinite precision
17779 and is not subject to the FCSR Flush to Zero bit. This may be
17780 undesirable in some circumstances. On other processors the result
17781 is numerically identical to the equivalent computation using
17782 separate multiply, add, subtract and negate instructions.
17786 Tell the MIPS assembler to not run its preprocessor over user
17787 assembler files (with a @samp{.s} suffix) when assembling them.
17792 @opindex mno-fix-24k
17793 Work around the 24K E48 (lost data on stores during refill) errata.
17794 The workarounds are implemented by the assembler rather than by GCC@.
17797 @itemx -mno-fix-r4000
17798 @opindex mfix-r4000
17799 @opindex mno-fix-r4000
17800 Work around certain R4000 CPU errata:
17803 A double-word or a variable shift may give an incorrect result if executed
17804 immediately after starting an integer division.
17806 A double-word or a variable shift may give an incorrect result if executed
17807 while an integer multiplication is in progress.
17809 An integer division may give an incorrect result if started in a delay slot
17810 of a taken branch or a jump.
17814 @itemx -mno-fix-r4400
17815 @opindex mfix-r4400
17816 @opindex mno-fix-r4400
17817 Work around certain R4400 CPU errata:
17820 A double-word or a variable shift may give an incorrect result if executed
17821 immediately after starting an integer division.
17825 @itemx -mno-fix-r10000
17826 @opindex mfix-r10000
17827 @opindex mno-fix-r10000
17828 Work around certain R10000 errata:
17831 @code{ll}/@code{sc} sequences may not behave atomically on revisions
17832 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
17835 This option can only be used if the target architecture supports
17836 branch-likely instructions. @option{-mfix-r10000} is the default when
17837 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
17841 @itemx -mno-fix-rm7000
17842 @opindex mfix-rm7000
17843 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
17844 workarounds are implemented by the assembler rather than by GCC@.
17847 @itemx -mno-fix-vr4120
17848 @opindex mfix-vr4120
17849 Work around certain VR4120 errata:
17852 @code{dmultu} does not always produce the correct result.
17854 @code{div} and @code{ddiv} do not always produce the correct result if one
17855 of the operands is negative.
17857 The workarounds for the division errata rely on special functions in
17858 @file{libgcc.a}. At present, these functions are only provided by
17859 the @code{mips64vr*-elf} configurations.
17861 Other VR4120 errata require a NOP to be inserted between certain pairs of
17862 instructions. These errata are handled by the assembler, not by GCC itself.
17865 @opindex mfix-vr4130
17866 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
17867 workarounds are implemented by the assembler rather than by GCC,
17868 although GCC avoids using @code{mflo} and @code{mfhi} if the
17869 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
17870 instructions are available instead.
17873 @itemx -mno-fix-sb1
17875 Work around certain SB-1 CPU core errata.
17876 (This flag currently works around the SB-1 revision 2
17877 ``F1'' and ``F2'' floating-point errata.)
17879 @item -mr10k-cache-barrier=@var{setting}
17880 @opindex mr10k-cache-barrier
17881 Specify whether GCC should insert cache barriers to avoid the
17882 side-effects of speculation on R10K processors.
17884 In common with many processors, the R10K tries to predict the outcome
17885 of a conditional branch and speculatively executes instructions from
17886 the ``taken'' branch. It later aborts these instructions if the
17887 predicted outcome is wrong. However, on the R10K, even aborted
17888 instructions can have side effects.
17890 This problem only affects kernel stores and, depending on the system,
17891 kernel loads. As an example, a speculatively-executed store may load
17892 the target memory into cache and mark the cache line as dirty, even if
17893 the store itself is later aborted. If a DMA operation writes to the
17894 same area of memory before the ``dirty'' line is flushed, the cached
17895 data overwrites the DMA-ed data. See the R10K processor manual
17896 for a full description, including other potential problems.
17898 One workaround is to insert cache barrier instructions before every memory
17899 access that might be speculatively executed and that might have side
17900 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
17901 controls GCC's implementation of this workaround. It assumes that
17902 aborted accesses to any byte in the following regions does not have
17907 the memory occupied by the current function's stack frame;
17910 the memory occupied by an incoming stack argument;
17913 the memory occupied by an object with a link-time-constant address.
17916 It is the kernel's responsibility to ensure that speculative
17917 accesses to these regions are indeed safe.
17919 If the input program contains a function declaration such as:
17925 then the implementation of @code{foo} must allow @code{j foo} and
17926 @code{jal foo} to be executed speculatively. GCC honors this
17927 restriction for functions it compiles itself. It expects non-GCC
17928 functions (such as hand-written assembly code) to do the same.
17930 The option has three forms:
17933 @item -mr10k-cache-barrier=load-store
17934 Insert a cache barrier before a load or store that might be
17935 speculatively executed and that might have side effects even
17938 @item -mr10k-cache-barrier=store
17939 Insert a cache barrier before a store that might be speculatively
17940 executed and that might have side effects even if aborted.
17942 @item -mr10k-cache-barrier=none
17943 Disable the insertion of cache barriers. This is the default setting.
17946 @item -mflush-func=@var{func}
17947 @itemx -mno-flush-func
17948 @opindex mflush-func
17949 Specifies the function to call to flush the I and D caches, or to not
17950 call any such function. If called, the function must take the same
17951 arguments as the common @code{_flush_func}, that is, the address of the
17952 memory range for which the cache is being flushed, the size of the
17953 memory range, and the number 3 (to flush both caches). The default
17954 depends on the target GCC was configured for, but commonly is either
17955 @code{_flush_func} or @code{__cpu_flush}.
17957 @item mbranch-cost=@var{num}
17958 @opindex mbranch-cost
17959 Set the cost of branches to roughly @var{num} ``simple'' instructions.
17960 This cost is only a heuristic and is not guaranteed to produce
17961 consistent results across releases. A zero cost redundantly selects
17962 the default, which is based on the @option{-mtune} setting.
17964 @item -mbranch-likely
17965 @itemx -mno-branch-likely
17966 @opindex mbranch-likely
17967 @opindex mno-branch-likely
17968 Enable or disable use of Branch Likely instructions, regardless of the
17969 default for the selected architecture. By default, Branch Likely
17970 instructions may be generated if they are supported by the selected
17971 architecture. An exception is for the MIPS32 and MIPS64 architectures
17972 and processors that implement those architectures; for those, Branch
17973 Likely instructions are not be generated by default because the MIPS32
17974 and MIPS64 architectures specifically deprecate their use.
17976 @item -mfp-exceptions
17977 @itemx -mno-fp-exceptions
17978 @opindex mfp-exceptions
17979 Specifies whether FP exceptions are enabled. This affects how
17980 FP instructions are scheduled for some processors.
17981 The default is that FP exceptions are
17984 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
17985 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
17988 @item -mvr4130-align
17989 @itemx -mno-vr4130-align
17990 @opindex mvr4130-align
17991 The VR4130 pipeline is two-way superscalar, but can only issue two
17992 instructions together if the first one is 8-byte aligned. When this
17993 option is enabled, GCC aligns pairs of instructions that it
17994 thinks should execute in parallel.
17996 This option only has an effect when optimizing for the VR4130.
17997 It normally makes code faster, but at the expense of making it bigger.
17998 It is enabled by default at optimization level @option{-O3}.
18003 Enable (disable) generation of @code{synci} instructions on
18004 architectures that support it. The @code{synci} instructions (if
18005 enabled) are generated when @code{__builtin___clear_cache} is
18008 This option defaults to @option{-mno-synci}, but the default can be
18009 overridden by configuring GCC with @option{--with-synci}.
18011 When compiling code for single processor systems, it is generally safe
18012 to use @code{synci}. However, on many multi-core (SMP) systems, it
18013 does not invalidate the instruction caches on all cores and may lead
18014 to undefined behavior.
18016 @item -mrelax-pic-calls
18017 @itemx -mno-relax-pic-calls
18018 @opindex mrelax-pic-calls
18019 Try to turn PIC calls that are normally dispatched via register
18020 @code{$25} into direct calls. This is only possible if the linker can
18021 resolve the destination at link-time and if the destination is within
18022 range for a direct call.
18024 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18025 an assembler and a linker that support the @code{.reloc} assembly
18026 directive and @option{-mexplicit-relocs} is in effect. With
18027 @option{-mno-explicit-relocs}, this optimization can be performed by the
18028 assembler and the linker alone without help from the compiler.
18030 @item -mmcount-ra-address
18031 @itemx -mno-mcount-ra-address
18032 @opindex mmcount-ra-address
18033 @opindex mno-mcount-ra-address
18034 Emit (do not emit) code that allows @code{_mcount} to modify the
18035 calling function's return address. When enabled, this option extends
18036 the usual @code{_mcount} interface with a new @var{ra-address}
18037 parameter, which has type @code{intptr_t *} and is passed in register
18038 @code{$12}. @code{_mcount} can then modify the return address by
18039 doing both of the following:
18042 Returning the new address in register @code{$31}.
18044 Storing the new address in @code{*@var{ra-address}},
18045 if @var{ra-address} is nonnull.
18048 The default is @option{-mno-mcount-ra-address}.
18053 @subsection MMIX Options
18054 @cindex MMIX Options
18056 These options are defined for the MMIX:
18060 @itemx -mno-libfuncs
18062 @opindex mno-libfuncs
18063 Specify that intrinsic library functions are being compiled, passing all
18064 values in registers, no matter the size.
18067 @itemx -mno-epsilon
18069 @opindex mno-epsilon
18070 Generate floating-point comparison instructions that compare with respect
18071 to the @code{rE} epsilon register.
18073 @item -mabi=mmixware
18075 @opindex mabi=mmixware
18077 Generate code that passes function parameters and return values that (in
18078 the called function) are seen as registers @code{$0} and up, as opposed to
18079 the GNU ABI which uses global registers @code{$231} and up.
18081 @item -mzero-extend
18082 @itemx -mno-zero-extend
18083 @opindex mzero-extend
18084 @opindex mno-zero-extend
18085 When reading data from memory in sizes shorter than 64 bits, use (do not
18086 use) zero-extending load instructions by default, rather than
18087 sign-extending ones.
18090 @itemx -mno-knuthdiv
18092 @opindex mno-knuthdiv
18093 Make the result of a division yielding a remainder have the same sign as
18094 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18095 remainder follows the sign of the dividend. Both methods are
18096 arithmetically valid, the latter being almost exclusively used.
18098 @item -mtoplevel-symbols
18099 @itemx -mno-toplevel-symbols
18100 @opindex mtoplevel-symbols
18101 @opindex mno-toplevel-symbols
18102 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18103 code can be used with the @code{PREFIX} assembly directive.
18107 Generate an executable in the ELF format, rather than the default
18108 @samp{mmo} format used by the @command{mmix} simulator.
18110 @item -mbranch-predict
18111 @itemx -mno-branch-predict
18112 @opindex mbranch-predict
18113 @opindex mno-branch-predict
18114 Use (do not use) the probable-branch instructions, when static branch
18115 prediction indicates a probable branch.
18117 @item -mbase-addresses
18118 @itemx -mno-base-addresses
18119 @opindex mbase-addresses
18120 @opindex mno-base-addresses
18121 Generate (do not generate) code that uses @emph{base addresses}. Using a
18122 base address automatically generates a request (handled by the assembler
18123 and the linker) for a constant to be set up in a global register. The
18124 register is used for one or more base address requests within the range 0
18125 to 255 from the value held in the register. The generally leads to short
18126 and fast code, but the number of different data items that can be
18127 addressed is limited. This means that a program that uses lots of static
18128 data may require @option{-mno-base-addresses}.
18130 @item -msingle-exit
18131 @itemx -mno-single-exit
18132 @opindex msingle-exit
18133 @opindex mno-single-exit
18134 Force (do not force) generated code to have a single exit point in each
18138 @node MN10300 Options
18139 @subsection MN10300 Options
18140 @cindex MN10300 options
18142 These @option{-m} options are defined for Matsushita MN10300 architectures:
18147 Generate code to avoid bugs in the multiply instructions for the MN10300
18148 processors. This is the default.
18150 @item -mno-mult-bug
18151 @opindex mno-mult-bug
18152 Do not generate code to avoid bugs in the multiply instructions for the
18153 MN10300 processors.
18157 Generate code using features specific to the AM33 processor.
18161 Do not generate code using features specific to the AM33 processor. This
18166 Generate code using features specific to the AM33/2.0 processor.
18170 Generate code using features specific to the AM34 processor.
18172 @item -mtune=@var{cpu-type}
18174 Use the timing characteristics of the indicated CPU type when
18175 scheduling instructions. This does not change the targeted processor
18176 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18177 @samp{am33-2} or @samp{am34}.
18179 @item -mreturn-pointer-on-d0
18180 @opindex mreturn-pointer-on-d0
18181 When generating a function that returns a pointer, return the pointer
18182 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18183 only in @code{a0}, and attempts to call such functions without a prototype
18184 result in errors. Note that this option is on by default; use
18185 @option{-mno-return-pointer-on-d0} to disable it.
18189 Do not link in the C run-time initialization object file.
18193 Indicate to the linker that it should perform a relaxation optimization pass
18194 to shorten branches, calls and absolute memory addresses. This option only
18195 has an effect when used on the command line for the final link step.
18197 This option makes symbolic debugging impossible.
18201 Allow the compiler to generate @emph{Long Instruction Word}
18202 instructions if the target is the @samp{AM33} or later. This is the
18203 default. This option defines the preprocessor macro @code{__LIW__}.
18207 Do not allow the compiler to generate @emph{Long Instruction Word}
18208 instructions. This option defines the preprocessor macro
18213 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18214 instructions if the target is the @samp{AM33} or later. This is the
18215 default. This option defines the preprocessor macro @code{__SETLB__}.
18219 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18220 instructions. This option defines the preprocessor macro
18221 @code{__NO_SETLB__}.
18225 @node Moxie Options
18226 @subsection Moxie Options
18227 @cindex Moxie Options
18233 Generate big-endian code. This is the default for @samp{moxie-*-*}
18238 Generate little-endian code.
18242 Generate mul.x and umul.x instructions. This is the default for
18243 @samp{moxiebox-*-*} configurations.
18247 Do not link in the C run-time initialization object file.
18251 @node MSP430 Options
18252 @subsection MSP430 Options
18253 @cindex MSP430 Options
18255 These options are defined for the MSP430:
18261 Force assembly output to always use hex constants. Normally such
18262 constants are signed decimals, but this option is available for
18263 testsuite and/or aesthetic purposes.
18267 Select the MCU to target. This is used to create a C preprocessor
18268 symbol based upon the MCU name, converted to upper case and pre- and
18269 post-fixed with @samp{__}. This in turn is used by the
18270 @file{msp430.h} header file to select an MCU-specific supplementary
18273 The option also sets the ISA to use. If the MCU name is one that is
18274 known to only support the 430 ISA then that is selected, otherwise the
18275 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18276 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18277 name selects the 430X ISA.
18279 In addition an MCU-specific linker script is added to the linker
18280 command line. The script's name is the name of the MCU with
18281 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18282 command line defines the C preprocessor symbol @code{__XXX__} and
18283 cause the linker to search for a script called @file{xxx.ld}.
18285 This option is also passed on to the assembler.
18289 Specifies the ISA to use. Accepted values are @samp{msp430},
18290 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18291 @option{-mmcu=} option should be used to select the ISA.
18295 Link to the simulator runtime libraries and linker script. Overrides
18296 any scripts that would be selected by the @option{-mmcu=} option.
18300 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18304 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18308 This option is passed to the assembler and linker, and allows the
18309 linker to perform certain optimizations that cannot be done until
18314 Describes the type of hardware multiply supported by the target.
18315 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18316 for the original 16-bit-only multiply supported by early MCUs.
18317 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18318 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18319 A value of @samp{auto} can also be given. This tells GCC to deduce
18320 the hardware multiply support based upon the MCU name provided by the
18321 @option{-mmcu} option. If no @option{-mmcu} option is specified then
18322 @samp{32bit} hardware multiply support is assumed. @samp{auto} is the
18325 Hardware multiplies are normally performed by calling a library
18326 routine. This saves space in the generated code. When compiling at
18327 @option{-O3} or higher however the hardware multiplier is invoked
18328 inline. This makes for bigger, but faster code.
18330 The hardware multiply routines disable interrupts whilst running and
18331 restore the previous interrupt state when they finish. This makes
18332 them safe to use inside interrupt handlers as well as in normal code.
18336 Enable the use of a minimum runtime environment - no static
18337 initializers or constructors. This is intended for memory-constrained
18338 devices. The compiler includes special symbols in some objects
18339 that tell the linker and runtime which code fragments are required.
18341 @item -mcode-region=
18342 @itemx -mdata-region=
18343 @opindex mcode-region
18344 @opindex mdata-region
18345 These options tell the compiler where to place functions and data that
18346 do not have one of the @code{lower}, @code{upper}, @code{either} or
18347 @code{section} attributes. Possible values are @code{lower},
18348 @code{upper}, @code{either} or @code{any}. The first three behave
18349 like the corresponding attribute. The fourth possible value -
18350 @code{any} - is the default. It leaves placement entirely up to the
18351 linker script and how it assigns the standard sections (.text, .data
18352 etc) to the memory regions.
18356 @node NDS32 Options
18357 @subsection NDS32 Options
18358 @cindex NDS32 Options
18360 These options are defined for NDS32 implementations:
18365 @opindex mbig-endian
18366 Generate code in big-endian mode.
18368 @item -mlittle-endian
18369 @opindex mlittle-endian
18370 Generate code in little-endian mode.
18372 @item -mreduced-regs
18373 @opindex mreduced-regs
18374 Use reduced-set registers for register allocation.
18377 @opindex mfull-regs
18378 Use full-set registers for register allocation.
18382 Generate conditional move instructions.
18386 Do not generate conditional move instructions.
18390 Generate performance extension instructions.
18392 @item -mno-perf-ext
18393 @opindex mno-perf-ext
18394 Do not generate performance extension instructions.
18398 Generate v3 push25/pop25 instructions.
18401 @opindex mno-v3push
18402 Do not generate v3 push25/pop25 instructions.
18406 Generate 16-bit instructions.
18409 @opindex mno-16-bit
18410 Do not generate 16-bit instructions.
18412 @item -misr-vector-size=@var{num}
18413 @opindex misr-vector-size
18414 Specify the size of each interrupt vector, which must be 4 or 16.
18416 @item -mcache-block-size=@var{num}
18417 @opindex mcache-block-size
18418 Specify the size of each cache block,
18419 which must be a power of 2 between 4 and 512.
18421 @item -march=@var{arch}
18423 Specify the name of the target architecture.
18425 @item -mcmodel=@var{code-model}
18427 Set the code model to one of
18430 All the data and read-only data segments must be within 512KB addressing space.
18431 The text segment must be within 16MB addressing space.
18432 @item @samp{medium}
18433 The data segment must be within 512KB while the read-only data segment can be
18434 within 4GB addressing space. The text segment should be still within 16MB
18437 All the text and data segments can be within 4GB addressing space.
18441 @opindex mctor-dtor
18442 Enable constructor/destructor feature.
18446 Guide linker to relax instructions.
18450 @node Nios II Options
18451 @subsection Nios II Options
18452 @cindex Nios II options
18453 @cindex Altera Nios II options
18455 These are the options defined for the Altera Nios II processor.
18461 @cindex smaller data references
18462 Put global and static objects less than or equal to @var{num} bytes
18463 into the small data or BSS sections instead of the normal data or BSS
18464 sections. The default value of @var{num} is 8.
18466 @item -mgpopt=@var{option}
18471 Generate (do not generate) GP-relative accesses. The following
18472 @var{option} names are recognized:
18477 Do not generate GP-relative accesses.
18480 Generate GP-relative accesses for small data objects that are not
18481 external or weak. Also use GP-relative addressing for objects that
18482 have been explicitly placed in a small data section via a @code{section}
18486 As for @samp{local}, but also generate GP-relative accesses for
18487 small data objects that are external or weak. If you use this option,
18488 you must ensure that all parts of your program (including libraries) are
18489 compiled with the same @option{-G} setting.
18492 Generate GP-relative accesses for all data objects in the program. If you
18493 use this option, the entire data and BSS segments
18494 of your program must fit in 64K of memory and you must use an appropriate
18495 linker script to allocate them within the addressible range of the
18499 Generate GP-relative addresses for function pointers as well as data
18500 pointers. If you use this option, the entire text, data, and BSS segments
18501 of your program must fit in 64K of memory and you must use an appropriate
18502 linker script to allocate them within the addressible range of the
18507 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18508 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18510 The default is @option{-mgpopt} except when @option{-fpic} or
18511 @option{-fPIC} is specified to generate position-independent code.
18512 Note that the Nios II ABI does not permit GP-relative accesses from
18515 You may need to specify @option{-mno-gpopt} explicitly when building
18516 programs that include large amounts of small data, including large
18517 GOT data sections. In this case, the 16-bit offset for GP-relative
18518 addressing may not be large enough to allow access to the entire
18519 small data section.
18525 Generate little-endian (default) or big-endian (experimental) code,
18528 @item -march=@var{arch}
18530 This specifies the name of the target Nios II architecture. GCC uses this
18531 name to determine what kind of instructions it can emit when generating
18532 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
18534 The preprocessor macro @code{__nios2_arch__} is available to programs,
18535 with value 1 or 2, indicating the targeted ISA level.
18537 @item -mbypass-cache
18538 @itemx -mno-bypass-cache
18539 @opindex mno-bypass-cache
18540 @opindex mbypass-cache
18541 Force all load and store instructions to always bypass cache by
18542 using I/O variants of the instructions. The default is not to
18545 @item -mno-cache-volatile
18546 @itemx -mcache-volatile
18547 @opindex mcache-volatile
18548 @opindex mno-cache-volatile
18549 Volatile memory access bypass the cache using the I/O variants of
18550 the load and store instructions. The default is not to bypass the cache.
18552 @item -mno-fast-sw-div
18553 @itemx -mfast-sw-div
18554 @opindex mno-fast-sw-div
18555 @opindex mfast-sw-div
18556 Do not use table-based fast divide for small numbers. The default
18557 is to use the fast divide at @option{-O3} and above.
18561 @itemx -mno-hw-mulx
18565 @opindex mno-hw-mul
18567 @opindex mno-hw-mulx
18569 @opindex mno-hw-div
18571 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18572 instructions by the compiler. The default is to emit @code{mul}
18573 and not emit @code{div} and @code{mulx}.
18579 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
18580 CDX (code density) instructions. Enabling these instructions also
18581 requires @option{-march=r2}. Since these instructions are optional
18582 extensions to the R2 architecture, the default is not to emit them.
18584 @item -mcustom-@var{insn}=@var{N}
18585 @itemx -mno-custom-@var{insn}
18586 @opindex mcustom-@var{insn}
18587 @opindex mno-custom-@var{insn}
18588 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18589 custom instruction with encoding @var{N} when generating code that uses
18590 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18591 instruction 253 for single-precision floating-point add operations instead
18592 of the default behavior of using a library call.
18594 The following values of @var{insn} are supported. Except as otherwise
18595 noted, floating-point operations are expected to be implemented with
18596 normal IEEE 754 semantics and correspond directly to the C operators or the
18597 equivalent GCC built-in functions (@pxref{Other Builtins}).
18599 Single-precision floating point:
18602 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18603 Binary arithmetic operations.
18609 Unary absolute value.
18611 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18612 Comparison operations.
18614 @item @samp{fmins}, @samp{fmaxs}
18615 Floating-point minimum and maximum. These instructions are only
18616 generated if @option{-ffinite-math-only} is specified.
18618 @item @samp{fsqrts}
18619 Unary square root operation.
18621 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18622 Floating-point trigonometric and exponential functions. These instructions
18623 are only generated if @option{-funsafe-math-optimizations} is also specified.
18627 Double-precision floating point:
18630 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18631 Binary arithmetic operations.
18637 Unary absolute value.
18639 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18640 Comparison operations.
18642 @item @samp{fmind}, @samp{fmaxd}
18643 Double-precision minimum and maximum. These instructions are only
18644 generated if @option{-ffinite-math-only} is specified.
18646 @item @samp{fsqrtd}
18647 Unary square root operation.
18649 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18650 Double-precision trigonometric and exponential functions. These instructions
18651 are only generated if @option{-funsafe-math-optimizations} is also specified.
18657 @item @samp{fextsd}
18658 Conversion from single precision to double precision.
18660 @item @samp{ftruncds}
18661 Conversion from double precision to single precision.
18663 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18664 Conversion from floating point to signed or unsigned integer types, with
18665 truncation towards zero.
18668 Conversion from single-precision floating point to signed integer,
18669 rounding to the nearest integer and ties away from zero.
18670 This corresponds to the @code{__builtin_lroundf} function when
18671 @option{-fno-math-errno} is used.
18673 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18674 Conversion from signed or unsigned integer types to floating-point types.
18678 In addition, all of the following transfer instructions for internal
18679 registers X and Y must be provided to use any of the double-precision
18680 floating-point instructions. Custom instructions taking two
18681 double-precision source operands expect the first operand in the
18682 64-bit register X. The other operand (or only operand of a unary
18683 operation) is given to the custom arithmetic instruction with the
18684 least significant half in source register @var{src1} and the most
18685 significant half in @var{src2}. A custom instruction that returns a
18686 double-precision result returns the most significant 32 bits in the
18687 destination register and the other half in 32-bit register Y.
18688 GCC automatically generates the necessary code sequences to write
18689 register X and/or read register Y when double-precision floating-point
18690 instructions are used.
18695 Write @var{src1} into the least significant half of X and @var{src2} into
18696 the most significant half of X.
18699 Write @var{src1} into Y.
18701 @item @samp{frdxhi}, @samp{frdxlo}
18702 Read the most or least (respectively) significant half of X and store it in
18706 Read the value of Y and store it into @var{dest}.
18709 Note that you can gain more local control over generation of Nios II custom
18710 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18711 and @code{target("no-custom-@var{insn}")} function attributes
18712 (@pxref{Function Attributes})
18713 or pragmas (@pxref{Function Specific Option Pragmas}).
18715 @item -mcustom-fpu-cfg=@var{name}
18716 @opindex mcustom-fpu-cfg
18718 This option enables a predefined, named set of custom instruction encodings
18719 (see @option{-mcustom-@var{insn}} above).
18720 Currently, the following sets are defined:
18722 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18723 @gccoptlist{-mcustom-fmuls=252 @gol
18724 -mcustom-fadds=253 @gol
18725 -mcustom-fsubs=254 @gol
18726 -fsingle-precision-constant}
18728 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18729 @gccoptlist{-mcustom-fmuls=252 @gol
18730 -mcustom-fadds=253 @gol
18731 -mcustom-fsubs=254 @gol
18732 -mcustom-fdivs=255 @gol
18733 -fsingle-precision-constant}
18735 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
18736 @gccoptlist{-mcustom-floatus=243 @gol
18737 -mcustom-fixsi=244 @gol
18738 -mcustom-floatis=245 @gol
18739 -mcustom-fcmpgts=246 @gol
18740 -mcustom-fcmples=249 @gol
18741 -mcustom-fcmpeqs=250 @gol
18742 -mcustom-fcmpnes=251 @gol
18743 -mcustom-fmuls=252 @gol
18744 -mcustom-fadds=253 @gol
18745 -mcustom-fsubs=254 @gol
18746 -mcustom-fdivs=255 @gol
18747 -fsingle-precision-constant}
18749 Custom instruction assignments given by individual
18750 @option{-mcustom-@var{insn}=} options override those given by
18751 @option{-mcustom-fpu-cfg=}, regardless of the
18752 order of the options on the command line.
18754 Note that you can gain more local control over selection of a FPU
18755 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
18756 function attribute (@pxref{Function Attributes})
18757 or pragma (@pxref{Function Specific Option Pragmas}).
18761 These additional @samp{-m} options are available for the Altera Nios II
18762 ELF (bare-metal) target:
18768 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
18769 startup and termination code, and is typically used in conjunction with
18770 @option{-msys-crt0=} to specify the location of the alternate startup code
18771 provided by the HAL BSP.
18775 Link with a limited version of the C library, @option{-lsmallc}, rather than
18778 @item -msys-crt0=@var{startfile}
18780 @var{startfile} is the file name of the startfile (crt0) to use
18781 when linking. This option is only useful in conjunction with @option{-mhal}.
18783 @item -msys-lib=@var{systemlib}
18785 @var{systemlib} is the library name of the library that provides
18786 low-level system calls required by the C library,
18787 e.g. @code{read} and @code{write}.
18788 This option is typically used to link with a library provided by a HAL BSP.
18792 @node Nvidia PTX Options
18793 @subsection Nvidia PTX Options
18794 @cindex Nvidia PTX options
18795 @cindex nvptx options
18797 These options are defined for Nvidia PTX:
18805 Generate code for 32-bit or 64-bit ABI.
18808 @opindex mmainkernel
18809 Link in code for a __main kernel. This is for stand-alone instead of
18810 offloading execution.
18814 @node PDP-11 Options
18815 @subsection PDP-11 Options
18816 @cindex PDP-11 Options
18818 These options are defined for the PDP-11:
18823 Use hardware FPP floating point. This is the default. (FIS floating
18824 point on the PDP-11/40 is not supported.)
18827 @opindex msoft-float
18828 Do not use hardware floating point.
18832 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
18836 Return floating-point results in memory. This is the default.
18840 Generate code for a PDP-11/40.
18844 Generate code for a PDP-11/45. This is the default.
18848 Generate code for a PDP-11/10.
18850 @item -mbcopy-builtin
18851 @opindex mbcopy-builtin
18852 Use inline @code{movmemhi} patterns for copying memory. This is the
18857 Do not use inline @code{movmemhi} patterns for copying memory.
18863 Use 16-bit @code{int}. This is the default.
18869 Use 32-bit @code{int}.
18872 @itemx -mno-float32
18874 @opindex mno-float32
18875 Use 64-bit @code{float}. This is the default.
18878 @itemx -mno-float64
18880 @opindex mno-float64
18881 Use 32-bit @code{float}.
18885 Use @code{abshi2} pattern. This is the default.
18889 Do not use @code{abshi2} pattern.
18891 @item -mbranch-expensive
18892 @opindex mbranch-expensive
18893 Pretend that branches are expensive. This is for experimenting with
18894 code generation only.
18896 @item -mbranch-cheap
18897 @opindex mbranch-cheap
18898 Do not pretend that branches are expensive. This is the default.
18902 Use Unix assembler syntax. This is the default when configured for
18903 @samp{pdp11-*-bsd}.
18907 Use DEC assembler syntax. This is the default when configured for any
18908 PDP-11 target other than @samp{pdp11-*-bsd}.
18911 @node picoChip Options
18912 @subsection picoChip Options
18913 @cindex picoChip options
18915 These @samp{-m} options are defined for picoChip implementations:
18919 @item -mae=@var{ae_type}
18921 Set the instruction set, register set, and instruction scheduling
18922 parameters for array element type @var{ae_type}. Supported values
18923 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
18925 @option{-mae=ANY} selects a completely generic AE type. Code
18926 generated with this option runs on any of the other AE types. The
18927 code is not as efficient as it would be if compiled for a specific
18928 AE type, and some types of operation (e.g., multiplication) do not
18929 work properly on all types of AE.
18931 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
18932 for compiled code, and is the default.
18934 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
18935 option may suffer from poor performance of byte (char) manipulation,
18936 since the DSP AE does not provide hardware support for byte load/stores.
18938 @item -msymbol-as-address
18939 Enable the compiler to directly use a symbol name as an address in a
18940 load/store instruction, without first loading it into a
18941 register. Typically, the use of this option generates larger
18942 programs, which run faster than when the option isn't used. However, the
18943 results vary from program to program, so it is left as a user option,
18944 rather than being permanently enabled.
18946 @item -mno-inefficient-warnings
18947 Disables warnings about the generation of inefficient code. These
18948 warnings can be generated, for example, when compiling code that
18949 performs byte-level memory operations on the MAC AE type. The MAC AE has
18950 no hardware support for byte-level memory operations, so all byte
18951 load/stores must be synthesized from word load/store operations. This is
18952 inefficient and a warning is generated to indicate
18953 that you should rewrite the code to avoid byte operations, or to target
18954 an AE type that has the necessary hardware support. This option disables
18959 @node PowerPC Options
18960 @subsection PowerPC Options
18961 @cindex PowerPC options
18963 These are listed under @xref{RS/6000 and PowerPC Options}.
18966 @subsection RL78 Options
18967 @cindex RL78 Options
18973 Links in additional target libraries to support operation within a
18982 Specifies the type of hardware multiplication and division support to
18983 be used. The simplest is @code{none}, which uses software for both
18984 multiplication and division. This is the default. The @code{g13}
18985 value is for the hardware multiply/divide peripheral found on the
18986 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
18987 the multiplication and division instructions supported by the RL78/G14
18988 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
18989 the value @code{mg10} is an alias for @code{none}.
18991 In addition a C preprocessor macro is defined, based upon the setting
18992 of this option. Possible values are: @code{__RL78_MUL_NONE__},
18993 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19000 Specifies the RL78 core to target. The default is the G14 core, also
19001 known as an S3 core or just RL78. The G13 or S2 core does not have
19002 multiply or divide instructions, instead it uses a hardware peripheral
19003 for these operations. The G10 or S1 core does not have register
19004 banks, so it uses a different calling convention.
19006 If this option is set it also selects the type of hardware multiply
19007 support to use, unless this is overridden by an explicit
19008 @option{-mmul=none} option on the command line. Thus specifying
19009 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19010 peripheral and specifying @option{-mcpu=g10} disables the use of
19011 hardware multipications altogether.
19013 Note, although the RL78/G14 core is the default target, specifying
19014 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19015 change the behaviour of the toolchain since it also enables G14
19016 hardware multiply support. If these options are not specified on the
19017 command line then software multiplication routines will be used even
19018 though the code targets the RL78 core. This is for backwards
19019 compatibility with older toolchains which did not have hardware
19020 multiply and divide support.
19022 In addition a C preprocessor macro is defined, based upon the setting
19023 of this option. Possible values are: @code{__RL78_G10__},
19024 @code{__RL78_G13__} or @code{__RL78_G14__}.
19034 These are aliases for the corresponding @option{-mcpu=} option. They
19035 are provided for backwards compatibility.
19039 Allow the compiler to use all of the available registers. By default
19040 registers @code{r24..r31} are reserved for use in interrupt handlers.
19041 With this option enabled these registers can be used in ordinary
19044 @item -m64bit-doubles
19045 @itemx -m32bit-doubles
19046 @opindex m64bit-doubles
19047 @opindex m32bit-doubles
19048 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19049 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19050 @option{-m32bit-doubles}.
19054 @node RS/6000 and PowerPC Options
19055 @subsection IBM RS/6000 and PowerPC Options
19056 @cindex RS/6000 and PowerPC Options
19057 @cindex IBM RS/6000 and PowerPC Options
19059 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19061 @item -mpowerpc-gpopt
19062 @itemx -mno-powerpc-gpopt
19063 @itemx -mpowerpc-gfxopt
19064 @itemx -mno-powerpc-gfxopt
19067 @itemx -mno-powerpc64
19071 @itemx -mno-popcntb
19073 @itemx -mno-popcntd
19082 @itemx -mno-hard-dfp
19083 @opindex mpowerpc-gpopt
19084 @opindex mno-powerpc-gpopt
19085 @opindex mpowerpc-gfxopt
19086 @opindex mno-powerpc-gfxopt
19087 @opindex mpowerpc64
19088 @opindex mno-powerpc64
19092 @opindex mno-popcntb
19094 @opindex mno-popcntd
19100 @opindex mno-mfpgpr
19102 @opindex mno-hard-dfp
19103 You use these options to specify which instructions are available on the
19104 processor you are using. The default value of these options is
19105 determined when configuring GCC@. Specifying the
19106 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19107 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19108 rather than the options listed above.
19110 Specifying @option{-mpowerpc-gpopt} allows
19111 GCC to use the optional PowerPC architecture instructions in the
19112 General Purpose group, including floating-point square root. Specifying
19113 @option{-mpowerpc-gfxopt} allows GCC to
19114 use the optional PowerPC architecture instructions in the Graphics
19115 group, including floating-point select.
19117 The @option{-mmfcrf} option allows GCC to generate the move from
19118 condition register field instruction implemented on the POWER4
19119 processor and other processors that support the PowerPC V2.01
19121 The @option{-mpopcntb} option allows GCC to generate the popcount and
19122 double-precision FP reciprocal estimate instruction implemented on the
19123 POWER5 processor and other processors that support the PowerPC V2.02
19125 The @option{-mpopcntd} option allows GCC to generate the popcount
19126 instruction implemented on the POWER7 processor and other processors
19127 that support the PowerPC V2.06 architecture.
19128 The @option{-mfprnd} option allows GCC to generate the FP round to
19129 integer instructions implemented on the POWER5+ processor and other
19130 processors that support the PowerPC V2.03 architecture.
19131 The @option{-mcmpb} option allows GCC to generate the compare bytes
19132 instruction implemented on the POWER6 processor and other processors
19133 that support the PowerPC V2.05 architecture.
19134 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19135 general-purpose register instructions implemented on the POWER6X
19136 processor and other processors that support the extended PowerPC V2.05
19138 The @option{-mhard-dfp} option allows GCC to generate the decimal
19139 floating-point instructions implemented on some POWER processors.
19141 The @option{-mpowerpc64} option allows GCC to generate the additional
19142 64-bit instructions that are found in the full PowerPC64 architecture
19143 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19144 @option{-mno-powerpc64}.
19146 @item -mcpu=@var{cpu_type}
19148 Set architecture type, register usage, and
19149 instruction scheduling parameters for machine type @var{cpu_type}.
19150 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19151 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19152 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19153 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19154 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19155 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19156 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19157 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19158 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19159 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8}, @samp{powerpc},
19160 @samp{powerpc64}, @samp{powerpc64le}, and @samp{rs64}.
19162 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19163 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19164 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19165 architecture machine types, with an appropriate, generic processor
19166 model assumed for scheduling purposes.
19168 The other options specify a specific processor. Code generated under
19169 those options runs best on that processor, and may not run at all on
19172 The @option{-mcpu} options automatically enable or disable the
19175 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19176 -mpopcntb -mpopcntd -mpowerpc64 @gol
19177 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19178 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19179 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19180 -mquad-memory -mquad-memory-atomic}
19182 The particular options set for any particular CPU varies between
19183 compiler versions, depending on what setting seems to produce optimal
19184 code for that CPU; it doesn't necessarily reflect the actual hardware's
19185 capabilities. If you wish to set an individual option to a particular
19186 value, you may specify it after the @option{-mcpu} option, like
19187 @option{-mcpu=970 -mno-altivec}.
19189 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19190 not enabled or disabled by the @option{-mcpu} option at present because
19191 AIX does not have full support for these options. You may still
19192 enable or disable them individually if you're sure it'll work in your
19195 @item -mtune=@var{cpu_type}
19197 Set the instruction scheduling parameters for machine type
19198 @var{cpu_type}, but do not set the architecture type or register usage,
19199 as @option{-mcpu=@var{cpu_type}} does. The same
19200 values for @var{cpu_type} are used for @option{-mtune} as for
19201 @option{-mcpu}. If both are specified, the code generated uses the
19202 architecture and registers set by @option{-mcpu}, but the
19203 scheduling parameters set by @option{-mtune}.
19205 @item -mcmodel=small
19206 @opindex mcmodel=small
19207 Generate PowerPC64 code for the small model: The TOC is limited to
19210 @item -mcmodel=medium
19211 @opindex mcmodel=medium
19212 Generate PowerPC64 code for the medium model: The TOC and other static
19213 data may be up to a total of 4G in size.
19215 @item -mcmodel=large
19216 @opindex mcmodel=large
19217 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19218 in size. Other data and code is only limited by the 64-bit address
19222 @itemx -mno-altivec
19224 @opindex mno-altivec
19225 Generate code that uses (does not use) AltiVec instructions, and also
19226 enable the use of built-in functions that allow more direct access to
19227 the AltiVec instruction set. You may also need to set
19228 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19231 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19232 @option{-maltivec=be}, the element order for Altivec intrinsics such
19233 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19234 match array element order corresponding to the endianness of the
19235 target. That is, element zero identifies the leftmost element in a
19236 vector register when targeting a big-endian platform, and identifies
19237 the rightmost element in a vector register when targeting a
19238 little-endian platform.
19241 @opindex maltivec=be
19242 Generate Altivec instructions using big-endian element order,
19243 regardless of whether the target is big- or little-endian. This is
19244 the default when targeting a big-endian platform.
19246 The element order is used to interpret element numbers in Altivec
19247 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19248 @code{vec_insert}. By default, these match array element order
19249 corresponding to the endianness for the target.
19252 @opindex maltivec=le
19253 Generate Altivec instructions using little-endian element order,
19254 regardless of whether the target is big- or little-endian. This is
19255 the default when targeting a little-endian platform. This option is
19256 currently ignored when targeting a big-endian platform.
19258 The element order is used to interpret element numbers in Altivec
19259 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19260 @code{vec_insert}. By default, these match array element order
19261 corresponding to the endianness for the target.
19266 @opindex mno-vrsave
19267 Generate VRSAVE instructions when generating AltiVec code.
19269 @item -mgen-cell-microcode
19270 @opindex mgen-cell-microcode
19271 Generate Cell microcode instructions.
19273 @item -mwarn-cell-microcode
19274 @opindex mwarn-cell-microcode
19275 Warn when a Cell microcode instruction is emitted. An example
19276 of a Cell microcode instruction is a variable shift.
19279 @opindex msecure-plt
19280 Generate code that allows @command{ld} and @command{ld.so}
19281 to build executables and shared
19282 libraries with non-executable @code{.plt} and @code{.got} sections.
19284 32-bit SYSV ABI option.
19288 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19290 requires @code{.plt} and @code{.got}
19291 sections that are both writable and executable.
19292 This is a PowerPC 32-bit SYSV ABI option.
19298 This switch enables or disables the generation of ISEL instructions.
19300 @item -misel=@var{yes/no}
19301 This switch has been deprecated. Use @option{-misel} and
19302 @option{-mno-isel} instead.
19308 This switch enables or disables the generation of SPE simd
19314 @opindex mno-paired
19315 This switch enables or disables the generation of PAIRED simd
19318 @item -mspe=@var{yes/no}
19319 This option has been deprecated. Use @option{-mspe} and
19320 @option{-mno-spe} instead.
19326 Generate code that uses (does not use) vector/scalar (VSX)
19327 instructions, and also enable the use of built-in functions that allow
19328 more direct access to the VSX instruction set.
19333 @opindex mno-crypto
19334 Enable the use (disable) of the built-in functions that allow direct
19335 access to the cryptographic instructions that were added in version
19336 2.07 of the PowerPC ISA.
19338 @item -mdirect-move
19339 @itemx -mno-direct-move
19340 @opindex mdirect-move
19341 @opindex mno-direct-move
19342 Generate code that uses (does not use) the instructions to move data
19343 between the general purpose registers and the vector/scalar (VSX)
19344 registers that were added in version 2.07 of the PowerPC ISA.
19346 @item -mpower8-fusion
19347 @itemx -mno-power8-fusion
19348 @opindex mpower8-fusion
19349 @opindex mno-power8-fusion
19350 Generate code that keeps (does not keeps) some integer operations
19351 adjacent so that the instructions can be fused together on power8 and
19354 @item -mpower8-vector
19355 @itemx -mno-power8-vector
19356 @opindex mpower8-vector
19357 @opindex mno-power8-vector
19358 Generate code that uses (does not use) the vector and scalar
19359 instructions that were added in version 2.07 of the PowerPC ISA. Also
19360 enable the use of built-in functions that allow more direct access to
19361 the vector instructions.
19363 @item -mquad-memory
19364 @itemx -mno-quad-memory
19365 @opindex mquad-memory
19366 @opindex mno-quad-memory
19367 Generate code that uses (does not use) the non-atomic quad word memory
19368 instructions. The @option{-mquad-memory} option requires use of
19371 @item -mquad-memory-atomic
19372 @itemx -mno-quad-memory-atomic
19373 @opindex mquad-memory-atomic
19374 @opindex mno-quad-memory-atomic
19375 Generate code that uses (does not use) the atomic quad word memory
19376 instructions. The @option{-mquad-memory-atomic} option requires use of
19379 @item -mupper-regs-df
19380 @itemx -mno-upper-regs-df
19381 @opindex mupper-regs-df
19382 @opindex mno-upper-regs-df
19383 Generate code that uses (does not use) the scalar double precision
19384 instructions that target all 64 registers in the vector/scalar
19385 floating point register set that were added in version 2.06 of the
19386 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19387 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19388 @option{-mvsx} options.
19390 @item -mupper-regs-sf
19391 @itemx -mno-upper-regs-sf
19392 @opindex mupper-regs-sf
19393 @opindex mno-upper-regs-sf
19394 Generate code that uses (does not use) the scalar single precision
19395 instructions that target all 64 registers in the vector/scalar
19396 floating point register set that were added in version 2.07 of the
19397 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19398 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19402 @itemx -mno-upper-regs
19403 @opindex mupper-regs
19404 @opindex mno-upper-regs
19405 Generate code that uses (does not use) the scalar
19406 instructions that target all 64 registers in the vector/scalar
19407 floating point register set, depending on the model of the machine.
19409 If the @option{-mno-upper-regs} option is used, it turns off both
19410 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19412 @item -mfloat-gprs=@var{yes/single/double/no}
19413 @itemx -mfloat-gprs
19414 @opindex mfloat-gprs
19415 This switch enables or disables the generation of floating-point
19416 operations on the general-purpose registers for architectures that
19419 The argument @samp{yes} or @samp{single} enables the use of
19420 single-precision floating-point operations.
19422 The argument @samp{double} enables the use of single and
19423 double-precision floating-point operations.
19425 The argument @samp{no} disables floating-point operations on the
19426 general-purpose registers.
19428 This option is currently only available on the MPC854x.
19434 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19435 targets (including GNU/Linux). The 32-bit environment sets int, long
19436 and pointer to 32 bits and generates code that runs on any PowerPC
19437 variant. The 64-bit environment sets int to 32 bits and long and
19438 pointer to 64 bits, and generates code for PowerPC64, as for
19439 @option{-mpowerpc64}.
19442 @itemx -mno-fp-in-toc
19443 @itemx -mno-sum-in-toc
19444 @itemx -mminimal-toc
19446 @opindex mno-fp-in-toc
19447 @opindex mno-sum-in-toc
19448 @opindex mminimal-toc
19449 Modify generation of the TOC (Table Of Contents), which is created for
19450 every executable file. The @option{-mfull-toc} option is selected by
19451 default. In that case, GCC allocates at least one TOC entry for
19452 each unique non-automatic variable reference in your program. GCC
19453 also places floating-point constants in the TOC@. However, only
19454 16,384 entries are available in the TOC@.
19456 If you receive a linker error message that saying you have overflowed
19457 the available TOC space, you can reduce the amount of TOC space used
19458 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19459 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19460 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19461 generate code to calculate the sum of an address and a constant at
19462 run time instead of putting that sum into the TOC@. You may specify one
19463 or both of these options. Each causes GCC to produce very slightly
19464 slower and larger code at the expense of conserving TOC space.
19466 If you still run out of space in the TOC even when you specify both of
19467 these options, specify @option{-mminimal-toc} instead. This option causes
19468 GCC to make only one TOC entry for every file. When you specify this
19469 option, GCC produces code that is slower and larger but which
19470 uses extremely little TOC space. You may wish to use this option
19471 only on files that contain less frequently-executed code.
19477 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19478 @code{long} type, and the infrastructure needed to support them.
19479 Specifying @option{-maix64} implies @option{-mpowerpc64},
19480 while @option{-maix32} disables the 64-bit ABI and
19481 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19484 @itemx -mno-xl-compat
19485 @opindex mxl-compat
19486 @opindex mno-xl-compat
19487 Produce code that conforms more closely to IBM XL compiler semantics
19488 when using AIX-compatible ABI@. Pass floating-point arguments to
19489 prototyped functions beyond the register save area (RSA) on the stack
19490 in addition to argument FPRs. Do not assume that most significant
19491 double in 128-bit long double value is properly rounded when comparing
19492 values and converting to double. Use XL symbol names for long double
19495 The AIX calling convention was extended but not initially documented to
19496 handle an obscure K&R C case of calling a function that takes the
19497 address of its arguments with fewer arguments than declared. IBM XL
19498 compilers access floating-point arguments that do not fit in the
19499 RSA from the stack when a subroutine is compiled without
19500 optimization. Because always storing floating-point arguments on the
19501 stack is inefficient and rarely needed, this option is not enabled by
19502 default and only is necessary when calling subroutines compiled by IBM
19503 XL compilers without optimization.
19507 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19508 application written to use message passing with special startup code to
19509 enable the application to run. The system must have PE installed in the
19510 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19511 must be overridden with the @option{-specs=} option to specify the
19512 appropriate directory location. The Parallel Environment does not
19513 support threads, so the @option{-mpe} option and the @option{-pthread}
19514 option are incompatible.
19516 @item -malign-natural
19517 @itemx -malign-power
19518 @opindex malign-natural
19519 @opindex malign-power
19520 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19521 @option{-malign-natural} overrides the ABI-defined alignment of larger
19522 types, such as floating-point doubles, on their natural size-based boundary.
19523 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19524 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19526 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19530 @itemx -mhard-float
19531 @opindex msoft-float
19532 @opindex mhard-float
19533 Generate code that does not use (uses) the floating-point register set.
19534 Software floating-point emulation is provided if you use the
19535 @option{-msoft-float} option, and pass the option to GCC when linking.
19537 @item -msingle-float
19538 @itemx -mdouble-float
19539 @opindex msingle-float
19540 @opindex mdouble-float
19541 Generate code for single- or double-precision floating-point operations.
19542 @option{-mdouble-float} implies @option{-msingle-float}.
19545 @opindex msimple-fpu
19546 Do not generate @code{sqrt} and @code{div} instructions for hardware
19547 floating-point unit.
19549 @item -mfpu=@var{name}
19551 Specify type of floating-point unit. Valid values for @var{name} are
19552 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19553 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19554 @samp{sp_full} (equivalent to @option{-msingle-float}),
19555 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19558 @opindex mxilinx-fpu
19559 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19562 @itemx -mno-multiple
19564 @opindex mno-multiple
19565 Generate code that uses (does not use) the load multiple word
19566 instructions and the store multiple word instructions. These
19567 instructions are generated by default on POWER systems, and not
19568 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19569 PowerPC systems, since those instructions do not work when the
19570 processor is in little-endian mode. The exceptions are PPC740 and
19571 PPC750 which permit these instructions in little-endian mode.
19576 @opindex mno-string
19577 Generate code that uses (does not use) the load string instructions
19578 and the store string word instructions to save multiple registers and
19579 do small block moves. These instructions are generated by default on
19580 POWER systems, and not generated on PowerPC systems. Do not use
19581 @option{-mstring} on little-endian PowerPC systems, since those
19582 instructions do not work when the processor is in little-endian mode.
19583 The exceptions are PPC740 and PPC750 which permit these instructions
19584 in little-endian mode.
19589 @opindex mno-update
19590 Generate code that uses (does not use) the load or store instructions
19591 that update the base register to the address of the calculated memory
19592 location. These instructions are generated by default. If you use
19593 @option{-mno-update}, there is a small window between the time that the
19594 stack pointer is updated and the address of the previous frame is
19595 stored, which means code that walks the stack frame across interrupts or
19596 signals may get corrupted data.
19598 @item -mavoid-indexed-addresses
19599 @itemx -mno-avoid-indexed-addresses
19600 @opindex mavoid-indexed-addresses
19601 @opindex mno-avoid-indexed-addresses
19602 Generate code that tries to avoid (not avoid) the use of indexed load
19603 or store instructions. These instructions can incur a performance
19604 penalty on Power6 processors in certain situations, such as when
19605 stepping through large arrays that cross a 16M boundary. This option
19606 is enabled by default when targeting Power6 and disabled otherwise.
19609 @itemx -mno-fused-madd
19610 @opindex mfused-madd
19611 @opindex mno-fused-madd
19612 Generate code that uses (does not use) the floating-point multiply and
19613 accumulate instructions. These instructions are generated by default
19614 if hardware floating point is used. The machine-dependent
19615 @option{-mfused-madd} option is now mapped to the machine-independent
19616 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19617 mapped to @option{-ffp-contract=off}.
19623 Generate code that uses (does not use) the half-word multiply and
19624 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19625 These instructions are generated by default when targeting those
19632 Generate code that uses (does not use) the string-search @samp{dlmzb}
19633 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19634 generated by default when targeting those processors.
19636 @item -mno-bit-align
19638 @opindex mno-bit-align
19639 @opindex mbit-align
19640 On System V.4 and embedded PowerPC systems do not (do) force structures
19641 and unions that contain bit-fields to be aligned to the base type of the
19644 For example, by default a structure containing nothing but 8
19645 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19646 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19647 the structure is aligned to a 1-byte boundary and is 1 byte in
19650 @item -mno-strict-align
19651 @itemx -mstrict-align
19652 @opindex mno-strict-align
19653 @opindex mstrict-align
19654 On System V.4 and embedded PowerPC systems do not (do) assume that
19655 unaligned memory references are handled by the system.
19657 @item -mrelocatable
19658 @itemx -mno-relocatable
19659 @opindex mrelocatable
19660 @opindex mno-relocatable
19661 Generate code that allows (does not allow) a static executable to be
19662 relocated to a different address at run time. A simple embedded
19663 PowerPC system loader should relocate the entire contents of
19664 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19665 a table of 32-bit addresses generated by this option. For this to
19666 work, all objects linked together must be compiled with
19667 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19668 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19670 @item -mrelocatable-lib
19671 @itemx -mno-relocatable-lib
19672 @opindex mrelocatable-lib
19673 @opindex mno-relocatable-lib
19674 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19675 @code{.fixup} section to allow static executables to be relocated at
19676 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19677 alignment of @option{-mrelocatable}. Objects compiled with
19678 @option{-mrelocatable-lib} may be linked with objects compiled with
19679 any combination of the @option{-mrelocatable} options.
19685 On System V.4 and embedded PowerPC systems do not (do) assume that
19686 register 2 contains a pointer to a global area pointing to the addresses
19687 used in the program.
19690 @itemx -mlittle-endian
19692 @opindex mlittle-endian
19693 On System V.4 and embedded PowerPC systems compile code for the
19694 processor in little-endian mode. The @option{-mlittle-endian} option is
19695 the same as @option{-mlittle}.
19698 @itemx -mbig-endian
19700 @opindex mbig-endian
19701 On System V.4 and embedded PowerPC systems compile code for the
19702 processor in big-endian mode. The @option{-mbig-endian} option is
19703 the same as @option{-mbig}.
19705 @item -mdynamic-no-pic
19706 @opindex mdynamic-no-pic
19707 On Darwin and Mac OS X systems, compile code so that it is not
19708 relocatable, but that its external references are relocatable. The
19709 resulting code is suitable for applications, but not shared
19712 @item -msingle-pic-base
19713 @opindex msingle-pic-base
19714 Treat the register used for PIC addressing as read-only, rather than
19715 loading it in the prologue for each function. The runtime system is
19716 responsible for initializing this register with an appropriate value
19717 before execution begins.
19719 @item -mprioritize-restricted-insns=@var{priority}
19720 @opindex mprioritize-restricted-insns
19721 This option controls the priority that is assigned to
19722 dispatch-slot restricted instructions during the second scheduling
19723 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
19724 or @samp{2} to assign no, highest, or second-highest (respectively)
19725 priority to dispatch-slot restricted
19728 @item -msched-costly-dep=@var{dependence_type}
19729 @opindex msched-costly-dep
19730 This option controls which dependences are considered costly
19731 by the target during instruction scheduling. The argument
19732 @var{dependence_type} takes one of the following values:
19736 No dependence is costly.
19739 All dependences are costly.
19741 @item @samp{true_store_to_load}
19742 A true dependence from store to load is costly.
19744 @item @samp{store_to_load}
19745 Any dependence from store to load is costly.
19748 Any dependence for which the latency is greater than or equal to
19749 @var{number} is costly.
19752 @item -minsert-sched-nops=@var{scheme}
19753 @opindex minsert-sched-nops
19754 This option controls which NOP insertion scheme is used during
19755 the second scheduling pass. The argument @var{scheme} takes one of the
19763 Pad with NOPs any dispatch group that has vacant issue slots,
19764 according to the scheduler's grouping.
19766 @item @samp{regroup_exact}
19767 Insert NOPs to force costly dependent insns into
19768 separate groups. Insert exactly as many NOPs as needed to force an insn
19769 to a new group, according to the estimated processor grouping.
19772 Insert NOPs to force costly dependent insns into
19773 separate groups. Insert @var{number} NOPs to force an insn to a new group.
19777 @opindex mcall-sysv
19778 On System V.4 and embedded PowerPC systems compile code using calling
19779 conventions that adhere to the March 1995 draft of the System V
19780 Application Binary Interface, PowerPC processor supplement. This is the
19781 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
19783 @item -mcall-sysv-eabi
19785 @opindex mcall-sysv-eabi
19786 @opindex mcall-eabi
19787 Specify both @option{-mcall-sysv} and @option{-meabi} options.
19789 @item -mcall-sysv-noeabi
19790 @opindex mcall-sysv-noeabi
19791 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
19793 @item -mcall-aixdesc
19795 On System V.4 and embedded PowerPC systems compile code for the AIX
19799 @opindex mcall-linux
19800 On System V.4 and embedded PowerPC systems compile code for the
19801 Linux-based GNU system.
19803 @item -mcall-freebsd
19804 @opindex mcall-freebsd
19805 On System V.4 and embedded PowerPC systems compile code for the
19806 FreeBSD operating system.
19808 @item -mcall-netbsd
19809 @opindex mcall-netbsd
19810 On System V.4 and embedded PowerPC systems compile code for the
19811 NetBSD operating system.
19813 @item -mcall-openbsd
19814 @opindex mcall-netbsd
19815 On System V.4 and embedded PowerPC systems compile code for the
19816 OpenBSD operating system.
19818 @item -maix-struct-return
19819 @opindex maix-struct-return
19820 Return all structures in memory (as specified by the AIX ABI)@.
19822 @item -msvr4-struct-return
19823 @opindex msvr4-struct-return
19824 Return structures smaller than 8 bytes in registers (as specified by the
19827 @item -mabi=@var{abi-type}
19829 Extend the current ABI with a particular extension, or remove such extension.
19830 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
19831 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
19832 @samp{elfv1}, @samp{elfv2}@.
19836 Extend the current ABI with SPE ABI extensions. This does not change
19837 the default ABI, instead it adds the SPE ABI extensions to the current
19841 @opindex mabi=no-spe
19842 Disable Book-E SPE ABI extensions for the current ABI@.
19844 @item -mabi=ibmlongdouble
19845 @opindex mabi=ibmlongdouble
19846 Change the current ABI to use IBM extended-precision long double.
19847 This is a PowerPC 32-bit SYSV ABI option.
19849 @item -mabi=ieeelongdouble
19850 @opindex mabi=ieeelongdouble
19851 Change the current ABI to use IEEE extended-precision long double.
19852 This is a PowerPC 32-bit Linux ABI option.
19855 @opindex mabi=elfv1
19856 Change the current ABI to use the ELFv1 ABI.
19857 This is the default ABI for big-endian PowerPC 64-bit Linux.
19858 Overriding the default ABI requires special system support and is
19859 likely to fail in spectacular ways.
19862 @opindex mabi=elfv2
19863 Change the current ABI to use the ELFv2 ABI.
19864 This is the default ABI for little-endian PowerPC 64-bit Linux.
19865 Overriding the default ABI requires special system support and is
19866 likely to fail in spectacular ways.
19869 @itemx -mno-prototype
19870 @opindex mprototype
19871 @opindex mno-prototype
19872 On System V.4 and embedded PowerPC systems assume that all calls to
19873 variable argument functions are properly prototyped. Otherwise, the
19874 compiler must insert an instruction before every non-prototyped call to
19875 set or clear bit 6 of the condition code register (@code{CR}) to
19876 indicate whether floating-point values are passed in the floating-point
19877 registers in case the function takes variable arguments. With
19878 @option{-mprototype}, only calls to prototyped variable argument functions
19879 set or clear the bit.
19883 On embedded PowerPC systems, assume that the startup module is called
19884 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
19885 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
19890 On embedded PowerPC systems, assume that the startup module is called
19891 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
19896 On embedded PowerPC systems, assume that the startup module is called
19897 @file{crt0.o} and the standard C libraries are @file{libads.a} and
19900 @item -myellowknife
19901 @opindex myellowknife
19902 On embedded PowerPC systems, assume that the startup module is called
19903 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
19908 On System V.4 and embedded PowerPC systems, specify that you are
19909 compiling for a VxWorks system.
19913 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
19914 header to indicate that @samp{eabi} extended relocations are used.
19920 On System V.4 and embedded PowerPC systems do (do not) adhere to the
19921 Embedded Applications Binary Interface (EABI), which is a set of
19922 modifications to the System V.4 specifications. Selecting @option{-meabi}
19923 means that the stack is aligned to an 8-byte boundary, a function
19924 @code{__eabi} is called from @code{main} to set up the EABI
19925 environment, and the @option{-msdata} option can use both @code{r2} and
19926 @code{r13} to point to two separate small data areas. Selecting
19927 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
19928 no EABI initialization function is called from @code{main}, and the
19929 @option{-msdata} option only uses @code{r13} to point to a single
19930 small data area. The @option{-meabi} option is on by default if you
19931 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
19934 @opindex msdata=eabi
19935 On System V.4 and embedded PowerPC systems, put small initialized
19936 @code{const} global and static data in the @code{.sdata2} section, which
19937 is pointed to by register @code{r2}. Put small initialized
19938 non-@code{const} global and static data in the @code{.sdata} section,
19939 which is pointed to by register @code{r13}. Put small uninitialized
19940 global and static data in the @code{.sbss} section, which is adjacent to
19941 the @code{.sdata} section. The @option{-msdata=eabi} option is
19942 incompatible with the @option{-mrelocatable} option. The
19943 @option{-msdata=eabi} option also sets the @option{-memb} option.
19946 @opindex msdata=sysv
19947 On System V.4 and embedded PowerPC systems, put small global and static
19948 data in the @code{.sdata} section, which is pointed to by register
19949 @code{r13}. Put small uninitialized global and static data in the
19950 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
19951 The @option{-msdata=sysv} option is incompatible with the
19952 @option{-mrelocatable} option.
19954 @item -msdata=default
19956 @opindex msdata=default
19958 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
19959 compile code the same as @option{-msdata=eabi}, otherwise compile code the
19960 same as @option{-msdata=sysv}.
19963 @opindex msdata=data
19964 On System V.4 and embedded PowerPC systems, put small global
19965 data in the @code{.sdata} section. Put small uninitialized global
19966 data in the @code{.sbss} section. Do not use register @code{r13}
19967 to address small data however. This is the default behavior unless
19968 other @option{-msdata} options are used.
19972 @opindex msdata=none
19974 On embedded PowerPC systems, put all initialized global and static data
19975 in the @code{.data} section, and all uninitialized data in the
19976 @code{.bss} section.
19978 @item -mblock-move-inline-limit=@var{num}
19979 @opindex mblock-move-inline-limit
19980 Inline all block moves (such as calls to @code{memcpy} or structure
19981 copies) less than or equal to @var{num} bytes. The minimum value for
19982 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
19983 targets. The default value is target-specific.
19987 @cindex smaller data references (PowerPC)
19988 @cindex .sdata/.sdata2 references (PowerPC)
19989 On embedded PowerPC systems, put global and static items less than or
19990 equal to @var{num} bytes into the small data or BSS sections instead of
19991 the normal data or BSS section. By default, @var{num} is 8. The
19992 @option{-G @var{num}} switch is also passed to the linker.
19993 All modules should be compiled with the same @option{-G @var{num}} value.
19996 @itemx -mno-regnames
19998 @opindex mno-regnames
19999 On System V.4 and embedded PowerPC systems do (do not) emit register
20000 names in the assembly language output using symbolic forms.
20003 @itemx -mno-longcall
20005 @opindex mno-longcall
20006 By default assume that all calls are far away so that a longer and more
20007 expensive calling sequence is required. This is required for calls
20008 farther than 32 megabytes (33,554,432 bytes) from the current location.
20009 A short call is generated if the compiler knows
20010 the call cannot be that far away. This setting can be overridden by
20011 the @code{shortcall} function attribute, or by @code{#pragma
20014 Some linkers are capable of detecting out-of-range calls and generating
20015 glue code on the fly. On these systems, long calls are unnecessary and
20016 generate slower code. As of this writing, the AIX linker can do this,
20017 as can the GNU linker for PowerPC/64. It is planned to add this feature
20018 to the GNU linker for 32-bit PowerPC systems as well.
20020 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20021 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20022 addresses represent the callee and the branch island. The
20023 Darwin/PPC linker prefers the first address and generates a @code{bl
20024 callee} if the PPC @code{bl} instruction reaches the callee directly;
20025 otherwise, the linker generates @code{bl L42} to call the branch
20026 island. The branch island is appended to the body of the
20027 calling function; it computes the full 32-bit address of the callee
20030 On Mach-O (Darwin) systems, this option directs the compiler emit to
20031 the glue for every direct call, and the Darwin linker decides whether
20032 to use or discard it.
20034 In the future, GCC may ignore all longcall specifications
20035 when the linker is known to generate glue.
20037 @item -mtls-markers
20038 @itemx -mno-tls-markers
20039 @opindex mtls-markers
20040 @opindex mno-tls-markers
20041 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20042 specifying the function argument. The relocation allows the linker to
20043 reliably associate function call with argument setup instructions for
20044 TLS optimization, which in turn allows GCC to better schedule the
20049 Adds support for multithreading with the @dfn{pthreads} library.
20050 This option sets flags for both the preprocessor and linker.
20055 This option enables use of the reciprocal estimate and
20056 reciprocal square root estimate instructions with additional
20057 Newton-Raphson steps to increase precision instead of doing a divide or
20058 square root and divide for floating-point arguments. You should use
20059 the @option{-ffast-math} option when using @option{-mrecip} (or at
20060 least @option{-funsafe-math-optimizations},
20061 @option{-finite-math-only}, @option{-freciprocal-math} and
20062 @option{-fno-trapping-math}). Note that while the throughput of the
20063 sequence is generally higher than the throughput of the non-reciprocal
20064 instruction, the precision of the sequence can be decreased by up to 2
20065 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20068 @item -mrecip=@var{opt}
20069 @opindex mrecip=opt
20070 This option controls which reciprocal estimate instructions
20071 may be used. @var{opt} is a comma-separated list of options, which may
20072 be preceded by a @code{!} to invert the option:
20077 Enable all estimate instructions.
20080 Enable the default instructions, equivalent to @option{-mrecip}.
20083 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20086 Enable the reciprocal approximation instructions for both
20087 single and double precision.
20090 Enable the single-precision reciprocal approximation instructions.
20093 Enable the double-precision reciprocal approximation instructions.
20096 Enable the reciprocal square root approximation instructions for both
20097 single and double precision.
20100 Enable the single-precision reciprocal square root approximation instructions.
20103 Enable the double-precision reciprocal square root approximation instructions.
20107 So, for example, @option{-mrecip=all,!rsqrtd} enables
20108 all of the reciprocal estimate instructions, except for the
20109 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20110 which handle the double-precision reciprocal square root calculations.
20112 @item -mrecip-precision
20113 @itemx -mno-recip-precision
20114 @opindex mrecip-precision
20115 Assume (do not assume) that the reciprocal estimate instructions
20116 provide higher-precision estimates than is mandated by the PowerPC
20117 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20118 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20119 The double-precision square root estimate instructions are not generated by
20120 default on low-precision machines, since they do not provide an
20121 estimate that converges after three steps.
20123 @item -mveclibabi=@var{type}
20124 @opindex mveclibabi
20125 Specifies the ABI type to use for vectorizing intrinsics using an
20126 external library. The only type supported at present is @samp{mass},
20127 which specifies to use IBM's Mathematical Acceleration Subsystem
20128 (MASS) libraries for vectorizing intrinsics using external libraries.
20129 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20130 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20131 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20132 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20133 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20134 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20135 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20136 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20137 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20138 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20139 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20140 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20141 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20142 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20143 for power7. Both @option{-ftree-vectorize} and
20144 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20145 libraries must be specified at link time.
20150 Generate (do not generate) the @code{friz} instruction when the
20151 @option{-funsafe-math-optimizations} option is used to optimize
20152 rounding of floating-point values to 64-bit integer and back to floating
20153 point. The @code{friz} instruction does not return the same value if
20154 the floating-point number is too large to fit in an integer.
20156 @item -mpointers-to-nested-functions
20157 @itemx -mno-pointers-to-nested-functions
20158 @opindex mpointers-to-nested-functions
20159 Generate (do not generate) code to load up the static chain register
20160 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20161 systems where a function pointer points to a 3-word descriptor giving
20162 the function address, TOC value to be loaded in register @code{r2}, and
20163 static chain value to be loaded in register @code{r11}. The
20164 @option{-mpointers-to-nested-functions} is on by default. You cannot
20165 call through pointers to nested functions or pointers
20166 to functions compiled in other languages that use the static chain if
20167 you use @option{-mno-pointers-to-nested-functions}.
20169 @item -msave-toc-indirect
20170 @itemx -mno-save-toc-indirect
20171 @opindex msave-toc-indirect
20172 Generate (do not generate) code to save the TOC value in the reserved
20173 stack location in the function prologue if the function calls through
20174 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20175 saved in the prologue, it is saved just before the call through the
20176 pointer. The @option{-mno-save-toc-indirect} option is the default.
20178 @item -mcompat-align-parm
20179 @itemx -mno-compat-align-parm
20180 @opindex mcompat-align-parm
20181 Generate (do not generate) code to pass structure parameters with a
20182 maximum alignment of 64 bits, for compatibility with older versions
20185 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20186 structure parameter on a 128-bit boundary when that structure contained
20187 a member requiring 128-bit alignment. This is corrected in more
20188 recent versions of GCC. This option may be used to generate code
20189 that is compatible with functions compiled with older versions of
20192 The @option{-mno-compat-align-parm} option is the default.
20196 @subsection RX Options
20199 These command-line options are defined for RX targets:
20202 @item -m64bit-doubles
20203 @itemx -m32bit-doubles
20204 @opindex m64bit-doubles
20205 @opindex m32bit-doubles
20206 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20207 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20208 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20209 works on 32-bit values, which is why the default is
20210 @option{-m32bit-doubles}.
20216 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20217 floating-point hardware. The default is enabled for the RX600
20218 series and disabled for the RX200 series.
20220 Floating-point instructions are only generated for 32-bit floating-point
20221 values, however, so the FPU hardware is not used for doubles if the
20222 @option{-m64bit-doubles} option is used.
20224 @emph{Note} If the @option{-fpu} option is enabled then
20225 @option{-funsafe-math-optimizations} is also enabled automatically.
20226 This is because the RX FPU instructions are themselves unsafe.
20228 @item -mcpu=@var{name}
20230 Selects the type of RX CPU to be targeted. Currently three types are
20231 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20232 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20234 The only difference between @samp{RX600} and @samp{RX610} is that the
20235 @samp{RX610} does not support the @code{MVTIPL} instruction.
20237 The @samp{RX200} series does not have a hardware floating-point unit
20238 and so @option{-nofpu} is enabled by default when this type is
20241 @item -mbig-endian-data
20242 @itemx -mlittle-endian-data
20243 @opindex mbig-endian-data
20244 @opindex mlittle-endian-data
20245 Store data (but not code) in the big-endian format. The default is
20246 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20249 @item -msmall-data-limit=@var{N}
20250 @opindex msmall-data-limit
20251 Specifies the maximum size in bytes of global and static variables
20252 which can be placed into the small data area. Using the small data
20253 area can lead to smaller and faster code, but the size of area is
20254 limited and it is up to the programmer to ensure that the area does
20255 not overflow. Also when the small data area is used one of the RX's
20256 registers (usually @code{r13}) is reserved for use pointing to this
20257 area, so it is no longer available for use by the compiler. This
20258 could result in slower and/or larger code if variables are pushed onto
20259 the stack instead of being held in this register.
20261 Note, common variables (variables that have not been initialized) and
20262 constants are not placed into the small data area as they are assigned
20263 to other sections in the output executable.
20265 The default value is zero, which disables this feature. Note, this
20266 feature is not enabled by default with higher optimization levels
20267 (@option{-O2} etc) because of the potentially detrimental effects of
20268 reserving a register. It is up to the programmer to experiment and
20269 discover whether this feature is of benefit to their program. See the
20270 description of the @option{-mpid} option for a description of how the
20271 actual register to hold the small data area pointer is chosen.
20277 Use the simulator runtime. The default is to use the libgloss
20278 board-specific runtime.
20280 @item -mas100-syntax
20281 @itemx -mno-as100-syntax
20282 @opindex mas100-syntax
20283 @opindex mno-as100-syntax
20284 When generating assembler output use a syntax that is compatible with
20285 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20286 assembler, but it has some restrictions so it is not generated by default.
20288 @item -mmax-constant-size=@var{N}
20289 @opindex mmax-constant-size
20290 Specifies the maximum size, in bytes, of a constant that can be used as
20291 an operand in a RX instruction. Although the RX instruction set does
20292 allow constants of up to 4 bytes in length to be used in instructions,
20293 a longer value equates to a longer instruction. Thus in some
20294 circumstances it can be beneficial to restrict the size of constants
20295 that are used in instructions. Constants that are too big are instead
20296 placed into a constant pool and referenced via register indirection.
20298 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20299 or 4 means that constants of any size are allowed.
20303 Enable linker relaxation. Linker relaxation is a process whereby the
20304 linker attempts to reduce the size of a program by finding shorter
20305 versions of various instructions. Disabled by default.
20307 @item -mint-register=@var{N}
20308 @opindex mint-register
20309 Specify the number of registers to reserve for fast interrupt handler
20310 functions. The value @var{N} can be between 0 and 4. A value of 1
20311 means that register @code{r13} is reserved for the exclusive use
20312 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20313 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20314 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20315 A value of 0, the default, does not reserve any registers.
20317 @item -msave-acc-in-interrupts
20318 @opindex msave-acc-in-interrupts
20319 Specifies that interrupt handler functions should preserve the
20320 accumulator register. This is only necessary if normal code might use
20321 the accumulator register, for example because it performs 64-bit
20322 multiplications. The default is to ignore the accumulator as this
20323 makes the interrupt handlers faster.
20329 Enables the generation of position independent data. When enabled any
20330 access to constant data is done via an offset from a base address
20331 held in a register. This allows the location of constant data to be
20332 determined at run time without requiring the executable to be
20333 relocated, which is a benefit to embedded applications with tight
20334 memory constraints. Data that can be modified is not affected by this
20337 Note, using this feature reserves a register, usually @code{r13}, for
20338 the constant data base address. This can result in slower and/or
20339 larger code, especially in complicated functions.
20341 The actual register chosen to hold the constant data base address
20342 depends upon whether the @option{-msmall-data-limit} and/or the
20343 @option{-mint-register} command-line options are enabled. Starting
20344 with register @code{r13} and proceeding downwards, registers are
20345 allocated first to satisfy the requirements of @option{-mint-register},
20346 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20347 is possible for the small data area register to be @code{r8} if both
20348 @option{-mint-register=4} and @option{-mpid} are specified on the
20351 By default this feature is not enabled. The default can be restored
20352 via the @option{-mno-pid} command-line option.
20354 @item -mno-warn-multiple-fast-interrupts
20355 @itemx -mwarn-multiple-fast-interrupts
20356 @opindex mno-warn-multiple-fast-interrupts
20357 @opindex mwarn-multiple-fast-interrupts
20358 Prevents GCC from issuing a warning message if it finds more than one
20359 fast interrupt handler when it is compiling a file. The default is to
20360 issue a warning for each extra fast interrupt handler found, as the RX
20361 only supports one such interrupt.
20363 @item -mallow-string-insns
20364 @itemx -mno-allow-string-insns
20365 @opindex mallow-string-insns
20366 @opindex mno-allow-string-insns
20367 Enables or disables the use of the string manipulation instructions
20368 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20369 @code{SWHILE} and also the @code{RMPA} instruction. These
20370 instructions may prefetch data, which is not safe to do if accessing
20371 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20372 for more information).
20374 The default is to allow these instructions, but it is not possible for
20375 GCC to reliably detect all circumstances where a string instruction
20376 might be used to access an I/O register, so their use cannot be
20377 disabled automatically. Instead it is reliant upon the programmer to
20378 use the @option{-mno-allow-string-insns} option if their program
20379 accesses I/O space.
20381 When the instructions are enabled GCC defines the C preprocessor
20382 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20383 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20386 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20387 has special significance to the RX port when used with the
20388 @code{interrupt} function attribute. This attribute indicates a
20389 function intended to process fast interrupts. GCC ensures
20390 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20391 and/or @code{r13} and only provided that the normal use of the
20392 corresponding registers have been restricted via the
20393 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20396 @node S/390 and zSeries Options
20397 @subsection S/390 and zSeries Options
20398 @cindex S/390 and zSeries Options
20400 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20404 @itemx -msoft-float
20405 @opindex mhard-float
20406 @opindex msoft-float
20407 Use (do not use) the hardware floating-point instructions and registers
20408 for floating-point operations. When @option{-msoft-float} is specified,
20409 functions in @file{libgcc.a} are used to perform floating-point
20410 operations. When @option{-mhard-float} is specified, the compiler
20411 generates IEEE floating-point instructions. This is the default.
20414 @itemx -mno-hard-dfp
20416 @opindex mno-hard-dfp
20417 Use (do not use) the hardware decimal-floating-point instructions for
20418 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20419 specified, functions in @file{libgcc.a} are used to perform
20420 decimal-floating-point operations. When @option{-mhard-dfp} is
20421 specified, the compiler generates decimal-floating-point hardware
20422 instructions. This is the default for @option{-march=z9-ec} or higher.
20424 @item -mlong-double-64
20425 @itemx -mlong-double-128
20426 @opindex mlong-double-64
20427 @opindex mlong-double-128
20428 These switches control the size of @code{long double} type. A size
20429 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20430 type. This is the default.
20433 @itemx -mno-backchain
20434 @opindex mbackchain
20435 @opindex mno-backchain
20436 Store (do not store) the address of the caller's frame as backchain pointer
20437 into the callee's stack frame.
20438 A backchain may be needed to allow debugging using tools that do not understand
20439 DWARF 2 call frame information.
20440 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20441 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20442 the backchain is placed into the topmost word of the 96/160 byte register
20445 In general, code compiled with @option{-mbackchain} is call-compatible with
20446 code compiled with @option{-mmo-backchain}; however, use of the backchain
20447 for debugging purposes usually requires that the whole binary is built with
20448 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20449 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20450 to build a linux kernel use @option{-msoft-float}.
20452 The default is to not maintain the backchain.
20454 @item -mpacked-stack
20455 @itemx -mno-packed-stack
20456 @opindex mpacked-stack
20457 @opindex mno-packed-stack
20458 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20459 specified, the compiler uses the all fields of the 96/160 byte register save
20460 area only for their default purpose; unused fields still take up stack space.
20461 When @option{-mpacked-stack} is specified, register save slots are densely
20462 packed at the top of the register save area; unused space is reused for other
20463 purposes, allowing for more efficient use of the available stack space.
20464 However, when @option{-mbackchain} is also in effect, the topmost word of
20465 the save area is always used to store the backchain, and the return address
20466 register is always saved two words below the backchain.
20468 As long as the stack frame backchain is not used, code generated with
20469 @option{-mpacked-stack} is call-compatible with code generated with
20470 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20471 S/390 or zSeries generated code that uses the stack frame backchain at run
20472 time, not just for debugging purposes. Such code is not call-compatible
20473 with code compiled with @option{-mpacked-stack}. Also, note that the
20474 combination of @option{-mbackchain},
20475 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20476 to build a linux kernel use @option{-msoft-float}.
20478 The default is to not use the packed stack layout.
20481 @itemx -mno-small-exec
20482 @opindex msmall-exec
20483 @opindex mno-small-exec
20484 Generate (or do not generate) code using the @code{bras} instruction
20485 to do subroutine calls.
20486 This only works reliably if the total executable size does not
20487 exceed 64k. The default is to use the @code{basr} instruction instead,
20488 which does not have this limitation.
20494 When @option{-m31} is specified, generate code compliant to the
20495 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20496 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20497 particular to generate 64-bit instructions. For the @samp{s390}
20498 targets, the default is @option{-m31}, while the @samp{s390x}
20499 targets default to @option{-m64}.
20505 When @option{-mzarch} is specified, generate code using the
20506 instructions available on z/Architecture.
20507 When @option{-mesa} is specified, generate code using the
20508 instructions available on ESA/390. Note that @option{-mesa} is
20509 not possible with @option{-m64}.
20510 When generating code compliant to the GNU/Linux for S/390 ABI,
20511 the default is @option{-mesa}. When generating code compliant
20512 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20518 Generate (or do not generate) code using the @code{mvcle} instruction
20519 to perform block moves. When @option{-mno-mvcle} is specified,
20520 use a @code{mvc} loop instead. This is the default unless optimizing for
20527 Print (or do not print) additional debug information when compiling.
20528 The default is to not print debug information.
20530 @item -march=@var{cpu-type}
20532 Generate code that runs on @var{cpu-type}, which is the name of a system
20533 representing a certain processor type. Possible values for
20534 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, @samp{z990},
20535 @samp{z9-109}, @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12},
20537 When generating code using the instructions available on z/Architecture,
20538 the default is @option{-march=z900}. Otherwise, the default is
20539 @option{-march=g5}.
20541 @item -mtune=@var{cpu-type}
20543 Tune to @var{cpu-type} everything applicable about the generated code,
20544 except for the ABI and the set of available instructions.
20545 The list of @var{cpu-type} values is the same as for @option{-march}.
20546 The default is the value used for @option{-march}.
20549 @itemx -mno-tpf-trace
20550 @opindex mtpf-trace
20551 @opindex mno-tpf-trace
20552 Generate code that adds (does not add) in TPF OS specific branches to trace
20553 routines in the operating system. This option is off by default, even
20554 when compiling for the TPF OS@.
20557 @itemx -mno-fused-madd
20558 @opindex mfused-madd
20559 @opindex mno-fused-madd
20560 Generate code that uses (does not use) the floating-point multiply and
20561 accumulate instructions. These instructions are generated by default if
20562 hardware floating point is used.
20564 @item -mwarn-framesize=@var{framesize}
20565 @opindex mwarn-framesize
20566 Emit a warning if the current function exceeds the given frame size. Because
20567 this is a compile-time check it doesn't need to be a real problem when the program
20568 runs. It is intended to identify functions that most probably cause
20569 a stack overflow. It is useful to be used in an environment with limited stack
20570 size e.g.@: the linux kernel.
20572 @item -mwarn-dynamicstack
20573 @opindex mwarn-dynamicstack
20574 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20575 arrays. This is generally a bad idea with a limited stack size.
20577 @item -mstack-guard=@var{stack-guard}
20578 @itemx -mstack-size=@var{stack-size}
20579 @opindex mstack-guard
20580 @opindex mstack-size
20581 If these options are provided the S/390 back end emits additional instructions in
20582 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20583 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20584 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20585 the frame size of the compiled function is chosen.
20586 These options are intended to be used to help debugging stack overflow problems.
20587 The additionally emitted code causes only little overhead and hence can also be
20588 used in production-like systems without greater performance degradation. The given
20589 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20590 @var{stack-guard} without exceeding 64k.
20591 In order to be efficient the extra code makes the assumption that the stack starts
20592 at an address aligned to the value given by @var{stack-size}.
20593 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20595 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20597 If the hotpatch option is enabled, a ``hot-patching'' function
20598 prologue is generated for all functions in the compilation unit.
20599 The funtion label is prepended with the given number of two-byte
20600 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20601 the label, 2 * @var{post-halfwords} bytes are appended, using the
20602 largest NOP like instructions the architecture allows (maximum
20605 If both arguments are zero, hotpatching is disabled.
20607 This option can be overridden for individual functions with the
20608 @code{hotpatch} attribute.
20611 @node Score Options
20612 @subsection Score Options
20613 @cindex Score Options
20615 These options are defined for Score implementations:
20620 Compile code for big-endian mode. This is the default.
20624 Compile code for little-endian mode.
20628 Disable generation of @code{bcnz} instructions.
20632 Enable generation of unaligned load and store instructions.
20636 Enable the use of multiply-accumulate instructions. Disabled by default.
20640 Specify the SCORE5 as the target architecture.
20644 Specify the SCORE5U of the target architecture.
20648 Specify the SCORE7 as the target architecture. This is the default.
20652 Specify the SCORE7D as the target architecture.
20656 @subsection SH Options
20658 These @samp{-m} options are defined for the SH implementations:
20663 Generate code for the SH1.
20667 Generate code for the SH2.
20670 Generate code for the SH2e.
20674 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
20675 that the floating-point unit is not used.
20677 @item -m2a-single-only
20678 @opindex m2a-single-only
20679 Generate code for the SH2a-FPU, in such a way that no double-precision
20680 floating-point operations are used.
20683 @opindex m2a-single
20684 Generate code for the SH2a-FPU assuming the floating-point unit is in
20685 single-precision mode by default.
20689 Generate code for the SH2a-FPU assuming the floating-point unit is in
20690 double-precision mode by default.
20694 Generate code for the SH3.
20698 Generate code for the SH3e.
20702 Generate code for the SH4 without a floating-point unit.
20704 @item -m4-single-only
20705 @opindex m4-single-only
20706 Generate code for the SH4 with a floating-point unit that only
20707 supports single-precision arithmetic.
20711 Generate code for the SH4 assuming the floating-point unit is in
20712 single-precision mode by default.
20716 Generate code for the SH4.
20720 Generate code for SH4-100.
20722 @item -m4-100-nofpu
20723 @opindex m4-100-nofpu
20724 Generate code for SH4-100 in such a way that the
20725 floating-point unit is not used.
20727 @item -m4-100-single
20728 @opindex m4-100-single
20729 Generate code for SH4-100 assuming the floating-point unit is in
20730 single-precision mode by default.
20732 @item -m4-100-single-only
20733 @opindex m4-100-single-only
20734 Generate code for SH4-100 in such a way that no double-precision
20735 floating-point operations are used.
20739 Generate code for SH4-200.
20741 @item -m4-200-nofpu
20742 @opindex m4-200-nofpu
20743 Generate code for SH4-200 without in such a way that the
20744 floating-point unit is not used.
20746 @item -m4-200-single
20747 @opindex m4-200-single
20748 Generate code for SH4-200 assuming the floating-point unit is in
20749 single-precision mode by default.
20751 @item -m4-200-single-only
20752 @opindex m4-200-single-only
20753 Generate code for SH4-200 in such a way that no double-precision
20754 floating-point operations are used.
20758 Generate code for SH4-300.
20760 @item -m4-300-nofpu
20761 @opindex m4-300-nofpu
20762 Generate code for SH4-300 without in such a way that the
20763 floating-point unit is not used.
20765 @item -m4-300-single
20766 @opindex m4-300-single
20767 Generate code for SH4-300 in such a way that no double-precision
20768 floating-point operations are used.
20770 @item -m4-300-single-only
20771 @opindex m4-300-single-only
20772 Generate code for SH4-300 in such a way that no double-precision
20773 floating-point operations are used.
20777 Generate code for SH4-340 (no MMU, no FPU).
20781 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
20786 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
20787 floating-point unit is not used.
20789 @item -m4a-single-only
20790 @opindex m4a-single-only
20791 Generate code for the SH4a, in such a way that no double-precision
20792 floating-point operations are used.
20795 @opindex m4a-single
20796 Generate code for the SH4a assuming the floating-point unit is in
20797 single-precision mode by default.
20801 Generate code for the SH4a.
20805 Same as @option{-m4a-nofpu}, except that it implicitly passes
20806 @option{-dsp} to the assembler. GCC doesn't generate any DSP
20807 instructions at the moment.
20810 @opindex m5-32media
20811 Generate 32-bit code for SHmedia.
20813 @item -m5-32media-nofpu
20814 @opindex m5-32media-nofpu
20815 Generate 32-bit code for SHmedia in such a way that the
20816 floating-point unit is not used.
20819 @opindex m5-64media
20820 Generate 64-bit code for SHmedia.
20822 @item -m5-64media-nofpu
20823 @opindex m5-64media-nofpu
20824 Generate 64-bit code for SHmedia in such a way that the
20825 floating-point unit is not used.
20828 @opindex m5-compact
20829 Generate code for SHcompact.
20831 @item -m5-compact-nofpu
20832 @opindex m5-compact-nofpu
20833 Generate code for SHcompact in such a way that the
20834 floating-point unit is not used.
20838 Compile code for the processor in big-endian mode.
20842 Compile code for the processor in little-endian mode.
20846 Align doubles at 64-bit boundaries. Note that this changes the calling
20847 conventions, and thus some functions from the standard C library do
20848 not work unless you recompile it first with @option{-mdalign}.
20852 Shorten some address references at link time, when possible; uses the
20853 linker option @option{-relax}.
20857 Use 32-bit offsets in @code{switch} tables. The default is to use
20862 Enable the use of bit manipulation instructions on SH2A.
20866 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
20867 alignment constraints.
20871 Comply with the calling conventions defined by Renesas.
20874 @opindex mno-renesas
20875 Comply with the calling conventions defined for GCC before the Renesas
20876 conventions were available. This option is the default for all
20877 targets of the SH toolchain.
20880 @opindex mnomacsave
20881 Mark the @code{MAC} register as call-clobbered, even if
20882 @option{-mrenesas} is given.
20888 Control the IEEE compliance of floating-point comparisons, which affects the
20889 handling of cases where the result of a comparison is unordered. By default
20890 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
20891 enabled @option{-mno-ieee} is implicitly set, which results in faster
20892 floating-point greater-equal and less-equal comparisons. The implcit settings
20893 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
20895 @item -minline-ic_invalidate
20896 @opindex minline-ic_invalidate
20897 Inline code to invalidate instruction cache entries after setting up
20898 nested function trampolines.
20899 This option has no effect if @option{-musermode} is in effect and the selected
20900 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
20902 If the selected code generation option does not allow the use of the @code{icbi}
20903 instruction, and @option{-musermode} is not in effect, the inlined code
20904 manipulates the instruction cache address array directly with an associative
20905 write. This not only requires privileged mode at run time, but it also
20906 fails if the cache line had been mapped via the TLB and has become unmapped.
20910 Dump instruction size and location in the assembly code.
20913 @opindex mpadstruct
20914 This option is deprecated. It pads structures to multiple of 4 bytes,
20915 which is incompatible with the SH ABI@.
20917 @item -matomic-model=@var{model}
20918 @opindex matomic-model=@var{model}
20919 Sets the model of atomic operations and additional parameters as a comma
20920 separated list. For details on the atomic built-in functions see
20921 @ref{__atomic Builtins}. The following models and parameters are supported:
20926 Disable compiler generated atomic sequences and emit library calls for atomic
20927 operations. This is the default if the target is not @code{sh*-*-linux*}.
20930 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
20931 built-in functions. The generated atomic sequences require additional support
20932 from the interrupt/exception handling code of the system and are only suitable
20933 for SH3* and SH4* single-core systems. This option is enabled by default when
20934 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
20935 this option also partially utilizes the hardware atomic instructions
20936 @code{movli.l} and @code{movco.l} to create more efficient code, unless
20937 @samp{strict} is specified.
20940 Generate software atomic sequences that use a variable in the thread control
20941 block. This is a variation of the gUSA sequences which can also be used on
20942 SH1* and SH2* targets. The generated atomic sequences require additional
20943 support from the interrupt/exception handling code of the system and are only
20944 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
20945 parameter has to be specified as well.
20948 Generate software atomic sequences that temporarily disable interrupts by
20949 setting @code{SR.IMASK = 1111}. This model works only when the program runs
20950 in privileged mode and is only suitable for single-core systems. Additional
20951 support from the interrupt/exception handling code of the system is not
20952 required. This model is enabled by default when the target is
20953 @code{sh*-*-linux*} and SH1* or SH2*.
20956 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
20957 instructions only. This is only available on SH4A and is suitable for
20958 multi-core systems. Since the hardware instructions support only 32 bit atomic
20959 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
20960 Code compiled with this option is also compatible with other software
20961 atomic model interrupt/exception handling systems if executed on an SH4A
20962 system. Additional support from the interrupt/exception handling code of the
20963 system is not required for this model.
20966 This parameter specifies the offset in bytes of the variable in the thread
20967 control block structure that should be used by the generated atomic sequences
20968 when the @samp{soft-tcb} model has been selected. For other models this
20969 parameter is ignored. The specified value must be an integer multiple of four
20970 and in the range 0-1020.
20973 This parameter prevents mixed usage of multiple atomic models, even if they
20974 are compatible, and makes the compiler generate atomic sequences of the
20975 specified model only.
20981 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
20982 Notice that depending on the particular hardware and software configuration
20983 this can degrade overall performance due to the operand cache line flushes
20984 that are implied by the @code{tas.b} instruction. On multi-core SH4A
20985 processors the @code{tas.b} instruction must be used with caution since it
20986 can result in data corruption for certain cache configurations.
20989 @opindex mprefergot
20990 When generating position-independent code, emit function calls using
20991 the Global Offset Table instead of the Procedure Linkage Table.
20994 @itemx -mno-usermode
20996 @opindex mno-usermode
20997 Don't allow (allow) the compiler generating privileged mode code. Specifying
20998 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
20999 inlined code would not work in user mode. @option{-musermode} is the default
21000 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21001 @option{-musermode} has no effect, since there is no user mode.
21003 @item -multcost=@var{number}
21004 @opindex multcost=@var{number}
21005 Set the cost to assume for a multiply insn.
21007 @item -mdiv=@var{strategy}
21008 @opindex mdiv=@var{strategy}
21009 Set the division strategy to be used for integer division operations.
21010 For SHmedia @var{strategy} can be one of:
21015 Performs the operation in floating point. This has a very high latency,
21016 but needs only a few instructions, so it might be a good choice if
21017 your code has enough easily-exploitable ILP to allow the compiler to
21018 schedule the floating-point instructions together with other instructions.
21019 Division by zero causes a floating-point exception.
21022 Uses integer operations to calculate the inverse of the divisor,
21023 and then multiplies the dividend with the inverse. This strategy allows
21024 CSE and hoisting of the inverse calculation. Division by zero calculates
21025 an unspecified result, but does not trap.
21028 A variant of @samp{inv} where, if no CSE or hoisting opportunities
21029 have been found, or if the entire operation has been hoisted to the same
21030 place, the last stages of the inverse calculation are intertwined with the
21031 final multiply to reduce the overall latency, at the expense of using a few
21032 more instructions, and thus offering fewer scheduling opportunities with
21036 Calls a library function that usually implements the @samp{inv:minlat}
21038 This gives high code density for @code{m5-*media-nofpu} compilations.
21041 Uses a different entry point of the same library function, where it
21042 assumes that a pointer to a lookup table has already been set up, which
21043 exposes the pointer load to CSE and code hoisting optimizations.
21048 Use the @samp{inv} algorithm for initial
21049 code generation, but if the code stays unoptimized, revert to the @samp{call},
21050 @samp{call2}, or @samp{fp} strategies, respectively. Note that the
21051 potentially-trapping side effect of division by zero is carried by a
21052 separate instruction, so it is possible that all the integer instructions
21053 are hoisted out, but the marker for the side effect stays where it is.
21054 A recombination to floating-point operations or a call is not possible
21059 Variants of the @samp{inv:minlat} strategy. In the case
21060 that the inverse calculation is not separated from the multiply, they speed
21061 up division where the dividend fits into 20 bits (plus sign where applicable)
21062 by inserting a test to skip a number of operations in this case; this test
21063 slows down the case of larger dividends. @samp{inv20u} assumes the case of a such
21064 a small dividend to be unlikely, and @samp{inv20l} assumes it to be likely.
21068 For targets other than SHmedia @var{strategy} can be one of:
21073 Calls a library function that uses the single-step division instruction
21074 @code{div1} to perform the operation. Division by zero calculates an
21075 unspecified result and does not trap. This is the default except for SH4,
21076 SH2A and SHcompact.
21079 Calls a library function that performs the operation in double precision
21080 floating point. Division by zero causes a floating-point exception. This is
21081 the default for SHcompact with FPU. Specifying this for targets that do not
21082 have a double precision FPU defaults to @code{call-div1}.
21085 Calls a library function that uses a lookup table for small divisors and
21086 the @code{div1} instruction with case distinction for larger divisors. Division
21087 by zero calculates an unspecified result and does not trap. This is the default
21088 for SH4. Specifying this for targets that do not have dynamic shift
21089 instructions defaults to @code{call-div1}.
21093 When a division strategy has not been specified the default strategy is
21094 selected based on the current target. For SH2A the default strategy is to
21095 use the @code{divs} and @code{divu} instructions instead of library function
21098 @item -maccumulate-outgoing-args
21099 @opindex maccumulate-outgoing-args
21100 Reserve space once for outgoing arguments in the function prologue rather
21101 than around each call. Generally beneficial for performance and size. Also
21102 needed for unwinding to avoid changing the stack frame around conditional code.
21104 @item -mdivsi3_libfunc=@var{name}
21105 @opindex mdivsi3_libfunc=@var{name}
21106 Set the name of the library function used for 32-bit signed division to
21108 This only affects the name used in the @samp{call} and @samp{inv:call}
21109 division strategies, and the compiler still expects the same
21110 sets of input/output/clobbered registers as if this option were not present.
21112 @item -mfixed-range=@var{register-range}
21113 @opindex mfixed-range
21114 Generate code treating the given register range as fixed registers.
21115 A fixed register is one that the register allocator can not use. This is
21116 useful when compiling kernel code. A register range is specified as
21117 two registers separated by a dash. Multiple register ranges can be
21118 specified separated by a comma.
21120 @item -mindexed-addressing
21121 @opindex mindexed-addressing
21122 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
21123 This is only safe if the hardware and/or OS implement 32-bit wrap-around
21124 semantics for the indexed addressing mode. The architecture allows the
21125 implementation of processors with 64-bit MMU, which the OS could use to
21126 get 32-bit addressing, but since no current hardware implementation supports
21127 this or any other way to make the indexed addressing mode safe to use in
21128 the 32-bit ABI, the default is @option{-mno-indexed-addressing}.
21130 @item -mgettrcost=@var{number}
21131 @opindex mgettrcost=@var{number}
21132 Set the cost assumed for the @code{gettr} instruction to @var{number}.
21133 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
21137 Assume @code{pt*} instructions won't trap. This generally generates
21138 better-scheduled code, but is unsafe on current hardware.
21139 The current architecture
21140 definition says that @code{ptabs} and @code{ptrel} trap when the target
21142 This has the unintentional effect of making it unsafe to schedule these
21143 instructions before a branch, or hoist them out of a loop. For example,
21144 @code{__do_global_ctors}, a part of @file{libgcc}
21145 that runs constructors at program
21146 startup, calls functions in a list which is delimited by @minus{}1. With the
21147 @option{-mpt-fixed} option, the @code{ptabs} is done before testing against @minus{}1.
21148 That means that all the constructors run a bit more quickly, but when
21149 the loop comes to the end of the list, the program crashes because @code{ptabs}
21150 loads @minus{}1 into a target register.
21152 Since this option is unsafe for any
21153 hardware implementing the current architecture specification, the default
21154 is @option{-mno-pt-fixed}. Unless specified explicitly with
21155 @option{-mgettrcost}, @option{-mno-pt-fixed} also implies @option{-mgettrcost=100};
21156 this deters register allocation from using target registers for storing
21159 @item -minvalid-symbols
21160 @opindex minvalid-symbols
21161 Assume symbols might be invalid. Ordinary function symbols generated by
21162 the compiler are always valid to load with
21163 @code{movi}/@code{shori}/@code{ptabs} or
21164 @code{movi}/@code{shori}/@code{ptrel},
21165 but with assembler and/or linker tricks it is possible
21166 to generate symbols that cause @code{ptabs} or @code{ptrel} to trap.
21167 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
21168 It prevents cross-basic-block CSE, hoisting and most scheduling
21169 of symbol loads. The default is @option{-mno-invalid-symbols}.
21171 @item -mbranch-cost=@var{num}
21172 @opindex mbranch-cost=@var{num}
21173 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21174 make the compiler try to generate more branch-free code if possible.
21175 If not specified the value is selected depending on the processor type that
21176 is being compiled for.
21179 @itemx -mno-zdcbranch
21180 @opindex mzdcbranch
21181 @opindex mno-zdcbranch
21182 Assume (do not assume) that zero displacement conditional branch instructions
21183 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21184 compiler prefers zero displacement branch code sequences. This is
21185 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21186 disabled by specifying @option{-mno-zdcbranch}.
21188 @item -mcbranch-force-delay-slot
21189 @opindex mcbranch-force-delay-slot
21190 Force the usage of delay slots for conditional branches, which stuffs the delay
21191 slot with a @code{nop} if a suitable instruction can't be found. By default
21192 this option is disabled. It can be enabled to work around hardware bugs as
21193 found in the original SH7055.
21196 @itemx -mno-fused-madd
21197 @opindex mfused-madd
21198 @opindex mno-fused-madd
21199 Generate code that uses (does not use) the floating-point multiply and
21200 accumulate instructions. These instructions are generated by default
21201 if hardware floating point is used. The machine-dependent
21202 @option{-mfused-madd} option is now mapped to the machine-independent
21203 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21204 mapped to @option{-ffp-contract=off}.
21210 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21211 and cosine approximations. The option @option{-mfsca} must be used in
21212 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21213 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21214 approximations even if @option{-funsafe-math-optimizations} is in effect.
21220 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21221 reciprocal square root approximations. The option @option{-mfsrra} must be used
21222 in combination with @option{-funsafe-math-optimizations} and
21223 @option{-ffinite-math-only}. It is enabled by default when generating code for
21224 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21225 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21228 @item -mpretend-cmove
21229 @opindex mpretend-cmove
21230 Prefer zero-displacement conditional branches for conditional move instruction
21231 patterns. This can result in faster code on the SH4 processor.
21235 @node Solaris 2 Options
21236 @subsection Solaris 2 Options
21237 @cindex Solaris 2 options
21239 These @samp{-m} options are supported on Solaris 2:
21242 @item -mclear-hwcap
21243 @opindex mclear-hwcap
21244 @option{-mclear-hwcap} tells the compiler to remove the hardware
21245 capabilities generated by the Solaris assembler. This is only necessary
21246 when object files use ISA extensions not supported by the current
21247 machine, but check at runtime whether or not to use them.
21249 @item -mimpure-text
21250 @opindex mimpure-text
21251 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21252 the compiler to not pass @option{-z text} to the linker when linking a
21253 shared object. Using this option, you can link position-dependent
21254 code into a shared object.
21256 @option{-mimpure-text} suppresses the ``relocations remain against
21257 allocatable but non-writable sections'' linker error message.
21258 However, the necessary relocations trigger copy-on-write, and the
21259 shared object is not actually shared across processes. Instead of
21260 using @option{-mimpure-text}, you should compile all source code with
21261 @option{-fpic} or @option{-fPIC}.
21265 These switches are supported in addition to the above on Solaris 2:
21270 Add support for multithreading using the POSIX threads library. This
21271 option sets flags for both the preprocessor and linker. This option does
21272 not affect the thread safety of object code produced by the compiler or
21273 that of libraries supplied with it.
21277 This is a synonym for @option{-pthreads}.
21280 @node SPARC Options
21281 @subsection SPARC Options
21282 @cindex SPARC options
21284 These @samp{-m} options are supported on the SPARC:
21287 @item -mno-app-regs
21289 @opindex mno-app-regs
21291 Specify @option{-mapp-regs} to generate output using the global registers
21292 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21293 global register 1, each global register 2 through 4 is then treated as an
21294 allocable register that is clobbered by function calls. This is the default.
21296 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21297 specify @option{-mno-app-regs}. You should compile libraries and system
21298 software with this option.
21304 With @option{-mflat}, the compiler does not generate save/restore instructions
21305 and uses a ``flat'' or single register window model. This model is compatible
21306 with the regular register window model. The local registers and the input
21307 registers (0--5) are still treated as ``call-saved'' registers and are
21308 saved on the stack as needed.
21310 With @option{-mno-flat} (the default), the compiler generates save/restore
21311 instructions (except for leaf functions). This is the normal operating mode.
21314 @itemx -mhard-float
21316 @opindex mhard-float
21317 Generate output containing floating-point instructions. This is the
21321 @itemx -msoft-float
21323 @opindex msoft-float
21324 Generate output containing library calls for floating point.
21325 @strong{Warning:} the requisite libraries are not available for all SPARC
21326 targets. Normally the facilities of the machine's usual C compiler are
21327 used, but this cannot be done directly in cross-compilation. You must make
21328 your own arrangements to provide suitable library functions for
21329 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21330 @samp{sparclite-*-*} do provide software floating-point support.
21332 @option{-msoft-float} changes the calling convention in the output file;
21333 therefore, it is only useful if you compile @emph{all} of a program with
21334 this option. In particular, you need to compile @file{libgcc.a}, the
21335 library that comes with GCC, with @option{-msoft-float} in order for
21338 @item -mhard-quad-float
21339 @opindex mhard-quad-float
21340 Generate output containing quad-word (long double) floating-point
21343 @item -msoft-quad-float
21344 @opindex msoft-quad-float
21345 Generate output containing library calls for quad-word (long double)
21346 floating-point instructions. The functions called are those specified
21347 in the SPARC ABI@. This is the default.
21349 As of this writing, there are no SPARC implementations that have hardware
21350 support for the quad-word floating-point instructions. They all invoke
21351 a trap handler for one of these instructions, and then the trap handler
21352 emulates the effect of the instruction. Because of the trap handler overhead,
21353 this is much slower than calling the ABI library routines. Thus the
21354 @option{-msoft-quad-float} option is the default.
21356 @item -mno-unaligned-doubles
21357 @itemx -munaligned-doubles
21358 @opindex mno-unaligned-doubles
21359 @opindex munaligned-doubles
21360 Assume that doubles have 8-byte alignment. This is the default.
21362 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21363 alignment only if they are contained in another type, or if they have an
21364 absolute address. Otherwise, it assumes they have 4-byte alignment.
21365 Specifying this option avoids some rare compatibility problems with code
21366 generated by other compilers. It is not the default because it results
21367 in a performance loss, especially for floating-point code.
21370 @itemx -mno-user-mode
21371 @opindex muser-mode
21372 @opindex mno-user-mode
21373 Do not generate code that can only run in supervisor mode. This is relevant
21374 only for the @code{casa} instruction emitted for the LEON3 processor. The
21375 default is @option{-mno-user-mode}.
21377 @item -mno-faster-structs
21378 @itemx -mfaster-structs
21379 @opindex mno-faster-structs
21380 @opindex mfaster-structs
21381 With @option{-mfaster-structs}, the compiler assumes that structures
21382 should have 8-byte alignment. This enables the use of pairs of
21383 @code{ldd} and @code{std} instructions for copies in structure
21384 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21385 However, the use of this changed alignment directly violates the SPARC
21386 ABI@. Thus, it's intended only for use on targets where the developer
21387 acknowledges that their resulting code is not directly in line with
21388 the rules of the ABI@.
21390 @item -mcpu=@var{cpu_type}
21392 Set the instruction set, register set, and instruction scheduling parameters
21393 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21394 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21395 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21396 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21397 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21398 @samp{niagara3} and @samp{niagara4}.
21400 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21401 which selects the best architecture option for the host processor.
21402 @option{-mcpu=native} has no effect if GCC does not recognize
21405 Default instruction scheduling parameters are used for values that select
21406 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21407 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21409 Here is a list of each supported architecture and their supported
21417 supersparc, hypersparc, leon, leon3
21420 f930, f934, sparclite86x
21426 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21429 By default (unless configured otherwise), GCC generates code for the V7
21430 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21431 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21432 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21433 SPARCStation 1, 2, IPX etc.
21435 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21436 architecture. The only difference from V7 code is that the compiler emits
21437 the integer multiply and integer divide instructions which exist in SPARC-V8
21438 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21439 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21442 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21443 the SPARC architecture. This adds the integer multiply, integer divide step
21444 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21445 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21446 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21447 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21448 MB86934 chip, which is the more recent SPARClite with FPU@.
21450 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21451 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21452 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21453 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21454 optimizes it for the TEMIC SPARClet chip.
21456 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21457 architecture. This adds 64-bit integer and floating-point move instructions,
21458 3 additional floating-point condition code registers and conditional move
21459 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21460 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21461 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21462 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21463 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21464 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21465 additionally optimizes it for Sun UltraSPARC T2 chips. With
21466 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21467 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21468 additionally optimizes it for Sun UltraSPARC T4 chips.
21470 @item -mtune=@var{cpu_type}
21472 Set the instruction scheduling parameters for machine type
21473 @var{cpu_type}, but do not set the instruction set or register set that the
21474 option @option{-mcpu=@var{cpu_type}} does.
21476 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21477 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21478 that select a particular CPU implementation. Those are @samp{cypress},
21479 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21480 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21481 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21482 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21483 toolchains, @samp{native} can also be used.
21488 @opindex mno-v8plus
21489 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21490 difference from the V8 ABI is that the global and out registers are
21491 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21492 mode for all SPARC-V9 processors.
21498 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21499 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21505 With @option{-mvis2}, GCC generates code that takes advantage of
21506 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21507 default is @option{-mvis2} when targeting a cpu that supports such
21508 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21509 also sets @option{-mvis}.
21515 With @option{-mvis3}, GCC generates code that takes advantage of
21516 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21517 default is @option{-mvis3} when targeting a cpu that supports such
21518 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21519 also sets @option{-mvis2} and @option{-mvis}.
21524 @opindex mno-cbcond
21525 With @option{-mcbcond}, GCC generates code that takes advantage of
21526 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21527 The default is @option{-mcbcond} when targeting a cpu that supports such
21528 instructions, such as niagara-4 and later.
21534 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21535 population count instruction. The default is @option{-mpopc}
21536 when targeting a cpu that supports such instructions, such as Niagara-2 and
21543 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21544 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21545 when targeting a cpu that supports such instructions, such as Niagara-3 and
21549 @opindex mfix-at697f
21550 Enable the documented workaround for the single erratum of the Atmel AT697F
21551 processor (which corresponds to erratum #13 of the AT697E processor).
21554 @opindex mfix-ut699
21555 Enable the documented workarounds for the floating-point errata and the data
21556 cache nullify errata of the UT699 processor.
21559 These @samp{-m} options are supported in addition to the above
21560 on SPARC-V9 processors in 64-bit environments:
21567 Generate code for a 32-bit or 64-bit environment.
21568 The 32-bit environment sets int, long and pointer to 32 bits.
21569 The 64-bit environment sets int to 32 bits and long and pointer
21572 @item -mcmodel=@var{which}
21574 Set the code model to one of
21578 The Medium/Low code model: 64-bit addresses, programs
21579 must be linked in the low 32 bits of memory. Programs can be statically
21580 or dynamically linked.
21583 The Medium/Middle code model: 64-bit addresses, programs
21584 must be linked in the low 44 bits of memory, the text and data segments must
21585 be less than 2GB in size and the data segment must be located within 2GB of
21589 The Medium/Anywhere code model: 64-bit addresses, programs
21590 may be linked anywhere in memory, the text and data segments must be less
21591 than 2GB in size and the data segment must be located within 2GB of the
21595 The Medium/Anywhere code model for embedded systems:
21596 64-bit addresses, the text and data segments must be less than 2GB in
21597 size, both starting anywhere in memory (determined at link time). The
21598 global register %g4 points to the base of the data segment. Programs
21599 are statically linked and PIC is not supported.
21602 @item -mmemory-model=@var{mem-model}
21603 @opindex mmemory-model
21604 Set the memory model in force on the processor to one of
21608 The default memory model for the processor and operating system.
21611 Relaxed Memory Order
21614 Partial Store Order
21620 Sequential Consistency
21623 These memory models are formally defined in Appendix D of the Sparc V9
21624 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21627 @itemx -mno-stack-bias
21628 @opindex mstack-bias
21629 @opindex mno-stack-bias
21630 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21631 frame pointer if present, are offset by @minus{}2047 which must be added back
21632 when making stack frame references. This is the default in 64-bit mode.
21633 Otherwise, assume no such offset is present.
21637 @subsection SPU Options
21638 @cindex SPU options
21640 These @samp{-m} options are supported on the SPU:
21644 @itemx -merror-reloc
21645 @opindex mwarn-reloc
21646 @opindex merror-reloc
21648 The loader for SPU does not handle dynamic relocations. By default, GCC
21649 gives an error when it generates code that requires a dynamic
21650 relocation. @option{-mno-error-reloc} disables the error,
21651 @option{-mwarn-reloc} generates a warning instead.
21654 @itemx -munsafe-dma
21656 @opindex munsafe-dma
21658 Instructions that initiate or test completion of DMA must not be
21659 reordered with respect to loads and stores of the memory that is being
21661 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21662 memory accesses, but that can lead to inefficient code in places where the
21663 memory is known to not change. Rather than mark the memory as volatile,
21664 you can use @option{-msafe-dma} to tell the compiler to treat
21665 the DMA instructions as potentially affecting all memory.
21667 @item -mbranch-hints
21668 @opindex mbranch-hints
21670 By default, GCC generates a branch hint instruction to avoid
21671 pipeline stalls for always-taken or probably-taken branches. A hint
21672 is not generated closer than 8 instructions away from its branch.
21673 There is little reason to disable them, except for debugging purposes,
21674 or to make an object a little bit smaller.
21678 @opindex msmall-mem
21679 @opindex mlarge-mem
21681 By default, GCC generates code assuming that addresses are never larger
21682 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21683 a full 32-bit address.
21688 By default, GCC links against startup code that assumes the SPU-style
21689 main function interface (which has an unconventional parameter list).
21690 With @option{-mstdmain}, GCC links your program against startup
21691 code that assumes a C99-style interface to @code{main}, including a
21692 local copy of @code{argv} strings.
21694 @item -mfixed-range=@var{register-range}
21695 @opindex mfixed-range
21696 Generate code treating the given register range as fixed registers.
21697 A fixed register is one that the register allocator cannot use. This is
21698 useful when compiling kernel code. A register range is specified as
21699 two registers separated by a dash. Multiple register ranges can be
21700 specified separated by a comma.
21706 Compile code assuming that pointers to the PPU address space accessed
21707 via the @code{__ea} named address space qualifier are either 32 or 64
21708 bits wide. The default is 32 bits. As this is an ABI-changing option,
21709 all object code in an executable must be compiled with the same setting.
21711 @item -maddress-space-conversion
21712 @itemx -mno-address-space-conversion
21713 @opindex maddress-space-conversion
21714 @opindex mno-address-space-conversion
21715 Allow/disallow treating the @code{__ea} address space as superset
21716 of the generic address space. This enables explicit type casts
21717 between @code{__ea} and generic pointer as well as implicit
21718 conversions of generic pointers to @code{__ea} pointers. The
21719 default is to allow address space pointer conversions.
21721 @item -mcache-size=@var{cache-size}
21722 @opindex mcache-size
21723 This option controls the version of libgcc that the compiler links to an
21724 executable and selects a software-managed cache for accessing variables
21725 in the @code{__ea} address space with a particular cache size. Possible
21726 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21727 and @samp{128}. The default cache size is 64KB.
21729 @item -matomic-updates
21730 @itemx -mno-atomic-updates
21731 @opindex matomic-updates
21732 @opindex mno-atomic-updates
21733 This option controls the version of libgcc that the compiler links to an
21734 executable and selects whether atomic updates to the software-managed
21735 cache of PPU-side variables are used. If you use atomic updates, changes
21736 to a PPU variable from SPU code using the @code{__ea} named address space
21737 qualifier do not interfere with changes to other PPU variables residing
21738 in the same cache line from PPU code. If you do not use atomic updates,
21739 such interference may occur; however, writing back cache lines is
21740 more efficient. The default behavior is to use atomic updates.
21743 @itemx -mdual-nops=@var{n}
21744 @opindex mdual-nops
21745 By default, GCC inserts nops to increase dual issue when it expects
21746 it to increase performance. @var{n} can be a value from 0 to 10. A
21747 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21748 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21750 @item -mhint-max-nops=@var{n}
21751 @opindex mhint-max-nops
21752 Maximum number of nops to insert for a branch hint. A branch hint must
21753 be at least 8 instructions away from the branch it is affecting. GCC
21754 inserts up to @var{n} nops to enforce this, otherwise it does not
21755 generate the branch hint.
21757 @item -mhint-max-distance=@var{n}
21758 @opindex mhint-max-distance
21759 The encoding of the branch hint instruction limits the hint to be within
21760 256 instructions of the branch it is affecting. By default, GCC makes
21761 sure it is within 125.
21764 @opindex msafe-hints
21765 Work around a hardware bug that causes the SPU to stall indefinitely.
21766 By default, GCC inserts the @code{hbrp} instruction to make sure
21767 this stall won't happen.
21771 @node System V Options
21772 @subsection Options for System V
21774 These additional options are available on System V Release 4 for
21775 compatibility with other compilers on those systems:
21780 Create a shared object.
21781 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
21785 Identify the versions of each tool used by the compiler, in a
21786 @code{.ident} assembler directive in the output.
21790 Refrain from adding @code{.ident} directives to the output file (this is
21793 @item -YP,@var{dirs}
21795 Search the directories @var{dirs}, and no others, for libraries
21796 specified with @option{-l}.
21798 @item -Ym,@var{dir}
21800 Look in the directory @var{dir} to find the M4 preprocessor.
21801 The assembler uses this option.
21802 @c This is supposed to go with a -Yd for predefined M4 macro files, but
21803 @c the generic assembler that comes with Solaris takes just -Ym.
21806 @node TILE-Gx Options
21807 @subsection TILE-Gx Options
21808 @cindex TILE-Gx options
21810 These @samp{-m} options are supported on the TILE-Gx:
21813 @item -mcmodel=small
21814 @opindex mcmodel=small
21815 Generate code for the small model. The distance for direct calls is
21816 limited to 500M in either direction. PC-relative addresses are 32
21817 bits. Absolute addresses support the full address range.
21819 @item -mcmodel=large
21820 @opindex mcmodel=large
21821 Generate code for the large model. There is no limitation on call
21822 distance, pc-relative addresses, or absolute addresses.
21824 @item -mcpu=@var{name}
21826 Selects the type of CPU to be targeted. Currently the only supported
21827 type is @samp{tilegx}.
21833 Generate code for a 32-bit or 64-bit environment. The 32-bit
21834 environment sets int, long, and pointer to 32 bits. The 64-bit
21835 environment sets int to 32 bits and long and pointer to 64 bits.
21838 @itemx -mlittle-endian
21839 @opindex mbig-endian
21840 @opindex mlittle-endian
21841 Generate code in big/little endian mode, respectively.
21844 @node TILEPro Options
21845 @subsection TILEPro Options
21846 @cindex TILEPro options
21848 These @samp{-m} options are supported on the TILEPro:
21851 @item -mcpu=@var{name}
21853 Selects the type of CPU to be targeted. Currently the only supported
21854 type is @samp{tilepro}.
21858 Generate code for a 32-bit environment, which sets int, long, and
21859 pointer to 32 bits. This is the only supported behavior so the flag
21860 is essentially ignored.
21864 @subsection V850 Options
21865 @cindex V850 Options
21867 These @samp{-m} options are defined for V850 implementations:
21871 @itemx -mno-long-calls
21872 @opindex mlong-calls
21873 @opindex mno-long-calls
21874 Treat all calls as being far away (near). If calls are assumed to be
21875 far away, the compiler always loads the function's address into a
21876 register, and calls indirect through the pointer.
21882 Do not optimize (do optimize) basic blocks that use the same index
21883 pointer 4 or more times to copy pointer into the @code{ep} register, and
21884 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
21885 option is on by default if you optimize.
21887 @item -mno-prolog-function
21888 @itemx -mprolog-function
21889 @opindex mno-prolog-function
21890 @opindex mprolog-function
21891 Do not use (do use) external functions to save and restore registers
21892 at the prologue and epilogue of a function. The external functions
21893 are slower, but use less code space if more than one function saves
21894 the same number of registers. The @option{-mprolog-function} option
21895 is on by default if you optimize.
21899 Try to make the code as small as possible. At present, this just turns
21900 on the @option{-mep} and @option{-mprolog-function} options.
21902 @item -mtda=@var{n}
21904 Put static or global variables whose size is @var{n} bytes or less into
21905 the tiny data area that register @code{ep} points to. The tiny data
21906 area can hold up to 256 bytes in total (128 bytes for byte references).
21908 @item -msda=@var{n}
21910 Put static or global variables whose size is @var{n} bytes or less into
21911 the small data area that register @code{gp} points to. The small data
21912 area can hold up to 64 kilobytes.
21914 @item -mzda=@var{n}
21916 Put static or global variables whose size is @var{n} bytes or less into
21917 the first 32 kilobytes of memory.
21921 Specify that the target processor is the V850.
21925 Specify that the target processor is the V850E3V5. The preprocessor
21926 constant @code{__v850e3v5__} is defined if this option is used.
21930 Specify that the target processor is the V850E3V5. This is an alias for
21931 the @option{-mv850e3v5} option.
21935 Specify that the target processor is the V850E2V3. The preprocessor
21936 constant @code{__v850e2v3__} is defined if this option is used.
21940 Specify that the target processor is the V850E2. The preprocessor
21941 constant @code{__v850e2__} is defined if this option is used.
21945 Specify that the target processor is the V850E1. The preprocessor
21946 constants @code{__v850e1__} and @code{__v850e__} are defined if
21947 this option is used.
21951 Specify that the target processor is the V850ES. This is an alias for
21952 the @option{-mv850e1} option.
21956 Specify that the target processor is the V850E@. The preprocessor
21957 constant @code{__v850e__} is defined if this option is used.
21959 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
21960 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
21961 are defined then a default target processor is chosen and the
21962 relevant @samp{__v850*__} preprocessor constant is defined.
21964 The preprocessor constants @code{__v850} and @code{__v851__} are always
21965 defined, regardless of which processor variant is the target.
21967 @item -mdisable-callt
21968 @itemx -mno-disable-callt
21969 @opindex mdisable-callt
21970 @opindex mno-disable-callt
21971 This option suppresses generation of the @code{CALLT} instruction for the
21972 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
21975 This option is enabled by default when the RH850 ABI is
21976 in use (see @option{-mrh850-abi}), and disabled by default when the
21977 GCC ABI is in use. If @code{CALLT} instructions are being generated
21978 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
21984 Pass on (or do not pass on) the @option{-mrelax} command-line option
21988 @itemx -mno-long-jumps
21989 @opindex mlong-jumps
21990 @opindex mno-long-jumps
21991 Disable (or re-enable) the generation of PC-relative jump instructions.
21994 @itemx -mhard-float
21995 @opindex msoft-float
21996 @opindex mhard-float
21997 Disable (or re-enable) the generation of hardware floating point
21998 instructions. This option is only significant when the target
21999 architecture is @samp{V850E2V3} or higher. If hardware floating point
22000 instructions are being generated then the C preprocessor symbol
22001 @code{__FPU_OK__} is defined, otherwise the symbol
22002 @code{__NO_FPU__} is defined.
22006 Enables the use of the e3v5 LOOP instruction. The use of this
22007 instruction is not enabled by default when the e3v5 architecture is
22008 selected because its use is still experimental.
22012 @opindex mrh850-abi
22014 Enables support for the RH850 version of the V850 ABI. This is the
22015 default. With this version of the ABI the following rules apply:
22019 Integer sized structures and unions are returned via a memory pointer
22020 rather than a register.
22023 Large structures and unions (more than 8 bytes in size) are passed by
22027 Functions are aligned to 16-bit boundaries.
22030 The @option{-m8byte-align} command-line option is supported.
22033 The @option{-mdisable-callt} command-line option is enabled by
22034 default. The @option{-mno-disable-callt} command-line option is not
22038 When this version of the ABI is enabled the C preprocessor symbol
22039 @code{__V850_RH850_ABI__} is defined.
22043 Enables support for the old GCC version of the V850 ABI. With this
22044 version of the ABI the following rules apply:
22048 Integer sized structures and unions are returned in register @code{r10}.
22051 Large structures and unions (more than 8 bytes in size) are passed by
22055 Functions are aligned to 32-bit boundaries, unless optimizing for
22059 The @option{-m8byte-align} command-line option is not supported.
22062 The @option{-mdisable-callt} command-line option is supported but not
22063 enabled by default.
22066 When this version of the ABI is enabled the C preprocessor symbol
22067 @code{__V850_GCC_ABI__} is defined.
22069 @item -m8byte-align
22070 @itemx -mno-8byte-align
22071 @opindex m8byte-align
22072 @opindex mno-8byte-align
22073 Enables support for @code{double} and @code{long long} types to be
22074 aligned on 8-byte boundaries. The default is to restrict the
22075 alignment of all objects to at most 4-bytes. When
22076 @option{-m8byte-align} is in effect the C preprocessor symbol
22077 @code{__V850_8BYTE_ALIGN__} is defined.
22080 @opindex mbig-switch
22081 Generate code suitable for big switch tables. Use this option only if
22082 the assembler/linker complain about out of range branches within a switch
22087 This option causes r2 and r5 to be used in the code generated by
22088 the compiler. This setting is the default.
22090 @item -mno-app-regs
22091 @opindex mno-app-regs
22092 This option causes r2 and r5 to be treated as fixed registers.
22097 @subsection VAX Options
22098 @cindex VAX options
22100 These @samp{-m} options are defined for the VAX:
22105 Do not output certain jump instructions (@code{aobleq} and so on)
22106 that the Unix assembler for the VAX cannot handle across long
22111 Do output those jump instructions, on the assumption that the
22112 GNU assembler is being used.
22116 Output code for G-format floating-point numbers instead of D-format.
22119 @node Visium Options
22120 @subsection Visium Options
22121 @cindex Visium options
22127 A program which performs file I/O and is destined to run on an MCM target
22128 should be linked with this option. It causes the libraries libc.a and
22129 libdebug.a to be linked. The program should be run on the target under
22130 the control of the GDB remote debugging stub.
22134 A program which performs file I/O and is destined to run on the simulator
22135 should be linked with option. This causes libraries libc.a and libsim.a to
22139 @itemx -mhard-float
22141 @opindex mhard-float
22142 Generate code containing floating-point instructions. This is the
22146 @itemx -msoft-float
22148 @opindex msoft-float
22149 Generate code containing library calls for floating-point.
22151 @option{-msoft-float} changes the calling convention in the output file;
22152 therefore, it is only useful if you compile @emph{all} of a program with
22153 this option. In particular, you need to compile @file{libgcc.a}, the
22154 library that comes with GCC, with @option{-msoft-float} in order for
22157 @item -mcpu=@var{cpu_type}
22159 Set the instruction set, register set, and instruction scheduling parameters
22160 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22161 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22163 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22165 By default (unless configured otherwise), GCC generates code for the GR5
22166 variant of the Visium architecture.
22168 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22169 architecture. The only difference from GR5 code is that the compiler will
22170 generate block move instructions.
22172 @item -mtune=@var{cpu_type}
22174 Set the instruction scheduling parameters for machine type @var{cpu_type},
22175 but do not set the instruction set or register set that the option
22176 @option{-mcpu=@var{cpu_type}} would.
22180 Generate code for the supervisor mode, where there are no restrictions on
22181 the access to general registers. This is the default.
22184 @opindex muser-mode
22185 Generate code for the user mode, where the access to some general registers
22186 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22187 mode; on the GR6, only registers r29 to r31 are affected.
22191 @subsection VMS Options
22193 These @samp{-m} options are defined for the VMS implementations:
22196 @item -mvms-return-codes
22197 @opindex mvms-return-codes
22198 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22199 condition (e.g.@ error) codes.
22201 @item -mdebug-main=@var{prefix}
22202 @opindex mdebug-main=@var{prefix}
22203 Flag the first routine whose name starts with @var{prefix} as the main
22204 routine for the debugger.
22208 Default to 64-bit memory allocation routines.
22210 @item -mpointer-size=@var{size}
22211 @opindex mpointer-size=@var{size}
22212 Set the default size of pointers. Possible options for @var{size} are
22213 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22214 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22215 The later option disables @code{pragma pointer_size}.
22218 @node VxWorks Options
22219 @subsection VxWorks Options
22220 @cindex VxWorks Options
22222 The options in this section are defined for all VxWorks targets.
22223 Options specific to the target hardware are listed with the other
22224 options for that target.
22229 GCC can generate code for both VxWorks kernels and real time processes
22230 (RTPs). This option switches from the former to the latter. It also
22231 defines the preprocessor macro @code{__RTP__}.
22234 @opindex non-static
22235 Link an RTP executable against shared libraries rather than static
22236 libraries. The options @option{-static} and @option{-shared} can
22237 also be used for RTPs (@pxref{Link Options}); @option{-static}
22244 These options are passed down to the linker. They are defined for
22245 compatibility with Diab.
22248 @opindex Xbind-lazy
22249 Enable lazy binding of function calls. This option is equivalent to
22250 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22254 Disable lazy binding of function calls. This option is the default and
22255 is defined for compatibility with Diab.
22259 @subsection x86 Options
22260 @cindex x86 Options
22262 These @samp{-m} options are defined for the x86 family of computers.
22266 @item -march=@var{cpu-type}
22268 Generate instructions for the machine type @var{cpu-type}. In contrast to
22269 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22270 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22271 to generate code that may not run at all on processors other than the one
22272 indicated. Specifying @option{-march=@var{cpu-type}} implies
22273 @option{-mtune=@var{cpu-type}}.
22275 The choices for @var{cpu-type} are:
22279 This selects the CPU to generate code for at compilation time by determining
22280 the processor type of the compiling machine. Using @option{-march=native}
22281 enables all instruction subsets supported by the local machine (hence
22282 the result might not run on different machines). Using @option{-mtune=native}
22283 produces code optimized for the local machine under the constraints
22284 of the selected instruction set.
22287 Original Intel i386 CPU@.
22290 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22294 Intel Pentium CPU with no MMX support.
22297 Intel MCU, based on Intel Pentium CPU.
22300 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22303 Intel Pentium Pro CPU@.
22306 When used with @option{-march}, the Pentium Pro
22307 instruction set is used, so the code runs on all i686 family chips.
22308 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22311 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22316 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22320 Intel Pentium M; low-power version of Intel Pentium III CPU
22321 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22325 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22328 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22332 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22333 SSE2 and SSE3 instruction set support.
22336 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22337 instruction set support.
22340 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22341 SSE4.1, SSE4.2 and POPCNT instruction set support.
22344 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22345 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22348 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22349 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22352 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22353 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22354 instruction set support.
22357 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22358 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22359 BMI, BMI2 and F16C instruction set support.
22362 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22363 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22364 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22367 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22368 instruction set support.
22371 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22372 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22375 Intel Knight's Landing 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, AVX512F, AVX512PF, AVX512ER and
22378 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 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22440 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22441 instruction set extensions.)
22444 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22445 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22446 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22449 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22453 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22454 instruction set support.
22457 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22458 implemented for this chip.)
22461 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22463 implemented for this chip.)
22466 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22469 @item -mtune=@var{cpu-type}
22471 Tune to @var{cpu-type} everything applicable about the generated code, except
22472 for the ABI and the set of available instructions.
22473 While picking a specific @var{cpu-type} schedules things appropriately
22474 for that particular chip, the compiler does not generate any code that
22475 cannot run on the default machine type unless you use a
22476 @option{-march=@var{cpu-type}} option.
22477 For example, if GCC is configured for i686-pc-linux-gnu
22478 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22479 but still runs on i686 machines.
22481 The choices for @var{cpu-type} are the same as for @option{-march}.
22482 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22486 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22487 If you know the CPU on which your code will run, then you should use
22488 the corresponding @option{-mtune} or @option{-march} option instead of
22489 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22490 of your application will have, then you should use this option.
22492 As new processors are deployed in the marketplace, the behavior of this
22493 option will change. Therefore, if you upgrade to a newer version of
22494 GCC, code generation controlled by this option will change to reflect
22496 that are most common at the time that version of GCC is released.
22498 There is no @option{-march=generic} option because @option{-march}
22499 indicates the instruction set the compiler can use, and there is no
22500 generic instruction set applicable to all processors. In contrast,
22501 @option{-mtune} indicates the processor (or, in this case, collection of
22502 processors) for which the code is optimized.
22505 Produce code optimized for the most current Intel processors, which are
22506 Haswell and Silvermont for this version of GCC. If you know the CPU
22507 on which your code will run, then you should use the corresponding
22508 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22509 But, if you want your application performs better on both Haswell and
22510 Silvermont, then you should use this option.
22512 As new Intel processors are deployed in the marketplace, the behavior of
22513 this option will change. Therefore, if you upgrade to a newer version of
22514 GCC, code generation controlled by this option will change to reflect
22515 the most current Intel processors at the time that version of GCC is
22518 There is no @option{-march=intel} option because @option{-march} indicates
22519 the instruction set the compiler can use, and there is no common
22520 instruction set applicable to all processors. In contrast,
22521 @option{-mtune} indicates the processor (or, in this case, collection of
22522 processors) for which the code is optimized.
22525 @item -mcpu=@var{cpu-type}
22527 A deprecated synonym for @option{-mtune}.
22529 @item -mfpmath=@var{unit}
22531 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22532 for @var{unit} are:
22536 Use the standard 387 floating-point coprocessor present on the majority of chips and
22537 emulated otherwise. Code compiled with this option runs almost everywhere.
22538 The temporary results are computed in 80-bit precision instead of the precision
22539 specified by the type, resulting in slightly different results compared to most
22540 of other chips. See @option{-ffloat-store} for more detailed description.
22542 This is the default choice for x86-32 targets.
22545 Use scalar floating-point instructions present in the SSE instruction set.
22546 This instruction set is supported by Pentium III and newer chips,
22547 and in the AMD line
22548 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22549 instruction set supports only single-precision arithmetic, thus the double and
22550 extended-precision arithmetic are still done using 387. A later version, present
22551 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22554 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22555 or @option{-msse2} switches to enable SSE extensions and make this option
22556 effective. For the x86-64 compiler, these extensions are enabled by default.
22558 The resulting code should be considerably faster in the majority of cases and avoid
22559 the numerical instability problems of 387 code, but may break some existing
22560 code that expects temporaries to be 80 bits.
22562 This is the default choice for the x86-64 compiler.
22567 Attempt to utilize both instruction sets at once. This effectively doubles the
22568 amount of available registers, and on chips with separate execution units for
22569 387 and SSE the execution resources too. Use this option with care, as it is
22570 still experimental, because the GCC register allocator does not model separate
22571 functional units well, resulting in unstable performance.
22574 @item -masm=@var{dialect}
22575 @opindex masm=@var{dialect}
22576 Output assembly instructions using selected @var{dialect}. Also affects
22577 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22578 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22579 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22580 not support @samp{intel}.
22583 @itemx -mno-ieee-fp
22585 @opindex mno-ieee-fp
22586 Control whether or not the compiler uses IEEE floating-point
22587 comparisons. These correctly handle the case where the result of a
22588 comparison is unordered.
22591 @opindex msoft-float
22592 Generate output containing library calls for floating point.
22594 @strong{Warning:} the requisite libraries are not part of GCC@.
22595 Normally the facilities of the machine's usual C compiler are used, but
22596 this can't be done directly in cross-compilation. You must make your
22597 own arrangements to provide suitable library functions for
22600 On machines where a function returns floating-point results in the 80387
22601 register stack, some floating-point opcodes may be emitted even if
22602 @option{-msoft-float} is used.
22604 @item -mno-fp-ret-in-387
22605 @opindex mno-fp-ret-in-387
22606 Do not use the FPU registers for return values of functions.
22608 The usual calling convention has functions return values of types
22609 @code{float} and @code{double} in an FPU register, even if there
22610 is no FPU@. The idea is that the operating system should emulate
22613 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22614 in ordinary CPU registers instead.
22616 @item -mno-fancy-math-387
22617 @opindex mno-fancy-math-387
22618 Some 387 emulators do not support the @code{sin}, @code{cos} and
22619 @code{sqrt} instructions for the 387. Specify this option to avoid
22620 generating those instructions. This option is the default on
22621 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22622 indicates that the target CPU always has an FPU and so the
22623 instruction does not need emulation. These
22624 instructions are not generated unless you also use the
22625 @option{-funsafe-math-optimizations} switch.
22627 @item -malign-double
22628 @itemx -mno-align-double
22629 @opindex malign-double
22630 @opindex mno-align-double
22631 Control whether GCC aligns @code{double}, @code{long double}, and
22632 @code{long long} variables on a two-word boundary or a one-word
22633 boundary. Aligning @code{double} variables on a two-word boundary
22634 produces code that runs somewhat faster on a Pentium at the
22635 expense of more memory.
22637 On x86-64, @option{-malign-double} is enabled by default.
22639 @strong{Warning:} if you use the @option{-malign-double} switch,
22640 structures containing the above types are aligned differently than
22641 the published application binary interface specifications for the x86-32
22642 and are not binary compatible with structures in code compiled
22643 without that switch.
22645 @item -m96bit-long-double
22646 @itemx -m128bit-long-double
22647 @opindex m96bit-long-double
22648 @opindex m128bit-long-double
22649 These switches control the size of @code{long double} type. The x86-32
22650 application binary interface specifies the size to be 96 bits,
22651 so @option{-m96bit-long-double} is the default in 32-bit mode.
22653 Modern architectures (Pentium and newer) prefer @code{long double}
22654 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22655 conforming to the ABI, this is not possible. So specifying
22656 @option{-m128bit-long-double} aligns @code{long double}
22657 to a 16-byte boundary by padding the @code{long double} with an additional
22660 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22661 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22663 Notice that neither of these options enable any extra precision over the x87
22664 standard of 80 bits for a @code{long double}.
22666 @strong{Warning:} if you override the default value for your target ABI, this
22667 changes the size of
22668 structures and arrays containing @code{long double} variables,
22669 as well as modifying the function calling convention for functions taking
22670 @code{long double}. Hence they are not binary-compatible
22671 with code compiled without that switch.
22673 @item -mlong-double-64
22674 @itemx -mlong-double-80
22675 @itemx -mlong-double-128
22676 @opindex mlong-double-64
22677 @opindex mlong-double-80
22678 @opindex mlong-double-128
22679 These switches control the size of @code{long double} type. A size
22680 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22681 type. This is the default for 32-bit Bionic C library. A size
22682 of 128 bits makes the @code{long double} type equivalent to the
22683 @code{__float128} type. This is the default for 64-bit Bionic C library.
22685 @strong{Warning:} if you override the default value for your target ABI, this
22686 changes the size of
22687 structures and arrays containing @code{long double} variables,
22688 as well as modifying the function calling convention for functions taking
22689 @code{long double}. Hence they are not binary-compatible
22690 with code compiled without that switch.
22692 @item -malign-data=@var{type}
22693 @opindex malign-data
22694 Control how GCC aligns variables. Supported values for @var{type} are
22695 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22696 and earlier, @samp{abi} uses alignment value as specified by the
22697 psABI, and @samp{cacheline} uses increased alignment value to match
22698 the cache line size. @samp{compat} is the default.
22700 @item -mlarge-data-threshold=@var{threshold}
22701 @opindex mlarge-data-threshold
22702 When @option{-mcmodel=medium} is specified, data objects larger than
22703 @var{threshold} are placed in the large data section. This value must be the
22704 same across all objects linked into the binary, and defaults to 65535.
22708 Use a different function-calling convention, in which functions that
22709 take a fixed number of arguments return with the @code{ret @var{num}}
22710 instruction, which pops their arguments while returning. This saves one
22711 instruction in the caller since there is no need to pop the arguments
22714 You can specify that an individual function is called with this calling
22715 sequence with the function attribute @code{stdcall}. You can also
22716 override the @option{-mrtd} option by using the function attribute
22717 @code{cdecl}. @xref{Function Attributes}.
22719 @strong{Warning:} this calling convention is incompatible with the one
22720 normally used on Unix, so you cannot use it if you need to call
22721 libraries compiled with the Unix compiler.
22723 Also, you must provide function prototypes for all functions that
22724 take variable numbers of arguments (including @code{printf});
22725 otherwise incorrect code is generated for calls to those
22728 In addition, seriously incorrect code results if you call a
22729 function with too many arguments. (Normally, extra arguments are
22730 harmlessly ignored.)
22732 @item -mregparm=@var{num}
22734 Control how many registers are used to pass integer arguments. By
22735 default, no registers are used to pass arguments, and at most 3
22736 registers can be used. You can control this behavior for a specific
22737 function by using the function attribute @code{regparm}.
22738 @xref{Function Attributes}.
22740 @strong{Warning:} if you use this switch, and
22741 @var{num} is nonzero, then you must build all modules with the same
22742 value, including any libraries. This includes the system libraries and
22746 @opindex msseregparm
22747 Use SSE register passing conventions for float and double arguments
22748 and return values. You can control this behavior for a specific
22749 function by using the function attribute @code{sseregparm}.
22750 @xref{Function Attributes}.
22752 @strong{Warning:} if you use this switch then you must build all
22753 modules with the same value, including any libraries. This includes
22754 the system libraries and startup modules.
22756 @item -mvect8-ret-in-mem
22757 @opindex mvect8-ret-in-mem
22758 Return 8-byte vectors in memory instead of MMX registers. This is the
22759 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
22760 Studio compilers until version 12. Later compiler versions (starting
22761 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
22762 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
22763 you need to remain compatible with existing code produced by those
22764 previous compiler versions or older versions of GCC@.
22773 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
22774 is specified, the significands of results of floating-point operations are
22775 rounded to 24 bits (single precision); @option{-mpc64} rounds the
22776 significands of results of floating-point operations to 53 bits (double
22777 precision) and @option{-mpc80} rounds the significands of results of
22778 floating-point operations to 64 bits (extended double precision), which is
22779 the default. When this option is used, floating-point operations in higher
22780 precisions are not available to the programmer without setting the FPU
22781 control word explicitly.
22783 Setting the rounding of floating-point operations to less than the default
22784 80 bits can speed some programs by 2% or more. Note that some mathematical
22785 libraries assume that extended-precision (80-bit) floating-point operations
22786 are enabled by default; routines in such libraries could suffer significant
22787 loss of accuracy, typically through so-called ``catastrophic cancellation'',
22788 when this option is used to set the precision to less than extended precision.
22790 @item -mstackrealign
22791 @opindex mstackrealign
22792 Realign the stack at entry. On the x86, the @option{-mstackrealign}
22793 option generates an alternate prologue and epilogue that realigns the
22794 run-time stack if necessary. This supports mixing legacy codes that keep
22795 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
22796 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
22797 applicable to individual functions.
22799 @item -mpreferred-stack-boundary=@var{num}
22800 @opindex mpreferred-stack-boundary
22801 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
22802 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
22803 the default is 4 (16 bytes or 128 bits).
22805 @strong{Warning:} When generating code for the x86-64 architecture with
22806 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
22807 used to keep the stack boundary aligned to 8 byte boundary. Since
22808 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
22809 intended to be used in controlled environment where stack space is
22810 important limitation. This option leads to wrong code when functions
22811 compiled with 16 byte stack alignment (such as functions from a standard
22812 library) are called with misaligned stack. In this case, SSE
22813 instructions may lead to misaligned memory access traps. In addition,
22814 variable arguments are handled incorrectly for 16 byte aligned
22815 objects (including x87 long double and __int128), leading to wrong
22816 results. You must build all modules with
22817 @option{-mpreferred-stack-boundary=3}, including any libraries. This
22818 includes the system libraries and startup modules.
22820 @item -mincoming-stack-boundary=@var{num}
22821 @opindex mincoming-stack-boundary
22822 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
22823 boundary. If @option{-mincoming-stack-boundary} is not specified,
22824 the one specified by @option{-mpreferred-stack-boundary} is used.
22826 On Pentium and Pentium Pro, @code{double} and @code{long double} values
22827 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
22828 suffer significant run time performance penalties. On Pentium III, the
22829 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
22830 properly if it is not 16-byte aligned.
22832 To ensure proper alignment of this values on the stack, the stack boundary
22833 must be as aligned as that required by any value stored on the stack.
22834 Further, every function must be generated such that it keeps the stack
22835 aligned. Thus calling a function compiled with a higher preferred
22836 stack boundary from a function compiled with a lower preferred stack
22837 boundary most likely misaligns the stack. It is recommended that
22838 libraries that use callbacks always use the default setting.
22840 This extra alignment does consume extra stack space, and generally
22841 increases code size. Code that is sensitive to stack space usage, such
22842 as embedded systems and operating system kernels, may want to reduce the
22843 preferred alignment to @option{-mpreferred-stack-boundary=2}.
22889 @opindex mclfushopt
22907 @itemx -mprefetchwt1
22908 @opindex mprefetchwt1
22959 These switches enable the use of instructions in the MMX, SSE,
22960 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
22961 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
22962 BMI, BMI2, FXSR, XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX or 3DNow!@:
22963 extended instruction sets. Each has a corresponding @option{-mno-} option
22964 to disable use of these instructions.
22966 These extensions are also available as built-in functions: see
22967 @ref{x86 Built-in Functions}, for details of the functions enabled and
22968 disabled by these switches.
22970 To generate SSE/SSE2 instructions automatically from floating-point
22971 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
22973 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
22974 generates new AVX instructions or AVX equivalence for all SSEx instructions
22977 These options enable GCC to use these extended instructions in
22978 generated code, even without @option{-mfpmath=sse}. Applications that
22979 perform run-time CPU detection must compile separate files for each
22980 supported architecture, using the appropriate flags. In particular,
22981 the file containing the CPU detection code should be compiled without
22984 @item -mdump-tune-features
22985 @opindex mdump-tune-features
22986 This option instructs GCC to dump the names of the x86 performance
22987 tuning features and default settings. The names can be used in
22988 @option{-mtune-ctrl=@var{feature-list}}.
22990 @item -mtune-ctrl=@var{feature-list}
22991 @opindex mtune-ctrl=@var{feature-list}
22992 This option is used to do fine grain control of x86 code generation features.
22993 @var{feature-list} is a comma separated list of @var{feature} names. See also
22994 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
22995 on if it is not preceded with @samp{^}, otherwise, it is turned off.
22996 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
22997 developers. Using it may lead to code paths not covered by testing and can
22998 potentially result in compiler ICEs or runtime errors.
23001 @opindex mno-default
23002 This option instructs GCC to turn off all tunable features. See also
23003 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23007 This option instructs GCC to emit a @code{cld} instruction in the prologue
23008 of functions that use string instructions. String instructions depend on
23009 the DF flag to select between autoincrement or autodecrement mode. While the
23010 ABI specifies the DF flag to be cleared on function entry, some operating
23011 systems violate this specification by not clearing the DF flag in their
23012 exception dispatchers. The exception handler can be invoked with the DF flag
23013 set, which leads to wrong direction mode when string instructions are used.
23014 This option can be enabled by default on 32-bit x86 targets by configuring
23015 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23016 instructions can be suppressed with the @option{-mno-cld} compiler option
23020 @opindex mvzeroupper
23021 This option instructs GCC to emit a @code{vzeroupper} instruction
23022 before a transfer of control flow out of the function to minimize
23023 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23026 @item -mprefer-avx128
23027 @opindex mprefer-avx128
23028 This option instructs GCC to use 128-bit AVX instructions instead of
23029 256-bit AVX instructions in the auto-vectorizer.
23033 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23034 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23035 (or oword) data types.
23036 This is useful for high-resolution counters that can be updated
23037 by multiple processors (or cores). This instruction is generated as part of
23038 atomic built-in functions: see @ref{__sync Builtins} or
23039 @ref{__atomic Builtins} for details.
23043 This option enables generation of @code{SAHF} instructions in 64-bit code.
23044 Early Intel Pentium 4 CPUs with Intel 64 support,
23045 prior to the introduction of Pentium 4 G1 step in December 2005,
23046 lacked the @code{LAHF} and @code{SAHF} instructions
23047 which are supported by AMD64.
23048 These are load and store instructions, respectively, for certain status flags.
23049 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23050 @code{drem}, and @code{remainder} built-in functions;
23051 see @ref{Other Builtins} for details.
23055 This option enables use of the @code{movbe} instruction to implement
23056 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23060 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23061 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23062 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23066 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23067 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23068 with an additional Newton-Raphson step
23069 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23070 (and their vectorized
23071 variants) for single-precision floating-point arguments. These instructions
23072 are generated only when @option{-funsafe-math-optimizations} is enabled
23073 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
23074 Note that while the throughput of the sequence is higher than the throughput
23075 of the non-reciprocal instruction, the precision of the sequence can be
23076 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23078 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23079 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23080 combination), and doesn't need @option{-mrecip}.
23082 Also note that GCC emits the above sequence with additional Newton-Raphson step
23083 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23084 already with @option{-ffast-math} (or the above option combination), and
23085 doesn't need @option{-mrecip}.
23087 @item -mrecip=@var{opt}
23088 @opindex mrecip=opt
23089 This option controls which reciprocal estimate instructions
23090 may be used. @var{opt} is a comma-separated list of options, which may
23091 be preceded by a @samp{!} to invert the option:
23095 Enable all estimate instructions.
23098 Enable the default instructions, equivalent to @option{-mrecip}.
23101 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23104 Enable the approximation for scalar division.
23107 Enable the approximation for vectorized division.
23110 Enable the approximation for scalar square root.
23113 Enable the approximation for vectorized square root.
23116 So, for example, @option{-mrecip=all,!sqrt} enables
23117 all of the reciprocal approximations, except for square root.
23119 @item -mveclibabi=@var{type}
23120 @opindex mveclibabi
23121 Specifies the ABI type to use for vectorizing intrinsics using an
23122 external library. Supported values for @var{type} are @samp{svml}
23123 for the Intel short
23124 vector math library and @samp{acml} for the AMD math core library.
23125 To use this option, both @option{-ftree-vectorize} and
23126 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23127 ABI-compatible library must be specified at link time.
23129 GCC currently emits calls to @code{vmldExp2},
23130 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23131 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23132 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23133 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23134 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23135 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23136 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23137 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23138 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23139 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23140 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23141 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23142 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23143 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23144 when @option{-mveclibabi=acml} is used.
23146 @item -mabi=@var{name}
23148 Generate code for the specified calling convention. Permissible values
23149 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23150 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23151 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23152 You can control this behavior for specific functions by
23153 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23154 @xref{Function Attributes}.
23156 @item -mtls-dialect=@var{type}
23157 @opindex mtls-dialect
23158 Generate code to access thread-local storage using the @samp{gnu} or
23159 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23160 @samp{gnu2} is more efficient, but it may add compile- and run-time
23161 requirements that cannot be satisfied on all systems.
23164 @itemx -mno-push-args
23165 @opindex mpush-args
23166 @opindex mno-push-args
23167 Use PUSH operations to store outgoing parameters. This method is shorter
23168 and usually equally fast as method using SUB/MOV operations and is enabled
23169 by default. In some cases disabling it may improve performance because of
23170 improved scheduling and reduced dependencies.
23172 @item -maccumulate-outgoing-args
23173 @opindex maccumulate-outgoing-args
23174 If enabled, the maximum amount of space required for outgoing arguments is
23175 computed in the function prologue. This is faster on most modern CPUs
23176 because of reduced dependencies, improved scheduling and reduced stack usage
23177 when the preferred stack boundary is not equal to 2. The drawback is a notable
23178 increase in code size. This switch implies @option{-mno-push-args}.
23182 Support thread-safe exception handling on MinGW. Programs that rely
23183 on thread-safe exception handling must compile and link all code with the
23184 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23185 @option{-D_MT}; when linking, it links in a special thread helper library
23186 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23188 @item -mno-align-stringops
23189 @opindex mno-align-stringops
23190 Do not align the destination of inlined string operations. This switch reduces
23191 code size and improves performance in case the destination is already aligned,
23192 but GCC doesn't know about it.
23194 @item -minline-all-stringops
23195 @opindex minline-all-stringops
23196 By default GCC inlines string operations only when the destination is
23197 known to be aligned to least a 4-byte boundary.
23198 This enables more inlining and increases code
23199 size, but may improve performance of code that depends on fast
23200 @code{memcpy}, @code{strlen},
23201 and @code{memset} for short lengths.
23203 @item -minline-stringops-dynamically
23204 @opindex minline-stringops-dynamically
23205 For string operations of unknown size, use run-time checks with
23206 inline code for small blocks and a library call for large blocks.
23208 @item -mstringop-strategy=@var{alg}
23209 @opindex mstringop-strategy=@var{alg}
23210 Override the internal decision heuristic for the particular algorithm to use
23211 for inlining string operations. The allowed values for @var{alg} are:
23217 Expand using i386 @code{rep} prefix of the specified size.
23221 @itemx unrolled_loop
23222 Expand into an inline loop.
23225 Always use a library call.
23228 @item -mmemcpy-strategy=@var{strategy}
23229 @opindex mmemcpy-strategy=@var{strategy}
23230 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23231 should be inlined and what inline algorithm to use when the expected size
23232 of the copy operation is known. @var{strategy}
23233 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23234 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23235 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23236 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23237 in the list must be specified in increasing order. The minimal byte size for
23238 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23241 @item -mmemset-strategy=@var{strategy}
23242 @opindex mmemset-strategy=@var{strategy}
23243 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23244 @code{__builtin_memset} expansion.
23246 @item -momit-leaf-frame-pointer
23247 @opindex momit-leaf-frame-pointer
23248 Don't keep the frame pointer in a register for leaf functions. This
23249 avoids the instructions to save, set up, and restore frame pointers and
23250 makes an extra register available in leaf functions. The option
23251 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23252 which might make debugging harder.
23254 @item -mtls-direct-seg-refs
23255 @itemx -mno-tls-direct-seg-refs
23256 @opindex mtls-direct-seg-refs
23257 Controls whether TLS variables may be accessed with offsets from the
23258 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23259 or whether the thread base pointer must be added. Whether or not this
23260 is valid depends on the operating system, and whether it maps the
23261 segment to cover the entire TLS area.
23263 For systems that use the GNU C Library, the default is on.
23266 @itemx -mno-sse2avx
23268 Specify that the assembler should encode SSE instructions with VEX
23269 prefix. The option @option{-mavx} turns this on by default.
23274 If profiling is active (@option{-pg}), put the profiling
23275 counter call before the prologue.
23276 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23277 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23279 @item -mrecord-mcount
23280 @itemx -mno-record-mcount
23281 @opindex mrecord-mcount
23282 If profiling is active (@option{-pg}), generate a __mcount_loc section
23283 that contains pointers to each profiling call. This is useful for
23284 automatically patching and out calls.
23287 @itemx -mno-nop-mcount
23288 @opindex mnop-mcount
23289 If profiling is active (@option{-pg}), generate the calls to
23290 the profiling functions as nops. This is useful when they
23291 should be patched in later dynamically. This is likely only
23292 useful together with @option{-mrecord-mcount}.
23294 @item -mskip-rax-setup
23295 @itemx -mno-skip-rax-setup
23296 @opindex mskip-rax-setup
23297 When generating code for the x86-64 architecture with SSE extensions
23298 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23299 register when there are no variable arguments passed in vector registers.
23301 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23302 saving vector registers on stack when passing variable arguments, the
23303 impacts of this option are callees may waste some stack space,
23304 misbehave or jump to a random location. GCC 4.4 or newer don't have
23305 those issues, regardless the RAX register value.
23308 @itemx -mno-8bit-idiv
23309 @opindex m8bit-idiv
23310 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23311 much faster than 32-bit/64-bit integer divide. This option generates a
23312 run-time check. If both dividend and divisor are within range of 0
23313 to 255, 8-bit unsigned integer divide is used instead of
23314 32-bit/64-bit integer divide.
23316 @item -mavx256-split-unaligned-load
23317 @itemx -mavx256-split-unaligned-store
23318 @opindex mavx256-split-unaligned-load
23319 @opindex mavx256-split-unaligned-store
23320 Split 32-byte AVX unaligned load and store.
23322 @item -mstack-protector-guard=@var{guard}
23323 @opindex mstack-protector-guard=@var{guard}
23324 Generate stack protection code using canary at @var{guard}. Supported
23325 locations are @samp{global} for global canary or @samp{tls} for per-thread
23326 canary in the TLS block (the default). This option has effect only when
23327 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23331 These @samp{-m} switches are supported in addition to the above
23332 on x86-64 processors in 64-bit environments.
23345 Generate code for a 16-bit, 32-bit or 64-bit environment.
23346 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23348 generates code that runs on any i386 system.
23350 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23351 types to 64 bits, and generates code for the x86-64 architecture.
23352 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23353 and @option{-mdynamic-no-pic} options.
23355 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23357 generates code for the x86-64 architecture.
23359 The @option{-m16} option is the same as @option{-m32}, except for that
23360 it outputs the @code{.code16gcc} assembly directive at the beginning of
23361 the assembly output so that the binary can run in 16-bit mode.
23363 The @option{-miamcu} option generates code which conforms to Intel MCU
23364 psABI. It requires the @option{-m32} option to be turned on.
23366 @item -mno-red-zone
23367 @opindex mno-red-zone
23368 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23369 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23370 stack pointer that is not modified by signal or interrupt handlers
23371 and therefore can be used for temporary data without adjusting the stack
23372 pointer. The flag @option{-mno-red-zone} disables this red zone.
23374 @item -mcmodel=small
23375 @opindex mcmodel=small
23376 Generate code for the small code model: the program and its symbols must
23377 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23378 Programs can be statically or dynamically linked. This is the default
23381 @item -mcmodel=kernel
23382 @opindex mcmodel=kernel
23383 Generate code for the kernel code model. The kernel runs in the
23384 negative 2 GB of the address space.
23385 This model has to be used for Linux kernel code.
23387 @item -mcmodel=medium
23388 @opindex mcmodel=medium
23389 Generate code for the medium model: the program is linked in the lower 2
23390 GB of the address space. Small symbols are also placed there. Symbols
23391 with sizes larger than @option{-mlarge-data-threshold} are put into
23392 large data or BSS sections and can be located above 2GB. Programs can
23393 be statically or dynamically linked.
23395 @item -mcmodel=large
23396 @opindex mcmodel=large
23397 Generate code for the large model. This model makes no assumptions
23398 about addresses and sizes of sections.
23400 @item -maddress-mode=long
23401 @opindex maddress-mode=long
23402 Generate code for long address mode. This is only supported for 64-bit
23403 and x32 environments. It is the default address mode for 64-bit
23406 @item -maddress-mode=short
23407 @opindex maddress-mode=short
23408 Generate code for short address mode. This is only supported for 32-bit
23409 and x32 environments. It is the default address mode for 32-bit and
23413 @node x86 Windows Options
23414 @subsection x86 Windows Options
23415 @cindex x86 Windows Options
23416 @cindex Windows Options for x86
23418 These additional options are available for Microsoft Windows targets:
23424 specifies that a console application is to be generated, by
23425 instructing the linker to set the PE header subsystem type
23426 required for console applications.
23427 This option is available for Cygwin and MinGW targets and is
23428 enabled by default on those targets.
23432 This option is available for Cygwin and MinGW targets. It
23433 specifies that a DLL---a dynamic link library---is to be
23434 generated, enabling the selection of the required runtime
23435 startup object and entry point.
23437 @item -mnop-fun-dllimport
23438 @opindex mnop-fun-dllimport
23439 This option is available for Cygwin and MinGW targets. It
23440 specifies that the @code{dllimport} attribute should be ignored.
23444 This option is available for MinGW targets. It specifies
23445 that MinGW-specific thread support is to be used.
23449 This option is available for MinGW-w64 targets. It causes
23450 the @code{UNICODE} preprocessor macro to be predefined, and
23451 chooses Unicode-capable runtime startup code.
23455 This option is available for Cygwin and MinGW targets. It
23456 specifies that the typical Microsoft Windows predefined macros are to
23457 be set in the pre-processor, but does not influence the choice
23458 of runtime library/startup code.
23462 This option is available for Cygwin and MinGW targets. It
23463 specifies that a GUI application is to be generated by
23464 instructing the linker to set the PE header subsystem type
23467 @item -fno-set-stack-executable
23468 @opindex fno-set-stack-executable
23469 This option is available for MinGW targets. It specifies that
23470 the executable flag for the stack used by nested functions isn't
23471 set. This is necessary for binaries running in kernel mode of
23472 Microsoft Windows, as there the User32 API, which is used to set executable
23473 privileges, isn't available.
23475 @item -fwritable-relocated-rdata
23476 @opindex fno-writable-relocated-rdata
23477 This option is available for MinGW and Cygwin targets. It specifies
23478 that relocated-data in read-only section is put into .data
23479 section. This is a necessary for older runtimes not supporting
23480 modification of .rdata sections for pseudo-relocation.
23482 @item -mpe-aligned-commons
23483 @opindex mpe-aligned-commons
23484 This option is available for Cygwin and MinGW targets. It
23485 specifies that the GNU extension to the PE file format that
23486 permits the correct alignment of COMMON variables should be
23487 used when generating code. It is enabled by default if
23488 GCC detects that the target assembler found during configuration
23489 supports the feature.
23492 See also under @ref{x86 Options} for standard options.
23494 @node Xstormy16 Options
23495 @subsection Xstormy16 Options
23496 @cindex Xstormy16 Options
23498 These options are defined for Xstormy16:
23503 Choose startup files and linker script suitable for the simulator.
23506 @node Xtensa Options
23507 @subsection Xtensa Options
23508 @cindex Xtensa Options
23510 These options are supported for Xtensa targets:
23514 @itemx -mno-const16
23516 @opindex mno-const16
23517 Enable or disable use of @code{CONST16} instructions for loading
23518 constant values. The @code{CONST16} instruction is currently not a
23519 standard option from Tensilica. When enabled, @code{CONST16}
23520 instructions are always used in place of the standard @code{L32R}
23521 instructions. The use of @code{CONST16} is enabled by default only if
23522 the @code{L32R} instruction is not available.
23525 @itemx -mno-fused-madd
23526 @opindex mfused-madd
23527 @opindex mno-fused-madd
23528 Enable or disable use of fused multiply/add and multiply/subtract
23529 instructions in the floating-point option. This has no effect if the
23530 floating-point option is not also enabled. Disabling fused multiply/add
23531 and multiply/subtract instructions forces the compiler to use separate
23532 instructions for the multiply and add/subtract operations. This may be
23533 desirable in some cases where strict IEEE 754-compliant results are
23534 required: the fused multiply add/subtract instructions do not round the
23535 intermediate result, thereby producing results with @emph{more} bits of
23536 precision than specified by the IEEE standard. Disabling fused multiply
23537 add/subtract instructions also ensures that the program output is not
23538 sensitive to the compiler's ability to combine multiply and add/subtract
23541 @item -mserialize-volatile
23542 @itemx -mno-serialize-volatile
23543 @opindex mserialize-volatile
23544 @opindex mno-serialize-volatile
23545 When this option is enabled, GCC inserts @code{MEMW} instructions before
23546 @code{volatile} memory references to guarantee sequential consistency.
23547 The default is @option{-mserialize-volatile}. Use
23548 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23550 @item -mforce-no-pic
23551 @opindex mforce-no-pic
23552 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23553 position-independent code (PIC), this option disables PIC for compiling
23556 @item -mtext-section-literals
23557 @itemx -mno-text-section-literals
23558 @opindex mtext-section-literals
23559 @opindex mno-text-section-literals
23560 These options control the treatment of literal pools. The default is
23561 @option{-mno-text-section-literals}, which places literals in a separate
23562 section in the output file. This allows the literal pool to be placed
23563 in a data RAM/ROM, and it also allows the linker to combine literal
23564 pools from separate object files to remove redundant literals and
23565 improve code size. With @option{-mtext-section-literals}, the literals
23566 are interspersed in the text section in order to keep them as close as
23567 possible to their references. This may be necessary for large assembly
23570 @item -mtarget-align
23571 @itemx -mno-target-align
23572 @opindex mtarget-align
23573 @opindex mno-target-align
23574 When this option is enabled, GCC instructs the assembler to
23575 automatically align instructions to reduce branch penalties at the
23576 expense of some code density. The assembler attempts to widen density
23577 instructions to align branch targets and the instructions following call
23578 instructions. If there are not enough preceding safe density
23579 instructions to align a target, no widening is performed. The
23580 default is @option{-mtarget-align}. These options do not affect the
23581 treatment of auto-aligned instructions like @code{LOOP}, which the
23582 assembler always aligns, either by widening density instructions or
23583 by inserting NOP instructions.
23586 @itemx -mno-longcalls
23587 @opindex mlongcalls
23588 @opindex mno-longcalls
23589 When this option is enabled, GCC instructs the assembler to translate
23590 direct calls to indirect calls unless it can determine that the target
23591 of a direct call is in the range allowed by the call instruction. This
23592 translation typically occurs for calls to functions in other source
23593 files. Specifically, the assembler translates a direct @code{CALL}
23594 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
23595 The default is @option{-mno-longcalls}. This option should be used in
23596 programs where the call target can potentially be out of range. This
23597 option is implemented in the assembler, not the compiler, so the
23598 assembly code generated by GCC still shows direct call
23599 instructions---look at the disassembled object code to see the actual
23600 instructions. Note that the assembler uses an indirect call for
23601 every cross-file call, not just those that really are out of range.
23604 @node zSeries Options
23605 @subsection zSeries Options
23606 @cindex zSeries options
23608 These are listed under @xref{S/390 and zSeries Options}.
23610 @node Code Gen Options
23611 @section Options for Code Generation Conventions
23612 @cindex code generation conventions
23613 @cindex options, code generation
23614 @cindex run-time options
23616 These machine-independent options control the interface conventions
23617 used in code generation.
23619 Most of them have both positive and negative forms; the negative form
23620 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
23621 one of the forms is listed---the one that is not the default. You
23622 can figure out the other form by either removing @samp{no-} or adding
23626 @item -fbounds-check
23627 @opindex fbounds-check
23628 For front ends that support it, generate additional code to check that
23629 indices used to access arrays are within the declared range. This is
23630 currently only supported by the Java and Fortran front ends, where
23631 this option defaults to true and false respectively.
23633 @item -fstack-reuse=@var{reuse-level}
23634 @opindex fstack_reuse
23635 This option controls stack space reuse for user declared local/auto variables
23636 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
23637 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
23638 local variables and temporaries, @samp{named_vars} enables the reuse only for
23639 user defined local variables with names, and @samp{none} disables stack reuse
23640 completely. The default value is @samp{all}. The option is needed when the
23641 program extends the lifetime of a scoped local variable or a compiler generated
23642 temporary beyond the end point defined by the language. When a lifetime of
23643 a variable ends, and if the variable lives in memory, the optimizing compiler
23644 has the freedom to reuse its stack space with other temporaries or scoped
23645 local variables whose live range does not overlap with it. Legacy code extending
23646 local lifetime is likely to break with the stack reuse optimization.
23665 if (*p == 10) // out of scope use of local1
23676 A(int k) : i(k), j(k) @{ @}
23683 void foo(const A& ar)
23690 foo(A(10)); // temp object's lifetime ends when foo returns
23696 ap->i+= 10; // ap references out of scope temp whose space
23697 // is reused with a. What is the value of ap->i?
23702 The lifetime of a compiler generated temporary is well defined by the C++
23703 standard. When a lifetime of a temporary ends, and if the temporary lives
23704 in memory, the optimizing compiler has the freedom to reuse its stack
23705 space with other temporaries or scoped local variables whose live range
23706 does not overlap with it. However some of the legacy code relies on
23707 the behavior of older compilers in which temporaries' stack space is
23708 not reused, the aggressive stack reuse can lead to runtime errors. This
23709 option is used to control the temporary stack reuse optimization.
23713 This option generates traps for signed overflow on addition, subtraction,
23714 multiplication operations.
23718 This option instructs the compiler to assume that signed arithmetic
23719 overflow of addition, subtraction and multiplication wraps around
23720 using twos-complement representation. This flag enables some optimizations
23721 and disables others. This option is enabled by default for the Java
23722 front end, as required by the Java language specification.
23725 @opindex fexceptions
23726 Enable exception handling. Generates extra code needed to propagate
23727 exceptions. For some targets, this implies GCC generates frame
23728 unwind information for all functions, which can produce significant data
23729 size overhead, although it does not affect execution. If you do not
23730 specify this option, GCC enables it by default for languages like
23731 C++ that normally require exception handling, and disables it for
23732 languages like C that do not normally require it. However, you may need
23733 to enable this option when compiling C code that needs to interoperate
23734 properly with exception handlers written in C++. You may also wish to
23735 disable this option if you are compiling older C++ programs that don't
23736 use exception handling.
23738 @item -fnon-call-exceptions
23739 @opindex fnon-call-exceptions
23740 Generate code that allows trapping instructions to throw exceptions.
23741 Note that this requires platform-specific runtime support that does
23742 not exist everywhere. Moreover, it only allows @emph{trapping}
23743 instructions to throw exceptions, i.e.@: memory references or floating-point
23744 instructions. It does not allow exceptions to be thrown from
23745 arbitrary signal handlers such as @code{SIGALRM}.
23747 @item -fdelete-dead-exceptions
23748 @opindex fdelete-dead-exceptions
23749 Consider that instructions that may throw exceptions but don't otherwise
23750 contribute to the execution of the program can be optimized away.
23751 This option is enabled by default for the Ada front end, as permitted by
23752 the Ada language specification.
23753 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
23755 @item -funwind-tables
23756 @opindex funwind-tables
23757 Similar to @option{-fexceptions}, except that it just generates any needed
23758 static data, but does not affect the generated code in any other way.
23759 You normally do not need to enable this option; instead, a language processor
23760 that needs this handling enables it on your behalf.
23762 @item -fasynchronous-unwind-tables
23763 @opindex fasynchronous-unwind-tables
23764 Generate unwind table in DWARF 2 format, if supported by target machine. The
23765 table is exact at each instruction boundary, so it can be used for stack
23766 unwinding from asynchronous events (such as debugger or garbage collector).
23768 @item -fno-gnu-unique
23769 @opindex fno-gnu-unique
23770 On systems with recent GNU assembler and C library, the C++ compiler
23771 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
23772 of template static data members and static local variables in inline
23773 functions are unique even in the presence of @code{RTLD_LOCAL}; this
23774 is necessary to avoid problems with a library used by two different
23775 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
23776 therefore disagreeing with the other one about the binding of the
23777 symbol. But this causes @code{dlclose} to be ignored for affected
23778 DSOs; if your program relies on reinitialization of a DSO via
23779 @code{dlclose} and @code{dlopen}, you can use
23780 @option{-fno-gnu-unique}.
23782 @item -fpcc-struct-return
23783 @opindex fpcc-struct-return
23784 Return ``short'' @code{struct} and @code{union} values in memory like
23785 longer ones, rather than in registers. This convention is less
23786 efficient, but it has the advantage of allowing intercallability between
23787 GCC-compiled files and files compiled with other compilers, particularly
23788 the Portable C Compiler (pcc).
23790 The precise convention for returning structures in memory depends
23791 on the target configuration macros.
23793 Short structures and unions are those whose size and alignment match
23794 that of some integer type.
23796 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
23797 switch is not binary compatible with code compiled with the
23798 @option{-freg-struct-return} switch.
23799 Use it to conform to a non-default application binary interface.
23801 @item -freg-struct-return
23802 @opindex freg-struct-return
23803 Return @code{struct} and @code{union} values in registers when possible.
23804 This is more efficient for small structures than
23805 @option{-fpcc-struct-return}.
23807 If you specify neither @option{-fpcc-struct-return} nor
23808 @option{-freg-struct-return}, GCC defaults to whichever convention is
23809 standard for the target. If there is no standard convention, GCC
23810 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
23811 the principal compiler. In those cases, we can choose the standard, and
23812 we chose the more efficient register return alternative.
23814 @strong{Warning:} code compiled with the @option{-freg-struct-return}
23815 switch is not binary compatible with code compiled with the
23816 @option{-fpcc-struct-return} switch.
23817 Use it to conform to a non-default application binary interface.
23819 @item -fshort-enums
23820 @opindex fshort-enums
23821 Allocate to an @code{enum} type only as many bytes as it needs for the
23822 declared range of possible values. Specifically, the @code{enum} type
23823 is equivalent to the smallest integer type that has enough room.
23825 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
23826 code that is not binary compatible with code generated without that switch.
23827 Use it to conform to a non-default application binary interface.
23829 @item -fshort-double
23830 @opindex fshort-double
23831 Use the same size for @code{double} as for @code{float}.
23833 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
23834 code that is not binary compatible with code generated without that switch.
23835 Use it to conform to a non-default application binary interface.
23837 @item -fshort-wchar
23838 @opindex fshort-wchar
23839 Override the underlying type for @code{wchar_t} to be @code{short
23840 unsigned int} instead of the default for the target. This option is
23841 useful for building programs to run under WINE@.
23843 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
23844 code that is not binary compatible with code generated without that switch.
23845 Use it to conform to a non-default application binary interface.
23848 @opindex fno-common
23849 In C code, controls the placement of uninitialized global variables.
23850 Unix C compilers have traditionally permitted multiple definitions of
23851 such variables in different compilation units by placing the variables
23853 This is the behavior specified by @option{-fcommon}, and is the default
23854 for GCC on most targets.
23855 On the other hand, this behavior is not required by ISO C, and on some
23856 targets may carry a speed or code size penalty on variable references.
23857 The @option{-fno-common} option specifies that the compiler should place
23858 uninitialized global variables in the data section of the object file,
23859 rather than generating them as common blocks.
23860 This has the effect that if the same variable is declared
23861 (without @code{extern}) in two different compilations,
23862 you get a multiple-definition error when you link them.
23863 In this case, you must compile with @option{-fcommon} instead.
23864 Compiling with @option{-fno-common} is useful on targets for which
23865 it provides better performance, or if you wish to verify that the
23866 program will work on other systems that always treat uninitialized
23867 variable declarations this way.
23871 Ignore the @code{#ident} directive.
23873 @item -finhibit-size-directive
23874 @opindex finhibit-size-directive
23875 Don't output a @code{.size} assembler directive, or anything else that
23876 would cause trouble if the function is split in the middle, and the
23877 two halves are placed at locations far apart in memory. This option is
23878 used when compiling @file{crtstuff.c}; you should not need to use it
23881 @item -fverbose-asm
23882 @opindex fverbose-asm
23883 Put extra commentary information in the generated assembly code to
23884 make it more readable. This option is generally only of use to those
23885 who actually need to read the generated assembly code (perhaps while
23886 debugging the compiler itself).
23888 @option{-fno-verbose-asm}, the default, causes the
23889 extra information to be omitted and is useful when comparing two assembler
23892 @item -frecord-gcc-switches
23893 @opindex frecord-gcc-switches
23894 This switch causes the command line used to invoke the
23895 compiler to be recorded into the object file that is being created.
23896 This switch is only implemented on some targets and the exact format
23897 of the recording is target and binary file format dependent, but it
23898 usually takes the form of a section containing ASCII text. This
23899 switch is related to the @option{-fverbose-asm} switch, but that
23900 switch only records information in the assembler output file as
23901 comments, so it never reaches the object file.
23902 See also @option{-grecord-gcc-switches} for another
23903 way of storing compiler options into the object file.
23907 @cindex global offset table
23909 Generate position-independent code (PIC) suitable for use in a shared
23910 library, if supported for the target machine. Such code accesses all
23911 constant addresses through a global offset table (GOT)@. The dynamic
23912 loader resolves the GOT entries when the program starts (the dynamic
23913 loader is not part of GCC; it is part of the operating system). If
23914 the GOT size for the linked executable exceeds a machine-specific
23915 maximum size, you get an error message from the linker indicating that
23916 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
23917 instead. (These maximums are 8k on the SPARC and 32k
23918 on the m68k and RS/6000. The x86 has no such limit.)
23920 Position-independent code requires special support, and therefore works
23921 only on certain machines. For the x86, GCC supports PIC for System V
23922 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
23923 position-independent.
23925 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23930 If supported for the target machine, emit position-independent code,
23931 suitable for dynamic linking and avoiding any limit on the size of the
23932 global offset table. This option makes a difference on the m68k,
23933 PowerPC and SPARC@.
23935 Position-independent code requires special support, and therefore works
23936 only on certain machines.
23938 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
23945 These options are similar to @option{-fpic} and @option{-fPIC}, but
23946 generated position independent code can be only linked into executables.
23947 Usually these options are used when @option{-pie} GCC option is
23948 used during linking.
23950 @option{-fpie} and @option{-fPIE} both define the macros
23951 @code{__pie__} and @code{__PIE__}. The macros have the value 1
23952 for @option{-fpie} and 2 for @option{-fPIE}.
23956 Do not use PLT for external function calls in position-independent code.
23957 Instead, load callee address at call site from GOT and branch to it.
23958 This leads to more efficient code by eliminating PLT stubs and exposing
23959 GOT load to optimizations. On architectures such as 32-bit x86 where
23960 PLT stubs expect GOT pointer in a specific register, this gives more
23961 register allocation freedom to the compiler. Lazy binding requires PLT:
23962 with @option{-fno-plt} all external symbols are resolved at load time.
23964 Alternatively, function attribute @code{noplt} can be used to avoid PLT
23965 for calls to specific external functions by marking those functions with
23968 Additionally, a few targets also convert calls to those functions that are
23969 marked to not use the PLT to use the GOT instead for non-position independent
23972 @item -fno-jump-tables
23973 @opindex fno-jump-tables
23974 Do not use jump tables for switch statements even where it would be
23975 more efficient than other code generation strategies. This option is
23976 of use in conjunction with @option{-fpic} or @option{-fPIC} for
23977 building code that forms part of a dynamic linker and cannot
23978 reference the address of a jump table. On some targets, jump tables
23979 do not require a GOT and this option is not needed.
23981 @item -ffixed-@var{reg}
23983 Treat the register named @var{reg} as a fixed register; generated code
23984 should never refer to it (except perhaps as a stack pointer, frame
23985 pointer or in some other fixed role).
23987 @var{reg} must be the name of a register. The register names accepted
23988 are machine-specific and are defined in the @code{REGISTER_NAMES}
23989 macro in the machine description macro file.
23991 This flag does not have a negative form, because it specifies a
23994 @item -fcall-used-@var{reg}
23995 @opindex fcall-used
23996 Treat the register named @var{reg} as an allocable register that is
23997 clobbered by function calls. It may be allocated for temporaries or
23998 variables that do not live across a call. Functions compiled this way
23999 do not save and restore the register @var{reg}.
24001 It is an error to use this flag with the frame pointer or stack pointer.
24002 Use of this flag for other registers that have fixed pervasive roles in
24003 the machine's execution model produces disastrous results.
24005 This flag does not have a negative form, because it specifies a
24008 @item -fcall-saved-@var{reg}
24009 @opindex fcall-saved
24010 Treat the register named @var{reg} as an allocable register saved by
24011 functions. It may be allocated even for temporaries or variables that
24012 live across a call. Functions compiled this way save and restore
24013 the register @var{reg} if they use it.
24015 It is an error to use this flag with the frame pointer or stack pointer.
24016 Use of this flag for other registers that have fixed pervasive roles in
24017 the machine's execution model produces disastrous results.
24019 A different sort of disaster results from the use of this flag for
24020 a register in which function values may be returned.
24022 This flag does not have a negative form, because it specifies a
24025 @item -fpack-struct[=@var{n}]
24026 @opindex fpack-struct
24027 Without a value specified, pack all structure members together without
24028 holes. When a value is specified (which must be a small power of two), pack
24029 structure members according to this value, representing the maximum
24030 alignment (that is, objects with default alignment requirements larger than
24031 this are output potentially unaligned at the next fitting location.
24033 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
24034 code that is not binary compatible with code generated without that switch.
24035 Additionally, it makes the code suboptimal.
24036 Use it to conform to a non-default application binary interface.
24038 @item -finstrument-functions
24039 @opindex finstrument-functions
24040 Generate instrumentation calls for entry and exit to functions. Just
24041 after function entry and just before function exit, the following
24042 profiling functions are called with the address of the current
24043 function and its call site. (On some platforms,
24044 @code{__builtin_return_address} does not work beyond the current
24045 function, so the call site information may not be available to the
24046 profiling functions otherwise.)
24049 void __cyg_profile_func_enter (void *this_fn,
24051 void __cyg_profile_func_exit (void *this_fn,
24055 The first argument is the address of the start of the current function,
24056 which may be looked up exactly in the symbol table.
24058 This instrumentation is also done for functions expanded inline in other
24059 functions. The profiling calls indicate where, conceptually, the
24060 inline function is entered and exited. This means that addressable
24061 versions of such functions must be available. If all your uses of a
24062 function are expanded inline, this may mean an additional expansion of
24063 code size. If you use @code{extern inline} in your C code, an
24064 addressable version of such functions must be provided. (This is
24065 normally the case anyway, but if you get lucky and the optimizer always
24066 expands the functions inline, you might have gotten away without
24067 providing static copies.)
24069 A function may be given the attribute @code{no_instrument_function}, in
24070 which case this instrumentation is not done. This can be used, for
24071 example, for the profiling functions listed above, high-priority
24072 interrupt routines, and any functions from which the profiling functions
24073 cannot safely be called (perhaps signal handlers, if the profiling
24074 routines generate output or allocate memory).
24076 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
24077 @opindex finstrument-functions-exclude-file-list
24079 Set the list of functions that are excluded from instrumentation (see
24080 the description of @option{-finstrument-functions}). If the file that
24081 contains a function definition matches with one of @var{file}, then
24082 that function is not instrumented. The match is done on substrings:
24083 if the @var{file} parameter is a substring of the file name, it is
24084 considered to be a match.
24089 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
24093 excludes any inline function defined in files whose pathnames
24094 contain @file{/bits/stl} or @file{include/sys}.
24096 If, for some reason, you want to include letter @samp{,} in one of
24097 @var{sym}, write @samp{\,}. For example,
24098 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
24099 (note the single quote surrounding the option).
24101 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
24102 @opindex finstrument-functions-exclude-function-list
24104 This is similar to @option{-finstrument-functions-exclude-file-list},
24105 but this option sets the list of function names to be excluded from
24106 instrumentation. The function name to be matched is its user-visible
24107 name, such as @code{vector<int> blah(const vector<int> &)}, not the
24108 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
24109 match is done on substrings: if the @var{sym} parameter is a substring
24110 of the function name, it is considered to be a match. For C99 and C++
24111 extended identifiers, the function name must be given in UTF-8, not
24112 using universal character names.
24114 @item -fstack-check
24115 @opindex fstack-check
24116 Generate code to verify that you do not go beyond the boundary of the
24117 stack. You should specify this flag if you are running in an
24118 environment with multiple threads, but you only rarely need to specify it in
24119 a single-threaded environment since stack overflow is automatically
24120 detected on nearly all systems if there is only one stack.
24122 Note that this switch does not actually cause checking to be done; the
24123 operating system or the language runtime must do that. The switch causes
24124 generation of code to ensure that they see the stack being extended.
24126 You can additionally specify a string parameter: @samp{no} means no
24127 checking, @samp{generic} means force the use of old-style checking,
24128 @samp{specific} means use the best checking method and is equivalent
24129 to bare @option{-fstack-check}.
24131 Old-style checking is a generic mechanism that requires no specific
24132 target support in the compiler but comes with the following drawbacks:
24136 Modified allocation strategy for large objects: they are always
24137 allocated dynamically if their size exceeds a fixed threshold.
24140 Fixed limit on the size of the static frame of functions: when it is
24141 topped by a particular function, stack checking is not reliable and
24142 a warning is issued by the compiler.
24145 Inefficiency: because of both the modified allocation strategy and the
24146 generic implementation, code performance is hampered.
24149 Note that old-style stack checking is also the fallback method for
24150 @samp{specific} if no target support has been added in the compiler.
24152 @item -fstack-limit-register=@var{reg}
24153 @itemx -fstack-limit-symbol=@var{sym}
24154 @itemx -fno-stack-limit
24155 @opindex fstack-limit-register
24156 @opindex fstack-limit-symbol
24157 @opindex fno-stack-limit
24158 Generate code to ensure that the stack does not grow beyond a certain value,
24159 either the value of a register or the address of a symbol. If a larger
24160 stack is required, a signal is raised at run time. For most targets,
24161 the signal is raised before the stack overruns the boundary, so
24162 it is possible to catch the signal without taking special precautions.
24164 For instance, if the stack starts at absolute address @samp{0x80000000}
24165 and grows downwards, you can use the flags
24166 @option{-fstack-limit-symbol=__stack_limit} and
24167 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
24168 of 128KB@. Note that this may only work with the GNU linker.
24170 @item -fsplit-stack
24171 @opindex fsplit-stack
24172 Generate code to automatically split the stack before it overflows.
24173 The resulting program has a discontiguous stack which can only
24174 overflow if the program is unable to allocate any more memory. This
24175 is most useful when running threaded programs, as it is no longer
24176 necessary to calculate a good stack size to use for each thread. This
24177 is currently only implemented for the x86 targets running
24180 When code compiled with @option{-fsplit-stack} calls code compiled
24181 without @option{-fsplit-stack}, there may not be much stack space
24182 available for the latter code to run. If compiling all code,
24183 including library code, with @option{-fsplit-stack} is not an option,
24184 then the linker can fix up these calls so that the code compiled
24185 without @option{-fsplit-stack} always has a large stack. Support for
24186 this is implemented in the gold linker in GNU binutils release 2.21
24189 @item -fleading-underscore
24190 @opindex fleading-underscore
24191 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
24192 change the way C symbols are represented in the object file. One use
24193 is to help link with legacy assembly code.
24195 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
24196 generate code that is not binary compatible with code generated without that
24197 switch. Use it to conform to a non-default application binary interface.
24198 Not all targets provide complete support for this switch.
24200 @item -ftls-model=@var{model}
24201 @opindex ftls-model
24202 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
24203 The @var{model} argument should be one of @samp{global-dynamic},
24204 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
24205 Note that the choice is subject to optimization: the compiler may use
24206 a more efficient model for symbols not visible outside of the translation
24207 unit, or if @option{-fpic} is not given on the command line.
24209 The default without @option{-fpic} is @samp{initial-exec}; with
24210 @option{-fpic} the default is @samp{global-dynamic}.
24212 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
24213 @opindex fvisibility
24214 Set the default ELF image symbol visibility to the specified option---all
24215 symbols are marked with this unless overridden within the code.
24216 Using this feature can very substantially improve linking and
24217 load times of shared object libraries, produce more optimized
24218 code, provide near-perfect API export and prevent symbol clashes.
24219 It is @strong{strongly} recommended that you use this in any shared objects
24222 Despite the nomenclature, @samp{default} always means public; i.e.,
24223 available to be linked against from outside the shared object.
24224 @samp{protected} and @samp{internal} are pretty useless in real-world
24225 usage so the only other commonly used option is @samp{hidden}.
24226 The default if @option{-fvisibility} isn't specified is
24227 @samp{default}, i.e., make every symbol public.
24229 A good explanation of the benefits offered by ensuring ELF
24230 symbols have the correct visibility is given by ``How To Write
24231 Shared Libraries'' by Ulrich Drepper (which can be found at
24232 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
24233 solution made possible by this option to marking things hidden when
24234 the default is public is to make the default hidden and mark things
24235 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
24236 and @code{__attribute__ ((visibility("default")))} instead of
24237 @code{__declspec(dllexport)} you get almost identical semantics with
24238 identical syntax. This is a great boon to those working with
24239 cross-platform projects.
24241 For those adding visibility support to existing code, you may find
24242 @code{#pragma GCC visibility} of use. This works by you enclosing
24243 the declarations you wish to set visibility for with (for example)
24244 @code{#pragma GCC visibility push(hidden)} and
24245 @code{#pragma GCC visibility pop}.
24246 Bear in mind that symbol visibility should be viewed @strong{as
24247 part of the API interface contract} and thus all new code should
24248 always specify visibility when it is not the default; i.e., declarations
24249 only for use within the local DSO should @strong{always} be marked explicitly
24250 as hidden as so to avoid PLT indirection overheads---making this
24251 abundantly clear also aids readability and self-documentation of the code.
24252 Note that due to ISO C++ specification requirements, @code{operator new} and
24253 @code{operator delete} must always be of default visibility.
24255 Be aware that headers from outside your project, in particular system
24256 headers and headers from any other library you use, may not be
24257 expecting to be compiled with visibility other than the default. You
24258 may need to explicitly say @code{#pragma GCC visibility push(default)}
24259 before including any such headers.
24261 @code{extern} declarations are not affected by @option{-fvisibility}, so
24262 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24263 no modifications. However, this means that calls to @code{extern}
24264 functions with no explicit visibility use the PLT, so it is more
24265 effective to use @code{__attribute ((visibility))} and/or
24266 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24267 declarations should be treated as hidden.
24269 Note that @option{-fvisibility} does affect C++ vague linkage
24270 entities. This means that, for instance, an exception class that is
24271 be thrown between DSOs must be explicitly marked with default
24272 visibility so that the @samp{type_info} nodes are unified between
24275 An overview of these techniques, their benefits and how to use them
24276 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24278 @item -fstrict-volatile-bitfields
24279 @opindex fstrict-volatile-bitfields
24280 This option should be used if accesses to volatile bit-fields (or other
24281 structure fields, although the compiler usually honors those types
24282 anyway) should use a single access of the width of the
24283 field's type, aligned to a natural alignment if possible. For
24284 example, targets with memory-mapped peripheral registers might require
24285 all such accesses to be 16 bits wide; with this flag you can
24286 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24287 is 16 bits on these targets) to force GCC to use 16-bit accesses
24288 instead of, perhaps, a more efficient 32-bit access.
24290 If this option is disabled, the compiler uses the most efficient
24291 instruction. In the previous example, that might be a 32-bit load
24292 instruction, even though that accesses bytes that do not contain
24293 any portion of the bit-field, or memory-mapped registers unrelated to
24294 the one being updated.
24296 In some cases, such as when the @code{packed} attribute is applied to a
24297 structure field, it may not be possible to access the field with a single
24298 read or write that is correctly aligned for the target machine. In this
24299 case GCC falls back to generating multiple accesses rather than code that
24300 will fault or truncate the result at run time.
24302 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24303 not allowed to touch non bit-field members. It is therefore recommended
24304 to define all bits of the field's type as bit-field members.
24306 The default value of this option is determined by the application binary
24307 interface for the target processor.
24309 @item -fsync-libcalls
24310 @opindex fsync-libcalls
24311 This option controls whether any out-of-line instance of the @code{__sync}
24312 family of functions may be used to implement the C++11 @code{__atomic}
24313 family of functions.
24315 The default value of this option is enabled, thus the only useful form
24316 of the option is @option{-fno-sync-libcalls}. This option is used in
24317 the implementation of the @file{libatomic} runtime library.
24323 @node Environment Variables
24324 @section Environment Variables Affecting GCC
24325 @cindex environment variables
24327 @c man begin ENVIRONMENT
24328 This section describes several environment variables that affect how GCC
24329 operates. Some of them work by specifying directories or prefixes to use
24330 when searching for various kinds of files. Some are used to specify other
24331 aspects of the compilation environment.
24333 Note that you can also specify places to search using options such as
24334 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24335 take precedence over places specified using environment variables, which
24336 in turn take precedence over those specified by the configuration of GCC@.
24337 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24338 GNU Compiler Collection (GCC) Internals}.
24343 @c @itemx LC_COLLATE
24345 @c @itemx LC_MONETARY
24346 @c @itemx LC_NUMERIC
24351 @c @findex LC_COLLATE
24352 @findex LC_MESSAGES
24353 @c @findex LC_MONETARY
24354 @c @findex LC_NUMERIC
24358 These environment variables control the way that GCC uses
24359 localization information which allows GCC to work with different
24360 national conventions. GCC inspects the locale categories
24361 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24362 so. These locale categories can be set to any value supported by your
24363 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24364 Kingdom encoded in UTF-8.
24366 The @env{LC_CTYPE} environment variable specifies character
24367 classification. GCC uses it to determine the character boundaries in
24368 a string; this is needed for some multibyte encodings that contain quote
24369 and escape characters that are otherwise interpreted as a string
24372 The @env{LC_MESSAGES} environment variable specifies the language to
24373 use in diagnostic messages.
24375 If the @env{LC_ALL} environment variable is set, it overrides the value
24376 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24377 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24378 environment variable. If none of these variables are set, GCC
24379 defaults to traditional C English behavior.
24383 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24384 files. GCC uses temporary files to hold the output of one stage of
24385 compilation which is to be used as input to the next stage: for example,
24386 the output of the preprocessor, which is the input to the compiler
24389 @item GCC_COMPARE_DEBUG
24390 @findex GCC_COMPARE_DEBUG
24391 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24392 @option{-fcompare-debug} to the compiler driver. See the documentation
24393 of this option for more details.
24395 @item GCC_EXEC_PREFIX
24396 @findex GCC_EXEC_PREFIX
24397 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24398 names of the subprograms executed by the compiler. No slash is added
24399 when this prefix is combined with the name of a subprogram, but you can
24400 specify a prefix that ends with a slash if you wish.
24402 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24403 an appropriate prefix to use based on the pathname it is invoked with.
24405 If GCC cannot find the subprogram using the specified prefix, it
24406 tries looking in the usual places for the subprogram.
24408 The default value of @env{GCC_EXEC_PREFIX} is
24409 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24410 the installed compiler. In many cases @var{prefix} is the value
24411 of @code{prefix} when you ran the @file{configure} script.
24413 Other prefixes specified with @option{-B} take precedence over this prefix.
24415 This prefix is also used for finding files such as @file{crt0.o} that are
24418 In addition, the prefix is used in an unusual way in finding the
24419 directories to search for header files. For each of the standard
24420 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24421 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24422 replacing that beginning with the specified prefix to produce an
24423 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24424 @file{foo/bar} just before it searches the standard directory
24425 @file{/usr/local/lib/bar}.
24426 If a standard directory begins with the configured
24427 @var{prefix} then the value of @var{prefix} is replaced by
24428 @env{GCC_EXEC_PREFIX} when looking for header files.
24430 @item COMPILER_PATH
24431 @findex COMPILER_PATH
24432 The value of @env{COMPILER_PATH} is a colon-separated list of
24433 directories, much like @env{PATH}. GCC tries the directories thus
24434 specified when searching for subprograms, if it can't find the
24435 subprograms using @env{GCC_EXEC_PREFIX}.
24438 @findex LIBRARY_PATH
24439 The value of @env{LIBRARY_PATH} is a colon-separated list of
24440 directories, much like @env{PATH}. When configured as a native compiler,
24441 GCC tries the directories thus specified when searching for special
24442 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24443 using GCC also uses these directories when searching for ordinary
24444 libraries for the @option{-l} option (but directories specified with
24445 @option{-L} come first).
24449 @cindex locale definition
24450 This variable is used to pass locale information to the compiler. One way in
24451 which this information is used is to determine the character set to be used
24452 when character literals, string literals and comments are parsed in C and C++.
24453 When the compiler is configured to allow multibyte characters,
24454 the following values for @env{LANG} are recognized:
24458 Recognize JIS characters.
24460 Recognize SJIS characters.
24462 Recognize EUCJP characters.
24465 If @env{LANG} is not defined, or if it has some other value, then the
24466 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24467 recognize and translate multibyte characters.
24471 Some additional environment variables affect the behavior of the
24474 @include cppenv.texi
24478 @node Precompiled Headers
24479 @section Using Precompiled Headers
24480 @cindex precompiled headers
24481 @cindex speed of compilation
24483 Often large projects have many header files that are included in every
24484 source file. The time the compiler takes to process these header files
24485 over and over again can account for nearly all of the time required to
24486 build the project. To make builds faster, GCC allows you to
24487 @dfn{precompile} a header file.
24489 To create a precompiled header file, simply compile it as you would any
24490 other file, if necessary using the @option{-x} option to make the driver
24491 treat it as a C or C++ header file. You may want to use a
24492 tool like @command{make} to keep the precompiled header up-to-date when
24493 the headers it contains change.
24495 A precompiled header file is searched for when @code{#include} is
24496 seen in the compilation. As it searches for the included file
24497 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24498 compiler looks for a precompiled header in each directory just before it
24499 looks for the include file in that directory. The name searched for is
24500 the name specified in the @code{#include} with @samp{.gch} appended. If
24501 the precompiled header file can't be used, it is ignored.
24503 For instance, if you have @code{#include "all.h"}, and you have
24504 @file{all.h.gch} in the same directory as @file{all.h}, then the
24505 precompiled header file is used if possible, and the original
24506 header is used otherwise.
24508 Alternatively, you might decide to put the precompiled header file in a
24509 directory and use @option{-I} to ensure that directory is searched
24510 before (or instead of) the directory containing the original header.
24511 Then, if you want to check that the precompiled header file is always
24512 used, you can put a file of the same name as the original header in this
24513 directory containing an @code{#error} command.
24515 This also works with @option{-include}. So yet another way to use
24516 precompiled headers, good for projects not designed with precompiled
24517 header files in mind, is to simply take most of the header files used by
24518 a project, include them from another header file, precompile that header
24519 file, and @option{-include} the precompiled header. If the header files
24520 have guards against multiple inclusion, they are skipped because
24521 they've already been included (in the precompiled header).
24523 If you need to precompile the same header file for different
24524 languages, targets, or compiler options, you can instead make a
24525 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24526 header in the directory, perhaps using @option{-o}. It doesn't matter
24527 what you call the files in the directory; every precompiled header in
24528 the directory is considered. The first precompiled header
24529 encountered in the directory that is valid for this compilation is
24530 used; they're searched in no particular order.
24532 There are many other possibilities, limited only by your imagination,
24533 good sense, and the constraints of your build system.
24535 A precompiled header file can be used only when these conditions apply:
24539 Only one precompiled header can be used in a particular compilation.
24542 A precompiled header can't be used once the first C token is seen. You
24543 can have preprocessor directives before a precompiled header; you cannot
24544 include a precompiled header from inside another header.
24547 The precompiled header file must be produced for the same language as
24548 the current compilation. You can't use a C precompiled header for a C++
24552 The precompiled header file must have been produced by the same compiler
24553 binary as the current compilation is using.
24556 Any macros defined before the precompiled header is included must
24557 either be defined in the same way as when the precompiled header was
24558 generated, or must not affect the precompiled header, which usually
24559 means that they don't appear in the precompiled header at all.
24561 The @option{-D} option is one way to define a macro before a
24562 precompiled header is included; using a @code{#define} can also do it.
24563 There are also some options that define macros implicitly, like
24564 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
24567 @item If debugging information is output when using the precompiled
24568 header, using @option{-g} or similar, the same kind of debugging information
24569 must have been output when building the precompiled header. However,
24570 a precompiled header built using @option{-g} can be used in a compilation
24571 when no debugging information is being output.
24573 @item The same @option{-m} options must generally be used when building
24574 and using the precompiled header. @xref{Submodel Options},
24575 for any cases where this rule is relaxed.
24577 @item Each of the following options must be the same when building and using
24578 the precompiled header:
24580 @gccoptlist{-fexceptions}
24583 Some other command-line options starting with @option{-f},
24584 @option{-p}, or @option{-O} must be defined in the same way as when
24585 the precompiled header was generated. At present, it's not clear
24586 which options are safe to change and which are not; the safest choice
24587 is to use exactly the same options when generating and using the
24588 precompiled header. The following are known to be safe:
24590 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
24591 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
24592 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
24597 For all of these except the last, the compiler automatically
24598 ignores the precompiled header if the conditions aren't met. If you
24599 find an option combination that doesn't work and doesn't cause the
24600 precompiled header to be ignored, please consider filing a bug report,
24603 If you do use differing options when generating and using the
24604 precompiled header, the actual behavior is a mixture of the
24605 behavior for the options. For instance, if you use @option{-g} to
24606 generate the precompiled header but not when using it, you may or may
24607 not get debugging information for routines in the precompiled header.