1 @c Copyright (C) 1988-2016 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{} 1988-2016 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.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
142 * Optimize Options:: How much optimization?
143 * Preprocessor Options:: Controlling header files and macro definitions.
144 Also, getting dependency information for Make.
145 * Assembler Options:: Passing options to the assembler.
146 * Link Options:: Specifying libraries and so on.
147 * Directory Options:: Where to find header files and libraries.
148 Where to find the compiler executable files.
149 * Spec Files:: How to pass switches to sub-processes.
150 * Submodel Options:: Specifying minor hardware or convention variations,
151 such as 68010 vs 68020.
152 * Code Gen Options:: Specifying conventions for function calls, data layout
154 * Environment Variables:: Env vars that affect GCC.
155 * Precompiled Headers:: Compiling a header once, and using it many times.
161 @section Option Summary
163 Here is a summary of all the options, grouped by type. Explanations are
164 in the following sections.
167 @item Overall Options
168 @xref{Overall Options,,Options Controlling the Kind of Output}.
169 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
170 -pipe -pass-exit-codes @gol
171 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
172 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
173 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
175 @item C Language Options
176 @xref{C Dialect Options,,Options Controlling C Dialect}.
177 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
178 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
179 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
180 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
181 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness}
182 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
183 -fsigned-bitfields -fsigned-char @gol
184 -funsigned-bitfields -funsigned-char @gol
185 -trigraphs -traditional -traditional-cpp}
187 @item C++ Language Options
188 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
189 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
190 -fconstexpr-depth=@var{n} -ffriend-injection @gol
191 -fno-elide-constructors @gol
192 -fno-enforce-eh-specs @gol
193 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
194 -fno-implicit-templates @gol
195 -fno-implicit-inline-templates @gol
196 -fno-implement-inlines -fms-extensions @gol
197 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
198 -fno-optional-diags -fpermissive @gol
199 -fno-pretty-templates @gol
200 -frepo -fno-rtti -fsized-deallocation @gol
201 -fstats -ftemplate-backtrace-limit=@var{n} @gol
202 -ftemplate-depth=@var{n} @gol
203 -fno-threadsafe-statics -fuse-cxa-atexit @gol
204 -fno-weak -nostdinc++ @gol
205 -fvisibility-inlines-hidden @gol
206 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
207 -fvtv-counts -fvtv-debug @gol
208 -fvisibility-ms-compat @gol
209 -fext-numeric-literals @gol
210 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
211 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
212 -Wnamespaces -Wnarrowing @gol
213 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
214 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
215 -Wno-non-template-friend -Wold-style-cast @gol
216 -Woverloaded-virtual -Wno-pmf-conversions @gol
217 -Wsign-promo -Wvirtual-inheritance}
219 @item Objective-C and Objective-C++ Language Options
220 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
221 Objective-C and Objective-C++ Dialects}.
222 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
223 -fgnu-runtime -fnext-runtime @gol
224 -fno-nil-receivers @gol
225 -fobjc-abi-version=@var{n} @gol
226 -fobjc-call-cxx-cdtors @gol
227 -fobjc-direct-dispatch @gol
228 -fobjc-exceptions @gol
231 -fobjc-std=objc1 @gol
232 -fno-local-ivars @gol
233 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
234 -freplace-objc-classes @gol
237 -Wassign-intercept @gol
238 -Wno-protocol -Wselector @gol
239 -Wstrict-selector-match @gol
240 -Wundeclared-selector}
242 @item Diagnostic Message Formatting Options
243 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
244 @gccoptlist{-fmessage-length=@var{n} @gol
245 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
246 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
247 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
249 @item Warning Options
250 @xref{Warning Options,,Options to Request or Suppress Warnings}.
251 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
252 -pedantic-errors @gol
253 -w -Wextra -Wall -Waddress -Waggregate-return @gol
254 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
255 -Wno-attributes -Wbool-compare -Wno-builtin-macro-redefined @gol
256 -Wc90-c99-compat -Wc99-c11-compat @gol
257 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
258 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
259 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdate-time -Wdelete-incomplete @gol
260 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
261 -Wdisabled-optimization @gol
262 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
263 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
264 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
265 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
266 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
267 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
268 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
269 -Wignored-qualifiers -Wincompatible-pointer-types @gol
270 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
271 -Winit-self -Winline -Wno-int-conversion @gol
272 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
273 -Winvalid-pch -Wlarger-than=@var{len} @gol
274 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
275 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args @gol
276 -Wmisleading-indentation -Wmissing-braces @gol
277 -Wmissing-field-initializers -Wmissing-include-dirs @gol
278 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
279 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
280 -Woverride-init-side-effects -Woverlength-strings @gol
281 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
282 -Wparentheses -Wno-pedantic-ms-format @gol
283 -Wplacement-new -Wpointer-arith -Wno-pointer-to-int-cast @gol
284 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
285 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
286 -Wshift-overflow -Wshift-overflow=@var{n} @gol
287 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
288 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
289 -Wno-scalar-storage-order @gol
290 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
291 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
292 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
293 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
294 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
295 -Wmissing-format-attribute -Wsubobject-linkage @gol
296 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
297 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
298 -Wtype-limits -Wundef @gol
299 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
300 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
301 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
302 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
303 -Wunused-const-variable @gol
304 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
305 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
306 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
307 -Wzero-as-null-pointer-constant}
309 @item C and Objective-C-only Warning Options
310 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
311 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
312 -Wold-style-declaration -Wold-style-definition @gol
313 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
314 -Wdeclaration-after-statement -Wpointer-sign}
316 @item Debugging Options
317 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
318 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
319 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
320 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
321 -fsanitize-undefined-trap-on-error @gol
322 -fcheck-pointer-bounds -fchecking -fchkp-check-incomplete-type @gol
323 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
324 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
325 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
326 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
327 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
328 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
329 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
330 -fchkp-use-wrappers @gol
331 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
332 -fdisable-ipa-@var{pass_name} @gol
333 -fdisable-rtl-@var{pass_name} @gol
334 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
335 -fdisable-tree-@var{pass_name} @gol
336 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
337 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
338 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
339 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
340 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
342 -fdump-statistics @gol
344 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-cfg -fdump-tree-alias @gol
348 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
349 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
350 -fdump-tree-gimple@r{[}-raw@r{]} @gol
351 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
352 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
353 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
354 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
355 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
356 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
357 -fdump-tree-nrv -fdump-tree-vect @gol
358 -fdump-tree-sink @gol
359 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
360 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
361 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
362 -fdump-tree-vtable-verify @gol
363 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
364 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
365 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
366 -fdump-final-insns=@var{file} @gol
367 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
368 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
369 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
370 -fenable-@var{kind}-@var{pass} @gol
371 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
372 -fdebug-types-section -fmem-report-wpa @gol
373 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
375 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
376 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
377 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
378 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
379 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
380 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
381 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
382 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
383 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
384 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
385 -fdebug-prefix-map=@var{old}=@var{new} @gol
386 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
387 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
388 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
389 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
390 -print-prog-name=@var{program} -print-search-dirs -Q @gol
391 -print-sysroot -print-sysroot-headers-suffix @gol
392 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
394 @item Optimization Options
395 @xref{Optimize Options,,Options that Control Optimization}.
396 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
397 -falign-jumps[=@var{n}] @gol
398 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
399 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
400 -fauto-inc-dec -fbranch-probabilities @gol
401 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
402 -fbtr-bb-exclusive -fcaller-saves @gol
403 -fcombine-stack-adjustments -fconserve-stack @gol
404 -fcompare-elim -fcprop-registers -fcrossjumping @gol
405 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
406 -fcx-limited-range @gol
407 -fdata-sections -fdce -fdelayed-branch @gol
408 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
409 -fdevirtualize-at-ltrans -fdse @gol
410 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
411 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
412 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
413 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
414 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
415 -fif-conversion2 -findirect-inlining @gol
416 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
417 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
418 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
419 -fira-algorithm=@var{algorithm} @gol
420 -fira-region=@var{region} -fira-hoist-pressure @gol
421 -fira-loop-pressure -fno-ira-share-save-slots @gol
422 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
423 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
424 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
425 -fkeep-static-consts -flive-range-shrinkage @gol
426 -floop-block -floop-interchange -floop-strip-mine @gol
427 -floop-unroll-and-jam -floop-nest-optimize @gol
428 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
429 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
430 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
431 -fmove-loop-invariants -fno-branch-count-reg @gol
432 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
433 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
434 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
435 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
436 -fomit-frame-pointer -foptimize-sibling-calls @gol
437 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
438 -fprefetch-loop-arrays -fprofile-report @gol
439 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
440 -fprofile-generate=@var{path} @gol
441 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
442 -fprofile-reorder-functions @gol
443 -freciprocal-math -free -frename-registers -freorder-blocks @gol
444 -freorder-blocks-algorithm=@var{algorithm} @gol
445 -freorder-blocks-and-partition -freorder-functions @gol
446 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
447 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
448 -fsched-spec-load -fsched-spec-load-dangerous @gol
449 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
450 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
451 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
452 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
453 -fschedule-fusion @gol
454 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
455 -fselective-scheduling -fselective-scheduling2 @gol
456 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
457 -fsemantic-interposition -fshrink-wrap -fsignaling-nans @gol
458 -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
460 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
461 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
462 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
463 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
464 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
465 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
466 -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
467 -ftree-loop-if-convert-stores -ftree-loop-im @gol
468 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
469 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
470 -ftree-loop-vectorize @gol
471 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
472 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
473 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
474 -ftree-vectorize -ftree-vrp @gol
475 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
476 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
477 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
478 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
479 --param @var{name}=@var{value}
480 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
482 @item Preprocessor Options
483 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
484 @gccoptlist{-A@var{question}=@var{answer} @gol
485 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
486 -C -dD -dI -dM -dN @gol
487 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
488 -idirafter @var{dir} @gol
489 -include @var{file} -imacros @var{file} @gol
490 -iprefix @var{file} -iwithprefix @var{dir} @gol
491 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
492 -imultilib @var{dir} -isysroot @var{dir} @gol
493 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
494 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
495 -remap -trigraphs -undef -U@var{macro} @gol
496 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
498 @item Assembler Option
499 @xref{Assembler Options,,Passing Options to the Assembler}.
500 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
503 @xref{Link Options,,Options for Linking}.
504 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
505 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
506 -s -static -static-libgcc -static-libstdc++ @gol
507 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
508 -static-libmpx -static-libmpxwrappers @gol
509 -shared -shared-libgcc -symbolic @gol
510 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
511 -u @var{symbol} -z @var{keyword}}
513 @item Directory Options
514 @xref{Directory Options,,Options for Directory Search}.
515 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
516 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
517 --sysroot=@var{dir} --no-sysroot-suffix}
519 @item Machine Dependent Options
520 @xref{Submodel Options,,Hardware Models and Configurations}.
521 @c This list is ordered alphanumerically by subsection name.
522 @c Try and put the significant identifier (CPU or system) first,
523 @c so users have a clue at guessing where the ones they want will be.
525 @emph{AArch64 Options}
526 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
527 -mgeneral-regs-only @gol
528 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
530 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
531 -mtls-dialect=desc -mtls-dialect=traditional @gol
532 -mtls-size=@var{size} @gol
533 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
534 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
535 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
536 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
538 @emph{Adapteva Epiphany Options}
539 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
540 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
541 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
542 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
543 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
544 -msplit-vecmove-early -m1reg-@var{reg}}
547 @gccoptlist{-mbarrel-shifter @gol
548 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
549 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
550 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
551 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
552 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
553 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
554 -mlong-calls -mmedium-calls -msdata @gol
555 -mucb-mcount -mvolatile-cache @gol
556 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
557 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
558 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
559 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
560 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
561 -mtune=@var{cpu} -mmultcost=@var{num} @gol
562 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
563 -mdiv-rem -mcode-density}
566 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
567 -mabi=@var{name} @gol
568 -mapcs-stack-check -mno-apcs-stack-check @gol
569 -mapcs-float -mno-apcs-float @gol
570 -mapcs-reentrant -mno-apcs-reentrant @gol
571 -msched-prolog -mno-sched-prolog @gol
572 -mlittle-endian -mbig-endian @gol
573 -mfloat-abi=@var{name} @gol
574 -mfp16-format=@var{name}
575 -mthumb-interwork -mno-thumb-interwork @gol
576 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
577 -mtune=@var{name} -mprint-tune-info @gol
578 -mstructure-size-boundary=@var{n} @gol
579 -mabort-on-noreturn @gol
580 -mlong-calls -mno-long-calls @gol
581 -msingle-pic-base -mno-single-pic-base @gol
582 -mpic-register=@var{reg} @gol
583 -mnop-fun-dllimport @gol
584 -mpoke-function-name @gol
586 -mtpcs-frame -mtpcs-leaf-frame @gol
587 -mcaller-super-interworking -mcallee-super-interworking @gol
588 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
589 -mword-relocations @gol
590 -mfix-cortex-m3-ldrd @gol
591 -munaligned-access @gol
592 -mneon-for-64bits @gol
593 -mslow-flash-data @gol
594 -masm-syntax-unified @gol
598 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
599 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
600 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
602 @emph{Blackfin Options}
603 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
604 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
605 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
606 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
607 -mno-id-shared-library -mshared-library-id=@var{n} @gol
608 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
609 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
610 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
614 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
615 -msim -msdata=@var{sdata-type}}
618 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
619 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
620 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
621 -mstack-align -mdata-align -mconst-align @gol
622 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
623 -melf -maout -melinux -mlinux -sim -sim2 @gol
624 -mmul-bug-workaround -mno-mul-bug-workaround}
627 @gccoptlist{-mmac @gol
628 -mcr16cplus -mcr16c @gol
629 -msim -mint32 -mbit-ops
630 -mdata-model=@var{model}}
632 @emph{Darwin Options}
633 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
634 -arch_only -bind_at_load -bundle -bundle_loader @gol
635 -client_name -compatibility_version -current_version @gol
637 -dependency-file -dylib_file -dylinker_install_name @gol
638 -dynamic -dynamiclib -exported_symbols_list @gol
639 -filelist -flat_namespace -force_cpusubtype_ALL @gol
640 -force_flat_namespace -headerpad_max_install_names @gol
642 -image_base -init -install_name -keep_private_externs @gol
643 -multi_module -multiply_defined -multiply_defined_unused @gol
644 -noall_load -no_dead_strip_inits_and_terms @gol
645 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
646 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
647 -private_bundle -read_only_relocs -sectalign @gol
648 -sectobjectsymbols -whyload -seg1addr @gol
649 -sectcreate -sectobjectsymbols -sectorder @gol
650 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
651 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
652 -segprot -segs_read_only_addr -segs_read_write_addr @gol
653 -single_module -static -sub_library -sub_umbrella @gol
654 -twolevel_namespace -umbrella -undefined @gol
655 -unexported_symbols_list -weak_reference_mismatches @gol
656 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
657 -mkernel -mone-byte-bool}
659 @emph{DEC Alpha Options}
660 @gccoptlist{-mno-fp-regs -msoft-float @gol
661 -mieee -mieee-with-inexact -mieee-conformant @gol
662 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
663 -mtrap-precision=@var{mode} -mbuild-constants @gol
664 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
665 -mbwx -mmax -mfix -mcix @gol
666 -mfloat-vax -mfloat-ieee @gol
667 -mexplicit-relocs -msmall-data -mlarge-data @gol
668 -msmall-text -mlarge-text @gol
669 -mmemory-latency=@var{time}}
672 @gccoptlist{-msmall-model -mno-lsim}
675 @gccoptlist{-msim -mlra}
678 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
679 -mhard-float -msoft-float @gol
680 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
681 -mdouble -mno-double @gol
682 -mmedia -mno-media -mmuladd -mno-muladd @gol
683 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
684 -mlinked-fp -mlong-calls -malign-labels @gol
685 -mlibrary-pic -macc-4 -macc-8 @gol
686 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
687 -moptimize-membar -mno-optimize-membar @gol
688 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
689 -mvliw-branch -mno-vliw-branch @gol
690 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
691 -mno-nested-cond-exec -mtomcat-stats @gol
695 @emph{GNU/Linux Options}
696 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
697 -tno-android-cc -tno-android-ld}
699 @emph{H8/300 Options}
700 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
703 @gccoptlist{-march=@var{architecture-type} @gol
704 -mdisable-fpregs -mdisable-indexing @gol
705 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
706 -mfixed-range=@var{register-range} @gol
707 -mjump-in-delay -mlinker-opt -mlong-calls @gol
708 -mlong-load-store -mno-disable-fpregs @gol
709 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
710 -mno-jump-in-delay -mno-long-load-store @gol
711 -mno-portable-runtime -mno-soft-float @gol
712 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
713 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
714 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
715 -munix=@var{unix-std} -nolibdld -static -threads}
718 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
719 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
720 -mconstant-gp -mauto-pic -mfused-madd @gol
721 -minline-float-divide-min-latency @gol
722 -minline-float-divide-max-throughput @gol
723 -mno-inline-float-divide @gol
724 -minline-int-divide-min-latency @gol
725 -minline-int-divide-max-throughput @gol
726 -mno-inline-int-divide @gol
727 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
728 -mno-inline-sqrt @gol
729 -mdwarf2-asm -mearly-stop-bits @gol
730 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
731 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
732 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
733 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
734 -msched-spec-ldc -msched-spec-control-ldc @gol
735 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
736 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
737 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
738 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
741 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
742 -msign-extend-enabled -muser-enabled}
744 @emph{M32R/D Options}
745 @gccoptlist{-m32r2 -m32rx -m32r @gol
747 -malign-loops -mno-align-loops @gol
748 -missue-rate=@var{number} @gol
749 -mbranch-cost=@var{number} @gol
750 -mmodel=@var{code-size-model-type} @gol
751 -msdata=@var{sdata-type} @gol
752 -mno-flush-func -mflush-func=@var{name} @gol
753 -mno-flush-trap -mflush-trap=@var{number} @gol
757 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
759 @emph{M680x0 Options}
760 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
761 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
762 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
763 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
764 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
765 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
766 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
767 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
771 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
772 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
773 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
774 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
775 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
778 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
779 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
780 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
781 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
784 @emph{MicroBlaze Options}
785 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
786 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
787 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
788 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
789 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
792 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
793 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
794 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
795 -mips16 -mno-mips16 -mflip-mips16 @gol
796 -minterlink-compressed -mno-interlink-compressed @gol
797 -minterlink-mips16 -mno-interlink-mips16 @gol
798 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
799 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
800 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
801 -mno-float -msingle-float -mdouble-float @gol
802 -modd-spreg -mno-odd-spreg @gol
803 -mabs=@var{mode} -mnan=@var{encoding} @gol
804 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
807 -mvirt -mno-virt @gol
809 -mmicromips -mno-micromips @gol
810 -mfpu=@var{fpu-type} @gol
811 -msmartmips -mno-smartmips @gol
812 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
813 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
814 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
815 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
816 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
817 -membedded-data -mno-embedded-data @gol
818 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
819 -mcode-readable=@var{setting} @gol
820 -msplit-addresses -mno-split-addresses @gol
821 -mexplicit-relocs -mno-explicit-relocs @gol
822 -mcheck-zero-division -mno-check-zero-division @gol
823 -mdivide-traps -mdivide-breaks @gol
824 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
825 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
826 -mfix-24k -mno-fix-24k @gol
827 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
828 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
829 -mfix-vr4120 -mno-fix-vr4120 @gol
830 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
831 -mflush-func=@var{func} -mno-flush-func @gol
832 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
833 -mcompact-branches=@var{policy} @gol
834 -mfp-exceptions -mno-fp-exceptions @gol
835 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
836 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
837 -mframe-header-opt -mno-frame-header-opt}
840 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
841 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
842 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
843 -mno-base-addresses -msingle-exit -mno-single-exit}
845 @emph{MN10300 Options}
846 @gccoptlist{-mmult-bug -mno-mult-bug @gol
847 -mno-am33 -mam33 -mam33-2 -mam34 @gol
848 -mtune=@var{cpu-type} @gol
849 -mreturn-pointer-on-d0 @gol
850 -mno-crt0 -mrelax -mliw -msetlb}
853 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
855 @emph{MSP430 Options}
856 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
858 -mcode-region= -mdata-region= @gol
859 -msilicon-errata= -msilicon-errata-warn= @gol
863 @gccoptlist{-mbig-endian -mlittle-endian @gol
864 -mreduced-regs -mfull-regs @gol
865 -mcmov -mno-cmov @gol
866 -mperf-ext -mno-perf-ext @gol
867 -mv3push -mno-v3push @gol
868 -m16bit -mno-16bit @gol
869 -misr-vector-size=@var{num} @gol
870 -mcache-block-size=@var{num} @gol
871 -march=@var{arch} @gol
872 -mcmodel=@var{code-model} @gol
875 @emph{Nios II Options}
876 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
878 -mno-bypass-cache -mbypass-cache @gol
879 -mno-cache-volatile -mcache-volatile @gol
880 -mno-fast-sw-div -mfast-sw-div @gol
881 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
882 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
883 -mcustom-fpu-cfg=@var{name} @gol
884 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
885 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
887 @emph{Nvidia PTX Options}
888 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
890 @emph{PDP-11 Options}
891 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
892 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
893 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
894 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
895 -mbranch-expensive -mbranch-cheap @gol
896 -munix-asm -mdec-asm}
898 @emph{picoChip Options}
899 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
900 -msymbol-as-address -mno-inefficient-warnings}
902 @emph{PowerPC Options}
903 See RS/6000 and PowerPC Options.
906 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
907 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
908 -m64bit-doubles -m32bit-doubles}
910 @emph{RS/6000 and PowerPC Options}
911 @gccoptlist{-mcpu=@var{cpu-type} @gol
912 -mtune=@var{cpu-type} @gol
913 -mcmodel=@var{code-model} @gol
915 -maltivec -mno-altivec @gol
916 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
917 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
918 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
919 -mfprnd -mno-fprnd @gol
920 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
921 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
922 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
923 -malign-power -malign-natural @gol
924 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
925 -msingle-float -mdouble-float -msimple-fpu @gol
926 -mstring -mno-string -mupdate -mno-update @gol
927 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
928 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
929 -mstrict-align -mno-strict-align -mrelocatable @gol
930 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
931 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
932 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
933 -mprioritize-restricted-insns=@var{priority} @gol
934 -msched-costly-dep=@var{dependence_type} @gol
935 -minsert-sched-nops=@var{scheme} @gol
936 -mcall-sysv -mcall-netbsd @gol
937 -maix-struct-return -msvr4-struct-return @gol
938 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
939 -mblock-move-inline-limit=@var{num} @gol
940 -misel -mno-isel @gol
941 -misel=yes -misel=no @gol
943 -mspe=yes -mspe=no @gol
945 -mgen-cell-microcode -mwarn-cell-microcode @gol
946 -mvrsave -mno-vrsave @gol
947 -mmulhw -mno-mulhw @gol
948 -mdlmzb -mno-dlmzb @gol
949 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
950 -mprototype -mno-prototype @gol
951 -msim -mmvme -mads -myellowknife -memb -msdata @gol
952 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
953 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
954 -mno-recip-precision @gol
955 -mveclibabi=@var{type} -mfriz -mno-friz @gol
956 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
957 -msave-toc-indirect -mno-save-toc-indirect @gol
958 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
959 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
960 -mquad-memory -mno-quad-memory @gol
961 -mquad-memory-atomic -mno-quad-memory-atomic @gol
962 -mcompat-align-parm -mno-compat-align-parm @gol
963 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
964 -mupper-regs -mno-upper-regs -mmodulo -mno-modulo @gol
965 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
966 -mpower9-fusion -mno-mpower9-fusion -mpower9-vector -mno-power9-vector}
969 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
971 -mbig-endian-data -mlittle-endian-data @gol
974 -mas100-syntax -mno-as100-syntax@gol
976 -mmax-constant-size=@gol
979 -mallow-string-insns -mno-allow-string-insns@gol
981 -mno-warn-multiple-fast-interrupts@gol
982 -msave-acc-in-interrupts}
984 @emph{S/390 and zSeries Options}
985 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
986 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
987 -mlong-double-64 -mlong-double-128 @gol
988 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
989 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
990 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
991 -mhtm -mvx -mzvector @gol
992 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
993 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
994 -mhotpatch=@var{halfwords},@var{halfwords}}
997 @gccoptlist{-meb -mel @gol
1001 -mscore5 -mscore5u -mscore7 -mscore7d}
1004 @gccoptlist{-m1 -m2 -m2e @gol
1005 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1007 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1008 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1009 -mb -ml -mdalign -mrelax @gol
1010 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1011 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1012 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1013 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1014 -maccumulate-outgoing-args @gol
1015 -matomic-model=@var{atomic-model} @gol
1016 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1017 -mcbranch-force-delay-slot @gol
1018 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1019 -mpretend-cmove -mtas}
1021 @emph{Solaris 2 Options}
1022 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1025 @emph{SPARC Options}
1026 @gccoptlist{-mcpu=@var{cpu-type} @gol
1027 -mtune=@var{cpu-type} @gol
1028 -mcmodel=@var{code-model} @gol
1029 -mmemory-model=@var{mem-model} @gol
1030 -m32 -m64 -mapp-regs -mno-app-regs @gol
1031 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1032 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1033 -mhard-quad-float -msoft-quad-float @gol
1034 -mstack-bias -mno-stack-bias @gol
1035 -mstd-struct-return -mno-std-struct-return @gol
1036 -munaligned-doubles -mno-unaligned-doubles @gol
1037 -muser-mode -mno-user-mode @gol
1038 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1039 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1040 -mcbcond -mno-cbcond @gol
1041 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1042 -mfix-at697f -mfix-ut699}
1045 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1046 -msafe-dma -munsafe-dma @gol
1048 -msmall-mem -mlarge-mem -mstdmain @gol
1049 -mfixed-range=@var{register-range} @gol
1051 -maddress-space-conversion -mno-address-space-conversion @gol
1052 -mcache-size=@var{cache-size} @gol
1053 -matomic-updates -mno-atomic-updates}
1055 @emph{System V Options}
1056 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1058 @emph{TILE-Gx Options}
1059 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1060 -mcmodel=@var{code-model}}
1062 @emph{TILEPro Options}
1063 @gccoptlist{-mcpu=@var{cpu} -m32}
1066 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1067 -mprolog-function -mno-prolog-function -mspace @gol
1068 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1069 -mapp-regs -mno-app-regs @gol
1070 -mdisable-callt -mno-disable-callt @gol
1071 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1072 -mv850e -mv850 -mv850e3v5 @gol
1083 @gccoptlist{-mg -mgnu -munix}
1085 @emph{Visium Options}
1086 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1087 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1090 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1091 -mpointer-size=@var{size}}
1093 @emph{VxWorks Options}
1094 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1095 -Xbind-lazy -Xbind-now}
1098 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1099 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1100 -mfpmath=@var{unit} @gol
1101 -masm=@var{dialect} -mno-fancy-math-387 @gol
1102 -mno-fp-ret-in-387 -msoft-float @gol
1103 -mno-wide-multiply -mrtd -malign-double @gol
1104 -mpreferred-stack-boundary=@var{num} @gol
1105 -mincoming-stack-boundary=@var{num} @gol
1106 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1107 -mrecip -mrecip=@var{opt} @gol
1108 -mvzeroupper -mprefer-avx128 @gol
1109 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1110 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1111 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1112 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1113 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1114 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1115 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1116 -mpku -mthreads @gol
1117 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1118 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1119 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1120 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1121 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1122 -mregparm=@var{num} -msseregparm @gol
1123 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1124 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1125 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1126 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1127 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1128 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1129 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1130 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1133 @emph{x86 Windows Options}
1134 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1135 -mnop-fun-dllimport -mthread @gol
1136 -municode -mwin32 -mwindows -fno-set-stack-executable}
1138 @emph{Xstormy16 Options}
1141 @emph{Xtensa Options}
1142 @gccoptlist{-mconst16 -mno-const16 @gol
1143 -mfused-madd -mno-fused-madd @gol
1145 -mserialize-volatile -mno-serialize-volatile @gol
1146 -mtext-section-literals -mno-text-section-literals @gol
1147 -mauto-litpools -mno-auto-litpools @gol
1148 -mtarget-align -mno-target-align @gol
1149 -mlongcalls -mno-longcalls}
1151 @emph{zSeries Options}
1152 See S/390 and zSeries Options.
1154 @item Code Generation Options
1155 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1156 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1157 -ffixed-@var{reg} -fexceptions @gol
1158 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1159 -fasynchronous-unwind-tables @gol
1160 -fno-gnu-unique @gol
1161 -finhibit-size-directive -finstrument-functions @gol
1162 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1163 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1164 -fno-common -fno-ident @gol
1165 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
1166 -fno-jump-tables @gol
1167 -frecord-gcc-switches @gol
1168 -freg-struct-return -fshort-enums @gol
1169 -fshort-double -fshort-wchar @gol
1170 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1171 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1172 -fno-stack-limit -fsplit-stack @gol
1173 -fleading-underscore -ftls-model=@var{model} @gol
1174 -fstack-reuse=@var{reuse_level} @gol
1175 -ftrapv -fwrapv -fbounds-check @gol
1176 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1177 -fstrict-volatile-bitfields -fsync-libcalls}
1181 @node Overall Options
1182 @section Options Controlling the Kind of Output
1184 Compilation can involve up to four stages: preprocessing, compilation
1185 proper, assembly and linking, always in that order. GCC is capable of
1186 preprocessing and compiling several files either into several
1187 assembler input files, or into one assembler input file; then each
1188 assembler input file produces an object file, and linking combines all
1189 the object files (those newly compiled, and those specified as input)
1190 into an executable file.
1192 @cindex file name suffix
1193 For any given input file, the file name suffix determines what kind of
1194 compilation is done:
1198 C source code that must be preprocessed.
1201 C source code that should not be preprocessed.
1204 C++ source code that should not be preprocessed.
1207 Objective-C source code. Note that you must link with the @file{libobjc}
1208 library to make an Objective-C program work.
1211 Objective-C source code that should not be preprocessed.
1215 Objective-C++ source code. Note that you must link with the @file{libobjc}
1216 library to make an Objective-C++ program work. Note that @samp{.M} refers
1217 to a literal capital M@.
1219 @item @var{file}.mii
1220 Objective-C++ source code that should not be preprocessed.
1223 C, C++, Objective-C or Objective-C++ header file to be turned into a
1224 precompiled header (default), or C, C++ header file to be turned into an
1225 Ada spec (via the @option{-fdump-ada-spec} switch).
1228 @itemx @var{file}.cp
1229 @itemx @var{file}.cxx
1230 @itemx @var{file}.cpp
1231 @itemx @var{file}.CPP
1232 @itemx @var{file}.c++
1234 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1235 the last two letters must both be literally @samp{x}. Likewise,
1236 @samp{.C} refers to a literal capital C@.
1240 Objective-C++ source code that must be preprocessed.
1242 @item @var{file}.mii
1243 Objective-C++ source code that should not be preprocessed.
1247 @itemx @var{file}.hp
1248 @itemx @var{file}.hxx
1249 @itemx @var{file}.hpp
1250 @itemx @var{file}.HPP
1251 @itemx @var{file}.h++
1252 @itemx @var{file}.tcc
1253 C++ header file to be turned into a precompiled header or Ada spec.
1256 @itemx @var{file}.for
1257 @itemx @var{file}.ftn
1258 Fixed form Fortran source code that should not be preprocessed.
1261 @itemx @var{file}.FOR
1262 @itemx @var{file}.fpp
1263 @itemx @var{file}.FPP
1264 @itemx @var{file}.FTN
1265 Fixed form Fortran source code that must be preprocessed (with the traditional
1268 @item @var{file}.f90
1269 @itemx @var{file}.f95
1270 @itemx @var{file}.f03
1271 @itemx @var{file}.f08
1272 Free form Fortran source code that should not be preprocessed.
1274 @item @var{file}.F90
1275 @itemx @var{file}.F95
1276 @itemx @var{file}.F03
1277 @itemx @var{file}.F08
1278 Free form Fortran source code that must be preprocessed (with the
1279 traditional preprocessor).
1284 @c FIXME: Descriptions of Java file types.
1290 @item @var{file}.ads
1291 Ada source code file that contains a library unit declaration (a
1292 declaration of a package, subprogram, or generic, or a generic
1293 instantiation), or a library unit renaming declaration (a package,
1294 generic, or subprogram renaming declaration). Such files are also
1297 @item @var{file}.adb
1298 Ada source code file containing a library unit body (a subprogram or
1299 package body). Such files are also called @dfn{bodies}.
1301 @c GCC also knows about some suffixes for languages not yet included:
1312 @itemx @var{file}.sx
1313 Assembler code that must be preprocessed.
1316 An object file to be fed straight into linking.
1317 Any file name with no recognized suffix is treated this way.
1321 You can specify the input language explicitly with the @option{-x} option:
1324 @item -x @var{language}
1325 Specify explicitly the @var{language} for the following input files
1326 (rather than letting the compiler choose a default based on the file
1327 name suffix). This option applies to all following input files until
1328 the next @option{-x} option. Possible values for @var{language} are:
1330 c c-header cpp-output
1331 c++ c++-header c++-cpp-output
1332 objective-c objective-c-header objective-c-cpp-output
1333 objective-c++ objective-c++-header objective-c++-cpp-output
1334 assembler assembler-with-cpp
1336 f77 f77-cpp-input f95 f95-cpp-input
1342 Turn off any specification of a language, so that subsequent files are
1343 handled according to their file name suffixes (as they are if @option{-x}
1344 has not been used at all).
1346 @item -pass-exit-codes
1347 @opindex pass-exit-codes
1348 Normally the @command{gcc} program exits with the code of 1 if any
1349 phase of the compiler returns a non-success return code. If you specify
1350 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1351 the numerically highest error produced by any phase returning an error
1352 indication. The C, C++, and Fortran front ends return 4 if an internal
1353 compiler error is encountered.
1356 If you only want some of the stages of compilation, you can use
1357 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1358 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1359 @command{gcc} is to stop. Note that some combinations (for example,
1360 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1365 Compile or assemble the source files, but do not link. The linking
1366 stage simply is not done. The ultimate output is in the form of an
1367 object file for each source file.
1369 By default, the object file name for a source file is made by replacing
1370 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1372 Unrecognized input files, not requiring compilation or assembly, are
1377 Stop after the stage of compilation proper; do not assemble. The output
1378 is in the form of an assembler code file for each non-assembler input
1381 By default, the assembler file name for a source file is made by
1382 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1384 Input files that don't require compilation are ignored.
1388 Stop after the preprocessing stage; do not run the compiler proper. The
1389 output is in the form of preprocessed source code, which is sent to the
1392 Input files that don't require preprocessing are ignored.
1394 @cindex output file option
1397 Place output in file @var{file}. This applies to whatever
1398 sort of output is being produced, whether it be an executable file,
1399 an object file, an assembler file or preprocessed C code.
1401 If @option{-o} is not specified, the default is to put an executable
1402 file in @file{a.out}, the object file for
1403 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1404 assembler file in @file{@var{source}.s}, a precompiled header file in
1405 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1410 Print (on standard error output) the commands executed to run the stages
1411 of compilation. Also print the version number of the compiler driver
1412 program and of the preprocessor and the compiler proper.
1416 Like @option{-v} except the commands are not executed and arguments
1417 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1418 This is useful for shell scripts to capture the driver-generated command lines.
1422 Use pipes rather than temporary files for communication between the
1423 various stages of compilation. This fails to work on some systems where
1424 the assembler is unable to read from a pipe; but the GNU assembler has
1429 Print (on the standard output) a description of the command-line options
1430 understood by @command{gcc}. If the @option{-v} option is also specified
1431 then @option{--help} is also passed on to the various processes
1432 invoked by @command{gcc}, so that they can display the command-line options
1433 they accept. If the @option{-Wextra} option has also been specified
1434 (prior to the @option{--help} option), then command-line options that
1435 have no documentation associated with them are also displayed.
1438 @opindex target-help
1439 Print (on the standard output) a description of target-specific command-line
1440 options for each tool. For some targets extra target-specific
1441 information may also be printed.
1443 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1444 Print (on the standard output) a description of the command-line
1445 options understood by the compiler that fit into all specified classes
1446 and qualifiers. These are the supported classes:
1449 @item @samp{optimizers}
1450 Display all of the optimization options supported by the
1453 @item @samp{warnings}
1454 Display all of the options controlling warning messages
1455 produced by the compiler.
1458 Display target-specific options. Unlike the
1459 @option{--target-help} option however, target-specific options of the
1460 linker and assembler are not displayed. This is because those
1461 tools do not currently support the extended @option{--help=} syntax.
1464 Display the values recognized by the @option{--param}
1467 @item @var{language}
1468 Display the options supported for @var{language}, where
1469 @var{language} is the name of one of the languages supported in this
1473 Display the options that are common to all languages.
1476 These are the supported qualifiers:
1479 @item @samp{undocumented}
1480 Display only those options that are undocumented.
1483 Display options taking an argument that appears after an equal
1484 sign in the same continuous piece of text, such as:
1485 @samp{--help=target}.
1487 @item @samp{separate}
1488 Display options taking an argument that appears as a separate word
1489 following the original option, such as: @samp{-o output-file}.
1492 Thus for example to display all the undocumented target-specific
1493 switches supported by the compiler, use:
1496 --help=target,undocumented
1499 The sense of a qualifier can be inverted by prefixing it with the
1500 @samp{^} character, so for example to display all binary warning
1501 options (i.e., ones that are either on or off and that do not take an
1502 argument) that have a description, use:
1505 --help=warnings,^joined,^undocumented
1508 The argument to @option{--help=} should not consist solely of inverted
1511 Combining several classes is possible, although this usually
1512 restricts the output so much that there is nothing to display. One
1513 case where it does work, however, is when one of the classes is
1514 @var{target}. For example, to display all the target-specific
1515 optimization options, use:
1518 --help=target,optimizers
1521 The @option{--help=} option can be repeated on the command line. Each
1522 successive use displays its requested class of options, skipping
1523 those that have already been displayed.
1525 If the @option{-Q} option appears on the command line before the
1526 @option{--help=} option, then the descriptive text displayed by
1527 @option{--help=} is changed. Instead of describing the displayed
1528 options, an indication is given as to whether the option is enabled,
1529 disabled or set to a specific value (assuming that the compiler
1530 knows this at the point where the @option{--help=} option is used).
1532 Here is a truncated example from the ARM port of @command{gcc}:
1535 % gcc -Q -mabi=2 --help=target -c
1536 The following options are target specific:
1538 -mabort-on-noreturn [disabled]
1542 The output is sensitive to the effects of previous command-line
1543 options, so for example it is possible to find out which optimizations
1544 are enabled at @option{-O2} by using:
1547 -Q -O2 --help=optimizers
1550 Alternatively you can discover which binary optimizations are enabled
1551 by @option{-O3} by using:
1554 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1555 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1556 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1559 @item -no-canonical-prefixes
1560 @opindex no-canonical-prefixes
1561 Do not expand any symbolic links, resolve references to @samp{/../}
1562 or @samp{/./}, or make the path absolute when generating a relative
1567 Display the version number and copyrights of the invoked GCC@.
1571 Invoke all subcommands under a wrapper program. The name of the
1572 wrapper program and its parameters are passed as a comma separated
1576 gcc -c t.c -wrapper gdb,--args
1580 This invokes all subprograms of @command{gcc} under
1581 @samp{gdb --args}, thus the invocation of @command{cc1} is
1582 @samp{gdb --args cc1 @dots{}}.
1584 @item -fplugin=@var{name}.so
1586 Load the plugin code in file @var{name}.so, assumed to be a
1587 shared object to be dlopen'd by the compiler. The base name of
1588 the shared object file is used to identify the plugin for the
1589 purposes of argument parsing (See
1590 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1591 Each plugin should define the callback functions specified in the
1594 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1595 @opindex fplugin-arg
1596 Define an argument called @var{key} with a value of @var{value}
1597 for the plugin called @var{name}.
1599 @item -fdump-ada-spec@r{[}-slim@r{]}
1600 @opindex fdump-ada-spec
1601 For C and C++ source and include files, generate corresponding Ada specs.
1602 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1603 GNAT User's Guide}, which provides detailed documentation on this feature.
1605 @item -fada-spec-parent=@var{unit}
1606 @opindex fada-spec-parent
1607 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1608 Ada specs as child units of parent @var{unit}.
1610 @item -fdump-go-spec=@var{file}
1611 @opindex fdump-go-spec
1612 For input files in any language, generate corresponding Go
1613 declarations in @var{file}. This generates Go @code{const},
1614 @code{type}, @code{var}, and @code{func} declarations which may be a
1615 useful way to start writing a Go interface to code written in some
1618 @include @value{srcdir}/../libiberty/at-file.texi
1622 @section Compiling C++ Programs
1624 @cindex suffixes for C++ source
1625 @cindex C++ source file suffixes
1626 C++ source files conventionally use one of the suffixes @samp{.C},
1627 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1628 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1629 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1630 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1631 files with these names and compiles them as C++ programs even if you
1632 call the compiler the same way as for compiling C programs (usually
1633 with the name @command{gcc}).
1637 However, the use of @command{gcc} does not add the C++ library.
1638 @command{g++} is a program that calls GCC and automatically specifies linking
1639 against the C++ library. It treats @samp{.c},
1640 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1641 files unless @option{-x} is used. This program is also useful when
1642 precompiling a C header file with a @samp{.h} extension for use in C++
1643 compilations. On many systems, @command{g++} is also installed with
1644 the name @command{c++}.
1646 @cindex invoking @command{g++}
1647 When you compile C++ programs, you may specify many of the same
1648 command-line options that you use for compiling programs in any
1649 language; or command-line options meaningful for C and related
1650 languages; or options that are meaningful only for C++ programs.
1651 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1652 explanations of options for languages related to C@.
1653 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1654 explanations of options that are meaningful only for C++ programs.
1656 @node C Dialect Options
1657 @section Options Controlling C Dialect
1658 @cindex dialect options
1659 @cindex language dialect options
1660 @cindex options, dialect
1662 The following options control the dialect of C (or languages derived
1663 from C, such as C++, Objective-C and Objective-C++) that the compiler
1667 @cindex ANSI support
1671 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1672 equivalent to @option{-std=c++98}.
1674 This turns off certain features of GCC that are incompatible with ISO
1675 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1676 such as the @code{asm} and @code{typeof} keywords, and
1677 predefined macros such as @code{unix} and @code{vax} that identify the
1678 type of system you are using. It also enables the undesirable and
1679 rarely used ISO trigraph feature. For the C compiler,
1680 it disables recognition of C++ style @samp{//} comments as well as
1681 the @code{inline} keyword.
1683 The alternate keywords @code{__asm__}, @code{__extension__},
1684 @code{__inline__} and @code{__typeof__} continue to work despite
1685 @option{-ansi}. You would not want to use them in an ISO C program, of
1686 course, but it is useful to put them in header files that might be included
1687 in compilations done with @option{-ansi}. Alternate predefined macros
1688 such as @code{__unix__} and @code{__vax__} are also available, with or
1689 without @option{-ansi}.
1691 The @option{-ansi} option does not cause non-ISO programs to be
1692 rejected gratuitously. For that, @option{-Wpedantic} is required in
1693 addition to @option{-ansi}. @xref{Warning Options}.
1695 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1696 option is used. Some header files may notice this macro and refrain
1697 from declaring certain functions or defining certain macros that the
1698 ISO standard doesn't call for; this is to avoid interfering with any
1699 programs that might use these names for other things.
1701 Functions that are normally built in but do not have semantics
1702 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1703 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1704 built-in functions provided by GCC}, for details of the functions
1709 Determine the language standard. @xref{Standards,,Language Standards
1710 Supported by GCC}, for details of these standard versions. This option
1711 is currently only supported when compiling C or C++.
1713 The compiler can accept several base standards, such as @samp{c90} or
1714 @samp{c++98}, and GNU dialects of those standards, such as
1715 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1716 compiler accepts all programs following that standard plus those
1717 using GNU extensions that do not contradict it. For example,
1718 @option{-std=c90} turns off certain features of GCC that are
1719 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1720 keywords, but not other GNU extensions that do not have a meaning in
1721 ISO C90, such as omitting the middle term of a @code{?:}
1722 expression. On the other hand, when a GNU dialect of a standard is
1723 specified, all features supported by the compiler are enabled, even when
1724 those features change the meaning of the base standard. As a result, some
1725 strict-conforming programs may be rejected. The particular standard
1726 is used by @option{-Wpedantic} to identify which features are GNU
1727 extensions given that version of the standard. For example
1728 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1729 comments, while @option{-std=gnu99 -Wpedantic} does not.
1731 A value for this option must be provided; possible values are
1737 Support all ISO C90 programs (certain GNU extensions that conflict
1738 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1740 @item iso9899:199409
1741 ISO C90 as modified in amendment 1.
1747 ISO C99. This standard is substantially completely supported, modulo
1748 bugs and floating-point issues
1749 (mainly but not entirely relating to optional C99 features from
1750 Annexes F and G). See
1751 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1752 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1757 ISO C11, the 2011 revision of the ISO C standard. This standard is
1758 substantially completely supported, modulo bugs, floating-point issues
1759 (mainly but not entirely relating to optional C11 features from
1760 Annexes F and G) and the optional Annexes K (Bounds-checking
1761 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1765 GNU dialect of ISO C90 (including some C99 features).
1769 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1773 GNU dialect of ISO C11. This is the default for C code.
1774 The name @samp{gnu1x} is deprecated.
1778 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1779 additional defect reports. Same as @option{-ansi} for C++ code.
1783 GNU dialect of @option{-std=c++98}. This is the default for
1788 The 2011 ISO C++ standard plus amendments.
1789 The name @samp{c++0x} is deprecated.
1793 GNU dialect of @option{-std=c++11}.
1794 The name @samp{gnu++0x} is deprecated.
1798 The 2014 ISO C++ standard plus amendments.
1799 The name @samp{c++1y} is deprecated.
1803 GNU dialect of @option{-std=c++14}.
1804 The name @samp{gnu++1y} is deprecated.
1807 The next revision of the ISO C++ standard, tentatively planned for
1808 2017. Support is highly experimental, and will almost certainly
1809 change in incompatible ways in future releases.
1812 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1813 and will almost certainly change in incompatible ways in future
1817 @item -fgnu89-inline
1818 @opindex fgnu89-inline
1819 The option @option{-fgnu89-inline} tells GCC to use the traditional
1820 GNU semantics for @code{inline} functions when in C99 mode.
1821 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1822 Using this option is roughly equivalent to adding the
1823 @code{gnu_inline} function attribute to all inline functions
1824 (@pxref{Function Attributes}).
1826 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1827 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1828 specifies the default behavior).
1829 This option is not supported in @option{-std=c90} or
1830 @option{-std=gnu90} mode.
1832 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1833 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1834 in effect for @code{inline} functions. @xref{Common Predefined
1835 Macros,,,cpp,The C Preprocessor}.
1837 @item -aux-info @var{filename}
1839 Output to the given filename prototyped declarations for all functions
1840 declared and/or defined in a translation unit, including those in header
1841 files. This option is silently ignored in any language other than C@.
1843 Besides declarations, the file indicates, in comments, the origin of
1844 each declaration (source file and line), whether the declaration was
1845 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1846 @samp{O} for old, respectively, in the first character after the line
1847 number and the colon), and whether it came from a declaration or a
1848 definition (@samp{C} or @samp{F}, respectively, in the following
1849 character). In the case of function definitions, a K&R-style list of
1850 arguments followed by their declarations is also provided, inside
1851 comments, after the declaration.
1853 @item -fallow-parameterless-variadic-functions
1854 @opindex fallow-parameterless-variadic-functions
1855 Accept variadic functions without named parameters.
1857 Although it is possible to define such a function, this is not very
1858 useful as it is not possible to read the arguments. This is only
1859 supported for C as this construct is allowed by C++.
1863 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1864 keyword, so that code can use these words as identifiers. You can use
1865 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1866 instead. @option{-ansi} implies @option{-fno-asm}.
1868 In C++, this switch only affects the @code{typeof} keyword, since
1869 @code{asm} and @code{inline} are standard keywords. You may want to
1870 use the @option{-fno-gnu-keywords} flag instead, which has the same
1871 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1872 switch only affects the @code{asm} and @code{typeof} keywords, since
1873 @code{inline} is a standard keyword in ISO C99.
1876 @itemx -fno-builtin-@var{function}
1877 @opindex fno-builtin
1878 @cindex built-in functions
1879 Don't recognize built-in functions that do not begin with
1880 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1881 functions provided by GCC}, for details of the functions affected,
1882 including those which are not built-in functions when @option{-ansi} or
1883 @option{-std} options for strict ISO C conformance are used because they
1884 do not have an ISO standard meaning.
1886 GCC normally generates special code to handle certain built-in functions
1887 more efficiently; for instance, calls to @code{alloca} may become single
1888 instructions which adjust the stack directly, and calls to @code{memcpy}
1889 may become inline copy loops. The resulting code is often both smaller
1890 and faster, but since the function calls no longer appear as such, you
1891 cannot set a breakpoint on those calls, nor can you change the behavior
1892 of the functions by linking with a different library. In addition,
1893 when a function is recognized as a built-in function, GCC may use
1894 information about that function to warn about problems with calls to
1895 that function, or to generate more efficient code, even if the
1896 resulting code still contains calls to that function. For example,
1897 warnings are given with @option{-Wformat} for bad calls to
1898 @code{printf} when @code{printf} is built in and @code{strlen} is
1899 known not to modify global memory.
1901 With the @option{-fno-builtin-@var{function}} option
1902 only the built-in function @var{function} is
1903 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1904 function is named that is not built-in in this version of GCC, this
1905 option is ignored. There is no corresponding
1906 @option{-fbuiltin-@var{function}} option; if you wish to enable
1907 built-in functions selectively when using @option{-fno-builtin} or
1908 @option{-ffreestanding}, you may define macros such as:
1911 #define abs(n) __builtin_abs ((n))
1912 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1917 @cindex hosted environment
1919 Assert that compilation targets a hosted environment. This implies
1920 @option{-fbuiltin}. A hosted environment is one in which the
1921 entire standard library is available, and in which @code{main} has a return
1922 type of @code{int}. Examples are nearly everything except a kernel.
1923 This is equivalent to @option{-fno-freestanding}.
1925 @item -ffreestanding
1926 @opindex ffreestanding
1927 @cindex hosted environment
1929 Assert that compilation targets a freestanding environment. This
1930 implies @option{-fno-builtin}. A freestanding environment
1931 is one in which the standard library may not exist, and program startup may
1932 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1933 This is equivalent to @option{-fno-hosted}.
1935 @xref{Standards,,Language Standards Supported by GCC}, for details of
1936 freestanding and hosted environments.
1940 @cindex OpenACC accelerator programming
1941 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1942 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1943 compiler generates accelerated code according to the OpenACC Application
1944 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1945 implies @option{-pthread}, and thus is only supported on targets that
1946 have support for @option{-pthread}.
1948 Note that this is an experimental feature, incomplete, and subject to
1949 change in future versions of GCC. See
1950 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1954 @cindex OpenMP parallel
1955 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1956 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1957 compiler generates parallel code according to the OpenMP Application
1958 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1959 implies @option{-pthread}, and thus is only supported on targets that
1960 have support for @option{-pthread}. @option{-fopenmp} implies
1961 @option{-fopenmp-simd}.
1964 @opindex fopenmp-simd
1967 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1968 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1973 @cindex Enable Cilk Plus
1974 Enable the usage of Cilk Plus language extension features for C/C++.
1975 When the option @option{-fcilkplus} is specified, enable the usage of
1976 the Cilk Plus Language extension features for C/C++. The present
1977 implementation follows ABI version 1.2. This is an experimental
1978 feature that is only partially complete, and whose interface may
1979 change in future versions of GCC as the official specification
1980 changes. Currently, all features but @code{_Cilk_for} have been
1985 When the option @option{-fgnu-tm} is specified, the compiler
1986 generates code for the Linux variant of Intel's current Transactional
1987 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1988 an experimental feature whose interface may change in future versions
1989 of GCC, as the official specification changes. Please note that not
1990 all architectures are supported for this feature.
1992 For more information on GCC's support for transactional memory,
1993 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1994 Transactional Memory Library}.
1996 Note that the transactional memory feature is not supported with
1997 non-call exceptions (@option{-fnon-call-exceptions}).
1999 @item -fms-extensions
2000 @opindex fms-extensions
2001 Accept some non-standard constructs used in Microsoft header files.
2003 In C++ code, this allows member names in structures to be similar
2004 to previous types declarations.
2013 Some cases of unnamed fields in structures and unions are only
2014 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2015 fields within structs/unions}, for details.
2017 Note that this option is off for all targets but x86
2018 targets using ms-abi.
2020 @item -fplan9-extensions
2021 @opindex fplan9-extensions
2022 Accept some non-standard constructs used in Plan 9 code.
2024 This enables @option{-fms-extensions}, permits passing pointers to
2025 structures with anonymous fields to functions that expect pointers to
2026 elements of the type of the field, and permits referring to anonymous
2027 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2028 struct/union fields within structs/unions}, for details. This is only
2029 supported for C, not C++.
2033 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2034 options for strict ISO C conformance) implies @option{-trigraphs}.
2036 @cindex traditional C language
2037 @cindex C language, traditional
2039 @itemx -traditional-cpp
2040 @opindex traditional-cpp
2041 @opindex traditional
2042 Formerly, these options caused GCC to attempt to emulate a pre-standard
2043 C compiler. They are now only supported with the @option{-E} switch.
2044 The preprocessor continues to support a pre-standard mode. See the GNU
2045 CPP manual for details.
2047 @item -fcond-mismatch
2048 @opindex fcond-mismatch
2049 Allow conditional expressions with mismatched types in the second and
2050 third arguments. The value of such an expression is void. This option
2051 is not supported for C++.
2053 @item -flax-vector-conversions
2054 @opindex flax-vector-conversions
2055 Allow implicit conversions between vectors with differing numbers of
2056 elements and/or incompatible element types. This option should not be
2059 @item -funsigned-char
2060 @opindex funsigned-char
2061 Let the type @code{char} be unsigned, like @code{unsigned char}.
2063 Each kind of machine has a default for what @code{char} should
2064 be. It is either like @code{unsigned char} by default or like
2065 @code{signed char} by default.
2067 Ideally, a portable program should always use @code{signed char} or
2068 @code{unsigned char} when it depends on the signedness of an object.
2069 But many programs have been written to use plain @code{char} and
2070 expect it to be signed, or expect it to be unsigned, depending on the
2071 machines they were written for. This option, and its inverse, let you
2072 make such a program work with the opposite default.
2074 The type @code{char} is always a distinct type from each of
2075 @code{signed char} or @code{unsigned char}, even though its behavior
2076 is always just like one of those two.
2079 @opindex fsigned-char
2080 Let the type @code{char} be signed, like @code{signed char}.
2082 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2083 the negative form of @option{-funsigned-char}. Likewise, the option
2084 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2086 @item -fsigned-bitfields
2087 @itemx -funsigned-bitfields
2088 @itemx -fno-signed-bitfields
2089 @itemx -fno-unsigned-bitfields
2090 @opindex fsigned-bitfields
2091 @opindex funsigned-bitfields
2092 @opindex fno-signed-bitfields
2093 @opindex fno-unsigned-bitfields
2094 These options control whether a bit-field is signed or unsigned, when the
2095 declaration does not use either @code{signed} or @code{unsigned}. By
2096 default, such a bit-field is signed, because this is consistent: the
2097 basic integer types such as @code{int} are signed types.
2099 @item -fsso-struct=@var{endianness}
2100 @opindex fsso-struct
2101 Set the default scalar storage order of structures and unions to the
2102 specified endianness. The accepted values are @samp{big-endian} and
2103 @samp{little-endian}. If the option is not passed, the compiler uses
2104 the native endianness of the target. This option is not supported for C++.
2106 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2107 code that is not binary compatible with code generated without it if the
2108 specified endianness is not the native endianness of the target.
2111 @node C++ Dialect Options
2112 @section Options Controlling C++ Dialect
2114 @cindex compiler options, C++
2115 @cindex C++ options, command-line
2116 @cindex options, C++
2117 This section describes the command-line options that are only meaningful
2118 for C++ programs. You can also use most of the GNU compiler options
2119 regardless of what language your program is in. For example, you
2120 might compile a file @file{firstClass.C} like this:
2123 g++ -g -fstrict-enums -O -c firstClass.C
2127 In this example, only @option{-fstrict-enums} is an option meant
2128 only for C++ programs; you can use the other options with any
2129 language supported by GCC@.
2131 Here is a list of options that are @emph{only} for compiling C++ programs:
2135 @item -fabi-version=@var{n}
2136 @opindex fabi-version
2137 Use version @var{n} of the C++ ABI@. The default is version 0.
2139 Version 0 refers to the version conforming most closely to
2140 the C++ ABI specification. Therefore, the ABI obtained using version 0
2141 will change in different versions of G++ as ABI bugs are fixed.
2143 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2145 Version 2 is the version of the C++ ABI that first appeared in G++
2146 3.4, and was the default through G++ 4.9.
2148 Version 3 corrects an error in mangling a constant address as a
2151 Version 4, which first appeared in G++ 4.5, implements a standard
2152 mangling for vector types.
2154 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2155 attribute const/volatile on function pointer types, decltype of a
2156 plain decl, and use of a function parameter in the declaration of
2159 Version 6, which first appeared in G++ 4.7, corrects the promotion
2160 behavior of C++11 scoped enums and the mangling of template argument
2161 packs, const/static_cast, prefix ++ and --, and a class scope function
2162 used as a template argument.
2164 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2165 builtin type and corrects the mangling of lambdas in default argument
2168 Version 8, which first appeared in G++ 4.9, corrects the substitution
2169 behavior of function types with function-cv-qualifiers.
2171 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2174 Version 10, which first appeared in G++ 6.1, adds mangling of
2175 attributes that affect type identity, such as ia32 calling convention
2176 attributes (e.g. @samp{stdcall}).
2178 See also @option{-Wabi}.
2180 @item -fabi-compat-version=@var{n}
2181 @opindex fabi-compat-version
2182 On targets that support strong aliases, G++
2183 works around mangling changes by creating an alias with the correct
2184 mangled name when defining a symbol with an incorrect mangled name.
2185 This switch specifies which ABI version to use for the alias.
2187 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2188 compatibility). If another ABI version is explicitly selected, this
2189 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2190 use @option{-fabi-compat-version=2}.
2192 If this option is not provided but @option{-Wabi=@var{n}} is, that
2193 version is used for compatibility aliases. If this option is provided
2194 along with @option{-Wabi} (without the version), the version from this
2195 option is used for the warning.
2197 @item -fno-access-control
2198 @opindex fno-access-control
2199 Turn off all access checking. This switch is mainly useful for working
2200 around bugs in the access control code.
2204 Check that the pointer returned by @code{operator new} is non-null
2205 before attempting to modify the storage allocated. This check is
2206 normally unnecessary because the C++ standard specifies that
2207 @code{operator new} only returns @code{0} if it is declared
2208 @code{throw()}, in which case the compiler always checks the
2209 return value even without this option. In all other cases, when
2210 @code{operator new} has a non-empty exception specification, memory
2211 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2212 @samp{new (nothrow)}.
2214 @item -fconstexpr-depth=@var{n}
2215 @opindex fconstexpr-depth
2216 Set the maximum nested evaluation depth for C++11 constexpr functions
2217 to @var{n}. A limit is needed to detect endless recursion during
2218 constant expression evaluation. The minimum specified by the standard
2221 @item -fdeduce-init-list
2222 @opindex fdeduce-init-list
2223 Enable deduction of a template type parameter as
2224 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2227 template <class T> auto forward(T t) -> decltype (realfn (t))
2234 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2238 This deduction was implemented as a possible extension to the
2239 originally proposed semantics for the C++11 standard, but was not part
2240 of the final standard, so it is disabled by default. This option is
2241 deprecated, and may be removed in a future version of G++.
2243 @item -ffriend-injection
2244 @opindex ffriend-injection
2245 Inject friend functions into the enclosing namespace, so that they are
2246 visible outside the scope of the class in which they are declared.
2247 Friend functions were documented to work this way in the old Annotated
2248 C++ Reference Manual.
2249 However, in ISO C++ a friend function that is not declared
2250 in an enclosing scope can only be found using argument dependent
2251 lookup. GCC defaults to the standard behavior.
2253 This option is for compatibility, and may be removed in a future
2256 @item -fno-elide-constructors
2257 @opindex fno-elide-constructors
2258 The C++ standard allows an implementation to omit creating a temporary
2259 that is only used to initialize another object of the same type.
2260 Specifying this option disables that optimization, and forces G++ to
2261 call the copy constructor in all cases.
2263 @item -fno-enforce-eh-specs
2264 @opindex fno-enforce-eh-specs
2265 Don't generate code to check for violation of exception specifications
2266 at run time. This option violates the C++ standard, but may be useful
2267 for reducing code size in production builds, much like defining
2268 @code{NDEBUG}. This does not give user code permission to throw
2269 exceptions in violation of the exception specifications; the compiler
2270 still optimizes based on the specifications, so throwing an
2271 unexpected exception results in undefined behavior at run time.
2273 @item -fextern-tls-init
2274 @itemx -fno-extern-tls-init
2275 @opindex fextern-tls-init
2276 @opindex fno-extern-tls-init
2277 The C++11 and OpenMP standards allow @code{thread_local} and
2278 @code{threadprivate} variables to have dynamic (runtime)
2279 initialization. To support this, any use of such a variable goes
2280 through a wrapper function that performs any necessary initialization.
2281 When the use and definition of the variable are in the same
2282 translation unit, this overhead can be optimized away, but when the
2283 use is in a different translation unit there is significant overhead
2284 even if the variable doesn't actually need dynamic initialization. If
2285 the programmer can be sure that no use of the variable in a
2286 non-defining TU needs to trigger dynamic initialization (either
2287 because the variable is statically initialized, or a use of the
2288 variable in the defining TU will be executed before any uses in
2289 another TU), they can avoid this overhead with the
2290 @option{-fno-extern-tls-init} option.
2292 On targets that support symbol aliases, the default is
2293 @option{-fextern-tls-init}. On targets that do not support symbol
2294 aliases, the default is @option{-fno-extern-tls-init}.
2297 @itemx -fno-for-scope
2299 @opindex fno-for-scope
2300 If @option{-ffor-scope} is specified, the scope of variables declared in
2301 a @i{for-init-statement} is limited to the @code{for} loop itself,
2302 as specified by the C++ standard.
2303 If @option{-fno-for-scope} is specified, the scope of variables declared in
2304 a @i{for-init-statement} extends to the end of the enclosing scope,
2305 as was the case in old versions of G++, and other (traditional)
2306 implementations of C++.
2308 If neither flag is given, the default is to follow the standard,
2309 but to allow and give a warning for old-style code that would
2310 otherwise be invalid, or have different behavior.
2312 @item -fno-gnu-keywords
2313 @opindex fno-gnu-keywords
2314 Do not recognize @code{typeof} as a keyword, so that code can use this
2315 word as an identifier. You can use the keyword @code{__typeof__} instead.
2316 @option{-ansi} implies @option{-fno-gnu-keywords}.
2318 @item -fno-implicit-templates
2319 @opindex fno-implicit-templates
2320 Never emit code for non-inline templates that are instantiated
2321 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2322 @xref{Template Instantiation}, for more information.
2324 @item -fno-implicit-inline-templates
2325 @opindex fno-implicit-inline-templates
2326 Don't emit code for implicit instantiations of inline templates, either.
2327 The default is to handle inlines differently so that compiles with and
2328 without optimization need the same set of explicit instantiations.
2330 @item -fno-implement-inlines
2331 @opindex fno-implement-inlines
2332 To save space, do not emit out-of-line copies of inline functions
2333 controlled by @code{#pragma implementation}. This causes linker
2334 errors if these functions are not inlined everywhere they are called.
2336 @item -fms-extensions
2337 @opindex fms-extensions
2338 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2339 int and getting a pointer to member function via non-standard syntax.
2341 @item -fno-nonansi-builtins
2342 @opindex fno-nonansi-builtins
2343 Disable built-in declarations of functions that are not mandated by
2344 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2345 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2348 @opindex fnothrow-opt
2349 Treat a @code{throw()} exception specification as if it were a
2350 @code{noexcept} specification to reduce or eliminate the text size
2351 overhead relative to a function with no exception specification. If
2352 the function has local variables of types with non-trivial
2353 destructors, the exception specification actually makes the
2354 function smaller because the EH cleanups for those variables can be
2355 optimized away. The semantic effect is that an exception thrown out of
2356 a function with such an exception specification results in a call
2357 to @code{terminate} rather than @code{unexpected}.
2359 @item -fno-operator-names
2360 @opindex fno-operator-names
2361 Do not treat the operator name keywords @code{and}, @code{bitand},
2362 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2363 synonyms as keywords.
2365 @item -fno-optional-diags
2366 @opindex fno-optional-diags
2367 Disable diagnostics that the standard says a compiler does not need to
2368 issue. Currently, the only such diagnostic issued by G++ is the one for
2369 a name having multiple meanings within a class.
2372 @opindex fpermissive
2373 Downgrade some diagnostics about nonconformant code from errors to
2374 warnings. Thus, using @option{-fpermissive} allows some
2375 nonconforming code to compile.
2377 @item -fno-pretty-templates
2378 @opindex fno-pretty-templates
2379 When an error message refers to a specialization of a function
2380 template, the compiler normally prints the signature of the
2381 template followed by the template arguments and any typedefs or
2382 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2383 rather than @code{void f(int)}) so that it's clear which template is
2384 involved. When an error message refers to a specialization of a class
2385 template, the compiler omits any template arguments that match
2386 the default template arguments for that template. If either of these
2387 behaviors make it harder to understand the error message rather than
2388 easier, you can use @option{-fno-pretty-templates} to disable them.
2392 Enable automatic template instantiation at link time. This option also
2393 implies @option{-fno-implicit-templates}. @xref{Template
2394 Instantiation}, for more information.
2398 Disable generation of information about every class with virtual
2399 functions for use by the C++ run-time type identification features
2400 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2401 of the language, you can save some space by using this flag. Note that
2402 exception handling uses the same information, but G++ generates it as
2403 needed. The @code{dynamic_cast} operator can still be used for casts that
2404 do not require run-time type information, i.e.@: casts to @code{void *} or to
2405 unambiguous base classes.
2407 @item -fsized-deallocation
2408 @opindex fsized-deallocation
2409 Enable the built-in global declarations
2411 void operator delete (void *, std::size_t) noexcept;
2412 void operator delete[] (void *, std::size_t) noexcept;
2414 as introduced in C++14. This is useful for user-defined replacement
2415 deallocation functions that, for example, use the size of the object
2416 to make deallocation faster. Enabled by default under
2417 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2418 warns about places that might want to add a definition.
2422 Emit statistics about front-end processing at the end of the compilation.
2423 This information is generally only useful to the G++ development team.
2425 @item -fstrict-enums
2426 @opindex fstrict-enums
2427 Allow the compiler to optimize using the assumption that a value of
2428 enumerated type can only be one of the values of the enumeration (as
2429 defined in the C++ standard; basically, a value that can be
2430 represented in the minimum number of bits needed to represent all the
2431 enumerators). This assumption may not be valid if the program uses a
2432 cast to convert an arbitrary integer value to the enumerated type.
2434 @item -ftemplate-backtrace-limit=@var{n}
2435 @opindex ftemplate-backtrace-limit
2436 Set the maximum number of template instantiation notes for a single
2437 warning or error to @var{n}. The default value is 10.
2439 @item -ftemplate-depth=@var{n}
2440 @opindex ftemplate-depth
2441 Set the maximum instantiation depth for template classes to @var{n}.
2442 A limit on the template instantiation depth is needed to detect
2443 endless recursions during template class instantiation. ANSI/ISO C++
2444 conforming programs must not rely on a maximum depth greater than 17
2445 (changed to 1024 in C++11). The default value is 900, as the compiler
2446 can run out of stack space before hitting 1024 in some situations.
2448 @item -fno-threadsafe-statics
2449 @opindex fno-threadsafe-statics
2450 Do not emit the extra code to use the routines specified in the C++
2451 ABI for thread-safe initialization of local statics. You can use this
2452 option to reduce code size slightly in code that doesn't need to be
2455 @item -fuse-cxa-atexit
2456 @opindex fuse-cxa-atexit
2457 Register destructors for objects with static storage duration with the
2458 @code{__cxa_atexit} function rather than the @code{atexit} function.
2459 This option is required for fully standards-compliant handling of static
2460 destructors, but only works if your C library supports
2461 @code{__cxa_atexit}.
2463 @item -fno-use-cxa-get-exception-ptr
2464 @opindex fno-use-cxa-get-exception-ptr
2465 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2466 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2467 if the runtime routine is not available.
2469 @item -fvisibility-inlines-hidden
2470 @opindex fvisibility-inlines-hidden
2471 This switch declares that the user does not attempt to compare
2472 pointers to inline functions or methods where the addresses of the two functions
2473 are taken in different shared objects.
2475 The effect of this is that GCC may, effectively, mark inline methods with
2476 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2477 appear in the export table of a DSO and do not require a PLT indirection
2478 when used within the DSO@. Enabling this option can have a dramatic effect
2479 on load and link times of a DSO as it massively reduces the size of the
2480 dynamic export table when the library makes heavy use of templates.
2482 The behavior of this switch is not quite the same as marking the
2483 methods as hidden directly, because it does not affect static variables
2484 local to the function or cause the compiler to deduce that
2485 the function is defined in only one shared object.
2487 You may mark a method as having a visibility explicitly to negate the
2488 effect of the switch for that method. For example, if you do want to
2489 compare pointers to a particular inline method, you might mark it as
2490 having default visibility. Marking the enclosing class with explicit
2491 visibility has no effect.
2493 Explicitly instantiated inline methods are unaffected by this option
2494 as their linkage might otherwise cross a shared library boundary.
2495 @xref{Template Instantiation}.
2497 @item -fvisibility-ms-compat
2498 @opindex fvisibility-ms-compat
2499 This flag attempts to use visibility settings to make GCC's C++
2500 linkage model compatible with that of Microsoft Visual Studio.
2502 The flag makes these changes to GCC's linkage model:
2506 It sets the default visibility to @code{hidden}, like
2507 @option{-fvisibility=hidden}.
2510 Types, but not their members, are not hidden by default.
2513 The One Definition Rule is relaxed for types without explicit
2514 visibility specifications that are defined in more than one
2515 shared object: those declarations are permitted if they are
2516 permitted when this option is not used.
2519 In new code it is better to use @option{-fvisibility=hidden} and
2520 export those classes that are intended to be externally visible.
2521 Unfortunately it is possible for code to rely, perhaps accidentally,
2522 on the Visual Studio behavior.
2524 Among the consequences of these changes are that static data members
2525 of the same type with the same name but defined in different shared
2526 objects are different, so changing one does not change the other;
2527 and that pointers to function members defined in different shared
2528 objects may not compare equal. When this flag is given, it is a
2529 violation of the ODR to define types with the same name differently.
2531 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2532 @opindex fvtable-verify
2533 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2534 feature that verifies at run time, for every virtual call, that
2535 the vtable pointer through which the call is made is valid for the type of
2536 the object, and has not been corrupted or overwritten. If an invalid vtable
2537 pointer is detected at run time, an error is reported and execution of the
2538 program is immediately halted.
2540 This option causes run-time data structures to be built at program startup,
2541 which are used for verifying the vtable pointers.
2542 The options @samp{std} and @samp{preinit}
2543 control the timing of when these data structures are built. In both cases the
2544 data structures are built before execution reaches @code{main}. Using
2545 @option{-fvtable-verify=std} causes the data structures to be built after
2546 shared libraries have been loaded and initialized.
2547 @option{-fvtable-verify=preinit} causes them to be built before shared
2548 libraries have been loaded and initialized.
2550 If this option appears multiple times in the command line with different
2551 values specified, @samp{none} takes highest priority over both @samp{std} and
2552 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2556 When used in conjunction with @option{-fvtable-verify=std} or
2557 @option{-fvtable-verify=preinit}, causes debug versions of the
2558 runtime functions for the vtable verification feature to be called.
2559 This flag also causes the compiler to log information about which
2560 vtable pointers it finds for each class.
2561 This information is written to a file named @file{vtv_set_ptr_data.log}
2562 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2563 if that is defined or the current working directory otherwise.
2565 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2566 file, be sure to delete any existing one.
2569 @opindex fvtv-counts
2570 This is a debugging flag. When used in conjunction with
2571 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2572 causes the compiler to keep track of the total number of virtual calls
2573 it encounters and the number of verifications it inserts. It also
2574 counts the number of calls to certain run-time library functions
2575 that it inserts and logs this information for each compilation unit.
2576 The compiler writes this information to a file named
2577 @file{vtv_count_data.log} in the directory named by the environment
2578 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2579 directory otherwise. It also counts the size of the vtable pointer sets
2580 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2581 in the same directory.
2583 Note: This feature @emph{appends} data to the log files. To get fresh log
2584 files, be sure to delete any existing ones.
2588 Do not use weak symbol support, even if it is provided by the linker.
2589 By default, G++ uses weak symbols if they are available. This
2590 option exists only for testing, and should not be used by end-users;
2591 it results in inferior code and has no benefits. This option may
2592 be removed in a future release of G++.
2596 Do not search for header files in the standard directories specific to
2597 C++, but do still search the other standard directories. (This option
2598 is used when building the C++ library.)
2601 In addition, these optimization, warning, and code generation options
2602 have meanings only for C++ programs:
2605 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2608 Warn when G++ it generates code that is probably not compatible with
2609 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2610 ABI with each major release, normally @option{-Wabi} will warn only if
2611 there is a check added later in a release series for an ABI issue
2612 discovered since the initial release. @option{-Wabi} will warn about
2613 more things if an older ABI version is selected (with
2614 @option{-fabi-version=@var{n}}).
2616 @option{-Wabi} can also be used with an explicit version number to
2617 warn about compatibility with a particular @option{-fabi-version}
2618 level, e.g. @option{-Wabi=2} to warn about changes relative to
2619 @option{-fabi-version=2}.
2621 If an explicit version number is provided and
2622 @option{-fabi-compat-version} is not specified, the version number
2623 from this option is used for compatibility aliases. If no explicit
2624 version number is provided with this option, but
2625 @option{-fabi-compat-version} is specified, that version number is
2626 used for ABI warnings.
2628 Although an effort has been made to warn about
2629 all such cases, there are probably some cases that are not warned about,
2630 even though G++ is generating incompatible code. There may also be
2631 cases where warnings are emitted even though the code that is generated
2634 You should rewrite your code to avoid these warnings if you are
2635 concerned about the fact that code generated by G++ may not be binary
2636 compatible with code generated by other compilers.
2638 Known incompatibilities in @option{-fabi-version=2} (which was the
2639 default from GCC 3.4 to 4.9) include:
2644 A template with a non-type template parameter of reference type was
2645 mangled incorrectly:
2648 template <int &> struct S @{@};
2652 This was fixed in @option{-fabi-version=3}.
2655 SIMD vector types declared using @code{__attribute ((vector_size))} were
2656 mangled in a non-standard way that does not allow for overloading of
2657 functions taking vectors of different sizes.
2659 The mangling was changed in @option{-fabi-version=4}.
2662 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2663 qualifiers, and @code{decltype} of a plain declaration was folded away.
2665 These mangling issues were fixed in @option{-fabi-version=5}.
2668 Scoped enumerators passed as arguments to a variadic function are
2669 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2670 On most targets this does not actually affect the parameter passing
2671 ABI, as there is no way to pass an argument smaller than @code{int}.
2673 Also, the ABI changed the mangling of template argument packs,
2674 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2675 a class scope function used as a template argument.
2677 These issues were corrected in @option{-fabi-version=6}.
2680 Lambdas in default argument scope were mangled incorrectly, and the
2681 ABI changed the mangling of @code{nullptr_t}.
2683 These issues were corrected in @option{-fabi-version=7}.
2686 When mangling a function type with function-cv-qualifiers, the
2687 un-qualified function type was incorrectly treated as a substitution
2690 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2693 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2694 unaligned accesses. Note that this did not affect the ABI of a
2695 function with a @code{nullptr_t} parameter, as parameters have a
2698 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2701 Target-specific attributes that affect the identity of a type, such as
2702 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2703 did not affect the mangled name, leading to name collisions when
2704 function pointers were used as template arguments.
2706 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2710 It also warns about psABI-related changes. The known psABI changes at this
2716 For SysV/x86-64, unions with @code{long double} members are
2717 passed in memory as specified in psABI. For example:
2727 @code{union U} is always passed in memory.
2731 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2734 Warn when a type with an ABI tag is used in a context that does not
2735 have that ABI tag. See @ref{C++ Attributes} for more information
2738 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2739 @opindex Wctor-dtor-privacy
2740 @opindex Wno-ctor-dtor-privacy
2741 Warn when a class seems unusable because all the constructors or
2742 destructors in that class are private, and it has neither friends nor
2743 public static member functions. Also warn if there are no non-private
2744 methods, and there's at least one private member function that isn't
2745 a constructor or destructor.
2747 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2748 @opindex Wdelete-non-virtual-dtor
2749 @opindex Wno-delete-non-virtual-dtor
2750 Warn when @code{delete} is used to destroy an instance of a class that
2751 has virtual functions and non-virtual destructor. It is unsafe to delete
2752 an instance of a derived class through a pointer to a base class if the
2753 base class does not have a virtual destructor. This warning is enabled
2756 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2757 @opindex Wliteral-suffix
2758 @opindex Wno-literal-suffix
2759 Warn when a string or character literal is followed by a ud-suffix which does
2760 not begin with an underscore. As a conforming extension, GCC treats such
2761 suffixes as separate preprocessing tokens in order to maintain backwards
2762 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2766 #define __STDC_FORMAT_MACROS
2767 #include <inttypes.h>
2772 printf("My int64: %" PRId64"\n", i64);
2776 In this case, @code{PRId64} is treated as a separate preprocessing token.
2778 This warning is enabled by default.
2780 @item -Wlto-type-mismatch
2781 @opindex Wlto-type-mismatch
2782 @opindex Wno-lto-type-mistmach
2784 During the link-time optimization warn about type mismatches in between
2785 global declarations from different compilation units.
2786 Requires @option{-flto} to be enabled. Enabled by default.
2788 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2790 @opindex Wno-narrowing
2791 Warn when a narrowing conversion prohibited by C++11 occurs within
2795 int i = @{ 2.2 @}; // error: narrowing from double to int
2798 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2800 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2801 required by the standard. Note that this does not affect the meaning
2802 of well-formed code; narrowing conversions are still considered
2803 ill-formed in SFINAE context.
2805 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2807 @opindex Wno-noexcept
2808 Warn when a noexcept-expression evaluates to false because of a call
2809 to a function that does not have a non-throwing exception
2810 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2811 the compiler to never throw an exception.
2813 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2814 @opindex Wnon-virtual-dtor
2815 @opindex Wno-non-virtual-dtor
2816 Warn when a class has virtual functions and an accessible non-virtual
2817 destructor itself or in an accessible polymorphic base class, in which
2818 case it is possible but unsafe to delete an instance of a derived
2819 class through a pointer to the class itself or base class. This
2820 warning is automatically enabled if @option{-Weffc++} is specified.
2822 @item -Wreorder @r{(C++ and Objective-C++ only)}
2824 @opindex Wno-reorder
2825 @cindex reordering, warning
2826 @cindex warning for reordering of member initializers
2827 Warn when the order of member initializers given in the code does not
2828 match the order in which they must be executed. For instance:
2834 A(): j (0), i (1) @{ @}
2839 The compiler rearranges the member initializers for @code{i}
2840 and @code{j} to match the declaration order of the members, emitting
2841 a warning to that effect. This warning is enabled by @option{-Wall}.
2843 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2844 @opindex fext-numeric-literals
2845 @opindex fno-ext-numeric-literals
2846 Accept imaginary, fixed-point, or machine-defined
2847 literal number suffixes as GNU extensions.
2848 When this option is turned off these suffixes are treated
2849 as C++11 user-defined literal numeric suffixes.
2850 This is on by default for all pre-C++11 dialects and all GNU dialects:
2851 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2852 @option{-std=gnu++14}.
2853 This option is off by default
2854 for ISO C++11 onwards (@option{-std=c++11}, ...).
2857 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2860 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2863 Warn about violations of the following style guidelines from Scott Meyers'
2864 @cite{Effective C++} series of books:
2868 Define a copy constructor and an assignment operator for classes
2869 with dynamically-allocated memory.
2872 Prefer initialization to assignment in constructors.
2875 Have @code{operator=} return a reference to @code{*this}.
2878 Don't try to return a reference when you must return an object.
2881 Distinguish between prefix and postfix forms of increment and
2882 decrement operators.
2885 Never overload @code{&&}, @code{||}, or @code{,}.
2889 This option also enables @option{-Wnon-virtual-dtor}, which is also
2890 one of the effective C++ recommendations. However, the check is
2891 extended to warn about the lack of virtual destructor in accessible
2892 non-polymorphic bases classes too.
2894 When selecting this option, be aware that the standard library
2895 headers do not obey all of these guidelines; use @samp{grep -v}
2896 to filter out those warnings.
2898 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2899 @opindex Wstrict-null-sentinel
2900 @opindex Wno-strict-null-sentinel
2901 Warn about the use of an uncasted @code{NULL} as sentinel. When
2902 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2903 to @code{__null}. Although it is a null pointer constant rather than a
2904 null pointer, it is guaranteed to be of the same size as a pointer.
2905 But this use is not portable across different compilers.
2907 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2908 @opindex Wno-non-template-friend
2909 @opindex Wnon-template-friend
2910 Disable warnings when non-templatized friend functions are declared
2911 within a template. Since the advent of explicit template specification
2912 support in G++, if the name of the friend is an unqualified-id (i.e.,
2913 @samp{friend foo(int)}), the C++ language specification demands that the
2914 friend declare or define an ordinary, nontemplate function. (Section
2915 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2916 could be interpreted as a particular specialization of a templatized
2917 function. Because this non-conforming behavior is no longer the default
2918 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2919 check existing code for potential trouble spots and is on by default.
2920 This new compiler behavior can be turned off with
2921 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2922 but disables the helpful warning.
2924 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2925 @opindex Wold-style-cast
2926 @opindex Wno-old-style-cast
2927 Warn if an old-style (C-style) cast to a non-void type is used within
2928 a C++ program. The new-style casts (@code{dynamic_cast},
2929 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2930 less vulnerable to unintended effects and much easier to search for.
2932 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2933 @opindex Woverloaded-virtual
2934 @opindex Wno-overloaded-virtual
2935 @cindex overloaded virtual function, warning
2936 @cindex warning for overloaded virtual function
2937 Warn when a function declaration hides virtual functions from a
2938 base class. For example, in:
2945 struct B: public A @{
2950 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2961 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2962 @opindex Wno-pmf-conversions
2963 @opindex Wpmf-conversions
2964 Disable the diagnostic for converting a bound pointer to member function
2967 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2968 @opindex Wsign-promo
2969 @opindex Wno-sign-promo
2970 Warn when overload resolution chooses a promotion from unsigned or
2971 enumerated type to a signed type, over a conversion to an unsigned type of
2972 the same size. Previous versions of G++ tried to preserve
2973 unsignedness, but the standard mandates the current behavior.
2975 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2977 Warn when a primary template declaration is encountered. Some coding
2978 rules disallow templates, and this may be used to enforce that rule.
2979 The warning is inactive inside a system header file, such as the STL, so
2980 one can still use the STL. One may also instantiate or specialize
2983 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2984 @opindex Wmultiple-inheritance
2985 Warn when a class is defined with multiple direct base classes. Some
2986 coding rules disallow multiple inheritance, and this may be used to
2987 enforce that rule. The warning is inactive inside a system header file,
2988 such as the STL, so one can still use the STL. One may also define
2989 classes that indirectly use multiple inheritance.
2991 @item -Wvirtual-inheritance
2992 @opindex Wvirtual-inheritance
2993 Warn when a class is defined with a virtual direct base classe. Some
2994 coding rules disallow multiple inheritance, and this may be used to
2995 enforce that rule. The warning is inactive inside a system header file,
2996 such as the STL, so one can still use the STL. One may also define
2997 classes that indirectly use virtual inheritance.
3000 @opindex Wnamespaces
3001 Warn when a namespace definition is opened. Some coding rules disallow
3002 namespaces, and this may be used to enforce that rule. The warning is
3003 inactive inside a system header file, such as the STL, so one can still
3004 use the STL. One may also use using directives and qualified names.
3006 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3008 @opindex Wno-terminate
3009 Disable the warning about a throw-expression that will immediately
3010 result in a call to @code{terminate}.
3013 @node Objective-C and Objective-C++ Dialect Options
3014 @section Options Controlling Objective-C and Objective-C++ Dialects
3016 @cindex compiler options, Objective-C and Objective-C++
3017 @cindex Objective-C and Objective-C++ options, command-line
3018 @cindex options, Objective-C and Objective-C++
3019 (NOTE: This manual does not describe the Objective-C and Objective-C++
3020 languages themselves. @xref{Standards,,Language Standards
3021 Supported by GCC}, for references.)
3023 This section describes the command-line options that are only meaningful
3024 for Objective-C and Objective-C++ programs. You can also use most of
3025 the language-independent GNU compiler options.
3026 For example, you might compile a file @file{some_class.m} like this:
3029 gcc -g -fgnu-runtime -O -c some_class.m
3033 In this example, @option{-fgnu-runtime} is an option meant only for
3034 Objective-C and Objective-C++ programs; you can use the other options with
3035 any language supported by GCC@.
3037 Note that since Objective-C is an extension of the C language, Objective-C
3038 compilations may also use options specific to the C front-end (e.g.,
3039 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3040 C++-specific options (e.g., @option{-Wabi}).
3042 Here is a list of options that are @emph{only} for compiling Objective-C
3043 and Objective-C++ programs:
3046 @item -fconstant-string-class=@var{class-name}
3047 @opindex fconstant-string-class
3048 Use @var{class-name} as the name of the class to instantiate for each
3049 literal string specified with the syntax @code{@@"@dots{}"}. The default
3050 class name is @code{NXConstantString} if the GNU runtime is being used, and
3051 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3052 @option{-fconstant-cfstrings} option, if also present, overrides the
3053 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3054 to be laid out as constant CoreFoundation strings.
3057 @opindex fgnu-runtime
3058 Generate object code compatible with the standard GNU Objective-C
3059 runtime. This is the default for most types of systems.
3061 @item -fnext-runtime
3062 @opindex fnext-runtime
3063 Generate output compatible with the NeXT runtime. This is the default
3064 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3065 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3068 @item -fno-nil-receivers
3069 @opindex fno-nil-receivers
3070 Assume that all Objective-C message dispatches (@code{[receiver
3071 message:arg]}) in this translation unit ensure that the receiver is
3072 not @code{nil}. This allows for more efficient entry points in the
3073 runtime to be used. This option is only available in conjunction with
3074 the NeXT runtime and ABI version 0 or 1.
3076 @item -fobjc-abi-version=@var{n}
3077 @opindex fobjc-abi-version
3078 Use version @var{n} of the Objective-C ABI for the selected runtime.
3079 This option is currently supported only for the NeXT runtime. In that
3080 case, Version 0 is the traditional (32-bit) ABI without support for
3081 properties and other Objective-C 2.0 additions. Version 1 is the
3082 traditional (32-bit) ABI with support for properties and other
3083 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3084 nothing is specified, the default is Version 0 on 32-bit target
3085 machines, and Version 2 on 64-bit target machines.
3087 @item -fobjc-call-cxx-cdtors
3088 @opindex fobjc-call-cxx-cdtors
3089 For each Objective-C class, check if any of its instance variables is a
3090 C++ object with a non-trivial default constructor. If so, synthesize a
3091 special @code{- (id) .cxx_construct} instance method which runs
3092 non-trivial default constructors on any such instance variables, in order,
3093 and then return @code{self}. Similarly, check if any instance variable
3094 is a C++ object with a non-trivial destructor, and if so, synthesize a
3095 special @code{- (void) .cxx_destruct} method which runs
3096 all such default destructors, in reverse order.
3098 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3099 methods thusly generated only operate on instance variables
3100 declared in the current Objective-C class, and not those inherited
3101 from superclasses. It is the responsibility of the Objective-C
3102 runtime to invoke all such methods in an object's inheritance
3103 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3104 by the runtime immediately after a new object instance is allocated;
3105 the @code{- (void) .cxx_destruct} methods are invoked immediately
3106 before the runtime deallocates an object instance.
3108 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3109 support for invoking the @code{- (id) .cxx_construct} and
3110 @code{- (void) .cxx_destruct} methods.
3112 @item -fobjc-direct-dispatch
3113 @opindex fobjc-direct-dispatch
3114 Allow fast jumps to the message dispatcher. On Darwin this is
3115 accomplished via the comm page.
3117 @item -fobjc-exceptions
3118 @opindex fobjc-exceptions
3119 Enable syntactic support for structured exception handling in
3120 Objective-C, similar to what is offered by C++ and Java. This option
3121 is required to use the Objective-C keywords @code{@@try},
3122 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3123 @code{@@synchronized}. This option is available with both the GNU
3124 runtime and the NeXT runtime (but not available in conjunction with
3125 the NeXT runtime on Mac OS X 10.2 and earlier).
3129 Enable garbage collection (GC) in Objective-C and Objective-C++
3130 programs. This option is only available with the NeXT runtime; the
3131 GNU runtime has a different garbage collection implementation that
3132 does not require special compiler flags.
3134 @item -fobjc-nilcheck
3135 @opindex fobjc-nilcheck
3136 For the NeXT runtime with version 2 of the ABI, check for a nil
3137 receiver in method invocations before doing the actual method call.
3138 This is the default and can be disabled using
3139 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3140 checked for nil in this way no matter what this flag is set to.
3141 Currently this flag does nothing when the GNU runtime, or an older
3142 version of the NeXT runtime ABI, is used.
3144 @item -fobjc-std=objc1
3146 Conform to the language syntax of Objective-C 1.0, the language
3147 recognized by GCC 4.0. This only affects the Objective-C additions to
3148 the C/C++ language; it does not affect conformance to C/C++ standards,
3149 which is controlled by the separate C/C++ dialect option flags. When
3150 this option is used with the Objective-C or Objective-C++ compiler,
3151 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3152 This is useful if you need to make sure that your Objective-C code can
3153 be compiled with older versions of GCC@.
3155 @item -freplace-objc-classes
3156 @opindex freplace-objc-classes
3157 Emit a special marker instructing @command{ld(1)} not to statically link in
3158 the resulting object file, and allow @command{dyld(1)} to load it in at
3159 run time instead. This is used in conjunction with the Fix-and-Continue
3160 debugging mode, where the object file in question may be recompiled and
3161 dynamically reloaded in the course of program execution, without the need
3162 to restart the program itself. Currently, Fix-and-Continue functionality
3163 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3168 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3169 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3170 compile time) with static class references that get initialized at load time,
3171 which improves run-time performance. Specifying the @option{-fzero-link} flag
3172 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3173 to be retained. This is useful in Zero-Link debugging mode, since it allows
3174 for individual class implementations to be modified during program execution.
3175 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3176 regardless of command-line options.
3178 @item -fno-local-ivars
3179 @opindex fno-local-ivars
3180 @opindex flocal-ivars
3181 By default instance variables in Objective-C can be accessed as if
3182 they were local variables from within the methods of the class they're
3183 declared in. This can lead to shadowing between instance variables
3184 and other variables declared either locally inside a class method or
3185 globally with the same name. Specifying the @option{-fno-local-ivars}
3186 flag disables this behavior thus avoiding variable shadowing issues.
3188 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3189 @opindex fivar-visibility
3190 Set the default instance variable visibility to the specified option
3191 so that instance variables declared outside the scope of any access
3192 modifier directives default to the specified visibility.
3196 Dump interface declarations for all classes seen in the source file to a
3197 file named @file{@var{sourcename}.decl}.
3199 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3200 @opindex Wassign-intercept
3201 @opindex Wno-assign-intercept
3202 Warn whenever an Objective-C assignment is being intercepted by the
3205 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3206 @opindex Wno-protocol
3208 If a class is declared to implement a protocol, a warning is issued for
3209 every method in the protocol that is not implemented by the class. The
3210 default behavior is to issue a warning for every method not explicitly
3211 implemented in the class, even if a method implementation is inherited
3212 from the superclass. If you use the @option{-Wno-protocol} option, then
3213 methods inherited from the superclass are considered to be implemented,
3214 and no warning is issued for them.
3216 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3218 @opindex Wno-selector
3219 Warn if multiple methods of different types for the same selector are
3220 found during compilation. The check is performed on the list of methods
3221 in the final stage of compilation. Additionally, a check is performed
3222 for each selector appearing in a @code{@@selector(@dots{})}
3223 expression, and a corresponding method for that selector has been found
3224 during compilation. Because these checks scan the method table only at
3225 the end of compilation, these warnings are not produced if the final
3226 stage of compilation is not reached, for example because an error is
3227 found during compilation, or because the @option{-fsyntax-only} option is
3230 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3231 @opindex Wstrict-selector-match
3232 @opindex Wno-strict-selector-match
3233 Warn if multiple methods with differing argument and/or return types are
3234 found for a given selector when attempting to send a message using this
3235 selector to a receiver of type @code{id} or @code{Class}. When this flag
3236 is off (which is the default behavior), the compiler omits such warnings
3237 if any differences found are confined to types that share the same size
3240 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3241 @opindex Wundeclared-selector
3242 @opindex Wno-undeclared-selector
3243 Warn if a @code{@@selector(@dots{})} expression referring to an
3244 undeclared selector is found. A selector is considered undeclared if no
3245 method with that name has been declared before the
3246 @code{@@selector(@dots{})} expression, either explicitly in an
3247 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3248 an @code{@@implementation} section. This option always performs its
3249 checks as soon as a @code{@@selector(@dots{})} expression is found,
3250 while @option{-Wselector} only performs its checks in the final stage of
3251 compilation. This also enforces the coding style convention
3252 that methods and selectors must be declared before being used.
3254 @item -print-objc-runtime-info
3255 @opindex print-objc-runtime-info
3256 Generate C header describing the largest structure that is passed by
3261 @node Diagnostic Message Formatting Options
3262 @section Options to Control Diagnostic Messages Formatting
3263 @cindex options to control diagnostics formatting
3264 @cindex diagnostic messages
3265 @cindex message formatting
3267 Traditionally, diagnostic messages have been formatted irrespective of
3268 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3269 options described below
3270 to control the formatting algorithm for diagnostic messages,
3271 e.g.@: how many characters per line, how often source location
3272 information should be reported. Note that some language front ends may not
3273 honor these options.
3276 @item -fmessage-length=@var{n}
3277 @opindex fmessage-length
3278 Try to format error messages so that they fit on lines of about
3279 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3280 done; each error message appears on a single line. This is the
3281 default for all front ends.
3283 @item -fdiagnostics-show-location=once
3284 @opindex fdiagnostics-show-location
3285 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3286 reporter to emit source location information @emph{once}; that is, in
3287 case the message is too long to fit on a single physical line and has to
3288 be wrapped, the source location won't be emitted (as prefix) again,
3289 over and over, in subsequent continuation lines. This is the default
3292 @item -fdiagnostics-show-location=every-line
3293 Only meaningful in line-wrapping mode. Instructs the diagnostic
3294 messages reporter to emit the same source location information (as
3295 prefix) for physical lines that result from the process of breaking
3296 a message which is too long to fit on a single line.
3298 @item -fdiagnostics-color[=@var{WHEN}]
3299 @itemx -fno-diagnostics-color
3300 @opindex fdiagnostics-color
3301 @cindex highlight, color, colour
3302 @vindex GCC_COLORS @r{environment variable}
3303 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3304 or @samp{auto}. The default depends on how the compiler has been configured,
3305 it can be any of the above @var{WHEN} options or also @samp{never}
3306 if @env{GCC_COLORS} environment variable isn't present in the environment,
3307 and @samp{auto} otherwise.
3308 @samp{auto} means to use color only when the standard error is a terminal.
3309 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3310 aliases for @option{-fdiagnostics-color=always} and
3311 @option{-fdiagnostics-color=never}, respectively.
3313 The colors are defined by the environment variable @env{GCC_COLORS}.
3314 Its value is a colon-separated list of capabilities and Select Graphic
3315 Rendition (SGR) substrings. SGR commands are interpreted by the
3316 terminal or terminal emulator. (See the section in the documentation
3317 of your text terminal for permitted values and their meanings as
3318 character attributes.) These substring values are integers in decimal
3319 representation and can be concatenated with semicolons.
3320 Common values to concatenate include
3322 @samp{4} for underline,
3324 @samp{7} for inverse,
3325 @samp{39} for default foreground color,
3326 @samp{30} to @samp{37} for foreground colors,
3327 @samp{90} to @samp{97} for 16-color mode foreground colors,
3328 @samp{38;5;0} to @samp{38;5;255}
3329 for 88-color and 256-color modes foreground colors,
3330 @samp{49} for default background color,
3331 @samp{40} to @samp{47} for background colors,
3332 @samp{100} to @samp{107} for 16-color mode background colors,
3333 and @samp{48;5;0} to @samp{48;5;255}
3334 for 88-color and 256-color modes background colors.
3336 The default @env{GCC_COLORS} is
3338 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3341 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3342 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3343 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3344 string disables colors.
3345 Supported capabilities are as follows.
3349 @vindex error GCC_COLORS @r{capability}
3350 SGR substring for error: markers.
3353 @vindex warning GCC_COLORS @r{capability}
3354 SGR substring for warning: markers.
3357 @vindex note GCC_COLORS @r{capability}
3358 SGR substring for note: markers.
3361 @vindex caret GCC_COLORS @r{capability}
3362 SGR substring for caret line.
3365 @vindex locus GCC_COLORS @r{capability}
3366 SGR substring for location information, @samp{file:line} or
3367 @samp{file:line:column} etc.
3370 @vindex quote GCC_COLORS @r{capability}
3371 SGR substring for information printed within quotes.
3374 @item -fno-diagnostics-show-option
3375 @opindex fno-diagnostics-show-option
3376 @opindex fdiagnostics-show-option
3377 By default, each diagnostic emitted includes text indicating the
3378 command-line option that directly controls the diagnostic (if such an
3379 option is known to the diagnostic machinery). Specifying the
3380 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3382 @item -fno-diagnostics-show-caret
3383 @opindex fno-diagnostics-show-caret
3384 @opindex fdiagnostics-show-caret
3385 By default, each diagnostic emitted includes the original source line
3386 and a caret '^' indicating the column. This option suppresses this
3387 information. The source line is truncated to @var{n} characters, if
3388 the @option{-fmessage-length=n} option is given. When the output is done
3389 to the terminal, the width is limited to the width given by the
3390 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3394 @node Warning Options
3395 @section Options to Request or Suppress Warnings
3396 @cindex options to control warnings
3397 @cindex warning messages
3398 @cindex messages, warning
3399 @cindex suppressing warnings
3401 Warnings are diagnostic messages that report constructions that
3402 are not inherently erroneous but that are risky or suggest there
3403 may have been an error.
3405 The following language-independent options do not enable specific
3406 warnings but control the kinds of diagnostics produced by GCC@.
3409 @cindex syntax checking
3411 @opindex fsyntax-only
3412 Check the code for syntax errors, but don't do anything beyond that.
3414 @item -fmax-errors=@var{n}
3415 @opindex fmax-errors
3416 Limits the maximum number of error messages to @var{n}, at which point
3417 GCC bails out rather than attempting to continue processing the source
3418 code. If @var{n} is 0 (the default), there is no limit on the number
3419 of error messages produced. If @option{-Wfatal-errors} is also
3420 specified, then @option{-Wfatal-errors} takes precedence over this
3425 Inhibit all warning messages.
3430 Make all warnings into errors.
3435 Make the specified warning into an error. The specifier for a warning
3436 is appended; for example @option{-Werror=switch} turns the warnings
3437 controlled by @option{-Wswitch} into errors. This switch takes a
3438 negative form, to be used to negate @option{-Werror} for specific
3439 warnings; for example @option{-Wno-error=switch} makes
3440 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3443 The warning message for each controllable warning includes the
3444 option that controls the warning. That option can then be used with
3445 @option{-Werror=} and @option{-Wno-error=} as described above.
3446 (Printing of the option in the warning message can be disabled using the
3447 @option{-fno-diagnostics-show-option} flag.)
3449 Note that specifying @option{-Werror=}@var{foo} automatically implies
3450 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3453 @item -Wfatal-errors
3454 @opindex Wfatal-errors
3455 @opindex Wno-fatal-errors
3456 This option causes the compiler to abort compilation on the first error
3457 occurred rather than trying to keep going and printing further error
3462 You can request many specific warnings with options beginning with
3463 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3464 implicit declarations. Each of these specific warning options also
3465 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3466 example, @option{-Wno-implicit}. This manual lists only one of the
3467 two forms, whichever is not the default. For further
3468 language-specific options also refer to @ref{C++ Dialect Options} and
3469 @ref{Objective-C and Objective-C++ Dialect Options}.
3471 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3472 options, such as @option{-Wunused}, which may turn on further options,
3473 such as @option{-Wunused-value}. The combined effect of positive and
3474 negative forms is that more specific options have priority over less
3475 specific ones, independently of their position in the command-line. For
3476 options of the same specificity, the last one takes effect. Options
3477 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3478 as if they appeared at the end of the command-line.
3480 When an unrecognized warning option is requested (e.g.,
3481 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3482 that the option is not recognized. However, if the @option{-Wno-} form
3483 is used, the behavior is slightly different: no diagnostic is
3484 produced for @option{-Wno-unknown-warning} unless other diagnostics
3485 are being produced. This allows the use of new @option{-Wno-} options
3486 with old compilers, but if something goes wrong, the compiler
3487 warns that an unrecognized option is present.
3494 Issue all the warnings demanded by strict ISO C and ISO C++;
3495 reject all programs that use forbidden extensions, and some other
3496 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3497 version of the ISO C standard specified by any @option{-std} option used.
3499 Valid ISO C and ISO C++ programs should compile properly with or without
3500 this option (though a rare few require @option{-ansi} or a
3501 @option{-std} option specifying the required version of ISO C)@. However,
3502 without this option, certain GNU extensions and traditional C and C++
3503 features are supported as well. With this option, they are rejected.
3505 @option{-Wpedantic} does not cause warning messages for use of the
3506 alternate keywords whose names begin and end with @samp{__}. Pedantic
3507 warnings are also disabled in the expression that follows
3508 @code{__extension__}. However, only system header files should use
3509 these escape routes; application programs should avoid them.
3510 @xref{Alternate Keywords}.
3512 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3513 C conformance. They soon find that it does not do quite what they want:
3514 it finds some non-ISO practices, but not all---only those for which
3515 ISO C @emph{requires} a diagnostic, and some others for which
3516 diagnostics have been added.
3518 A feature to report any failure to conform to ISO C might be useful in
3519 some instances, but would require considerable additional work and would
3520 be quite different from @option{-Wpedantic}. We don't have plans to
3521 support such a feature in the near future.
3523 Where the standard specified with @option{-std} represents a GNU
3524 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3525 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3526 extended dialect is based. Warnings from @option{-Wpedantic} are given
3527 where they are required by the base standard. (It does not make sense
3528 for such warnings to be given only for features not in the specified GNU
3529 C dialect, since by definition the GNU dialects of C include all
3530 features the compiler supports with the given option, and there would be
3531 nothing to warn about.)
3533 @item -pedantic-errors
3534 @opindex pedantic-errors
3535 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3536 requires a diagnostic, in some cases where there is undefined behavior
3537 at compile-time and in some other cases that do not prevent compilation
3538 of programs that are valid according to the standard. This is not
3539 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3540 by this option and not enabled by the latter and vice versa.
3545 This enables all the warnings about constructions that some users
3546 consider questionable, and that are easy to avoid (or modify to
3547 prevent the warning), even in conjunction with macros. This also
3548 enables some language-specific warnings described in @ref{C++ Dialect
3549 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3551 @option{-Wall} turns on the following warning flags:
3553 @gccoptlist{-Waddress @gol
3554 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3556 -Wc++11-compat -Wc++14-compat@gol
3557 -Wchar-subscripts @gol
3559 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3561 -Wimplicit @r{(C and Objective-C only)} @gol
3562 -Wimplicit-int @r{(C and Objective-C only)} @gol
3563 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3564 -Winit-self @r{(only for C++)} @gol
3565 -Wlogical-not-parentheses
3566 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3567 -Wmaybe-uninitialized @gol
3568 -Wmemset-transposed-args @gol
3569 -Wmisleading-indentation @r{(only for C/C++)} @gol
3570 -Wmissing-braces @r{(only for C/ObjC)} @gol
3571 -Wnarrowing @r{(only for C++)} @gol
3578 -Wsequence-point @gol
3579 -Wsign-compare @r{(only in C++)} @gol
3580 -Wsizeof-pointer-memaccess @gol
3581 -Wstrict-aliasing @gol
3582 -Wstrict-overflow=1 @gol
3584 -Wtautological-compare @gol
3586 -Wuninitialized @gol
3587 -Wunknown-pragmas @gol
3588 -Wunused-function @gol
3591 -Wunused-variable @gol
3592 -Wvolatile-register-var @gol
3595 Note that some warning flags are not implied by @option{-Wall}. Some of
3596 them warn about constructions that users generally do not consider
3597 questionable, but which occasionally you might wish to check for;
3598 others warn about constructions that are necessary or hard to avoid in
3599 some cases, and there is no simple way to modify the code to suppress
3600 the warning. Some of them are enabled by @option{-Wextra} but many of
3601 them must be enabled individually.
3607 This enables some extra warning flags that are not enabled by
3608 @option{-Wall}. (This option used to be called @option{-W}. The older
3609 name is still supported, but the newer name is more descriptive.)
3611 @gccoptlist{-Wclobbered @gol
3613 -Wignored-qualifiers @gol
3614 -Wmissing-field-initializers @gol
3615 -Wmissing-parameter-type @r{(C only)} @gol
3616 -Wold-style-declaration @r{(C only)} @gol
3617 -Woverride-init @gol
3618 -Wsign-compare @r{(C only)} @gol
3620 -Wuninitialized @gol
3621 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3622 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3623 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3626 The option @option{-Wextra} also prints warning messages for the
3632 A pointer is compared against integer zero with @code{<}, @code{<=},
3633 @code{>}, or @code{>=}.
3636 (C++ only) An enumerator and a non-enumerator both appear in a
3637 conditional expression.
3640 (C++ only) Ambiguous virtual bases.
3643 (C++ only) Subscripting an array that has been declared @code{register}.
3646 (C++ only) Taking the address of a variable that has been declared
3650 (C++ only) A base class is not initialized in a derived class's copy
3655 @item -Wchar-subscripts
3656 @opindex Wchar-subscripts
3657 @opindex Wno-char-subscripts
3658 Warn if an array subscript has type @code{char}. This is a common cause
3659 of error, as programmers often forget that this type is signed on some
3661 This warning is enabled by @option{-Wall}.
3665 @opindex Wno-comment
3666 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3667 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3668 This warning is enabled by @option{-Wall}.
3670 @item -Wno-coverage-mismatch
3671 @opindex Wno-coverage-mismatch
3672 Warn if feedback profiles do not match when using the
3673 @option{-fprofile-use} option.
3674 If a source file is changed between compiling with @option{-fprofile-gen} and
3675 with @option{-fprofile-use}, the files with the profile feedback can fail
3676 to match the source file and GCC cannot use the profile feedback
3677 information. By default, this warning is enabled and is treated as an
3678 error. @option{-Wno-coverage-mismatch} can be used to disable the
3679 warning or @option{-Wno-error=coverage-mismatch} can be used to
3680 disable the error. Disabling the error for this warning can result in
3681 poorly optimized code and is useful only in the
3682 case of very minor changes such as bug fixes to an existing code-base.
3683 Completely disabling the warning is not recommended.
3686 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3688 Suppress warning messages emitted by @code{#warning} directives.
3690 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3691 @opindex Wdouble-promotion
3692 @opindex Wno-double-promotion
3693 Give a warning when a value of type @code{float} is implicitly
3694 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3695 floating-point unit implement @code{float} in hardware, but emulate
3696 @code{double} in software. On such a machine, doing computations
3697 using @code{double} values is much more expensive because of the
3698 overhead required for software emulation.
3700 It is easy to accidentally do computations with @code{double} because
3701 floating-point literals are implicitly of type @code{double}. For
3705 float area(float radius)
3707 return 3.14159 * radius * radius;
3711 the compiler performs the entire computation with @code{double}
3712 because the floating-point literal is a @code{double}.
3715 @itemx -Wformat=@var{n}
3718 @opindex ffreestanding
3719 @opindex fno-builtin
3721 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3722 the arguments supplied have types appropriate to the format string
3723 specified, and that the conversions specified in the format string make
3724 sense. This includes standard functions, and others specified by format
3725 attributes (@pxref{Function Attributes}), in the @code{printf},
3726 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3727 not in the C standard) families (or other target-specific families).
3728 Which functions are checked without format attributes having been
3729 specified depends on the standard version selected, and such checks of
3730 functions without the attribute specified are disabled by
3731 @option{-ffreestanding} or @option{-fno-builtin}.
3733 The formats are checked against the format features supported by GNU
3734 libc version 2.2. These include all ISO C90 and C99 features, as well
3735 as features from the Single Unix Specification and some BSD and GNU
3736 extensions. Other library implementations may not support all these
3737 features; GCC does not support warning about features that go beyond a
3738 particular library's limitations. However, if @option{-Wpedantic} is used
3739 with @option{-Wformat}, warnings are given about format features not
3740 in the selected standard version (but not for @code{strfmon} formats,
3741 since those are not in any version of the C standard). @xref{C Dialect
3742 Options,,Options Controlling C Dialect}.
3749 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3750 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3751 @option{-Wformat} also checks for null format arguments for several
3752 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3753 aspects of this level of format checking can be disabled by the
3754 options: @option{-Wno-format-contains-nul},
3755 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3756 @option{-Wformat} is enabled by @option{-Wall}.
3758 @item -Wno-format-contains-nul
3759 @opindex Wno-format-contains-nul
3760 @opindex Wformat-contains-nul
3761 If @option{-Wformat} is specified, do not warn about format strings that
3764 @item -Wno-format-extra-args
3765 @opindex Wno-format-extra-args
3766 @opindex Wformat-extra-args
3767 If @option{-Wformat} is specified, do not warn about excess arguments to a
3768 @code{printf} or @code{scanf} format function. The C standard specifies
3769 that such arguments are ignored.
3771 Where the unused arguments lie between used arguments that are
3772 specified with @samp{$} operand number specifications, normally
3773 warnings are still given, since the implementation could not know what
3774 type to pass to @code{va_arg} to skip the unused arguments. However,
3775 in the case of @code{scanf} formats, this option suppresses the
3776 warning if the unused arguments are all pointers, since the Single
3777 Unix Specification says that such unused arguments are allowed.
3779 @item -Wno-format-zero-length
3780 @opindex Wno-format-zero-length
3781 @opindex Wformat-zero-length
3782 If @option{-Wformat} is specified, do not warn about zero-length formats.
3783 The C standard specifies that zero-length formats are allowed.
3788 Enable @option{-Wformat} plus additional format checks. Currently
3789 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3792 @item -Wformat-nonliteral
3793 @opindex Wformat-nonliteral
3794 @opindex Wno-format-nonliteral
3795 If @option{-Wformat} is specified, also warn if the format string is not a
3796 string literal and so cannot be checked, unless the format function
3797 takes its format arguments as a @code{va_list}.
3799 @item -Wformat-security
3800 @opindex Wformat-security
3801 @opindex Wno-format-security
3802 If @option{-Wformat} is specified, also warn about uses of format
3803 functions that represent possible security problems. At present, this
3804 warns about calls to @code{printf} and @code{scanf} functions where the
3805 format string is not a string literal and there are no format arguments,
3806 as in @code{printf (foo);}. This may be a security hole if the format
3807 string came from untrusted input and contains @samp{%n}. (This is
3808 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3809 in future warnings may be added to @option{-Wformat-security} that are not
3810 included in @option{-Wformat-nonliteral}.)
3812 @item -Wformat-signedness
3813 @opindex Wformat-signedness
3814 @opindex Wno-format-signedness
3815 If @option{-Wformat} is specified, also warn if the format string
3816 requires an unsigned argument and the argument is signed and vice versa.
3819 @opindex Wformat-y2k
3820 @opindex Wno-format-y2k
3821 If @option{-Wformat} is specified, also warn about @code{strftime}
3822 formats that may yield only a two-digit year.
3827 @opindex Wno-nonnull
3828 Warn about passing a null pointer for arguments marked as
3829 requiring a non-null value by the @code{nonnull} function attribute.
3831 Also warns when comparing an argument marked with the @code{nonnull}
3832 function attribute against null inside the function.
3834 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3835 can be disabled with the @option{-Wno-nonnull} option.
3837 @item -Wnull-dereference
3838 @opindex Wnull-dereference
3839 @opindex Wno-null-dereference
3840 Warn if the compiler detects paths that trigger erroneous or
3841 undefined behavior due to dereferencing a null pointer. This option
3842 is only active when @option{-fdelete-null-pointer-checks} is active,
3843 which is enabled by optimizations in most targets. The precision of
3844 the warnings depends on the optimization options used.
3846 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3848 @opindex Wno-init-self
3849 Warn about uninitialized variables that are initialized with themselves.
3850 Note this option can only be used with the @option{-Wuninitialized} option.
3852 For example, GCC warns about @code{i} being uninitialized in the
3853 following snippet only when @option{-Winit-self} has been specified:
3864 This warning is enabled by @option{-Wall} in C++.
3866 @item -Wimplicit-int @r{(C and Objective-C only)}
3867 @opindex Wimplicit-int
3868 @opindex Wno-implicit-int
3869 Warn when a declaration does not specify a type.
3870 This warning is enabled by @option{-Wall}.
3872 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3873 @opindex Wimplicit-function-declaration
3874 @opindex Wno-implicit-function-declaration
3875 Give a warning whenever a function is used before being declared. In
3876 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3877 enabled by default and it is made into an error by
3878 @option{-pedantic-errors}. This warning is also enabled by
3881 @item -Wimplicit @r{(C and Objective-C only)}
3883 @opindex Wno-implicit
3884 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3885 This warning is enabled by @option{-Wall}.
3887 @item -Wignored-qualifiers @r{(C and C++ only)}
3888 @opindex Wignored-qualifiers
3889 @opindex Wno-ignored-qualifiers
3890 Warn if the return type of a function has a type qualifier
3891 such as @code{const}. For ISO C such a type qualifier has no effect,
3892 since the value returned by a function is not an lvalue.
3893 For C++, the warning is only emitted for scalar types or @code{void}.
3894 ISO C prohibits qualified @code{void} return types on function
3895 definitions, so such return types always receive a warning
3896 even without this option.
3898 This warning is also enabled by @option{-Wextra}.
3903 Warn if the type of @code{main} is suspicious. @code{main} should be
3904 a function with external linkage, returning int, taking either zero
3905 arguments, two, or three arguments of appropriate types. This warning
3906 is enabled by default in C++ and is enabled by either @option{-Wall}
3907 or @option{-Wpedantic}.
3909 @item -Wmisleading-indentation @r{(C and C++ only)}
3910 @opindex Wmisleading-indentation
3911 @opindex Wno-misleading-indentation
3912 Warn when the indentation of the code does not reflect the block structure.
3913 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3914 @code{for} clauses with a guarded statement that does not use braces,
3915 followed by an unguarded statement with the same indentation.
3917 In the following example, the call to ``bar'' is misleadingly indented as
3918 if it were guarded by the ``if'' conditional.
3921 if (some_condition ())
3923 bar (); /* Gotcha: this is not guarded by the "if". */
3926 In the case of mixed tabs and spaces, the warning uses the
3927 @option{-ftabstop=} option to determine if the statements line up
3930 The warning is not issued for code involving multiline preprocessor logic
3931 such as the following example.
3936 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3942 The warning is not issued after a @code{#line} directive, since this
3943 typically indicates autogenerated code, and no assumptions can be made
3944 about the layout of the file that the directive references.
3946 This warning is enabled by @option{-Wall} in C and C++.
3948 @item -Wmissing-braces
3949 @opindex Wmissing-braces
3950 @opindex Wno-missing-braces
3951 Warn if an aggregate or union initializer is not fully bracketed. In
3952 the following example, the initializer for @code{a} is not fully
3953 bracketed, but that for @code{b} is fully bracketed. This warning is
3954 enabled by @option{-Wall} in C.
3957 int a[2][2] = @{ 0, 1, 2, 3 @};
3958 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3961 This warning is enabled by @option{-Wall}.
3963 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3964 @opindex Wmissing-include-dirs
3965 @opindex Wno-missing-include-dirs
3966 Warn if a user-supplied include directory does not exist.
3969 @opindex Wparentheses
3970 @opindex Wno-parentheses
3971 Warn if parentheses are omitted in certain contexts, such
3972 as when there is an assignment in a context where a truth value
3973 is expected, or when operators are nested whose precedence people
3974 often get confused about.
3976 Also warn if a comparison like @code{x<=y<=z} appears; this is
3977 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3978 interpretation from that of ordinary mathematical notation.
3980 Also warn about constructions where there may be confusion to which
3981 @code{if} statement an @code{else} branch belongs. Here is an example of
3996 In C/C++, every @code{else} branch belongs to the innermost possible
3997 @code{if} statement, which in this example is @code{if (b)}. This is
3998 often not what the programmer expected, as illustrated in the above
3999 example by indentation the programmer chose. When there is the
4000 potential for this confusion, GCC issues a warning when this flag
4001 is specified. To eliminate the warning, add explicit braces around
4002 the innermost @code{if} statement so there is no way the @code{else}
4003 can belong to the enclosing @code{if}. The resulting code
4020 Also warn for dangerous uses of the GNU extension to
4021 @code{?:} with omitted middle operand. When the condition
4022 in the @code{?}: operator is a boolean expression, the omitted value is
4023 always 1. Often programmers expect it to be a value computed
4024 inside the conditional expression instead.
4026 This warning is enabled by @option{-Wall}.
4028 @item -Wsequence-point
4029 @opindex Wsequence-point
4030 @opindex Wno-sequence-point
4031 Warn about code that may have undefined semantics because of violations
4032 of sequence point rules in the C and C++ standards.
4034 The C and C++ standards define the order in which expressions in a C/C++
4035 program are evaluated in terms of @dfn{sequence points}, which represent
4036 a partial ordering between the execution of parts of the program: those
4037 executed before the sequence point, and those executed after it. These
4038 occur after the evaluation of a full expression (one which is not part
4039 of a larger expression), after the evaluation of the first operand of a
4040 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4041 function is called (but after the evaluation of its arguments and the
4042 expression denoting the called function), and in certain other places.
4043 Other than as expressed by the sequence point rules, the order of
4044 evaluation of subexpressions of an expression is not specified. All
4045 these rules describe only a partial order rather than a total order,
4046 since, for example, if two functions are called within one expression
4047 with no sequence point between them, the order in which the functions
4048 are called is not specified. However, the standards committee have
4049 ruled that function calls do not overlap.
4051 It is not specified when between sequence points modifications to the
4052 values of objects take effect. Programs whose behavior depends on this
4053 have undefined behavior; the C and C++ standards specify that ``Between
4054 the previous and next sequence point an object shall have its stored
4055 value modified at most once by the evaluation of an expression.
4056 Furthermore, the prior value shall be read only to determine the value
4057 to be stored.''. If a program breaks these rules, the results on any
4058 particular implementation are entirely unpredictable.
4060 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4061 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4062 diagnosed by this option, and it may give an occasional false positive
4063 result, but in general it has been found fairly effective at detecting
4064 this sort of problem in programs.
4066 The standard is worded confusingly, therefore there is some debate
4067 over the precise meaning of the sequence point rules in subtle cases.
4068 Links to discussions of the problem, including proposed formal
4069 definitions, may be found on the GCC readings page, at
4070 @uref{http://gcc.gnu.org/@/readings.html}.
4072 This warning is enabled by @option{-Wall} for C and C++.
4074 @item -Wno-return-local-addr
4075 @opindex Wno-return-local-addr
4076 @opindex Wreturn-local-addr
4077 Do not warn about returning a pointer (or in C++, a reference) to a
4078 variable that goes out of scope after the function returns.
4081 @opindex Wreturn-type
4082 @opindex Wno-return-type
4083 Warn whenever a function is defined with a return type that defaults
4084 to @code{int}. Also warn about any @code{return} statement with no
4085 return value in a function whose return type is not @code{void}
4086 (falling off the end of the function body is considered returning
4087 without a value), and about a @code{return} statement with an
4088 expression in a function whose return type is @code{void}.
4090 For C++, a function without return type always produces a diagnostic
4091 message, even when @option{-Wno-return-type} is specified. The only
4092 exceptions are @code{main} and functions defined in system headers.
4094 This warning is enabled by @option{-Wall}.
4096 @item -Wshift-count-negative
4097 @opindex Wshift-count-negative
4098 @opindex Wno-shift-count-negative
4099 Warn if shift count is negative. This warning is enabled by default.
4101 @item -Wshift-count-overflow
4102 @opindex Wshift-count-overflow
4103 @opindex Wno-shift-count-overflow
4104 Warn if shift count >= width of type. This warning is enabled by default.
4106 @item -Wshift-negative-value
4107 @opindex Wshift-negative-value
4108 @opindex Wno-shift-negative-value
4109 Warn if left shifting a negative value. This warning is enabled by
4110 @option{-Wextra} in C99 and C++11 modes (and newer).
4112 @item -Wshift-overflow
4113 @itemx -Wshift-overflow=@var{n}
4114 @opindex Wshift-overflow
4115 @opindex Wno-shift-overflow
4116 Warn about left shift overflows. This warning is enabled by
4117 default in C99 and C++11 modes (and newer).
4120 @item -Wshift-overflow=1
4121 This is the warning level of @option{-Wshift-overflow} and is enabled
4122 by default in C99 and C++11 modes (and newer). This warning level does
4123 not warn about left-shifting 1 into the sign bit. (However, in C, such
4124 an overflow is still rejected in contexts where an integer constant expression
4127 @item -Wshift-overflow=2
4128 This warning level also warns about left-shifting 1 into the sign bit,
4129 unless C++14 mode is active.
4135 Warn whenever a @code{switch} statement has an index of enumerated type
4136 and lacks a @code{case} for one or more of the named codes of that
4137 enumeration. (The presence of a @code{default} label prevents this
4138 warning.) @code{case} labels outside the enumeration range also
4139 provoke warnings when this option is used (even if there is a
4140 @code{default} label).
4141 This warning is enabled by @option{-Wall}.
4143 @item -Wswitch-default
4144 @opindex Wswitch-default
4145 @opindex Wno-switch-default
4146 Warn whenever a @code{switch} statement does not have a @code{default}
4150 @opindex Wswitch-enum
4151 @opindex Wno-switch-enum
4152 Warn whenever a @code{switch} statement has an index of enumerated type
4153 and lacks a @code{case} for one or more of the named codes of that
4154 enumeration. @code{case} labels outside the enumeration range also
4155 provoke warnings when this option is used. The only difference
4156 between @option{-Wswitch} and this option is that this option gives a
4157 warning about an omitted enumeration code even if there is a
4158 @code{default} label.
4161 @opindex Wswitch-bool
4162 @opindex Wno-switch-bool
4163 Warn whenever a @code{switch} statement has an index of boolean type
4164 and the case values are outside the range of a boolean type.
4165 It is possible to suppress this warning by casting the controlling
4166 expression to a type other than @code{bool}. For example:
4169 switch ((int) (a == 4))
4175 This warning is enabled by default for C and C++ programs.
4177 @item -Wsync-nand @r{(C and C++ only)}
4179 @opindex Wno-sync-nand
4180 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4181 built-in functions are used. These functions changed semantics in GCC 4.4.
4185 @opindex Wno-trigraphs
4186 Warn if any trigraphs are encountered that might change the meaning of
4187 the program (trigraphs within comments are not warned about).
4188 This warning is enabled by @option{-Wall}.
4190 @item -Wunused-but-set-parameter
4191 @opindex Wunused-but-set-parameter
4192 @opindex Wno-unused-but-set-parameter
4193 Warn whenever a function parameter is assigned to, but otherwise unused
4194 (aside from its declaration).
4196 To suppress this warning use the @code{unused} attribute
4197 (@pxref{Variable Attributes}).
4199 This warning is also enabled by @option{-Wunused} together with
4202 @item -Wunused-but-set-variable
4203 @opindex Wunused-but-set-variable
4204 @opindex Wno-unused-but-set-variable
4205 Warn whenever a local variable is assigned to, but otherwise unused
4206 (aside from its declaration).
4207 This warning is enabled by @option{-Wall}.
4209 To suppress this warning use the @code{unused} attribute
4210 (@pxref{Variable Attributes}).
4212 This warning is also enabled by @option{-Wunused}, which is enabled
4215 @item -Wunused-function
4216 @opindex Wunused-function
4217 @opindex Wno-unused-function
4218 Warn whenever a static function is declared but not defined or a
4219 non-inline static function is unused.
4220 This warning is enabled by @option{-Wall}.
4222 @item -Wunused-label
4223 @opindex Wunused-label
4224 @opindex Wno-unused-label
4225 Warn whenever a label is declared but not used.
4226 This warning is enabled by @option{-Wall}.
4228 To suppress this warning use the @code{unused} attribute
4229 (@pxref{Variable Attributes}).
4231 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4232 @opindex Wunused-local-typedefs
4233 Warn when a typedef locally defined in a function is not used.
4234 This warning is enabled by @option{-Wall}.
4236 @item -Wunused-parameter
4237 @opindex Wunused-parameter
4238 @opindex Wno-unused-parameter
4239 Warn whenever a function parameter is unused aside from its declaration.
4241 To suppress this warning use the @code{unused} attribute
4242 (@pxref{Variable Attributes}).
4244 @item -Wno-unused-result
4245 @opindex Wunused-result
4246 @opindex Wno-unused-result
4247 Do not warn if a caller of a function marked with attribute
4248 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4249 its return value. The default is @option{-Wunused-result}.
4251 @item -Wunused-variable
4252 @opindex Wunused-variable
4253 @opindex Wno-unused-variable
4254 Warn whenever a local or static variable is unused aside from its
4255 declaration. This option implies @option{-Wunused-const-variable} for C,
4256 but not for C++. This warning is enabled by @option{-Wall}.
4258 To suppress this warning use the @code{unused} attribute
4259 (@pxref{Variable Attributes}).
4261 @item -Wunused-const-variable
4262 @opindex Wunused-const-variable
4263 @opindex Wno-unused-const-variable
4264 Warn whenever a constant static variable is unused aside from its declaration.
4265 This warning is enabled by @option{-Wunused-variable} for C, but not for C++.
4266 In C++ this is normally not an error since const variables take the place of
4267 @code{#define}s in C++.
4269 To suppress this warning use the @code{unused} attribute
4270 (@pxref{Variable Attributes}).
4272 @item -Wunused-value
4273 @opindex Wunused-value
4274 @opindex Wno-unused-value
4275 Warn whenever a statement computes a result that is explicitly not
4276 used. To suppress this warning cast the unused expression to
4277 @code{void}. This includes an expression-statement or the left-hand
4278 side of a comma expression that contains no side effects. For example,
4279 an expression such as @code{x[i,j]} causes a warning, while
4280 @code{x[(void)i,j]} does not.
4282 This warning is enabled by @option{-Wall}.
4287 All the above @option{-Wunused} options combined.
4289 In order to get a warning about an unused function parameter, you must
4290 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4291 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4293 @item -Wuninitialized
4294 @opindex Wuninitialized
4295 @opindex Wno-uninitialized
4296 Warn if an automatic variable is used without first being initialized
4297 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4298 warn if a non-static reference or non-static @code{const} member
4299 appears in a class without constructors.
4301 If you want to warn about code that uses the uninitialized value of the
4302 variable in its own initializer, use the @option{-Winit-self} option.
4304 These warnings occur for individual uninitialized or clobbered
4305 elements of structure, union or array variables as well as for
4306 variables that are uninitialized or clobbered as a whole. They do
4307 not occur for variables or elements declared @code{volatile}. Because
4308 these warnings depend on optimization, the exact variables or elements
4309 for which there are warnings depends on the precise optimization
4310 options and version of GCC used.
4312 Note that there may be no warning about a variable that is used only
4313 to compute a value that itself is never used, because such
4314 computations may be deleted by data flow analysis before the warnings
4317 @item -Winvalid-memory-model
4318 @opindex Winvalid-memory-model
4319 @opindex Wno-invalid-memory-model
4320 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4321 and the C11 atomic generic functions with a memory consistency argument
4322 that is either invalid for the operation or outside the range of values
4323 of the @code{memory_order} enumeration. For example, since the
4324 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4325 defined for the relaxed, release, and sequentially consistent memory
4326 orders the following code is diagnosed:
4331 __atomic_store_n (i, 0, memory_order_consume);
4335 @option{-Winvalid-memory-model} is enabled by default.
4337 @item -Wmaybe-uninitialized
4338 @opindex Wmaybe-uninitialized
4339 @opindex Wno-maybe-uninitialized
4340 For an automatic variable, if there exists a path from the function
4341 entry to a use of the variable that is initialized, but there exist
4342 some other paths for which the variable is not initialized, the compiler
4343 emits a warning if it cannot prove the uninitialized paths are not
4344 executed at run time. These warnings are made optional because GCC is
4345 not smart enough to see all the reasons why the code might be correct
4346 in spite of appearing to have an error. Here is one example of how
4367 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4368 always initialized, but GCC doesn't know this. To suppress the
4369 warning, you need to provide a default case with assert(0) or
4372 @cindex @code{longjmp} warnings
4373 This option also warns when a non-volatile automatic variable might be
4374 changed by a call to @code{longjmp}. These warnings as well are possible
4375 only in optimizing compilation.
4377 The compiler sees only the calls to @code{setjmp}. It cannot know
4378 where @code{longjmp} will be called; in fact, a signal handler could
4379 call it at any point in the code. As a result, you may get a warning
4380 even when there is in fact no problem because @code{longjmp} cannot
4381 in fact be called at the place that would cause a problem.
4383 Some spurious warnings can be avoided if you declare all the functions
4384 you use that never return as @code{noreturn}. @xref{Function
4387 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4389 @item -Wunknown-pragmas
4390 @opindex Wunknown-pragmas
4391 @opindex Wno-unknown-pragmas
4392 @cindex warning for unknown pragmas
4393 @cindex unknown pragmas, warning
4394 @cindex pragmas, warning of unknown
4395 Warn when a @code{#pragma} directive is encountered that is not understood by
4396 GCC@. If this command-line option is used, warnings are even issued
4397 for unknown pragmas in system header files. This is not the case if
4398 the warnings are only enabled by the @option{-Wall} command-line option.
4401 @opindex Wno-pragmas
4403 Do not warn about misuses of pragmas, such as incorrect parameters,
4404 invalid syntax, or conflicts between pragmas. See also
4405 @option{-Wunknown-pragmas}.
4407 @item -Wstrict-aliasing
4408 @opindex Wstrict-aliasing
4409 @opindex Wno-strict-aliasing
4410 This option is only active when @option{-fstrict-aliasing} is active.
4411 It warns about code that might break the strict aliasing rules that the
4412 compiler is using for optimization. The warning does not catch all
4413 cases, but does attempt to catch the more common pitfalls. It is
4414 included in @option{-Wall}.
4415 It is equivalent to @option{-Wstrict-aliasing=3}
4417 @item -Wstrict-aliasing=n
4418 @opindex Wstrict-aliasing=n
4419 This option is only active when @option{-fstrict-aliasing} is active.
4420 It warns about code that might break the strict aliasing rules that the
4421 compiler is using for optimization.
4422 Higher levels correspond to higher accuracy (fewer false positives).
4423 Higher levels also correspond to more effort, similar to the way @option{-O}
4425 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4427 Level 1: Most aggressive, quick, least accurate.
4428 Possibly useful when higher levels
4429 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4430 false negatives. However, it has many false positives.
4431 Warns for all pointer conversions between possibly incompatible types,
4432 even if never dereferenced. Runs in the front end only.
4434 Level 2: Aggressive, quick, not too precise.
4435 May still have many false positives (not as many as level 1 though),
4436 and few false negatives (but possibly more than level 1).
4437 Unlike level 1, it only warns when an address is taken. Warns about
4438 incomplete types. Runs in the front end only.
4440 Level 3 (default for @option{-Wstrict-aliasing}):
4441 Should have very few false positives and few false
4442 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4443 Takes care of the common pun+dereference pattern in the front end:
4444 @code{*(int*)&some_float}.
4445 If optimization is enabled, it also runs in the back end, where it deals
4446 with multiple statement cases using flow-sensitive points-to information.
4447 Only warns when the converted pointer is dereferenced.
4448 Does not warn about incomplete types.
4450 @item -Wstrict-overflow
4451 @itemx -Wstrict-overflow=@var{n}
4452 @opindex Wstrict-overflow
4453 @opindex Wno-strict-overflow
4454 This option is only active when @option{-fstrict-overflow} is active.
4455 It warns about cases where the compiler optimizes based on the
4456 assumption that signed overflow does not occur. Note that it does not
4457 warn about all cases where the code might overflow: it only warns
4458 about cases where the compiler implements some optimization. Thus
4459 this warning depends on the optimization level.
4461 An optimization that assumes that signed overflow does not occur is
4462 perfectly safe if the values of the variables involved are such that
4463 overflow never does, in fact, occur. Therefore this warning can
4464 easily give a false positive: a warning about code that is not
4465 actually a problem. To help focus on important issues, several
4466 warning levels are defined. No warnings are issued for the use of
4467 undefined signed overflow when estimating how many iterations a loop
4468 requires, in particular when determining whether a loop will be
4472 @item -Wstrict-overflow=1
4473 Warn about cases that are both questionable and easy to avoid. For
4474 example, with @option{-fstrict-overflow}, the compiler simplifies
4475 @code{x + 1 > x} to @code{1}. This level of
4476 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4477 are not, and must be explicitly requested.
4479 @item -Wstrict-overflow=2
4480 Also warn about other cases where a comparison is simplified to a
4481 constant. For example: @code{abs (x) >= 0}. This can only be
4482 simplified when @option{-fstrict-overflow} is in effect, because
4483 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4484 zero. @option{-Wstrict-overflow} (with no level) is the same as
4485 @option{-Wstrict-overflow=2}.
4487 @item -Wstrict-overflow=3
4488 Also warn about other cases where a comparison is simplified. For
4489 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4491 @item -Wstrict-overflow=4
4492 Also warn about other simplifications not covered by the above cases.
4493 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4495 @item -Wstrict-overflow=5
4496 Also warn about cases where the compiler reduces the magnitude of a
4497 constant involved in a comparison. For example: @code{x + 2 > y} is
4498 simplified to @code{x + 1 >= y}. This is reported only at the
4499 highest warning level because this simplification applies to many
4500 comparisons, so this warning level gives a very large number of
4504 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4505 @opindex Wsuggest-attribute=
4506 @opindex Wno-suggest-attribute=
4507 Warn for cases where adding an attribute may be beneficial. The
4508 attributes currently supported are listed below.
4511 @item -Wsuggest-attribute=pure
4512 @itemx -Wsuggest-attribute=const
4513 @itemx -Wsuggest-attribute=noreturn
4514 @opindex Wsuggest-attribute=pure
4515 @opindex Wno-suggest-attribute=pure
4516 @opindex Wsuggest-attribute=const
4517 @opindex Wno-suggest-attribute=const
4518 @opindex Wsuggest-attribute=noreturn
4519 @opindex Wno-suggest-attribute=noreturn
4521 Warn about functions that might be candidates for attributes
4522 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4523 functions visible in other compilation units or (in the case of @code{pure} and
4524 @code{const}) if it cannot prove that the function returns normally. A function
4525 returns normally if it doesn't contain an infinite loop or return abnormally
4526 by throwing, calling @code{abort} or trapping. This analysis requires option
4527 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4528 higher. Higher optimization levels improve the accuracy of the analysis.
4530 @item -Wsuggest-attribute=format
4531 @itemx -Wmissing-format-attribute
4532 @opindex Wsuggest-attribute=format
4533 @opindex Wmissing-format-attribute
4534 @opindex Wno-suggest-attribute=format
4535 @opindex Wno-missing-format-attribute
4539 Warn about function pointers that might be candidates for @code{format}
4540 attributes. Note these are only possible candidates, not absolute ones.
4541 GCC guesses that function pointers with @code{format} attributes that
4542 are used in assignment, initialization, parameter passing or return
4543 statements should have a corresponding @code{format} attribute in the
4544 resulting type. I.e.@: the left-hand side of the assignment or
4545 initialization, the type of the parameter variable, or the return type
4546 of the containing function respectively should also have a @code{format}
4547 attribute to avoid the warning.
4549 GCC also warns about function definitions that might be
4550 candidates for @code{format} attributes. Again, these are only
4551 possible candidates. GCC guesses that @code{format} attributes
4552 might be appropriate for any function that calls a function like
4553 @code{vprintf} or @code{vscanf}, but this might not always be the
4554 case, and some functions for which @code{format} attributes are
4555 appropriate may not be detected.
4558 @item -Wsuggest-final-types
4559 @opindex Wno-suggest-final-types
4560 @opindex Wsuggest-final-types
4561 Warn about types with virtual methods where code quality would be improved
4562 if the type were declared with the C++11 @code{final} specifier,
4564 declared in an anonymous namespace. This allows GCC to more aggressively
4565 devirtualize the polymorphic calls. This warning is more effective with link
4566 time optimization, where the information about the class hierarchy graph is
4569 @item -Wsuggest-final-methods
4570 @opindex Wno-suggest-final-methods
4571 @opindex Wsuggest-final-methods
4572 Warn about virtual methods where code quality would be improved if the method
4573 were declared with the C++11 @code{final} specifier,
4574 or, if possible, its type were
4575 declared in an anonymous namespace or with the @code{final} specifier.
4577 more effective with link time optimization, where the information about the
4578 class hierarchy graph is more complete. It is recommended to first consider
4579 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4582 @item -Wsuggest-override
4583 Warn about overriding virtual functions that are not marked with the override
4586 @item -Warray-bounds
4587 @itemx -Warray-bounds=@var{n}
4588 @opindex Wno-array-bounds
4589 @opindex Warray-bounds
4590 This option is only active when @option{-ftree-vrp} is active
4591 (default for @option{-O2} and above). It warns about subscripts to arrays
4592 that are always out of bounds. This warning is enabled by @option{-Wall}.
4595 @item -Warray-bounds=1
4596 This is the warning level of @option{-Warray-bounds} and is enabled
4597 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4599 @item -Warray-bounds=2
4600 This warning level also warns about out of bounds access for
4601 arrays at the end of a struct and for arrays accessed through
4602 pointers. This warning level may give a larger number of
4603 false positives and is deactivated by default.
4606 @item -Wbool-compare
4607 @opindex Wno-bool-compare
4608 @opindex Wbool-compare
4609 Warn about boolean expression compared with an integer value different from
4610 @code{true}/@code{false}. For instance, the following comparison is
4615 if ((n > 1) == 2) @{ @dots{} @}
4617 This warning is enabled by @option{-Wall}.
4619 @item -Wduplicated-cond
4620 @opindex Wno-duplicated-cond
4621 @opindex Wduplicated-cond
4622 Warn about duplicated conditions in an if-else-if chain. For instance,
4623 warn for the following code:
4625 if (p->q != NULL) @{ @dots{} @}
4626 else if (p->q != NULL) @{ @dots{} @}
4629 @item -Wframe-address
4630 @opindex Wno-frame-address
4631 @opindex Wframe-address
4632 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4633 is called with an argument greater than 0. Such calls may return indeterminate
4634 values or crash the program. The warning is included in @option{-Wall}.
4636 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4637 @opindex Wno-discarded-qualifiers
4638 @opindex Wdiscarded-qualifiers
4639 Do not warn if type qualifiers on pointers are being discarded.
4640 Typically, the compiler warns if a @code{const char *} variable is
4641 passed to a function that takes a @code{char *} parameter. This option
4642 can be used to suppress such a warning.
4644 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4645 @opindex Wno-discarded-array-qualifiers
4646 @opindex Wdiscarded-array-qualifiers
4647 Do not warn if type qualifiers on arrays which are pointer targets
4648 are being discarded. Typically, the compiler warns if a
4649 @code{const int (*)[]} variable is passed to a function that
4650 takes a @code{int (*)[]} parameter. This option can be used to
4651 suppress such a warning.
4653 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4654 @opindex Wno-incompatible-pointer-types
4655 @opindex Wincompatible-pointer-types
4656 Do not warn when there is a conversion between pointers that have incompatible
4657 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4658 which warns for pointer argument passing or assignment with different
4661 @item -Wno-int-conversion @r{(C and Objective-C only)}
4662 @opindex Wno-int-conversion
4663 @opindex Wint-conversion
4664 Do not warn about incompatible integer to pointer and pointer to integer
4665 conversions. This warning is about implicit conversions; for explicit
4666 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4667 @option{-Wno-pointer-to-int-cast} may be used.
4669 @item -Wno-div-by-zero
4670 @opindex Wno-div-by-zero
4671 @opindex Wdiv-by-zero
4672 Do not warn about compile-time integer division by zero. Floating-point
4673 division by zero is not warned about, as it can be a legitimate way of
4674 obtaining infinities and NaNs.
4676 @item -Wsystem-headers
4677 @opindex Wsystem-headers
4678 @opindex Wno-system-headers
4679 @cindex warnings from system headers
4680 @cindex system headers, warnings from
4681 Print warning messages for constructs found in system header files.
4682 Warnings from system headers are normally suppressed, on the assumption
4683 that they usually do not indicate real problems and would only make the
4684 compiler output harder to read. Using this command-line option tells
4685 GCC to emit warnings from system headers as if they occurred in user
4686 code. However, note that using @option{-Wall} in conjunction with this
4687 option does @emph{not} warn about unknown pragmas in system
4688 headers---for that, @option{-Wunknown-pragmas} must also be used.
4690 @item -Wtautological-compare
4691 @opindex Wtautological-compare
4692 @opindex Wno-tautological-compare
4693 Warn if a self-comparison always evaluates to true or false. This
4694 warning detects various mistakes such as:
4698 if (i > i) @{ @dots{} @}
4700 This warning is enabled by @option{-Wall}.
4703 @opindex Wtrampolines
4704 @opindex Wno-trampolines
4705 Warn about trampolines generated for pointers to nested functions.
4706 A trampoline is a small piece of data or code that is created at run
4707 time on the stack when the address of a nested function is taken, and is
4708 used to call the nested function indirectly. For some targets, it is
4709 made up of data only and thus requires no special treatment. But, for
4710 most targets, it is made up of code and thus requires the stack to be
4711 made executable in order for the program to work properly.
4714 @opindex Wfloat-equal
4715 @opindex Wno-float-equal
4716 Warn if floating-point values are used in equality comparisons.
4718 The idea behind this is that sometimes it is convenient (for the
4719 programmer) to consider floating-point values as approximations to
4720 infinitely precise real numbers. If you are doing this, then you need
4721 to compute (by analyzing the code, or in some other way) the maximum or
4722 likely maximum error that the computation introduces, and allow for it
4723 when performing comparisons (and when producing output, but that's a
4724 different problem). In particular, instead of testing for equality, you
4725 should check to see whether the two values have ranges that overlap; and
4726 this is done with the relational operators, so equality comparisons are
4729 @item -Wtraditional @r{(C and Objective-C only)}
4730 @opindex Wtraditional
4731 @opindex Wno-traditional
4732 Warn about certain constructs that behave differently in traditional and
4733 ISO C@. Also warn about ISO C constructs that have no traditional C
4734 equivalent, and/or problematic constructs that should be avoided.
4738 Macro parameters that appear within string literals in the macro body.
4739 In traditional C macro replacement takes place within string literals,
4740 but in ISO C it does not.
4743 In traditional C, some preprocessor directives did not exist.
4744 Traditional preprocessors only considered a line to be a directive
4745 if the @samp{#} appeared in column 1 on the line. Therefore
4746 @option{-Wtraditional} warns about directives that traditional C
4747 understands but ignores because the @samp{#} does not appear as the
4748 first character on the line. It also suggests you hide directives like
4749 @code{#pragma} not understood by traditional C by indenting them. Some
4750 traditional implementations do not recognize @code{#elif}, so this option
4751 suggests avoiding it altogether.
4754 A function-like macro that appears without arguments.
4757 The unary plus operator.
4760 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4761 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4762 constants.) Note, these suffixes appear in macros defined in the system
4763 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4764 Use of these macros in user code might normally lead to spurious
4765 warnings, however GCC's integrated preprocessor has enough context to
4766 avoid warning in these cases.
4769 A function declared external in one block and then used after the end of
4773 A @code{switch} statement has an operand of type @code{long}.
4776 A non-@code{static} function declaration follows a @code{static} one.
4777 This construct is not accepted by some traditional C compilers.
4780 The ISO type of an integer constant has a different width or
4781 signedness from its traditional type. This warning is only issued if
4782 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4783 typically represent bit patterns, are not warned about.
4786 Usage of ISO string concatenation is detected.
4789 Initialization of automatic aggregates.
4792 Identifier conflicts with labels. Traditional C lacks a separate
4793 namespace for labels.
4796 Initialization of unions. If the initializer is zero, the warning is
4797 omitted. This is done under the assumption that the zero initializer in
4798 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4799 initializer warnings and relies on default initialization to zero in the
4803 Conversions by prototypes between fixed/floating-point values and vice
4804 versa. The absence of these prototypes when compiling with traditional
4805 C causes serious problems. This is a subset of the possible
4806 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4809 Use of ISO C style function definitions. This warning intentionally is
4810 @emph{not} issued for prototype declarations or variadic functions
4811 because these ISO C features appear in your code when using
4812 libiberty's traditional C compatibility macros, @code{PARAMS} and
4813 @code{VPARAMS}. This warning is also bypassed for nested functions
4814 because that feature is already a GCC extension and thus not relevant to
4815 traditional C compatibility.
4818 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4819 @opindex Wtraditional-conversion
4820 @opindex Wno-traditional-conversion
4821 Warn if a prototype causes a type conversion that is different from what
4822 would happen to the same argument in the absence of a prototype. This
4823 includes conversions of fixed point to floating and vice versa, and
4824 conversions changing the width or signedness of a fixed-point argument
4825 except when the same as the default promotion.
4827 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4828 @opindex Wdeclaration-after-statement
4829 @opindex Wno-declaration-after-statement
4830 Warn when a declaration is found after a statement in a block. This
4831 construct, known from C++, was introduced with ISO C99 and is by default
4832 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4837 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4839 @item -Wno-endif-labels
4840 @opindex Wno-endif-labels
4841 @opindex Wendif-labels
4842 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4847 Warn whenever a local variable or type declaration shadows another
4848 variable, parameter, type, class member (in C++), or instance variable
4849 (in Objective-C) or whenever a built-in function is shadowed. Note
4850 that in C++, the compiler warns if a local variable shadows an
4851 explicit typedef, but not if it shadows a struct/class/enum.
4853 @item -Wno-shadow-ivar @r{(Objective-C only)}
4854 @opindex Wno-shadow-ivar
4855 @opindex Wshadow-ivar
4856 Do not warn whenever a local variable shadows an instance variable in an
4859 @item -Wlarger-than=@var{len}
4860 @opindex Wlarger-than=@var{len}
4861 @opindex Wlarger-than-@var{len}
4862 Warn whenever an object of larger than @var{len} bytes is defined.
4864 @item -Wframe-larger-than=@var{len}
4865 @opindex Wframe-larger-than
4866 Warn if the size of a function frame is larger than @var{len} bytes.
4867 The computation done to determine the stack frame size is approximate
4868 and not conservative.
4869 The actual requirements may be somewhat greater than @var{len}
4870 even if you do not get a warning. In addition, any space allocated
4871 via @code{alloca}, variable-length arrays, or related constructs
4872 is not included by the compiler when determining
4873 whether or not to issue a warning.
4875 @item -Wno-free-nonheap-object
4876 @opindex Wno-free-nonheap-object
4877 @opindex Wfree-nonheap-object
4878 Do not warn when attempting to free an object that was not allocated
4881 @item -Wstack-usage=@var{len}
4882 @opindex Wstack-usage
4883 Warn if the stack usage of a function might be larger than @var{len} bytes.
4884 The computation done to determine the stack usage is conservative.
4885 Any space allocated via @code{alloca}, variable-length arrays, or related
4886 constructs is included by the compiler when determining whether or not to
4889 The message is in keeping with the output of @option{-fstack-usage}.
4893 If the stack usage is fully static but exceeds the specified amount, it's:
4896 warning: stack usage is 1120 bytes
4899 If the stack usage is (partly) dynamic but bounded, it's:
4902 warning: stack usage might be 1648 bytes
4905 If the stack usage is (partly) dynamic and not bounded, it's:
4908 warning: stack usage might be unbounded
4912 @item -Wunsafe-loop-optimizations
4913 @opindex Wunsafe-loop-optimizations
4914 @opindex Wno-unsafe-loop-optimizations
4915 Warn if the loop cannot be optimized because the compiler cannot
4916 assume anything on the bounds of the loop indices. With
4917 @option{-funsafe-loop-optimizations} warn if the compiler makes
4920 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4921 @opindex Wno-pedantic-ms-format
4922 @opindex Wpedantic-ms-format
4923 When used in combination with @option{-Wformat}
4924 and @option{-pedantic} without GNU extensions, this option
4925 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4926 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4927 which depend on the MS runtime.
4929 @item -Wplacement-new
4930 @opindex Wplacement-new
4931 @opindex Wno-placement-new
4932 Warn about placement new expressions with undefined behavior, such as
4933 constructing an object in a buffer that is smaller than the type of
4934 the object. For example, the placement new expression below is diagnosed
4935 because it attempts to construct an array of 64 integers in a buffer only
4941 This warning is enabled by default.
4943 @item -Wpointer-arith
4944 @opindex Wpointer-arith
4945 @opindex Wno-pointer-arith
4946 Warn about anything that depends on the ``size of'' a function type or
4947 of @code{void}. GNU C assigns these types a size of 1, for
4948 convenience in calculations with @code{void *} pointers and pointers
4949 to functions. In C++, warn also when an arithmetic operation involves
4950 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4953 @opindex Wtype-limits
4954 @opindex Wno-type-limits
4955 Warn if a comparison is always true or always false due to the limited
4956 range of the data type, but do not warn for constant expressions. For
4957 example, warn if an unsigned variable is compared against zero with
4958 @code{<} or @code{>=}. This warning is also enabled by
4961 @item -Wbad-function-cast @r{(C and Objective-C only)}
4962 @opindex Wbad-function-cast
4963 @opindex Wno-bad-function-cast
4964 Warn when a function call is cast to a non-matching type.
4965 For example, warn if a call to a function returning an integer type
4966 is cast to a pointer type.
4968 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4969 @opindex Wc90-c99-compat
4970 @opindex Wno-c90-c99-compat
4971 Warn about features not present in ISO C90, but present in ISO C99.
4972 For instance, warn about use of variable length arrays, @code{long long}
4973 type, @code{bool} type, compound literals, designated initializers, and so
4974 on. This option is independent of the standards mode. Warnings are disabled
4975 in the expression that follows @code{__extension__}.
4977 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4978 @opindex Wc99-c11-compat
4979 @opindex Wno-c99-c11-compat
4980 Warn about features not present in ISO C99, but present in ISO C11.
4981 For instance, warn about use of anonymous structures and unions,
4982 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4983 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4984 and so on. This option is independent of the standards mode. Warnings are
4985 disabled in the expression that follows @code{__extension__}.
4987 @item -Wc++-compat @r{(C and Objective-C only)}
4988 @opindex Wc++-compat
4989 Warn about ISO C constructs that are outside of the common subset of
4990 ISO C and ISO C++, e.g.@: request for implicit conversion from
4991 @code{void *} to a pointer to non-@code{void} type.
4993 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4994 @opindex Wc++11-compat
4995 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4996 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4997 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4998 enabled by @option{-Wall}.
5000 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5001 @opindex Wc++14-compat
5002 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5003 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5007 @opindex Wno-cast-qual
5008 Warn whenever a pointer is cast so as to remove a type qualifier from
5009 the target type. For example, warn if a @code{const char *} is cast
5010 to an ordinary @code{char *}.
5012 Also warn when making a cast that introduces a type qualifier in an
5013 unsafe way. For example, casting @code{char **} to @code{const char **}
5014 is unsafe, as in this example:
5017 /* p is char ** value. */
5018 const char **q = (const char **) p;
5019 /* Assignment of readonly string to const char * is OK. */
5021 /* Now char** pointer points to read-only memory. */
5026 @opindex Wcast-align
5027 @opindex Wno-cast-align
5028 Warn whenever a pointer is cast such that the required alignment of the
5029 target is increased. For example, warn if a @code{char *} is cast to
5030 an @code{int *} on machines where integers can only be accessed at
5031 two- or four-byte boundaries.
5033 @item -Wwrite-strings
5034 @opindex Wwrite-strings
5035 @opindex Wno-write-strings
5036 When compiling C, give string constants the type @code{const
5037 char[@var{length}]} so that copying the address of one into a
5038 non-@code{const} @code{char *} pointer produces a warning. These
5039 warnings help you find at compile time code that can try to write
5040 into a string constant, but only if you have been very careful about
5041 using @code{const} in declarations and prototypes. Otherwise, it is
5042 just a nuisance. This is why we did not make @option{-Wall} request
5045 When compiling C++, warn about the deprecated conversion from string
5046 literals to @code{char *}. This warning is enabled by default for C++
5051 @opindex Wno-clobbered
5052 Warn for variables that might be changed by @code{longjmp} or
5053 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5055 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5056 @opindex Wconditionally-supported
5057 @opindex Wno-conditionally-supported
5058 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5061 @opindex Wconversion
5062 @opindex Wno-conversion
5063 Warn for implicit conversions that may alter a value. This includes
5064 conversions between real and integer, like @code{abs (x)} when
5065 @code{x} is @code{double}; conversions between signed and unsigned,
5066 like @code{unsigned ui = -1}; and conversions to smaller types, like
5067 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5068 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5069 changed by the conversion like in @code{abs (2.0)}. Warnings about
5070 conversions between signed and unsigned integers can be disabled by
5071 using @option{-Wno-sign-conversion}.
5073 For C++, also warn for confusing overload resolution for user-defined
5074 conversions; and conversions that never use a type conversion
5075 operator: conversions to @code{void}, the same type, a base class or a
5076 reference to them. Warnings about conversions between signed and
5077 unsigned integers are disabled by default in C++ unless
5078 @option{-Wsign-conversion} is explicitly enabled.
5080 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5081 @opindex Wconversion-null
5082 @opindex Wno-conversion-null
5083 Do not warn for conversions between @code{NULL} and non-pointer
5084 types. @option{-Wconversion-null} is enabled by default.
5086 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5087 @opindex Wzero-as-null-pointer-constant
5088 @opindex Wno-zero-as-null-pointer-constant
5089 Warn when a literal '0' is used as null pointer constant. This can
5090 be useful to facilitate the conversion to @code{nullptr} in C++11.
5092 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5093 @opindex Wsubobject-linkage
5094 @opindex Wno-subobject-linkage
5095 Warn if a class type has a base or a field whose type uses the anonymous
5096 namespace or depends on a type with no linkage. If a type A depends on
5097 a type B with no or internal linkage, defining it in multiple
5098 translation units would be an ODR violation because the meaning of B
5099 is different in each translation unit. If A only appears in a single
5100 translation unit, the best way to silence the warning is to give it
5101 internal linkage by putting it in an anonymous namespace as well. The
5102 compiler doesn't give this warning for types defined in the main .C
5103 file, as those are unlikely to have multiple definitions.
5104 @option{-Wsubobject-linkage} is enabled by default.
5108 @opindex Wno-date-time
5109 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5110 are encountered as they might prevent bit-wise-identical reproducible
5113 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5114 @opindex Wdelete-incomplete
5115 @opindex Wno-delete-incomplete
5116 Warn when deleting a pointer to incomplete type, which may cause
5117 undefined behavior at runtime. This warning is enabled by default.
5119 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5120 @opindex Wuseless-cast
5121 @opindex Wno-useless-cast
5122 Warn when an expression is casted to its own type.
5125 @opindex Wempty-body
5126 @opindex Wno-empty-body
5127 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5128 while} statement. This warning is also enabled by @option{-Wextra}.
5130 @item -Wenum-compare
5131 @opindex Wenum-compare
5132 @opindex Wno-enum-compare
5133 Warn about a comparison between values of different enumerated types.
5134 In C++ enumeral mismatches in conditional expressions are also
5135 diagnosed and the warning is enabled by default. In C this warning is
5136 enabled by @option{-Wall}.
5138 @item -Wjump-misses-init @r{(C, Objective-C only)}
5139 @opindex Wjump-misses-init
5140 @opindex Wno-jump-misses-init
5141 Warn if a @code{goto} statement or a @code{switch} statement jumps
5142 forward across the initialization of a variable, or jumps backward to a
5143 label after the variable has been initialized. This only warns about
5144 variables that are initialized when they are declared. This warning is
5145 only supported for C and Objective-C; in C++ this sort of branch is an
5148 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5149 can be disabled with the @option{-Wno-jump-misses-init} option.
5151 @item -Wsign-compare
5152 @opindex Wsign-compare
5153 @opindex Wno-sign-compare
5154 @cindex warning for comparison of signed and unsigned values
5155 @cindex comparison of signed and unsigned values, warning
5156 @cindex signed and unsigned values, comparison warning
5157 Warn when a comparison between signed and unsigned values could produce
5158 an incorrect result when the signed value is converted to unsigned.
5159 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5160 also enabled by @option{-Wextra}.
5162 @item -Wsign-conversion
5163 @opindex Wsign-conversion
5164 @opindex Wno-sign-conversion
5165 Warn for implicit conversions that may change the sign of an integer
5166 value, like assigning a signed integer expression to an unsigned
5167 integer variable. An explicit cast silences the warning. In C, this
5168 option is enabled also by @option{-Wconversion}.
5170 @item -Wfloat-conversion
5171 @opindex Wfloat-conversion
5172 @opindex Wno-float-conversion
5173 Warn for implicit conversions that reduce the precision of a real value.
5174 This includes conversions from real to integer, and from higher precision
5175 real to lower precision real values. This option is also enabled by
5176 @option{-Wconversion}.
5178 @item -Wno-scalar-storage-order
5179 @opindex -Wno-scalar-storage-order
5180 @opindex -Wscalar-storage-order
5181 Do not warn on suspicious constructs involving reverse scalar storage order.
5183 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5184 @opindex Wsized-deallocation
5185 @opindex Wno-sized-deallocation
5186 Warn about a definition of an unsized deallocation function
5188 void operator delete (void *) noexcept;
5189 void operator delete[] (void *) noexcept;
5191 without a definition of the corresponding sized deallocation function
5193 void operator delete (void *, std::size_t) noexcept;
5194 void operator delete[] (void *, std::size_t) noexcept;
5196 or vice versa. Enabled by @option{-Wextra} along with
5197 @option{-fsized-deallocation}.
5199 @item -Wsizeof-pointer-memaccess
5200 @opindex Wsizeof-pointer-memaccess
5201 @opindex Wno-sizeof-pointer-memaccess
5202 Warn for suspicious length parameters to certain string and memory built-in
5203 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5204 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5205 but a pointer, and suggests a possible fix, or about
5206 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5209 @item -Wsizeof-array-argument
5210 @opindex Wsizeof-array-argument
5211 @opindex Wno-sizeof-array-argument
5212 Warn when the @code{sizeof} operator is applied to a parameter that is
5213 declared as an array in a function definition. This warning is enabled by
5214 default for C and C++ programs.
5216 @item -Wmemset-transposed-args
5217 @opindex Wmemset-transposed-args
5218 @opindex Wno-memset-transposed-args
5219 Warn for suspicious calls to the @code{memset} built-in function, if the
5220 second argument is not zero and the third argument is zero. This warns e.g.@
5221 about @code{memset (buf, sizeof buf, 0)} where most probably
5222 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5223 is only emitted if the third argument is literal zero. If it is some
5224 expression that is folded to zero, a cast of zero to some type, etc.,
5225 it is far less likely that the user has mistakenly exchanged the arguments
5226 and no warning is emitted. This warning is enabled by @option{-Wall}.
5230 @opindex Wno-address
5231 Warn about suspicious uses of memory addresses. These include using
5232 the address of a function in a conditional expression, such as
5233 @code{void func(void); if (func)}, and comparisons against the memory
5234 address of a string literal, such as @code{if (x == "abc")}. Such
5235 uses typically indicate a programmer error: the address of a function
5236 always evaluates to true, so their use in a conditional usually
5237 indicate that the programmer forgot the parentheses in a function
5238 call; and comparisons against string literals result in unspecified
5239 behavior and are not portable in C, so they usually indicate that the
5240 programmer intended to use @code{strcmp}. This warning is enabled by
5244 @opindex Wlogical-op
5245 @opindex Wno-logical-op
5246 Warn about suspicious uses of logical operators in expressions.
5247 This includes using logical operators in contexts where a
5248 bit-wise operator is likely to be expected. Also warns when
5249 the operands of a logical operator are the same:
5252 if (a < 0 && a < 0) @{ @dots{} @}
5255 @item -Wlogical-not-parentheses
5256 @opindex Wlogical-not-parentheses
5257 @opindex Wno-logical-not-parentheses
5258 Warn about logical not used on the left hand side operand of a comparison.
5259 This option does not warn if the RHS operand is of a boolean type. Its
5260 purpose is to detect suspicious code like the following:
5264 if (!a > 1) @{ @dots{} @}
5267 It is possible to suppress the warning by wrapping the LHS into
5270 if ((!a) > 1) @{ @dots{} @}
5273 This warning is enabled by @option{-Wall}.
5275 @item -Waggregate-return
5276 @opindex Waggregate-return
5277 @opindex Wno-aggregate-return
5278 Warn if any functions that return structures or unions are defined or
5279 called. (In languages where you can return an array, this also elicits
5282 @item -Wno-aggressive-loop-optimizations
5283 @opindex Wno-aggressive-loop-optimizations
5284 @opindex Waggressive-loop-optimizations
5285 Warn if in a loop with constant number of iterations the compiler detects
5286 undefined behavior in some statement during one or more of the iterations.
5288 @item -Wno-attributes
5289 @opindex Wno-attributes
5290 @opindex Wattributes
5291 Do not warn if an unexpected @code{__attribute__} is used, such as
5292 unrecognized attributes, function attributes applied to variables,
5293 etc. This does not stop errors for incorrect use of supported
5296 @item -Wno-builtin-macro-redefined
5297 @opindex Wno-builtin-macro-redefined
5298 @opindex Wbuiltin-macro-redefined
5299 Do not warn if certain built-in macros are redefined. This suppresses
5300 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5301 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5303 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5304 @opindex Wstrict-prototypes
5305 @opindex Wno-strict-prototypes
5306 Warn if a function is declared or defined without specifying the
5307 argument types. (An old-style function definition is permitted without
5308 a warning if preceded by a declaration that specifies the argument
5311 @item -Wold-style-declaration @r{(C and Objective-C only)}
5312 @opindex Wold-style-declaration
5313 @opindex Wno-old-style-declaration
5314 Warn for obsolescent usages, according to the C Standard, in a
5315 declaration. For example, warn if storage-class specifiers like
5316 @code{static} are not the first things in a declaration. This warning
5317 is also enabled by @option{-Wextra}.
5319 @item -Wold-style-definition @r{(C and Objective-C only)}
5320 @opindex Wold-style-definition
5321 @opindex Wno-old-style-definition
5322 Warn if an old-style function definition is used. A warning is given
5323 even if there is a previous prototype.
5325 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5326 @opindex Wmissing-parameter-type
5327 @opindex Wno-missing-parameter-type
5328 A function parameter is declared without a type specifier in K&R-style
5335 This warning is also enabled by @option{-Wextra}.
5337 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5338 @opindex Wmissing-prototypes
5339 @opindex Wno-missing-prototypes
5340 Warn if a global function is defined without a previous prototype
5341 declaration. This warning is issued even if the definition itself
5342 provides a prototype. Use this option to detect global functions
5343 that do not have a matching prototype declaration in a header file.
5344 This option is not valid for C++ because all function declarations
5345 provide prototypes and a non-matching declaration declares an
5346 overload rather than conflict with an earlier declaration.
5347 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5349 @item -Wmissing-declarations
5350 @opindex Wmissing-declarations
5351 @opindex Wno-missing-declarations
5352 Warn if a global function is defined without a previous declaration.
5353 Do so even if the definition itself provides a prototype.
5354 Use this option to detect global functions that are not declared in
5355 header files. In C, no warnings are issued for functions with previous
5356 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5357 missing prototypes. In C++, no warnings are issued for function templates,
5358 or for inline functions, or for functions in anonymous namespaces.
5360 @item -Wmissing-field-initializers
5361 @opindex Wmissing-field-initializers
5362 @opindex Wno-missing-field-initializers
5366 Warn if a structure's initializer has some fields missing. For
5367 example, the following code causes such a warning, because
5368 @code{x.h} is implicitly zero:
5371 struct s @{ int f, g, h; @};
5372 struct s x = @{ 3, 4 @};
5375 This option does not warn about designated initializers, so the following
5376 modification does not trigger a warning:
5379 struct s @{ int f, g, h; @};
5380 struct s x = @{ .f = 3, .g = 4 @};
5383 In C++ this option does not warn either about the empty @{ @}
5384 initializer, for example:
5387 struct s @{ int f, g, h; @};
5391 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5392 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5394 @item -Wno-multichar
5395 @opindex Wno-multichar
5397 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5398 Usually they indicate a typo in the user's code, as they have
5399 implementation-defined values, and should not be used in portable code.
5401 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5402 @opindex Wnormalized=
5403 @opindex Wnormalized
5404 @opindex Wno-normalized
5407 @cindex character set, input normalization
5408 In ISO C and ISO C++, two identifiers are different if they are
5409 different sequences of characters. However, sometimes when characters
5410 outside the basic ASCII character set are used, you can have two
5411 different character sequences that look the same. To avoid confusion,
5412 the ISO 10646 standard sets out some @dfn{normalization rules} which
5413 when applied ensure that two sequences that look the same are turned into
5414 the same sequence. GCC can warn you if you are using identifiers that
5415 have not been normalized; this option controls that warning.
5417 There are four levels of warning supported by GCC@. The default is
5418 @option{-Wnormalized=nfc}, which warns about any identifier that is
5419 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5420 recommended form for most uses. It is equivalent to
5421 @option{-Wnormalized}.
5423 Unfortunately, there are some characters allowed in identifiers by
5424 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5425 identifiers. That is, there's no way to use these symbols in portable
5426 ISO C or C++ and have all your identifiers in NFC@.
5427 @option{-Wnormalized=id} suppresses the warning for these characters.
5428 It is hoped that future versions of the standards involved will correct
5429 this, which is why this option is not the default.
5431 You can switch the warning off for all characters by writing
5432 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5433 only do this if you are using some other normalization scheme (like
5434 ``D''), because otherwise you can easily create bugs that are
5435 literally impossible to see.
5437 Some characters in ISO 10646 have distinct meanings but look identical
5438 in some fonts or display methodologies, especially once formatting has
5439 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5440 LETTER N'', displays just like a regular @code{n} that has been
5441 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5442 normalization scheme to convert all these into a standard form as
5443 well, and GCC warns if your code is not in NFKC if you use
5444 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5445 about every identifier that contains the letter O because it might be
5446 confused with the digit 0, and so is not the default, but may be
5447 useful as a local coding convention if the programming environment
5448 cannot be fixed to display these characters distinctly.
5450 @item -Wno-deprecated
5451 @opindex Wno-deprecated
5452 @opindex Wdeprecated
5453 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5455 @item -Wno-deprecated-declarations
5456 @opindex Wno-deprecated-declarations
5457 @opindex Wdeprecated-declarations
5458 Do not warn about uses of functions (@pxref{Function Attributes}),
5459 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5460 Attributes}) marked as deprecated by using the @code{deprecated}
5464 @opindex Wno-overflow
5466 Do not warn about compile-time overflow in constant expressions.
5471 Warn about One Definition Rule violations during link-time optimization.
5472 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5475 @opindex Wopenm-simd
5476 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5477 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5478 option can be used to relax the cost model.
5480 @item -Woverride-init @r{(C and Objective-C only)}
5481 @opindex Woverride-init
5482 @opindex Wno-override-init
5486 Warn if an initialized field without side effects is overridden when
5487 using designated initializers (@pxref{Designated Inits, , Designated
5490 This warning is included in @option{-Wextra}. To get other
5491 @option{-Wextra} warnings without this one, use @option{-Wextra
5492 -Wno-override-init}.
5494 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5495 @opindex Woverride-init-side-effects
5496 @opindex Wno-override-init-side-effects
5497 Warn if an initialized field with side effects is overridden when
5498 using designated initializers (@pxref{Designated Inits, , Designated
5499 Initializers}). This warning is enabled by default.
5504 Warn if a structure is given the packed attribute, but the packed
5505 attribute has no effect on the layout or size of the structure.
5506 Such structures may be mis-aligned for little benefit. For
5507 instance, in this code, the variable @code{f.x} in @code{struct bar}
5508 is misaligned even though @code{struct bar} does not itself
5509 have the packed attribute:
5516 @} __attribute__((packed));
5524 @item -Wpacked-bitfield-compat
5525 @opindex Wpacked-bitfield-compat
5526 @opindex Wno-packed-bitfield-compat
5527 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5528 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5529 the change can lead to differences in the structure layout. GCC
5530 informs you when the offset of such a field has changed in GCC 4.4.
5531 For example there is no longer a 4-bit padding between field @code{a}
5532 and @code{b} in this structure:
5539 @} __attribute__ ((packed));
5542 This warning is enabled by default. Use
5543 @option{-Wno-packed-bitfield-compat} to disable this warning.
5548 Warn if padding is included in a structure, either to align an element
5549 of the structure or to align the whole structure. Sometimes when this
5550 happens it is possible to rearrange the fields of the structure to
5551 reduce the padding and so make the structure smaller.
5553 @item -Wredundant-decls
5554 @opindex Wredundant-decls
5555 @opindex Wno-redundant-decls
5556 Warn if anything is declared more than once in the same scope, even in
5557 cases where multiple declaration is valid and changes nothing.
5559 @item -Wnested-externs @r{(C and Objective-C only)}
5560 @opindex Wnested-externs
5561 @opindex Wno-nested-externs
5562 Warn if an @code{extern} declaration is encountered within a function.
5564 @item -Wno-inherited-variadic-ctor
5565 @opindex Winherited-variadic-ctor
5566 @opindex Wno-inherited-variadic-ctor
5567 Suppress warnings about use of C++11 inheriting constructors when the
5568 base class inherited from has a C variadic constructor; the warning is
5569 on by default because the ellipsis is not inherited.
5574 Warn if a function that is declared as inline cannot be inlined.
5575 Even with this option, the compiler does not warn about failures to
5576 inline functions declared in system headers.
5578 The compiler uses a variety of heuristics to determine whether or not
5579 to inline a function. For example, the compiler takes into account
5580 the size of the function being inlined and the amount of inlining
5581 that has already been done in the current function. Therefore,
5582 seemingly insignificant changes in the source program can cause the
5583 warnings produced by @option{-Winline} to appear or disappear.
5585 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5586 @opindex Wno-invalid-offsetof
5587 @opindex Winvalid-offsetof
5588 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5589 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5590 to a non-standard-layout type is undefined. In existing C++ implementations,
5591 however, @code{offsetof} typically gives meaningful results.
5592 This flag is for users who are aware that they are
5593 writing nonportable code and who have deliberately chosen to ignore the
5596 The restrictions on @code{offsetof} may be relaxed in a future version
5597 of the C++ standard.
5599 @item -Wno-int-to-pointer-cast
5600 @opindex Wno-int-to-pointer-cast
5601 @opindex Wint-to-pointer-cast
5602 Suppress warnings from casts to pointer type of an integer of a
5603 different size. In C++, casting to a pointer type of smaller size is
5604 an error. @option{Wint-to-pointer-cast} is enabled by default.
5607 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5608 @opindex Wno-pointer-to-int-cast
5609 @opindex Wpointer-to-int-cast
5610 Suppress warnings from casts from a pointer to an integer type of a
5614 @opindex Winvalid-pch
5615 @opindex Wno-invalid-pch
5616 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5617 the search path but can't be used.
5621 @opindex Wno-long-long
5622 Warn if @code{long long} type is used. This is enabled by either
5623 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5624 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5626 @item -Wvariadic-macros
5627 @opindex Wvariadic-macros
5628 @opindex Wno-variadic-macros
5629 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5630 alternate syntax is used in ISO C99 mode. This is enabled by either
5631 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5632 messages, use @option{-Wno-variadic-macros}.
5636 @opindex Wno-varargs
5637 Warn upon questionable usage of the macros used to handle variable
5638 arguments like @code{va_start}. This is default. To inhibit the
5639 warning messages, use @option{-Wno-varargs}.
5641 @item -Wvector-operation-performance
5642 @opindex Wvector-operation-performance
5643 @opindex Wno-vector-operation-performance
5644 Warn if vector operation is not implemented via SIMD capabilities of the
5645 architecture. Mainly useful for the performance tuning.
5646 Vector operation can be implemented @code{piecewise}, which means that the
5647 scalar operation is performed on every vector element;
5648 @code{in parallel}, which means that the vector operation is implemented
5649 using scalars of wider type, which normally is more performance efficient;
5650 and @code{as a single scalar}, which means that vector fits into a
5653 @item -Wno-virtual-move-assign
5654 @opindex Wvirtual-move-assign
5655 @opindex Wno-virtual-move-assign
5656 Suppress warnings about inheriting from a virtual base with a
5657 non-trivial C++11 move assignment operator. This is dangerous because
5658 if the virtual base is reachable along more than one path, it is
5659 moved multiple times, which can mean both objects end up in the
5660 moved-from state. If the move assignment operator is written to avoid
5661 moving from a moved-from object, this warning can be disabled.
5666 Warn if variable length array is used in the code.
5667 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5668 the variable length array.
5670 @item -Wvolatile-register-var
5671 @opindex Wvolatile-register-var
5672 @opindex Wno-volatile-register-var
5673 Warn if a register variable is declared volatile. The volatile
5674 modifier does not inhibit all optimizations that may eliminate reads
5675 and/or writes to register variables. This warning is enabled by
5678 @item -Wdisabled-optimization
5679 @opindex Wdisabled-optimization
5680 @opindex Wno-disabled-optimization
5681 Warn if a requested optimization pass is disabled. This warning does
5682 not generally indicate that there is anything wrong with your code; it
5683 merely indicates that GCC's optimizers are unable to handle the code
5684 effectively. Often, the problem is that your code is too big or too
5685 complex; GCC refuses to optimize programs when the optimization
5686 itself is likely to take inordinate amounts of time.
5688 @item -Wpointer-sign @r{(C and Objective-C only)}
5689 @opindex Wpointer-sign
5690 @opindex Wno-pointer-sign
5691 Warn for pointer argument passing or assignment with different signedness.
5692 This option is only supported for C and Objective-C@. It is implied by
5693 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5694 @option{-Wno-pointer-sign}.
5696 @item -Wstack-protector
5697 @opindex Wstack-protector
5698 @opindex Wno-stack-protector
5699 This option is only active when @option{-fstack-protector} is active. It
5700 warns about functions that are not protected against stack smashing.
5702 @item -Woverlength-strings
5703 @opindex Woverlength-strings
5704 @opindex Wno-overlength-strings
5705 Warn about string constants that are longer than the ``minimum
5706 maximum'' length specified in the C standard. Modern compilers
5707 generally allow string constants that are much longer than the
5708 standard's minimum limit, but very portable programs should avoid
5709 using longer strings.
5711 The limit applies @emph{after} string constant concatenation, and does
5712 not count the trailing NUL@. In C90, the limit was 509 characters; in
5713 C99, it was raised to 4095. C++98 does not specify a normative
5714 minimum maximum, so we do not diagnose overlength strings in C++@.
5716 This option is implied by @option{-Wpedantic}, and can be disabled with
5717 @option{-Wno-overlength-strings}.
5719 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5720 @opindex Wunsuffixed-float-constants
5722 Issue a warning for any floating constant that does not have
5723 a suffix. When used together with @option{-Wsystem-headers} it
5724 warns about such constants in system header files. This can be useful
5725 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5726 from the decimal floating-point extension to C99.
5728 @item -Wno-designated-init @r{(C and Objective-C only)}
5729 Suppress warnings when a positional initializer is used to initialize
5730 a structure that has been marked with the @code{designated_init}
5735 @node Debugging Options
5736 @section Options for Debugging Your Program or GCC
5737 @cindex options, debugging
5738 @cindex debugging information options
5740 GCC has various special options that are used for debugging
5741 either your program or GCC:
5746 Produce debugging information in the operating system's native format
5747 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5750 On most systems that use stabs format, @option{-g} enables use of extra
5751 debugging information that only GDB can use; this extra information
5752 makes debugging work better in GDB but probably makes other debuggers
5754 refuse to read the program. If you want to control for certain whether
5755 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5756 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5758 GCC allows you to use @option{-g} with
5759 @option{-O}. The shortcuts taken by optimized code may occasionally
5760 produce surprising results: some variables you declared may not exist
5761 at all; flow of control may briefly move where you did not expect it;
5762 some statements may not be executed because they compute constant
5763 results or their values are already at hand; some statements may
5764 execute in different places because they have been moved out of loops.
5766 Nevertheless it proves possible to debug optimized output. This makes
5767 it reasonable to use the optimizer for programs that might have bugs.
5769 The following options are useful when GCC is generated with the
5770 capability for more than one debugging format.
5773 @opindex gsplit-dwarf
5774 Separate as much dwarf debugging information as possible into a
5775 separate output file with the extension .dwo. This option allows
5776 the build system to avoid linking files with debug information. To
5777 be useful, this option requires a debugger capable of reading .dwo
5782 Produce debugging information for use by GDB@. This means to use the
5783 most expressive format available (DWARF 2, stabs, or the native format
5784 if neither of those are supported), including GDB extensions if at all
5789 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5791 @item -ggnu-pubnames
5792 @opindex ggnu-pubnames
5793 Generate .debug_pubnames and .debug_pubtypes sections in a format
5794 suitable for conversion into a GDB@ index. This option is only useful
5795 with a linker that can produce GDB@ index version 7.
5799 Produce debugging information in stabs format (if that is supported),
5800 without GDB extensions. This is the format used by DBX on most BSD
5801 systems. On MIPS, Alpha and System V Release 4 systems this option
5802 produces stabs debugging output that is not understood by DBX or SDB@.
5803 On System V Release 4 systems this option requires the GNU assembler.
5805 @item -feliminate-unused-debug-symbols
5806 @opindex feliminate-unused-debug-symbols
5807 Produce debugging information in stabs format (if that is supported),
5808 for only symbols that are actually used.
5810 @item -femit-class-debug-always
5811 @opindex femit-class-debug-always
5812 Instead of emitting debugging information for a C++ class in only one
5813 object file, emit it in all object files using the class. This option
5814 should be used only with debuggers that are unable to handle the way GCC
5815 normally emits debugging information for classes because using this
5816 option increases the size of debugging information by as much as a
5819 @item -fdebug-types-section
5820 @opindex fdebug-types-section
5821 @opindex fno-debug-types-section
5822 When using DWARF Version 4 or higher, type DIEs can be put into
5823 their own @code{.debug_types} section instead of making them part of the
5824 @code{.debug_info} section. It is more efficient to put them in a separate
5825 comdat sections since the linker can then remove duplicates.
5826 But not all DWARF consumers support @code{.debug_types} sections yet
5827 and on some objects @code{.debug_types} produces larger instead of smaller
5828 debugging information.
5832 Produce debugging information in stabs format (if that is supported),
5833 using GNU extensions understood only by the GNU debugger (GDB)@. The
5834 use of these extensions is likely to make other debuggers crash or
5835 refuse to read the program.
5839 Produce debugging information in COFF format (if that is supported).
5840 This is the format used by SDB on most System V systems prior to
5845 Produce debugging information in XCOFF format (if that is supported).
5846 This is the format used by the DBX debugger on IBM RS/6000 systems.
5850 Produce debugging information in XCOFF format (if that is supported),
5851 using GNU extensions understood only by the GNU debugger (GDB)@. The
5852 use of these extensions is likely to make other debuggers crash or
5853 refuse to read the program, and may cause assemblers other than the GNU
5854 assembler (GAS) to fail with an error.
5856 @item -gdwarf-@var{version}
5857 @opindex gdwarf-@var{version}
5858 Produce debugging information in DWARF format (if that is supported).
5859 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5860 for most targets is 4. DWARF Version 5 is only experimental.
5862 Note that with DWARF Version 2, some ports require and always
5863 use some non-conflicting DWARF 3 extensions in the unwind tables.
5865 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5866 for maximum benefit.
5868 @item -grecord-gcc-switches
5869 @opindex grecord-gcc-switches
5870 This switch causes the command-line options used to invoke the
5871 compiler that may affect code generation to be appended to the
5872 DW_AT_producer attribute in DWARF debugging information. The options
5873 are concatenated with spaces separating them from each other and from
5874 the compiler version. See also @option{-frecord-gcc-switches} for another
5875 way of storing compiler options into the object file. This is the default.
5877 @item -gno-record-gcc-switches
5878 @opindex gno-record-gcc-switches
5879 Disallow appending command-line options to the DW_AT_producer attribute
5880 in DWARF debugging information.
5882 @item -gstrict-dwarf
5883 @opindex gstrict-dwarf
5884 Disallow using extensions of later DWARF standard version than selected
5885 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5886 DWARF extensions from later standard versions is allowed.
5888 @item -gno-strict-dwarf
5889 @opindex gno-strict-dwarf
5890 Allow using extensions of later DWARF standard version than selected with
5891 @option{-gdwarf-@var{version}}.
5893 @item -gz@r{[}=@var{type}@r{]}
5895 Produce compressed debug sections in DWARF format, if that is supported.
5896 If @var{type} is not given, the default type depends on the capabilities
5897 of the assembler and linker used. @var{type} may be one of
5898 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5899 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5900 compression in traditional GNU format). If the linker doesn't support
5901 writing compressed debug sections, the option is rejected. Otherwise,
5902 if the assembler does not support them, @option{-gz} is silently ignored
5903 when producing object files.
5907 Produce debugging information in Alpha/VMS debug format (if that is
5908 supported). This is the format used by DEBUG on Alpha/VMS systems.
5911 @itemx -ggdb@var{level}
5912 @itemx -gstabs@var{level}
5913 @itemx -gcoff@var{level}
5914 @itemx -gxcoff@var{level}
5915 @itemx -gvms@var{level}
5916 Request debugging information and also use @var{level} to specify how
5917 much information. The default level is 2.
5919 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5922 Level 1 produces minimal information, enough for making backtraces in
5923 parts of the program that you don't plan to debug. This includes
5924 descriptions of functions and external variables, and line number
5925 tables, but no information about local variables.
5927 Level 3 includes extra information, such as all the macro definitions
5928 present in the program. Some debuggers support macro expansion when
5929 you use @option{-g3}.
5931 @option{-gdwarf-2} does not accept a concatenated debug level, because
5932 GCC used to support an option @option{-gdwarf} that meant to generate
5933 debug information in version 1 of the DWARF format (which is very
5934 different from version 2), and it would have been too confusing. That
5935 debug format is long obsolete, but the option cannot be changed now.
5936 Instead use an additional @option{-g@var{level}} option to change the
5937 debug level for DWARF.
5941 Turn off generation of debug info, if leaving out this option
5942 generates it, or turn it on at level 2 otherwise. The position of this
5943 argument in the command line does not matter; it takes effect after all
5944 other options are processed, and it does so only once, no matter how
5945 many times it is given. This is mainly intended to be used with
5946 @option{-fcompare-debug}.
5948 @item -fsanitize=address
5949 @opindex fsanitize=address
5950 Enable AddressSanitizer, a fast memory error detector.
5951 Memory access instructions are instrumented to detect
5952 out-of-bounds and use-after-free bugs.
5953 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
5954 more details. The run-time behavior can be influenced using the
5955 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
5956 the available options are shown at startup of the instrumended program. See
5957 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
5958 for a list of supported options.
5960 @item -fsanitize=kernel-address
5961 @opindex fsanitize=kernel-address
5962 Enable AddressSanitizer for Linux kernel.
5963 See @uref{https://github.com/google/kasan/wiki} for more details.
5965 @item -fsanitize=thread
5966 @opindex fsanitize=thread
5967 Enable ThreadSanitizer, a fast data race detector.
5968 Memory access instructions are instrumented to detect
5969 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
5970 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5971 environment variable; see
5972 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
5975 @item -fsanitize=leak
5976 @opindex fsanitize=leak
5977 Enable LeakSanitizer, a memory leak detector.
5978 This option only matters for linking of executables and if neither
5979 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5980 case the executable is linked against a library that overrides @code{malloc}
5981 and other allocator functions. See
5982 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
5983 details. The run-time behavior can be influenced using the
5984 @env{LSAN_OPTIONS} environment variable.
5986 @item -fsanitize=undefined
5987 @opindex fsanitize=undefined
5988 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5989 Various computations are instrumented to detect undefined behavior
5990 at runtime. Current suboptions are:
5994 @item -fsanitize=shift
5995 @opindex fsanitize=shift
5996 This option enables checking that the result of a shift operation is
5997 not undefined. Note that what exactly is considered undefined differs
5998 slightly between C and C++, as well as between ISO C90 and C99, etc.
6000 @item -fsanitize=integer-divide-by-zero
6001 @opindex fsanitize=integer-divide-by-zero
6002 Detect integer division by zero as well as @code{INT_MIN / -1} division.
6004 @item -fsanitize=unreachable
6005 @opindex fsanitize=unreachable
6006 With this option, the compiler turns the @code{__builtin_unreachable}
6007 call into a diagnostics message call instead. When reaching the
6008 @code{__builtin_unreachable} call, the behavior is undefined.
6010 @item -fsanitize=vla-bound
6011 @opindex fsanitize=vla-bound
6012 This option instructs the compiler to check that the size of a variable
6013 length array is positive.
6015 @item -fsanitize=null
6016 @opindex fsanitize=null
6017 This option enables pointer checking. Particularly, the application
6018 built with this option turned on will issue an error message when it
6019 tries to dereference a NULL pointer, or if a reference (possibly an
6020 rvalue reference) is bound to a NULL pointer, or if a method is invoked
6021 on an object pointed by a NULL pointer.
6023 @item -fsanitize=return
6024 @opindex fsanitize=return
6025 This option enables return statement checking. Programs
6026 built with this option turned on will issue an error message
6027 when the end of a non-void function is reached without actually
6028 returning a value. This option works in C++ only.
6030 @item -fsanitize=signed-integer-overflow
6031 @opindex fsanitize=signed-integer-overflow
6032 This option enables signed integer overflow checking. We check that
6033 the result of @code{+}, @code{*}, and both unary and binary @code{-}
6034 does not overflow in the signed arithmetics. Note, integer promotion
6035 rules must be taken into account. That is, the following is not an
6038 signed char a = SCHAR_MAX;
6042 @item -fsanitize=bounds
6043 @opindex fsanitize=bounds
6044 This option enables instrumentation of array bounds. Various out of bounds
6045 accesses are detected. Flexible array members, flexible array member-like
6046 arrays, and initializers of variables with static storage are not instrumented.
6048 @item -fsanitize=bounds-strict
6049 @opindex fsanitize=bounds-strict
6050 This option enables strict instrumentation of array bounds. Most out of bounds
6051 accesses are detected, including flexible array members and flexible array
6052 member-like arrays. Initializers of variables with static storage are not
6055 @item -fsanitize=alignment
6056 @opindex fsanitize=alignment
6058 This option enables checking of alignment of pointers when they are
6059 dereferenced, or when a reference is bound to insufficiently aligned target,
6060 or when a method or constructor is invoked on insufficiently aligned object.
6062 @item -fsanitize=object-size
6063 @opindex fsanitize=object-size
6064 This option enables instrumentation of memory references using the
6065 @code{__builtin_object_size} function. Various out of bounds pointer
6066 accesses are detected.
6068 @item -fsanitize=float-divide-by-zero
6069 @opindex fsanitize=float-divide-by-zero
6070 Detect floating-point division by zero. Unlike other similar options,
6071 @option{-fsanitize=float-divide-by-zero} is not enabled by
6072 @option{-fsanitize=undefined}, since floating-point division by zero can
6073 be a legitimate way of obtaining infinities and NaNs.
6075 @item -fsanitize=float-cast-overflow
6076 @opindex fsanitize=float-cast-overflow
6077 This option enables floating-point type to integer conversion checking.
6078 We check that the result of the conversion does not overflow.
6079 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
6080 not enabled by @option{-fsanitize=undefined}.
6081 This option does not work well with @code{FE_INVALID} exceptions enabled.
6083 @item -fsanitize=nonnull-attribute
6084 @opindex fsanitize=nonnull-attribute
6086 This option enables instrumentation of calls, checking whether null values
6087 are not passed to arguments marked as requiring a non-null value by the
6088 @code{nonnull} function attribute.
6090 @item -fsanitize=returns-nonnull-attribute
6091 @opindex fsanitize=returns-nonnull-attribute
6093 This option enables instrumentation of return statements in functions
6094 marked with @code{returns_nonnull} function attribute, to detect returning
6095 of null values from such functions.
6097 @item -fsanitize=bool
6098 @opindex fsanitize=bool
6100 This option enables instrumentation of loads from bool. If a value other
6101 than 0/1 is loaded, a run-time error is issued.
6103 @item -fsanitize=enum
6104 @opindex fsanitize=enum
6106 This option enables instrumentation of loads from an enum type. If
6107 a value outside the range of values for the enum type is loaded,
6108 a run-time error is issued.
6110 @item -fsanitize=vptr
6111 @opindex fsanitize=vptr
6113 This option enables instrumentation of C++ member function calls, member
6114 accesses and some conversions between pointers to base and derived classes,
6115 to verify the referenced object has the correct dynamic type.
6119 While @option{-ftrapv} causes traps for signed overflows to be emitted,
6120 @option{-fsanitize=undefined} gives a diagnostic message.
6121 This currently works only for the C family of languages.
6123 @item -fno-sanitize=all
6124 @opindex fno-sanitize=all
6126 This option disables all previously enabled sanitizers.
6127 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
6130 @item -fasan-shadow-offset=@var{number}
6131 @opindex fasan-shadow-offset
6132 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
6133 It is useful for experimenting with different shadow memory layouts in
6134 Kernel AddressSanitizer.
6136 @item -fsanitize-sections=@var{s1},@var{s2},...
6137 @opindex fsanitize-sections
6138 Sanitize global variables in selected user-defined sections. @var{si} may
6141 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
6142 @opindex fsanitize-recover
6143 @opindex fno-sanitize-recover
6144 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
6145 mentioned in comma-separated list of @var{opts}. Enabling this option
6146 for a sanitizer component causes it to attempt to continue
6147 running the program as if no error happened. This means multiple
6148 runtime errors can be reported in a single program run, and the exit
6149 code of the program may indicate success even when errors
6150 have been reported. The @option{-fno-sanitize-recover=} option
6151 can be used to alter
6152 this behavior: only the first detected error is reported
6153 and program then exits with a non-zero exit code.
6155 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
6156 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
6157 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
6158 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
6159 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
6160 for which this feature is experimental.
6161 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
6162 accepted, the former enables recovery for all sanitizers that support it,
6163 the latter disables recovery for all sanitizers that support it.
6165 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
6167 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6170 Similarly @option{-fno-sanitize-recover} is equivalent to
6172 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6175 @item -fsanitize-undefined-trap-on-error
6176 @opindex fsanitize-undefined-trap-on-error
6177 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
6178 report undefined behavior using @code{__builtin_trap} rather than
6179 a @code{libubsan} library routine. The advantage of this is that the
6180 @code{libubsan} library is not needed and is not linked in, so this
6181 is usable even in freestanding environments.
6183 @item -fsanitize-coverage=trace-pc
6184 @opindex fsanitize-coverage=trace-pc
6185 Enable coverage-guided fuzzing code instrumentation.
6186 Inserts call to __sanitizer_cov_trace_pc into every basic block.
6188 @item -fcheck-pointer-bounds
6189 @opindex fcheck-pointer-bounds
6190 @opindex fno-check-pointer-bounds
6191 @cindex Pointer Bounds Checker options
6192 Enable Pointer Bounds Checker instrumentation. Each memory reference
6193 is instrumented with checks of the pointer used for memory access against
6194 bounds associated with that pointer.
6197 is only an implementation for Intel MPX available, thus x86 target
6198 and @option{-mmpx} are required to enable this feature.
6199 MPX-based instrumentation requires
6200 a runtime library to enable MPX in hardware and handle bounds
6201 violation signals. By default when @option{-fcheck-pointer-bounds}
6202 and @option{-mmpx} options are used to link a program, the GCC driver
6203 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
6204 library. It also passes '-z bndplt' to a linker in case it supports this
6205 option (which is checked on libmpx configuration). Note that old versions
6206 of linker may ignore option. Gold linker doesn't support '-z bndplt'
6207 option. With no '-z bndplt' support in linker all calls to dynamic libraries
6208 lose passed bounds reducing overall protection level. It's highly
6209 recommended to use linker with '-z bndplt' support. In case such linker
6210 is not available it is adviced to always use @option{-static-libmpxwrappers}
6211 for better protection level or use @option{-static} to completely avoid
6212 external calls to dynamic libraries. MPX-based instrumentation
6213 may be used for debugging and also may be included in production code
6214 to increase program security. Depending on usage, you may
6215 have different requirements for the runtime library. The current version
6216 of the MPX runtime library is more oriented for use as a debugging
6217 tool. MPX runtime library usage implies @option{-lpthread}. See
6218 also @option{-static-libmpx}. The runtime library behavior can be
6219 influenced using various @env{CHKP_RT_*} environment variables. See
6220 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
6223 Generated instrumentation may be controlled by various
6224 @option{-fchkp-*} options and by the @code{bnd_variable_size}
6225 structure field attribute (@pxref{Type Attributes}) and
6226 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
6227 (@pxref{Function Attributes}). GCC also provides a number of built-in
6228 functions for controlling the Pointer Bounds Checker. @xref{Pointer
6229 Bounds Checker builtins}, for more information.
6233 @opindex fno-checking
6234 Enable internal consistency checking. The default depends on
6235 the compiler configuration.
6237 @item -fchkp-check-incomplete-type
6238 @opindex fchkp-check-incomplete-type
6239 @opindex fno-chkp-check-incomplete-type
6240 Generate pointer bounds checks for variables with incomplete type.
6243 @item -fchkp-narrow-bounds
6244 @opindex fchkp-narrow-bounds
6245 @opindex fno-chkp-narrow-bounds
6246 Controls bounds used by Pointer Bounds Checker for pointers to object
6247 fields. If narrowing is enabled then field bounds are used. Otherwise
6248 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
6249 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
6251 @item -fchkp-first-field-has-own-bounds
6252 @opindex fchkp-first-field-has-own-bounds
6253 @opindex fno-chkp-first-field-has-own-bounds
6254 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
6255 first field in the structure. By default a pointer to the first field has
6256 the same bounds as a pointer to the whole structure.
6258 @item -fchkp-narrow-to-innermost-array
6259 @opindex fchkp-narrow-to-innermost-array
6260 @opindex fno-chkp-narrow-to-innermost-array
6261 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
6262 case of nested static array access. By default this option is disabled and
6263 bounds of the outermost array are used.
6265 @item -fchkp-optimize
6266 @opindex fchkp-optimize
6267 @opindex fno-chkp-optimize
6268 Enables Pointer Bounds Checker optimizations. Enabled by default at
6269 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
6271 @item -fchkp-use-fast-string-functions
6272 @opindex fchkp-use-fast-string-functions
6273 @opindex fno-chkp-use-fast-string-functions
6274 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
6275 by Pointer Bounds Checker. Disabled by default.
6277 @item -fchkp-use-nochk-string-functions
6278 @opindex fchkp-use-nochk-string-functions
6279 @opindex fno-chkp-use-nochk-string-functions
6280 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
6281 by Pointer Bounds Checker. Disabled by default.
6283 @item -fchkp-use-static-bounds
6284 @opindex fchkp-use-static-bounds
6285 @opindex fno-chkp-use-static-bounds
6286 Allow Pointer Bounds Checker to generate static bounds holding
6287 bounds of static variables. Enabled by default.
6289 @item -fchkp-use-static-const-bounds
6290 @opindex fchkp-use-static-const-bounds
6291 @opindex fno-chkp-use-static-const-bounds
6292 Use statically-initialized bounds for constant bounds instead of
6293 generating them each time they are required. By default enabled when
6294 @option{-fchkp-use-static-bounds} is enabled.
6296 @item -fchkp-treat-zero-dynamic-size-as-infinite
6297 @opindex fchkp-treat-zero-dynamic-size-as-infinite
6298 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
6299 With this option, objects with incomplete type whose
6300 dynamically-obtained size is zero are treated as having infinite size
6301 instead by Pointer Bounds
6302 Checker. This option may be helpful if a program is linked with a library
6303 missing size information for some symbols. Disabled by default.
6305 @item -fchkp-check-read
6306 @opindex fchkp-check-read
6307 @opindex fno-chkp-check-read
6308 Instructs Pointer Bounds Checker to generate checks for all read
6309 accesses to memory. Enabled by default.
6311 @item -fchkp-check-write
6312 @opindex fchkp-check-write
6313 @opindex fno-chkp-check-write
6314 Instructs Pointer Bounds Checker to generate checks for all write
6315 accesses to memory. Enabled by default.
6317 @item -fchkp-store-bounds
6318 @opindex fchkp-store-bounds
6319 @opindex fno-chkp-store-bounds
6320 Instructs Pointer Bounds Checker to generate bounds stores for
6321 pointer writes. Enabled by default.
6323 @item -fchkp-instrument-calls
6324 @opindex fchkp-instrument-calls
6325 @opindex fno-chkp-instrument-calls
6326 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
6329 @item -fchkp-instrument-marked-only
6330 @opindex fchkp-instrument-marked-only
6331 @opindex fno-chkp-instrument-marked-only
6332 Instructs Pointer Bounds Checker to instrument only functions
6333 marked with the @code{bnd_instrument} attribute
6334 (@pxref{Function Attributes}). Disabled by default.
6336 @item -fchkp-use-wrappers
6337 @opindex fchkp-use-wrappers
6338 @opindex fno-chkp-use-wrappers
6339 Allows Pointer Bounds Checker to replace calls to built-in functions
6340 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6341 is used to link a program, the GCC driver automatically links
6342 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6345 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6346 @opindex fdump-final-insns
6347 Dump the final internal representation (RTL) to @var{file}. If the
6348 optional argument is omitted (or if @var{file} is @code{.}), the name
6349 of the dump file is determined by appending @code{.gkd} to the
6350 compilation output file name.
6352 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6353 @opindex fcompare-debug
6354 @opindex fno-compare-debug
6355 If no error occurs during compilation, run the compiler a second time,
6356 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6357 passed to the second compilation. Dump the final internal
6358 representation in both compilations, and print an error if they differ.
6360 If the equal sign is omitted, the default @option{-gtoggle} is used.
6362 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6363 and nonzero, implicitly enables @option{-fcompare-debug}. If
6364 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6365 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6368 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6369 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6370 of the final representation and the second compilation, preventing even
6371 @env{GCC_COMPARE_DEBUG} from taking effect.
6373 To verify full coverage during @option{-fcompare-debug} testing, set
6374 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6375 which GCC rejects as an invalid option in any actual compilation
6376 (rather than preprocessing, assembly or linking). To get just a
6377 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6378 not overridden} will do.
6380 @item -fcompare-debug-second
6381 @opindex fcompare-debug-second
6382 This option is implicitly passed to the compiler for the second
6383 compilation requested by @option{-fcompare-debug}, along with options to
6384 silence warnings, and omitting other options that would cause
6385 side-effect compiler outputs to files or to the standard output. Dump
6386 files and preserved temporary files are renamed so as to contain the
6387 @code{.gk} additional extension during the second compilation, to avoid
6388 overwriting those generated by the first.
6390 When this option is passed to the compiler driver, it causes the
6391 @emph{first} compilation to be skipped, which makes it useful for little
6392 other than debugging the compiler proper.
6394 @item -feliminate-dwarf2-dups
6395 @opindex feliminate-dwarf2-dups
6396 Compress DWARF 2 debugging information by eliminating duplicated
6397 information about each symbol. This option only makes sense when
6398 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6400 @item -femit-struct-debug-baseonly
6401 @opindex femit-struct-debug-baseonly
6402 Emit debug information for struct-like types
6403 only when the base name of the compilation source file
6404 matches the base name of file in which the struct is defined.
6406 This option substantially reduces the size of debugging information,
6407 but at significant potential loss in type information to the debugger.
6408 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6409 See @option{-femit-struct-debug-detailed} for more detailed control.
6411 This option works only with DWARF 2.
6413 @item -femit-struct-debug-reduced
6414 @opindex femit-struct-debug-reduced
6415 Emit debug information for struct-like types
6416 only when the base name of the compilation source file
6417 matches the base name of file in which the type is defined,
6418 unless the struct is a template or defined in a system header.
6420 This option significantly reduces the size of debugging information,
6421 with some potential loss in type information to the debugger.
6422 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6423 See @option{-femit-struct-debug-detailed} for more detailed control.
6425 This option works only with DWARF 2.
6427 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6428 @opindex femit-struct-debug-detailed
6429 Specify the struct-like types
6430 for which the compiler generates debug information.
6431 The intent is to reduce duplicate struct debug information
6432 between different object files within the same program.
6434 This option is a detailed version of
6435 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6436 which serves for most needs.
6438 A specification has the syntax@*
6439 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6441 The optional first word limits the specification to
6442 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6443 A struct type is used directly when it is the type of a variable, member.
6444 Indirect uses arise through pointers to structs.
6445 That is, when use of an incomplete struct is valid, the use is indirect.
6447 @samp{struct one direct; struct two * indirect;}.
6449 The optional second word limits the specification to
6450 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6451 Generic structs are a bit complicated to explain.
6452 For C++, these are non-explicit specializations of template classes,
6453 or non-template classes within the above.
6454 Other programming languages have generics,
6455 but @option{-femit-struct-debug-detailed} does not yet implement them.
6457 The third word specifies the source files for those
6458 structs for which the compiler should emit debug information.
6459 The values @samp{none} and @samp{any} have the normal meaning.
6460 The value @samp{base} means that
6461 the base of name of the file in which the type declaration appears
6462 must match the base of the name of the main compilation file.
6463 In practice, this means that when compiling @file{foo.c}, debug information
6464 is generated for types declared in that file and @file{foo.h},
6465 but not other header files.
6466 The value @samp{sys} means those types satisfying @samp{base}
6467 or declared in system or compiler headers.
6469 You may need to experiment to determine the best settings for your application.
6471 The default is @option{-femit-struct-debug-detailed=all}.
6473 This option works only with DWARF 2.
6475 @item -fno-merge-debug-strings
6476 @opindex fmerge-debug-strings
6477 @opindex fno-merge-debug-strings
6478 Direct the linker to not merge together strings in the debugging
6479 information that are identical in different object files. Merging is
6480 not supported by all assemblers or linkers. Merging decreases the size
6481 of the debug information in the output file at the cost of increasing
6482 link processing time. Merging is enabled by default.
6484 @item -fdebug-prefix-map=@var{old}=@var{new}
6485 @opindex fdebug-prefix-map
6486 When compiling files in directory @file{@var{old}}, record debugging
6487 information describing them as in @file{@var{new}} instead.
6489 @item -fno-dwarf2-cfi-asm
6490 @opindex fdwarf2-cfi-asm
6491 @opindex fno-dwarf2-cfi-asm
6492 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6493 instead of using GAS @code{.cfi_*} directives.
6495 @cindex @command{prof}
6498 Generate extra code to write profile information suitable for the
6499 analysis program @command{prof}. You must use this option when compiling
6500 the source files you want data about, and you must also use it when
6503 @cindex @command{gprof}
6506 Generate extra code to write profile information suitable for the
6507 analysis program @command{gprof}. You must use this option when compiling
6508 the source files you want data about, and you must also use it when
6513 Makes the compiler print out each function name as it is compiled, and
6514 print some statistics about each pass when it finishes.
6517 @opindex ftime-report
6518 Makes the compiler print some statistics about the time consumed by each
6519 pass when it finishes.
6522 @opindex fmem-report
6523 Makes the compiler print some statistics about permanent memory
6524 allocation when it finishes.
6526 @item -fmem-report-wpa
6527 @opindex fmem-report-wpa
6528 Makes the compiler print some statistics about permanent memory
6529 allocation for the WPA phase only.
6531 @item -fpre-ipa-mem-report
6532 @opindex fpre-ipa-mem-report
6533 @item -fpost-ipa-mem-report
6534 @opindex fpost-ipa-mem-report
6535 Makes the compiler print some statistics about permanent memory
6536 allocation before or after interprocedural optimization.
6538 @item -fprofile-report
6539 @opindex fprofile-report
6540 Makes the compiler print some statistics about consistency of the
6541 (estimated) profile and effect of individual passes.
6544 @opindex fstack-usage
6545 Makes the compiler output stack usage information for the program, on a
6546 per-function basis. The filename for the dump is made by appending
6547 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6548 the output file, if explicitly specified and it is not an executable,
6549 otherwise it is the basename of the source file. An entry is made up
6554 The name of the function.
6558 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6561 The qualifier @code{static} means that the function manipulates the stack
6562 statically: a fixed number of bytes are allocated for the frame on function
6563 entry and released on function exit; no stack adjustments are otherwise made
6564 in the function. The second field is this fixed number of bytes.
6566 The qualifier @code{dynamic} means that the function manipulates the stack
6567 dynamically: in addition to the static allocation described above, stack
6568 adjustments are made in the body of the function, for example to push/pop
6569 arguments around function calls. If the qualifier @code{bounded} is also
6570 present, the amount of these adjustments is bounded at compile time and
6571 the second field is an upper bound of the total amount of stack used by
6572 the function. If it is not present, the amount of these adjustments is
6573 not bounded at compile time and the second field only represents the
6576 @item -fprofile-arcs
6577 @opindex fprofile-arcs
6578 Add code so that program flow @dfn{arcs} are instrumented. During
6579 execution the program records how many times each branch and call is
6580 executed and how many times it is taken or returns. When the compiled
6581 program exits it saves this data to a file called
6582 @file{@var{auxname}.gcda} for each source file. The data may be used for
6583 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6584 test coverage analysis (@option{-ftest-coverage}). Each object file's
6585 @var{auxname} is generated from the name of the output file, if
6586 explicitly specified and it is not the final executable, otherwise it is
6587 the basename of the source file. In both cases any suffix is removed
6588 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6589 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6590 @xref{Cross-profiling}.
6592 @cindex @command{gcov}
6596 This option is used to compile and link code instrumented for coverage
6597 analysis. The option is a synonym for @option{-fprofile-arcs}
6598 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6599 linking). See the documentation for those options for more details.
6604 Compile the source files with @option{-fprofile-arcs} plus optimization
6605 and code generation options. For test coverage analysis, use the
6606 additional @option{-ftest-coverage} option. You do not need to profile
6607 every source file in a program.
6610 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6611 (the latter implies the former).
6614 Run the program on a representative workload to generate the arc profile
6615 information. This may be repeated any number of times. You can run
6616 concurrent instances of your program, and provided that the file system
6617 supports locking, the data files will be correctly updated. Also
6618 @code{fork} calls are detected and correctly handled (double counting
6622 For profile-directed optimizations, compile the source files again with
6623 the same optimization and code generation options plus
6624 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6625 Control Optimization}).
6628 For test coverage analysis, use @command{gcov} to produce human readable
6629 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6630 @command{gcov} documentation for further information.
6634 With @option{-fprofile-arcs}, for each function of your program GCC
6635 creates a program flow graph, then finds a spanning tree for the graph.
6636 Only arcs that are not on the spanning tree have to be instrumented: the
6637 compiler adds code to count the number of times that these arcs are
6638 executed. When an arc is the only exit or only entrance to a block, the
6639 instrumentation code can be added to the block; otherwise, a new basic
6640 block must be created to hold the instrumentation code.
6643 @item -ftest-coverage
6644 @opindex ftest-coverage
6645 Produce a notes file that the @command{gcov} code-coverage utility
6646 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6647 show program coverage. Each source file's note file is called
6648 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6649 above for a description of @var{auxname} and instructions on how to
6650 generate test coverage data. Coverage data matches the source files
6651 more closely if you do not optimize.
6653 @item -fdbg-cnt-list
6654 @opindex fdbg-cnt-list
6655 Print the name and the counter upper bound for all debug counters.
6658 @item -fdbg-cnt=@var{counter-value-list}
6660 Set the internal debug counter upper bound. @var{counter-value-list}
6661 is a comma-separated list of @var{name}:@var{value} pairs
6662 which sets the upper bound of each debug counter @var{name} to @var{value}.
6663 All debug counters have the initial upper bound of @code{UINT_MAX};
6664 thus @code{dbg_cnt} returns true always unless the upper bound
6665 is set by this option.
6666 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6667 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6669 @item -fenable-@var{kind}-@var{pass}
6670 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6674 This is a set of options that are used to explicitly disable/enable
6675 optimization passes. These options are intended for use for debugging GCC.
6676 Compiler users should use regular options for enabling/disabling
6681 @item -fdisable-ipa-@var{pass}
6682 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6683 statically invoked in the compiler multiple times, the pass name should be
6684 appended with a sequential number starting from 1.
6686 @item -fdisable-rtl-@var{pass}
6687 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6688 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6689 statically invoked in the compiler multiple times, the pass name should be
6690 appended with a sequential number starting from 1. @var{range-list} is a
6691 comma-separated list of function ranges or assembler names. Each range is a number
6692 pair separated by a colon. The range is inclusive in both ends. If the range
6693 is trivial, the number pair can be simplified as a single number. If the
6694 function's call graph node's @var{uid} falls within one of the specified ranges,
6695 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6696 function header of a dump file, and the pass names can be dumped by using
6697 option @option{-fdump-passes}.
6699 @item -fdisable-tree-@var{pass}
6700 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6701 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6704 @item -fenable-ipa-@var{pass}
6705 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6706 statically invoked in the compiler multiple times, the pass name should be
6707 appended with a sequential number starting from 1.
6709 @item -fenable-rtl-@var{pass}
6710 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6711 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6712 description and examples.
6714 @item -fenable-tree-@var{pass}
6715 @itemx -fenable-tree-@var{pass}=@var{range-list}
6716 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6717 of option arguments.
6721 Here are some examples showing uses of these options.
6725 # disable ccp1 for all functions
6727 # disable complete unroll for function whose cgraph node uid is 1
6728 -fenable-tree-cunroll=1
6729 # disable gcse2 for functions at the following ranges [1,1],
6730 # [300,400], and [400,1000]
6731 # disable gcse2 for functions foo and foo2
6732 -fdisable-rtl-gcse2=foo,foo2
6733 # disable early inlining
6734 -fdisable-tree-einline
6735 # disable ipa inlining
6736 -fdisable-ipa-inline
6737 # enable tree full unroll
6738 -fenable-tree-unroll
6742 @item -d@var{letters}
6743 @itemx -fdump-rtl-@var{pass}
6744 @itemx -fdump-rtl-@var{pass}=@var{filename}
6746 @opindex fdump-rtl-@var{pass}
6747 Says to make debugging dumps during compilation at times specified by
6748 @var{letters}. This is used for debugging the RTL-based passes of the
6749 compiler. The file names for most of the dumps are made by appending
6750 a pass number and a word to the @var{dumpname}, and the files are
6751 created in the directory of the output file. In case of
6752 @option{=@var{filename}} option, the dump is output on the given file
6753 instead of the pass numbered dump files. Note that the pass number is
6754 assigned as passes are registered into the pass manager. Most passes
6755 are registered in the order that they will execute and for these passes
6756 the number corresponds to the pass execution order. However, passes
6757 registered by plugins, passes specific to compilation targets, or
6758 passes that are otherwise registered after all the other passes are
6759 numbered higher than a pass named "final", even if they are executed
6760 earlier. @var{dumpname} is generated from the name of the output
6761 file if explicitly specified and not an executable, otherwise it is
6762 the basename of the source file. These switches may have different
6763 effects when @option{-E} is used for preprocessing.
6765 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6766 @option{-d} option @var{letters}. Here are the possible
6767 letters for use in @var{pass} and @var{letters}, and their meanings:
6771 @item -fdump-rtl-alignments
6772 @opindex fdump-rtl-alignments
6773 Dump after branch alignments have been computed.
6775 @item -fdump-rtl-asmcons
6776 @opindex fdump-rtl-asmcons
6777 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6779 @item -fdump-rtl-auto_inc_dec
6780 @opindex fdump-rtl-auto_inc_dec
6781 Dump after auto-inc-dec discovery. This pass is only run on
6782 architectures that have auto inc or auto dec instructions.
6784 @item -fdump-rtl-barriers
6785 @opindex fdump-rtl-barriers
6786 Dump after cleaning up the barrier instructions.
6788 @item -fdump-rtl-bbpart
6789 @opindex fdump-rtl-bbpart
6790 Dump after partitioning hot and cold basic blocks.
6792 @item -fdump-rtl-bbro
6793 @opindex fdump-rtl-bbro
6794 Dump after block reordering.
6796 @item -fdump-rtl-btl1
6797 @itemx -fdump-rtl-btl2
6798 @opindex fdump-rtl-btl2
6799 @opindex fdump-rtl-btl2
6800 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6801 after the two branch
6802 target load optimization passes.
6804 @item -fdump-rtl-bypass
6805 @opindex fdump-rtl-bypass
6806 Dump after jump bypassing and control flow optimizations.
6808 @item -fdump-rtl-combine
6809 @opindex fdump-rtl-combine
6810 Dump after the RTL instruction combination pass.
6812 @item -fdump-rtl-compgotos
6813 @opindex fdump-rtl-compgotos
6814 Dump after duplicating the computed gotos.
6816 @item -fdump-rtl-ce1
6817 @itemx -fdump-rtl-ce2
6818 @itemx -fdump-rtl-ce3
6819 @opindex fdump-rtl-ce1
6820 @opindex fdump-rtl-ce2
6821 @opindex fdump-rtl-ce3
6822 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6823 @option{-fdump-rtl-ce3} enable dumping after the three
6824 if conversion passes.
6826 @item -fdump-rtl-cprop_hardreg
6827 @opindex fdump-rtl-cprop_hardreg
6828 Dump after hard register copy propagation.
6830 @item -fdump-rtl-csa
6831 @opindex fdump-rtl-csa
6832 Dump after combining stack adjustments.
6834 @item -fdump-rtl-cse1
6835 @itemx -fdump-rtl-cse2
6836 @opindex fdump-rtl-cse1
6837 @opindex fdump-rtl-cse2
6838 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6839 the two common subexpression elimination passes.
6841 @item -fdump-rtl-dce
6842 @opindex fdump-rtl-dce
6843 Dump after the standalone dead code elimination passes.
6845 @item -fdump-rtl-dbr
6846 @opindex fdump-rtl-dbr
6847 Dump after delayed branch scheduling.
6849 @item -fdump-rtl-dce1
6850 @itemx -fdump-rtl-dce2
6851 @opindex fdump-rtl-dce1
6852 @opindex fdump-rtl-dce2
6853 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6854 the two dead store elimination passes.
6857 @opindex fdump-rtl-eh
6858 Dump after finalization of EH handling code.
6860 @item -fdump-rtl-eh_ranges
6861 @opindex fdump-rtl-eh_ranges
6862 Dump after conversion of EH handling range regions.
6864 @item -fdump-rtl-expand
6865 @opindex fdump-rtl-expand
6866 Dump after RTL generation.
6868 @item -fdump-rtl-fwprop1
6869 @itemx -fdump-rtl-fwprop2
6870 @opindex fdump-rtl-fwprop1
6871 @opindex fdump-rtl-fwprop2
6872 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6873 dumping after the two forward propagation passes.
6875 @item -fdump-rtl-gcse1
6876 @itemx -fdump-rtl-gcse2
6877 @opindex fdump-rtl-gcse1
6878 @opindex fdump-rtl-gcse2
6879 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6880 after global common subexpression elimination.
6882 @item -fdump-rtl-init-regs
6883 @opindex fdump-rtl-init-regs
6884 Dump after the initialization of the registers.
6886 @item -fdump-rtl-initvals
6887 @opindex fdump-rtl-initvals
6888 Dump after the computation of the initial value sets.
6890 @item -fdump-rtl-into_cfglayout
6891 @opindex fdump-rtl-into_cfglayout
6892 Dump after converting to cfglayout mode.
6894 @item -fdump-rtl-ira
6895 @opindex fdump-rtl-ira
6896 Dump after iterated register allocation.
6898 @item -fdump-rtl-jump
6899 @opindex fdump-rtl-jump
6900 Dump after the second jump optimization.
6902 @item -fdump-rtl-loop2
6903 @opindex fdump-rtl-loop2
6904 @option{-fdump-rtl-loop2} enables dumping after the rtl
6905 loop optimization passes.
6907 @item -fdump-rtl-mach
6908 @opindex fdump-rtl-mach
6909 Dump after performing the machine dependent reorganization pass, if that
6912 @item -fdump-rtl-mode_sw
6913 @opindex fdump-rtl-mode_sw
6914 Dump after removing redundant mode switches.
6916 @item -fdump-rtl-rnreg
6917 @opindex fdump-rtl-rnreg
6918 Dump after register renumbering.
6920 @item -fdump-rtl-outof_cfglayout
6921 @opindex fdump-rtl-outof_cfglayout
6922 Dump after converting from cfglayout mode.
6924 @item -fdump-rtl-peephole2
6925 @opindex fdump-rtl-peephole2
6926 Dump after the peephole pass.
6928 @item -fdump-rtl-postreload
6929 @opindex fdump-rtl-postreload
6930 Dump after post-reload optimizations.
6932 @item -fdump-rtl-pro_and_epilogue
6933 @opindex fdump-rtl-pro_and_epilogue
6934 Dump after generating the function prologues and epilogues.
6936 @item -fdump-rtl-sched1
6937 @itemx -fdump-rtl-sched2
6938 @opindex fdump-rtl-sched1
6939 @opindex fdump-rtl-sched2
6940 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6941 after the basic block scheduling passes.
6943 @item -fdump-rtl-ree
6944 @opindex fdump-rtl-ree
6945 Dump after sign/zero extension elimination.
6947 @item -fdump-rtl-seqabstr
6948 @opindex fdump-rtl-seqabstr
6949 Dump after common sequence discovery.
6951 @item -fdump-rtl-shorten
6952 @opindex fdump-rtl-shorten
6953 Dump after shortening branches.
6955 @item -fdump-rtl-sibling
6956 @opindex fdump-rtl-sibling
6957 Dump after sibling call optimizations.
6959 @item -fdump-rtl-split1
6960 @itemx -fdump-rtl-split2
6961 @itemx -fdump-rtl-split3
6962 @itemx -fdump-rtl-split4
6963 @itemx -fdump-rtl-split5
6964 @opindex fdump-rtl-split1
6965 @opindex fdump-rtl-split2
6966 @opindex fdump-rtl-split3
6967 @opindex fdump-rtl-split4
6968 @opindex fdump-rtl-split5
6969 These options enable dumping after five rounds of
6970 instruction splitting.
6972 @item -fdump-rtl-sms
6973 @opindex fdump-rtl-sms
6974 Dump after modulo scheduling. This pass is only run on some
6977 @item -fdump-rtl-stack
6978 @opindex fdump-rtl-stack
6979 Dump after conversion from GCC's ``flat register file'' registers to the
6980 x87's stack-like registers. This pass is only run on x86 variants.
6982 @item -fdump-rtl-subreg1
6983 @itemx -fdump-rtl-subreg2
6984 @opindex fdump-rtl-subreg1
6985 @opindex fdump-rtl-subreg2
6986 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6987 the two subreg expansion passes.
6989 @item -fdump-rtl-unshare
6990 @opindex fdump-rtl-unshare
6991 Dump after all rtl has been unshared.
6993 @item -fdump-rtl-vartrack
6994 @opindex fdump-rtl-vartrack
6995 Dump after variable tracking.
6997 @item -fdump-rtl-vregs
6998 @opindex fdump-rtl-vregs
6999 Dump after converting virtual registers to hard registers.
7001 @item -fdump-rtl-web
7002 @opindex fdump-rtl-web
7003 Dump after live range splitting.
7005 @item -fdump-rtl-regclass
7006 @itemx -fdump-rtl-subregs_of_mode_init
7007 @itemx -fdump-rtl-subregs_of_mode_finish
7008 @itemx -fdump-rtl-dfinit
7009 @itemx -fdump-rtl-dfinish
7010 @opindex fdump-rtl-regclass
7011 @opindex fdump-rtl-subregs_of_mode_init
7012 @opindex fdump-rtl-subregs_of_mode_finish
7013 @opindex fdump-rtl-dfinit
7014 @opindex fdump-rtl-dfinish
7015 These dumps are defined but always produce empty files.
7018 @itemx -fdump-rtl-all
7020 @opindex fdump-rtl-all
7021 Produce all the dumps listed above.
7025 Annotate the assembler output with miscellaneous debugging information.
7029 Dump all macro definitions, at the end of preprocessing, in addition to
7034 Produce a core dump whenever an error occurs.
7038 Annotate the assembler output with a comment indicating which
7039 pattern and alternative is used. The length of each instruction is
7044 Dump the RTL in the assembler output as a comment before each instruction.
7045 Also turns on @option{-dp} annotation.
7049 Just generate RTL for a function instead of compiling it. Usually used
7050 with @option{-fdump-rtl-expand}.
7054 @opindex fdump-noaddr
7055 When doing debugging dumps, suppress address output. This makes it more
7056 feasible to use diff on debugging dumps for compiler invocations with
7057 different compiler binaries and/or different
7058 text / bss / data / heap / stack / dso start locations.
7061 @opindex freport-bug
7062 Collect and dump debug information into temporary file if ICE in C/C++
7065 @item -fdump-unnumbered
7066 @opindex fdump-unnumbered
7067 When doing debugging dumps, suppress instruction numbers and address output.
7068 This makes it more feasible to use diff on debugging dumps for compiler
7069 invocations with different options, in particular with and without
7072 @item -fdump-unnumbered-links
7073 @opindex fdump-unnumbered-links
7074 When doing debugging dumps (see @option{-d} option above), suppress
7075 instruction numbers for the links to the previous and next instructions
7078 @item -fdump-translation-unit @r{(C++ only)}
7079 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
7080 @opindex fdump-translation-unit
7081 Dump a representation of the tree structure for the entire translation
7082 unit to a file. The file name is made by appending @file{.tu} to the
7083 source file name, and the file is created in the same directory as the
7084 output file. If the @samp{-@var{options}} form is used, @var{options}
7085 controls the details of the dump as described for the
7086 @option{-fdump-tree} options.
7088 @item -fdump-class-hierarchy @r{(C++ only)}
7089 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
7090 @opindex fdump-class-hierarchy
7091 Dump a representation of each class's hierarchy and virtual function
7092 table layout to a file. The file name is made by appending
7093 @file{.class} to the source file name, and the file is created in the
7094 same directory as the output file. If the @samp{-@var{options}} form
7095 is used, @var{options} controls the details of the dump as described
7096 for the @option{-fdump-tree} options.
7098 @item -fdump-ipa-@var{switch}
7100 Control the dumping at various stages of inter-procedural analysis
7101 language tree to a file. The file name is generated by appending a
7102 switch specific suffix to the source file name, and the file is created
7103 in the same directory as the output file. The following dumps are
7108 Enables all inter-procedural analysis dumps.
7111 Dumps information about call-graph optimization, unused function removal,
7112 and inlining decisions.
7115 Dump after function inlining.
7120 @opindex fdump-passes
7121 Dump the list of optimization passes that are turned on and off by
7122 the current command-line options.
7124 @item -fdump-statistics-@var{option}
7125 @opindex fdump-statistics
7126 Enable and control dumping of pass statistics in a separate file. The
7127 file name is generated by appending a suffix ending in
7128 @samp{.statistics} to the source file name, and the file is created in
7129 the same directory as the output file. If the @samp{-@var{option}}
7130 form is used, @samp{-stats} causes counters to be summed over the
7131 whole compilation unit while @samp{-details} dumps every event as
7132 the passes generate them. The default with no option is to sum
7133 counters for each function compiled.
7135 @item -fdump-tree-@var{switch}
7136 @itemx -fdump-tree-@var{switch}-@var{options}
7137 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
7139 Control the dumping at various stages of processing the intermediate
7140 language tree to a file. The file name is generated by appending a
7141 switch-specific suffix to the source file name, and the file is
7142 created in the same directory as the output file. In case of
7143 @option{=@var{filename}} option, the dump is output on the given file
7144 instead of the auto named dump files. If the @samp{-@var{options}}
7145 form is used, @var{options} is a list of @samp{-} separated options
7146 which control the details of the dump. Not all options are applicable
7147 to all dumps; those that are not meaningful are ignored. The
7148 following options are available
7152 Print the address of each node. Usually this is not meaningful as it
7153 changes according to the environment and source file. Its primary use
7154 is for tying up a dump file with a debug environment.
7156 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
7157 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
7158 use working backward from mangled names in the assembly file.
7160 When dumping front-end intermediate representations, inhibit dumping
7161 of members of a scope or body of a function merely because that scope
7162 has been reached. Only dump such items when they are directly reachable
7165 When dumping pretty-printed trees, this option inhibits dumping the
7166 bodies of control structures.
7168 When dumping RTL, print the RTL in slim (condensed) form instead of
7169 the default LISP-like representation.
7171 Print a raw representation of the tree. By default, trees are
7172 pretty-printed into a C-like representation.
7174 Enable more detailed dumps (not honored by every dump option). Also
7175 include information from the optimization passes.
7177 Enable dumping various statistics about the pass (not honored by every dump
7180 Enable showing basic block boundaries (disabled in raw dumps).
7182 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
7183 dump a representation of the control flow graph suitable for viewing with
7184 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
7185 the file is pretty-printed as a subgraph, so that GraphViz can render them
7186 all in a single plot.
7188 This option currently only works for RTL dumps, and the RTL is always
7189 dumped in slim form.
7191 Enable showing virtual operands for every statement.
7193 Enable showing line numbers for statements.
7195 Enable showing the unique ID (@code{DECL_UID}) for each variable.
7197 Enable showing the tree dump for each statement.
7199 Enable showing the EH region number holding each statement.
7201 Enable showing scalar evolution analysis details.
7203 Enable showing optimization information (only available in certain
7206 Enable showing missed optimization information (only available in certain
7209 Enable other detailed optimization information (only available in
7211 @item =@var{filename}
7212 Instead of an auto named dump file, output into the given file
7213 name. The file names @file{stdout} and @file{stderr} are treated
7214 specially and are considered already open standard streams. For
7218 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
7219 -fdump-tree-pre=stderr file.c
7222 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
7223 output on to @file{stderr}. If two conflicting dump filenames are
7224 given for the same pass, then the latter option overrides the earlier
7228 @opindex fdump-tree-split-paths
7229 Dump each function after splitting paths to loop backedges. The file
7230 name is made by appending @file{.split-paths} to the source file name.
7233 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
7234 and @option{lineno}.
7237 Turn on all optimization options, i.e., @option{optimized},
7238 @option{missed}, and @option{note}.
7241 The following tree dumps are possible:
7245 @opindex fdump-tree-original
7246 Dump before any tree based optimization, to @file{@var{file}.original}.
7249 @opindex fdump-tree-optimized
7250 Dump after all tree based optimization, to @file{@var{file}.optimized}.
7253 @opindex fdump-tree-gimple
7254 Dump each function before and after the gimplification pass to a file. The
7255 file name is made by appending @file{.gimple} to the source file name.
7258 @opindex fdump-tree-cfg
7259 Dump the control flow graph of each function to a file. The file name is
7260 made by appending @file{.cfg} to the source file name.
7263 @opindex fdump-tree-ch
7264 Dump each function after copying loop headers. The file name is made by
7265 appending @file{.ch} to the source file name.
7268 @opindex fdump-tree-ssa
7269 Dump SSA related information to a file. The file name is made by appending
7270 @file{.ssa} to the source file name.
7273 @opindex fdump-tree-alias
7274 Dump aliasing information for each function. The file name is made by
7275 appending @file{.alias} to the source file name.
7278 @opindex fdump-tree-ccp
7279 Dump each function after CCP@. The file name is made by appending
7280 @file{.ccp} to the source file name.
7283 @opindex fdump-tree-storeccp
7284 Dump each function after STORE-CCP@. The file name is made by appending
7285 @file{.storeccp} to the source file name.
7288 @opindex fdump-tree-pre
7289 Dump trees after partial redundancy elimination. The file name is made
7290 by appending @file{.pre} to the source file name.
7293 @opindex fdump-tree-fre
7294 Dump trees after full redundancy elimination. The file name is made
7295 by appending @file{.fre} to the source file name.
7298 @opindex fdump-tree-copyprop
7299 Dump trees after copy propagation. The file name is made
7300 by appending @file{.copyprop} to the source file name.
7302 @item store_copyprop
7303 @opindex fdump-tree-store_copyprop
7304 Dump trees after store copy-propagation. The file name is made
7305 by appending @file{.store_copyprop} to the source file name.
7308 @opindex fdump-tree-dce
7309 Dump each function after dead code elimination. The file name is made by
7310 appending @file{.dce} to the source file name.
7313 @opindex fdump-tree-sra
7314 Dump each function after performing scalar replacement of aggregates. The
7315 file name is made by appending @file{.sra} to the source file name.
7318 @opindex fdump-tree-sink
7319 Dump each function after performing code sinking. The file name is made
7320 by appending @file{.sink} to the source file name.
7323 @opindex fdump-tree-dom
7324 Dump each function after applying dominator tree optimizations. The file
7325 name is made by appending @file{.dom} to the source file name.
7328 @opindex fdump-tree-dse
7329 Dump each function after applying dead store elimination. The file
7330 name is made by appending @file{.dse} to the source file name.
7333 @opindex fdump-tree-phiopt
7334 Dump each function after optimizing PHI nodes into straightline code. The file
7335 name is made by appending @file{.phiopt} to the source file name.
7338 @opindex fdump-tree-backprop
7339 Dump each function after back-propagating use information up the definition
7340 chain. The file name is made by appending @file{.backprop} to the
7344 @opindex fdump-tree-forwprop
7345 Dump each function after forward propagating single use variables. The file
7346 name is made by appending @file{.forwprop} to the source file name.
7349 @opindex fdump-tree-nrv
7350 Dump each function after applying the named return value optimization on
7351 generic trees. The file name is made by appending @file{.nrv} to the source
7355 @opindex fdump-tree-vect
7356 Dump each function after applying vectorization of loops. The file name is
7357 made by appending @file{.vect} to the source file name.
7360 @opindex fdump-tree-slp
7361 Dump each function after applying vectorization of basic blocks. The file name
7362 is made by appending @file{.slp} to the source file name.
7365 @opindex fdump-tree-vrp
7366 Dump each function after Value Range Propagation (VRP). The file name
7367 is made by appending @file{.vrp} to the source file name.
7370 @opindex fdump-tree-oaccdevlow
7371 Dump each function after applying device-specific OpenACC transformations.
7372 The file name is made by appending @file{.oaccdevlow} to the source file name.
7375 @opindex fdump-tree-all
7376 Enable all the available tree dumps with the flags provided in this option.
7380 @itemx -fopt-info-@var{options}
7381 @itemx -fopt-info-@var{options}=@var{filename}
7383 Controls optimization dumps from various optimization passes. If the
7384 @samp{-@var{options}} form is used, @var{options} is a list of
7385 @samp{-} separated option keywords to select the dump details and
7388 The @var{options} can be divided into two groups: options describing the
7389 verbosity of the dump, and options describing which optimizations
7390 should be included. The options from both the groups can be freely
7391 mixed as they are non-overlapping. However, in case of any conflicts,
7392 the later options override the earlier options on the command
7395 The following options control the dump verbosity:
7399 Print information when an optimization is successfully applied. It is
7400 up to a pass to decide which information is relevant. For example, the
7401 vectorizer passes print the source location of loops which are
7402 successfully vectorized.
7404 Print information about missed optimizations. Individual passes
7405 control which information to include in the output.
7407 Print verbose information about optimizations, such as certain
7408 transformations, more detailed messages about decisions etc.
7410 Print detailed optimization information. This includes
7411 @samp{optimized}, @samp{missed}, and @samp{note}.
7414 One or more of the following option keywords can be used to describe a
7415 group of optimizations:
7419 Enable dumps from all interprocedural optimizations.
7421 Enable dumps from all loop optimizations.
7423 Enable dumps from all inlining optimizations.
7425 Enable dumps from all vectorization optimizations.
7427 Enable dumps from all optimizations. This is a superset of
7428 the optimization groups listed above.
7432 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7433 info about successful optimizations from all the passes.
7435 If the @var{filename} is provided, then the dumps from all the
7436 applicable optimizations are concatenated into the @var{filename}.
7437 Otherwise the dump is output onto @file{stderr}. Though multiple
7438 @option{-fopt-info} options are accepted, only one of them can include
7439 a @var{filename}. If other filenames are provided then all but the
7440 first such option are ignored.
7442 Note that the output @var{filename} is overwritten
7443 in case of multiple translation units. If a combined output from
7444 multiple translation units is desired, @file{stderr} should be used
7447 In the following example, the optimization info is output to
7456 gcc -O3 -fopt-info-missed=missed.all
7460 outputs missed optimization report from all the passes into
7461 @file{missed.all}, and this one:
7464 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7468 prints information about missed optimization opportunities from
7469 vectorization passes on @file{stderr}.
7470 Note that @option{-fopt-info-vec-missed} is equivalent to
7471 @option{-fopt-info-missed-vec}.
7475 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7479 outputs information about missed optimizations as well as
7480 optimized locations from all the inlining passes into
7486 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7490 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7491 in conflict since only one output file is allowed. In this case, only
7492 the first option takes effect and the subsequent options are
7493 ignored. Thus only @file{vec.miss} is produced which contains
7494 dumps from the vectorizer about missed opportunities.
7496 @item -frandom-seed=@var{string}
7497 @opindex frandom-seed
7498 This option provides a seed that GCC uses in place of
7499 random numbers in generating certain symbol names
7500 that have to be different in every compiled file. It is also used to
7501 place unique stamps in coverage data files and the object files that
7502 produce them. You can use the @option{-frandom-seed} option to produce
7503 reproducibly identical object files.
7505 The @var{string} can either be a number (decimal, octal or hex) or an
7506 arbitrary string (in which case it's converted to a number by
7509 The @var{string} should be different for every file you compile.
7511 @item -fsched-verbose=@var{n}
7512 @opindex fsched-verbose
7513 On targets that use instruction scheduling, this option controls the
7514 amount of debugging output the scheduler prints to the dump files.
7516 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7517 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7518 For @var{n} greater than one, it also output basic block probabilities,
7519 detailed ready list information and unit/insn info. For @var{n} greater
7520 than two, it includes RTL at abort point, control-flow and regions info.
7521 And for @var{n} over four, @option{-fsched-verbose} also includes
7525 @itemx -save-temps=cwd
7527 Store the usual ``temporary'' intermediate files permanently; place them
7528 in the current directory and name them based on the source file. Thus,
7529 compiling @file{foo.c} with @option{-c -save-temps} produces files
7530 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7531 preprocessed @file{foo.i} output file even though the compiler now
7532 normally uses an integrated preprocessor.
7534 When used in combination with the @option{-x} command-line option,
7535 @option{-save-temps} is sensible enough to avoid over writing an
7536 input source file with the same extension as an intermediate file.
7537 The corresponding intermediate file may be obtained by renaming the
7538 source file before using @option{-save-temps}.
7540 If you invoke GCC in parallel, compiling several different source
7541 files that share a common base name in different subdirectories or the
7542 same source file compiled for multiple output destinations, it is
7543 likely that the different parallel compilers will interfere with each
7544 other, and overwrite the temporary files. For instance:
7547 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7548 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7551 may result in @file{foo.i} and @file{foo.o} being written to
7552 simultaneously by both compilers.
7554 @item -save-temps=obj
7555 @opindex save-temps=obj
7556 Store the usual ``temporary'' intermediate files permanently. If the
7557 @option{-o} option is used, the temporary files are based on the
7558 object file. If the @option{-o} option is not used, the
7559 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7564 gcc -save-temps=obj -c foo.c
7565 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7566 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7570 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7571 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7572 @file{dir2/yfoobar.o}.
7574 @item -time@r{[}=@var{file}@r{]}
7576 Report the CPU time taken by each subprocess in the compilation
7577 sequence. For C source files, this is the compiler proper and assembler
7578 (plus the linker if linking is done).
7580 Without the specification of an output file, the output looks like this:
7587 The first number on each line is the ``user time'', that is time spent
7588 executing the program itself. The second number is ``system time'',
7589 time spent executing operating system routines on behalf of the program.
7590 Both numbers are in seconds.
7592 With the specification of an output file, the output is appended to the
7593 named file, and it looks like this:
7596 0.12 0.01 cc1 @var{options}
7597 0.00 0.01 as @var{options}
7600 The ``user time'' and the ``system time'' are moved before the program
7601 name, and the options passed to the program are displayed, so that one
7602 can later tell what file was being compiled, and with which options.
7604 @item -fvar-tracking
7605 @opindex fvar-tracking
7606 Run variable tracking pass. It computes where variables are stored at each
7607 position in code. Better debugging information is then generated
7608 (if the debugging information format supports this information).
7610 It is enabled by default when compiling with optimization (@option{-Os},
7611 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7612 the debug info format supports it.
7614 @item -fvar-tracking-assignments
7615 @opindex fvar-tracking-assignments
7616 @opindex fno-var-tracking-assignments
7617 Annotate assignments to user variables early in the compilation and
7618 attempt to carry the annotations over throughout the compilation all the
7619 way to the end, in an attempt to improve debug information while
7620 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7622 It can be enabled even if var-tracking is disabled, in which case
7623 annotations are created and maintained, but discarded at the end.
7624 By default, this flag is enabled together with @option{-fvar-tracking},
7625 except when selective scheduling is enabled.
7627 @item -fvar-tracking-assignments-toggle
7628 @opindex fvar-tracking-assignments-toggle
7629 @opindex fno-var-tracking-assignments-toggle
7630 Toggle @option{-fvar-tracking-assignments}, in the same way that
7631 @option{-gtoggle} toggles @option{-g}.
7633 @item -print-file-name=@var{library}
7634 @opindex print-file-name
7635 Print the full absolute name of the library file @var{library} that
7636 would be used when linking---and don't do anything else. With this
7637 option, GCC does not compile or link anything; it just prints the
7640 @item -print-multi-directory
7641 @opindex print-multi-directory
7642 Print the directory name corresponding to the multilib selected by any
7643 other switches present in the command line. This directory is supposed
7644 to exist in @env{GCC_EXEC_PREFIX}.
7646 @item -print-multi-lib
7647 @opindex print-multi-lib
7648 Print the mapping from multilib directory names to compiler switches
7649 that enable them. The directory name is separated from the switches by
7650 @samp{;}, and each switch starts with an @samp{@@} instead of the
7651 @samp{-}, without spaces between multiple switches. This is supposed to
7652 ease shell processing.
7654 @item -print-multi-os-directory
7655 @opindex print-multi-os-directory
7656 Print the path to OS libraries for the selected
7657 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7658 present in the @file{lib} subdirectory and no multilibs are used, this is
7659 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7660 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7661 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7662 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7664 @item -print-multiarch
7665 @opindex print-multiarch
7666 Print the path to OS libraries for the selected multiarch,
7667 relative to some @file{lib} subdirectory.
7669 @item -print-prog-name=@var{program}
7670 @opindex print-prog-name
7671 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7673 @item -print-libgcc-file-name
7674 @opindex print-libgcc-file-name
7675 Same as @option{-print-file-name=libgcc.a}.
7677 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7678 but you do want to link with @file{libgcc.a}. You can do:
7681 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7684 @item -print-search-dirs
7685 @opindex print-search-dirs
7686 Print the name of the configured installation directory and a list of
7687 program and library directories @command{gcc} searches---and don't do anything else.
7689 This is useful when @command{gcc} prints the error message
7690 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7691 To resolve this you either need to put @file{cpp0} and the other compiler
7692 components where @command{gcc} expects to find them, or you can set the environment
7693 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7694 Don't forget the trailing @samp{/}.
7695 @xref{Environment Variables}.
7697 @item -print-sysroot
7698 @opindex print-sysroot
7699 Print the target sysroot directory that is used during
7700 compilation. This is the target sysroot specified either at configure
7701 time or using the @option{--sysroot} option, possibly with an extra
7702 suffix that depends on compilation options. If no target sysroot is
7703 specified, the option prints nothing.
7705 @item -print-sysroot-headers-suffix
7706 @opindex print-sysroot-headers-suffix
7707 Print the suffix added to the target sysroot when searching for
7708 headers, or give an error if the compiler is not configured with such
7709 a suffix---and don't do anything else.
7712 @opindex dumpmachine
7713 Print the compiler's target machine (for example,
7714 @samp{i686-pc-linux-gnu})---and don't do anything else.
7717 @opindex dumpversion
7718 Print the compiler version (for example, @code{3.0})---and don't do
7723 Print the compiler's built-in specs---and don't do anything else. (This
7724 is used when GCC itself is being built.) @xref{Spec Files}.
7726 @item -fno-eliminate-unused-debug-types
7727 @opindex feliminate-unused-debug-types
7728 @opindex fno-eliminate-unused-debug-types
7729 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7730 output for types that are nowhere used in the source file being compiled.
7731 Sometimes it is useful to have GCC emit debugging
7732 information for all types declared in a compilation
7733 unit, regardless of whether or not they are actually used
7734 in that compilation unit, for example
7735 if, in the debugger, you want to cast a value to a type that is
7736 not actually used in your program (but is declared). More often,
7737 however, this results in a significant amount of wasted space.
7740 @node Optimize Options
7741 @section Options That Control Optimization
7742 @cindex optimize options
7743 @cindex options, optimization
7745 These options control various sorts of optimizations.
7747 Without any optimization option, the compiler's goal is to reduce the
7748 cost of compilation and to make debugging produce the expected
7749 results. Statements are independent: if you stop the program with a
7750 breakpoint between statements, you can then assign a new value to any
7751 variable or change the program counter to any other statement in the
7752 function and get exactly the results you expect from the source
7755 Turning on optimization flags makes the compiler attempt to improve
7756 the performance and/or code size at the expense of compilation time
7757 and possibly the ability to debug the program.
7759 The compiler performs optimization based on the knowledge it has of the
7760 program. Compiling multiple files at once to a single output file mode allows
7761 the compiler to use information gained from all of the files when compiling
7764 Not all optimizations are controlled directly by a flag. Only
7765 optimizations that have a flag are listed in this section.
7767 Most optimizations are only enabled if an @option{-O} level is set on
7768 the command line. Otherwise they are disabled, even if individual
7769 optimization flags are specified.
7771 Depending on the target and how GCC was configured, a slightly different
7772 set of optimizations may be enabled at each @option{-O} level than
7773 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7774 to find out the exact set of optimizations that are enabled at each level.
7775 @xref{Overall Options}, for examples.
7782 Optimize. Optimizing compilation takes somewhat more time, and a lot
7783 more memory for a large function.
7785 With @option{-O}, the compiler tries to reduce code size and execution
7786 time, without performing any optimizations that take a great deal of
7789 @option{-O} turns on the following optimization flags:
7792 -fbranch-count-reg @gol
7793 -fcombine-stack-adjustments @gol
7795 -fcprop-registers @gol
7798 -fdelayed-branch @gol
7800 -fforward-propagate @gol
7801 -fguess-branch-probability @gol
7802 -fif-conversion2 @gol
7803 -fif-conversion @gol
7804 -finline-functions-called-once @gol
7805 -fipa-pure-const @gol
7807 -fipa-reference @gol
7808 -fmerge-constants @gol
7809 -fmove-loop-invariants @gol
7810 -freorder-blocks @gol
7812 -fsplit-wide-types @gol
7818 -ftree-coalesce-vars @gol
7819 -ftree-copy-prop @gol
7821 -ftree-dominator-opts @gol
7823 -ftree-forwprop @gol
7833 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7834 where doing so does not interfere with debugging.
7838 Optimize even more. GCC performs nearly all supported optimizations
7839 that do not involve a space-speed tradeoff.
7840 As compared to @option{-O}, this option increases both compilation time
7841 and the performance of the generated code.
7843 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7844 also turns on the following optimization flags:
7845 @gccoptlist{-fthread-jumps @gol
7846 -falign-functions -falign-jumps @gol
7847 -falign-loops -falign-labels @gol
7850 -fcse-follow-jumps -fcse-skip-blocks @gol
7851 -fdelete-null-pointer-checks @gol
7852 -fdevirtualize -fdevirtualize-speculatively @gol
7853 -fexpensive-optimizations @gol
7854 -fgcse -fgcse-lm @gol
7855 -fhoist-adjacent-loads @gol
7856 -finline-small-functions @gol
7857 -findirect-inlining @gol
7859 -fipa-cp-alignment @gol
7862 -fisolate-erroneous-paths-dereference @gol
7864 -foptimize-sibling-calls @gol
7865 -foptimize-strlen @gol
7866 -fpartial-inlining @gol
7868 -freorder-blocks-algorithm=stc @gol
7869 -freorder-blocks-and-partition -freorder-functions @gol
7870 -frerun-cse-after-loop @gol
7871 -fsched-interblock -fsched-spec @gol
7872 -fschedule-insns -fschedule-insns2 @gol
7873 -fstrict-aliasing -fstrict-overflow @gol
7874 -ftree-builtin-call-dce @gol
7875 -ftree-switch-conversion -ftree-tail-merge @gol
7880 Please note the warning under @option{-fgcse} about
7881 invoking @option{-O2} on programs that use computed gotos.
7885 Optimize yet more. @option{-O3} turns on all optimizations specified
7886 by @option{-O2} and also turns on the @option{-finline-functions},
7887 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7888 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7889 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
7890 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7891 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7895 Reduce compilation time and make debugging produce the expected
7896 results. This is the default.
7900 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7901 do not typically increase code size. It also performs further
7902 optimizations designed to reduce code size.
7904 @option{-Os} disables the following optimization flags:
7905 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7906 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7907 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7911 Disregard strict standards compliance. @option{-Ofast} enables all
7912 @option{-O3} optimizations. It also enables optimizations that are not
7913 valid for all standard-compliant programs.
7914 It turns on @option{-ffast-math} and the Fortran-specific
7915 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7919 Optimize debugging experience. @option{-Og} enables optimizations
7920 that do not interfere with debugging. It should be the optimization
7921 level of choice for the standard edit-compile-debug cycle, offering
7922 a reasonable level of optimization while maintaining fast compilation
7923 and a good debugging experience.
7926 If you use multiple @option{-O} options, with or without level numbers,
7927 the last such option is the one that is effective.
7929 Options of the form @option{-f@var{flag}} specify machine-independent
7930 flags. Most flags have both positive and negative forms; the negative
7931 form of @option{-ffoo} is @option{-fno-foo}. In the table
7932 below, only one of the forms is listed---the one you typically
7933 use. You can figure out the other form by either removing @samp{no-}
7936 The following options control specific optimizations. They are either
7937 activated by @option{-O} options or are related to ones that are. You
7938 can use the following flags in the rare cases when ``fine-tuning'' of
7939 optimizations to be performed is desired.
7942 @item -fno-defer-pop
7943 @opindex fno-defer-pop
7944 Always pop the arguments to each function call as soon as that function
7945 returns. For machines that must pop arguments after a function call,
7946 the compiler normally lets arguments accumulate on the stack for several
7947 function calls and pops them all at once.
7949 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7951 @item -fforward-propagate
7952 @opindex fforward-propagate
7953 Perform a forward propagation pass on RTL@. The pass tries to combine two
7954 instructions and checks if the result can be simplified. If loop unrolling
7955 is active, two passes are performed and the second is scheduled after
7958 This option is enabled by default at optimization levels @option{-O},
7959 @option{-O2}, @option{-O3}, @option{-Os}.
7961 @item -ffp-contract=@var{style}
7962 @opindex ffp-contract
7963 @option{-ffp-contract=off} disables floating-point expression contraction.
7964 @option{-ffp-contract=fast} enables floating-point expression contraction
7965 such as forming of fused multiply-add operations if the target has
7966 native support for them.
7967 @option{-ffp-contract=on} enables floating-point expression contraction
7968 if allowed by the language standard. This is currently not implemented
7969 and treated equal to @option{-ffp-contract=off}.
7971 The default is @option{-ffp-contract=fast}.
7973 @item -fomit-frame-pointer
7974 @opindex fomit-frame-pointer
7975 Don't keep the frame pointer in a register for functions that
7976 don't need one. This avoids the instructions to save, set up and
7977 restore frame pointers; it also makes an extra register available
7978 in many functions. @strong{It also makes debugging impossible on
7981 On some machines, such as the VAX, this flag has no effect, because
7982 the standard calling sequence automatically handles the frame pointer
7983 and nothing is saved by pretending it doesn't exist. The
7984 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7985 whether a target machine supports this flag. @xref{Registers,,Register
7986 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7988 The default setting (when not optimizing for
7989 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7990 @option{-fomit-frame-pointer}. You can configure GCC with the
7991 @option{--enable-frame-pointer} configure option to change the default.
7993 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7995 @item -foptimize-sibling-calls
7996 @opindex foptimize-sibling-calls
7997 Optimize sibling and tail recursive calls.
7999 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8001 @item -foptimize-strlen
8002 @opindex foptimize-strlen
8003 Optimize various standard C string functions (e.g. @code{strlen},
8004 @code{strchr} or @code{strcpy}) and
8005 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
8007 Enabled at levels @option{-O2}, @option{-O3}.
8011 Do not expand any functions inline apart from those marked with
8012 the @code{always_inline} attribute. This is the default when not
8015 Single functions can be exempted from inlining by marking them
8016 with the @code{noinline} attribute.
8018 @item -finline-small-functions
8019 @opindex finline-small-functions
8020 Integrate functions into their callers when their body is smaller than expected
8021 function call code (so overall size of program gets smaller). The compiler
8022 heuristically decides which functions are simple enough to be worth integrating
8023 in this way. This inlining applies to all functions, even those not declared
8026 Enabled at level @option{-O2}.
8028 @item -findirect-inlining
8029 @opindex findirect-inlining
8030 Inline also indirect calls that are discovered to be known at compile
8031 time thanks to previous inlining. This option has any effect only
8032 when inlining itself is turned on by the @option{-finline-functions}
8033 or @option{-finline-small-functions} options.
8035 Enabled at level @option{-O2}.
8037 @item -finline-functions
8038 @opindex finline-functions
8039 Consider all functions for inlining, even if they are not declared inline.
8040 The compiler heuristically decides which functions are worth integrating
8043 If all calls to a given function are integrated, and the function is
8044 declared @code{static}, then the function is normally not output as
8045 assembler code in its own right.
8047 Enabled at level @option{-O3}.
8049 @item -finline-functions-called-once
8050 @opindex finline-functions-called-once
8051 Consider all @code{static} functions called once for inlining into their
8052 caller even if they are not marked @code{inline}. If a call to a given
8053 function is integrated, then the function is not output as assembler code
8056 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
8058 @item -fearly-inlining
8059 @opindex fearly-inlining
8060 Inline functions marked by @code{always_inline} and functions whose body seems
8061 smaller than the function call overhead early before doing
8062 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8063 makes profiling significantly cheaper and usually inlining faster on programs
8064 having large chains of nested wrapper functions.
8070 Perform interprocedural scalar replacement of aggregates, removal of
8071 unused parameters and replacement of parameters passed by reference
8072 by parameters passed by value.
8074 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8076 @item -finline-limit=@var{n}
8077 @opindex finline-limit
8078 By default, GCC limits the size of functions that can be inlined. This flag
8079 allows coarse control of this limit. @var{n} is the size of functions that
8080 can be inlined in number of pseudo instructions.
8082 Inlining is actually controlled by a number of parameters, which may be
8083 specified individually by using @option{--param @var{name}=@var{value}}.
8084 The @option{-finline-limit=@var{n}} option sets some of these parameters
8088 @item max-inline-insns-single
8089 is set to @var{n}/2.
8090 @item max-inline-insns-auto
8091 is set to @var{n}/2.
8094 See below for a documentation of the individual
8095 parameters controlling inlining and for the defaults of these parameters.
8097 @emph{Note:} there may be no value to @option{-finline-limit} that results
8098 in default behavior.
8100 @emph{Note:} pseudo instruction represents, in this particular context, an
8101 abstract measurement of function's size. In no way does it represent a count
8102 of assembly instructions and as such its exact meaning might change from one
8103 release to an another.
8105 @item -fno-keep-inline-dllexport
8106 @opindex fno-keep-inline-dllexport
8107 This is a more fine-grained version of @option{-fkeep-inline-functions},
8108 which applies only to functions that are declared using the @code{dllexport}
8109 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
8112 @item -fkeep-inline-functions
8113 @opindex fkeep-inline-functions
8114 In C, emit @code{static} functions that are declared @code{inline}
8115 into the object file, even if the function has been inlined into all
8116 of its callers. This switch does not affect functions using the
8117 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8118 inline functions into the object file.
8120 @item -fkeep-static-functions
8121 @opindex fkeep-static-functions
8122 Emit @code{static} functions into the object file, even if the function
8125 @item -fkeep-static-consts
8126 @opindex fkeep-static-consts
8127 Emit variables declared @code{static const} when optimization isn't turned
8128 on, even if the variables aren't referenced.
8130 GCC enables this option by default. If you want to force the compiler to
8131 check if a variable is referenced, regardless of whether or not
8132 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8134 @item -fmerge-constants
8135 @opindex fmerge-constants
8136 Attempt to merge identical constants (string constants and floating-point
8137 constants) across compilation units.
8139 This option is the default for optimized compilation if the assembler and
8140 linker support it. Use @option{-fno-merge-constants} to inhibit this
8143 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8145 @item -fmerge-all-constants
8146 @opindex fmerge-all-constants
8147 Attempt to merge identical constants and identical variables.
8149 This option implies @option{-fmerge-constants}. In addition to
8150 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8151 arrays or initialized constant variables with integral or floating-point
8152 types. Languages like C or C++ require each variable, including multiple
8153 instances of the same variable in recursive calls, to have distinct locations,
8154 so using this option results in non-conforming
8157 @item -fmodulo-sched
8158 @opindex fmodulo-sched
8159 Perform swing modulo scheduling immediately before the first scheduling
8160 pass. This pass looks at innermost loops and reorders their
8161 instructions by overlapping different iterations.
8163 @item -fmodulo-sched-allow-regmoves
8164 @opindex fmodulo-sched-allow-regmoves
8165 Perform more aggressive SMS-based modulo scheduling with register moves
8166 allowed. By setting this flag certain anti-dependences edges are
8167 deleted, which triggers the generation of reg-moves based on the
8168 life-range analysis. This option is effective only with
8169 @option{-fmodulo-sched} enabled.
8171 @item -fno-branch-count-reg
8172 @opindex fno-branch-count-reg
8173 Do not use ``decrement and branch'' instructions on a count register,
8174 but instead generate a sequence of instructions that decrement a
8175 register, compare it against zero, then branch based upon the result.
8176 This option is only meaningful on architectures that support such
8177 instructions, which include x86, PowerPC, IA-64 and S/390.
8179 Enabled by default at @option{-O1} and higher.
8181 The default is @option{-fbranch-count-reg}.
8183 @item -fno-function-cse
8184 @opindex fno-function-cse
8185 Do not put function addresses in registers; make each instruction that
8186 calls a constant function contain the function's address explicitly.
8188 This option results in less efficient code, but some strange hacks
8189 that alter the assembler output may be confused by the optimizations
8190 performed when this option is not used.
8192 The default is @option{-ffunction-cse}
8194 @item -fno-zero-initialized-in-bss
8195 @opindex fno-zero-initialized-in-bss
8196 If the target supports a BSS section, GCC by default puts variables that
8197 are initialized to zero into BSS@. This can save space in the resulting
8200 This option turns off this behavior because some programs explicitly
8201 rely on variables going to the data section---e.g., so that the
8202 resulting executable can find the beginning of that section and/or make
8203 assumptions based on that.
8205 The default is @option{-fzero-initialized-in-bss}.
8207 @item -fthread-jumps
8208 @opindex fthread-jumps
8209 Perform optimizations that check to see if a jump branches to a
8210 location where another comparison subsumed by the first is found. If
8211 so, the first branch is redirected to either the destination of the
8212 second branch or a point immediately following it, depending on whether
8213 the condition is known to be true or false.
8215 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8217 @item -fsplit-wide-types
8218 @opindex fsplit-wide-types
8219 When using a type that occupies multiple registers, such as @code{long
8220 long} on a 32-bit system, split the registers apart and allocate them
8221 independently. This normally generates better code for those types,
8222 but may make debugging more difficult.
8224 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8227 @item -fcse-follow-jumps
8228 @opindex fcse-follow-jumps
8229 In common subexpression elimination (CSE), scan through jump instructions
8230 when the target of the jump is not reached by any other path. For
8231 example, when CSE encounters an @code{if} statement with an
8232 @code{else} clause, CSE follows the jump when the condition
8235 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8237 @item -fcse-skip-blocks
8238 @opindex fcse-skip-blocks
8239 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8240 follow jumps that conditionally skip over blocks. When CSE
8241 encounters a simple @code{if} statement with no else clause,
8242 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8243 body of the @code{if}.
8245 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8247 @item -frerun-cse-after-loop
8248 @opindex frerun-cse-after-loop
8249 Re-run common subexpression elimination after loop optimizations are
8252 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8256 Perform a global common subexpression elimination pass.
8257 This pass also performs global constant and copy propagation.
8259 @emph{Note:} When compiling a program using computed gotos, a GCC
8260 extension, you may get better run-time performance if you disable
8261 the global common subexpression elimination pass by adding
8262 @option{-fno-gcse} to the command line.
8264 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8268 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8269 attempts to move loads that are only killed by stores into themselves. This
8270 allows a loop containing a load/store sequence to be changed to a load outside
8271 the loop, and a copy/store within the loop.
8273 Enabled by default when @option{-fgcse} is enabled.
8277 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8278 global common subexpression elimination. This pass attempts to move
8279 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8280 loops containing a load/store sequence can be changed to a load before
8281 the loop and a store after the loop.
8283 Not enabled at any optimization level.
8287 When @option{-fgcse-las} is enabled, the global common subexpression
8288 elimination pass eliminates redundant loads that come after stores to the
8289 same memory location (both partial and full redundancies).
8291 Not enabled at any optimization level.
8293 @item -fgcse-after-reload
8294 @opindex fgcse-after-reload
8295 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8296 pass is performed after reload. The purpose of this pass is to clean up
8299 @item -faggressive-loop-optimizations
8300 @opindex faggressive-loop-optimizations
8301 This option tells the loop optimizer to use language constraints to
8302 derive bounds for the number of iterations of a loop. This assumes that
8303 loop code does not invoke undefined behavior by for example causing signed
8304 integer overflows or out-of-bound array accesses. The bounds for the
8305 number of iterations of a loop are used to guide loop unrolling and peeling
8306 and loop exit test optimizations.
8307 This option is enabled by default.
8309 @item -funsafe-loop-optimizations
8310 @opindex funsafe-loop-optimizations
8311 This option tells the loop optimizer to assume that loop indices do not
8312 overflow, and that loops with nontrivial exit condition are not
8313 infinite. This enables a wider range of loop optimizations even if
8314 the loop optimizer itself cannot prove that these assumptions are valid.
8315 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
8316 if it finds this kind of loop.
8318 @item -fcrossjumping
8319 @opindex fcrossjumping
8320 Perform cross-jumping transformation.
8321 This transformation unifies equivalent code and saves code size. The
8322 resulting code may or may not perform better than without cross-jumping.
8324 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8326 @item -fauto-inc-dec
8327 @opindex fauto-inc-dec
8328 Combine increments or decrements of addresses with memory accesses.
8329 This pass is always skipped on architectures that do not have
8330 instructions to support this. Enabled by default at @option{-O} and
8331 higher on architectures that support this.
8335 Perform dead code elimination (DCE) on RTL@.
8336 Enabled by default at @option{-O} and higher.
8340 Perform dead store elimination (DSE) on RTL@.
8341 Enabled by default at @option{-O} and higher.
8343 @item -fif-conversion
8344 @opindex fif-conversion
8345 Attempt to transform conditional jumps into branch-less equivalents. This
8346 includes use of conditional moves, min, max, set flags and abs instructions, and
8347 some tricks doable by standard arithmetics. The use of conditional execution
8348 on chips where it is available is controlled by @option{-fif-conversion2}.
8350 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8352 @item -fif-conversion2
8353 @opindex fif-conversion2
8354 Use conditional execution (where available) to transform conditional jumps into
8355 branch-less equivalents.
8357 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8359 @item -fdeclone-ctor-dtor
8360 @opindex fdeclone-ctor-dtor
8361 The C++ ABI requires multiple entry points for constructors and
8362 destructors: one for a base subobject, one for a complete object, and
8363 one for a virtual destructor that calls operator delete afterwards.
8364 For a hierarchy with virtual bases, the base and complete variants are
8365 clones, which means two copies of the function. With this option, the
8366 base and complete variants are changed to be thunks that call a common
8369 Enabled by @option{-Os}.
8371 @item -fdelete-null-pointer-checks
8372 @opindex fdelete-null-pointer-checks
8373 Assume that programs cannot safely dereference null pointers, and that
8374 no code or data element resides at address zero.
8375 This option enables simple constant
8376 folding optimizations at all optimization levels. In addition, other
8377 optimization passes in GCC use this flag to control global dataflow
8378 analyses that eliminate useless checks for null pointers; these assume
8379 that a memory access to address zero always results in a trap, so
8380 that if a pointer is checked after it has already been dereferenced,
8383 Note however that in some environments this assumption is not true.
8384 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8385 for programs that depend on that behavior.
8387 This option is enabled by default on most targets. On Nios II ELF, it
8388 defaults to off. On AVR and CR16, this option is completely disabled.
8390 Passes that use the dataflow information
8391 are enabled independently at different optimization levels.
8393 @item -fdevirtualize
8394 @opindex fdevirtualize
8395 Attempt to convert calls to virtual functions to direct calls. This
8396 is done both within a procedure and interprocedurally as part of
8397 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8398 propagation (@option{-fipa-cp}).
8399 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8401 @item -fdevirtualize-speculatively
8402 @opindex fdevirtualize-speculatively
8403 Attempt to convert calls to virtual functions to speculative direct calls.
8404 Based on the analysis of the type inheritance graph, determine for a given call
8405 the set of likely targets. If the set is small, preferably of size 1, change
8406 the call into a conditional deciding between direct and indirect calls. The
8407 speculative calls enable more optimizations, such as inlining. When they seem
8408 useless after further optimization, they are converted back into original form.
8410 @item -fdevirtualize-at-ltrans
8411 @opindex fdevirtualize-at-ltrans
8412 Stream extra information needed for aggressive devirtualization when running
8413 the link-time optimizer in local transformation mode.
8414 This option enables more devirtualization but
8415 significantly increases the size of streamed data. For this reason it is
8416 disabled by default.
8418 @item -fexpensive-optimizations
8419 @opindex fexpensive-optimizations
8420 Perform a number of minor optimizations that are relatively expensive.
8422 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8426 Attempt to remove redundant extension instructions. This is especially
8427 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8428 registers after writing to their lower 32-bit half.
8430 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8431 @option{-O3}, @option{-Os}.
8433 @item -fno-lifetime-dse
8434 @opindex fno-lifetime-dse
8435 In C++ the value of an object is only affected by changes within its
8436 lifetime: when the constructor begins, the object has an indeterminate
8437 value, and any changes during the lifetime of the object are dead when
8438 the object is destroyed. Normally dead store elimination will take
8439 advantage of this; if your code relies on the value of the object
8440 storage persisting beyond the lifetime of the object, you can use this
8441 flag to disable this optimization.
8443 @item -flive-range-shrinkage
8444 @opindex flive-range-shrinkage
8445 Attempt to decrease register pressure through register live range
8446 shrinkage. This is helpful for fast processors with small or moderate
8449 @item -fira-algorithm=@var{algorithm}
8450 @opindex fira-algorithm
8451 Use the specified coloring algorithm for the integrated register
8452 allocator. The @var{algorithm} argument can be @samp{priority}, which
8453 specifies Chow's priority coloring, or @samp{CB}, which specifies
8454 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8455 for all architectures, but for those targets that do support it, it is
8456 the default because it generates better code.
8458 @item -fira-region=@var{region}
8459 @opindex fira-region
8460 Use specified regions for the integrated register allocator. The
8461 @var{region} argument should be one of the following:
8466 Use all loops as register allocation regions.
8467 This can give the best results for machines with a small and/or
8468 irregular register set.
8471 Use all loops except for loops with small register pressure
8472 as the regions. This value usually gives
8473 the best results in most cases and for most architectures,
8474 and is enabled by default when compiling with optimization for speed
8475 (@option{-O}, @option{-O2}, @dots{}).
8478 Use all functions as a single region.
8479 This typically results in the smallest code size, and is enabled by default for
8480 @option{-Os} or @option{-O0}.
8484 @item -fira-hoist-pressure
8485 @opindex fira-hoist-pressure
8486 Use IRA to evaluate register pressure in the code hoisting pass for
8487 decisions to hoist expressions. This option usually results in smaller
8488 code, but it can slow the compiler down.
8490 This option is enabled at level @option{-Os} for all targets.
8492 @item -fira-loop-pressure
8493 @opindex fira-loop-pressure
8494 Use IRA to evaluate register pressure in loops for decisions to move
8495 loop invariants. This option usually results in generation
8496 of faster and smaller code on machines with large register files (>= 32
8497 registers), but it can slow the compiler down.
8499 This option is enabled at level @option{-O3} for some targets.
8501 @item -fno-ira-share-save-slots
8502 @opindex fno-ira-share-save-slots
8503 Disable sharing of stack slots used for saving call-used hard
8504 registers living through a call. Each hard register gets a
8505 separate stack slot, and as a result function stack frames are
8508 @item -fno-ira-share-spill-slots
8509 @opindex fno-ira-share-spill-slots
8510 Disable sharing of stack slots allocated for pseudo-registers. Each
8511 pseudo-register that does not get a hard register gets a separate
8512 stack slot, and as a result function stack frames are larger.
8514 @item -fira-verbose=@var{n}
8515 @opindex fira-verbose
8516 Control the verbosity of the dump file for the integrated register allocator.
8517 The default value is 5. If the value @var{n} is greater or equal to 10,
8518 the dump output is sent to stderr using the same format as @var{n} minus 10.
8522 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8523 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8524 values if it is profitable.
8526 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8528 @item -fdelayed-branch
8529 @opindex fdelayed-branch
8530 If supported for the target machine, attempt to reorder instructions
8531 to exploit instruction slots available after delayed branch
8534 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8536 @item -fschedule-insns
8537 @opindex fschedule-insns
8538 If supported for the target machine, attempt to reorder instructions to
8539 eliminate execution stalls due to required data being unavailable. This
8540 helps machines that have slow floating point or memory load instructions
8541 by allowing other instructions to be issued until the result of the load
8542 or floating-point instruction is required.
8544 Enabled at levels @option{-O2}, @option{-O3}.
8546 @item -fschedule-insns2
8547 @opindex fschedule-insns2
8548 Similar to @option{-fschedule-insns}, but requests an additional pass of
8549 instruction scheduling after register allocation has been done. This is
8550 especially useful on machines with a relatively small number of
8551 registers and where memory load instructions take more than one cycle.
8553 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8555 @item -fno-sched-interblock
8556 @opindex fno-sched-interblock
8557 Don't schedule instructions across basic blocks. This is normally
8558 enabled by default when scheduling before register allocation, i.e.@:
8559 with @option{-fschedule-insns} or at @option{-O2} or higher.
8561 @item -fno-sched-spec
8562 @opindex fno-sched-spec
8563 Don't allow speculative motion of non-load instructions. This is normally
8564 enabled by default when scheduling before register allocation, i.e.@:
8565 with @option{-fschedule-insns} or at @option{-O2} or higher.
8567 @item -fsched-pressure
8568 @opindex fsched-pressure
8569 Enable register pressure sensitive insn scheduling before register
8570 allocation. This only makes sense when scheduling before register
8571 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8572 @option{-O2} or higher. Usage of this option can improve the
8573 generated code and decrease its size by preventing register pressure
8574 increase above the number of available hard registers and subsequent
8575 spills in register allocation.
8577 @item -fsched-spec-load
8578 @opindex fsched-spec-load
8579 Allow speculative motion of some load instructions. This only makes
8580 sense when scheduling before register allocation, i.e.@: with
8581 @option{-fschedule-insns} or at @option{-O2} or higher.
8583 @item -fsched-spec-load-dangerous
8584 @opindex fsched-spec-load-dangerous
8585 Allow speculative motion of more load instructions. This only makes
8586 sense when scheduling before register allocation, i.e.@: with
8587 @option{-fschedule-insns} or at @option{-O2} or higher.
8589 @item -fsched-stalled-insns
8590 @itemx -fsched-stalled-insns=@var{n}
8591 @opindex fsched-stalled-insns
8592 Define how many insns (if any) can be moved prematurely from the queue
8593 of stalled insns into the ready list during the second scheduling pass.
8594 @option{-fno-sched-stalled-insns} means that no insns are moved
8595 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8596 on how many queued insns can be moved prematurely.
8597 @option{-fsched-stalled-insns} without a value is equivalent to
8598 @option{-fsched-stalled-insns=1}.
8600 @item -fsched-stalled-insns-dep
8601 @itemx -fsched-stalled-insns-dep=@var{n}
8602 @opindex fsched-stalled-insns-dep
8603 Define how many insn groups (cycles) are examined for a dependency
8604 on a stalled insn that is a candidate for premature removal from the queue
8605 of stalled insns. This has an effect only during the second scheduling pass,
8606 and only if @option{-fsched-stalled-insns} is used.
8607 @option{-fno-sched-stalled-insns-dep} is equivalent to
8608 @option{-fsched-stalled-insns-dep=0}.
8609 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8610 @option{-fsched-stalled-insns-dep=1}.
8612 @item -fsched2-use-superblocks
8613 @opindex fsched2-use-superblocks
8614 When scheduling after register allocation, use superblock scheduling.
8615 This allows motion across basic block boundaries,
8616 resulting in faster schedules. This option is experimental, as not all machine
8617 descriptions used by GCC model the CPU closely enough to avoid unreliable
8618 results from the algorithm.
8620 This only makes sense when scheduling after register allocation, i.e.@: with
8621 @option{-fschedule-insns2} or at @option{-O2} or higher.
8623 @item -fsched-group-heuristic
8624 @opindex fsched-group-heuristic
8625 Enable the group heuristic in the scheduler. This heuristic favors
8626 the instruction that belongs to a schedule group. This is enabled
8627 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8628 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8630 @item -fsched-critical-path-heuristic
8631 @opindex fsched-critical-path-heuristic
8632 Enable the critical-path heuristic in the scheduler. This heuristic favors
8633 instructions on the critical path. This is enabled by default when
8634 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8635 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8637 @item -fsched-spec-insn-heuristic
8638 @opindex fsched-spec-insn-heuristic
8639 Enable the speculative instruction heuristic in the scheduler. This
8640 heuristic favors speculative instructions with greater dependency weakness.
8641 This is enabled by default when scheduling is enabled, i.e.@:
8642 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8643 or at @option{-O2} or higher.
8645 @item -fsched-rank-heuristic
8646 @opindex fsched-rank-heuristic
8647 Enable the rank heuristic in the scheduler. This heuristic favors
8648 the instruction belonging to a basic block with greater size or frequency.
8649 This is enabled by default when scheduling is enabled, i.e.@:
8650 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8651 at @option{-O2} or higher.
8653 @item -fsched-last-insn-heuristic
8654 @opindex fsched-last-insn-heuristic
8655 Enable the last-instruction heuristic in the scheduler. This heuristic
8656 favors the instruction that is less dependent on the last instruction
8657 scheduled. This is enabled by default when scheduling is enabled,
8658 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8659 at @option{-O2} or higher.
8661 @item -fsched-dep-count-heuristic
8662 @opindex fsched-dep-count-heuristic
8663 Enable the dependent-count heuristic in the scheduler. This heuristic
8664 favors the instruction that has more instructions depending on it.
8665 This is enabled by default when scheduling is enabled, i.e.@:
8666 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8667 at @option{-O2} or higher.
8669 @item -freschedule-modulo-scheduled-loops
8670 @opindex freschedule-modulo-scheduled-loops
8671 Modulo scheduling is performed before traditional scheduling. If a loop
8672 is modulo scheduled, later scheduling passes may change its schedule.
8673 Use this option to control that behavior.
8675 @item -fselective-scheduling
8676 @opindex fselective-scheduling
8677 Schedule instructions using selective scheduling algorithm. Selective
8678 scheduling runs instead of the first scheduler pass.
8680 @item -fselective-scheduling2
8681 @opindex fselective-scheduling2
8682 Schedule instructions using selective scheduling algorithm. Selective
8683 scheduling runs instead of the second scheduler pass.
8685 @item -fsel-sched-pipelining
8686 @opindex fsel-sched-pipelining
8687 Enable software pipelining of innermost loops during selective scheduling.
8688 This option has no effect unless one of @option{-fselective-scheduling} or
8689 @option{-fselective-scheduling2} is turned on.
8691 @item -fsel-sched-pipelining-outer-loops
8692 @opindex fsel-sched-pipelining-outer-loops
8693 When pipelining loops during selective scheduling, also pipeline outer loops.
8694 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8696 @item -fsemantic-interposition
8697 @opindex fsemantic-interposition
8698 Some object formats, like ELF, allow interposing of symbols by the
8700 This means that for symbols exported from the DSO, the compiler cannot perform
8701 interprocedural propagation, inlining and other optimizations in anticipation
8702 that the function or variable in question may change. While this feature is
8703 useful, for example, to rewrite memory allocation functions by a debugging
8704 implementation, it is expensive in the terms of code quality.
8705 With @option{-fno-semantic-interposition} the compiler assumes that
8706 if interposition happens for functions the overwriting function will have
8707 precisely the same semantics (and side effects).
8708 Similarly if interposition happens
8709 for variables, the constructor of the variable will be the same. The flag
8710 has no effect for functions explicitly declared inline
8711 (where it is never allowed for interposition to change semantics)
8712 and for symbols explicitly declared weak.
8715 @opindex fshrink-wrap
8716 Emit function prologues only before parts of the function that need it,
8717 rather than at the top of the function. This flag is enabled by default at
8718 @option{-O} and higher.
8720 @item -fcaller-saves
8721 @opindex fcaller-saves
8722 Enable allocation of values to registers that are clobbered by
8723 function calls, by emitting extra instructions to save and restore the
8724 registers around such calls. Such allocation is done only when it
8725 seems to result in better code.
8727 This option is always enabled by default on certain machines, usually
8728 those which have no call-preserved registers to use instead.
8730 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8732 @item -fcombine-stack-adjustments
8733 @opindex fcombine-stack-adjustments
8734 Tracks stack adjustments (pushes and pops) and stack memory references
8735 and then tries to find ways to combine them.
8737 Enabled by default at @option{-O1} and higher.
8741 Use caller save registers for allocation if those registers are not used by
8742 any called function. In that case it is not necessary to save and restore
8743 them around calls. This is only possible if called functions are part of
8744 same compilation unit as current function and they are compiled before it.
8746 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8748 @item -fconserve-stack
8749 @opindex fconserve-stack
8750 Attempt to minimize stack usage. The compiler attempts to use less
8751 stack space, even if that makes the program slower. This option
8752 implies setting the @option{large-stack-frame} parameter to 100
8753 and the @option{large-stack-frame-growth} parameter to 400.
8755 @item -ftree-reassoc
8756 @opindex ftree-reassoc
8757 Perform reassociation on trees. This flag is enabled by default
8758 at @option{-O} and higher.
8762 Perform partial redundancy elimination (PRE) on trees. This flag is
8763 enabled by default at @option{-O2} and @option{-O3}.
8765 @item -ftree-partial-pre
8766 @opindex ftree-partial-pre
8767 Make partial redundancy elimination (PRE) more aggressive. This flag is
8768 enabled by default at @option{-O3}.
8770 @item -ftree-forwprop
8771 @opindex ftree-forwprop
8772 Perform forward propagation on trees. This flag is enabled by default
8773 at @option{-O} and higher.
8777 Perform full redundancy elimination (FRE) on trees. The difference
8778 between FRE and PRE is that FRE only considers expressions
8779 that are computed on all paths leading to the redundant computation.
8780 This analysis is faster than PRE, though it exposes fewer redundancies.
8781 This flag is enabled by default at @option{-O} and higher.
8783 @item -ftree-phiprop
8784 @opindex ftree-phiprop
8785 Perform hoisting of loads from conditional pointers on trees. This
8786 pass is enabled by default at @option{-O} and higher.
8788 @item -fhoist-adjacent-loads
8789 @opindex fhoist-adjacent-loads
8790 Speculatively hoist loads from both branches of an if-then-else if the
8791 loads are from adjacent locations in the same structure and the target
8792 architecture has a conditional move instruction. This flag is enabled
8793 by default at @option{-O2} and higher.
8795 @item -ftree-copy-prop
8796 @opindex ftree-copy-prop
8797 Perform copy propagation on trees. This pass eliminates unnecessary
8798 copy operations. This flag is enabled by default at @option{-O} and
8801 @item -fipa-pure-const
8802 @opindex fipa-pure-const
8803 Discover which functions are pure or constant.
8804 Enabled by default at @option{-O} and higher.
8806 @item -fipa-reference
8807 @opindex fipa-reference
8808 Discover which static variables do not escape the
8810 Enabled by default at @option{-O} and higher.
8814 Perform interprocedural pointer analysis and interprocedural modification
8815 and reference analysis. This option can cause excessive memory and
8816 compile-time usage on large compilation units. It is not enabled by
8817 default at any optimization level.
8820 @opindex fipa-profile
8821 Perform interprocedural profile propagation. The functions called only from
8822 cold functions are marked as cold. Also functions executed once (such as
8823 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8824 functions and loop less parts of functions executed once are then optimized for
8826 Enabled by default at @option{-O} and higher.
8830 Perform interprocedural constant propagation.
8831 This optimization analyzes the program to determine when values passed
8832 to functions are constants and then optimizes accordingly.
8833 This optimization can substantially increase performance
8834 if the application has constants passed to functions.
8835 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8837 @item -fipa-cp-clone
8838 @opindex fipa-cp-clone
8839 Perform function cloning to make interprocedural constant propagation stronger.
8840 When enabled, interprocedural constant propagation performs function cloning
8841 when externally visible function can be called with constant arguments.
8842 Because this optimization can create multiple copies of functions,
8843 it may significantly increase code size
8844 (see @option{--param ipcp-unit-growth=@var{value}}).
8845 This flag is enabled by default at @option{-O3}.
8847 @item -fipa-cp-alignment
8848 @opindex -fipa-cp-alignment
8849 When enabled, this optimization propagates alignment of function
8850 parameters to support better vectorization and string operations.
8852 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8853 requires that @option{-fipa-cp} is enabled.
8857 Perform Identical Code Folding for functions and read-only variables.
8858 The optimization reduces code size and may disturb unwind stacks by replacing
8859 a function by equivalent one with a different name. The optimization works
8860 more effectively with link time optimization enabled.
8862 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8863 works on different levels and thus the optimizations are not same - there are
8864 equivalences that are found only by GCC and equivalences found only by Gold.
8866 This flag is enabled by default at @option{-O2} and @option{-Os}.
8868 @item -fisolate-erroneous-paths-dereference
8869 @opindex fisolate-erroneous-paths-dereference
8870 Detect paths that trigger erroneous or undefined behavior due to
8871 dereferencing a null pointer. Isolate those paths from the main control
8872 flow and turn the statement with erroneous or undefined behavior into a trap.
8873 This flag is enabled by default at @option{-O2} and higher and depends on
8874 @option{-fdelete-null-pointer-checks} also being enabled.
8876 @item -fisolate-erroneous-paths-attribute
8877 @opindex fisolate-erroneous-paths-attribute
8878 Detect paths that trigger erroneous or undefined behavior due a null value
8879 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8880 attribute. Isolate those paths from the main control flow and turn the
8881 statement with erroneous or undefined behavior into a trap. This is not
8882 currently enabled, but may be enabled by @option{-O2} in the future.
8886 Perform forward store motion on trees. This flag is
8887 enabled by default at @option{-O} and higher.
8889 @item -ftree-bit-ccp
8890 @opindex ftree-bit-ccp
8891 Perform sparse conditional bit constant propagation on trees and propagate
8892 pointer alignment information.
8893 This pass only operates on local scalar variables and is enabled by default
8894 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8898 Perform sparse conditional constant propagation (CCP) on trees. This
8899 pass only operates on local scalar variables and is enabled by default
8900 at @option{-O} and higher.
8902 @item -fssa-backprop
8903 @opindex fssa-backprop
8904 Propagate information about uses of a value up the definition chain
8905 in order to simplify the definitions. For example, this pass strips
8906 sign operations if the sign of a value never matters. The flag is
8907 enabled by default at @option{-O} and higher.
8910 @opindex fssa-phiopt
8911 Perform pattern matching on SSA PHI nodes to optimize conditional
8912 code. This pass is enabled by default at @option{-O} and higher.
8914 @item -ftree-switch-conversion
8915 @opindex ftree-switch-conversion
8916 Perform conversion of simple initializations in a switch to
8917 initializations from a scalar array. This flag is enabled by default
8918 at @option{-O2} and higher.
8920 @item -ftree-tail-merge
8921 @opindex ftree-tail-merge
8922 Look for identical code sequences. When found, replace one with a jump to the
8923 other. This optimization is known as tail merging or cross jumping. This flag
8924 is enabled by default at @option{-O2} and higher. The compilation time
8926 be limited using @option{max-tail-merge-comparisons} parameter and
8927 @option{max-tail-merge-iterations} parameter.
8931 Perform dead code elimination (DCE) on trees. This flag is enabled by
8932 default at @option{-O} and higher.
8934 @item -ftree-builtin-call-dce
8935 @opindex ftree-builtin-call-dce
8936 Perform conditional dead code elimination (DCE) for calls to built-in functions
8937 that may set @code{errno} but are otherwise side-effect free. This flag is
8938 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8941 @item -ftree-dominator-opts
8942 @opindex ftree-dominator-opts
8943 Perform a variety of simple scalar cleanups (constant/copy
8944 propagation, redundancy elimination, range propagation and expression
8945 simplification) based on a dominator tree traversal. This also
8946 performs jump threading (to reduce jumps to jumps). This flag is
8947 enabled by default at @option{-O} and higher.
8951 Perform dead store elimination (DSE) on trees. A dead store is a store into
8952 a memory location that is later overwritten by another store without
8953 any intervening loads. In this case the earlier store can be deleted. This
8954 flag is enabled by default at @option{-O} and higher.
8958 Perform loop header copying on trees. This is beneficial since it increases
8959 effectiveness of code motion optimizations. It also saves one jump. This flag
8960 is enabled by default at @option{-O} and higher. It is not enabled
8961 for @option{-Os}, since it usually increases code size.
8963 @item -ftree-loop-optimize
8964 @opindex ftree-loop-optimize
8965 Perform loop optimizations on trees. This flag is enabled by default
8966 at @option{-O} and higher.
8968 @item -ftree-loop-linear
8969 @itemx -floop-interchange
8970 @itemx -floop-strip-mine
8972 @itemx -floop-unroll-and-jam
8973 @opindex ftree-loop-linear
8974 @opindex floop-interchange
8975 @opindex floop-strip-mine
8976 @opindex floop-block
8977 @opindex floop-unroll-and-jam
8978 Perform loop nest optimizations. Same as
8979 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8980 to be configured with @option{--with-isl} to enable the Graphite loop
8981 transformation infrastructure.
8983 @item -fgraphite-identity
8984 @opindex fgraphite-identity
8985 Enable the identity transformation for graphite. For every SCoP we generate
8986 the polyhedral representation and transform it back to gimple. Using
8987 @option{-fgraphite-identity} we can check the costs or benefits of the
8988 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8989 are also performed by the code generator isl, like index splitting and
8990 dead code elimination in loops.
8992 @item -floop-nest-optimize
8993 @opindex floop-nest-optimize
8994 Enable the isl based loop nest optimizer. This is a generic loop nest
8995 optimizer based on the Pluto optimization algorithms. It calculates a loop
8996 structure optimized for data-locality and parallelism. This option
8999 @item -floop-parallelize-all
9000 @opindex floop-parallelize-all
9001 Use the Graphite data dependence analysis to identify loops that can
9002 be parallelized. Parallelize all the loops that can be analyzed to
9003 not contain loop carried dependences without checking that it is
9004 profitable to parallelize the loops.
9006 @item -ftree-coalesce-vars
9007 @opindex ftree-coalesce-vars
9008 While transforming the program out of the SSA representation, attempt to
9009 reduce copying by coalescing versions of different user-defined
9010 variables, instead of just compiler temporaries. This may severely
9011 limit the ability to debug an optimized program compiled with
9012 @option{-fno-var-tracking-assignments}. In the negated form, this flag
9013 prevents SSA coalescing of user variables. This option is enabled by
9014 default if optimization is enabled, and it does very little otherwise.
9016 @item -ftree-loop-if-convert
9017 @opindex ftree-loop-if-convert
9018 Attempt to transform conditional jumps in the innermost loops to
9019 branch-less equivalents. The intent is to remove control-flow from
9020 the innermost loops in order to improve the ability of the
9021 vectorization pass to handle these loops. This is enabled by default
9022 if vectorization is enabled.
9024 @item -ftree-loop-if-convert-stores
9025 @opindex ftree-loop-if-convert-stores
9026 Attempt to also if-convert conditional jumps containing memory writes.
9027 This transformation can be unsafe for multi-threaded programs as it
9028 transforms conditional memory writes into unconditional memory writes.
9031 for (i = 0; i < N; i++)
9037 for (i = 0; i < N; i++)
9038 A[i] = cond ? expr : A[i];
9040 potentially producing data races.
9042 @item -ftree-loop-distribution
9043 @opindex ftree-loop-distribution
9044 Perform loop distribution. This flag can improve cache performance on
9045 big loop bodies and allow further loop optimizations, like
9046 parallelization or vectorization, to take place. For example, the loop
9063 @item -ftree-loop-distribute-patterns
9064 @opindex ftree-loop-distribute-patterns
9065 Perform loop distribution of patterns that can be code generated with
9066 calls to a library. This flag is enabled by default at @option{-O3}.
9068 This pass distributes the initialization loops and generates a call to
9069 memset zero. For example, the loop
9085 and the initialization loop is transformed into a call to memset zero.
9087 @item -ftree-loop-im
9088 @opindex ftree-loop-im
9089 Perform loop invariant motion on trees. This pass moves only invariants that
9090 are hard to handle at RTL level (function calls, operations that expand to
9091 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9092 operands of conditions that are invariant out of the loop, so that we can use
9093 just trivial invariantness analysis in loop unswitching. The pass also includes
9096 @item -ftree-loop-ivcanon
9097 @opindex ftree-loop-ivcanon
9098 Create a canonical counter for number of iterations in loops for which
9099 determining number of iterations requires complicated analysis. Later
9100 optimizations then may determine the number easily. Useful especially
9101 in connection with unrolling.
9105 Perform induction variable optimizations (strength reduction, induction
9106 variable merging and induction variable elimination) on trees.
9108 @item -ftree-parallelize-loops=n
9109 @opindex ftree-parallelize-loops
9110 Parallelize loops, i.e., split their iteration space to run in n threads.
9111 This is only possible for loops whose iterations are independent
9112 and can be arbitrarily reordered. The optimization is only
9113 profitable on multiprocessor machines, for loops that are CPU-intensive,
9114 rather than constrained e.g.@: by memory bandwidth. This option
9115 implies @option{-pthread}, and thus is only supported on targets
9116 that have support for @option{-pthread}.
9120 Perform function-local points-to analysis on trees. This flag is
9121 enabled by default at @option{-O} and higher.
9125 Perform scalar replacement of aggregates. This pass replaces structure
9126 references with scalars to prevent committing structures to memory too
9127 early. This flag is enabled by default at @option{-O} and higher.
9131 Perform temporary expression replacement during the SSA->normal phase. Single
9132 use/single def temporaries are replaced at their use location with their
9133 defining expression. This results in non-GIMPLE code, but gives the expanders
9134 much more complex trees to work on resulting in better RTL generation. This is
9135 enabled by default at @option{-O} and higher.
9139 Perform straight-line strength reduction on trees. This recognizes related
9140 expressions involving multiplications and replaces them by less expensive
9141 calculations when possible. This is enabled by default at @option{-O} and
9144 @item -ftree-vectorize
9145 @opindex ftree-vectorize
9146 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9147 and @option{-ftree-slp-vectorize} if not explicitly specified.
9149 @item -ftree-loop-vectorize
9150 @opindex ftree-loop-vectorize
9151 Perform loop vectorization on trees. This flag is enabled by default at
9152 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9154 @item -ftree-slp-vectorize
9155 @opindex ftree-slp-vectorize
9156 Perform basic block vectorization on trees. This flag is enabled by default at
9157 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9159 @item -fvect-cost-model=@var{model}
9160 @opindex fvect-cost-model
9161 Alter the cost model used for vectorization. The @var{model} argument
9162 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9163 With the @samp{unlimited} model the vectorized code-path is assumed
9164 to be profitable while with the @samp{dynamic} model a runtime check
9165 guards the vectorized code-path to enable it only for iteration
9166 counts that will likely execute faster than when executing the original
9167 scalar loop. The @samp{cheap} model disables vectorization of
9168 loops where doing so would be cost prohibitive for example due to
9169 required runtime checks for data dependence or alignment but otherwise
9170 is equal to the @samp{dynamic} model.
9171 The default cost model depends on other optimization flags and is
9172 either @samp{dynamic} or @samp{cheap}.
9174 @item -fsimd-cost-model=@var{model}
9175 @opindex fsimd-cost-model
9176 Alter the cost model used for vectorization of loops marked with the OpenMP
9177 or Cilk Plus simd directive. The @var{model} argument should be one of
9178 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9179 have the same meaning as described in @option{-fvect-cost-model} and by
9180 default a cost model defined with @option{-fvect-cost-model} is used.
9184 Perform Value Range Propagation on trees. This is similar to the
9185 constant propagation pass, but instead of values, ranges of values are
9186 propagated. This allows the optimizers to remove unnecessary range
9187 checks like array bound checks and null pointer checks. This is
9188 enabled by default at @option{-O2} and higher. Null pointer check
9189 elimination is only done if @option{-fdelete-null-pointer-checks} is
9193 @opindex fsplit-paths
9194 Split paths leading to loop backedges. This can improve dead code
9195 elimination and common subexpression elimination. This is enabled by
9196 default at @option{-O2} and above.
9198 @item -fsplit-ivs-in-unroller
9199 @opindex fsplit-ivs-in-unroller
9200 Enables expression of values of induction variables in later iterations
9201 of the unrolled loop using the value in the first iteration. This breaks
9202 long dependency chains, thus improving efficiency of the scheduling passes.
9204 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9205 same effect. However, that is not reliable in cases where the loop body
9206 is more complicated than a single basic block. It also does not work at all
9207 on some architectures due to restrictions in the CSE pass.
9209 This optimization is enabled by default.
9211 @item -fvariable-expansion-in-unroller
9212 @opindex fvariable-expansion-in-unroller
9213 With this option, the compiler creates multiple copies of some
9214 local variables when unrolling a loop, which can result in superior code.
9216 @item -fpartial-inlining
9217 @opindex fpartial-inlining
9218 Inline parts of functions. This option has any effect only
9219 when inlining itself is turned on by the @option{-finline-functions}
9220 or @option{-finline-small-functions} options.
9222 Enabled at level @option{-O2}.
9224 @item -fpredictive-commoning
9225 @opindex fpredictive-commoning
9226 Perform predictive commoning optimization, i.e., reusing computations
9227 (especially memory loads and stores) performed in previous
9228 iterations of loops.
9230 This option is enabled at level @option{-O3}.
9232 @item -fprefetch-loop-arrays
9233 @opindex fprefetch-loop-arrays
9234 If supported by the target machine, generate instructions to prefetch
9235 memory to improve the performance of loops that access large arrays.
9237 This option may generate better or worse code; results are highly
9238 dependent on the structure of loops within the source code.
9240 Disabled at level @option{-Os}.
9243 @itemx -fno-peephole2
9244 @opindex fno-peephole
9245 @opindex fno-peephole2
9246 Disable any machine-specific peephole optimizations. The difference
9247 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9248 are implemented in the compiler; some targets use one, some use the
9249 other, a few use both.
9251 @option{-fpeephole} is enabled by default.
9252 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9254 @item -fno-guess-branch-probability
9255 @opindex fno-guess-branch-probability
9256 Do not guess branch probabilities using heuristics.
9258 GCC uses heuristics to guess branch probabilities if they are
9259 not provided by profiling feedback (@option{-fprofile-arcs}). These
9260 heuristics are based on the control flow graph. If some branch probabilities
9261 are specified by @code{__builtin_expect}, then the heuristics are
9262 used to guess branch probabilities for the rest of the control flow graph,
9263 taking the @code{__builtin_expect} info into account. The interactions
9264 between the heuristics and @code{__builtin_expect} can be complex, and in
9265 some cases, it may be useful to disable the heuristics so that the effects
9266 of @code{__builtin_expect} are easier to understand.
9268 The default is @option{-fguess-branch-probability} at levels
9269 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9271 @item -freorder-blocks
9272 @opindex freorder-blocks
9273 Reorder basic blocks in the compiled function in order to reduce number of
9274 taken branches and improve code locality.
9276 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9278 @item -freorder-blocks-algorithm=@var{algorithm}
9279 @opindex freorder-blocks-algorithm
9280 Use the specified algorithm for basic block reordering. The
9281 @var{algorithm} argument can be @samp{simple}, which does not increase
9282 code size (except sometimes due to secondary effects like alignment),
9283 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9284 put all often executed code together, minimizing the number of branches
9285 executed by making extra copies of code.
9287 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9288 @samp{stc} at levels @option{-O2}, @option{-O3}.
9290 @item -freorder-blocks-and-partition
9291 @opindex freorder-blocks-and-partition
9292 In addition to reordering basic blocks in the compiled function, in order
9293 to reduce number of taken branches, partitions hot and cold basic blocks
9294 into separate sections of the assembly and .o files, to improve
9295 paging and cache locality performance.
9297 This optimization is automatically turned off in the presence of
9298 exception handling, for linkonce sections, for functions with a user-defined
9299 section attribute and on any architecture that does not support named
9302 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9304 @item -freorder-functions
9305 @opindex freorder-functions
9306 Reorder functions in the object file in order to
9307 improve code locality. This is implemented by using special
9308 subsections @code{.text.hot} for most frequently executed functions and
9309 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9310 the linker so object file format must support named sections and linker must
9311 place them in a reasonable way.
9313 Also profile feedback must be available to make this option effective. See
9314 @option{-fprofile-arcs} for details.
9316 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9318 @item -fstrict-aliasing
9319 @opindex fstrict-aliasing
9320 Allow the compiler to assume the strictest aliasing rules applicable to
9321 the language being compiled. For C (and C++), this activates
9322 optimizations based on the type of expressions. In particular, an
9323 object of one type is assumed never to reside at the same address as an
9324 object of a different type, unless the types are almost the same. For
9325 example, an @code{unsigned int} can alias an @code{int}, but not a
9326 @code{void*} or a @code{double}. A character type may alias any other
9329 @anchor{Type-punning}Pay special attention to code like this:
9342 The practice of reading from a different union member than the one most
9343 recently written to (called ``type-punning'') is common. Even with
9344 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9345 is accessed through the union type. So, the code above works as
9346 expected. @xref{Structures unions enumerations and bit-fields
9347 implementation}. However, this code might not:
9358 Similarly, access by taking the address, casting the resulting pointer
9359 and dereferencing the result has undefined behavior, even if the cast
9360 uses a union type, e.g.:
9364 return ((union a_union *) &d)->i;
9368 The @option{-fstrict-aliasing} option is enabled at levels
9369 @option{-O2}, @option{-O3}, @option{-Os}.
9371 @item -fstrict-overflow
9372 @opindex fstrict-overflow
9373 Allow the compiler to assume strict signed overflow rules, depending
9374 on the language being compiled. For C (and C++) this means that
9375 overflow when doing arithmetic with signed numbers is undefined, which
9376 means that the compiler may assume that it does not happen. This
9377 permits various optimizations. For example, the compiler assumes
9378 that an expression like @code{i + 10 > i} is always true for
9379 signed @code{i}. This assumption is only valid if signed overflow is
9380 undefined, as the expression is false if @code{i + 10} overflows when
9381 using twos complement arithmetic. When this option is in effect any
9382 attempt to determine whether an operation on signed numbers
9383 overflows must be written carefully to not actually involve overflow.
9385 This option also allows the compiler to assume strict pointer
9386 semantics: given a pointer to an object, if adding an offset to that
9387 pointer does not produce a pointer to the same object, the addition is
9388 undefined. This permits the compiler to conclude that @code{p + u >
9389 p} is always true for a pointer @code{p} and unsigned integer
9390 @code{u}. This assumption is only valid because pointer wraparound is
9391 undefined, as the expression is false if @code{p + u} overflows using
9392 twos complement arithmetic.
9394 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9395 that integer signed overflow is fully defined: it wraps. When
9396 @option{-fwrapv} is used, there is no difference between
9397 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9398 integers. With @option{-fwrapv} certain types of overflow are
9399 permitted. For example, if the compiler gets an overflow when doing
9400 arithmetic on constants, the overflowed value can still be used with
9401 @option{-fwrapv}, but not otherwise.
9403 The @option{-fstrict-overflow} option is enabled at levels
9404 @option{-O2}, @option{-O3}, @option{-Os}.
9406 @item -falign-functions
9407 @itemx -falign-functions=@var{n}
9408 @opindex falign-functions
9409 Align the start of functions to the next power-of-two greater than
9410 @var{n}, skipping up to @var{n} bytes. For instance,
9411 @option{-falign-functions=32} aligns functions to the next 32-byte
9412 boundary, but @option{-falign-functions=24} aligns to the next
9413 32-byte boundary only if this can be done by skipping 23 bytes or less.
9415 @option{-fno-align-functions} and @option{-falign-functions=1} are
9416 equivalent and mean that functions are not aligned.
9418 Some assemblers only support this flag when @var{n} is a power of two;
9419 in that case, it is rounded up.
9421 If @var{n} is not specified or is zero, use a machine-dependent default.
9423 Enabled at levels @option{-O2}, @option{-O3}.
9425 @item -falign-labels
9426 @itemx -falign-labels=@var{n}
9427 @opindex falign-labels
9428 Align all branch targets to a power-of-two boundary, skipping up to
9429 @var{n} bytes like @option{-falign-functions}. This option can easily
9430 make code slower, because it must insert dummy operations for when the
9431 branch target is reached in the usual flow of the code.
9433 @option{-fno-align-labels} and @option{-falign-labels=1} are
9434 equivalent and mean that labels are not aligned.
9436 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9437 are greater than this value, then their values are used instead.
9439 If @var{n} is not specified or is zero, use a machine-dependent default
9440 which is very likely to be @samp{1}, meaning no alignment.
9442 Enabled at levels @option{-O2}, @option{-O3}.
9445 @itemx -falign-loops=@var{n}
9446 @opindex falign-loops
9447 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9448 like @option{-falign-functions}. If the loops are
9449 executed many times, this makes up for any execution of the dummy
9452 @option{-fno-align-loops} and @option{-falign-loops=1} are
9453 equivalent and mean that loops are not aligned.
9455 If @var{n} is not specified or is zero, use a machine-dependent default.
9457 Enabled at levels @option{-O2}, @option{-O3}.
9460 @itemx -falign-jumps=@var{n}
9461 @opindex falign-jumps
9462 Align branch targets to a power-of-two boundary, for branch targets
9463 where the targets can only be reached by jumping, skipping up to @var{n}
9464 bytes like @option{-falign-functions}. In this case, no dummy operations
9467 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9468 equivalent and mean that loops are not aligned.
9470 If @var{n} is not specified or is zero, use a machine-dependent default.
9472 Enabled at levels @option{-O2}, @option{-O3}.
9474 @item -funit-at-a-time
9475 @opindex funit-at-a-time
9476 This option is left for compatibility reasons. @option{-funit-at-a-time}
9477 has no effect, while @option{-fno-unit-at-a-time} implies
9478 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9482 @item -fno-toplevel-reorder
9483 @opindex fno-toplevel-reorder
9484 Do not reorder top-level functions, variables, and @code{asm}
9485 statements. Output them in the same order that they appear in the
9486 input file. When this option is used, unreferenced static variables
9487 are not removed. This option is intended to support existing code
9488 that relies on a particular ordering. For new code, it is better to
9489 use attributes when possible.
9491 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9492 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9497 Constructs webs as commonly used for register allocation purposes and assign
9498 each web individual pseudo register. This allows the register allocation pass
9499 to operate on pseudos directly, but also strengthens several other optimization
9500 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9501 however, make debugging impossible, since variables no longer stay in a
9504 Enabled by default with @option{-funroll-loops}.
9506 @item -fwhole-program
9507 @opindex fwhole-program
9508 Assume that the current compilation unit represents the whole program being
9509 compiled. All public functions and variables with the exception of @code{main}
9510 and those merged by attribute @code{externally_visible} become static functions
9511 and in effect are optimized more aggressively by interprocedural optimizers.
9513 This option should not be used in combination with @option{-flto}.
9514 Instead relying on a linker plugin should provide safer and more precise
9517 @item -flto[=@var{n}]
9519 This option runs the standard link-time optimizer. When invoked
9520 with source code, it generates GIMPLE (one of GCC's internal
9521 representations) and writes it to special ELF sections in the object
9522 file. When the object files are linked together, all the function
9523 bodies are read from these ELF sections and instantiated as if they
9524 had been part of the same translation unit.
9526 To use the link-time optimizer, @option{-flto} and optimization
9527 options should be specified at compile time and during the final link.
9531 gcc -c -O2 -flto foo.c
9532 gcc -c -O2 -flto bar.c
9533 gcc -o myprog -flto -O2 foo.o bar.o
9536 The first two invocations to GCC save a bytecode representation
9537 of GIMPLE into special ELF sections inside @file{foo.o} and
9538 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9539 @file{foo.o} and @file{bar.o}, merges the two files into a single
9540 internal image, and compiles the result as usual. Since both
9541 @file{foo.o} and @file{bar.o} are merged into a single image, this
9542 causes all the interprocedural analyses and optimizations in GCC to
9543 work across the two files as if they were a single one. This means,
9544 for example, that the inliner is able to inline functions in
9545 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9547 Another (simpler) way to enable link-time optimization is:
9550 gcc -o myprog -flto -O2 foo.c bar.c
9553 The above generates bytecode for @file{foo.c} and @file{bar.c},
9554 merges them together into a single GIMPLE representation and optimizes
9555 them as usual to produce @file{myprog}.
9557 The only important thing to keep in mind is that to enable link-time
9558 optimizations you need to use the GCC driver to perform the link-step.
9559 GCC then automatically performs link-time optimization if any of the
9560 objects involved were compiled with the @option{-flto} command-line option.
9562 should specify the optimization options to be used for link-time
9563 optimization though GCC tries to be clever at guessing an
9564 optimization level to use from the options used at compile-time
9565 if you fail to specify one at link-time. You can always override
9566 the automatic decision to do link-time optimization at link-time
9567 by passing @option{-fno-lto} to the link command.
9569 To make whole program optimization effective, it is necessary to make
9570 certain whole program assumptions. The compiler needs to know
9571 what functions and variables can be accessed by libraries and runtime
9572 outside of the link-time optimized unit. When supported by the linker,
9573 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9574 to the compiler about used and externally visible symbols. When
9575 the linker plugin is not available, @option{-fwhole-program} should be
9576 used to allow the compiler to make these assumptions, which leads
9577 to more aggressive optimization decisions.
9579 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9580 compiled with @option{-flto}, the generated object file is larger than
9581 a regular object file because it contains GIMPLE bytecodes and the usual
9582 final code (see @option{-ffat-lto-objects}. This means that
9583 object files with LTO information can be linked as normal object
9584 files; if @option{-fno-lto} is passed to the linker, no
9585 interprocedural optimizations are applied. Note that when
9586 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9587 but you cannot perform a regular, non-LTO link on them.
9589 Additionally, the optimization flags used to compile individual files
9590 are not necessarily related to those used at link time. For instance,
9593 gcc -c -O0 -ffat-lto-objects -flto foo.c
9594 gcc -c -O0 -ffat-lto-objects -flto bar.c
9595 gcc -o myprog -O3 foo.o bar.o
9598 This produces individual object files with unoptimized assembler
9599 code, but the resulting binary @file{myprog} is optimized at
9600 @option{-O3}. If, instead, the final binary is generated with
9601 @option{-fno-lto}, then @file{myprog} is not optimized.
9603 When producing the final binary, GCC only
9604 applies link-time optimizations to those files that contain bytecode.
9605 Therefore, you can mix and match object files and libraries with
9606 GIMPLE bytecodes and final object code. GCC automatically selects
9607 which files to optimize in LTO mode and which files to link without
9610 There are some code generation flags preserved by GCC when
9611 generating bytecodes, as they need to be used during the final link
9612 stage. Generally options specified at link-time override those
9613 specified at compile-time.
9615 If you do not specify an optimization level option @option{-O} at
9616 link-time then GCC computes one based on the optimization levels
9617 used when compiling the object files. The highest optimization
9620 Currently, the following options and their setting are take from
9621 the first object file that explicitely specified it:
9622 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9623 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9624 and all the @option{-m} target flags.
9626 Certain ABI changing flags are required to match in all compilation-units
9627 and trying to override this at link-time with a conflicting value
9628 is ignored. This includes options such as @option{-freg-struct-return}
9629 and @option{-fpcc-struct-return}.
9631 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9632 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9633 are passed through to the link stage and merged conservatively for
9634 conflicting translation units. Specifically
9635 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9636 precedence and for example @option{-ffp-contract=off} takes precedence
9637 over @option{-ffp-contract=fast}. You can override them at linke-time.
9639 It is recommended that you compile all the files participating in the
9640 same link with the same options and also specify those options at
9643 If LTO encounters objects with C linkage declared with incompatible
9644 types in separate translation units to be linked together (undefined
9645 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9646 issued. The behavior is still undefined at run time. Similar
9647 diagnostics may be raised for other languages.
9649 Another feature of LTO is that it is possible to apply interprocedural
9650 optimizations on files written in different languages:
9655 gfortran -c -flto baz.f90
9656 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9659 Notice that the final link is done with @command{g++} to get the C++
9660 runtime libraries and @option{-lgfortran} is added to get the Fortran
9661 runtime libraries. In general, when mixing languages in LTO mode, you
9662 should use the same link command options as when mixing languages in a
9663 regular (non-LTO) compilation.
9665 If object files containing GIMPLE bytecode are stored in a library archive, say
9666 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9667 are using a linker with plugin support. To create static libraries suitable
9668 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9669 and @command{ranlib};
9670 to show the symbols of object files with GIMPLE bytecode, use
9671 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9672 and @command{nm} have been compiled with plugin support. At link time, use the the
9673 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9674 the LTO optimization process:
9677 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9680 With the linker plugin enabled, the linker extracts the needed
9681 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9682 to make them part of the aggregated GIMPLE image to be optimized.
9684 If you are not using a linker with plugin support and/or do not
9685 enable the linker plugin, then the objects inside @file{libfoo.a}
9686 are extracted and linked as usual, but they do not participate
9687 in the LTO optimization process. In order to make a static library suitable
9688 for both LTO optimization and usual linkage, compile its object files with
9689 @option{-flto} @option{-ffat-lto-objects}.
9691 Link-time optimizations do not require the presence of the whole program to
9692 operate. If the program does not require any symbols to be exported, it is
9693 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9694 the interprocedural optimizers to use more aggressive assumptions which may
9695 lead to improved optimization opportunities.
9696 Use of @option{-fwhole-program} is not needed when linker plugin is
9697 active (see @option{-fuse-linker-plugin}).
9699 The current implementation of LTO makes no
9700 attempt to generate bytecode that is portable between different
9701 types of hosts. The bytecode files are versioned and there is a
9702 strict version check, so bytecode files generated in one version of
9703 GCC do not work with an older or newer version of GCC.
9705 Link-time optimization does not work well with generation of debugging
9706 information. Combining @option{-flto} with
9707 @option{-g} is currently experimental and expected to produce unexpected
9710 If you specify the optional @var{n}, the optimization and code
9711 generation done at link time is executed in parallel using @var{n}
9712 parallel jobs by utilizing an installed @command{make} program. The
9713 environment variable @env{MAKE} may be used to override the program
9714 used. The default value for @var{n} is 1.
9716 You can also specify @option{-flto=jobserver} to use GNU make's
9717 job server mode to determine the number of parallel jobs. This
9718 is useful when the Makefile calling GCC is already executing in parallel.
9719 You must prepend a @samp{+} to the command recipe in the parent Makefile
9720 for this to work. This option likely only works if @env{MAKE} is
9723 @item -flto-partition=@var{alg}
9724 @opindex flto-partition
9725 Specify the partitioning algorithm used by the link-time optimizer.
9726 The value is either @samp{1to1} to specify a partitioning mirroring
9727 the original source files or @samp{balanced} to specify partitioning
9728 into equally sized chunks (whenever possible) or @samp{max} to create
9729 new partition for every symbol where possible. Specifying @samp{none}
9730 as an algorithm disables partitioning and streaming completely.
9731 The default value is @samp{balanced}. While @samp{1to1} can be used
9732 as an workaround for various code ordering issues, the @samp{max}
9733 partitioning is intended for internal testing only.
9734 The value @samp{one} specifies that exactly one partition should be
9735 used while the value @samp{none} bypasses partitioning and executes
9736 the link-time optimization step directly from the WPA phase.
9738 @item -flto-odr-type-merging
9739 @opindex flto-odr-type-merging
9740 Enable streaming of mangled types names of C++ types and their unification
9741 at linktime. This increases size of LTO object files, but enable
9742 diagnostics about One Definition Rule violations.
9744 @item -flto-compression-level=@var{n}
9745 @opindex flto-compression-level
9746 This option specifies the level of compression used for intermediate
9747 language written to LTO object files, and is only meaningful in
9748 conjunction with LTO mode (@option{-flto}). Valid
9749 values are 0 (no compression) to 9 (maximum compression). Values
9750 outside this range are clamped to either 0 or 9. If the option is not
9751 given, a default balanced compression setting is used.
9754 @opindex flto-report
9755 Prints a report with internal details on the workings of the link-time
9756 optimizer. The contents of this report vary from version to version.
9757 It is meant to be useful to GCC developers when processing object
9758 files in LTO mode (via @option{-flto}).
9760 Disabled by default.
9762 @item -flto-report-wpa
9763 @opindex flto-report-wpa
9764 Like @option{-flto-report}, but only print for the WPA phase of Link
9767 @item -fuse-linker-plugin
9768 @opindex fuse-linker-plugin
9769 Enables the use of a linker plugin during link-time optimization. This
9770 option relies on plugin support in the linker, which is available in gold
9771 or in GNU ld 2.21 or newer.
9773 This option enables the extraction of object files with GIMPLE bytecode out
9774 of library archives. This improves the quality of optimization by exposing
9775 more code to the link-time optimizer. This information specifies what
9776 symbols can be accessed externally (by non-LTO object or during dynamic
9777 linking). Resulting code quality improvements on binaries (and shared
9778 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9779 See @option{-flto} for a description of the effect of this flag and how to
9782 This option is enabled by default when LTO support in GCC is enabled
9783 and GCC was configured for use with
9784 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9786 @item -ffat-lto-objects
9787 @opindex ffat-lto-objects
9788 Fat LTO objects are object files that contain both the intermediate language
9789 and the object code. This makes them usable for both LTO linking and normal
9790 linking. This option is effective only when compiling with @option{-flto}
9791 and is ignored at link time.
9793 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9794 requires the complete toolchain to be aware of LTO. It requires a linker with
9795 linker plugin support for basic functionality. Additionally,
9796 @command{nm}, @command{ar} and @command{ranlib}
9797 need to support linker plugins to allow a full-featured build environment
9798 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9799 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9800 to these tools. With non fat LTO makefiles need to be modified to use them.
9802 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9805 @item -fcompare-elim
9806 @opindex fcompare-elim
9807 After register allocation and post-register allocation instruction splitting,
9808 identify arithmetic instructions that compute processor flags similar to a
9809 comparison operation based on that arithmetic. If possible, eliminate the
9810 explicit comparison operation.
9812 This pass only applies to certain targets that cannot explicitly represent
9813 the comparison operation before register allocation is complete.
9815 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9817 @item -fcprop-registers
9818 @opindex fcprop-registers
9819 After register allocation and post-register allocation instruction splitting,
9820 perform a copy-propagation pass to try to reduce scheduling dependencies
9821 and occasionally eliminate the copy.
9823 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9825 @item -fprofile-correction
9826 @opindex fprofile-correction
9827 Profiles collected using an instrumented binary for multi-threaded programs may
9828 be inconsistent due to missed counter updates. When this option is specified,
9829 GCC uses heuristics to correct or smooth out such inconsistencies. By
9830 default, GCC emits an error message when an inconsistent profile is detected.
9832 @item -fprofile-dir=@var{path}
9833 @opindex fprofile-dir
9835 Set the directory to search for the profile data files in to @var{path}.
9836 This option affects only the profile data generated by
9837 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9838 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9839 and its related options. Both absolute and relative paths can be used.
9840 By default, GCC uses the current directory as @var{path}, thus the
9841 profile data file appears in the same directory as the object file.
9843 @item -fprofile-generate
9844 @itemx -fprofile-generate=@var{path}
9845 @opindex fprofile-generate
9847 Enable options usually used for instrumenting application to produce
9848 profile useful for later recompilation with profile feedback based
9849 optimization. You must use @option{-fprofile-generate} both when
9850 compiling and when linking your program.
9852 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9854 If @var{path} is specified, GCC looks at the @var{path} to find
9855 the profile feedback data files. See @option{-fprofile-dir}.
9858 @itemx -fprofile-use=@var{path}
9859 @opindex fprofile-use
9860 Enable profile feedback-directed optimizations,
9861 and the following optimizations
9862 which are generally profitable only with profile feedback available:
9863 @option{-fbranch-probabilities}, @option{-fvpt},
9864 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9865 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9867 By default, GCC emits an error message if the feedback profiles do not
9868 match the source code. This error can be turned into a warning by using
9869 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9872 If @var{path} is specified, GCC looks at the @var{path} to find
9873 the profile feedback data files. See @option{-fprofile-dir}.
9875 @item -fauto-profile
9876 @itemx -fauto-profile=@var{path}
9877 @opindex fauto-profile
9878 Enable sampling-based feedback-directed optimizations,
9879 and the following optimizations
9880 which are generally profitable only with profile feedback available:
9881 @option{-fbranch-probabilities}, @option{-fvpt},
9882 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9883 @option{-ftree-vectorize},
9884 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9885 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9886 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9888 @var{path} is the name of a file containing AutoFDO profile information.
9889 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9891 Producing an AutoFDO profile data file requires running your program
9892 with the @command{perf} utility on a supported GNU/Linux target system.
9893 For more information, see @uref{https://perf.wiki.kernel.org/}.
9897 perf record -e br_inst_retired:near_taken -b -o perf.data \
9901 Then use the @command{create_gcov} tool to convert the raw profile data
9902 to a format that can be used by GCC.@ You must also supply the
9903 unstripped binary for your program to this tool.
9904 See @uref{https://github.com/google/autofdo}.
9908 create_gcov --binary=your_program.unstripped --profile=perf.data \
9913 The following options control compiler behavior regarding floating-point
9914 arithmetic. These options trade off between speed and
9915 correctness. All must be specifically enabled.
9919 @opindex ffloat-store
9920 Do not store floating-point variables in registers, and inhibit other
9921 options that might change whether a floating-point value is taken from a
9924 @cindex floating-point precision
9925 This option prevents undesirable excess precision on machines such as
9926 the 68000 where the floating registers (of the 68881) keep more
9927 precision than a @code{double} is supposed to have. Similarly for the
9928 x86 architecture. For most programs, the excess precision does only
9929 good, but a few programs rely on the precise definition of IEEE floating
9930 point. Use @option{-ffloat-store} for such programs, after modifying
9931 them to store all pertinent intermediate computations into variables.
9933 @item -fexcess-precision=@var{style}
9934 @opindex fexcess-precision
9935 This option allows further control over excess precision on machines
9936 where floating-point registers have more precision than the IEEE
9937 @code{float} and @code{double} types and the processor does not
9938 support operations rounding to those types. By default,
9939 @option{-fexcess-precision=fast} is in effect; this means that
9940 operations are carried out in the precision of the registers and that
9941 it is unpredictable when rounding to the types specified in the source
9942 code takes place. When compiling C, if
9943 @option{-fexcess-precision=standard} is specified then excess
9944 precision follows the rules specified in ISO C99; in particular,
9945 both casts and assignments cause values to be rounded to their
9946 semantic types (whereas @option{-ffloat-store} only affects
9947 assignments). This option is enabled by default for C if a strict
9948 conformance option such as @option{-std=c99} is used.
9951 @option{-fexcess-precision=standard} is not implemented for languages
9952 other than C, and has no effect if
9953 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9954 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9955 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9956 semantics apply without excess precision, and in the latter, rounding
9961 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9962 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9963 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9965 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9967 This option is not turned on by any @option{-O} option besides
9968 @option{-Ofast} since it can result in incorrect output for programs
9969 that depend on an exact implementation of IEEE or ISO rules/specifications
9970 for math functions. It may, however, yield faster code for programs
9971 that do not require the guarantees of these specifications.
9973 @item -fno-math-errno
9974 @opindex fno-math-errno
9975 Do not set @code{errno} after calling math functions that are executed
9976 with a single instruction, e.g., @code{sqrt}. A program that relies on
9977 IEEE exceptions for math error handling may want to use this flag
9978 for speed while maintaining IEEE arithmetic compatibility.
9980 This option is not turned on by any @option{-O} option since
9981 it can result in incorrect output for programs that depend on
9982 an exact implementation of IEEE or ISO rules/specifications for
9983 math functions. It may, however, yield faster code for programs
9984 that do not require the guarantees of these specifications.
9986 The default is @option{-fmath-errno}.
9988 On Darwin systems, the math library never sets @code{errno}. There is
9989 therefore no reason for the compiler to consider the possibility that
9990 it might, and @option{-fno-math-errno} is the default.
9992 @item -funsafe-math-optimizations
9993 @opindex funsafe-math-optimizations
9995 Allow optimizations for floating-point arithmetic that (a) assume
9996 that arguments and results are valid and (b) may violate IEEE or
9997 ANSI standards. When used at link-time, it may include libraries
9998 or startup files that change the default FPU control word or other
9999 similar optimizations.
10001 This option is not turned on by any @option{-O} option since
10002 it can result in incorrect output for programs that depend on
10003 an exact implementation of IEEE or ISO rules/specifications for
10004 math functions. It may, however, yield faster code for programs
10005 that do not require the guarantees of these specifications.
10006 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
10007 @option{-fassociative-math} and @option{-freciprocal-math}.
10009 The default is @option{-fno-unsafe-math-optimizations}.
10011 @item -fassociative-math
10012 @opindex fassociative-math
10014 Allow re-association of operands in series of floating-point operations.
10015 This violates the ISO C and C++ language standard by possibly changing
10016 computation result. NOTE: re-ordering may change the sign of zero as
10017 well as ignore NaNs and inhibit or create underflow or overflow (and
10018 thus cannot be used on code that relies on rounding behavior like
10019 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
10020 and thus may not be used when ordered comparisons are required.
10021 This option requires that both @option{-fno-signed-zeros} and
10022 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
10023 much sense with @option{-frounding-math}. For Fortran the option
10024 is automatically enabled when both @option{-fno-signed-zeros} and
10025 @option{-fno-trapping-math} are in effect.
10027 The default is @option{-fno-associative-math}.
10029 @item -freciprocal-math
10030 @opindex freciprocal-math
10032 Allow the reciprocal of a value to be used instead of dividing by
10033 the value if this enables optimizations. For example @code{x / y}
10034 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10035 is subject to common subexpression elimination. Note that this loses
10036 precision and increases the number of flops operating on the value.
10038 The default is @option{-fno-reciprocal-math}.
10040 @item -ffinite-math-only
10041 @opindex ffinite-math-only
10042 Allow optimizations for floating-point arithmetic that assume
10043 that arguments and results are not NaNs or +-Infs.
10045 This option is not turned on by any @option{-O} option since
10046 it can result in incorrect output for programs that depend on
10047 an exact implementation of IEEE or ISO rules/specifications for
10048 math functions. It may, however, yield faster code for programs
10049 that do not require the guarantees of these specifications.
10051 The default is @option{-fno-finite-math-only}.
10053 @item -fno-signed-zeros
10054 @opindex fno-signed-zeros
10055 Allow optimizations for floating-point arithmetic that ignore the
10056 signedness of zero. IEEE arithmetic specifies the behavior of
10057 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10058 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10059 This option implies that the sign of a zero result isn't significant.
10061 The default is @option{-fsigned-zeros}.
10063 @item -fno-trapping-math
10064 @opindex fno-trapping-math
10065 Compile code assuming that floating-point operations cannot generate
10066 user-visible traps. These traps include division by zero, overflow,
10067 underflow, inexact result and invalid operation. This option requires
10068 that @option{-fno-signaling-nans} be in effect. Setting this option may
10069 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10071 This option should never be turned on by any @option{-O} option since
10072 it can result in incorrect output for programs that depend on
10073 an exact implementation of IEEE or ISO rules/specifications for
10076 The default is @option{-ftrapping-math}.
10078 @item -frounding-math
10079 @opindex frounding-math
10080 Disable transformations and optimizations that assume default floating-point
10081 rounding behavior. This is round-to-zero for all floating point
10082 to integer conversions, and round-to-nearest for all other arithmetic
10083 truncations. This option should be specified for programs that change
10084 the FP rounding mode dynamically, or that may be executed with a
10085 non-default rounding mode. This option disables constant folding of
10086 floating-point expressions at compile time (which may be affected by
10087 rounding mode) and arithmetic transformations that are unsafe in the
10088 presence of sign-dependent rounding modes.
10090 The default is @option{-fno-rounding-math}.
10092 This option is experimental and does not currently guarantee to
10093 disable all GCC optimizations that are affected by rounding mode.
10094 Future versions of GCC may provide finer control of this setting
10095 using C99's @code{FENV_ACCESS} pragma. This command-line option
10096 will be used to specify the default state for @code{FENV_ACCESS}.
10098 @item -fsignaling-nans
10099 @opindex fsignaling-nans
10100 Compile code assuming that IEEE signaling NaNs may generate user-visible
10101 traps during floating-point operations. Setting this option disables
10102 optimizations that may change the number of exceptions visible with
10103 signaling NaNs. This option implies @option{-ftrapping-math}.
10105 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10108 The default is @option{-fno-signaling-nans}.
10110 This option is experimental and does not currently guarantee to
10111 disable all GCC optimizations that affect signaling NaN behavior.
10113 @item -fsingle-precision-constant
10114 @opindex fsingle-precision-constant
10115 Treat floating-point constants as single precision instead of
10116 implicitly converting them to double-precision constants.
10118 @item -fcx-limited-range
10119 @opindex fcx-limited-range
10120 When enabled, this option states that a range reduction step is not
10121 needed when performing complex division. Also, there is no checking
10122 whether the result of a complex multiplication or division is @code{NaN
10123 + I*NaN}, with an attempt to rescue the situation in that case. The
10124 default is @option{-fno-cx-limited-range}, but is enabled by
10125 @option{-ffast-math}.
10127 This option controls the default setting of the ISO C99
10128 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10131 @item -fcx-fortran-rules
10132 @opindex fcx-fortran-rules
10133 Complex multiplication and division follow Fortran rules. Range
10134 reduction is done as part of complex division, but there is no checking
10135 whether the result of a complex multiplication or division is @code{NaN
10136 + I*NaN}, with an attempt to rescue the situation in that case.
10138 The default is @option{-fno-cx-fortran-rules}.
10142 The following options control optimizations that may improve
10143 performance, but are not enabled by any @option{-O} options. This
10144 section includes experimental options that may produce broken code.
10147 @item -fbranch-probabilities
10148 @opindex fbranch-probabilities
10149 After running a program compiled with @option{-fprofile-arcs}
10150 (@pxref{Debugging Options,, Options for Debugging Your Program or
10151 @command{gcc}}), you can compile it a second time using
10152 @option{-fbranch-probabilities}, to improve optimizations based on
10153 the number of times each branch was taken. When a program
10154 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10155 counts to a file called @file{@var{sourcename}.gcda} for each source
10156 file. The information in this data file is very dependent on the
10157 structure of the generated code, so you must use the same source code
10158 and the same optimization options for both compilations.
10160 With @option{-fbranch-probabilities}, GCC puts a
10161 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10162 These can be used to improve optimization. Currently, they are only
10163 used in one place: in @file{reorg.c}, instead of guessing which path a
10164 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10165 exactly determine which path is taken more often.
10167 @item -fprofile-values
10168 @opindex fprofile-values
10169 If combined with @option{-fprofile-arcs}, it adds code so that some
10170 data about values of expressions in the program is gathered.
10172 With @option{-fbranch-probabilities}, it reads back the data gathered
10173 from profiling values of expressions for usage in optimizations.
10175 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
10177 @item -fprofile-reorder-functions
10178 @opindex fprofile-reorder-functions
10179 Function reordering based on profile instrumentation collects
10180 first time of execution of a function and orders these functions
10181 in ascending order.
10183 Enabled with @option{-fprofile-use}.
10187 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10188 to add code to gather information about values of expressions.
10190 With @option{-fbranch-probabilities}, it reads back the data gathered
10191 and actually performs the optimizations based on them.
10192 Currently the optimizations include specialization of division operations
10193 using the knowledge about the value of the denominator.
10195 @item -frename-registers
10196 @opindex frename-registers
10197 Attempt to avoid false dependencies in scheduled code by making use
10198 of registers left over after register allocation. This optimization
10199 most benefits processors with lots of registers. Depending on the
10200 debug information format adopted by the target, however, it can
10201 make debugging impossible, since variables no longer stay in
10202 a ``home register''.
10204 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
10206 @item -fschedule-fusion
10207 @opindex fschedule-fusion
10208 Performs a target dependent pass over the instruction stream to schedule
10209 instructions of same type together because target machine can execute them
10210 more efficiently if they are adjacent to each other in the instruction flow.
10212 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10216 Perform tail duplication to enlarge superblock size. This transformation
10217 simplifies the control flow of the function allowing other optimizations to do
10220 Enabled with @option{-fprofile-use}.
10222 @item -funroll-loops
10223 @opindex funroll-loops
10224 Unroll loops whose number of iterations can be determined at compile time or
10225 upon entry to the loop. @option{-funroll-loops} implies
10226 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10227 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10228 a small constant number of iterations). This option makes code larger, and may
10229 or may not make it run faster.
10231 Enabled with @option{-fprofile-use}.
10233 @item -funroll-all-loops
10234 @opindex funroll-all-loops
10235 Unroll all loops, even if their number of iterations is uncertain when
10236 the loop is entered. This usually makes programs run more slowly.
10237 @option{-funroll-all-loops} implies the same options as
10238 @option{-funroll-loops}.
10241 @opindex fpeel-loops
10242 Peels loops for which there is enough information that they do not
10243 roll much (from profile feedback). It also turns on complete loop peeling
10244 (i.e.@: complete removal of loops with small constant number of iterations).
10246 Enabled with @option{-fprofile-use}.
10248 @item -fmove-loop-invariants
10249 @opindex fmove-loop-invariants
10250 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10251 at level @option{-O1}
10253 @item -funswitch-loops
10254 @opindex funswitch-loops
10255 Move branches with loop invariant conditions out of the loop, with duplicates
10256 of the loop on both branches (modified according to result of the condition).
10258 @item -ffunction-sections
10259 @itemx -fdata-sections
10260 @opindex ffunction-sections
10261 @opindex fdata-sections
10262 Place each function or data item into its own section in the output
10263 file if the target supports arbitrary sections. The name of the
10264 function or the name of the data item determines the section's name
10265 in the output file.
10267 Use these options on systems where the linker can perform optimizations
10268 to improve locality of reference in the instruction space. Most systems
10269 using the ELF object format and SPARC processors running Solaris 2 have
10270 linkers with such optimizations. AIX may have these optimizations in
10273 Only use these options when there are significant benefits from doing
10274 so. When you specify these options, the assembler and linker
10275 create larger object and executable files and are also slower.
10276 You cannot use @command{gprof} on all systems if you
10277 specify this option, and you may have problems with debugging if
10278 you specify both this option and @option{-g}.
10280 @item -fbranch-target-load-optimize
10281 @opindex fbranch-target-load-optimize
10282 Perform branch target register load optimization before prologue / epilogue
10284 The use of target registers can typically be exposed only during reload,
10285 thus hoisting loads out of loops and doing inter-block scheduling needs
10286 a separate optimization pass.
10288 @item -fbranch-target-load-optimize2
10289 @opindex fbranch-target-load-optimize2
10290 Perform branch target register load optimization after prologue / epilogue
10293 @item -fbtr-bb-exclusive
10294 @opindex fbtr-bb-exclusive
10295 When performing branch target register load optimization, don't reuse
10296 branch target registers within any basic block.
10298 @item -fstack-protector
10299 @opindex fstack-protector
10300 Emit extra code to check for buffer overflows, such as stack smashing
10301 attacks. This is done by adding a guard variable to functions with
10302 vulnerable objects. This includes functions that call @code{alloca}, and
10303 functions with buffers larger than 8 bytes. The guards are initialized
10304 when a function is entered and then checked when the function exits.
10305 If a guard check fails, an error message is printed and the program exits.
10307 @item -fstack-protector-all
10308 @opindex fstack-protector-all
10309 Like @option{-fstack-protector} except that all functions are protected.
10311 @item -fstack-protector-strong
10312 @opindex fstack-protector-strong
10313 Like @option{-fstack-protector} but includes additional functions to
10314 be protected --- those that have local array definitions, or have
10315 references to local frame addresses.
10317 @item -fstack-protector-explicit
10318 @opindex fstack-protector-explicit
10319 Like @option{-fstack-protector} but only protects those functions which
10320 have the @code{stack_protect} attribute.
10323 @opindex fstdarg-opt
10324 Optimize the prologue of variadic argument functions with respect to usage of
10327 @item -fsection-anchors
10328 @opindex fsection-anchors
10329 Try to reduce the number of symbolic address calculations by using
10330 shared ``anchor'' symbols to address nearby objects. This transformation
10331 can help to reduce the number of GOT entries and GOT accesses on some
10334 For example, the implementation of the following function @code{foo}:
10337 static int a, b, c;
10338 int foo (void) @{ return a + b + c; @}
10342 usually calculates the addresses of all three variables, but if you
10343 compile it with @option{-fsection-anchors}, it accesses the variables
10344 from a common anchor point instead. The effect is similar to the
10345 following pseudocode (which isn't valid C):
10350 register int *xr = &x;
10351 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10355 Not all targets support this option.
10357 @item --param @var{name}=@var{value}
10359 In some places, GCC uses various constants to control the amount of
10360 optimization that is done. For example, GCC does not inline functions
10361 that contain more than a certain number of instructions. You can
10362 control some of these constants on the command line using the
10363 @option{--param} option.
10365 The names of specific parameters, and the meaning of the values, are
10366 tied to the internals of the compiler, and are subject to change
10367 without notice in future releases.
10369 In each case, the @var{value} is an integer. The allowable choices for
10373 @item predictable-branch-outcome
10374 When branch is predicted to be taken with probability lower than this threshold
10375 (in percent), then it is considered well predictable. The default is 10.
10377 @item max-rtl-if-conversion-insns
10378 RTL if-conversion tries to remove conditional branches around a block and
10379 replace them with conditionally executed instructions. This parameter
10380 gives the maximum number of instructions in a block which should be
10381 considered for if-conversion. The default is 10, though the compiler will
10382 also use other heuristics to decide whether if-conversion is likely to be
10385 @item max-crossjump-edges
10386 The maximum number of incoming edges to consider for cross-jumping.
10387 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10388 the number of edges incoming to each block. Increasing values mean
10389 more aggressive optimization, making the compilation time increase with
10390 probably small improvement in executable size.
10392 @item min-crossjump-insns
10393 The minimum number of instructions that must be matched at the end
10394 of two blocks before cross-jumping is performed on them. This
10395 value is ignored in the case where all instructions in the block being
10396 cross-jumped from are matched. The default value is 5.
10398 @item max-grow-copy-bb-insns
10399 The maximum code size expansion factor when copying basic blocks
10400 instead of jumping. The expansion is relative to a jump instruction.
10401 The default value is 8.
10403 @item max-goto-duplication-insns
10404 The maximum number of instructions to duplicate to a block that jumps
10405 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10406 passes, GCC factors computed gotos early in the compilation process,
10407 and unfactors them as late as possible. Only computed jumps at the
10408 end of a basic blocks with no more than max-goto-duplication-insns are
10409 unfactored. The default value is 8.
10411 @item max-delay-slot-insn-search
10412 The maximum number of instructions to consider when looking for an
10413 instruction to fill a delay slot. If more than this arbitrary number of
10414 instructions are searched, the time savings from filling the delay slot
10415 are minimal, so stop searching. Increasing values mean more
10416 aggressive optimization, making the compilation time increase with probably
10417 small improvement in execution time.
10419 @item max-delay-slot-live-search
10420 When trying to fill delay slots, the maximum number of instructions to
10421 consider when searching for a block with valid live register
10422 information. Increasing this arbitrarily chosen value means more
10423 aggressive optimization, increasing the compilation time. This parameter
10424 should be removed when the delay slot code is rewritten to maintain the
10425 control-flow graph.
10427 @item max-gcse-memory
10428 The approximate maximum amount of memory that can be allocated in
10429 order to perform the global common subexpression elimination
10430 optimization. If more memory than specified is required, the
10431 optimization is not done.
10433 @item max-gcse-insertion-ratio
10434 If the ratio of expression insertions to deletions is larger than this value
10435 for any expression, then RTL PRE inserts or removes the expression and thus
10436 leaves partially redundant computations in the instruction stream. The default value is 20.
10438 @item max-pending-list-length
10439 The maximum number of pending dependencies scheduling allows
10440 before flushing the current state and starting over. Large functions
10441 with few branches or calls can create excessively large lists which
10442 needlessly consume memory and resources.
10444 @item max-modulo-backtrack-attempts
10445 The maximum number of backtrack attempts the scheduler should make
10446 when modulo scheduling a loop. Larger values can exponentially increase
10449 @item max-inline-insns-single
10450 Several parameters control the tree inliner used in GCC@.
10451 This number sets the maximum number of instructions (counted in GCC's
10452 internal representation) in a single function that the tree inliner
10453 considers for inlining. This only affects functions declared
10454 inline and methods implemented in a class declaration (C++).
10455 The default value is 400.
10457 @item max-inline-insns-auto
10458 When you use @option{-finline-functions} (included in @option{-O3}),
10459 a lot of functions that would otherwise not be considered for inlining
10460 by the compiler are investigated. To those functions, a different
10461 (more restrictive) limit compared to functions declared inline can
10463 The default value is 40.
10465 @item inline-min-speedup
10466 When estimated performance improvement of caller + callee runtime exceeds this
10467 threshold (in precent), the function can be inlined regardless the limit on
10468 @option{--param max-inline-insns-single} and @option{--param
10469 max-inline-insns-auto}.
10471 @item large-function-insns
10472 The limit specifying really large functions. For functions larger than this
10473 limit after inlining, inlining is constrained by
10474 @option{--param large-function-growth}. This parameter is useful primarily
10475 to avoid extreme compilation time caused by non-linear algorithms used by the
10477 The default value is 2700.
10479 @item large-function-growth
10480 Specifies maximal growth of large function caused by inlining in percents.
10481 The default value is 100 which limits large function growth to 2.0 times
10484 @item large-unit-insns
10485 The limit specifying large translation unit. Growth caused by inlining of
10486 units larger than this limit is limited by @option{--param inline-unit-growth}.
10487 For small units this might be too tight.
10488 For example, consider a unit consisting of function A
10489 that is inline and B that just calls A three times. If B is small relative to
10490 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10491 large units consisting of small inlineable functions, however, the overall unit
10492 growth limit is needed to avoid exponential explosion of code size. Thus for
10493 smaller units, the size is increased to @option{--param large-unit-insns}
10494 before applying @option{--param inline-unit-growth}. The default is 10000.
10496 @item inline-unit-growth
10497 Specifies maximal overall growth of the compilation unit caused by inlining.
10498 The default value is 20 which limits unit growth to 1.2 times the original
10499 size. Cold functions (either marked cold via an attribute or by profile
10500 feedback) are not accounted into the unit size.
10502 @item ipcp-unit-growth
10503 Specifies maximal overall growth of the compilation unit caused by
10504 interprocedural constant propagation. The default value is 10 which limits
10505 unit growth to 1.1 times the original size.
10507 @item large-stack-frame
10508 The limit specifying large stack frames. While inlining the algorithm is trying
10509 to not grow past this limit too much. The default value is 256 bytes.
10511 @item large-stack-frame-growth
10512 Specifies maximal growth of large stack frames caused by inlining in percents.
10513 The default value is 1000 which limits large stack frame growth to 11 times
10516 @item max-inline-insns-recursive
10517 @itemx max-inline-insns-recursive-auto
10518 Specifies the maximum number of instructions an out-of-line copy of a
10519 self-recursive inline
10520 function can grow into by performing recursive inlining.
10522 @option{--param max-inline-insns-recursive} applies to functions
10524 For functions not declared inline, recursive inlining
10525 happens only when @option{-finline-functions} (included in @option{-O3}) is
10526 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10527 default value is 450.
10529 @item max-inline-recursive-depth
10530 @itemx max-inline-recursive-depth-auto
10531 Specifies the maximum recursion depth used for recursive inlining.
10533 @option{--param max-inline-recursive-depth} applies to functions
10534 declared inline. For functions not declared inline, recursive inlining
10535 happens only when @option{-finline-functions} (included in @option{-O3}) is
10536 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10537 default value is 8.
10539 @item min-inline-recursive-probability
10540 Recursive inlining is profitable only for function having deep recursion
10541 in average and can hurt for function having little recursion depth by
10542 increasing the prologue size or complexity of function body to other
10545 When profile feedback is available (see @option{-fprofile-generate}) the actual
10546 recursion depth can be guessed from probability that function recurses via a
10547 given call expression. This parameter limits inlining only to call expressions
10548 whose probability exceeds the given threshold (in percents).
10549 The default value is 10.
10551 @item early-inlining-insns
10552 Specify growth that the early inliner can make. In effect it increases
10553 the amount of inlining for code having a large abstraction penalty.
10554 The default value is 14.
10556 @item max-early-inliner-iterations
10557 Limit of iterations of the early inliner. This basically bounds
10558 the number of nested indirect calls the early inliner can resolve.
10559 Deeper chains are still handled by late inlining.
10561 @item comdat-sharing-probability
10562 Probability (in percent) that C++ inline function with comdat visibility
10563 are shared across multiple compilation units. The default value is 20.
10565 @item profile-func-internal-id
10566 A parameter to control whether to use function internal id in profile
10567 database lookup. If the value is 0, the compiler uses an id that
10568 is based on function assembler name and filename, which makes old profile
10569 data more tolerant to source changes such as function reordering etc.
10570 The default value is 0.
10572 @item min-vect-loop-bound
10573 The minimum number of iterations under which loops are not vectorized
10574 when @option{-ftree-vectorize} is used. The number of iterations after
10575 vectorization needs to be greater than the value specified by this option
10576 to allow vectorization. The default value is 0.
10578 @item gcse-cost-distance-ratio
10579 Scaling factor in calculation of maximum distance an expression
10580 can be moved by GCSE optimizations. This is currently supported only in the
10581 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10582 is with simple expressions, i.e., the expressions that have cost
10583 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10584 hoisting of simple expressions. The default value is 10.
10586 @item gcse-unrestricted-cost
10587 Cost, roughly measured as the cost of a single typical machine
10588 instruction, at which GCSE optimizations do not constrain
10589 the distance an expression can travel. This is currently
10590 supported only in the code hoisting pass. The lesser the cost,
10591 the more aggressive code hoisting is. Specifying 0
10592 allows all expressions to travel unrestricted distances.
10593 The default value is 3.
10595 @item max-hoist-depth
10596 The depth of search in the dominator tree for expressions to hoist.
10597 This is used to avoid quadratic behavior in hoisting algorithm.
10598 The value of 0 does not limit on the search, but may slow down compilation
10599 of huge functions. The default value is 30.
10601 @item max-tail-merge-comparisons
10602 The maximum amount of similar bbs to compare a bb with. This is used to
10603 avoid quadratic behavior in tree tail merging. The default value is 10.
10605 @item max-tail-merge-iterations
10606 The maximum amount of iterations of the pass over the function. This is used to
10607 limit compilation time in tree tail merging. The default value is 2.
10609 @item max-unrolled-insns
10610 The maximum number of instructions that a loop may have to be unrolled.
10611 If a loop is unrolled, this parameter also determines how many times
10612 the loop code is unrolled.
10614 @item max-average-unrolled-insns
10615 The maximum number of instructions biased by probabilities of their execution
10616 that a loop may have to be unrolled. If a loop is unrolled,
10617 this parameter also determines how many times the loop code is unrolled.
10619 @item max-unroll-times
10620 The maximum number of unrollings of a single loop.
10622 @item max-peeled-insns
10623 The maximum number of instructions that a loop may have to be peeled.
10624 If a loop is peeled, this parameter also determines how many times
10625 the loop code is peeled.
10627 @item max-peel-times
10628 The maximum number of peelings of a single loop.
10630 @item max-peel-branches
10631 The maximum number of branches on the hot path through the peeled sequence.
10633 @item max-completely-peeled-insns
10634 The maximum number of insns of a completely peeled loop.
10636 @item max-completely-peel-times
10637 The maximum number of iterations of a loop to be suitable for complete peeling.
10639 @item max-completely-peel-loop-nest-depth
10640 The maximum depth of a loop nest suitable for complete peeling.
10642 @item max-unswitch-insns
10643 The maximum number of insns of an unswitched loop.
10645 @item max-unswitch-level
10646 The maximum number of branches unswitched in a single loop.
10648 @item lim-expensive
10649 The minimum cost of an expensive expression in the loop invariant motion.
10651 @item iv-consider-all-candidates-bound
10652 Bound on number of candidates for induction variables, below which
10653 all candidates are considered for each use in induction variable
10654 optimizations. If there are more candidates than this,
10655 only the most relevant ones are considered to avoid quadratic time complexity.
10657 @item iv-max-considered-uses
10658 The induction variable optimizations give up on loops that contain more
10659 induction variable uses.
10661 @item iv-always-prune-cand-set-bound
10662 If the number of candidates in the set is smaller than this value,
10663 always try to remove unnecessary ivs from the set
10664 when adding a new one.
10666 @item scev-max-expr-size
10667 Bound on size of expressions used in the scalar evolutions analyzer.
10668 Large expressions slow the analyzer.
10670 @item scev-max-expr-complexity
10671 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10672 Complex expressions slow the analyzer.
10674 @item vect-max-version-for-alignment-checks
10675 The maximum number of run-time checks that can be performed when
10676 doing loop versioning for alignment in the vectorizer.
10678 @item vect-max-version-for-alias-checks
10679 The maximum number of run-time checks that can be performed when
10680 doing loop versioning for alias in the vectorizer.
10682 @item vect-max-peeling-for-alignment
10683 The maximum number of loop peels to enhance access alignment
10684 for vectorizer. Value -1 means 'no limit'.
10686 @item max-iterations-to-track
10687 The maximum number of iterations of a loop the brute-force algorithm
10688 for analysis of the number of iterations of the loop tries to evaluate.
10690 @item hot-bb-count-ws-permille
10691 A basic block profile count is considered hot if it contributes to
10692 the given permillage (i.e. 0...1000) of the entire profiled execution.
10694 @item hot-bb-frequency-fraction
10695 Select fraction of the entry block frequency of executions of basic block in
10696 function given basic block needs to have to be considered hot.
10698 @item max-predicted-iterations
10699 The maximum number of loop iterations we predict statically. This is useful
10700 in cases where a function contains a single loop with known bound and
10701 another loop with unknown bound.
10702 The known number of iterations is predicted correctly, while
10703 the unknown number of iterations average to roughly 10. This means that the
10704 loop without bounds appears artificially cold relative to the other one.
10706 @item builtin-expect-probability
10707 Control the probability of the expression having the specified value. This
10708 parameter takes a percentage (i.e. 0 ... 100) as input.
10709 The default probability of 90 is obtained empirically.
10711 @item align-threshold
10713 Select fraction of the maximal frequency of executions of a basic block in
10714 a function to align the basic block.
10716 @item align-loop-iterations
10718 A loop expected to iterate at least the selected number of iterations is
10721 @item tracer-dynamic-coverage
10722 @itemx tracer-dynamic-coverage-feedback
10724 This value is used to limit superblock formation once the given percentage of
10725 executed instructions is covered. This limits unnecessary code size
10728 The @option{tracer-dynamic-coverage-feedback} parameter
10729 is used only when profile
10730 feedback is available. The real profiles (as opposed to statically estimated
10731 ones) are much less balanced allowing the threshold to be larger value.
10733 @item tracer-max-code-growth
10734 Stop tail duplication once code growth has reached given percentage. This is
10735 a rather artificial limit, as most of the duplicates are eliminated later in
10736 cross jumping, so it may be set to much higher values than is the desired code
10739 @item tracer-min-branch-ratio
10741 Stop reverse growth when the reverse probability of best edge is less than this
10742 threshold (in percent).
10744 @item tracer-min-branch-ratio
10745 @itemx tracer-min-branch-ratio-feedback
10747 Stop forward growth if the best edge has probability lower than this
10750 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10751 compilation for profile feedback and one for compilation without. The value
10752 for compilation with profile feedback needs to be more conservative (higher) in
10753 order to make tracer effective.
10755 @item max-cse-path-length
10757 The maximum number of basic blocks on path that CSE considers.
10760 @item max-cse-insns
10761 The maximum number of instructions CSE processes before flushing.
10762 The default is 1000.
10764 @item ggc-min-expand
10766 GCC uses a garbage collector to manage its own memory allocation. This
10767 parameter specifies the minimum percentage by which the garbage
10768 collector's heap should be allowed to expand between collections.
10769 Tuning this may improve compilation speed; it has no effect on code
10772 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10773 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10774 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10775 GCC is not able to calculate RAM on a particular platform, the lower
10776 bound of 30% is used. Setting this parameter and
10777 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10778 every opportunity. This is extremely slow, but can be useful for
10781 @item ggc-min-heapsize
10783 Minimum size of the garbage collector's heap before it begins bothering
10784 to collect garbage. The first collection occurs after the heap expands
10785 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10786 tuning this may improve compilation speed, and has no effect on code
10789 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10790 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10791 with a lower bound of 4096 (four megabytes) and an upper bound of
10792 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10793 particular platform, the lower bound is used. Setting this parameter
10794 very large effectively disables garbage collection. Setting this
10795 parameter and @option{ggc-min-expand} to zero causes a full collection
10796 to occur at every opportunity.
10798 @item max-reload-search-insns
10799 The maximum number of instruction reload should look backward for equivalent
10800 register. Increasing values mean more aggressive optimization, making the
10801 compilation time increase with probably slightly better performance.
10802 The default value is 100.
10804 @item max-cselib-memory-locations
10805 The maximum number of memory locations cselib should take into account.
10806 Increasing values mean more aggressive optimization, making the compilation time
10807 increase with probably slightly better performance. The default value is 500.
10809 @item reorder-blocks-duplicate
10810 @itemx reorder-blocks-duplicate-feedback
10812 Used by the basic block reordering pass to decide whether to use unconditional
10813 branch or duplicate the code on its destination. Code is duplicated when its
10814 estimated size is smaller than this value multiplied by the estimated size of
10815 unconditional jump in the hot spots of the program.
10817 The @option{reorder-block-duplicate-feedback} parameter
10818 is used only when profile
10819 feedback is available. It may be set to higher values than
10820 @option{reorder-block-duplicate} since information about the hot spots is more
10823 @item max-sched-ready-insns
10824 The maximum number of instructions ready to be issued the scheduler should
10825 consider at any given time during the first scheduling pass. Increasing
10826 values mean more thorough searches, making the compilation time increase
10827 with probably little benefit. The default value is 100.
10829 @item max-sched-region-blocks
10830 The maximum number of blocks in a region to be considered for
10831 interblock scheduling. The default value is 10.
10833 @item max-pipeline-region-blocks
10834 The maximum number of blocks in a region to be considered for
10835 pipelining in the selective scheduler. The default value is 15.
10837 @item max-sched-region-insns
10838 The maximum number of insns in a region to be considered for
10839 interblock scheduling. The default value is 100.
10841 @item max-pipeline-region-insns
10842 The maximum number of insns in a region to be considered for
10843 pipelining in the selective scheduler. The default value is 200.
10845 @item min-spec-prob
10846 The minimum probability (in percents) of reaching a source block
10847 for interblock speculative scheduling. The default value is 40.
10849 @item max-sched-extend-regions-iters
10850 The maximum number of iterations through CFG to extend regions.
10851 A value of 0 (the default) disables region extensions.
10853 @item max-sched-insn-conflict-delay
10854 The maximum conflict delay for an insn to be considered for speculative motion.
10855 The default value is 3.
10857 @item sched-spec-prob-cutoff
10858 The minimal probability of speculation success (in percents), so that
10859 speculative insns are scheduled.
10860 The default value is 40.
10862 @item sched-spec-state-edge-prob-cutoff
10863 The minimum probability an edge must have for the scheduler to save its
10865 The default value is 10.
10867 @item sched-mem-true-dep-cost
10868 Minimal distance (in CPU cycles) between store and load targeting same
10869 memory locations. The default value is 1.
10871 @item selsched-max-lookahead
10872 The maximum size of the lookahead window of selective scheduling. It is a
10873 depth of search for available instructions.
10874 The default value is 50.
10876 @item selsched-max-sched-times
10877 The maximum number of times that an instruction is scheduled during
10878 selective scheduling. This is the limit on the number of iterations
10879 through which the instruction may be pipelined. The default value is 2.
10881 @item selsched-max-insns-to-rename
10882 The maximum number of best instructions in the ready list that are considered
10883 for renaming in the selective scheduler. The default value is 2.
10886 The minimum value of stage count that swing modulo scheduler
10887 generates. The default value is 2.
10889 @item max-last-value-rtl
10890 The maximum size measured as number of RTLs that can be recorded in an expression
10891 in combiner for a pseudo register as last known value of that register. The default
10894 @item max-combine-insns
10895 The maximum number of instructions the RTL combiner tries to combine.
10896 The default value is 2 at @option{-Og} and 4 otherwise.
10898 @item integer-share-limit
10899 Small integer constants can use a shared data structure, reducing the
10900 compiler's memory usage and increasing its speed. This sets the maximum
10901 value of a shared integer constant. The default value is 256.
10903 @item ssp-buffer-size
10904 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10905 protection when @option{-fstack-protection} is used.
10907 @item min-size-for-stack-sharing
10908 The minimum size of variables taking part in stack slot sharing when not
10909 optimizing. The default value is 32.
10911 @item max-jump-thread-duplication-stmts
10912 Maximum number of statements allowed in a block that needs to be
10913 duplicated when threading jumps.
10915 @item max-fields-for-field-sensitive
10916 Maximum number of fields in a structure treated in
10917 a field sensitive manner during pointer analysis. The default is zero
10918 for @option{-O0} and @option{-O1},
10919 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10921 @item prefetch-latency
10922 Estimate on average number of instructions that are executed before
10923 prefetch finishes. The distance prefetched ahead is proportional
10924 to this constant. Increasing this number may also lead to less
10925 streams being prefetched (see @option{simultaneous-prefetches}).
10927 @item simultaneous-prefetches
10928 Maximum number of prefetches that can run at the same time.
10930 @item l1-cache-line-size
10931 The size of cache line in L1 cache, in bytes.
10933 @item l1-cache-size
10934 The size of L1 cache, in kilobytes.
10936 @item l2-cache-size
10937 The size of L2 cache, in kilobytes.
10939 @item min-insn-to-prefetch-ratio
10940 The minimum ratio between the number of instructions and the
10941 number of prefetches to enable prefetching in a loop.
10943 @item prefetch-min-insn-to-mem-ratio
10944 The minimum ratio between the number of instructions and the
10945 number of memory references to enable prefetching in a loop.
10947 @item use-canonical-types
10948 Whether the compiler should use the ``canonical'' type system. By
10949 default, this should always be 1, which uses a more efficient internal
10950 mechanism for comparing types in C++ and Objective-C++. However, if
10951 bugs in the canonical type system are causing compilation failures,
10952 set this value to 0 to disable canonical types.
10954 @item switch-conversion-max-branch-ratio
10955 Switch initialization conversion refuses to create arrays that are
10956 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10957 branches in the switch.
10959 @item max-partial-antic-length
10960 Maximum length of the partial antic set computed during the tree
10961 partial redundancy elimination optimization (@option{-ftree-pre}) when
10962 optimizing at @option{-O3} and above. For some sorts of source code
10963 the enhanced partial redundancy elimination optimization can run away,
10964 consuming all of the memory available on the host machine. This
10965 parameter sets a limit on the length of the sets that are computed,
10966 which prevents the runaway behavior. Setting a value of 0 for
10967 this parameter allows an unlimited set length.
10969 @item sccvn-max-scc-size
10970 Maximum size of a strongly connected component (SCC) during SCCVN
10971 processing. If this limit is hit, SCCVN processing for the whole
10972 function is not done and optimizations depending on it are
10973 disabled. The default maximum SCC size is 10000.
10975 @item sccvn-max-alias-queries-per-access
10976 Maximum number of alias-oracle queries we perform when looking for
10977 redundancies for loads and stores. If this limit is hit the search
10978 is aborted and the load or store is not considered redundant. The
10979 number of queries is algorithmically limited to the number of
10980 stores on all paths from the load to the function entry.
10981 The default maxmimum number of queries is 1000.
10983 @item ira-max-loops-num
10984 IRA uses regional register allocation by default. If a function
10985 contains more loops than the number given by this parameter, only at most
10986 the given number of the most frequently-executed loops form regions
10987 for regional register allocation. The default value of the
10990 @item ira-max-conflict-table-size
10991 Although IRA uses a sophisticated algorithm to compress the conflict
10992 table, the table can still require excessive amounts of memory for
10993 huge functions. If the conflict table for a function could be more
10994 than the size in MB given by this parameter, the register allocator
10995 instead uses a faster, simpler, and lower-quality
10996 algorithm that does not require building a pseudo-register conflict table.
10997 The default value of the parameter is 2000.
10999 @item ira-loop-reserved-regs
11000 IRA can be used to evaluate more accurate register pressure in loops
11001 for decisions to move loop invariants (see @option{-O3}). The number
11002 of available registers reserved for some other purposes is given
11003 by this parameter. The default value of the parameter is 2, which is
11004 the minimal number of registers needed by typical instructions.
11005 This value is the best found from numerous experiments.
11007 @item lra-inheritance-ebb-probability-cutoff
11008 LRA tries to reuse values reloaded in registers in subsequent insns.
11009 This optimization is called inheritance. EBB is used as a region to
11010 do this optimization. The parameter defines a minimal fall-through
11011 edge probability in percentage used to add BB to inheritance EBB in
11012 LRA. The default value of the parameter is 40. The value was chosen
11013 from numerous runs of SPEC2000 on x86-64.
11015 @item loop-invariant-max-bbs-in-loop
11016 Loop invariant motion can be very expensive, both in compilation time and
11017 in amount of needed compile-time memory, with very large loops. Loops
11018 with more basic blocks than this parameter won't have loop invariant
11019 motion optimization performed on them. The default value of the
11020 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
11022 @item loop-max-datarefs-for-datadeps
11023 Building data dapendencies is expensive for very large loops. This
11024 parameter limits the number of data references in loops that are
11025 considered for data dependence analysis. These large loops are no
11026 handled by the optimizations using loop data dependencies.
11027 The default value is 1000.
11029 @item max-vartrack-size
11030 Sets a maximum number of hash table slots to use during variable
11031 tracking dataflow analysis of any function. If this limit is exceeded
11032 with variable tracking at assignments enabled, analysis for that
11033 function is retried without it, after removing all debug insns from
11034 the function. If the limit is exceeded even without debug insns, var
11035 tracking analysis is completely disabled for the function. Setting
11036 the parameter to zero makes it unlimited.
11038 @item max-vartrack-expr-depth
11039 Sets a maximum number of recursion levels when attempting to map
11040 variable names or debug temporaries to value expressions. This trades
11041 compilation time for more complete debug information. If this is set too
11042 low, value expressions that are available and could be represented in
11043 debug information may end up not being used; setting this higher may
11044 enable the compiler to find more complex debug expressions, but compile
11045 time and memory use may grow. The default is 12.
11047 @item min-nondebug-insn-uid
11048 Use uids starting at this parameter for nondebug insns. The range below
11049 the parameter is reserved exclusively for debug insns created by
11050 @option{-fvar-tracking-assignments}, but debug insns may get
11051 (non-overlapping) uids above it if the reserved range is exhausted.
11053 @item ipa-sra-ptr-growth-factor
11054 IPA-SRA replaces a pointer to an aggregate with one or more new
11055 parameters only when their cumulative size is less or equal to
11056 @option{ipa-sra-ptr-growth-factor} times the size of the original
11059 @item sra-max-scalarization-size-Ospeed
11060 @item sra-max-scalarization-size-Osize
11061 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11062 replace scalar parts of aggregates with uses of independent scalar
11063 variables. These parameters control the maximum size, in storage units,
11064 of aggregate which is considered for replacement when compiling for
11066 (@option{sra-max-scalarization-size-Ospeed}) or size
11067 (@option{sra-max-scalarization-size-Osize}) respectively.
11069 @item tm-max-aggregate-size
11070 When making copies of thread-local variables in a transaction, this
11071 parameter specifies the size in bytes after which variables are
11072 saved with the logging functions as opposed to save/restore code
11073 sequence pairs. This option only applies when using
11076 @item graphite-max-nb-scop-params
11077 To avoid exponential effects in the Graphite loop transforms, the
11078 number of parameters in a Static Control Part (SCoP) is bounded. The
11079 default value is 10 parameters. A variable whose value is unknown at
11080 compilation time and defined outside a SCoP is a parameter of the SCoP.
11082 @item graphite-max-bbs-per-function
11083 To avoid exponential effects in the detection of SCoPs, the size of
11084 the functions analyzed by Graphite is bounded. The default value is
11087 @item loop-block-tile-size
11088 Loop blocking or strip mining transforms, enabled with
11089 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11090 loop in the loop nest by a given number of iterations. The strip
11091 length can be changed using the @option{loop-block-tile-size}
11092 parameter. The default value is 51 iterations.
11094 @item loop-unroll-jam-size
11095 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
11096 default value is 4.
11098 @item loop-unroll-jam-depth
11099 Specify the dimension to be unrolled (counting from the most inner loop)
11100 for the @option{-floop-unroll-and-jam}. The default value is 2.
11102 @item ipa-cp-value-list-size
11103 IPA-CP attempts to track all possible values and types passed to a function's
11104 parameter in order to propagate them and perform devirtualization.
11105 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11106 stores per one formal parameter of a function.
11108 @item ipa-cp-eval-threshold
11109 IPA-CP calculates its own score of cloning profitability heuristics
11110 and performs those cloning opportunities with scores that exceed
11111 @option{ipa-cp-eval-threshold}.
11113 @item ipa-cp-recursion-penalty
11114 Percentage penalty the recursive functions will receive when they
11115 are evaluated for cloning.
11117 @item ipa-cp-single-call-penalty
11118 Percentage penalty functions containg a single call to another
11119 function will receive when they are evaluated for cloning.
11122 @item ipa-max-agg-items
11123 IPA-CP is also capable to propagate a number of scalar values passed
11124 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11125 number of such values per one parameter.
11127 @item ipa-cp-loop-hint-bonus
11128 When IPA-CP determines that a cloning candidate would make the number
11129 of iterations of a loop known, it adds a bonus of
11130 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11133 @item ipa-cp-array-index-hint-bonus
11134 When IPA-CP determines that a cloning candidate would make the index of
11135 an array access known, it adds a bonus of
11136 @option{ipa-cp-array-index-hint-bonus} to the profitability
11137 score of the candidate.
11139 @item ipa-max-aa-steps
11140 During its analysis of function bodies, IPA-CP employs alias analysis
11141 in order to track values pointed to by function parameters. In order
11142 not spend too much time analyzing huge functions, it gives up and
11143 consider all memory clobbered after examining
11144 @option{ipa-max-aa-steps} statements modifying memory.
11146 @item lto-partitions
11147 Specify desired number of partitions produced during WHOPR compilation.
11148 The number of partitions should exceed the number of CPUs used for compilation.
11149 The default value is 32.
11151 @item lto-minpartition
11152 Size of minimal partition for WHOPR (in estimated instructions).
11153 This prevents expenses of splitting very small programs into too many
11156 @item cxx-max-namespaces-for-diagnostic-help
11157 The maximum number of namespaces to consult for suggestions when C++
11158 name lookup fails for an identifier. The default is 1000.
11160 @item sink-frequency-threshold
11161 The maximum relative execution frequency (in percents) of the target block
11162 relative to a statement's original block to allow statement sinking of a
11163 statement. Larger numbers result in more aggressive statement sinking.
11164 The default value is 75. A small positive adjustment is applied for
11165 statements with memory operands as those are even more profitable so sink.
11167 @item max-stores-to-sink
11168 The maximum number of conditional stores paires that can be sunk. Set to 0
11169 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11170 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
11172 @item allow-store-data-races
11173 Allow optimizers to introduce new data races on stores.
11174 Set to 1 to allow, otherwise to 0. This option is enabled by default
11175 at optimization level @option{-Ofast}.
11177 @item case-values-threshold
11178 The smallest number of different values for which it is best to use a
11179 jump-table instead of a tree of conditional branches. If the value is
11180 0, use the default for the machine. The default is 0.
11182 @item tree-reassoc-width
11183 Set the maximum number of instructions executed in parallel in
11184 reassociated tree. This parameter overrides target dependent
11185 heuristics used by default if has non zero value.
11187 @item sched-pressure-algorithm
11188 Choose between the two available implementations of
11189 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11190 and is the more likely to prevent instructions from being reordered.
11191 Algorithm 2 was designed to be a compromise between the relatively
11192 conservative approach taken by algorithm 1 and the rather aggressive
11193 approach taken by the default scheduler. It relies more heavily on
11194 having a regular register file and accurate register pressure classes.
11195 See @file{haifa-sched.c} in the GCC sources for more details.
11197 The default choice depends on the target.
11199 @item max-slsr-cand-scan
11200 Set the maximum number of existing candidates that are considered when
11201 seeking a basis for a new straight-line strength reduction candidate.
11204 Enable buffer overflow detection for global objects. This kind
11205 of protection is enabled by default if you are using
11206 @option{-fsanitize=address} option.
11207 To disable global objects protection use @option{--param asan-globals=0}.
11210 Enable buffer overflow detection for stack objects. This kind of
11211 protection is enabled by default when using @option{-fsanitize=address}.
11212 To disable stack protection use @option{--param asan-stack=0} option.
11214 @item asan-instrument-reads
11215 Enable buffer overflow detection for memory reads. This kind of
11216 protection is enabled by default when using @option{-fsanitize=address}.
11217 To disable memory reads protection use
11218 @option{--param asan-instrument-reads=0}.
11220 @item asan-instrument-writes
11221 Enable buffer overflow detection for memory writes. This kind of
11222 protection is enabled by default when using @option{-fsanitize=address}.
11223 To disable memory writes protection use
11224 @option{--param asan-instrument-writes=0} option.
11226 @item asan-memintrin
11227 Enable detection for built-in functions. This kind of protection
11228 is enabled by default when using @option{-fsanitize=address}.
11229 To disable built-in functions protection use
11230 @option{--param asan-memintrin=0}.
11232 @item asan-use-after-return
11233 Enable detection of use-after-return. This kind of protection
11234 is enabled by default when using @option{-fsanitize=address} option.
11235 To disable use-after-return detection use
11236 @option{--param asan-use-after-return=0}.
11238 @item asan-instrumentation-with-call-threshold
11239 If number of memory accesses in function being instrumented
11240 is greater or equal to this number, use callbacks instead of inline checks.
11241 E.g. to disable inline code use
11242 @option{--param asan-instrumentation-with-call-threshold=0}.
11244 @item chkp-max-ctor-size
11245 Static constructors generated by Pointer Bounds Checker may become very
11246 large and significantly increase compile time at optimization level
11247 @option{-O1} and higher. This parameter is a maximum nubmer of statements
11248 in a single generated constructor. Default value is 5000.
11250 @item max-fsm-thread-path-insns
11251 Maximum number of instructions to copy when duplicating blocks on a
11252 finite state automaton jump thread path. The default is 100.
11254 @item max-fsm-thread-length
11255 Maximum number of basic blocks on a finite state automaton jump thread
11256 path. The default is 10.
11258 @item max-fsm-thread-paths
11259 Maximum number of new jump thread paths to create for a finite state
11260 automaton. The default is 50.
11262 @item parloops-chunk-size
11263 Chunk size of omp schedule for loops parallelized by parloops. The default
11266 @item parloops-schedule
11267 Schedule type of omp schedule for loops parallelized by parloops (static,
11268 dynamic, guided, auto, runtime). The default is static.
11270 @item max-ssa-name-query-depth
11271 Maximum depth of recursion when querying properties of SSA names in things
11272 like fold routines. One level of recursion corresponds to following a
11277 @node Preprocessor Options
11278 @section Options Controlling the Preprocessor
11279 @cindex preprocessor options
11280 @cindex options, preprocessor
11282 These options control the C preprocessor, which is run on each C source
11283 file before actual compilation.
11285 If you use the @option{-E} option, nothing is done except preprocessing.
11286 Some of these options make sense only together with @option{-E} because
11287 they cause the preprocessor output to be unsuitable for actual
11291 @item -Wp,@var{option}
11293 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11294 and pass @var{option} directly through to the preprocessor. If
11295 @var{option} contains commas, it is split into multiple options at the
11296 commas. However, many options are modified, translated or interpreted
11297 by the compiler driver before being passed to the preprocessor, and
11298 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11299 interface is undocumented and subject to change, so whenever possible
11300 you should avoid using @option{-Wp} and let the driver handle the
11303 @item -Xpreprocessor @var{option}
11304 @opindex Xpreprocessor
11305 Pass @var{option} as an option to the preprocessor. You can use this to
11306 supply system-specific preprocessor options that GCC does not
11309 If you want to pass an option that takes an argument, you must use
11310 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11312 @item -no-integrated-cpp
11313 @opindex no-integrated-cpp
11314 Perform preprocessing as a separate pass before compilation.
11315 By default, GCC performs preprocessing as an integrated part of
11316 input tokenization and parsing.
11317 If this option is provided, the appropriate language front end
11318 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11319 and Objective-C, respectively) is instead invoked twice,
11320 once for preprocessing only and once for actual compilation
11321 of the preprocessed input.
11322 This option may be useful in conjunction with the @option{-B} or
11323 @option{-wrapper} options to specify an alternate preprocessor or
11324 perform additional processing of the program source between
11325 normal preprocessing and compilation.
11328 @include cppopts.texi
11330 @node Assembler Options
11331 @section Passing Options to the Assembler
11333 @c prevent bad page break with this line
11334 You can pass options to the assembler.
11337 @item -Wa,@var{option}
11339 Pass @var{option} as an option to the assembler. If @var{option}
11340 contains commas, it is split into multiple options at the commas.
11342 @item -Xassembler @var{option}
11343 @opindex Xassembler
11344 Pass @var{option} as an option to the assembler. You can use this to
11345 supply system-specific assembler options that GCC does not
11348 If you want to pass an option that takes an argument, you must use
11349 @option{-Xassembler} twice, once for the option and once for the argument.
11354 @section Options for Linking
11355 @cindex link options
11356 @cindex options, linking
11358 These options come into play when the compiler links object files into
11359 an executable output file. They are meaningless if the compiler is
11360 not doing a link step.
11364 @item @var{object-file-name}
11365 A file name that does not end in a special recognized suffix is
11366 considered to name an object file or library. (Object files are
11367 distinguished from libraries by the linker according to the file
11368 contents.) If linking is done, these object files are used as input
11377 If any of these options is used, then the linker is not run, and
11378 object file names should not be used as arguments. @xref{Overall
11382 @opindex fuse-ld=bfd
11383 Use the @command{bfd} linker instead of the default linker.
11385 @item -fuse-ld=gold
11386 @opindex fuse-ld=gold
11387 Use the @command{gold} linker instead of the default linker.
11390 @item -l@var{library}
11391 @itemx -l @var{library}
11393 Search the library named @var{library} when linking. (The second
11394 alternative with the library as a separate argument is only for
11395 POSIX compliance and is not recommended.)
11397 It makes a difference where in the command you write this option; the
11398 linker searches and processes libraries and object files in the order they
11399 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11400 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11401 to functions in @samp{z}, those functions may not be loaded.
11403 The linker searches a standard list of directories for the library,
11404 which is actually a file named @file{lib@var{library}.a}. The linker
11405 then uses this file as if it had been specified precisely by name.
11407 The directories searched include several standard system directories
11408 plus any that you specify with @option{-L}.
11410 Normally the files found this way are library files---archive files
11411 whose members are object files. The linker handles an archive file by
11412 scanning through it for members which define symbols that have so far
11413 been referenced but not defined. But if the file that is found is an
11414 ordinary object file, it is linked in the usual fashion. The only
11415 difference between using an @option{-l} option and specifying a file name
11416 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11417 and searches several directories.
11421 You need this special case of the @option{-l} option in order to
11422 link an Objective-C or Objective-C++ program.
11424 @item -nostartfiles
11425 @opindex nostartfiles
11426 Do not use the standard system startup files when linking.
11427 The standard system libraries are used normally, unless @option{-nostdlib}
11428 or @option{-nodefaultlibs} is used.
11430 @item -nodefaultlibs
11431 @opindex nodefaultlibs
11432 Do not use the standard system libraries when linking.
11433 Only the libraries you specify are passed to the linker, and options
11434 specifying linkage of the system libraries, such as @option{-static-libgcc}
11435 or @option{-shared-libgcc}, are ignored.
11436 The standard startup files are used normally, unless @option{-nostartfiles}
11439 The compiler may generate calls to @code{memcmp},
11440 @code{memset}, @code{memcpy} and @code{memmove}.
11441 These entries are usually resolved by entries in
11442 libc. These entry points should be supplied through some other
11443 mechanism when this option is specified.
11447 Do not use the standard system startup files or libraries when linking.
11448 No startup files and only the libraries you specify are passed to
11449 the linker, and options specifying linkage of the system libraries, such as
11450 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11452 The compiler may generate calls to @code{memcmp}, @code{memset},
11453 @code{memcpy} and @code{memmove}.
11454 These entries are usually resolved by entries in
11455 libc. These entry points should be supplied through some other
11456 mechanism when this option is specified.
11458 @cindex @option{-lgcc}, use with @option{-nostdlib}
11459 @cindex @option{-nostdlib} and unresolved references
11460 @cindex unresolved references and @option{-nostdlib}
11461 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11462 @cindex @option{-nodefaultlibs} and unresolved references
11463 @cindex unresolved references and @option{-nodefaultlibs}
11464 One of the standard libraries bypassed by @option{-nostdlib} and
11465 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11466 which GCC uses to overcome shortcomings of particular machines, or special
11467 needs for some languages.
11468 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11469 Collection (GCC) Internals},
11470 for more discussion of @file{libgcc.a}.)
11471 In most cases, you need @file{libgcc.a} even when you want to avoid
11472 other standard libraries. In other words, when you specify @option{-nostdlib}
11473 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11474 This ensures that you have no unresolved references to internal GCC
11475 library subroutines.
11476 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11477 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11478 GNU Compiler Collection (GCC) Internals}.)
11482 Produce a position independent executable on targets that support it.
11483 For predictable results, you must also specify the same set of options
11484 used for compilation (@option{-fpie}, @option{-fPIE},
11485 or model suboptions) when you specify this linker option.
11489 Don't produce a position independent executable.
11493 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11494 that support it. This instructs the linker to add all symbols, not
11495 only used ones, to the dynamic symbol table. This option is needed
11496 for some uses of @code{dlopen} or to allow obtaining backtraces
11497 from within a program.
11501 Remove all symbol table and relocation information from the executable.
11505 On systems that support dynamic linking, this prevents linking with the shared
11506 libraries. On other systems, this option has no effect.
11510 Produce a shared object which can then be linked with other objects to
11511 form an executable. Not all systems support this option. For predictable
11512 results, you must also specify the same set of options used for compilation
11513 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11514 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11515 needs to build supplementary stub code for constructors to work. On
11516 multi-libbed systems, @samp{gcc -shared} must select the correct support
11517 libraries to link against. Failing to supply the correct flags may lead
11518 to subtle defects. Supplying them in cases where they are not necessary
11521 @item -shared-libgcc
11522 @itemx -static-libgcc
11523 @opindex shared-libgcc
11524 @opindex static-libgcc
11525 On systems that provide @file{libgcc} as a shared library, these options
11526 force the use of either the shared or static version, respectively.
11527 If no shared version of @file{libgcc} was built when the compiler was
11528 configured, these options have no effect.
11530 There are several situations in which an application should use the
11531 shared @file{libgcc} instead of the static version. The most common
11532 of these is when the application wishes to throw and catch exceptions
11533 across different shared libraries. In that case, each of the libraries
11534 as well as the application itself should use the shared @file{libgcc}.
11536 Therefore, the G++ and GCJ drivers automatically add
11537 @option{-shared-libgcc} whenever you build a shared library or a main
11538 executable, because C++ and Java programs typically use exceptions, so
11539 this is the right thing to do.
11541 If, instead, you use the GCC driver to create shared libraries, you may
11542 find that they are not always linked with the shared @file{libgcc}.
11543 If GCC finds, at its configuration time, that you have a non-GNU linker
11544 or a GNU linker that does not support option @option{--eh-frame-hdr},
11545 it links the shared version of @file{libgcc} into shared libraries
11546 by default. Otherwise, it takes advantage of the linker and optimizes
11547 away the linking with the shared version of @file{libgcc}, linking with
11548 the static version of libgcc by default. This allows exceptions to
11549 propagate through such shared libraries, without incurring relocation
11550 costs at library load time.
11552 However, if a library or main executable is supposed to throw or catch
11553 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11554 for the languages used in the program, or using the option
11555 @option{-shared-libgcc}, such that it is linked with the shared
11558 @item -static-libasan
11559 @opindex static-libasan
11560 When the @option{-fsanitize=address} option is used to link a program,
11561 the GCC driver automatically links against @option{libasan}. If
11562 @file{libasan} is available as a shared library, and the @option{-static}
11563 option is not used, then this links against the shared version of
11564 @file{libasan}. The @option{-static-libasan} option directs the GCC
11565 driver to link @file{libasan} statically, without necessarily linking
11566 other libraries statically.
11568 @item -static-libtsan
11569 @opindex static-libtsan
11570 When the @option{-fsanitize=thread} option is used to link a program,
11571 the GCC driver automatically links against @option{libtsan}. If
11572 @file{libtsan} is available as a shared library, and the @option{-static}
11573 option is not used, then this links against the shared version of
11574 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11575 driver to link @file{libtsan} statically, without necessarily linking
11576 other libraries statically.
11578 @item -static-liblsan
11579 @opindex static-liblsan
11580 When the @option{-fsanitize=leak} option is used to link a program,
11581 the GCC driver automatically links against @option{liblsan}. If
11582 @file{liblsan} is available as a shared library, and the @option{-static}
11583 option is not used, then this links against the shared version of
11584 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11585 driver to link @file{liblsan} statically, without necessarily linking
11586 other libraries statically.
11588 @item -static-libubsan
11589 @opindex static-libubsan
11590 When the @option{-fsanitize=undefined} option is used to link a program,
11591 the GCC driver automatically links against @option{libubsan}. If
11592 @file{libubsan} is available as a shared library, and the @option{-static}
11593 option is not used, then this links against the shared version of
11594 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11595 driver to link @file{libubsan} statically, without necessarily linking
11596 other libraries statically.
11598 @item -static-libmpx
11599 @opindex static-libmpx
11600 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11601 used to link a program, the GCC driver automatically links against
11602 @file{libmpx}. If @file{libmpx} is available as a shared library,
11603 and the @option{-static} option is not used, then this links against
11604 the shared version of @file{libmpx}. The @option{-static-libmpx}
11605 option directs the GCC driver to link @file{libmpx} statically,
11606 without necessarily linking other libraries statically.
11608 @item -static-libmpxwrappers
11609 @opindex static-libmpxwrappers
11610 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11611 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11612 GCC driver automatically links against @file{libmpxwrappers}. If
11613 @file{libmpxwrappers} is available as a shared library, and the
11614 @option{-static} option is not used, then this links against the shared
11615 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11616 option directs the GCC driver to link @file{libmpxwrappers} statically,
11617 without necessarily linking other libraries statically.
11619 @item -static-libstdc++
11620 @opindex static-libstdc++
11621 When the @command{g++} program is used to link a C++ program, it
11622 normally automatically links against @option{libstdc++}. If
11623 @file{libstdc++} is available as a shared library, and the
11624 @option{-static} option is not used, then this links against the
11625 shared version of @file{libstdc++}. That is normally fine. However, it
11626 is sometimes useful to freeze the version of @file{libstdc++} used by
11627 the program without going all the way to a fully static link. The
11628 @option{-static-libstdc++} option directs the @command{g++} driver to
11629 link @file{libstdc++} statically, without necessarily linking other
11630 libraries statically.
11634 Bind references to global symbols when building a shared object. Warn
11635 about any unresolved references (unless overridden by the link editor
11636 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11639 @item -T @var{script}
11641 @cindex linker script
11642 Use @var{script} as the linker script. This option is supported by most
11643 systems using the GNU linker. On some targets, such as bare-board
11644 targets without an operating system, the @option{-T} option may be required
11645 when linking to avoid references to undefined symbols.
11647 @item -Xlinker @var{option}
11649 Pass @var{option} as an option to the linker. You can use this to
11650 supply system-specific linker options that GCC does not recognize.
11652 If you want to pass an option that takes a separate argument, you must use
11653 @option{-Xlinker} twice, once for the option and once for the argument.
11654 For example, to pass @option{-assert definitions}, you must write
11655 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11656 @option{-Xlinker "-assert definitions"}, because this passes the entire
11657 string as a single argument, which is not what the linker expects.
11659 When using the GNU linker, it is usually more convenient to pass
11660 arguments to linker options using the @option{@var{option}=@var{value}}
11661 syntax than as separate arguments. For example, you can specify
11662 @option{-Xlinker -Map=output.map} rather than
11663 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11664 this syntax for command-line options.
11666 @item -Wl,@var{option}
11668 Pass @var{option} as an option to the linker. If @var{option} contains
11669 commas, it is split into multiple options at the commas. You can use this
11670 syntax to pass an argument to the option.
11671 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11672 linker. When using the GNU linker, you can also get the same effect with
11673 @option{-Wl,-Map=output.map}.
11675 @item -u @var{symbol}
11677 Pretend the symbol @var{symbol} is undefined, to force linking of
11678 library modules to define it. You can use @option{-u} multiple times with
11679 different symbols to force loading of additional library modules.
11681 @item -z @var{keyword}
11683 @option{-z} is passed directly on to the linker along with the keyword
11684 @var{keyword}. See the section in the documentation of your linker for
11685 permitted values and their meanings.
11688 @node Directory Options
11689 @section Options for Directory Search
11690 @cindex directory options
11691 @cindex options, directory search
11692 @cindex search path
11694 These options specify directories to search for header files, for
11695 libraries and for parts of the compiler:
11700 Add the directory @var{dir} to the head of the list of directories to be
11701 searched for header files. This can be used to override a system header
11702 file, substituting your own version, since these directories are
11703 searched before the system header file directories. However, you should
11704 not use this option to add directories that contain vendor-supplied
11705 system header files (use @option{-isystem} for that). If you use more than
11706 one @option{-I} option, the directories are scanned in left-to-right
11707 order; the standard system directories come after.
11709 If a standard system include directory, or a directory specified with
11710 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11711 option is ignored. The directory is still searched but as a
11712 system directory at its normal position in the system include chain.
11713 This is to ensure that GCC's procedure to fix buggy system headers and
11714 the ordering for the @code{include_next} directive are not inadvertently changed.
11715 If you really need to change the search order for system directories,
11716 use the @option{-nostdinc} and/or @option{-isystem} options.
11718 @item -iplugindir=@var{dir}
11719 @opindex iplugindir=
11720 Set the directory to search for plugins that are passed
11721 by @option{-fplugin=@var{name}} instead of
11722 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11723 to be used by the user, but only passed by the driver.
11725 @item -iquote@var{dir}
11727 Add the directory @var{dir} to the head of the list of directories to
11728 be searched for header files only for the case of @code{#include
11729 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11730 otherwise just like @option{-I}.
11734 Add directory @var{dir} to the list of directories to be searched
11737 @item -B@var{prefix}
11739 This option specifies where to find the executables, libraries,
11740 include files, and data files of the compiler itself.
11742 The compiler driver program runs one or more of the subprograms
11743 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11744 @var{prefix} as a prefix for each program it tries to run, both with and
11745 without @samp{@var{machine}/@var{version}/} for the corresponding target
11746 machine and compiler version.
11748 For each subprogram to be run, the compiler driver first tries the
11749 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11750 is not specified, the driver tries two standard prefixes,
11751 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11752 those results in a file name that is found, the unmodified program
11753 name is searched for using the directories specified in your
11754 @env{PATH} environment variable.
11756 The compiler checks to see if the path provided by @option{-B}
11757 refers to a directory, and if necessary it adds a directory
11758 separator character at the end of the path.
11760 @option{-B} prefixes that effectively specify directory names also apply
11761 to libraries in the linker, because the compiler translates these
11762 options into @option{-L} options for the linker. They also apply to
11763 include files in the preprocessor, because the compiler translates these
11764 options into @option{-isystem} options for the preprocessor. In this case,
11765 the compiler appends @samp{include} to the prefix.
11767 The runtime support file @file{libgcc.a} can also be searched for using
11768 the @option{-B} prefix, if needed. If it is not found there, the two
11769 standard prefixes above are tried, and that is all. The file is left
11770 out of the link if it is not found by those means.
11772 Another way to specify a prefix much like the @option{-B} prefix is to use
11773 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11776 As a special kludge, if the path provided by @option{-B} is
11777 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11778 9, then it is replaced by @file{[dir/]include}. This is to help
11779 with boot-strapping the compiler.
11781 @item -specs=@var{file}
11783 Process @var{file} after the compiler reads in the standard @file{specs}
11784 file, in order to override the defaults which the @command{gcc} driver
11785 program uses when determining what switches to pass to @command{cc1},
11786 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11787 @option{-specs=@var{file}} can be specified on the command line, and they
11788 are processed in order, from left to right.
11790 @item --sysroot=@var{dir}
11792 Use @var{dir} as the logical root directory for headers and libraries.
11793 For example, if the compiler normally searches for headers in
11794 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11795 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11797 If you use both this option and the @option{-isysroot} option, then
11798 the @option{--sysroot} option applies to libraries, but the
11799 @option{-isysroot} option applies to header files.
11801 The GNU linker (beginning with version 2.16) has the necessary support
11802 for this option. If your linker does not support this option, the
11803 header file aspect of @option{--sysroot} still works, but the
11804 library aspect does not.
11806 @item --no-sysroot-suffix
11807 @opindex no-sysroot-suffix
11808 For some targets, a suffix is added to the root directory specified
11809 with @option{--sysroot}, depending on the other options used, so that
11810 headers may for example be found in
11811 @file{@var{dir}/@var{suffix}/usr/include} instead of
11812 @file{@var{dir}/usr/include}. This option disables the addition of
11817 This option has been deprecated. Please use @option{-iquote} instead for
11818 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11820 Any directories you specify with @option{-I} options before the @option{-I-}
11821 option are searched only for the case of @code{#include "@var{file}"};
11822 they are not searched for @code{#include <@var{file}>}.
11824 If additional directories are specified with @option{-I} options after
11825 the @option{-I-} option, these directories are searched for all @code{#include}
11826 directives. (Ordinarily @emph{all} @option{-I} directories are used
11829 In addition, the @option{-I-} option inhibits the use of the current
11830 directory (where the current input file came from) as the first search
11831 directory for @code{#include "@var{file}"}. There is no way to
11832 override this effect of @option{-I-}. With @option{-I.} you can specify
11833 searching the directory that is current when the compiler is
11834 invoked. That is not exactly the same as what the preprocessor does
11835 by default, but it is often satisfactory.
11837 @option{-I-} does not inhibit the use of the standard system directories
11838 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11845 @section Specifying Subprocesses and the Switches to Pass to Them
11848 @command{gcc} is a driver program. It performs its job by invoking a
11849 sequence of other programs to do the work of compiling, assembling and
11850 linking. GCC interprets its command-line parameters and uses these to
11851 deduce which programs it should invoke, and which command-line options
11852 it ought to place on their command lines. This behavior is controlled
11853 by @dfn{spec strings}. In most cases there is one spec string for each
11854 program that GCC can invoke, but a few programs have multiple spec
11855 strings to control their behavior. The spec strings built into GCC can
11856 be overridden by using the @option{-specs=} command-line switch to specify
11859 @dfn{Spec files} are plaintext files that are used to construct spec
11860 strings. They consist of a sequence of directives separated by blank
11861 lines. The type of directive is determined by the first non-whitespace
11862 character on the line, which can be one of the following:
11865 @item %@var{command}
11866 Issues a @var{command} to the spec file processor. The commands that can
11870 @item %include <@var{file}>
11871 @cindex @code{%include}
11872 Search for @var{file} and insert its text at the current point in the
11875 @item %include_noerr <@var{file}>
11876 @cindex @code{%include_noerr}
11877 Just like @samp{%include}, but do not generate an error message if the include
11878 file cannot be found.
11880 @item %rename @var{old_name} @var{new_name}
11881 @cindex @code{%rename}
11882 Rename the spec string @var{old_name} to @var{new_name}.
11886 @item *[@var{spec_name}]:
11887 This tells the compiler to create, override or delete the named spec
11888 string. All lines after this directive up to the next directive or
11889 blank line are considered to be the text for the spec string. If this
11890 results in an empty string then the spec is deleted. (Or, if the
11891 spec did not exist, then nothing happens.) Otherwise, if the spec
11892 does not currently exist a new spec is created. If the spec does
11893 exist then its contents are overridden by the text of this
11894 directive, unless the first character of that text is the @samp{+}
11895 character, in which case the text is appended to the spec.
11897 @item [@var{suffix}]:
11898 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11899 and up to the next directive or blank line are considered to make up the
11900 spec string for the indicated suffix. When the compiler encounters an
11901 input file with the named suffix, it processes the spec string in
11902 order to work out how to compile that file. For example:
11906 z-compile -input %i
11909 This says that any input file whose name ends in @samp{.ZZ} should be
11910 passed to the program @samp{z-compile}, which should be invoked with the
11911 command-line switch @option{-input} and with the result of performing the
11912 @samp{%i} substitution. (See below.)
11914 As an alternative to providing a spec string, the text following a
11915 suffix directive can be one of the following:
11918 @item @@@var{language}
11919 This says that the suffix is an alias for a known @var{language}. This is
11920 similar to using the @option{-x} command-line switch to GCC to specify a
11921 language explicitly. For example:
11928 Says that .ZZ files are, in fact, C++ source files.
11931 This causes an error messages saying:
11934 @var{name} compiler not installed on this system.
11938 GCC already has an extensive list of suffixes built into it.
11939 This directive adds an entry to the end of the list of suffixes, but
11940 since the list is searched from the end backwards, it is effectively
11941 possible to override earlier entries using this technique.
11945 GCC has the following spec strings built into it. Spec files can
11946 override these strings or create their own. Note that individual
11947 targets can also add their own spec strings to this list.
11950 asm Options to pass to the assembler
11951 asm_final Options to pass to the assembler post-processor
11952 cpp Options to pass to the C preprocessor
11953 cc1 Options to pass to the C compiler
11954 cc1plus Options to pass to the C++ compiler
11955 endfile Object files to include at the end of the link
11956 link Options to pass to the linker
11957 lib Libraries to include on the command line to the linker
11958 libgcc Decides which GCC support library to pass to the linker
11959 linker Sets the name of the linker
11960 predefines Defines to be passed to the C preprocessor
11961 signed_char Defines to pass to CPP to say whether @code{char} is signed
11963 startfile Object files to include at the start of the link
11966 Here is a small example of a spec file:
11969 %rename lib old_lib
11972 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11975 This example renames the spec called @samp{lib} to @samp{old_lib} and
11976 then overrides the previous definition of @samp{lib} with a new one.
11977 The new definition adds in some extra command-line options before
11978 including the text of the old definition.
11980 @dfn{Spec strings} are a list of command-line options to be passed to their
11981 corresponding program. In addition, the spec strings can contain
11982 @samp{%}-prefixed sequences to substitute variable text or to
11983 conditionally insert text into the command line. Using these constructs
11984 it is possible to generate quite complex command lines.
11986 Here is a table of all defined @samp{%}-sequences for spec
11987 strings. Note that spaces are not generated automatically around the
11988 results of expanding these sequences. Therefore you can concatenate them
11989 together or combine them with constant text in a single argument.
11993 Substitute one @samp{%} into the program name or argument.
11996 Substitute the name of the input file being processed.
11999 Substitute the basename of the input file being processed.
12000 This is the substring up to (and not including) the last period
12001 and not including the directory.
12004 This is the same as @samp{%b}, but include the file suffix (text after
12008 Marks the argument containing or following the @samp{%d} as a
12009 temporary file name, so that that file is deleted if GCC exits
12010 successfully. Unlike @samp{%g}, this contributes no text to the
12013 @item %g@var{suffix}
12014 Substitute a file name that has suffix @var{suffix} and is chosen
12015 once per compilation, and mark the argument in the same way as
12016 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
12017 name is now chosen in a way that is hard to predict even when previously
12018 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
12019 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
12020 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
12021 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
12022 was simply substituted with a file name chosen once per compilation,
12023 without regard to any appended suffix (which was therefore treated
12024 just like ordinary text), making such attacks more likely to succeed.
12026 @item %u@var{suffix}
12027 Like @samp{%g}, but generates a new temporary file name
12028 each time it appears instead of once per compilation.
12030 @item %U@var{suffix}
12031 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
12032 new one if there is no such last file name. In the absence of any
12033 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
12034 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
12035 involves the generation of two distinct file names, one
12036 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
12037 simply substituted with a file name chosen for the previous @samp{%u},
12038 without regard to any appended suffix.
12040 @item %j@var{suffix}
12041 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
12042 writable, and if @option{-save-temps} is not used;
12043 otherwise, substitute the name
12044 of a temporary file, just like @samp{%u}. This temporary file is not
12045 meant for communication between processes, but rather as a junk
12046 disposal mechanism.
12048 @item %|@var{suffix}
12049 @itemx %m@var{suffix}
12050 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
12051 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
12052 all. These are the two most common ways to instruct a program that it
12053 should read from standard input or write to standard output. If you
12054 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
12055 construct: see for example @file{f/lang-specs.h}.
12057 @item %.@var{SUFFIX}
12058 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
12059 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
12060 terminated by the next space or %.
12063 Marks the argument containing or following the @samp{%w} as the
12064 designated output file of this compilation. This puts the argument
12065 into the sequence of arguments that @samp{%o} substitutes.
12068 Substitutes the names of all the output files, with spaces
12069 automatically placed around them. You should write spaces
12070 around the @samp{%o} as well or the results are undefined.
12071 @samp{%o} is for use in the specs for running the linker.
12072 Input files whose names have no recognized suffix are not compiled
12073 at all, but they are included among the output files, so they are
12077 Substitutes the suffix for object files. Note that this is
12078 handled specially when it immediately follows @samp{%g, %u, or %U},
12079 because of the need for those to form complete file names. The
12080 handling is such that @samp{%O} is treated exactly as if it had already
12081 been substituted, except that @samp{%g, %u, and %U} do not currently
12082 support additional @var{suffix} characters following @samp{%O} as they do
12083 following, for example, @samp{.o}.
12086 Substitutes the standard macro predefinitions for the
12087 current target machine. Use this when running @command{cpp}.
12090 Like @samp{%p}, but puts @samp{__} before and after the name of each
12091 predefined macro, except for macros that start with @samp{__} or with
12092 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
12096 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
12097 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
12098 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
12099 and @option{-imultilib} as necessary.
12102 Current argument is the name of a library or startup file of some sort.
12103 Search for that file in a standard list of directories and substitute
12104 the full name found. The current working directory is included in the
12105 list of directories scanned.
12108 Current argument is the name of a linker script. Search for that file
12109 in the current list of directories to scan for libraries. If the file
12110 is located insert a @option{--script} option into the command line
12111 followed by the full path name found. If the file is not found then
12112 generate an error message. Note: the current working directory is not
12116 Print @var{str} as an error message. @var{str} is terminated by a newline.
12117 Use this when inconsistent options are detected.
12119 @item %(@var{name})
12120 Substitute the contents of spec string @var{name} at this point.
12122 @item %x@{@var{option}@}
12123 Accumulate an option for @samp{%X}.
12126 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
12130 Output the accumulated assembler options specified by @option{-Wa}.
12133 Output the accumulated preprocessor options specified by @option{-Wp}.
12136 Process the @code{asm} spec. This is used to compute the
12137 switches to be passed to the assembler.
12140 Process the @code{asm_final} spec. This is a spec string for
12141 passing switches to an assembler post-processor, if such a program is
12145 Process the @code{link} spec. This is the spec for computing the
12146 command line passed to the linker. Typically it makes use of the
12147 @samp{%L %G %S %D and %E} sequences.
12150 Dump out a @option{-L} option for each directory that GCC believes might
12151 contain startup files. If the target supports multilibs then the
12152 current multilib directory is prepended to each of these paths.
12155 Process the @code{lib} spec. This is a spec string for deciding which
12156 libraries are included on the command line to the linker.
12159 Process the @code{libgcc} spec. This is a spec string for deciding
12160 which GCC support library is included on the command line to the linker.
12163 Process the @code{startfile} spec. This is a spec for deciding which
12164 object files are the first ones passed to the linker. Typically
12165 this might be a file named @file{crt0.o}.
12168 Process the @code{endfile} spec. This is a spec string that specifies
12169 the last object files that are passed to the linker.
12172 Process the @code{cpp} spec. This is used to construct the arguments
12173 to be passed to the C preprocessor.
12176 Process the @code{cc1} spec. This is used to construct the options to be
12177 passed to the actual C compiler (@command{cc1}).
12180 Process the @code{cc1plus} spec. This is used to construct the options to be
12181 passed to the actual C++ compiler (@command{cc1plus}).
12184 Substitute the variable part of a matched option. See below.
12185 Note that each comma in the substituted string is replaced by
12189 Remove all occurrences of @code{-S} from the command line. Note---this
12190 command is position dependent. @samp{%} commands in the spec string
12191 before this one see @code{-S}, @samp{%} commands in the spec string
12192 after this one do not.
12194 @item %:@var{function}(@var{args})
12195 Call the named function @var{function}, passing it @var{args}.
12196 @var{args} is first processed as a nested spec string, then split
12197 into an argument vector in the usual fashion. The function returns
12198 a string which is processed as if it had appeared literally as part
12199 of the current spec.
12201 The following built-in spec functions are provided:
12204 @item @code{getenv}
12205 The @code{getenv} spec function takes two arguments: an environment
12206 variable name and a string. If the environment variable is not
12207 defined, a fatal error is issued. Otherwise, the return value is the
12208 value of the environment variable concatenated with the string. For
12209 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
12212 %:getenv(TOPDIR /include)
12215 expands to @file{/path/to/top/include}.
12217 @item @code{if-exists}
12218 The @code{if-exists} spec function takes one argument, an absolute
12219 pathname to a file. If the file exists, @code{if-exists} returns the
12220 pathname. Here is a small example of its usage:
12224 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
12227 @item @code{if-exists-else}
12228 The @code{if-exists-else} spec function is similar to the @code{if-exists}
12229 spec function, except that it takes two arguments. The first argument is
12230 an absolute pathname to a file. If the file exists, @code{if-exists-else}
12231 returns the pathname. If it does not exist, it returns the second argument.
12232 This way, @code{if-exists-else} can be used to select one file or another,
12233 based on the existence of the first. Here is a small example of its usage:
12237 crt0%O%s %:if-exists(crti%O%s) \
12238 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
12241 @item @code{replace-outfile}
12242 The @code{replace-outfile} spec function takes two arguments. It looks for the
12243 first argument in the outfiles array and replaces it with the second argument. Here
12244 is a small example of its usage:
12247 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
12250 @item @code{remove-outfile}
12251 The @code{remove-outfile} spec function takes one argument. It looks for the
12252 first argument in the outfiles array and removes it. Here is a small example
12256 %:remove-outfile(-lm)
12259 @item @code{pass-through-libs}
12260 The @code{pass-through-libs} spec function takes any number of arguments. It
12261 finds any @option{-l} options and any non-options ending in @file{.a} (which it
12262 assumes are the names of linker input library archive files) and returns a
12263 result containing all the found arguments each prepended by
12264 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
12265 intended to be passed to the LTO linker plugin.
12268 %:pass-through-libs(%G %L %G)
12271 @item @code{print-asm-header}
12272 The @code{print-asm-header} function takes no arguments and simply
12273 prints a banner like:
12279 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12282 It is used to separate compiler options from assembler options
12283 in the @option{--target-help} output.
12286 @item %@{@code{S}@}
12287 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12288 If that switch is not specified, this substitutes nothing. Note that
12289 the leading dash is omitted when specifying this option, and it is
12290 automatically inserted if the substitution is performed. Thus the spec
12291 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12292 and outputs the command-line option @option{-foo}.
12294 @item %W@{@code{S}@}
12295 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12296 deleted on failure.
12298 @item %@{@code{S}*@}
12299 Substitutes all the switches specified to GCC whose names start
12300 with @code{-S}, but which also take an argument. This is used for
12301 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12302 GCC considers @option{-o foo} as being
12303 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12304 text, including the space. Thus two arguments are generated.
12306 @item %@{@code{S}*&@code{T}*@}
12307 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12308 (the order of @code{S} and @code{T} in the spec is not significant).
12309 There can be any number of ampersand-separated variables; for each the
12310 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12312 @item %@{@code{S}:@code{X}@}
12313 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12315 @item %@{!@code{S}:@code{X}@}
12316 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12318 @item %@{@code{S}*:@code{X}@}
12319 Substitutes @code{X} if one or more switches whose names start with
12320 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12321 once, no matter how many such switches appeared. However, if @code{%*}
12322 appears somewhere in @code{X}, then @code{X} is substituted once
12323 for each matching switch, with the @code{%*} replaced by the part of
12324 that switch matching the @code{*}.
12326 If @code{%*} appears as the last part of a spec sequence then a space
12327 is added after the end of the last substitution. If there is more
12328 text in the sequence, however, then a space is not generated. This
12329 allows the @code{%*} substitution to be used as part of a larger
12330 string. For example, a spec string like this:
12333 %@{mcu=*:--script=%*/memory.ld@}
12337 when matching an option like @option{-mcu=newchip} produces:
12340 --script=newchip/memory.ld
12343 @item %@{.@code{S}:@code{X}@}
12344 Substitutes @code{X}, if processing a file with suffix @code{S}.
12346 @item %@{!.@code{S}:@code{X}@}
12347 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12349 @item %@{,@code{S}:@code{X}@}
12350 Substitutes @code{X}, if processing a file for language @code{S}.
12352 @item %@{!,@code{S}:@code{X}@}
12353 Substitutes @code{X}, if not processing a file for language @code{S}.
12355 @item %@{@code{S}|@code{P}:@code{X}@}
12356 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12357 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12358 @code{*} sequences as well, although they have a stronger binding than
12359 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12360 alternatives must be starred, and only the first matching alternative
12363 For example, a spec string like this:
12366 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12370 outputs the following command-line options from the following input
12371 command-line options:
12376 -d fred.c -foo -baz -boggle
12377 -d jim.d -bar -baz -boggle
12380 @item %@{S:X; T:Y; :D@}
12382 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12383 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12384 be as many clauses as you need. This may be combined with @code{.},
12385 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12390 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12391 construct may contain other nested @samp{%} constructs or spaces, or
12392 even newlines. They are processed as usual, as described above.
12393 Trailing white space in @code{X} is ignored. White space may also
12394 appear anywhere on the left side of the colon in these constructs,
12395 except between @code{.} or @code{*} and the corresponding word.
12397 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12398 handled specifically in these constructs. If another value of
12399 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12400 @option{-W} switch is found later in the command line, the earlier
12401 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12402 just one letter, which passes all matching options.
12404 The character @samp{|} at the beginning of the predicate text is used to
12405 indicate that a command should be piped to the following command, but
12406 only if @option{-pipe} is specified.
12408 It is built into GCC which switches take arguments and which do not.
12409 (You might think it would be useful to generalize this to allow each
12410 compiler's spec to say which switches take arguments. But this cannot
12411 be done in a consistent fashion. GCC cannot even decide which input
12412 files have been specified without knowing which switches take arguments,
12413 and it must know which input files to compile in order to tell which
12416 GCC also knows implicitly that arguments starting in @option{-l} are to be
12417 treated as compiler output files, and passed to the linker in their
12418 proper position among the other output files.
12420 @c man begin OPTIONS
12422 @node Submodel Options
12423 @section Hardware Models and Configurations
12424 @cindex submodel options
12425 @cindex specifying hardware config
12426 @cindex hardware models and configurations, specifying
12427 @cindex machine dependent options
12429 Each target machine types can have its own
12430 special options, starting with @samp{-m}, to choose among various
12431 hardware models or configurations---for example, 68010 vs 68020,
12432 floating coprocessor or none. A single installed version of the
12433 compiler can compile for any model or configuration, according to the
12436 Some configurations of the compiler also support additional special
12437 options, usually for compatibility with other compilers on the same
12440 @c This list is ordered alphanumerically by subsection name.
12441 @c It should be the same order and spelling as these options are listed
12442 @c in Machine Dependent Options
12445 * AArch64 Options::
12446 * Adapteva Epiphany Options::
12450 * Blackfin Options::
12455 * DEC Alpha Options::
12459 * GNU/Linux Options::
12469 * MicroBlaze Options::
12472 * MN10300 Options::
12476 * Nios II Options::
12477 * Nvidia PTX Options::
12479 * picoChip Options::
12480 * PowerPC Options::
12482 * RS/6000 and PowerPC Options::
12484 * S/390 and zSeries Options::
12487 * Solaris 2 Options::
12490 * System V Options::
12491 * TILE-Gx Options::
12492 * TILEPro Options::
12497 * VxWorks Options::
12499 * x86 Windows Options::
12500 * Xstormy16 Options::
12502 * zSeries Options::
12505 @node AArch64 Options
12506 @subsection AArch64 Options
12507 @cindex AArch64 Options
12509 These options are defined for AArch64 implementations:
12513 @item -mabi=@var{name}
12515 Generate code for the specified data model. Permissible values
12516 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12517 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12518 but long int and pointer are 64-bit.
12520 The default depends on the specific target configuration. Note that
12521 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12522 entire program with the same ABI, and link with a compatible set of libraries.
12525 @opindex mbig-endian
12526 Generate big-endian code. This is the default when GCC is configured for an
12527 @samp{aarch64_be-*-*} target.
12529 @item -mgeneral-regs-only
12530 @opindex mgeneral-regs-only
12531 Generate code which uses only the general-purpose registers. This will prevent
12532 the compiler from using floating-point and Advanced SIMD registers but will not
12533 impose any restrictions on the assembler.
12535 @item -mlittle-endian
12536 @opindex mlittle-endian
12537 Generate little-endian code. This is the default when GCC is configured for an
12538 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12540 @item -mcmodel=tiny
12541 @opindex mcmodel=tiny
12542 Generate code for the tiny code model. The program and its statically defined
12543 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12544 be statically or dynamically linked. This model is not fully implemented and
12545 mostly treated as @samp{small}.
12547 @item -mcmodel=small
12548 @opindex mcmodel=small
12549 Generate code for the small code model. The program and its statically defined
12550 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12551 be statically or dynamically linked. This is the default code model.
12553 @item -mcmodel=large
12554 @opindex mcmodel=large
12555 Generate code for the large code model. This makes no assumptions about
12556 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12557 statically linked only.
12559 @item -mstrict-align
12560 @opindex mstrict-align
12561 Do not assume that unaligned memory references are handled by the system.
12563 @item -momit-leaf-frame-pointer
12564 @itemx -mno-omit-leaf-frame-pointer
12565 @opindex momit-leaf-frame-pointer
12566 @opindex mno-omit-leaf-frame-pointer
12567 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12570 @item -mtls-dialect=desc
12571 @opindex mtls-dialect=desc
12572 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12573 of TLS variables. This is the default.
12575 @item -mtls-dialect=traditional
12576 @opindex mtls-dialect=traditional
12577 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12580 @item -mtls-size=@var{size}
12582 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
12583 This option depends on binutils higher than 2.25.
12585 @item -mfix-cortex-a53-835769
12586 @itemx -mno-fix-cortex-a53-835769
12587 @opindex mfix-cortex-a53-835769
12588 @opindex mno-fix-cortex-a53-835769
12589 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12590 This involves inserting a NOP instruction between memory instructions and
12591 64-bit integer multiply-accumulate instructions.
12593 @item -mfix-cortex-a53-843419
12594 @itemx -mno-fix-cortex-a53-843419
12595 @opindex mfix-cortex-a53-843419
12596 @opindex mno-fix-cortex-a53-843419
12597 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12598 This erratum workaround is made at link time and this will only pass the
12599 corresponding flag to the linker.
12601 @item -mlow-precision-recip-sqrt
12602 @item -mno-low-precision-recip-sqrt
12603 @opindex -mlow-precision-recip-sqrt
12604 @opindex -mno-low-precision-recip-sqrt
12605 The square root estimate uses two steps instead of three for double-precision,
12606 and one step instead of two for single-precision.
12607 Thus reducing latency and precision.
12608 This is only relevant if @option{-ffast-math} activates
12609 reciprocal square root estimate instructions.
12610 Which in turn depends on the target processor.
12612 @item -march=@var{name}
12614 Specify the name of the target architecture and, optionally, one or
12615 more feature modifiers. This option has the form
12616 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12618 The permissible values for @var{arch} are @samp{armv8-a},
12619 @samp{armv8.1-a} or @var{native}.
12621 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
12622 support for the ARMv8.1 architecture extension. In particular, it
12623 enables the @samp{+crc} and @samp{+lse} features.
12625 The value @samp{native} is available on native AArch64 GNU/Linux and
12626 causes the compiler to pick the architecture of the host system. This
12627 option has no effect if the compiler is unable to recognize the
12628 architecture of the host system,
12630 The permissible values for @var{feature} are listed in the sub-section
12631 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12632 Feature Modifiers}. Where conflicting feature modifiers are
12633 specified, the right-most feature is used.
12635 GCC uses @var{name} to determine what kind of instructions it can emit
12636 when generating assembly code. If @option{-march} is specified
12637 without either of @option{-mtune} or @option{-mcpu} also being
12638 specified, the code is tuned to perform well across a range of target
12639 processors implementing the target architecture.
12641 @item -mtune=@var{name}
12643 Specify the name of the target processor for which GCC should tune the
12644 performance of the code. Permissible values for this option are:
12645 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
12646 @samp{cortex-a72}, @samp{exynos-m1}, @samp{qdf24xx}, @samp{thunderx},
12649 Additionally, this option can specify that GCC should tune the performance
12650 of the code for a big.LITTLE system. Permissible values for this
12651 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12653 Additionally on native AArch64 GNU/Linux systems the value
12654 @samp{native} is available. This option causes the compiler to pick
12655 the architecture of and tune the performance of the code for the
12656 processor of the host system. This option has no effect if the
12657 compiler is unable to recognize the architecture of the host system.
12659 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12660 are specified, the code is tuned to perform well across a range
12661 of target processors.
12663 This option cannot be suffixed by feature modifiers.
12665 @item -mcpu=@var{name}
12667 Specify the name of the target processor, optionally suffixed by one
12668 or more feature modifiers. This option has the form
12669 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
12670 the permissible values for @var{cpu} are the same as those available
12671 for @option{-mtune}. The permissible values for @var{feature} are
12672 documented in the sub-section on
12673 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12674 Feature Modifiers}. Where conflicting feature modifiers are
12675 specified, the right-most feature is used.
12677 Additionally on native AArch64 GNU/Linux systems the value
12678 @samp{native} is available. This option causes the compiler to tune
12679 the performance of the code for the processor of the host system.
12680 This option has no effect if the compiler is unable to recognize the
12681 architecture of the host system.
12683 GCC uses @var{name} to determine what kind of instructions it can emit when
12684 generating assembly code (as if by @option{-march}) and to determine
12685 the target processor for which to tune for performance (as if
12686 by @option{-mtune}). Where this option is used in conjunction
12687 with @option{-march} or @option{-mtune}, those options take precedence
12688 over the appropriate part of this option.
12690 @item -moverride=@var{string}
12692 Override tuning decisions made by the back-end in response to a
12693 @option{-mtune=} switch. The syntax, semantics, and accepted values
12694 for @var{string} in this option are not guaranteed to be consistent
12697 This option is only intended to be useful when developing GCC.
12699 @item -mpc-relative-literal-loads
12700 @opindex mpcrelativeliteralloads
12701 Enable PC relative literal loads. If this option is used, literal
12702 pools are assumed to have a range of up to 1MiB and an appropriate
12703 instruction sequence is used. This option has no impact when used
12704 with @option{-mcmodel=tiny}.
12708 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12709 @anchor{aarch64-feature-modifiers}
12710 @cindex @option{-march} feature modifiers
12711 @cindex @option{-mcpu} feature modifiers
12712 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
12713 the following and their inverses @option{no@var{feature}}:
12717 Enable CRC extension. This is on by default for
12718 @option{-march=armv8.1-a}.
12720 Enable Crypto extension. This also enables Advanced SIMD and floating-point
12723 Enable floating-point instructions. This is on by default for all possible
12724 values for options @option{-march} and @option{-mcpu}.
12726 Enable Advanced SIMD instructions. This also enables floating-point
12727 instructions. This is on by default for all possible values for options
12728 @option{-march} and @option{-mcpu}.
12730 Enable Large System Extension instructions. This is on by default for
12731 @option{-march=armv8.1-a}.
12735 That is, @option{crypto} implies @option{simd} implies @option{fp}.
12736 Conversely, @option{nofp} (or equivalently, @option{-mgeneral-regs-only})
12737 implies @option{nosimd} implies @option{nocrypto}.
12739 @node Adapteva Epiphany Options
12740 @subsection Adapteva Epiphany Options
12742 These @samp{-m} options are defined for Adapteva Epiphany:
12745 @item -mhalf-reg-file
12746 @opindex mhalf-reg-file
12747 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12748 That allows code to run on hardware variants that lack these registers.
12750 @item -mprefer-short-insn-regs
12751 @opindex mprefer-short-insn-regs
12752 Preferrentially allocate registers that allow short instruction generation.
12753 This can result in increased instruction count, so this may either reduce or
12754 increase overall code size.
12756 @item -mbranch-cost=@var{num}
12757 @opindex mbranch-cost
12758 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12759 This cost is only a heuristic and is not guaranteed to produce
12760 consistent results across releases.
12764 Enable the generation of conditional moves.
12766 @item -mnops=@var{num}
12768 Emit @var{num} NOPs before every other generated instruction.
12770 @item -mno-soft-cmpsf
12771 @opindex mno-soft-cmpsf
12772 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12773 and test the flags. This is faster than a software comparison, but can
12774 get incorrect results in the presence of NaNs, or when two different small
12775 numbers are compared such that their difference is calculated as zero.
12776 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12777 software comparisons.
12779 @item -mstack-offset=@var{num}
12780 @opindex mstack-offset
12781 Set the offset between the top of the stack and the stack pointer.
12782 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12783 can be used by leaf functions without stack allocation.
12784 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12785 Note also that this option changes the ABI; compiling a program with a
12786 different stack offset than the libraries have been compiled with
12787 generally does not work.
12788 This option can be useful if you want to evaluate if a different stack
12789 offset would give you better code, but to actually use a different stack
12790 offset to build working programs, it is recommended to configure the
12791 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12793 @item -mno-round-nearest
12794 @opindex mno-round-nearest
12795 Make the scheduler assume that the rounding mode has been set to
12796 truncating. The default is @option{-mround-nearest}.
12799 @opindex mlong-calls
12800 If not otherwise specified by an attribute, assume all calls might be beyond
12801 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12802 function address into a register before performing a (otherwise direct) call.
12803 This is the default.
12805 @item -mshort-calls
12806 @opindex short-calls
12807 If not otherwise specified by an attribute, assume all direct calls are
12808 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12809 for direct calls. The default is @option{-mlong-calls}.
12813 Assume addresses can be loaded as 16-bit unsigned values. This does not
12814 apply to function addresses for which @option{-mlong-calls} semantics
12817 @item -mfp-mode=@var{mode}
12819 Set the prevailing mode of the floating-point unit.
12820 This determines the floating-point mode that is provided and expected
12821 at function call and return time. Making this mode match the mode you
12822 predominantly need at function start can make your programs smaller and
12823 faster by avoiding unnecessary mode switches.
12825 @var{mode} can be set to one the following values:
12829 Any mode at function entry is valid, and retained or restored when
12830 the function returns, and when it calls other functions.
12831 This mode is useful for compiling libraries or other compilation units
12832 you might want to incorporate into different programs with different
12833 prevailing FPU modes, and the convenience of being able to use a single
12834 object file outweighs the size and speed overhead for any extra
12835 mode switching that might be needed, compared with what would be needed
12836 with a more specific choice of prevailing FPU mode.
12839 This is the mode used for floating-point calculations with
12840 truncating (i.e.@: round towards zero) rounding mode. That includes
12841 conversion from floating point to integer.
12843 @item round-nearest
12844 This is the mode used for floating-point calculations with
12845 round-to-nearest-or-even rounding mode.
12848 This is the mode used to perform integer calculations in the FPU, e.g.@:
12849 integer multiply, or integer multiply-and-accumulate.
12852 The default is @option{-mfp-mode=caller}
12854 @item -mnosplit-lohi
12855 @itemx -mno-postinc
12856 @itemx -mno-postmodify
12857 @opindex mnosplit-lohi
12858 @opindex mno-postinc
12859 @opindex mno-postmodify
12860 Code generation tweaks that disable, respectively, splitting of 32-bit
12861 loads, generation of post-increment addresses, and generation of
12862 post-modify addresses. The defaults are @option{msplit-lohi},
12863 @option{-mpost-inc}, and @option{-mpost-modify}.
12865 @item -mnovect-double
12866 @opindex mno-vect-double
12867 Change the preferred SIMD mode to SImode. The default is
12868 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12870 @item -max-vect-align=@var{num}
12871 @opindex max-vect-align
12872 The maximum alignment for SIMD vector mode types.
12873 @var{num} may be 4 or 8. The default is 8.
12874 Note that this is an ABI change, even though many library function
12875 interfaces are unaffected if they don't use SIMD vector modes
12876 in places that affect size and/or alignment of relevant types.
12878 @item -msplit-vecmove-early
12879 @opindex msplit-vecmove-early
12880 Split vector moves into single word moves before reload. In theory this
12881 can give better register allocation, but so far the reverse seems to be
12882 generally the case.
12884 @item -m1reg-@var{reg}
12886 Specify a register to hold the constant @minus{}1, which makes loading small negative
12887 constants and certain bitmasks faster.
12888 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12889 which specify use of that register as a fixed register,
12890 and @samp{none}, which means that no register is used for this
12891 purpose. The default is @option{-m1reg-none}.
12896 @subsection ARC Options
12897 @cindex ARC options
12899 The following options control the architecture variant for which code
12902 @c architecture variants
12905 @item -mbarrel-shifter
12906 @opindex mbarrel-shifter
12907 Generate instructions supported by barrel shifter. This is the default
12908 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
12910 @item -mcpu=@var{cpu}
12912 Set architecture type, register usage, and instruction scheduling
12913 parameters for @var{cpu}. There are also shortcut alias options
12914 available for backward compatibility and convenience. Supported
12915 values for @var{cpu} are
12922 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12927 Compile for ARC601. Alias: @option{-mARC601}.
12933 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12934 This is the default when configured with @option{--with-cpu=arc700}@.
12938 Compile for ARC EM.
12942 Compile for ARC HS.
12947 @itemx -mdpfp-compact
12948 @opindex mdpfp-compact
12949 FPX: Generate Double Precision FPX instructions, tuned for the compact
12953 @opindex mdpfp-fast
12954 FPX: Generate Double Precision FPX instructions, tuned for the fast
12957 @item -mno-dpfp-lrsr
12958 @opindex mno-dpfp-lrsr
12959 Disable LR and SR instructions from using FPX extension aux registers.
12963 Generate Extended arithmetic instructions. Currently only
12964 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12965 supported. This is always enabled for @option{-mcpu=ARC700}.
12969 Do not generate mpy instructions for ARC700.
12973 Generate 32x16 bit multiply and mac instructions.
12977 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12981 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12986 @itemx -mspfp-compact
12987 @opindex mspfp-compact
12988 FPX: Generate Single Precision FPX instructions, tuned for the compact
12992 @opindex mspfp-fast
12993 FPX: Generate Single Precision FPX instructions, tuned for the fast
12998 Enable generation of ARC SIMD instructions via target-specific
12999 builtins. Only valid for @option{-mcpu=ARC700}.
13002 @opindex msoft-float
13003 This option ignored; it is provided for compatibility purposes only.
13004 Software floating point code is emitted by default, and this default
13005 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
13006 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
13007 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
13011 Generate swap instructions.
13015 This enables Locked Load/Store Conditional extension to implement
13016 atomic memopry built-in functions. Not available for ARC 6xx or ARC
13021 Enable DIV/REM instructions for ARCv2 cores.
13023 @item -mcode-density
13024 @opindex mcode-density
13025 Enable code density instructions for ARC EM, default on for ARC HS.
13027 @item -mmpy-option=@var{multo}
13028 @opindex mmpy-option
13029 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
13030 the default value. The recognized values for @var{multo} are:
13034 No multiplier available.
13038 The multiply option is set to w: 16x16 multiplier, fully pipelined.
13039 The following instructions are enabled: MPYW, and MPYUW.
13043 The multiply option is set to wlh1: 32x32 multiplier, fully
13044 pipelined (1 stage). The following instructions are additionaly
13045 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
13049 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
13050 (2 stages). The following instructions are additionaly enabled: MPY,
13051 MPYU, MPYM, MPYMU, and MPY_S.
13055 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
13056 sequential. The following instructions are additionaly enabled: MPY,
13057 MPYU, MPYM, MPYMU, and MPY_S.
13061 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
13062 sequential. The following instructions are additionaly enabled: MPY,
13063 MPYU, MPYM, MPYMU, and MPY_S.
13067 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
13068 sequential. The following instructions are additionaly enabled: MPY,
13069 MPYU, MPYM, MPYMU, and MPY_S.
13073 This option is only available for ARCv2 cores@.
13077 The following options are passed through to the assembler, and also
13078 define preprocessor macro symbols.
13080 @c Flags used by the assembler, but for which we define preprocessor
13081 @c macro symbols as well.
13084 @opindex mdsp-packa
13085 Passed down to the assembler to enable the DSP Pack A extensions.
13086 Also sets the preprocessor symbol @code{__Xdsp_packa}.
13090 Passed down to the assembler to enable the dual viterbi butterfly
13091 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
13093 @c ARC700 4.10 extension instruction
13096 Passed down to the assembler to enable the Locked Load/Store
13097 Conditional extension. Also sets the preprocessor symbol
13102 Passed down to the assembler. Also sets the preprocessor symbol
13103 @code{__Xxmac_d16}.
13107 Passed down to the assembler. Also sets the preprocessor symbol
13110 @c ARC700 4.10 extension instruction
13113 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
13114 extension instruction. Also sets the preprocessor symbol
13117 @c ARC700 4.10 extension instruction
13120 Passed down to the assembler to enable the swap byte ordering
13121 extension instruction. Also sets the preprocessor symbol
13125 @opindex mtelephony
13126 Passed down to the assembler to enable dual and single operand
13127 instructions for telephony. Also sets the preprocessor symbol
13128 @code{__Xtelephony}.
13132 Passed down to the assembler to enable the XY Memory extension. Also
13133 sets the preprocessor symbol @code{__Xxy}.
13137 The following options control how the assembly code is annotated:
13139 @c Assembly annotation options
13143 Annotate assembler instructions with estimated addresses.
13145 @item -mannotate-align
13146 @opindex mannotate-align
13147 Explain what alignment considerations lead to the decision to make an
13148 instruction short or long.
13152 The following options are passed through to the linker:
13154 @c options passed through to the linker
13158 Passed through to the linker, to specify use of the @code{arclinux} emulation.
13159 This option is enabled by default in tool chains built for
13160 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
13161 when profiling is not requested.
13163 @item -marclinux_prof
13164 @opindex marclinux_prof
13165 Passed through to the linker, to specify use of the
13166 @code{arclinux_prof} emulation. This option is enabled by default in
13167 tool chains built for @w{@code{arc-linux-uclibc}} and
13168 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
13172 The following options control the semantics of generated code:
13174 @c semantically relevant code generation options
13177 @opindex mlong-calls
13178 Generate call insns as register indirect calls, thus providing access
13179 to the full 32-bit address range.
13181 @item -mmedium-calls
13182 @opindex mmedium-calls
13183 Don't use less than 25 bit addressing range for calls, which is the
13184 offset available for an unconditional branch-and-link
13185 instruction. Conditional execution of function calls is suppressed, to
13186 allow use of the 25-bit range, rather than the 21-bit range with
13187 conditional branch-and-link. This is the default for tool chains built
13188 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13192 Do not generate sdata references. This is the default for tool chains
13193 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13197 @opindex mucb-mcount
13198 Instrument with mcount calls as used in UCB code. I.e. do the
13199 counting in the callee, not the caller. By default ARC instrumentation
13200 counts in the caller.
13202 @item -mvolatile-cache
13203 @opindex mvolatile-cache
13204 Use ordinarily cached memory accesses for volatile references. This is the
13207 @item -mno-volatile-cache
13208 @opindex mno-volatile-cache
13209 Enable cache bypass for volatile references.
13213 The following options fine tune code generation:
13214 @c code generation tuning options
13217 @opindex malign-call
13218 Do alignment optimizations for call instructions.
13220 @item -mauto-modify-reg
13221 @opindex mauto-modify-reg
13222 Enable the use of pre/post modify with register displacement.
13224 @item -mbbit-peephole
13225 @opindex mbbit-peephole
13226 Enable bbit peephole2.
13230 This option disables a target-specific pass in @file{arc_reorg} to
13231 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13232 generation driven by the combiner pass.
13234 @item -mcase-vector-pcrel
13235 @opindex mcase-vector-pcrel
13236 Use pc-relative switch case tables - this enables case table shortening.
13237 This is the default for @option{-Os}.
13239 @item -mcompact-casesi
13240 @opindex mcompact-casesi
13241 Enable compact casesi pattern.
13242 This is the default for @option{-Os}.
13244 @item -mno-cond-exec
13245 @opindex mno-cond-exec
13246 Disable ARCompact specific pass to generate conditional execution instructions.
13247 Due to delay slot scheduling and interactions between operand numbers,
13248 literal sizes, instruction lengths, and the support for conditional execution,
13249 the target-independent pass to generate conditional execution is often lacking,
13250 so the ARC port has kept a special pass around that tries to find more
13251 conditional execution generating opportunities after register allocation,
13252 branch shortening, and delay slot scheduling have been done. This pass
13253 generally, but not always, improves performance and code size, at the cost of
13254 extra compilation time, which is why there is an option to switch it off.
13255 If you have a problem with call instructions exceeding their allowable
13256 offset range because they are conditionalized, you should consider using
13257 @option{-mmedium-calls} instead.
13259 @item -mearly-cbranchsi
13260 @opindex mearly-cbranchsi
13261 Enable pre-reload use of the cbranchsi pattern.
13263 @item -mexpand-adddi
13264 @opindex mexpand-adddi
13265 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13266 @code{add.f}, @code{adc} etc.
13268 @item -mindexed-loads
13269 @opindex mindexed-loads
13270 Enable the use of indexed loads. This can be problematic because some
13271 optimizers then assume that indexed stores exist, which is not
13276 Enable Local Register Allocation. This is still experimental for ARC,
13277 so by default the compiler uses standard reload
13278 (i.e. @option{-mno-lra}).
13280 @item -mlra-priority-none
13281 @opindex mlra-priority-none
13282 Don't indicate any priority for target registers.
13284 @item -mlra-priority-compact
13285 @opindex mlra-priority-compact
13286 Indicate target register priority for r0..r3 / r12..r15.
13288 @item -mlra-priority-noncompact
13289 @opindex mlra-priority-noncompact
13290 Reduce target regsiter priority for r0..r3 / r12..r15.
13292 @item -mno-millicode
13293 @opindex mno-millicode
13294 When optimizing for size (using @option{-Os}), prologues and epilogues
13295 that have to save or restore a large number of registers are often
13296 shortened by using call to a special function in libgcc; this is
13297 referred to as a @emph{millicode} call. As these calls can pose
13298 performance issues, and/or cause linking issues when linking in a
13299 nonstandard way, this option is provided to turn off millicode call
13303 @opindex mmixed-code
13304 Tweak register allocation to help 16-bit instruction generation.
13305 This generally has the effect of decreasing the average instruction size
13306 while increasing the instruction count.
13310 Enable 'q' instruction alternatives.
13311 This is the default for @option{-Os}.
13315 Enable Rcq constraint handling - most short code generation depends on this.
13316 This is the default.
13320 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13321 This is the default.
13323 @item -msize-level=@var{level}
13324 @opindex msize-level
13325 Fine-tune size optimization with regards to instruction lengths and alignment.
13326 The recognized values for @var{level} are:
13329 No size optimization. This level is deprecated and treated like @samp{1}.
13332 Short instructions are used opportunistically.
13335 In addition, alignment of loops and of code after barriers are dropped.
13338 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13342 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13343 the behavior when this is not set is equivalent to level @samp{1}.
13345 @item -mtune=@var{cpu}
13347 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13348 by @option{-mcpu=}.
13350 Supported values for @var{cpu} are
13354 Tune for ARC600 cpu.
13357 Tune for ARC601 cpu.
13360 Tune for ARC700 cpu with standard multiplier block.
13363 Tune for ARC700 cpu with XMAC block.
13366 Tune for ARC725D cpu.
13369 Tune for ARC750D cpu.
13373 @item -mmultcost=@var{num}
13375 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13376 normal instruction.
13378 @item -munalign-prob-threshold=@var{probability}
13379 @opindex munalign-prob-threshold
13380 Set probability threshold for unaligning branches.
13381 When tuning for @samp{ARC700} and optimizing for speed, branches without
13382 filled delay slot are preferably emitted unaligned and long, unless
13383 profiling indicates that the probability for the branch to be taken
13384 is below @var{probability}. @xref{Cross-profiling}.
13385 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13389 The following options are maintained for backward compatibility, but
13390 are now deprecated and will be removed in a future release:
13392 @c Deprecated options
13400 @opindex mbig-endian
13403 Compile code for big endian targets. Use of these options is now
13404 deprecated. Users wanting big-endian code, should use the
13405 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13406 building the tool chain, for which big-endian is the default.
13408 @item -mlittle-endian
13409 @opindex mlittle-endian
13412 Compile code for little endian targets. Use of these options is now
13413 deprecated. Users wanting little-endian code should use the
13414 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13415 building the tool chain, for which little-endian is the default.
13417 @item -mbarrel_shifter
13418 @opindex mbarrel_shifter
13419 Replaced by @option{-mbarrel-shifter}.
13421 @item -mdpfp_compact
13422 @opindex mdpfp_compact
13423 Replaced by @option{-mdpfp-compact}.
13426 @opindex mdpfp_fast
13427 Replaced by @option{-mdpfp-fast}.
13430 @opindex mdsp_packa
13431 Replaced by @option{-mdsp-packa}.
13435 Replaced by @option{-mea}.
13439 Replaced by @option{-mmac-24}.
13443 Replaced by @option{-mmac-d16}.
13445 @item -mspfp_compact
13446 @opindex mspfp_compact
13447 Replaced by @option{-mspfp-compact}.
13450 @opindex mspfp_fast
13451 Replaced by @option{-mspfp-fast}.
13453 @item -mtune=@var{cpu}
13455 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13456 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13457 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13459 @item -multcost=@var{num}
13461 Replaced by @option{-mmultcost}.
13466 @subsection ARM Options
13467 @cindex ARM options
13469 These @samp{-m} options are defined for the ARM port:
13472 @item -mabi=@var{name}
13474 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13475 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13478 @opindex mapcs-frame
13479 Generate a stack frame that is compliant with the ARM Procedure Call
13480 Standard for all functions, even if this is not strictly necessary for
13481 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13482 with this option causes the stack frames not to be generated for
13483 leaf functions. The default is @option{-mno-apcs-frame}.
13484 This option is deprecated.
13488 This is a synonym for @option{-mapcs-frame} and is deprecated.
13491 @c not currently implemented
13492 @item -mapcs-stack-check
13493 @opindex mapcs-stack-check
13494 Generate code to check the amount of stack space available upon entry to
13495 every function (that actually uses some stack space). If there is
13496 insufficient space available then either the function
13497 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13498 called, depending upon the amount of stack space required. The runtime
13499 system is required to provide these functions. The default is
13500 @option{-mno-apcs-stack-check}, since this produces smaller code.
13502 @c not currently implemented
13504 @opindex mapcs-float
13505 Pass floating-point arguments using the floating-point registers. This is
13506 one of the variants of the APCS@. This option is recommended if the
13507 target hardware has a floating-point unit or if a lot of floating-point
13508 arithmetic is going to be performed by the code. The default is
13509 @option{-mno-apcs-float}, since the size of integer-only code is
13510 slightly increased if @option{-mapcs-float} is used.
13512 @c not currently implemented
13513 @item -mapcs-reentrant
13514 @opindex mapcs-reentrant
13515 Generate reentrant, position-independent code. The default is
13516 @option{-mno-apcs-reentrant}.
13519 @item -mthumb-interwork
13520 @opindex mthumb-interwork
13521 Generate code that supports calling between the ARM and Thumb
13522 instruction sets. Without this option, on pre-v5 architectures, the
13523 two instruction sets cannot be reliably used inside one program. The
13524 default is @option{-mno-thumb-interwork}, since slightly larger code
13525 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13526 configurations this option is meaningless.
13528 @item -mno-sched-prolog
13529 @opindex mno-sched-prolog
13530 Prevent the reordering of instructions in the function prologue, or the
13531 merging of those instruction with the instructions in the function's
13532 body. This means that all functions start with a recognizable set
13533 of instructions (or in fact one of a choice from a small set of
13534 different function prologues), and this information can be used to
13535 locate the start of functions inside an executable piece of code. The
13536 default is @option{-msched-prolog}.
13538 @item -mfloat-abi=@var{name}
13539 @opindex mfloat-abi
13540 Specifies which floating-point ABI to use. Permissible values
13541 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13543 Specifying @samp{soft} causes GCC to generate output containing
13544 library calls for floating-point operations.
13545 @samp{softfp} allows the generation of code using hardware floating-point
13546 instructions, but still uses the soft-float calling conventions.
13547 @samp{hard} allows generation of floating-point instructions
13548 and uses FPU-specific calling conventions.
13550 The default depends on the specific target configuration. Note that
13551 the hard-float and soft-float ABIs are not link-compatible; you must
13552 compile your entire program with the same ABI, and link with a
13553 compatible set of libraries.
13555 @item -mlittle-endian
13556 @opindex mlittle-endian
13557 Generate code for a processor running in little-endian mode. This is
13558 the default for all standard configurations.
13561 @opindex mbig-endian
13562 Generate code for a processor running in big-endian mode; the default is
13563 to compile code for a little-endian processor.
13565 @item -march=@var{name}
13567 This specifies the name of the target ARM architecture. GCC uses this
13568 name to determine what kind of instructions it can emit when generating
13569 assembly code. This option can be used in conjunction with or instead
13570 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13571 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13572 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13573 @samp{armv6}, @samp{armv6j},
13574 @samp{armv6t2}, @samp{armv6z}, @samp{armv6kz}, @samp{armv6-m},
13575 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13576 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
13577 @samp{armv8.1-a+crc}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13579 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13582 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13583 architecture together with the optional CRC32 extensions.
13585 @option{-march=native} causes the compiler to auto-detect the architecture
13586 of the build computer. At present, this feature is only supported on
13587 GNU/Linux, and not all architectures are recognized. If the auto-detect
13588 is unsuccessful the option has no effect.
13590 @item -mtune=@var{name}
13592 This option specifies the name of the target ARM processor for
13593 which GCC should tune the performance of the code.
13594 For some ARM implementations better performance can be obtained by using
13596 Permissible names are: @samp{arm2}, @samp{arm250},
13597 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13598 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13599 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13600 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13602 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13603 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13604 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13605 @samp{strongarm1110},
13606 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13607 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13608 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13609 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13610 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13611 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13612 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13613 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
13614 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
13615 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
13617 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13622 @samp{cortex-m0plus},
13623 @samp{cortex-m1.small-multiply},
13624 @samp{cortex-m0.small-multiply},
13625 @samp{cortex-m0plus.small-multiply},
13628 @samp{marvell-pj4},
13629 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13630 @samp{fa526}, @samp{fa626},
13631 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13634 Additionally, this option can specify that GCC should tune the performance
13635 of the code for a big.LITTLE system. Permissible names are:
13636 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
13637 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
13639 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13640 performance for a blend of processors within architecture @var{arch}.
13641 The aim is to generate code that run well on the current most popular
13642 processors, balancing between optimizations that benefit some CPUs in the
13643 range, and avoiding performance pitfalls of other CPUs. The effects of
13644 this option may change in future GCC versions as CPU models come and go.
13646 @option{-mtune=native} causes the compiler to auto-detect the CPU
13647 of the build computer. At present, this feature is only supported on
13648 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13649 unsuccessful the option has no effect.
13651 @item -mcpu=@var{name}
13653 This specifies the name of the target ARM processor. GCC uses this name
13654 to derive the name of the target ARM architecture (as if specified
13655 by @option{-march}) and the ARM processor type for which to tune for
13656 performance (as if specified by @option{-mtune}). Where this option
13657 is used in conjunction with @option{-march} or @option{-mtune},
13658 those options take precedence over the appropriate part of this option.
13660 Permissible names for this option are the same as those for
13663 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13664 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13665 See @option{-mtune} for more information.
13667 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13668 of the build computer. At present, this feature is only supported on
13669 GNU/Linux, and not all architectures are recognized. If the auto-detect
13670 is unsuccessful the option has no effect.
13672 @item -mfpu=@var{name}
13674 This specifies what floating-point hardware (or hardware emulation) is
13675 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13676 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13677 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13678 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13679 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13680 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
13682 If @option{-msoft-float} is specified this specifies the format of
13683 floating-point values.
13685 If the selected floating-point hardware includes the NEON extension
13686 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13687 operations are not generated by GCC's auto-vectorization pass unless
13688 @option{-funsafe-math-optimizations} is also specified. This is
13689 because NEON hardware does not fully implement the IEEE 754 standard for
13690 floating-point arithmetic (in particular denormal values are treated as
13691 zero), so the use of NEON instructions may lead to a loss of precision.
13693 You can also set the fpu name at function level by using the @code{target("fpu=")} function attributes (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13695 @item -mfp16-format=@var{name}
13696 @opindex mfp16-format
13697 Specify the format of the @code{__fp16} half-precision floating-point type.
13698 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13699 the default is @samp{none}, in which case the @code{__fp16} type is not
13700 defined. @xref{Half-Precision}, for more information.
13702 @item -mstructure-size-boundary=@var{n}
13703 @opindex mstructure-size-boundary
13704 The sizes of all structures and unions are rounded up to a multiple
13705 of the number of bits set by this option. Permissible values are 8, 32
13706 and 64. The default value varies for different toolchains. For the COFF
13707 targeted toolchain the default value is 8. A value of 64 is only allowed
13708 if the underlying ABI supports it.
13710 Specifying a larger number can produce faster, more efficient code, but
13711 can also increase the size of the program. Different values are potentially
13712 incompatible. Code compiled with one value cannot necessarily expect to
13713 work with code or libraries compiled with another value, if they exchange
13714 information using structures or unions.
13716 @item -mabort-on-noreturn
13717 @opindex mabort-on-noreturn
13718 Generate a call to the function @code{abort} at the end of a
13719 @code{noreturn} function. It is executed if the function tries to
13723 @itemx -mno-long-calls
13724 @opindex mlong-calls
13725 @opindex mno-long-calls
13726 Tells the compiler to perform function calls by first loading the
13727 address of the function into a register and then performing a subroutine
13728 call on this register. This switch is needed if the target function
13729 lies outside of the 64-megabyte addressing range of the offset-based
13730 version of subroutine call instruction.
13732 Even if this switch is enabled, not all function calls are turned
13733 into long calls. The heuristic is that static functions, functions
13734 that have the @code{short_call} attribute, functions that are inside
13735 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13736 definitions have already been compiled within the current compilation
13737 unit are not turned into long calls. The exceptions to this rule are
13738 that weak function definitions, functions with the @code{long_call}
13739 attribute or the @code{section} attribute, and functions that are within
13740 the scope of a @code{#pragma long_calls} directive are always
13741 turned into long calls.
13743 This feature is not enabled by default. Specifying
13744 @option{-mno-long-calls} restores the default behavior, as does
13745 placing the function calls within the scope of a @code{#pragma
13746 long_calls_off} directive. Note these switches have no effect on how
13747 the compiler generates code to handle function calls via function
13750 @item -msingle-pic-base
13751 @opindex msingle-pic-base
13752 Treat the register used for PIC addressing as read-only, rather than
13753 loading it in the prologue for each function. The runtime system is
13754 responsible for initializing this register with an appropriate value
13755 before execution begins.
13757 @item -mpic-register=@var{reg}
13758 @opindex mpic-register
13759 Specify the register to be used for PIC addressing.
13760 For standard PIC base case, the default is any suitable register
13761 determined by compiler. For single PIC base case, the default is
13762 @samp{R9} if target is EABI based or stack-checking is enabled,
13763 otherwise the default is @samp{R10}.
13765 @item -mpic-data-is-text-relative
13766 @opindex mpic-data-is-text-relative
13767 Assume that each data segments are relative to text segment at load time.
13768 Therefore, it permits addressing data using PC-relative operations.
13769 This option is on by default for targets other than VxWorks RTP.
13771 @item -mpoke-function-name
13772 @opindex mpoke-function-name
13773 Write the name of each function into the text section, directly
13774 preceding the function prologue. The generated code is similar to this:
13778 .ascii "arm_poke_function_name", 0
13781 .word 0xff000000 + (t1 - t0)
13782 arm_poke_function_name
13784 stmfd sp!, @{fp, ip, lr, pc@}
13788 When performing a stack backtrace, code can inspect the value of
13789 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13790 location @code{pc - 12} and the top 8 bits are set, then we know that
13791 there is a function name embedded immediately preceding this location
13792 and has length @code{((pc[-3]) & 0xff000000)}.
13799 Select between generating code that executes in ARM and Thumb
13800 states. The default for most configurations is to generate code
13801 that executes in ARM state, but the default can be changed by
13802 configuring GCC with the @option{--with-mode=}@var{state}
13805 You can also override the ARM and Thumb mode for each function
13806 by using the @code{target("thumb")} and @code{target("arm")} function attributes
13807 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13810 @opindex mtpcs-frame
13811 Generate a stack frame that is compliant with the Thumb Procedure Call
13812 Standard for all non-leaf functions. (A leaf function is one that does
13813 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13815 @item -mtpcs-leaf-frame
13816 @opindex mtpcs-leaf-frame
13817 Generate a stack frame that is compliant with the Thumb Procedure Call
13818 Standard for all leaf functions. (A leaf function is one that does
13819 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13821 @item -mcallee-super-interworking
13822 @opindex mcallee-super-interworking
13823 Gives all externally visible functions in the file being compiled an ARM
13824 instruction set header which switches to Thumb mode before executing the
13825 rest of the function. This allows these functions to be called from
13826 non-interworking code. This option is not valid in AAPCS configurations
13827 because interworking is enabled by default.
13829 @item -mcaller-super-interworking
13830 @opindex mcaller-super-interworking
13831 Allows calls via function pointers (including virtual functions) to
13832 execute correctly regardless of whether the target code has been
13833 compiled for interworking or not. There is a small overhead in the cost
13834 of executing a function pointer if this option is enabled. This option
13835 is not valid in AAPCS configurations because interworking is enabled
13838 @item -mtp=@var{name}
13840 Specify the access model for the thread local storage pointer. The valid
13841 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13842 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13843 (supported in the arm6k architecture), and @samp{auto}, which uses the
13844 best available method for the selected processor. The default setting is
13847 @item -mtls-dialect=@var{dialect}
13848 @opindex mtls-dialect
13849 Specify the dialect to use for accessing thread local storage. Two
13850 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13851 @samp{gnu} dialect selects the original GNU scheme for supporting
13852 local and global dynamic TLS models. The @samp{gnu2} dialect
13853 selects the GNU descriptor scheme, which provides better performance
13854 for shared libraries. The GNU descriptor scheme is compatible with
13855 the original scheme, but does require new assembler, linker and
13856 library support. Initial and local exec TLS models are unaffected by
13857 this option and always use the original scheme.
13859 @item -mword-relocations
13860 @opindex mword-relocations
13861 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13862 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13863 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13866 @item -mfix-cortex-m3-ldrd
13867 @opindex mfix-cortex-m3-ldrd
13868 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13869 with overlapping destination and base registers are used. This option avoids
13870 generating these instructions. This option is enabled by default when
13871 @option{-mcpu=cortex-m3} is specified.
13873 @item -munaligned-access
13874 @itemx -mno-unaligned-access
13875 @opindex munaligned-access
13876 @opindex mno-unaligned-access
13877 Enables (or disables) reading and writing of 16- and 32- bit values
13878 from addresses that are not 16- or 32- bit aligned. By default
13879 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13880 architectures, and enabled for all other architectures. If unaligned
13881 access is not enabled then words in packed data structures are
13882 accessed a byte at a time.
13884 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13885 generated object file to either true or false, depending upon the
13886 setting of this option. If unaligned access is enabled then the
13887 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13890 @item -mneon-for-64bits
13891 @opindex mneon-for-64bits
13892 Enables using Neon to handle scalar 64-bits operations. This is
13893 disabled by default since the cost of moving data from core registers
13896 @item -mslow-flash-data
13897 @opindex mslow-flash-data
13898 Assume loading data from flash is slower than fetching instruction.
13899 Therefore literal load is minimized for better performance.
13900 This option is only supported when compiling for ARMv7 M-profile and
13903 @item -masm-syntax-unified
13904 @opindex masm-syntax-unified
13905 Assume inline assembler is using unified asm syntax. The default is
13906 currently off which implies divided syntax. This option has no impact
13907 on Thumb2. However, this may change in future releases of GCC.
13908 Divided syntax should be considered deprecated.
13910 @item -mrestrict-it
13911 @opindex mrestrict-it
13912 Restricts generation of IT blocks to conform to the rules of ARMv8.
13913 IT blocks can only contain a single 16-bit instruction from a select
13914 set of instructions. This option is on by default for ARMv8 Thumb mode.
13916 @item -mprint-tune-info
13917 @opindex mprint-tune-info
13918 Print CPU tuning information as comment in assembler file. This is
13919 an option used only for regression testing of the compiler and not
13920 intended for ordinary use in compiling code. This option is disabled
13925 @subsection AVR Options
13926 @cindex AVR Options
13928 These options are defined for AVR implementations:
13931 @item -mmcu=@var{mcu}
13933 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13935 The default for this option is@tie{}@samp{avr2}.
13937 GCC supports the following AVR devices and ISAs:
13939 @include avr-mmcu.texi
13941 @item -maccumulate-args
13942 @opindex maccumulate-args
13943 Accumulate outgoing function arguments and acquire/release the needed
13944 stack space for outgoing function arguments once in function
13945 prologue/epilogue. Without this option, outgoing arguments are pushed
13946 before calling a function and popped afterwards.
13948 Popping the arguments after the function call can be expensive on
13949 AVR so that accumulating the stack space might lead to smaller
13950 executables because arguments need not to be removed from the
13951 stack after such a function call.
13953 This option can lead to reduced code size for functions that perform
13954 several calls to functions that get their arguments on the stack like
13955 calls to printf-like functions.
13957 @item -mbranch-cost=@var{cost}
13958 @opindex mbranch-cost
13959 Set the branch costs for conditional branch instructions to
13960 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13961 integers. The default branch cost is 0.
13963 @item -mcall-prologues
13964 @opindex mcall-prologues
13965 Functions prologues/epilogues are expanded as calls to appropriate
13966 subroutines. Code size is smaller.
13970 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13971 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13972 and @code{long long} is 4 bytes. Please note that this option does not
13973 conform to the C standards, but it results in smaller code
13976 @item -mn-flash=@var{num}
13978 Assume that the flash memory has a size of
13979 @var{num} times 64@tie{}KiB.
13981 @item -mno-interrupts
13982 @opindex mno-interrupts
13983 Generated code is not compatible with hardware interrupts.
13984 Code size is smaller.
13988 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13989 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13990 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13991 the assembler's command line and the @option{--relax} option to the
13992 linker's command line.
13994 Jump relaxing is performed by the linker because jump offsets are not
13995 known before code is located. Therefore, the assembler code generated by the
13996 compiler is the same, but the instructions in the executable may
13997 differ from instructions in the assembler code.
13999 Relaxing must be turned on if linker stubs are needed, see the
14000 section on @code{EIND} and linker stubs below.
14004 Assume that the device supports the Read-Modify-Write
14005 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
14009 Treat the stack pointer register as an 8-bit register,
14010 i.e.@: assume the high byte of the stack pointer is zero.
14011 In general, you don't need to set this option by hand.
14013 This option is used internally by the compiler to select and
14014 build multilibs for architectures @code{avr2} and @code{avr25}.
14015 These architectures mix devices with and without @code{SPH}.
14016 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
14017 the compiler driver adds or removes this option from the compiler
14018 proper's command line, because the compiler then knows if the device
14019 or architecture has an 8-bit stack pointer and thus no @code{SPH}
14024 Use address register @code{X} in a way proposed by the hardware. This means
14025 that @code{X} is only used in indirect, post-increment or
14026 pre-decrement addressing.
14028 Without this option, the @code{X} register may be used in the same way
14029 as @code{Y} or @code{Z} which then is emulated by additional
14031 For example, loading a value with @code{X+const} addressing with a
14032 small non-negative @code{const < 64} to a register @var{Rn} is
14036 adiw r26, const ; X += const
14037 ld @var{Rn}, X ; @var{Rn} = *X
14038 sbiw r26, const ; X -= const
14042 @opindex mtiny-stack
14043 Only change the lower 8@tie{}bits of the stack pointer.
14046 @opindex nodevicelib
14047 Don't link against AVR-LibC's device specific library @code{libdev.a}.
14049 @item -Waddr-space-convert
14050 @opindex Waddr-space-convert
14051 Warn about conversions between address spaces in the case where the
14052 resulting address space is not contained in the incoming address space.
14055 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
14056 @cindex @code{EIND}
14057 Pointers in the implementation are 16@tie{}bits wide.
14058 The address of a function or label is represented as word address so
14059 that indirect jumps and calls can target any code address in the
14060 range of 64@tie{}Ki words.
14062 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
14063 bytes of program memory space, there is a special function register called
14064 @code{EIND} that serves as most significant part of the target address
14065 when @code{EICALL} or @code{EIJMP} instructions are used.
14067 Indirect jumps and calls on these devices are handled as follows by
14068 the compiler and are subject to some limitations:
14073 The compiler never sets @code{EIND}.
14076 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
14077 instructions or might read @code{EIND} directly in order to emulate an
14078 indirect call/jump by means of a @code{RET} instruction.
14081 The compiler assumes that @code{EIND} never changes during the startup
14082 code or during the application. In particular, @code{EIND} is not
14083 saved/restored in function or interrupt service routine
14087 For indirect calls to functions and computed goto, the linker
14088 generates @emph{stubs}. Stubs are jump pads sometimes also called
14089 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
14090 The stub contains a direct jump to the desired address.
14093 Linker relaxation must be turned on so that the linker generates
14094 the stubs correctly in all situations. See the compiler option
14095 @option{-mrelax} and the linker option @option{--relax}.
14096 There are corner cases where the linker is supposed to generate stubs
14097 but aborts without relaxation and without a helpful error message.
14100 The default linker script is arranged for code with @code{EIND = 0}.
14101 If code is supposed to work for a setup with @code{EIND != 0}, a custom
14102 linker script has to be used in order to place the sections whose
14103 name start with @code{.trampolines} into the segment where @code{EIND}
14107 The startup code from libgcc never sets @code{EIND}.
14108 Notice that startup code is a blend of code from libgcc and AVR-LibC.
14109 For the impact of AVR-LibC on @code{EIND}, see the
14110 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
14113 It is legitimate for user-specific startup code to set up @code{EIND}
14114 early, for example by means of initialization code located in
14115 section @code{.init3}. Such code runs prior to general startup code
14116 that initializes RAM and calls constructors, but after the bit
14117 of startup code from AVR-LibC that sets @code{EIND} to the segment
14118 where the vector table is located.
14120 #include <avr/io.h>
14123 __attribute__((section(".init3"),naked,used,no_instrument_function))
14124 init3_set_eind (void)
14126 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
14127 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
14132 The @code{__trampolines_start} symbol is defined in the linker script.
14135 Stubs are generated automatically by the linker if
14136 the following two conditions are met:
14139 @item The address of a label is taken by means of the @code{gs} modifier
14140 (short for @emph{generate stubs}) like so:
14142 LDI r24, lo8(gs(@var{func}))
14143 LDI r25, hi8(gs(@var{func}))
14145 @item The final location of that label is in a code segment
14146 @emph{outside} the segment where the stubs are located.
14150 The compiler emits such @code{gs} modifiers for code labels in the
14151 following situations:
14153 @item Taking address of a function or code label.
14154 @item Computed goto.
14155 @item If prologue-save function is used, see @option{-mcall-prologues}
14156 command-line option.
14157 @item Switch/case dispatch tables. If you do not want such dispatch
14158 tables you can specify the @option{-fno-jump-tables} command-line option.
14159 @item C and C++ constructors/destructors called during startup/shutdown.
14160 @item If the tools hit a @code{gs()} modifier explained above.
14164 Jumping to non-symbolic addresses like so is @emph{not} supported:
14169 /* Call function at word address 0x2 */
14170 return ((int(*)(void)) 0x2)();
14174 Instead, a stub has to be set up, i.e.@: the function has to be called
14175 through a symbol (@code{func_4} in the example):
14180 extern int func_4 (void);
14182 /* Call function at byte address 0x4 */
14187 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
14188 Alternatively, @code{func_4} can be defined in the linker script.
14191 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14192 @cindex @code{RAMPD}
14193 @cindex @code{RAMPX}
14194 @cindex @code{RAMPY}
14195 @cindex @code{RAMPZ}
14196 Some AVR devices support memories larger than the 64@tie{}KiB range
14197 that can be accessed with 16-bit pointers. To access memory locations
14198 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14199 register is used as high part of the address:
14200 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14201 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14202 register, respectively, to get a wide address. Similarly,
14203 @code{RAMPD} is used together with direct addressing.
14207 The startup code initializes the @code{RAMP} special function
14208 registers with zero.
14211 If a @ref{AVR Named Address Spaces,named address space} other than
14212 generic or @code{__flash} is used, then @code{RAMPZ} is set
14213 as needed before the operation.
14216 If the device supports RAM larger than 64@tie{}KiB and the compiler
14217 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14218 is reset to zero after the operation.
14221 If the device comes with a specific @code{RAMP} register, the ISR
14222 prologue/epilogue saves/restores that SFR and initializes it with
14223 zero in case the ISR code might (implicitly) use it.
14226 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14227 If you use inline assembler to read from locations outside the
14228 16-bit address range and change one of the @code{RAMP} registers,
14229 you must reset it to zero after the access.
14233 @subsubsection AVR Built-in Macros
14235 GCC defines several built-in macros so that the user code can test
14236 for the presence or absence of features. Almost any of the following
14237 built-in macros are deduced from device capabilities and thus
14238 triggered by the @option{-mmcu=} command-line option.
14240 For even more AVR-specific built-in macros see
14241 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14246 Build-in macro that resolves to a decimal number that identifies the
14247 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14248 Possible values are:
14250 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14251 @code{4}, @code{5}, @code{51}, @code{6}
14253 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14254 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14258 @code{100}, @code{102}, @code{104},
14259 @code{105}, @code{106}, @code{107}
14261 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14262 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14263 If @var{mcu} specifies a device, this built-in macro is set
14264 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14265 defined to @code{4}.
14267 @item __AVR_@var{Device}__
14268 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14269 the device's name. For example, @option{-mmcu=atmega8} defines the
14270 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14271 @code{__AVR_ATtiny261A__}, etc.
14273 The built-in macros' names follow
14274 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14275 the device name as from the AVR user manual. The difference between
14276 @var{Device} in the built-in macro and @var{device} in
14277 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14279 If @var{device} is not a device but only a core architecture like
14280 @samp{avr51}, this macro is not defined.
14282 @item __AVR_DEVICE_NAME__
14283 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14284 the device's name. For example, with @option{-mmcu=atmega8} the macro
14285 is defined to @code{atmega8}.
14287 If @var{device} is not a device but only a core architecture like
14288 @samp{avr51}, this macro is not defined.
14290 @item __AVR_XMEGA__
14291 The device / architecture belongs to the XMEGA family of devices.
14293 @item __AVR_HAVE_ELPM__
14294 The device has the @code{ELPM} instruction.
14296 @item __AVR_HAVE_ELPMX__
14297 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14298 R@var{n},Z+} instructions.
14300 @item __AVR_HAVE_MOVW__
14301 The device has the @code{MOVW} instruction to perform 16-bit
14302 register-register moves.
14304 @item __AVR_HAVE_LPMX__
14305 The device has the @code{LPM R@var{n},Z} and
14306 @code{LPM R@var{n},Z+} instructions.
14308 @item __AVR_HAVE_MUL__
14309 The device has a hardware multiplier.
14311 @item __AVR_HAVE_JMP_CALL__
14312 The device has the @code{JMP} and @code{CALL} instructions.
14313 This is the case for devices with at least 16@tie{}KiB of program
14316 @item __AVR_HAVE_EIJMP_EICALL__
14317 @itemx __AVR_3_BYTE_PC__
14318 The device has the @code{EIJMP} and @code{EICALL} instructions.
14319 This is the case for devices with more than 128@tie{}KiB of program memory.
14320 This also means that the program counter
14321 (PC) is 3@tie{}bytes wide.
14323 @item __AVR_2_BYTE_PC__
14324 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14325 with up to 128@tie{}KiB of program memory.
14327 @item __AVR_HAVE_8BIT_SP__
14328 @itemx __AVR_HAVE_16BIT_SP__
14329 The stack pointer (SP) register is treated as 8-bit respectively
14330 16-bit register by the compiler.
14331 The definition of these macros is affected by @option{-mtiny-stack}.
14333 @item __AVR_HAVE_SPH__
14335 The device has the SPH (high part of stack pointer) special function
14336 register or has an 8-bit stack pointer, respectively.
14337 The definition of these macros is affected by @option{-mmcu=} and
14338 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14341 @item __AVR_HAVE_RAMPD__
14342 @itemx __AVR_HAVE_RAMPX__
14343 @itemx __AVR_HAVE_RAMPY__
14344 @itemx __AVR_HAVE_RAMPZ__
14345 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14346 @code{RAMPZ} special function register, respectively.
14348 @item __NO_INTERRUPTS__
14349 This macro reflects the @option{-mno-interrupts} command-line option.
14351 @item __AVR_ERRATA_SKIP__
14352 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14353 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14354 instructions because of a hardware erratum. Skip instructions are
14355 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14356 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14359 @item __AVR_ISA_RMW__
14360 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14362 @item __AVR_SFR_OFFSET__=@var{offset}
14363 Instructions that can address I/O special function registers directly
14364 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14365 address as if addressed by an instruction to access RAM like @code{LD}
14366 or @code{STS}. This offset depends on the device architecture and has
14367 to be subtracted from the RAM address in order to get the
14368 respective I/O@tie{}address.
14370 @item __WITH_AVRLIBC__
14371 The compiler is configured to be used together with AVR-Libc.
14372 See the @option{--with-avrlibc} configure option.
14376 @node Blackfin Options
14377 @subsection Blackfin Options
14378 @cindex Blackfin Options
14381 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14383 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14384 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14385 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14386 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14387 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14388 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14389 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14390 @samp{bf561}, @samp{bf592}.
14392 The optional @var{sirevision} specifies the silicon revision of the target
14393 Blackfin processor. Any workarounds available for the targeted silicon revision
14394 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14395 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14396 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14397 hexadecimal digits representing the major and minor numbers in the silicon
14398 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14399 is not defined. If @var{sirevision} is @samp{any}, the
14400 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14401 If this optional @var{sirevision} is not used, GCC assumes the latest known
14402 silicon revision of the targeted Blackfin processor.
14404 GCC defines a preprocessor macro for the specified @var{cpu}.
14405 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14406 provided by libgloss to be linked in if @option{-msim} is not given.
14408 Without this option, @samp{bf532} is used as the processor by default.
14410 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14411 only the preprocessor macro is defined.
14415 Specifies that the program will be run on the simulator. This causes
14416 the simulator BSP provided by libgloss to be linked in. This option
14417 has effect only for @samp{bfin-elf} toolchain.
14418 Certain other options, such as @option{-mid-shared-library} and
14419 @option{-mfdpic}, imply @option{-msim}.
14421 @item -momit-leaf-frame-pointer
14422 @opindex momit-leaf-frame-pointer
14423 Don't keep the frame pointer in a register for leaf functions. This
14424 avoids the instructions to save, set up and restore frame pointers and
14425 makes an extra register available in leaf functions. The option
14426 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14427 which might make debugging harder.
14429 @item -mspecld-anomaly
14430 @opindex mspecld-anomaly
14431 When enabled, the compiler ensures that the generated code does not
14432 contain speculative loads after jump instructions. If this option is used,
14433 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14435 @item -mno-specld-anomaly
14436 @opindex mno-specld-anomaly
14437 Don't generate extra code to prevent speculative loads from occurring.
14439 @item -mcsync-anomaly
14440 @opindex mcsync-anomaly
14441 When enabled, the compiler ensures that the generated code does not
14442 contain CSYNC or SSYNC instructions too soon after conditional branches.
14443 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14445 @item -mno-csync-anomaly
14446 @opindex mno-csync-anomaly
14447 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14448 occurring too soon after a conditional branch.
14452 When enabled, the compiler is free to take advantage of the knowledge that
14453 the entire program fits into the low 64k of memory.
14456 @opindex mno-low-64k
14457 Assume that the program is arbitrarily large. This is the default.
14459 @item -mstack-check-l1
14460 @opindex mstack-check-l1
14461 Do stack checking using information placed into L1 scratchpad memory by the
14464 @item -mid-shared-library
14465 @opindex mid-shared-library
14466 Generate code that supports shared libraries via the library ID method.
14467 This allows for execute in place and shared libraries in an environment
14468 without virtual memory management. This option implies @option{-fPIC}.
14469 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14471 @item -mno-id-shared-library
14472 @opindex mno-id-shared-library
14473 Generate code that doesn't assume ID-based shared libraries are being used.
14474 This is the default.
14476 @item -mleaf-id-shared-library
14477 @opindex mleaf-id-shared-library
14478 Generate code that supports shared libraries via the library ID method,
14479 but assumes that this library or executable won't link against any other
14480 ID shared libraries. That allows the compiler to use faster code for jumps
14483 @item -mno-leaf-id-shared-library
14484 @opindex mno-leaf-id-shared-library
14485 Do not assume that the code being compiled won't link against any ID shared
14486 libraries. Slower code is generated for jump and call insns.
14488 @item -mshared-library-id=n
14489 @opindex mshared-library-id
14490 Specifies the identification number of the ID-based shared library being
14491 compiled. Specifying a value of 0 generates more compact code; specifying
14492 other values forces the allocation of that number to the current
14493 library but is no more space- or time-efficient than omitting this option.
14497 Generate code that allows the data segment to be located in a different
14498 area of memory from the text segment. This allows for execute in place in
14499 an environment without virtual memory management by eliminating relocations
14500 against the text section.
14502 @item -mno-sep-data
14503 @opindex mno-sep-data
14504 Generate code that assumes that the data segment follows the text segment.
14505 This is the default.
14508 @itemx -mno-long-calls
14509 @opindex mlong-calls
14510 @opindex mno-long-calls
14511 Tells the compiler to perform function calls by first loading the
14512 address of the function into a register and then performing a subroutine
14513 call on this register. This switch is needed if the target function
14514 lies outside of the 24-bit addressing range of the offset-based
14515 version of subroutine call instruction.
14517 This feature is not enabled by default. Specifying
14518 @option{-mno-long-calls} restores the default behavior. Note these
14519 switches have no effect on how the compiler generates code to handle
14520 function calls via function pointers.
14524 Link with the fast floating-point library. This library relaxes some of
14525 the IEEE floating-point standard's rules for checking inputs against
14526 Not-a-Number (NAN), in the interest of performance.
14529 @opindex minline-plt
14530 Enable inlining of PLT entries in function calls to functions that are
14531 not known to bind locally. It has no effect without @option{-mfdpic}.
14534 @opindex mmulticore
14535 Build a standalone application for multicore Blackfin processors.
14536 This option causes proper start files and link scripts supporting
14537 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14538 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14540 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14541 selects the one-application-per-core programming model. Without
14542 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14543 programming model is used. In this model, the main function of Core B
14544 should be named as @code{coreb_main}.
14546 If this option is not used, the single-core application programming
14551 Build a standalone application for Core A of BF561 when using
14552 the one-application-per-core programming model. Proper start files
14553 and link scripts are used to support Core A, and the macro
14554 @code{__BFIN_COREA} is defined.
14555 This option can only be used in conjunction with @option{-mmulticore}.
14559 Build a standalone application for Core B of BF561 when using
14560 the one-application-per-core programming model. Proper start files
14561 and link scripts are used to support Core B, and the macro
14562 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14563 should be used instead of @code{main}.
14564 This option can only be used in conjunction with @option{-mmulticore}.
14568 Build a standalone application for SDRAM. Proper start files and
14569 link scripts are used to put the application into SDRAM, and the macro
14570 @code{__BFIN_SDRAM} is defined.
14571 The loader should initialize SDRAM before loading the application.
14575 Assume that ICPLBs are enabled at run time. This has an effect on certain
14576 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14577 are enabled; for standalone applications the default is off.
14581 @subsection C6X Options
14582 @cindex C6X Options
14585 @item -march=@var{name}
14587 This specifies the name of the target architecture. GCC uses this
14588 name to determine what kind of instructions it can emit when generating
14589 assembly code. Permissible names are: @samp{c62x},
14590 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14593 @opindex mbig-endian
14594 Generate code for a big-endian target.
14596 @item -mlittle-endian
14597 @opindex mlittle-endian
14598 Generate code for a little-endian target. This is the default.
14602 Choose startup files and linker script suitable for the simulator.
14604 @item -msdata=default
14605 @opindex msdata=default
14606 Put small global and static data in the @code{.neardata} section,
14607 which is pointed to by register @code{B14}. Put small uninitialized
14608 global and static data in the @code{.bss} section, which is adjacent
14609 to the @code{.neardata} section. Put small read-only data into the
14610 @code{.rodata} section. The corresponding sections used for large
14611 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14614 @opindex msdata=all
14615 Put all data, not just small objects, into the sections reserved for
14616 small data, and use addressing relative to the @code{B14} register to
14620 @opindex msdata=none
14621 Make no use of the sections reserved for small data, and use absolute
14622 addresses to access all data. Put all initialized global and static
14623 data in the @code{.fardata} section, and all uninitialized data in the
14624 @code{.far} section. Put all constant data into the @code{.const}
14629 @subsection CRIS Options
14630 @cindex CRIS Options
14632 These options are defined specifically for the CRIS ports.
14635 @item -march=@var{architecture-type}
14636 @itemx -mcpu=@var{architecture-type}
14639 Generate code for the specified architecture. The choices for
14640 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14641 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14642 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14645 @item -mtune=@var{architecture-type}
14647 Tune to @var{architecture-type} everything applicable about the generated
14648 code, except for the ABI and the set of available instructions. The
14649 choices for @var{architecture-type} are the same as for
14650 @option{-march=@var{architecture-type}}.
14652 @item -mmax-stack-frame=@var{n}
14653 @opindex mmax-stack-frame
14654 Warn when the stack frame of a function exceeds @var{n} bytes.
14660 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14661 @option{-march=v3} and @option{-march=v8} respectively.
14663 @item -mmul-bug-workaround
14664 @itemx -mno-mul-bug-workaround
14665 @opindex mmul-bug-workaround
14666 @opindex mno-mul-bug-workaround
14667 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14668 models where it applies. This option is active by default.
14672 Enable CRIS-specific verbose debug-related information in the assembly
14673 code. This option also has the effect of turning off the @samp{#NO_APP}
14674 formatted-code indicator to the assembler at the beginning of the
14679 Do not use condition-code results from previous instruction; always emit
14680 compare and test instructions before use of condition codes.
14682 @item -mno-side-effects
14683 @opindex mno-side-effects
14684 Do not emit instructions with side effects in addressing modes other than
14687 @item -mstack-align
14688 @itemx -mno-stack-align
14689 @itemx -mdata-align
14690 @itemx -mno-data-align
14691 @itemx -mconst-align
14692 @itemx -mno-const-align
14693 @opindex mstack-align
14694 @opindex mno-stack-align
14695 @opindex mdata-align
14696 @opindex mno-data-align
14697 @opindex mconst-align
14698 @opindex mno-const-align
14699 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14700 stack frame, individual data and constants to be aligned for the maximum
14701 single data access size for the chosen CPU model. The default is to
14702 arrange for 32-bit alignment. ABI details such as structure layout are
14703 not affected by these options.
14711 Similar to the stack- data- and const-align options above, these options
14712 arrange for stack frame, writable data and constants to all be 32-bit,
14713 16-bit or 8-bit aligned. The default is 32-bit alignment.
14715 @item -mno-prologue-epilogue
14716 @itemx -mprologue-epilogue
14717 @opindex mno-prologue-epilogue
14718 @opindex mprologue-epilogue
14719 With @option{-mno-prologue-epilogue}, the normal function prologue and
14720 epilogue which set up the stack frame are omitted and no return
14721 instructions or return sequences are generated in the code. Use this
14722 option only together with visual inspection of the compiled code: no
14723 warnings or errors are generated when call-saved registers must be saved,
14724 or storage for local variables needs to be allocated.
14728 @opindex mno-gotplt
14730 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14731 instruction sequences that load addresses for functions from the PLT part
14732 of the GOT rather than (traditional on other architectures) calls to the
14733 PLT@. The default is @option{-mgotplt}.
14737 Legacy no-op option only recognized with the cris-axis-elf and
14738 cris-axis-linux-gnu targets.
14742 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14746 This option, recognized for the cris-axis-elf, arranges
14747 to link with input-output functions from a simulator library. Code,
14748 initialized data and zero-initialized data are allocated consecutively.
14752 Like @option{-sim}, but pass linker options to locate initialized data at
14753 0x40000000 and zero-initialized data at 0x80000000.
14757 @subsection CR16 Options
14758 @cindex CR16 Options
14760 These options are defined specifically for the CR16 ports.
14766 Enable the use of multiply-accumulate instructions. Disabled by default.
14770 @opindex mcr16cplus
14772 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14777 Links the library libsim.a which is in compatible with simulator. Applicable
14778 to ELF compiler only.
14782 Choose integer type as 32-bit wide.
14786 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14788 @item -mdata-model=@var{model}
14789 @opindex mdata-model
14790 Choose a data model. The choices for @var{model} are @samp{near},
14791 @samp{far} or @samp{medium}. @samp{medium} is default.
14792 However, @samp{far} is not valid with @option{-mcr16c}, as the
14793 CR16C architecture does not support the far data model.
14796 @node Darwin Options
14797 @subsection Darwin Options
14798 @cindex Darwin options
14800 These options are defined for all architectures running the Darwin operating
14803 FSF GCC on Darwin does not create ``fat'' object files; it creates
14804 an object file for the single architecture that GCC was built to
14805 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14806 @option{-arch} options are used; it does so by running the compiler or
14807 linker multiple times and joining the results together with
14810 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14811 @samp{i686}) is determined by the flags that specify the ISA
14812 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14813 @option{-force_cpusubtype_ALL} option can be used to override this.
14815 The Darwin tools vary in their behavior when presented with an ISA
14816 mismatch. The assembler, @file{as}, only permits instructions to
14817 be used that are valid for the subtype of the file it is generating,
14818 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14819 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14820 and prints an error if asked to create a shared library with a less
14821 restrictive subtype than its input files (for instance, trying to put
14822 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14823 for executables, @command{ld}, quietly gives the executable the most
14824 restrictive subtype of any of its input files.
14829 Add the framework directory @var{dir} to the head of the list of
14830 directories to be searched for header files. These directories are
14831 interleaved with those specified by @option{-I} options and are
14832 scanned in a left-to-right order.
14834 A framework directory is a directory with frameworks in it. A
14835 framework is a directory with a @file{Headers} and/or
14836 @file{PrivateHeaders} directory contained directly in it that ends
14837 in @file{.framework}. The name of a framework is the name of this
14838 directory excluding the @file{.framework}. Headers associated with
14839 the framework are found in one of those two directories, with
14840 @file{Headers} being searched first. A subframework is a framework
14841 directory that is in a framework's @file{Frameworks} directory.
14842 Includes of subframework headers can only appear in a header of a
14843 framework that contains the subframework, or in a sibling subframework
14844 header. Two subframeworks are siblings if they occur in the same
14845 framework. A subframework should not have the same name as a
14846 framework; a warning is issued if this is violated. Currently a
14847 subframework cannot have subframeworks; in the future, the mechanism
14848 may be extended to support this. The standard frameworks can be found
14849 in @file{/System/Library/Frameworks} and
14850 @file{/Library/Frameworks}. An example include looks like
14851 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14852 the name of the framework and @file{header.h} is found in the
14853 @file{PrivateHeaders} or @file{Headers} directory.
14855 @item -iframework@var{dir}
14856 @opindex iframework
14857 Like @option{-F} except the directory is a treated as a system
14858 directory. The main difference between this @option{-iframework} and
14859 @option{-F} is that with @option{-iframework} the compiler does not
14860 warn about constructs contained within header files found via
14861 @var{dir}. This option is valid only for the C family of languages.
14865 Emit debugging information for symbols that are used. For stabs
14866 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14867 This is by default ON@.
14871 Emit debugging information for all symbols and types.
14873 @item -mmacosx-version-min=@var{version}
14874 The earliest version of MacOS X that this executable will run on
14875 is @var{version}. Typical values of @var{version} include @code{10.1},
14876 @code{10.2}, and @code{10.3.9}.
14878 If the compiler was built to use the system's headers by default,
14879 then the default for this option is the system version on which the
14880 compiler is running, otherwise the default is to make choices that
14881 are compatible with as many systems and code bases as possible.
14885 Enable kernel development mode. The @option{-mkernel} option sets
14886 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14887 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14888 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14889 applicable. This mode also sets @option{-mno-altivec},
14890 @option{-msoft-float}, @option{-fno-builtin} and
14891 @option{-mlong-branch} for PowerPC targets.
14893 @item -mone-byte-bool
14894 @opindex mone-byte-bool
14895 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14896 By default @code{sizeof(bool)} is @code{4} when compiling for
14897 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14898 option has no effect on x86.
14900 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14901 to generate code that is not binary compatible with code generated
14902 without that switch. Using this switch may require recompiling all
14903 other modules in a program, including system libraries. Use this
14904 switch to conform to a non-default data model.
14906 @item -mfix-and-continue
14907 @itemx -ffix-and-continue
14908 @itemx -findirect-data
14909 @opindex mfix-and-continue
14910 @opindex ffix-and-continue
14911 @opindex findirect-data
14912 Generate code suitable for fast turnaround development, such as to
14913 allow GDB to dynamically load @file{.o} files into already-running
14914 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14915 are provided for backwards compatibility.
14919 Loads all members of static archive libraries.
14920 See man ld(1) for more information.
14922 @item -arch_errors_fatal
14923 @opindex arch_errors_fatal
14924 Cause the errors having to do with files that have the wrong architecture
14927 @item -bind_at_load
14928 @opindex bind_at_load
14929 Causes the output file to be marked such that the dynamic linker will
14930 bind all undefined references when the file is loaded or launched.
14934 Produce a Mach-o bundle format file.
14935 See man ld(1) for more information.
14937 @item -bundle_loader @var{executable}
14938 @opindex bundle_loader
14939 This option specifies the @var{executable} that will load the build
14940 output file being linked. See man ld(1) for more information.
14943 @opindex dynamiclib
14944 When passed this option, GCC produces a dynamic library instead of
14945 an executable when linking, using the Darwin @file{libtool} command.
14947 @item -force_cpusubtype_ALL
14948 @opindex force_cpusubtype_ALL
14949 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14950 one controlled by the @option{-mcpu} or @option{-march} option.
14952 @item -allowable_client @var{client_name}
14953 @itemx -client_name
14954 @itemx -compatibility_version
14955 @itemx -current_version
14957 @itemx -dependency-file
14959 @itemx -dylinker_install_name
14961 @itemx -exported_symbols_list
14964 @itemx -flat_namespace
14965 @itemx -force_flat_namespace
14966 @itemx -headerpad_max_install_names
14969 @itemx -install_name
14970 @itemx -keep_private_externs
14971 @itemx -multi_module
14972 @itemx -multiply_defined
14973 @itemx -multiply_defined_unused
14976 @itemx -no_dead_strip_inits_and_terms
14977 @itemx -nofixprebinding
14978 @itemx -nomultidefs
14980 @itemx -noseglinkedit
14981 @itemx -pagezero_size
14983 @itemx -prebind_all_twolevel_modules
14984 @itemx -private_bundle
14986 @itemx -read_only_relocs
14988 @itemx -sectobjectsymbols
14992 @itemx -sectobjectsymbols
14995 @itemx -segs_read_only_addr
14997 @itemx -segs_read_write_addr
14998 @itemx -seg_addr_table
14999 @itemx -seg_addr_table_filename
15000 @itemx -seglinkedit
15002 @itemx -segs_read_only_addr
15003 @itemx -segs_read_write_addr
15004 @itemx -single_module
15006 @itemx -sub_library
15008 @itemx -sub_umbrella
15009 @itemx -twolevel_namespace
15012 @itemx -unexported_symbols_list
15013 @itemx -weak_reference_mismatches
15014 @itemx -whatsloaded
15015 @opindex allowable_client
15016 @opindex client_name
15017 @opindex compatibility_version
15018 @opindex current_version
15019 @opindex dead_strip
15020 @opindex dependency-file
15021 @opindex dylib_file
15022 @opindex dylinker_install_name
15024 @opindex exported_symbols_list
15026 @opindex flat_namespace
15027 @opindex force_flat_namespace
15028 @opindex headerpad_max_install_names
15029 @opindex image_base
15031 @opindex install_name
15032 @opindex keep_private_externs
15033 @opindex multi_module
15034 @opindex multiply_defined
15035 @opindex multiply_defined_unused
15036 @opindex noall_load
15037 @opindex no_dead_strip_inits_and_terms
15038 @opindex nofixprebinding
15039 @opindex nomultidefs
15041 @opindex noseglinkedit
15042 @opindex pagezero_size
15044 @opindex prebind_all_twolevel_modules
15045 @opindex private_bundle
15046 @opindex read_only_relocs
15048 @opindex sectobjectsymbols
15051 @opindex sectcreate
15052 @opindex sectobjectsymbols
15055 @opindex segs_read_only_addr
15056 @opindex segs_read_write_addr
15057 @opindex seg_addr_table
15058 @opindex seg_addr_table_filename
15059 @opindex seglinkedit
15061 @opindex segs_read_only_addr
15062 @opindex segs_read_write_addr
15063 @opindex single_module
15065 @opindex sub_library
15066 @opindex sub_umbrella
15067 @opindex twolevel_namespace
15070 @opindex unexported_symbols_list
15071 @opindex weak_reference_mismatches
15072 @opindex whatsloaded
15073 These options are passed to the Darwin linker. The Darwin linker man page
15074 describes them in detail.
15077 @node DEC Alpha Options
15078 @subsection DEC Alpha Options
15080 These @samp{-m} options are defined for the DEC Alpha implementations:
15083 @item -mno-soft-float
15084 @itemx -msoft-float
15085 @opindex mno-soft-float
15086 @opindex msoft-float
15087 Use (do not use) the hardware floating-point instructions for
15088 floating-point operations. When @option{-msoft-float} is specified,
15089 functions in @file{libgcc.a} are used to perform floating-point
15090 operations. Unless they are replaced by routines that emulate the
15091 floating-point operations, or compiled in such a way as to call such
15092 emulations routines, these routines issue floating-point
15093 operations. If you are compiling for an Alpha without floating-point
15094 operations, you must ensure that the library is built so as not to call
15097 Note that Alpha implementations without floating-point operations are
15098 required to have floating-point registers.
15101 @itemx -mno-fp-regs
15103 @opindex mno-fp-regs
15104 Generate code that uses (does not use) the floating-point register set.
15105 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
15106 register set is not used, floating-point operands are passed in integer
15107 registers as if they were integers and floating-point results are passed
15108 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
15109 so any function with a floating-point argument or return value called by code
15110 compiled with @option{-mno-fp-regs} must also be compiled with that
15113 A typical use of this option is building a kernel that does not use,
15114 and hence need not save and restore, any floating-point registers.
15118 The Alpha architecture implements floating-point hardware optimized for
15119 maximum performance. It is mostly compliant with the IEEE floating-point
15120 standard. However, for full compliance, software assistance is
15121 required. This option generates code fully IEEE-compliant code
15122 @emph{except} that the @var{inexact-flag} is not maintained (see below).
15123 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
15124 defined during compilation. The resulting code is less efficient but is
15125 able to correctly support denormalized numbers and exceptional IEEE
15126 values such as not-a-number and plus/minus infinity. Other Alpha
15127 compilers call this option @option{-ieee_with_no_inexact}.
15129 @item -mieee-with-inexact
15130 @opindex mieee-with-inexact
15131 This is like @option{-mieee} except the generated code also maintains
15132 the IEEE @var{inexact-flag}. Turning on this option causes the
15133 generated code to implement fully-compliant IEEE math. In addition to
15134 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
15135 macro. On some Alpha implementations the resulting code may execute
15136 significantly slower than the code generated by default. Since there is
15137 very little code that depends on the @var{inexact-flag}, you should
15138 normally not specify this option. Other Alpha compilers call this
15139 option @option{-ieee_with_inexact}.
15141 @item -mfp-trap-mode=@var{trap-mode}
15142 @opindex mfp-trap-mode
15143 This option controls what floating-point related traps are enabled.
15144 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
15145 The trap mode can be set to one of four values:
15149 This is the default (normal) setting. The only traps that are enabled
15150 are the ones that cannot be disabled in software (e.g., division by zero
15154 In addition to the traps enabled by @samp{n}, underflow traps are enabled
15158 Like @samp{u}, but the instructions are marked to be safe for software
15159 completion (see Alpha architecture manual for details).
15162 Like @samp{su}, but inexact traps are enabled as well.
15165 @item -mfp-rounding-mode=@var{rounding-mode}
15166 @opindex mfp-rounding-mode
15167 Selects the IEEE rounding mode. Other Alpha compilers call this option
15168 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
15173 Normal IEEE rounding mode. Floating-point numbers are rounded towards
15174 the nearest machine number or towards the even machine number in case
15178 Round towards minus infinity.
15181 Chopped rounding mode. Floating-point numbers are rounded towards zero.
15184 Dynamic rounding mode. A field in the floating-point control register
15185 (@var{fpcr}, see Alpha architecture reference manual) controls the
15186 rounding mode in effect. The C library initializes this register for
15187 rounding towards plus infinity. Thus, unless your program modifies the
15188 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15191 @item -mtrap-precision=@var{trap-precision}
15192 @opindex mtrap-precision
15193 In the Alpha architecture, floating-point traps are imprecise. This
15194 means without software assistance it is impossible to recover from a
15195 floating trap and program execution normally needs to be terminated.
15196 GCC can generate code that can assist operating system trap handlers
15197 in determining the exact location that caused a floating-point trap.
15198 Depending on the requirements of an application, different levels of
15199 precisions can be selected:
15203 Program precision. This option is the default and means a trap handler
15204 can only identify which program caused a floating-point exception.
15207 Function precision. The trap handler can determine the function that
15208 caused a floating-point exception.
15211 Instruction precision. The trap handler can determine the exact
15212 instruction that caused a floating-point exception.
15215 Other Alpha compilers provide the equivalent options called
15216 @option{-scope_safe} and @option{-resumption_safe}.
15218 @item -mieee-conformant
15219 @opindex mieee-conformant
15220 This option marks the generated code as IEEE conformant. You must not
15221 use this option unless you also specify @option{-mtrap-precision=i} and either
15222 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15223 is to emit the line @samp{.eflag 48} in the function prologue of the
15224 generated assembly file.
15226 @item -mbuild-constants
15227 @opindex mbuild-constants
15228 Normally GCC examines a 32- or 64-bit integer constant to
15229 see if it can construct it from smaller constants in two or three
15230 instructions. If it cannot, it outputs the constant as a literal and
15231 generates code to load it from the data segment at run time.
15233 Use this option to require GCC to construct @emph{all} integer constants
15234 using code, even if it takes more instructions (the maximum is six).
15236 You typically use this option to build a shared library dynamic
15237 loader. Itself a shared library, it must relocate itself in memory
15238 before it can find the variables and constants in its own data segment.
15256 Indicate whether GCC should generate code to use the optional BWX,
15257 CIX, FIX and MAX instruction sets. The default is to use the instruction
15258 sets supported by the CPU type specified via @option{-mcpu=} option or that
15259 of the CPU on which GCC was built if none is specified.
15262 @itemx -mfloat-ieee
15263 @opindex mfloat-vax
15264 @opindex mfloat-ieee
15265 Generate code that uses (does not use) VAX F and G floating-point
15266 arithmetic instead of IEEE single and double precision.
15268 @item -mexplicit-relocs
15269 @itemx -mno-explicit-relocs
15270 @opindex mexplicit-relocs
15271 @opindex mno-explicit-relocs
15272 Older Alpha assemblers provided no way to generate symbol relocations
15273 except via assembler macros. Use of these macros does not allow
15274 optimal instruction scheduling. GNU binutils as of version 2.12
15275 supports a new syntax that allows the compiler to explicitly mark
15276 which relocations should apply to which instructions. This option
15277 is mostly useful for debugging, as GCC detects the capabilities of
15278 the assembler when it is built and sets the default accordingly.
15281 @itemx -mlarge-data
15282 @opindex msmall-data
15283 @opindex mlarge-data
15284 When @option{-mexplicit-relocs} is in effect, static data is
15285 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15286 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15287 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15288 16-bit relocations off of the @code{$gp} register. This limits the
15289 size of the small data area to 64KB, but allows the variables to be
15290 directly accessed via a single instruction.
15292 The default is @option{-mlarge-data}. With this option the data area
15293 is limited to just below 2GB@. Programs that require more than 2GB of
15294 data must use @code{malloc} or @code{mmap} to allocate the data in the
15295 heap instead of in the program's data segment.
15297 When generating code for shared libraries, @option{-fpic} implies
15298 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15301 @itemx -mlarge-text
15302 @opindex msmall-text
15303 @opindex mlarge-text
15304 When @option{-msmall-text} is used, the compiler assumes that the
15305 code of the entire program (or shared library) fits in 4MB, and is
15306 thus reachable with a branch instruction. When @option{-msmall-data}
15307 is used, the compiler can assume that all local symbols share the
15308 same @code{$gp} value, and thus reduce the number of instructions
15309 required for a function call from 4 to 1.
15311 The default is @option{-mlarge-text}.
15313 @item -mcpu=@var{cpu_type}
15315 Set the instruction set and instruction scheduling parameters for
15316 machine type @var{cpu_type}. You can specify either the @samp{EV}
15317 style name or the corresponding chip number. GCC supports scheduling
15318 parameters for the EV4, EV5 and EV6 family of processors and
15319 chooses the default values for the instruction set from the processor
15320 you specify. If you do not specify a processor type, GCC defaults
15321 to the processor on which the compiler was built.
15323 Supported values for @var{cpu_type} are
15329 Schedules as an EV4 and has no instruction set extensions.
15333 Schedules as an EV5 and has no instruction set extensions.
15337 Schedules as an EV5 and supports the BWX extension.
15342 Schedules as an EV5 and supports the BWX and MAX extensions.
15346 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15350 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15353 Native toolchains also support the value @samp{native},
15354 which selects the best architecture option for the host processor.
15355 @option{-mcpu=native} has no effect if GCC does not recognize
15358 @item -mtune=@var{cpu_type}
15360 Set only the instruction scheduling parameters for machine type
15361 @var{cpu_type}. The instruction set is not changed.
15363 Native toolchains also support the value @samp{native},
15364 which selects the best architecture option for the host processor.
15365 @option{-mtune=native} has no effect if GCC does not recognize
15368 @item -mmemory-latency=@var{time}
15369 @opindex mmemory-latency
15370 Sets the latency the scheduler should assume for typical memory
15371 references as seen by the application. This number is highly
15372 dependent on the memory access patterns used by the application
15373 and the size of the external cache on the machine.
15375 Valid options for @var{time} are
15379 A decimal number representing clock cycles.
15385 The compiler contains estimates of the number of clock cycles for
15386 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15387 (also called Dcache, Scache, and Bcache), as well as to main memory.
15388 Note that L3 is only valid for EV5.
15394 @subsection FR30 Options
15395 @cindex FR30 Options
15397 These options are defined specifically for the FR30 port.
15401 @item -msmall-model
15402 @opindex msmall-model
15403 Use the small address space model. This can produce smaller code, but
15404 it does assume that all symbolic values and addresses fit into a
15409 Assume that runtime support has been provided and so there is no need
15410 to include the simulator library (@file{libsim.a}) on the linker
15416 @subsection FT32 Options
15417 @cindex FT32 Options
15419 These options are defined specifically for the FT32 port.
15425 Specifies that the program will be run on the simulator. This causes
15426 an alternate runtime startup and library to be linked.
15427 You must not use this option when generating programs that will run on
15428 real hardware; you must provide your own runtime library for whatever
15429 I/O functions are needed.
15433 Enable Local Register Allocation. This is still experimental for FT32,
15434 so by default the compiler uses standard reload.
15439 @subsection FRV Options
15440 @cindex FRV Options
15446 Only use the first 32 general-purpose registers.
15451 Use all 64 general-purpose registers.
15456 Use only the first 32 floating-point registers.
15461 Use all 64 floating-point registers.
15464 @opindex mhard-float
15466 Use hardware instructions for floating-point operations.
15469 @opindex msoft-float
15471 Use library routines for floating-point operations.
15476 Dynamically allocate condition code registers.
15481 Do not try to dynamically allocate condition code registers, only
15482 use @code{icc0} and @code{fcc0}.
15487 Change ABI to use double word insns.
15492 Do not use double word instructions.
15497 Use floating-point double instructions.
15500 @opindex mno-double
15502 Do not use floating-point double instructions.
15507 Use media instructions.
15512 Do not use media instructions.
15517 Use multiply and add/subtract instructions.
15520 @opindex mno-muladd
15522 Do not use multiply and add/subtract instructions.
15527 Select the FDPIC ABI, which uses function descriptors to represent
15528 pointers to functions. Without any PIC/PIE-related options, it
15529 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15530 assumes GOT entries and small data are within a 12-bit range from the
15531 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15532 are computed with 32 bits.
15533 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15536 @opindex minline-plt
15538 Enable inlining of PLT entries in function calls to functions that are
15539 not known to bind locally. It has no effect without @option{-mfdpic}.
15540 It's enabled by default if optimizing for speed and compiling for
15541 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15542 optimization option such as @option{-O3} or above is present in the
15548 Assume a large TLS segment when generating thread-local code.
15553 Do not assume a large TLS segment when generating thread-local code.
15558 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15559 that is known to be in read-only sections. It's enabled by default,
15560 except for @option{-fpic} or @option{-fpie}: even though it may help
15561 make the global offset table smaller, it trades 1 instruction for 4.
15562 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15563 one of which may be shared by multiple symbols, and it avoids the need
15564 for a GOT entry for the referenced symbol, so it's more likely to be a
15565 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15567 @item -multilib-library-pic
15568 @opindex multilib-library-pic
15570 Link with the (library, not FD) pic libraries. It's implied by
15571 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15572 @option{-fpic} without @option{-mfdpic}. You should never have to use
15576 @opindex mlinked-fp
15578 Follow the EABI requirement of always creating a frame pointer whenever
15579 a stack frame is allocated. This option is enabled by default and can
15580 be disabled with @option{-mno-linked-fp}.
15583 @opindex mlong-calls
15585 Use indirect addressing to call functions outside the current
15586 compilation unit. This allows the functions to be placed anywhere
15587 within the 32-bit address space.
15589 @item -malign-labels
15590 @opindex malign-labels
15592 Try to align labels to an 8-byte boundary by inserting NOPs into the
15593 previous packet. This option only has an effect when VLIW packing
15594 is enabled. It doesn't create new packets; it merely adds NOPs to
15597 @item -mlibrary-pic
15598 @opindex mlibrary-pic
15600 Generate position-independent EABI code.
15605 Use only the first four media accumulator registers.
15610 Use all eight media accumulator registers.
15615 Pack VLIW instructions.
15620 Do not pack VLIW instructions.
15623 @opindex mno-eflags
15625 Do not mark ABI switches in e_flags.
15628 @opindex mcond-move
15630 Enable the use of conditional-move instructions (default).
15632 This switch is mainly for debugging the compiler and will likely be removed
15633 in a future version.
15635 @item -mno-cond-move
15636 @opindex mno-cond-move
15638 Disable the use of conditional-move instructions.
15640 This switch is mainly for debugging the compiler and will likely be removed
15641 in a future version.
15646 Enable the use of conditional set instructions (default).
15648 This switch is mainly for debugging the compiler and will likely be removed
15649 in a future version.
15654 Disable the use of conditional set instructions.
15656 This switch is mainly for debugging the compiler and will likely be removed
15657 in a future version.
15660 @opindex mcond-exec
15662 Enable the use of conditional execution (default).
15664 This switch is mainly for debugging the compiler and will likely be removed
15665 in a future version.
15667 @item -mno-cond-exec
15668 @opindex mno-cond-exec
15670 Disable the use of conditional execution.
15672 This switch is mainly for debugging the compiler and will likely be removed
15673 in a future version.
15675 @item -mvliw-branch
15676 @opindex mvliw-branch
15678 Run a pass to pack branches into VLIW instructions (default).
15680 This switch is mainly for debugging the compiler and will likely be removed
15681 in a future version.
15683 @item -mno-vliw-branch
15684 @opindex mno-vliw-branch
15686 Do not run a pass to pack branches into VLIW instructions.
15688 This switch is mainly for debugging the compiler and will likely be removed
15689 in a future version.
15691 @item -mmulti-cond-exec
15692 @opindex mmulti-cond-exec
15694 Enable optimization of @code{&&} and @code{||} in conditional execution
15697 This switch is mainly for debugging the compiler and will likely be removed
15698 in a future version.
15700 @item -mno-multi-cond-exec
15701 @opindex mno-multi-cond-exec
15703 Disable optimization of @code{&&} and @code{||} in conditional execution.
15705 This switch is mainly for debugging the compiler and will likely be removed
15706 in a future version.
15708 @item -mnested-cond-exec
15709 @opindex mnested-cond-exec
15711 Enable nested conditional execution optimizations (default).
15713 This switch is mainly for debugging the compiler and will likely be removed
15714 in a future version.
15716 @item -mno-nested-cond-exec
15717 @opindex mno-nested-cond-exec
15719 Disable nested conditional execution optimizations.
15721 This switch is mainly for debugging the compiler and will likely be removed
15722 in a future version.
15724 @item -moptimize-membar
15725 @opindex moptimize-membar
15727 This switch removes redundant @code{membar} instructions from the
15728 compiler-generated code. It is enabled by default.
15730 @item -mno-optimize-membar
15731 @opindex mno-optimize-membar
15733 This switch disables the automatic removal of redundant @code{membar}
15734 instructions from the generated code.
15736 @item -mtomcat-stats
15737 @opindex mtomcat-stats
15739 Cause gas to print out tomcat statistics.
15741 @item -mcpu=@var{cpu}
15744 Select the processor type for which to generate code. Possible values are
15745 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15746 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15750 @node GNU/Linux Options
15751 @subsection GNU/Linux Options
15753 These @samp{-m} options are defined for GNU/Linux targets:
15758 Use the GNU C library. This is the default except
15759 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
15760 @samp{*-*-linux-*android*} targets.
15764 Use uClibc C library. This is the default on
15765 @samp{*-*-linux-*uclibc*} targets.
15769 Use the musl C library. This is the default on
15770 @samp{*-*-linux-*musl*} targets.
15774 Use Bionic C library. This is the default on
15775 @samp{*-*-linux-*android*} targets.
15779 Compile code compatible with Android platform. This is the default on
15780 @samp{*-*-linux-*android*} targets.
15782 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15783 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15784 this option makes the GCC driver pass Android-specific options to the linker.
15785 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15788 @item -tno-android-cc
15789 @opindex tno-android-cc
15790 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15791 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15792 @option{-fno-rtti} by default.
15794 @item -tno-android-ld
15795 @opindex tno-android-ld
15796 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15797 linking options to the linker.
15801 @node H8/300 Options
15802 @subsection H8/300 Options
15804 These @samp{-m} options are defined for the H8/300 implementations:
15809 Shorten some address references at link time, when possible; uses the
15810 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15811 ld, Using ld}, for a fuller description.
15815 Generate code for the H8/300H@.
15819 Generate code for the H8S@.
15823 Generate code for the H8S and H8/300H in the normal mode. This switch
15824 must be used either with @option{-mh} or @option{-ms}.
15828 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15832 Extended registers are stored on stack before execution of function
15833 with monitor attribute. Default option is @option{-mexr}.
15834 This option is valid only for H8S targets.
15838 Extended registers are not stored on stack before execution of function
15839 with monitor attribute. Default option is @option{-mno-exr}.
15840 This option is valid only for H8S targets.
15844 Make @code{int} data 32 bits by default.
15847 @opindex malign-300
15848 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15849 The default for the H8/300H and H8S is to align longs and floats on
15851 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15852 This option has no effect on the H8/300.
15856 @subsection HPPA Options
15857 @cindex HPPA Options
15859 These @samp{-m} options are defined for the HPPA family of computers:
15862 @item -march=@var{architecture-type}
15864 Generate code for the specified architecture. The choices for
15865 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15866 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15867 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15868 architecture option for your machine. Code compiled for lower numbered
15869 architectures runs on higher numbered architectures, but not the
15872 @item -mpa-risc-1-0
15873 @itemx -mpa-risc-1-1
15874 @itemx -mpa-risc-2-0
15875 @opindex mpa-risc-1-0
15876 @opindex mpa-risc-1-1
15877 @opindex mpa-risc-2-0
15878 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15880 @item -mjump-in-delay
15881 @opindex mjump-in-delay
15882 This option is ignored and provided for compatibility purposes only.
15884 @item -mdisable-fpregs
15885 @opindex mdisable-fpregs
15886 Prevent floating-point registers from being used in any manner. This is
15887 necessary for compiling kernels that perform lazy context switching of
15888 floating-point registers. If you use this option and attempt to perform
15889 floating-point operations, the compiler aborts.
15891 @item -mdisable-indexing
15892 @opindex mdisable-indexing
15893 Prevent the compiler from using indexing address modes. This avoids some
15894 rather obscure problems when compiling MIG generated code under MACH@.
15896 @item -mno-space-regs
15897 @opindex mno-space-regs
15898 Generate code that assumes the target has no space registers. This allows
15899 GCC to generate faster indirect calls and use unscaled index address modes.
15901 Such code is suitable for level 0 PA systems and kernels.
15903 @item -mfast-indirect-calls
15904 @opindex mfast-indirect-calls
15905 Generate code that assumes calls never cross space boundaries. This
15906 allows GCC to emit code that performs faster indirect calls.
15908 This option does not work in the presence of shared libraries or nested
15911 @item -mfixed-range=@var{register-range}
15912 @opindex mfixed-range
15913 Generate code treating the given register range as fixed registers.
15914 A fixed register is one that the register allocator cannot use. This is
15915 useful when compiling kernel code. A register range is specified as
15916 two registers separated by a dash. Multiple register ranges can be
15917 specified separated by a comma.
15919 @item -mlong-load-store
15920 @opindex mlong-load-store
15921 Generate 3-instruction load and store sequences as sometimes required by
15922 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15925 @item -mportable-runtime
15926 @opindex mportable-runtime
15927 Use the portable calling conventions proposed by HP for ELF systems.
15931 Enable the use of assembler directives only GAS understands.
15933 @item -mschedule=@var{cpu-type}
15935 Schedule code according to the constraints for the machine type
15936 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15937 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15938 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15939 proper scheduling option for your machine. The default scheduling is
15943 @opindex mlinker-opt
15944 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15945 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15946 linkers in which they give bogus error messages when linking some programs.
15949 @opindex msoft-float
15950 Generate output containing library calls for floating point.
15951 @strong{Warning:} the requisite libraries are not available for all HPPA
15952 targets. Normally the facilities of the machine's usual C compiler are
15953 used, but this cannot be done directly in cross-compilation. You must make
15954 your own arrangements to provide suitable library functions for
15957 @option{-msoft-float} changes the calling convention in the output file;
15958 therefore, it is only useful if you compile @emph{all} of a program with
15959 this option. In particular, you need to compile @file{libgcc.a}, the
15960 library that comes with GCC, with @option{-msoft-float} in order for
15965 Generate the predefine, @code{_SIO}, for server IO@. The default is
15966 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15967 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15968 options are available under HP-UX and HI-UX@.
15972 Use options specific to GNU @command{ld}.
15973 This passes @option{-shared} to @command{ld} when
15974 building a shared library. It is the default when GCC is configured,
15975 explicitly or implicitly, with the GNU linker. This option does not
15976 affect which @command{ld} is called; it only changes what parameters
15977 are passed to that @command{ld}.
15978 The @command{ld} that is called is determined by the
15979 @option{--with-ld} configure option, GCC's program search path, and
15980 finally by the user's @env{PATH}. The linker used by GCC can be printed
15981 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15982 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15986 Use options specific to HP @command{ld}.
15987 This passes @option{-b} to @command{ld} when building
15988 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15989 links. It is the default when GCC is configured, explicitly or
15990 implicitly, with the HP linker. This option does not affect
15991 which @command{ld} is called; it only changes what parameters are passed to that
15993 The @command{ld} that is called is determined by the @option{--with-ld}
15994 configure option, GCC's program search path, and finally by the user's
15995 @env{PATH}. The linker used by GCC can be printed using @samp{which
15996 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15997 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16000 @opindex mno-long-calls
16001 Generate code that uses long call sequences. This ensures that a call
16002 is always able to reach linker generated stubs. The default is to generate
16003 long calls only when the distance from the call site to the beginning
16004 of the function or translation unit, as the case may be, exceeds a
16005 predefined limit set by the branch type being used. The limits for
16006 normal calls are 7,600,000 and 240,000 bytes, respectively for the
16007 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
16010 Distances are measured from the beginning of functions when using the
16011 @option{-ffunction-sections} option, or when using the @option{-mgas}
16012 and @option{-mno-portable-runtime} options together under HP-UX with
16015 It is normally not desirable to use this option as it degrades
16016 performance. However, it may be useful in large applications,
16017 particularly when partial linking is used to build the application.
16019 The types of long calls used depends on the capabilities of the
16020 assembler and linker, and the type of code being generated. The
16021 impact on systems that support long absolute calls, and long pic
16022 symbol-difference or pc-relative calls should be relatively small.
16023 However, an indirect call is used on 32-bit ELF systems in pic code
16024 and it is quite long.
16026 @item -munix=@var{unix-std}
16028 Generate compiler predefines and select a startfile for the specified
16029 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
16030 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
16031 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
16032 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
16033 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
16036 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
16037 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
16038 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
16039 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
16040 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
16041 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
16043 It is @emph{important} to note that this option changes the interfaces
16044 for various library routines. It also affects the operational behavior
16045 of the C library. Thus, @emph{extreme} care is needed in using this
16048 Library code that is intended to operate with more than one UNIX
16049 standard must test, set and restore the variable @code{__xpg4_extended_mask}
16050 as appropriate. Most GNU software doesn't provide this capability.
16054 Suppress the generation of link options to search libdld.sl when the
16055 @option{-static} option is specified on HP-UX 10 and later.
16059 The HP-UX implementation of setlocale in libc has a dependency on
16060 libdld.sl. There isn't an archive version of libdld.sl. Thus,
16061 when the @option{-static} option is specified, special link options
16062 are needed to resolve this dependency.
16064 On HP-UX 10 and later, the GCC driver adds the necessary options to
16065 link with libdld.sl when the @option{-static} option is specified.
16066 This causes the resulting binary to be dynamic. On the 64-bit port,
16067 the linkers generate dynamic binaries by default in any case. The
16068 @option{-nolibdld} option can be used to prevent the GCC driver from
16069 adding these link options.
16073 Add support for multithreading with the @dfn{dce thread} library
16074 under HP-UX@. This option sets flags for both the preprocessor and
16078 @node IA-64 Options
16079 @subsection IA-64 Options
16080 @cindex IA-64 Options
16082 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16086 @opindex mbig-endian
16087 Generate code for a big-endian target. This is the default for HP-UX@.
16089 @item -mlittle-endian
16090 @opindex mlittle-endian
16091 Generate code for a little-endian target. This is the default for AIX5
16097 @opindex mno-gnu-as
16098 Generate (or don't) code for the GNU assembler. This is the default.
16099 @c Also, this is the default if the configure option @option{--with-gnu-as}
16105 @opindex mno-gnu-ld
16106 Generate (or don't) code for the GNU linker. This is the default.
16107 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16112 Generate code that does not use a global pointer register. The result
16113 is not position independent code, and violates the IA-64 ABI@.
16115 @item -mvolatile-asm-stop
16116 @itemx -mno-volatile-asm-stop
16117 @opindex mvolatile-asm-stop
16118 @opindex mno-volatile-asm-stop
16119 Generate (or don't) a stop bit immediately before and after volatile asm
16122 @item -mregister-names
16123 @itemx -mno-register-names
16124 @opindex mregister-names
16125 @opindex mno-register-names
16126 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16127 the stacked registers. This may make assembler output more readable.
16133 Disable (or enable) optimizations that use the small data section. This may
16134 be useful for working around optimizer bugs.
16136 @item -mconstant-gp
16137 @opindex mconstant-gp
16138 Generate code that uses a single constant global pointer value. This is
16139 useful when compiling kernel code.
16143 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16144 This is useful when compiling firmware code.
16146 @item -minline-float-divide-min-latency
16147 @opindex minline-float-divide-min-latency
16148 Generate code for inline divides of floating-point values
16149 using the minimum latency algorithm.
16151 @item -minline-float-divide-max-throughput
16152 @opindex minline-float-divide-max-throughput
16153 Generate code for inline divides of floating-point values
16154 using the maximum throughput algorithm.
16156 @item -mno-inline-float-divide
16157 @opindex mno-inline-float-divide
16158 Do not generate inline code for divides of floating-point values.
16160 @item -minline-int-divide-min-latency
16161 @opindex minline-int-divide-min-latency
16162 Generate code for inline divides of integer values
16163 using the minimum latency algorithm.
16165 @item -minline-int-divide-max-throughput
16166 @opindex minline-int-divide-max-throughput
16167 Generate code for inline divides of integer values
16168 using the maximum throughput algorithm.
16170 @item -mno-inline-int-divide
16171 @opindex mno-inline-int-divide
16172 Do not generate inline code for divides of integer values.
16174 @item -minline-sqrt-min-latency
16175 @opindex minline-sqrt-min-latency
16176 Generate code for inline square roots
16177 using the minimum latency algorithm.
16179 @item -minline-sqrt-max-throughput
16180 @opindex minline-sqrt-max-throughput
16181 Generate code for inline square roots
16182 using the maximum throughput algorithm.
16184 @item -mno-inline-sqrt
16185 @opindex mno-inline-sqrt
16186 Do not generate inline code for @code{sqrt}.
16189 @itemx -mno-fused-madd
16190 @opindex mfused-madd
16191 @opindex mno-fused-madd
16192 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16193 instructions. The default is to use these instructions.
16195 @item -mno-dwarf2-asm
16196 @itemx -mdwarf2-asm
16197 @opindex mno-dwarf2-asm
16198 @opindex mdwarf2-asm
16199 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16200 info. This may be useful when not using the GNU assembler.
16202 @item -mearly-stop-bits
16203 @itemx -mno-early-stop-bits
16204 @opindex mearly-stop-bits
16205 @opindex mno-early-stop-bits
16206 Allow stop bits to be placed earlier than immediately preceding the
16207 instruction that triggered the stop bit. This can improve instruction
16208 scheduling, but does not always do so.
16210 @item -mfixed-range=@var{register-range}
16211 @opindex mfixed-range
16212 Generate code treating the given register range as fixed registers.
16213 A fixed register is one that the register allocator cannot use. This is
16214 useful when compiling kernel code. A register range is specified as
16215 two registers separated by a dash. Multiple register ranges can be
16216 specified separated by a comma.
16218 @item -mtls-size=@var{tls-size}
16220 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16223 @item -mtune=@var{cpu-type}
16225 Tune the instruction scheduling for a particular CPU, Valid values are
16226 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16227 and @samp{mckinley}.
16233 Generate code for a 32-bit or 64-bit environment.
16234 The 32-bit environment sets int, long and pointer to 32 bits.
16235 The 64-bit environment sets int to 32 bits and long and pointer
16236 to 64 bits. These are HP-UX specific flags.
16238 @item -mno-sched-br-data-spec
16239 @itemx -msched-br-data-spec
16240 @opindex mno-sched-br-data-spec
16241 @opindex msched-br-data-spec
16242 (Dis/En)able data speculative scheduling before reload.
16243 This results in generation of @code{ld.a} instructions and
16244 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16245 The default is 'disable'.
16247 @item -msched-ar-data-spec
16248 @itemx -mno-sched-ar-data-spec
16249 @opindex msched-ar-data-spec
16250 @opindex mno-sched-ar-data-spec
16251 (En/Dis)able data speculative scheduling after reload.
16252 This results in generation of @code{ld.a} instructions and
16253 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16254 The default is 'enable'.
16256 @item -mno-sched-control-spec
16257 @itemx -msched-control-spec
16258 @opindex mno-sched-control-spec
16259 @opindex msched-control-spec
16260 (Dis/En)able control speculative scheduling. This feature is
16261 available only during region scheduling (i.e.@: before reload).
16262 This results in generation of the @code{ld.s} instructions and
16263 the corresponding check instructions @code{chk.s}.
16264 The default is 'disable'.
16266 @item -msched-br-in-data-spec
16267 @itemx -mno-sched-br-in-data-spec
16268 @opindex msched-br-in-data-spec
16269 @opindex mno-sched-br-in-data-spec
16270 (En/Dis)able speculative scheduling of the instructions that
16271 are dependent on the data speculative loads before reload.
16272 This is effective only with @option{-msched-br-data-spec} enabled.
16273 The default is 'enable'.
16275 @item -msched-ar-in-data-spec
16276 @itemx -mno-sched-ar-in-data-spec
16277 @opindex msched-ar-in-data-spec
16278 @opindex mno-sched-ar-in-data-spec
16279 (En/Dis)able speculative scheduling of the instructions that
16280 are dependent on the data speculative loads after reload.
16281 This is effective only with @option{-msched-ar-data-spec} enabled.
16282 The default is 'enable'.
16284 @item -msched-in-control-spec
16285 @itemx -mno-sched-in-control-spec
16286 @opindex msched-in-control-spec
16287 @opindex mno-sched-in-control-spec
16288 (En/Dis)able speculative scheduling of the instructions that
16289 are dependent on the control speculative loads.
16290 This is effective only with @option{-msched-control-spec} enabled.
16291 The default is 'enable'.
16293 @item -mno-sched-prefer-non-data-spec-insns
16294 @itemx -msched-prefer-non-data-spec-insns
16295 @opindex mno-sched-prefer-non-data-spec-insns
16296 @opindex msched-prefer-non-data-spec-insns
16297 If enabled, data-speculative instructions are chosen for schedule
16298 only if there are no other choices at the moment. This makes
16299 the use of the data speculation much more conservative.
16300 The default is 'disable'.
16302 @item -mno-sched-prefer-non-control-spec-insns
16303 @itemx -msched-prefer-non-control-spec-insns
16304 @opindex mno-sched-prefer-non-control-spec-insns
16305 @opindex msched-prefer-non-control-spec-insns
16306 If enabled, control-speculative instructions are chosen for schedule
16307 only if there are no other choices at the moment. This makes
16308 the use of the control speculation much more conservative.
16309 The default is 'disable'.
16311 @item -mno-sched-count-spec-in-critical-path
16312 @itemx -msched-count-spec-in-critical-path
16313 @opindex mno-sched-count-spec-in-critical-path
16314 @opindex msched-count-spec-in-critical-path
16315 If enabled, speculative dependencies are considered during
16316 computation of the instructions priorities. This makes the use of the
16317 speculation a bit more conservative.
16318 The default is 'disable'.
16320 @item -msched-spec-ldc
16321 @opindex msched-spec-ldc
16322 Use a simple data speculation check. This option is on by default.
16324 @item -msched-control-spec-ldc
16325 @opindex msched-spec-ldc
16326 Use a simple check for control speculation. This option is on by default.
16328 @item -msched-stop-bits-after-every-cycle
16329 @opindex msched-stop-bits-after-every-cycle
16330 Place a stop bit after every cycle when scheduling. This option is on
16333 @item -msched-fp-mem-deps-zero-cost
16334 @opindex msched-fp-mem-deps-zero-cost
16335 Assume that floating-point stores and loads are not likely to cause a conflict
16336 when placed into the same instruction group. This option is disabled by
16339 @item -msel-sched-dont-check-control-spec
16340 @opindex msel-sched-dont-check-control-spec
16341 Generate checks for control speculation in selective scheduling.
16342 This flag is disabled by default.
16344 @item -msched-max-memory-insns=@var{max-insns}
16345 @opindex msched-max-memory-insns
16346 Limit on the number of memory insns per instruction group, giving lower
16347 priority to subsequent memory insns attempting to schedule in the same
16348 instruction group. Frequently useful to prevent cache bank conflicts.
16349 The default value is 1.
16351 @item -msched-max-memory-insns-hard-limit
16352 @opindex msched-max-memory-insns-hard-limit
16353 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16354 disallowing more than that number in an instruction group.
16355 Otherwise, the limit is ``soft'', meaning that non-memory operations
16356 are preferred when the limit is reached, but memory operations may still
16362 @subsection LM32 Options
16363 @cindex LM32 options
16365 These @option{-m} options are defined for the LatticeMico32 architecture:
16368 @item -mbarrel-shift-enabled
16369 @opindex mbarrel-shift-enabled
16370 Enable barrel-shift instructions.
16372 @item -mdivide-enabled
16373 @opindex mdivide-enabled
16374 Enable divide and modulus instructions.
16376 @item -mmultiply-enabled
16377 @opindex multiply-enabled
16378 Enable multiply instructions.
16380 @item -msign-extend-enabled
16381 @opindex msign-extend-enabled
16382 Enable sign extend instructions.
16384 @item -muser-enabled
16385 @opindex muser-enabled
16386 Enable user-defined instructions.
16391 @subsection M32C Options
16392 @cindex M32C options
16395 @item -mcpu=@var{name}
16397 Select the CPU for which code is generated. @var{name} may be one of
16398 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16399 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16400 the M32C/80 series.
16404 Specifies that the program will be run on the simulator. This causes
16405 an alternate runtime library to be linked in which supports, for
16406 example, file I/O@. You must not use this option when generating
16407 programs that will run on real hardware; you must provide your own
16408 runtime library for whatever I/O functions are needed.
16410 @item -memregs=@var{number}
16412 Specifies the number of memory-based pseudo-registers GCC uses
16413 during code generation. These pseudo-registers are used like real
16414 registers, so there is a tradeoff between GCC's ability to fit the
16415 code into available registers, and the performance penalty of using
16416 memory instead of registers. Note that all modules in a program must
16417 be compiled with the same value for this option. Because of that, you
16418 must not use this option with GCC's default runtime libraries.
16422 @node M32R/D Options
16423 @subsection M32R/D Options
16424 @cindex M32R/D options
16426 These @option{-m} options are defined for Renesas M32R/D architectures:
16431 Generate code for the M32R/2@.
16435 Generate code for the M32R/X@.
16439 Generate code for the M32R@. This is the default.
16441 @item -mmodel=small
16442 @opindex mmodel=small
16443 Assume all objects live in the lower 16MB of memory (so that their addresses
16444 can be loaded with the @code{ld24} instruction), and assume all subroutines
16445 are reachable with the @code{bl} instruction.
16446 This is the default.
16448 The addressability of a particular object can be set with the
16449 @code{model} attribute.
16451 @item -mmodel=medium
16452 @opindex mmodel=medium
16453 Assume objects may be anywhere in the 32-bit address space (the compiler
16454 generates @code{seth/add3} instructions to load their addresses), and
16455 assume all subroutines are reachable with the @code{bl} instruction.
16457 @item -mmodel=large
16458 @opindex mmodel=large
16459 Assume objects may be anywhere in the 32-bit address space (the compiler
16460 generates @code{seth/add3} instructions to load their addresses), and
16461 assume subroutines may not be reachable with the @code{bl} instruction
16462 (the compiler generates the much slower @code{seth/add3/jl}
16463 instruction sequence).
16466 @opindex msdata=none
16467 Disable use of the small data area. Variables are put into
16468 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16469 @code{section} attribute has been specified).
16470 This is the default.
16472 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16473 Objects may be explicitly put in the small data area with the
16474 @code{section} attribute using one of these sections.
16476 @item -msdata=sdata
16477 @opindex msdata=sdata
16478 Put small global and static data in the small data area, but do not
16479 generate special code to reference them.
16482 @opindex msdata=use
16483 Put small global and static data in the small data area, and generate
16484 special instructions to reference them.
16488 @cindex smaller data references
16489 Put global and static objects less than or equal to @var{num} bytes
16490 into the small data or BSS sections instead of the normal data or BSS
16491 sections. The default value of @var{num} is 8.
16492 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16493 for this option to have any effect.
16495 All modules should be compiled with the same @option{-G @var{num}} value.
16496 Compiling with different values of @var{num} may or may not work; if it
16497 doesn't the linker gives an error message---incorrect code is not
16502 Makes the M32R-specific code in the compiler display some statistics
16503 that might help in debugging programs.
16505 @item -malign-loops
16506 @opindex malign-loops
16507 Align all loops to a 32-byte boundary.
16509 @item -mno-align-loops
16510 @opindex mno-align-loops
16511 Do not enforce a 32-byte alignment for loops. This is the default.
16513 @item -missue-rate=@var{number}
16514 @opindex missue-rate=@var{number}
16515 Issue @var{number} instructions per cycle. @var{number} can only be 1
16518 @item -mbranch-cost=@var{number}
16519 @opindex mbranch-cost=@var{number}
16520 @var{number} can only be 1 or 2. If it is 1 then branches are
16521 preferred over conditional code, if it is 2, then the opposite applies.
16523 @item -mflush-trap=@var{number}
16524 @opindex mflush-trap=@var{number}
16525 Specifies the trap number to use to flush the cache. The default is
16526 12. Valid numbers are between 0 and 15 inclusive.
16528 @item -mno-flush-trap
16529 @opindex mno-flush-trap
16530 Specifies that the cache cannot be flushed by using a trap.
16532 @item -mflush-func=@var{name}
16533 @opindex mflush-func=@var{name}
16534 Specifies the name of the operating system function to call to flush
16535 the cache. The default is @samp{_flush_cache}, but a function call
16536 is only used if a trap is not available.
16538 @item -mno-flush-func
16539 @opindex mno-flush-func
16540 Indicates that there is no OS function for flushing the cache.
16544 @node M680x0 Options
16545 @subsection M680x0 Options
16546 @cindex M680x0 options
16548 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16549 The default settings depend on which architecture was selected when
16550 the compiler was configured; the defaults for the most common choices
16554 @item -march=@var{arch}
16556 Generate code for a specific M680x0 or ColdFire instruction set
16557 architecture. Permissible values of @var{arch} for M680x0
16558 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16559 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16560 architectures are selected according to Freescale's ISA classification
16561 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16562 @samp{isab} and @samp{isac}.
16564 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16565 code for a ColdFire target. The @var{arch} in this macro is one of the
16566 @option{-march} arguments given above.
16568 When used together, @option{-march} and @option{-mtune} select code
16569 that runs on a family of similar processors but that is optimized
16570 for a particular microarchitecture.
16572 @item -mcpu=@var{cpu}
16574 Generate code for a specific M680x0 or ColdFire processor.
16575 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16576 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16577 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16578 below, which also classifies the CPUs into families:
16580 @multitable @columnfractions 0.20 0.80
16581 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16582 @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}
16583 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16584 @item @samp{5206e} @tab @samp{5206e}
16585 @item @samp{5208} @tab @samp{5207} @samp{5208}
16586 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16587 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16588 @item @samp{5216} @tab @samp{5214} @samp{5216}
16589 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16590 @item @samp{5225} @tab @samp{5224} @samp{5225}
16591 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16592 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16593 @item @samp{5249} @tab @samp{5249}
16594 @item @samp{5250} @tab @samp{5250}
16595 @item @samp{5271} @tab @samp{5270} @samp{5271}
16596 @item @samp{5272} @tab @samp{5272}
16597 @item @samp{5275} @tab @samp{5274} @samp{5275}
16598 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16599 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16600 @item @samp{5307} @tab @samp{5307}
16601 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16602 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16603 @item @samp{5407} @tab @samp{5407}
16604 @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}
16607 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16608 @var{arch} is compatible with @var{cpu}. Other combinations of
16609 @option{-mcpu} and @option{-march} are rejected.
16611 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16612 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16613 where the value of @var{family} is given by the table above.
16615 @item -mtune=@var{tune}
16617 Tune the code for a particular microarchitecture within the
16618 constraints set by @option{-march} and @option{-mcpu}.
16619 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16620 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16621 and @samp{cpu32}. The ColdFire microarchitectures
16622 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16624 You can also use @option{-mtune=68020-40} for code that needs
16625 to run relatively well on 68020, 68030 and 68040 targets.
16626 @option{-mtune=68020-60} is similar but includes 68060 targets
16627 as well. These two options select the same tuning decisions as
16628 @option{-m68020-40} and @option{-m68020-60} respectively.
16630 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16631 when tuning for 680x0 architecture @var{arch}. It also defines
16632 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16633 option is used. If GCC is tuning for a range of architectures,
16634 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16635 it defines the macros for every architecture in the range.
16637 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16638 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16639 of the arguments given above.
16645 Generate output for a 68000. This is the default
16646 when the compiler is configured for 68000-based systems.
16647 It is equivalent to @option{-march=68000}.
16649 Use this option for microcontrollers with a 68000 or EC000 core,
16650 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16654 Generate output for a 68010. This is the default
16655 when the compiler is configured for 68010-based systems.
16656 It is equivalent to @option{-march=68010}.
16662 Generate output for a 68020. This is the default
16663 when the compiler is configured for 68020-based systems.
16664 It is equivalent to @option{-march=68020}.
16668 Generate output for a 68030. This is the default when the compiler is
16669 configured for 68030-based systems. It is equivalent to
16670 @option{-march=68030}.
16674 Generate output for a 68040. This is the default when the compiler is
16675 configured for 68040-based systems. It is equivalent to
16676 @option{-march=68040}.
16678 This option inhibits the use of 68881/68882 instructions that have to be
16679 emulated by software on the 68040. Use this option if your 68040 does not
16680 have code to emulate those instructions.
16684 Generate output for a 68060. This is the default when the compiler is
16685 configured for 68060-based systems. It is equivalent to
16686 @option{-march=68060}.
16688 This option inhibits the use of 68020 and 68881/68882 instructions that
16689 have to be emulated by software on the 68060. Use this option if your 68060
16690 does not have code to emulate those instructions.
16694 Generate output for a CPU32. This is the default
16695 when the compiler is configured for CPU32-based systems.
16696 It is equivalent to @option{-march=cpu32}.
16698 Use this option for microcontrollers with a
16699 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16700 68336, 68340, 68341, 68349 and 68360.
16704 Generate output for a 520X ColdFire CPU@. This is the default
16705 when the compiler is configured for 520X-based systems.
16706 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16707 in favor of that option.
16709 Use this option for microcontroller with a 5200 core, including
16710 the MCF5202, MCF5203, MCF5204 and MCF5206.
16714 Generate output for a 5206e ColdFire CPU@. The option is now
16715 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16719 Generate output for a member of the ColdFire 528X family.
16720 The option is now deprecated in favor of the equivalent
16721 @option{-mcpu=528x}.
16725 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16726 in favor of the equivalent @option{-mcpu=5307}.
16730 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16731 in favor of the equivalent @option{-mcpu=5407}.
16735 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16736 This includes use of hardware floating-point instructions.
16737 The option is equivalent to @option{-mcpu=547x}, and is now
16738 deprecated in favor of that option.
16742 Generate output for a 68040, without using any of the new instructions.
16743 This results in code that can run relatively efficiently on either a
16744 68020/68881 or a 68030 or a 68040. The generated code does use the
16745 68881 instructions that are emulated on the 68040.
16747 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16751 Generate output for a 68060, without using any of the new instructions.
16752 This results in code that can run relatively efficiently on either a
16753 68020/68881 or a 68030 or a 68040. The generated code does use the
16754 68881 instructions that are emulated on the 68060.
16756 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16760 @opindex mhard-float
16762 Generate floating-point instructions. This is the default for 68020
16763 and above, and for ColdFire devices that have an FPU@. It defines the
16764 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16765 on ColdFire targets.
16768 @opindex msoft-float
16769 Do not generate floating-point instructions; use library calls instead.
16770 This is the default for 68000, 68010, and 68832 targets. It is also
16771 the default for ColdFire devices that have no FPU.
16777 Generate (do not generate) ColdFire hardware divide and remainder
16778 instructions. If @option{-march} is used without @option{-mcpu},
16779 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16780 architectures. Otherwise, the default is taken from the target CPU
16781 (either the default CPU, or the one specified by @option{-mcpu}). For
16782 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16783 @option{-mcpu=5206e}.
16785 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16789 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16790 Additionally, parameters passed on the stack are also aligned to a
16791 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16795 Do not consider type @code{int} to be 16 bits wide. This is the default.
16798 @itemx -mno-bitfield
16799 @opindex mnobitfield
16800 @opindex mno-bitfield
16801 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16802 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16806 Do use the bit-field instructions. The @option{-m68020} option implies
16807 @option{-mbitfield}. This is the default if you use a configuration
16808 designed for a 68020.
16812 Use a different function-calling convention, in which functions
16813 that take a fixed number of arguments return with the @code{rtd}
16814 instruction, which pops their arguments while returning. This
16815 saves one instruction in the caller since there is no need to pop
16816 the arguments there.
16818 This calling convention is incompatible with the one normally
16819 used on Unix, so you cannot use it if you need to call libraries
16820 compiled with the Unix compiler.
16822 Also, you must provide function prototypes for all functions that
16823 take variable numbers of arguments (including @code{printf});
16824 otherwise incorrect code is generated for calls to those
16827 In addition, seriously incorrect code results if you call a
16828 function with too many arguments. (Normally, extra arguments are
16829 harmlessly ignored.)
16831 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16832 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16836 Do not use the calling conventions selected by @option{-mrtd}.
16837 This is the default.
16840 @itemx -mno-align-int
16841 @opindex malign-int
16842 @opindex mno-align-int
16843 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16844 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16845 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16846 Aligning variables on 32-bit boundaries produces code that runs somewhat
16847 faster on processors with 32-bit busses at the expense of more memory.
16849 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16850 aligns structures containing the above types differently than
16851 most published application binary interface specifications for the m68k.
16855 Use the pc-relative addressing mode of the 68000 directly, instead of
16856 using a global offset table. At present, this option implies @option{-fpic},
16857 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16858 not presently supported with @option{-mpcrel}, though this could be supported for
16859 68020 and higher processors.
16861 @item -mno-strict-align
16862 @itemx -mstrict-align
16863 @opindex mno-strict-align
16864 @opindex mstrict-align
16865 Do not (do) assume that unaligned memory references are handled by
16869 Generate code that allows the data segment to be located in a different
16870 area of memory from the text segment. This allows for execute-in-place in
16871 an environment without virtual memory management. This option implies
16874 @item -mno-sep-data
16875 Generate code that assumes that the data segment follows the text segment.
16876 This is the default.
16878 @item -mid-shared-library
16879 Generate code that supports shared libraries via the library ID method.
16880 This allows for execute-in-place and shared libraries in an environment
16881 without virtual memory management. This option implies @option{-fPIC}.
16883 @item -mno-id-shared-library
16884 Generate code that doesn't assume ID-based shared libraries are being used.
16885 This is the default.
16887 @item -mshared-library-id=n
16888 Specifies the identification number of the ID-based shared library being
16889 compiled. Specifying a value of 0 generates more compact code; specifying
16890 other values forces the allocation of that number to the current
16891 library, but is no more space- or time-efficient than omitting this option.
16897 When generating position-independent code for ColdFire, generate code
16898 that works if the GOT has more than 8192 entries. This code is
16899 larger and slower than code generated without this option. On M680x0
16900 processors, this option is not needed; @option{-fPIC} suffices.
16902 GCC normally uses a single instruction to load values from the GOT@.
16903 While this is relatively efficient, it only works if the GOT
16904 is smaller than about 64k. Anything larger causes the linker
16905 to report an error such as:
16907 @cindex relocation truncated to fit (ColdFire)
16909 relocation truncated to fit: R_68K_GOT16O foobar
16912 If this happens, you should recompile your code with @option{-mxgot}.
16913 It should then work with very large GOTs. However, code generated with
16914 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16915 the value of a global symbol.
16917 Note that some linkers, including newer versions of the GNU linker,
16918 can create multiple GOTs and sort GOT entries. If you have such a linker,
16919 you should only need to use @option{-mxgot} when compiling a single
16920 object file that accesses more than 8192 GOT entries. Very few do.
16922 These options have no effect unless GCC is generating
16923 position-independent code.
16927 @node MCore Options
16928 @subsection MCore Options
16929 @cindex MCore options
16931 These are the @samp{-m} options defined for the Motorola M*Core
16937 @itemx -mno-hardlit
16939 @opindex mno-hardlit
16940 Inline constants into the code stream if it can be done in two
16941 instructions or less.
16947 Use the divide instruction. (Enabled by default).
16949 @item -mrelax-immediate
16950 @itemx -mno-relax-immediate
16951 @opindex mrelax-immediate
16952 @opindex mno-relax-immediate
16953 Allow arbitrary-sized immediates in bit operations.
16955 @item -mwide-bitfields
16956 @itemx -mno-wide-bitfields
16957 @opindex mwide-bitfields
16958 @opindex mno-wide-bitfields
16959 Always treat bit-fields as @code{int}-sized.
16961 @item -m4byte-functions
16962 @itemx -mno-4byte-functions
16963 @opindex m4byte-functions
16964 @opindex mno-4byte-functions
16965 Force all functions to be aligned to a 4-byte boundary.
16967 @item -mcallgraph-data
16968 @itemx -mno-callgraph-data
16969 @opindex mcallgraph-data
16970 @opindex mno-callgraph-data
16971 Emit callgraph information.
16974 @itemx -mno-slow-bytes
16975 @opindex mslow-bytes
16976 @opindex mno-slow-bytes
16977 Prefer word access when reading byte quantities.
16979 @item -mlittle-endian
16980 @itemx -mbig-endian
16981 @opindex mlittle-endian
16982 @opindex mbig-endian
16983 Generate code for a little-endian target.
16989 Generate code for the 210 processor.
16993 Assume that runtime support has been provided and so omit the
16994 simulator library (@file{libsim.a)} from the linker command line.
16996 @item -mstack-increment=@var{size}
16997 @opindex mstack-increment
16998 Set the maximum amount for a single stack increment operation. Large
16999 values can increase the speed of programs that contain functions
17000 that need a large amount of stack space, but they can also trigger a
17001 segmentation fault if the stack is extended too much. The default
17007 @subsection MeP Options
17008 @cindex MeP options
17014 Enables the @code{abs} instruction, which is the absolute difference
17015 between two registers.
17019 Enables all the optional instructions---average, multiply, divide, bit
17020 operations, leading zero, absolute difference, min/max, clip, and
17026 Enables the @code{ave} instruction, which computes the average of two
17029 @item -mbased=@var{n}
17031 Variables of size @var{n} bytes or smaller are placed in the
17032 @code{.based} section by default. Based variables use the @code{$tp}
17033 register as a base register, and there is a 128-byte limit to the
17034 @code{.based} section.
17038 Enables the bit operation instructions---bit test (@code{btstm}), set
17039 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17040 test-and-set (@code{tas}).
17042 @item -mc=@var{name}
17044 Selects which section constant data is placed in. @var{name} may
17045 be @samp{tiny}, @samp{near}, or @samp{far}.
17049 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17050 useful unless you also provide @option{-mminmax}.
17052 @item -mconfig=@var{name}
17054 Selects one of the built-in core configurations. Each MeP chip has
17055 one or more modules in it; each module has a core CPU and a variety of
17056 coprocessors, optional instructions, and peripherals. The
17057 @code{MeP-Integrator} tool, not part of GCC, provides these
17058 configurations through this option; using this option is the same as
17059 using all the corresponding command-line options. The default
17060 configuration is @samp{default}.
17064 Enables the coprocessor instructions. By default, this is a 32-bit
17065 coprocessor. Note that the coprocessor is normally enabled via the
17066 @option{-mconfig=} option.
17070 Enables the 32-bit coprocessor's instructions.
17074 Enables the 64-bit coprocessor's instructions.
17078 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17082 Causes constant variables to be placed in the @code{.near} section.
17086 Enables the @code{div} and @code{divu} instructions.
17090 Generate big-endian code.
17094 Generate little-endian code.
17096 @item -mio-volatile
17097 @opindex mio-volatile
17098 Tells the compiler that any variable marked with the @code{io}
17099 attribute is to be considered volatile.
17103 Causes variables to be assigned to the @code{.far} section by default.
17107 Enables the @code{leadz} (leading zero) instruction.
17111 Causes variables to be assigned to the @code{.near} section by default.
17115 Enables the @code{min} and @code{max} instructions.
17119 Enables the multiplication and multiply-accumulate instructions.
17123 Disables all the optional instructions enabled by @option{-mall-opts}.
17127 Enables the @code{repeat} and @code{erepeat} instructions, used for
17128 low-overhead looping.
17132 Causes all variables to default to the @code{.tiny} section. Note
17133 that there is a 65536-byte limit to this section. Accesses to these
17134 variables use the @code{%gp} base register.
17138 Enables the saturation instructions. Note that the compiler does not
17139 currently generate these itself, but this option is included for
17140 compatibility with other tools, like @code{as}.
17144 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17148 Link the simulator run-time libraries.
17152 Link the simulator runtime libraries, excluding built-in support
17153 for reset and exception vectors and tables.
17157 Causes all functions to default to the @code{.far} section. Without
17158 this option, functions default to the @code{.near} section.
17160 @item -mtiny=@var{n}
17162 Variables that are @var{n} bytes or smaller are allocated to the
17163 @code{.tiny} section. These variables use the @code{$gp} base
17164 register. The default for this option is 4, but note that there's a
17165 65536-byte limit to the @code{.tiny} section.
17169 @node MicroBlaze Options
17170 @subsection MicroBlaze Options
17171 @cindex MicroBlaze Options
17176 @opindex msoft-float
17177 Use software emulation for floating point (default).
17180 @opindex mhard-float
17181 Use hardware floating-point instructions.
17185 Do not optimize block moves, use @code{memcpy}.
17187 @item -mno-clearbss
17188 @opindex mno-clearbss
17189 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17191 @item -mcpu=@var{cpu-type}
17193 Use features of, and schedule code for, the given CPU.
17194 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17195 where @var{X} is a major version, @var{YY} is the minor version, and
17196 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17197 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17199 @item -mxl-soft-mul
17200 @opindex mxl-soft-mul
17201 Use software multiply emulation (default).
17203 @item -mxl-soft-div
17204 @opindex mxl-soft-div
17205 Use software emulation for divides (default).
17207 @item -mxl-barrel-shift
17208 @opindex mxl-barrel-shift
17209 Use the hardware barrel shifter.
17211 @item -mxl-pattern-compare
17212 @opindex mxl-pattern-compare
17213 Use pattern compare instructions.
17215 @item -msmall-divides
17216 @opindex msmall-divides
17217 Use table lookup optimization for small signed integer divisions.
17219 @item -mxl-stack-check
17220 @opindex mxl-stack-check
17221 This option is deprecated. Use @option{-fstack-check} instead.
17224 @opindex mxl-gp-opt
17225 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17227 @item -mxl-multiply-high
17228 @opindex mxl-multiply-high
17229 Use multiply high instructions for high part of 32x32 multiply.
17231 @item -mxl-float-convert
17232 @opindex mxl-float-convert
17233 Use hardware floating-point conversion instructions.
17235 @item -mxl-float-sqrt
17236 @opindex mxl-float-sqrt
17237 Use hardware floating-point square root instruction.
17240 @opindex mbig-endian
17241 Generate code for a big-endian target.
17243 @item -mlittle-endian
17244 @opindex mlittle-endian
17245 Generate code for a little-endian target.
17248 @opindex mxl-reorder
17249 Use reorder instructions (swap and byte reversed load/store).
17251 @item -mxl-mode-@var{app-model}
17252 Select application model @var{app-model}. Valid models are
17255 normal executable (default), uses startup code @file{crt0.o}.
17258 for use with Xilinx Microprocessor Debugger (XMD) based
17259 software intrusive debug agent called xmdstub. This uses startup file
17260 @file{crt1.o} and sets the start address of the program to 0x800.
17263 for applications that are loaded using a bootloader.
17264 This model uses startup file @file{crt2.o} which does not contain a processor
17265 reset vector handler. This is suitable for transferring control on a
17266 processor reset to the bootloader rather than the application.
17269 for applications that do not require any of the
17270 MicroBlaze vectors. This option may be useful for applications running
17271 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17274 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17275 @option{-mxl-mode-@var{app-model}}.
17280 @subsection MIPS Options
17281 @cindex MIPS options
17287 Generate big-endian code.
17291 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17294 @item -march=@var{arch}
17296 Generate code that runs on @var{arch}, which can be the name of a
17297 generic MIPS ISA, or the name of a particular processor.
17299 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17300 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17301 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17302 @samp{mips64r5} and @samp{mips64r6}.
17303 The processor names are:
17304 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17305 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17306 @samp{5kc}, @samp{5kf},
17308 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17309 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17310 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17311 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17312 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17315 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17317 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17318 @samp{m5100}, @samp{m5101},
17319 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17322 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17323 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17324 @samp{rm7000}, @samp{rm9000},
17325 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17328 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17329 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17330 @samp{xlr} and @samp{xlp}.
17331 The special value @samp{from-abi} selects the
17332 most compatible architecture for the selected ABI (that is,
17333 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17335 The native Linux/GNU toolchain also supports the value @samp{native},
17336 which selects the best architecture option for the host processor.
17337 @option{-march=native} has no effect if GCC does not recognize
17340 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17341 (for example, @option{-march=r2k}). Prefixes are optional, and
17342 @samp{vr} may be written @samp{r}.
17344 Names of the form @samp{@var{n}f2_1} refer to processors with
17345 FPUs clocked at half the rate of the core, names of the form
17346 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17347 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17348 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17349 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17350 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17351 accepted as synonyms for @samp{@var{n}f1_1}.
17353 GCC defines two macros based on the value of this option. The first
17354 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17355 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17356 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17357 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17358 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17360 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17361 above. In other words, it has the full prefix and does not
17362 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17363 the macro names the resolved architecture (either @code{"mips1"} or
17364 @code{"mips3"}). It names the default architecture when no
17365 @option{-march} option is given.
17367 @item -mtune=@var{arch}
17369 Optimize for @var{arch}. Among other things, this option controls
17370 the way instructions are scheduled, and the perceived cost of arithmetic
17371 operations. The list of @var{arch} values is the same as for
17374 When this option is not used, GCC optimizes for the processor
17375 specified by @option{-march}. By using @option{-march} and
17376 @option{-mtune} together, it is possible to generate code that
17377 runs on a family of processors, but optimize the code for one
17378 particular member of that family.
17380 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17381 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17382 @option{-march} ones described above.
17386 Equivalent to @option{-march=mips1}.
17390 Equivalent to @option{-march=mips2}.
17394 Equivalent to @option{-march=mips3}.
17398 Equivalent to @option{-march=mips4}.
17402 Equivalent to @option{-march=mips32}.
17406 Equivalent to @option{-march=mips32r3}.
17410 Equivalent to @option{-march=mips32r5}.
17414 Equivalent to @option{-march=mips32r6}.
17418 Equivalent to @option{-march=mips64}.
17422 Equivalent to @option{-march=mips64r2}.
17426 Equivalent to @option{-march=mips64r3}.
17430 Equivalent to @option{-march=mips64r5}.
17434 Equivalent to @option{-march=mips64r6}.
17439 @opindex mno-mips16
17440 Generate (do not generate) MIPS16 code. If GCC is targeting a
17441 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17443 MIPS16 code generation can also be controlled on a per-function basis
17444 by means of @code{mips16} and @code{nomips16} attributes.
17445 @xref{Function Attributes}, for more information.
17447 @item -mflip-mips16
17448 @opindex mflip-mips16
17449 Generate MIPS16 code on alternating functions. This option is provided
17450 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17451 not intended for ordinary use in compiling user code.
17453 @item -minterlink-compressed
17454 @item -mno-interlink-compressed
17455 @opindex minterlink-compressed
17456 @opindex mno-interlink-compressed
17457 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17458 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17460 For example, code using the standard ISA encoding cannot jump directly
17461 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17462 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17463 knows that the target of the jump is not compressed.
17465 @item -minterlink-mips16
17466 @itemx -mno-interlink-mips16
17467 @opindex minterlink-mips16
17468 @opindex mno-interlink-mips16
17469 Aliases of @option{-minterlink-compressed} and
17470 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17471 and are retained for backwards compatibility.
17483 Generate code for the given ABI@.
17485 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17486 generates 64-bit code when you select a 64-bit architecture, but you
17487 can use @option{-mgp32} to get 32-bit code instead.
17489 For information about the O64 ABI, see
17490 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17492 GCC supports a variant of the o32 ABI in which floating-point registers
17493 are 64 rather than 32 bits wide. You can select this combination with
17494 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17495 and @code{mfhc1} instructions and is therefore only supported for
17496 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17498 The register assignments for arguments and return values remain the
17499 same, but each scalar value is passed in a single 64-bit register
17500 rather than a pair of 32-bit registers. For example, scalar
17501 floating-point values are returned in @samp{$f0} only, not a
17502 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17503 remains the same in that the even-numbered double-precision registers
17506 Two additional variants of the o32 ABI are supported to enable
17507 a transition from 32-bit to 64-bit registers. These are FPXX
17508 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17509 The FPXX extension mandates that all code must execute correctly
17510 when run using 32-bit or 64-bit registers. The code can be interlinked
17511 with either FP32 or FP64, but not both.
17512 The FP64A extension is similar to the FP64 extension but forbids the
17513 use of odd-numbered single-precision registers. This can be used
17514 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17515 processors and allows both FP32 and FP64A code to interlink and
17516 run in the same process without changing FPU modes.
17519 @itemx -mno-abicalls
17521 @opindex mno-abicalls
17522 Generate (do not generate) code that is suitable for SVR4-style
17523 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17528 Generate (do not generate) code that is fully position-independent,
17529 and that can therefore be linked into shared libraries. This option
17530 only affects @option{-mabicalls}.
17532 All @option{-mabicalls} code has traditionally been position-independent,
17533 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17534 as an extension, the GNU toolchain allows executables to use absolute
17535 accesses for locally-binding symbols. It can also use shorter GP
17536 initialization sequences and generate direct calls to locally-defined
17537 functions. This mode is selected by @option{-mno-shared}.
17539 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17540 objects that can only be linked by the GNU linker. However, the option
17541 does not affect the ABI of the final executable; it only affects the ABI
17542 of relocatable objects. Using @option{-mno-shared} generally makes
17543 executables both smaller and quicker.
17545 @option{-mshared} is the default.
17551 Assume (do not assume) that the static and dynamic linkers
17552 support PLTs and copy relocations. This option only affects
17553 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17554 has no effect without @option{-msym32}.
17556 You can make @option{-mplt} the default by configuring
17557 GCC with @option{--with-mips-plt}. The default is
17558 @option{-mno-plt} otherwise.
17564 Lift (do not lift) the usual restrictions on the size of the global
17567 GCC normally uses a single instruction to load values from the GOT@.
17568 While this is relatively efficient, it only works if the GOT
17569 is smaller than about 64k. Anything larger causes the linker
17570 to report an error such as:
17572 @cindex relocation truncated to fit (MIPS)
17574 relocation truncated to fit: R_MIPS_GOT16 foobar
17577 If this happens, you should recompile your code with @option{-mxgot}.
17578 This works with very large GOTs, although the code is also
17579 less efficient, since it takes three instructions to fetch the
17580 value of a global symbol.
17582 Note that some linkers can create multiple GOTs. If you have such a
17583 linker, you should only need to use @option{-mxgot} when a single object
17584 file accesses more than 64k's worth of GOT entries. Very few do.
17586 These options have no effect unless GCC is generating position
17591 Assume that general-purpose registers are 32 bits wide.
17595 Assume that general-purpose registers are 64 bits wide.
17599 Assume that floating-point registers are 32 bits wide.
17603 Assume that floating-point registers are 64 bits wide.
17607 Do not assume the width of floating-point registers.
17610 @opindex mhard-float
17611 Use floating-point coprocessor instructions.
17614 @opindex msoft-float
17615 Do not use floating-point coprocessor instructions. Implement
17616 floating-point calculations using library calls instead.
17620 Equivalent to @option{-msoft-float}, but additionally asserts that the
17621 program being compiled does not perform any floating-point operations.
17622 This option is presently supported only by some bare-metal MIPS
17623 configurations, where it may select a special set of libraries
17624 that lack all floating-point support (including, for example, the
17625 floating-point @code{printf} formats).
17626 If code compiled with @option{-mno-float} accidentally contains
17627 floating-point operations, it is likely to suffer a link-time
17628 or run-time failure.
17630 @item -msingle-float
17631 @opindex msingle-float
17632 Assume that the floating-point coprocessor only supports single-precision
17635 @item -mdouble-float
17636 @opindex mdouble-float
17637 Assume that the floating-point coprocessor supports double-precision
17638 operations. This is the default.
17641 @itemx -mno-odd-spreg
17642 @opindex modd-spreg
17643 @opindex mno-odd-spreg
17644 Enable the use of odd-numbered single-precision floating-point registers
17645 for the o32 ABI. This is the default for processors that are known to
17646 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17650 @itemx -mabs=legacy
17652 @opindex mabs=legacy
17653 These options control the treatment of the special not-a-number (NaN)
17654 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17655 @code{neg.@i{fmt}} machine instructions.
17657 By default or when @option{-mabs=legacy} is used the legacy
17658 treatment is selected. In this case these instructions are considered
17659 arithmetic and avoided where correct operation is required and the
17660 input operand might be a NaN. A longer sequence of instructions that
17661 manipulate the sign bit of floating-point datum manually is used
17662 instead unless the @option{-ffinite-math-only} option has also been
17665 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17666 this case these instructions are considered non-arithmetic and therefore
17667 operating correctly in all cases, including in particular where the
17668 input operand is a NaN. These instructions are therefore always used
17669 for the respective operations.
17672 @itemx -mnan=legacy
17674 @opindex mnan=legacy
17675 These options control the encoding of the special not-a-number (NaN)
17676 IEEE 754 floating-point data.
17678 The @option{-mnan=legacy} option selects the legacy encoding. In this
17679 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17680 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17681 by the first bit of their trailing significand field being 1.
17683 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17684 this case qNaNs are denoted by the first bit of their trailing
17685 significand field being 1, whereas sNaNs are denoted by the first bit of
17686 their trailing significand field being 0.
17688 The default is @option{-mnan=legacy} unless GCC has been configured with
17689 @option{--with-nan=2008}.
17695 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17696 implement atomic memory built-in functions. When neither option is
17697 specified, GCC uses the instructions if the target architecture
17700 @option{-mllsc} is useful if the runtime environment can emulate the
17701 instructions and @option{-mno-llsc} can be useful when compiling for
17702 nonstandard ISAs. You can make either option the default by
17703 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17704 respectively. @option{--with-llsc} is the default for some
17705 configurations; see the installation documentation for details.
17711 Use (do not use) revision 1 of the MIPS DSP ASE@.
17712 @xref{MIPS DSP Built-in Functions}. This option defines the
17713 preprocessor macro @code{__mips_dsp}. It also defines
17714 @code{__mips_dsp_rev} to 1.
17720 Use (do not use) revision 2 of the MIPS DSP ASE@.
17721 @xref{MIPS DSP Built-in Functions}. This option defines the
17722 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17723 It also defines @code{__mips_dsp_rev} to 2.
17726 @itemx -mno-smartmips
17727 @opindex msmartmips
17728 @opindex mno-smartmips
17729 Use (do not use) the MIPS SmartMIPS ASE.
17731 @item -mpaired-single
17732 @itemx -mno-paired-single
17733 @opindex mpaired-single
17734 @opindex mno-paired-single
17735 Use (do not use) paired-single floating-point instructions.
17736 @xref{MIPS Paired-Single Support}. This option requires
17737 hardware floating-point support to be enabled.
17743 Use (do not use) MIPS Digital Media Extension instructions.
17744 This option can only be used when generating 64-bit code and requires
17745 hardware floating-point support to be enabled.
17750 @opindex mno-mips3d
17751 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17752 The option @option{-mips3d} implies @option{-mpaired-single}.
17755 @itemx -mno-micromips
17756 @opindex mmicromips
17757 @opindex mno-mmicromips
17758 Generate (do not generate) microMIPS code.
17760 MicroMIPS code generation can also be controlled on a per-function basis
17761 by means of @code{micromips} and @code{nomicromips} attributes.
17762 @xref{Function Attributes}, for more information.
17768 Use (do not use) MT Multithreading instructions.
17774 Use (do not use) the MIPS MCU ASE instructions.
17780 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17786 Use (do not use) the MIPS Virtualization Application Specific instructions.
17792 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17796 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17797 an explanation of the default and the way that the pointer size is
17802 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17804 The default size of @code{int}s, @code{long}s and pointers depends on
17805 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17806 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17807 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17808 or the same size as integer registers, whichever is smaller.
17814 Assume (do not assume) that all symbols have 32-bit values, regardless
17815 of the selected ABI@. This option is useful in combination with
17816 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17817 to generate shorter and faster references to symbolic addresses.
17821 Put definitions of externally-visible data in a small data section
17822 if that data is no bigger than @var{num} bytes. GCC can then generate
17823 more efficient accesses to the data; see @option{-mgpopt} for details.
17825 The default @option{-G} option depends on the configuration.
17827 @item -mlocal-sdata
17828 @itemx -mno-local-sdata
17829 @opindex mlocal-sdata
17830 @opindex mno-local-sdata
17831 Extend (do not extend) the @option{-G} behavior to local data too,
17832 such as to static variables in C@. @option{-mlocal-sdata} is the
17833 default for all configurations.
17835 If the linker complains that an application is using too much small data,
17836 you might want to try rebuilding the less performance-critical parts with
17837 @option{-mno-local-sdata}. You might also want to build large
17838 libraries with @option{-mno-local-sdata}, so that the libraries leave
17839 more room for the main program.
17841 @item -mextern-sdata
17842 @itemx -mno-extern-sdata
17843 @opindex mextern-sdata
17844 @opindex mno-extern-sdata
17845 Assume (do not assume) that externally-defined data is in
17846 a small data section if the size of that data is within the @option{-G} limit.
17847 @option{-mextern-sdata} is the default for all configurations.
17849 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17850 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17851 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17852 is placed in a small data section. If @var{Var} is defined by another
17853 module, you must either compile that module with a high-enough
17854 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17855 definition. If @var{Var} is common, you must link the application
17856 with a high-enough @option{-G} setting.
17858 The easiest way of satisfying these restrictions is to compile
17859 and link every module with the same @option{-G} option. However,
17860 you may wish to build a library that supports several different
17861 small data limits. You can do this by compiling the library with
17862 the highest supported @option{-G} setting and additionally using
17863 @option{-mno-extern-sdata} to stop the library from making assumptions
17864 about externally-defined data.
17870 Use (do not use) GP-relative accesses for symbols that are known to be
17871 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17872 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17875 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17876 might not hold the value of @code{_gp}. For example, if the code is
17877 part of a library that might be used in a boot monitor, programs that
17878 call boot monitor routines pass an unknown value in @code{$gp}.
17879 (In such situations, the boot monitor itself is usually compiled
17880 with @option{-G0}.)
17882 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17883 @option{-mno-extern-sdata}.
17885 @item -membedded-data
17886 @itemx -mno-embedded-data
17887 @opindex membedded-data
17888 @opindex mno-embedded-data
17889 Allocate variables to the read-only data section first if possible, then
17890 next in the small data section if possible, otherwise in data. This gives
17891 slightly slower code than the default, but reduces the amount of RAM required
17892 when executing, and thus may be preferred for some embedded systems.
17894 @item -muninit-const-in-rodata
17895 @itemx -mno-uninit-const-in-rodata
17896 @opindex muninit-const-in-rodata
17897 @opindex mno-uninit-const-in-rodata
17898 Put uninitialized @code{const} variables in the read-only data section.
17899 This option is only meaningful in conjunction with @option{-membedded-data}.
17901 @item -mcode-readable=@var{setting}
17902 @opindex mcode-readable
17903 Specify whether GCC may generate code that reads from executable sections.
17904 There are three possible settings:
17907 @item -mcode-readable=yes
17908 Instructions may freely access executable sections. This is the
17911 @item -mcode-readable=pcrel
17912 MIPS16 PC-relative load instructions can access executable sections,
17913 but other instructions must not do so. This option is useful on 4KSc
17914 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17915 It is also useful on processors that can be configured to have a dual
17916 instruction/data SRAM interface and that, like the M4K, automatically
17917 redirect PC-relative loads to the instruction RAM.
17919 @item -mcode-readable=no
17920 Instructions must not access executable sections. This option can be
17921 useful on targets that are configured to have a dual instruction/data
17922 SRAM interface but that (unlike the M4K) do not automatically redirect
17923 PC-relative loads to the instruction RAM.
17926 @item -msplit-addresses
17927 @itemx -mno-split-addresses
17928 @opindex msplit-addresses
17929 @opindex mno-split-addresses
17930 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17931 relocation operators. This option has been superseded by
17932 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17934 @item -mexplicit-relocs
17935 @itemx -mno-explicit-relocs
17936 @opindex mexplicit-relocs
17937 @opindex mno-explicit-relocs
17938 Use (do not use) assembler relocation operators when dealing with symbolic
17939 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17940 is to use assembler macros instead.
17942 @option{-mexplicit-relocs} is the default if GCC was configured
17943 to use an assembler that supports relocation operators.
17945 @item -mcheck-zero-division
17946 @itemx -mno-check-zero-division
17947 @opindex mcheck-zero-division
17948 @opindex mno-check-zero-division
17949 Trap (do not trap) on integer division by zero.
17951 The default is @option{-mcheck-zero-division}.
17953 @item -mdivide-traps
17954 @itemx -mdivide-breaks
17955 @opindex mdivide-traps
17956 @opindex mdivide-breaks
17957 MIPS systems check for division by zero by generating either a
17958 conditional trap or a break instruction. Using traps results in
17959 smaller code, but is only supported on MIPS II and later. Also, some
17960 versions of the Linux kernel have a bug that prevents trap from
17961 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17962 allow conditional traps on architectures that support them and
17963 @option{-mdivide-breaks} to force the use of breaks.
17965 The default is usually @option{-mdivide-traps}, but this can be
17966 overridden at configure time using @option{--with-divide=breaks}.
17967 Divide-by-zero checks can be completely disabled using
17968 @option{-mno-check-zero-division}.
17973 @opindex mno-memcpy
17974 Force (do not force) the use of @code{memcpy} for non-trivial block
17975 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17976 most constant-sized copies.
17979 @itemx -mno-long-calls
17980 @opindex mlong-calls
17981 @opindex mno-long-calls
17982 Disable (do not disable) use of the @code{jal} instruction. Calling
17983 functions using @code{jal} is more efficient but requires the caller
17984 and callee to be in the same 256 megabyte segment.
17986 This option has no effect on abicalls code. The default is
17987 @option{-mno-long-calls}.
17993 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17994 instructions, as provided by the R4650 ISA@.
18000 Enable (disable) use of the @code{madd} and @code{msub} integer
18001 instructions. The default is @option{-mimadd} on architectures
18002 that support @code{madd} and @code{msub} except for the 74k
18003 architecture where it was found to generate slower code.
18006 @itemx -mno-fused-madd
18007 @opindex mfused-madd
18008 @opindex mno-fused-madd
18009 Enable (disable) use of the floating-point multiply-accumulate
18010 instructions, when they are available. The default is
18011 @option{-mfused-madd}.
18013 On the R8000 CPU when multiply-accumulate instructions are used,
18014 the intermediate product is calculated to infinite precision
18015 and is not subject to the FCSR Flush to Zero bit. This may be
18016 undesirable in some circumstances. On other processors the result
18017 is numerically identical to the equivalent computation using
18018 separate multiply, add, subtract and negate instructions.
18022 Tell the MIPS assembler to not run its preprocessor over user
18023 assembler files (with a @samp{.s} suffix) when assembling them.
18028 @opindex mno-fix-24k
18029 Work around the 24K E48 (lost data on stores during refill) errata.
18030 The workarounds are implemented by the assembler rather than by GCC@.
18033 @itemx -mno-fix-r4000
18034 @opindex mfix-r4000
18035 @opindex mno-fix-r4000
18036 Work around certain R4000 CPU errata:
18039 A double-word or a variable shift may give an incorrect result if executed
18040 immediately after starting an integer division.
18042 A double-word or a variable shift may give an incorrect result if executed
18043 while an integer multiplication is in progress.
18045 An integer division may give an incorrect result if started in a delay slot
18046 of a taken branch or a jump.
18050 @itemx -mno-fix-r4400
18051 @opindex mfix-r4400
18052 @opindex mno-fix-r4400
18053 Work around certain R4400 CPU errata:
18056 A double-word or a variable shift may give an incorrect result if executed
18057 immediately after starting an integer division.
18061 @itemx -mno-fix-r10000
18062 @opindex mfix-r10000
18063 @opindex mno-fix-r10000
18064 Work around certain R10000 errata:
18067 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18068 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18071 This option can only be used if the target architecture supports
18072 branch-likely instructions. @option{-mfix-r10000} is the default when
18073 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18077 @itemx -mno-fix-rm7000
18078 @opindex mfix-rm7000
18079 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18080 workarounds are implemented by the assembler rather than by GCC@.
18083 @itemx -mno-fix-vr4120
18084 @opindex mfix-vr4120
18085 Work around certain VR4120 errata:
18088 @code{dmultu} does not always produce the correct result.
18090 @code{div} and @code{ddiv} do not always produce the correct result if one
18091 of the operands is negative.
18093 The workarounds for the division errata rely on special functions in
18094 @file{libgcc.a}. At present, these functions are only provided by
18095 the @code{mips64vr*-elf} configurations.
18097 Other VR4120 errata require a NOP to be inserted between certain pairs of
18098 instructions. These errata are handled by the assembler, not by GCC itself.
18101 @opindex mfix-vr4130
18102 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18103 workarounds are implemented by the assembler rather than by GCC,
18104 although GCC avoids using @code{mflo} and @code{mfhi} if the
18105 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18106 instructions are available instead.
18109 @itemx -mno-fix-sb1
18111 Work around certain SB-1 CPU core errata.
18112 (This flag currently works around the SB-1 revision 2
18113 ``F1'' and ``F2'' floating-point errata.)
18115 @item -mr10k-cache-barrier=@var{setting}
18116 @opindex mr10k-cache-barrier
18117 Specify whether GCC should insert cache barriers to avoid the
18118 side-effects of speculation on R10K processors.
18120 In common with many processors, the R10K tries to predict the outcome
18121 of a conditional branch and speculatively executes instructions from
18122 the ``taken'' branch. It later aborts these instructions if the
18123 predicted outcome is wrong. However, on the R10K, even aborted
18124 instructions can have side effects.
18126 This problem only affects kernel stores and, depending on the system,
18127 kernel loads. As an example, a speculatively-executed store may load
18128 the target memory into cache and mark the cache line as dirty, even if
18129 the store itself is later aborted. If a DMA operation writes to the
18130 same area of memory before the ``dirty'' line is flushed, the cached
18131 data overwrites the DMA-ed data. See the R10K processor manual
18132 for a full description, including other potential problems.
18134 One workaround is to insert cache barrier instructions before every memory
18135 access that might be speculatively executed and that might have side
18136 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18137 controls GCC's implementation of this workaround. It assumes that
18138 aborted accesses to any byte in the following regions does not have
18143 the memory occupied by the current function's stack frame;
18146 the memory occupied by an incoming stack argument;
18149 the memory occupied by an object with a link-time-constant address.
18152 It is the kernel's responsibility to ensure that speculative
18153 accesses to these regions are indeed safe.
18155 If the input program contains a function declaration such as:
18161 then the implementation of @code{foo} must allow @code{j foo} and
18162 @code{jal foo} to be executed speculatively. GCC honors this
18163 restriction for functions it compiles itself. It expects non-GCC
18164 functions (such as hand-written assembly code) to do the same.
18166 The option has three forms:
18169 @item -mr10k-cache-barrier=load-store
18170 Insert a cache barrier before a load or store that might be
18171 speculatively executed and that might have side effects even
18174 @item -mr10k-cache-barrier=store
18175 Insert a cache barrier before a store that might be speculatively
18176 executed and that might have side effects even if aborted.
18178 @item -mr10k-cache-barrier=none
18179 Disable the insertion of cache barriers. This is the default setting.
18182 @item -mflush-func=@var{func}
18183 @itemx -mno-flush-func
18184 @opindex mflush-func
18185 Specifies the function to call to flush the I and D caches, or to not
18186 call any such function. If called, the function must take the same
18187 arguments as the common @code{_flush_func}, that is, the address of the
18188 memory range for which the cache is being flushed, the size of the
18189 memory range, and the number 3 (to flush both caches). The default
18190 depends on the target GCC was configured for, but commonly is either
18191 @code{_flush_func} or @code{__cpu_flush}.
18193 @item mbranch-cost=@var{num}
18194 @opindex mbranch-cost
18195 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18196 This cost is only a heuristic and is not guaranteed to produce
18197 consistent results across releases. A zero cost redundantly selects
18198 the default, which is based on the @option{-mtune} setting.
18200 @item -mbranch-likely
18201 @itemx -mno-branch-likely
18202 @opindex mbranch-likely
18203 @opindex mno-branch-likely
18204 Enable or disable use of Branch Likely instructions, regardless of the
18205 default for the selected architecture. By default, Branch Likely
18206 instructions may be generated if they are supported by the selected
18207 architecture. An exception is for the MIPS32 and MIPS64 architectures
18208 and processors that implement those architectures; for those, Branch
18209 Likely instructions are not be generated by default because the MIPS32
18210 and MIPS64 architectures specifically deprecate their use.
18212 @item -mcompact-branches=never
18213 @itemx -mcompact-branches=optimal
18214 @itemx -mcompact-branches=always
18215 @opindex mcompact-branches=never
18216 @opindex mcompact-branches=optimal
18217 @opindex mcompact-branches=always
18218 These options control which form of branches will be generated. The
18219 default is @option{-mcompact-branches=optimal}.
18221 The @option{-mcompact-branches=never} option ensures that compact branch
18222 instructions will never be generated.
18224 The @option{-mcompact-branches=always} option ensures that a compact
18225 branch instruction will be generated if available. If a compact branch
18226 instruction is not available, a delay slot form of the branch will be
18229 This option is supported from MIPS Release 6 onwards.
18231 The @option{-mcompact-branches=optimal} option will cause a delay slot
18232 branch to be used if one is available in the current ISA and the delay
18233 slot is successfully filled. If the delay slot is not filled, a compact
18234 branch will be chosen if one is available.
18236 @item -mfp-exceptions
18237 @itemx -mno-fp-exceptions
18238 @opindex mfp-exceptions
18239 Specifies whether FP exceptions are enabled. This affects how
18240 FP instructions are scheduled for some processors.
18241 The default is that FP exceptions are
18244 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18245 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18248 @item -mvr4130-align
18249 @itemx -mno-vr4130-align
18250 @opindex mvr4130-align
18251 The VR4130 pipeline is two-way superscalar, but can only issue two
18252 instructions together if the first one is 8-byte aligned. When this
18253 option is enabled, GCC aligns pairs of instructions that it
18254 thinks should execute in parallel.
18256 This option only has an effect when optimizing for the VR4130.
18257 It normally makes code faster, but at the expense of making it bigger.
18258 It is enabled by default at optimization level @option{-O3}.
18263 Enable (disable) generation of @code{synci} instructions on
18264 architectures that support it. The @code{synci} instructions (if
18265 enabled) are generated when @code{__builtin___clear_cache} is
18268 This option defaults to @option{-mno-synci}, but the default can be
18269 overridden by configuring GCC with @option{--with-synci}.
18271 When compiling code for single processor systems, it is generally safe
18272 to use @code{synci}. However, on many multi-core (SMP) systems, it
18273 does not invalidate the instruction caches on all cores and may lead
18274 to undefined behavior.
18276 @item -mrelax-pic-calls
18277 @itemx -mno-relax-pic-calls
18278 @opindex mrelax-pic-calls
18279 Try to turn PIC calls that are normally dispatched via register
18280 @code{$25} into direct calls. This is only possible if the linker can
18281 resolve the destination at link-time and if the destination is within
18282 range for a direct call.
18284 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18285 an assembler and a linker that support the @code{.reloc} assembly
18286 directive and @option{-mexplicit-relocs} is in effect. With
18287 @option{-mno-explicit-relocs}, this optimization can be performed by the
18288 assembler and the linker alone without help from the compiler.
18290 @item -mmcount-ra-address
18291 @itemx -mno-mcount-ra-address
18292 @opindex mmcount-ra-address
18293 @opindex mno-mcount-ra-address
18294 Emit (do not emit) code that allows @code{_mcount} to modify the
18295 calling function's return address. When enabled, this option extends
18296 the usual @code{_mcount} interface with a new @var{ra-address}
18297 parameter, which has type @code{intptr_t *} and is passed in register
18298 @code{$12}. @code{_mcount} can then modify the return address by
18299 doing both of the following:
18302 Returning the new address in register @code{$31}.
18304 Storing the new address in @code{*@var{ra-address}},
18305 if @var{ra-address} is nonnull.
18308 The default is @option{-mno-mcount-ra-address}.
18310 @item -mframe-header-opt
18311 @itemx -mno-frame-header-opt
18312 @opindex mframe-header-opt
18313 Enable (disable) frame header optimization in the o32 ABI. When using the
18314 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18315 function to write out register arguments. When enabled, this optimization
18316 will suppress the allocation of the frame header if it can be determined that
18319 This optimization is off by default at all optimization levels.
18324 @subsection MMIX Options
18325 @cindex MMIX Options
18327 These options are defined for the MMIX:
18331 @itemx -mno-libfuncs
18333 @opindex mno-libfuncs
18334 Specify that intrinsic library functions are being compiled, passing all
18335 values in registers, no matter the size.
18338 @itemx -mno-epsilon
18340 @opindex mno-epsilon
18341 Generate floating-point comparison instructions that compare with respect
18342 to the @code{rE} epsilon register.
18344 @item -mabi=mmixware
18346 @opindex mabi=mmixware
18348 Generate code that passes function parameters and return values that (in
18349 the called function) are seen as registers @code{$0} and up, as opposed to
18350 the GNU ABI which uses global registers @code{$231} and up.
18352 @item -mzero-extend
18353 @itemx -mno-zero-extend
18354 @opindex mzero-extend
18355 @opindex mno-zero-extend
18356 When reading data from memory in sizes shorter than 64 bits, use (do not
18357 use) zero-extending load instructions by default, rather than
18358 sign-extending ones.
18361 @itemx -mno-knuthdiv
18363 @opindex mno-knuthdiv
18364 Make the result of a division yielding a remainder have the same sign as
18365 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18366 remainder follows the sign of the dividend. Both methods are
18367 arithmetically valid, the latter being almost exclusively used.
18369 @item -mtoplevel-symbols
18370 @itemx -mno-toplevel-symbols
18371 @opindex mtoplevel-symbols
18372 @opindex mno-toplevel-symbols
18373 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18374 code can be used with the @code{PREFIX} assembly directive.
18378 Generate an executable in the ELF format, rather than the default
18379 @samp{mmo} format used by the @command{mmix} simulator.
18381 @item -mbranch-predict
18382 @itemx -mno-branch-predict
18383 @opindex mbranch-predict
18384 @opindex mno-branch-predict
18385 Use (do not use) the probable-branch instructions, when static branch
18386 prediction indicates a probable branch.
18388 @item -mbase-addresses
18389 @itemx -mno-base-addresses
18390 @opindex mbase-addresses
18391 @opindex mno-base-addresses
18392 Generate (do not generate) code that uses @emph{base addresses}. Using a
18393 base address automatically generates a request (handled by the assembler
18394 and the linker) for a constant to be set up in a global register. The
18395 register is used for one or more base address requests within the range 0
18396 to 255 from the value held in the register. The generally leads to short
18397 and fast code, but the number of different data items that can be
18398 addressed is limited. This means that a program that uses lots of static
18399 data may require @option{-mno-base-addresses}.
18401 @item -msingle-exit
18402 @itemx -mno-single-exit
18403 @opindex msingle-exit
18404 @opindex mno-single-exit
18405 Force (do not force) generated code to have a single exit point in each
18409 @node MN10300 Options
18410 @subsection MN10300 Options
18411 @cindex MN10300 options
18413 These @option{-m} options are defined for Matsushita MN10300 architectures:
18418 Generate code to avoid bugs in the multiply instructions for the MN10300
18419 processors. This is the default.
18421 @item -mno-mult-bug
18422 @opindex mno-mult-bug
18423 Do not generate code to avoid bugs in the multiply instructions for the
18424 MN10300 processors.
18428 Generate code using features specific to the AM33 processor.
18432 Do not generate code using features specific to the AM33 processor. This
18437 Generate code using features specific to the AM33/2.0 processor.
18441 Generate code using features specific to the AM34 processor.
18443 @item -mtune=@var{cpu-type}
18445 Use the timing characteristics of the indicated CPU type when
18446 scheduling instructions. This does not change the targeted processor
18447 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18448 @samp{am33-2} or @samp{am34}.
18450 @item -mreturn-pointer-on-d0
18451 @opindex mreturn-pointer-on-d0
18452 When generating a function that returns a pointer, return the pointer
18453 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18454 only in @code{a0}, and attempts to call such functions without a prototype
18455 result in errors. Note that this option is on by default; use
18456 @option{-mno-return-pointer-on-d0} to disable it.
18460 Do not link in the C run-time initialization object file.
18464 Indicate to the linker that it should perform a relaxation optimization pass
18465 to shorten branches, calls and absolute memory addresses. This option only
18466 has an effect when used on the command line for the final link step.
18468 This option makes symbolic debugging impossible.
18472 Allow the compiler to generate @emph{Long Instruction Word}
18473 instructions if the target is the @samp{AM33} or later. This is the
18474 default. This option defines the preprocessor macro @code{__LIW__}.
18478 Do not allow the compiler to generate @emph{Long Instruction Word}
18479 instructions. This option defines the preprocessor macro
18484 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18485 instructions if the target is the @samp{AM33} or later. This is the
18486 default. This option defines the preprocessor macro @code{__SETLB__}.
18490 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18491 instructions. This option defines the preprocessor macro
18492 @code{__NO_SETLB__}.
18496 @node Moxie Options
18497 @subsection Moxie Options
18498 @cindex Moxie Options
18504 Generate big-endian code. This is the default for @samp{moxie-*-*}
18509 Generate little-endian code.
18513 Generate mul.x and umul.x instructions. This is the default for
18514 @samp{moxiebox-*-*} configurations.
18518 Do not link in the C run-time initialization object file.
18522 @node MSP430 Options
18523 @subsection MSP430 Options
18524 @cindex MSP430 Options
18526 These options are defined for the MSP430:
18532 Force assembly output to always use hex constants. Normally such
18533 constants are signed decimals, but this option is available for
18534 testsuite and/or aesthetic purposes.
18538 Select the MCU to target. This is used to create a C preprocessor
18539 symbol based upon the MCU name, converted to upper case and pre- and
18540 post-fixed with @samp{__}. This in turn is used by the
18541 @file{msp430.h} header file to select an MCU-specific supplementary
18544 The option also sets the ISA to use. If the MCU name is one that is
18545 known to only support the 430 ISA then that is selected, otherwise the
18546 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18547 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18548 name selects the 430X ISA.
18550 In addition an MCU-specific linker script is added to the linker
18551 command line. The script's name is the name of the MCU with
18552 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18553 command line defines the C preprocessor symbol @code{__XXX__} and
18554 cause the linker to search for a script called @file{xxx.ld}.
18556 This option is also passed on to the assembler.
18559 @itemx -mno-warn-mcu
18561 @opindex mno-warn-mcu
18562 This option enables or disables warnings about conflicts between the
18563 MCU name specified by the @option{-mmcu} option and the ISA set by the
18564 @option{-mcpu} option and/or the hardware multiply support set by the
18565 @option{-mhwmult} option. It also toggles warnings about unrecognised
18566 MCU names. This option is on by default.
18570 Specifies the ISA to use. Accepted values are @samp{msp430},
18571 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18572 @option{-mmcu=} option should be used to select the ISA.
18576 Link to the simulator runtime libraries and linker script. Overrides
18577 any scripts that would be selected by the @option{-mmcu=} option.
18581 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18585 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18589 This option is passed to the assembler and linker, and allows the
18590 linker to perform certain optimizations that cannot be done until
18595 Describes the type of hardware multiply supported by the target.
18596 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18597 for the original 16-bit-only multiply supported by early MCUs.
18598 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18599 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18600 A value of @samp{auto} can also be given. This tells GCC to deduce
18601 the hardware multiply support based upon the MCU name provided by the
18602 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
18603 the MCU name is not recognised then no hardware multiply support is
18604 assumed. @code{auto} is the default setting.
18606 Hardware multiplies are normally performed by calling a library
18607 routine. This saves space in the generated code. When compiling at
18608 @option{-O3} or higher however the hardware multiplier is invoked
18609 inline. This makes for bigger, but faster code.
18611 The hardware multiply routines disable interrupts whilst running and
18612 restore the previous interrupt state when they finish. This makes
18613 them safe to use inside interrupt handlers as well as in normal code.
18617 Enable the use of a minimum runtime environment - no static
18618 initializers or constructors. This is intended for memory-constrained
18619 devices. The compiler includes special symbols in some objects
18620 that tell the linker and runtime which code fragments are required.
18622 @item -mcode-region=
18623 @itemx -mdata-region=
18624 @opindex mcode-region
18625 @opindex mdata-region
18626 These options tell the compiler where to place functions and data that
18627 do not have one of the @code{lower}, @code{upper}, @code{either} or
18628 @code{section} attributes. Possible values are @code{lower},
18629 @code{upper}, @code{either} or @code{any}. The first three behave
18630 like the corresponding attribute. The fourth possible value -
18631 @code{any} - is the default. It leaves placement entirely up to the
18632 linker script and how it assigns the standard sections (.text, .data
18633 etc) to the memory regions.
18635 @item -msilicon-errata=
18636 @opindex msilicon-errata
18637 This option passes on a request to assembler to enable the fixes for
18638 the named silicon errata.
18640 @item -msilicon-errata-warn=
18641 @opindex msilicon-errata-warn
18642 This option passes on a request to the assembler to enable warning
18643 messages when a silicon errata might need to be applied.
18647 @node NDS32 Options
18648 @subsection NDS32 Options
18649 @cindex NDS32 Options
18651 These options are defined for NDS32 implementations:
18656 @opindex mbig-endian
18657 Generate code in big-endian mode.
18659 @item -mlittle-endian
18660 @opindex mlittle-endian
18661 Generate code in little-endian mode.
18663 @item -mreduced-regs
18664 @opindex mreduced-regs
18665 Use reduced-set registers for register allocation.
18668 @opindex mfull-regs
18669 Use full-set registers for register allocation.
18673 Generate conditional move instructions.
18677 Do not generate conditional move instructions.
18681 Generate performance extension instructions.
18683 @item -mno-perf-ext
18684 @opindex mno-perf-ext
18685 Do not generate performance extension instructions.
18689 Generate v3 push25/pop25 instructions.
18692 @opindex mno-v3push
18693 Do not generate v3 push25/pop25 instructions.
18697 Generate 16-bit instructions.
18700 @opindex mno-16-bit
18701 Do not generate 16-bit instructions.
18703 @item -misr-vector-size=@var{num}
18704 @opindex misr-vector-size
18705 Specify the size of each interrupt vector, which must be 4 or 16.
18707 @item -mcache-block-size=@var{num}
18708 @opindex mcache-block-size
18709 Specify the size of each cache block,
18710 which must be a power of 2 between 4 and 512.
18712 @item -march=@var{arch}
18714 Specify the name of the target architecture.
18716 @item -mcmodel=@var{code-model}
18718 Set the code model to one of
18721 All the data and read-only data segments must be within 512KB addressing space.
18722 The text segment must be within 16MB addressing space.
18723 @item @samp{medium}
18724 The data segment must be within 512KB while the read-only data segment can be
18725 within 4GB addressing space. The text segment should be still within 16MB
18728 All the text and data segments can be within 4GB addressing space.
18732 @opindex mctor-dtor
18733 Enable constructor/destructor feature.
18737 Guide linker to relax instructions.
18741 @node Nios II Options
18742 @subsection Nios II Options
18743 @cindex Nios II options
18744 @cindex Altera Nios II options
18746 These are the options defined for the Altera Nios II processor.
18752 @cindex smaller data references
18753 Put global and static objects less than or equal to @var{num} bytes
18754 into the small data or BSS sections instead of the normal data or BSS
18755 sections. The default value of @var{num} is 8.
18757 @item -mgpopt=@var{option}
18762 Generate (do not generate) GP-relative accesses. The following
18763 @var{option} names are recognized:
18768 Do not generate GP-relative accesses.
18771 Generate GP-relative accesses for small data objects that are not
18772 external, weak, or uninitialized common symbols.
18773 Also use GP-relative addressing for objects that
18774 have been explicitly placed in a small data section via a @code{section}
18778 As for @samp{local}, but also generate GP-relative accesses for
18779 small data objects that are external, weak, or common. If you use this option,
18780 you must ensure that all parts of your program (including libraries) are
18781 compiled with the same @option{-G} setting.
18784 Generate GP-relative accesses for all data objects in the program. If you
18785 use this option, the entire data and BSS segments
18786 of your program must fit in 64K of memory and you must use an appropriate
18787 linker script to allocate them within the addressible range of the
18791 Generate GP-relative addresses for function pointers as well as data
18792 pointers. If you use this option, the entire text, data, and BSS segments
18793 of your program must fit in 64K of memory and you must use an appropriate
18794 linker script to allocate them within the addressible range of the
18799 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18800 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18802 The default is @option{-mgpopt} except when @option{-fpic} or
18803 @option{-fPIC} is specified to generate position-independent code.
18804 Note that the Nios II ABI does not permit GP-relative accesses from
18807 You may need to specify @option{-mno-gpopt} explicitly when building
18808 programs that include large amounts of small data, including large
18809 GOT data sections. In this case, the 16-bit offset for GP-relative
18810 addressing may not be large enough to allow access to the entire
18811 small data section.
18817 Generate little-endian (default) or big-endian (experimental) code,
18820 @item -march=@var{arch}
18822 This specifies the name of the target Nios II architecture. GCC uses this
18823 name to determine what kind of instructions it can emit when generating
18824 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
18826 The preprocessor macro @code{__nios2_arch__} is available to programs,
18827 with value 1 or 2, indicating the targeted ISA level.
18829 @item -mbypass-cache
18830 @itemx -mno-bypass-cache
18831 @opindex mno-bypass-cache
18832 @opindex mbypass-cache
18833 Force all load and store instructions to always bypass cache by
18834 using I/O variants of the instructions. The default is not to
18837 @item -mno-cache-volatile
18838 @itemx -mcache-volatile
18839 @opindex mcache-volatile
18840 @opindex mno-cache-volatile
18841 Volatile memory access bypass the cache using the I/O variants of
18842 the load and store instructions. The default is not to bypass the cache.
18844 @item -mno-fast-sw-div
18845 @itemx -mfast-sw-div
18846 @opindex mno-fast-sw-div
18847 @opindex mfast-sw-div
18848 Do not use table-based fast divide for small numbers. The default
18849 is to use the fast divide at @option{-O3} and above.
18853 @itemx -mno-hw-mulx
18857 @opindex mno-hw-mul
18859 @opindex mno-hw-mulx
18861 @opindex mno-hw-div
18863 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18864 instructions by the compiler. The default is to emit @code{mul}
18865 and not emit @code{div} and @code{mulx}.
18871 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
18872 CDX (code density) instructions. Enabling these instructions also
18873 requires @option{-march=r2}. Since these instructions are optional
18874 extensions to the R2 architecture, the default is not to emit them.
18876 @item -mcustom-@var{insn}=@var{N}
18877 @itemx -mno-custom-@var{insn}
18878 @opindex mcustom-@var{insn}
18879 @opindex mno-custom-@var{insn}
18880 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18881 custom instruction with encoding @var{N} when generating code that uses
18882 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18883 instruction 253 for single-precision floating-point add operations instead
18884 of the default behavior of using a library call.
18886 The following values of @var{insn} are supported. Except as otherwise
18887 noted, floating-point operations are expected to be implemented with
18888 normal IEEE 754 semantics and correspond directly to the C operators or the
18889 equivalent GCC built-in functions (@pxref{Other Builtins}).
18891 Single-precision floating point:
18894 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18895 Binary arithmetic operations.
18901 Unary absolute value.
18903 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18904 Comparison operations.
18906 @item @samp{fmins}, @samp{fmaxs}
18907 Floating-point minimum and maximum. These instructions are only
18908 generated if @option{-ffinite-math-only} is specified.
18910 @item @samp{fsqrts}
18911 Unary square root operation.
18913 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18914 Floating-point trigonometric and exponential functions. These instructions
18915 are only generated if @option{-funsafe-math-optimizations} is also specified.
18919 Double-precision floating point:
18922 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18923 Binary arithmetic operations.
18929 Unary absolute value.
18931 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18932 Comparison operations.
18934 @item @samp{fmind}, @samp{fmaxd}
18935 Double-precision minimum and maximum. These instructions are only
18936 generated if @option{-ffinite-math-only} is specified.
18938 @item @samp{fsqrtd}
18939 Unary square root operation.
18941 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18942 Double-precision trigonometric and exponential functions. These instructions
18943 are only generated if @option{-funsafe-math-optimizations} is also specified.
18949 @item @samp{fextsd}
18950 Conversion from single precision to double precision.
18952 @item @samp{ftruncds}
18953 Conversion from double precision to single precision.
18955 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18956 Conversion from floating point to signed or unsigned integer types, with
18957 truncation towards zero.
18960 Conversion from single-precision floating point to signed integer,
18961 rounding to the nearest integer and ties away from zero.
18962 This corresponds to the @code{__builtin_lroundf} function when
18963 @option{-fno-math-errno} is used.
18965 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18966 Conversion from signed or unsigned integer types to floating-point types.
18970 In addition, all of the following transfer instructions for internal
18971 registers X and Y must be provided to use any of the double-precision
18972 floating-point instructions. Custom instructions taking two
18973 double-precision source operands expect the first operand in the
18974 64-bit register X. The other operand (or only operand of a unary
18975 operation) is given to the custom arithmetic instruction with the
18976 least significant half in source register @var{src1} and the most
18977 significant half in @var{src2}. A custom instruction that returns a
18978 double-precision result returns the most significant 32 bits in the
18979 destination register and the other half in 32-bit register Y.
18980 GCC automatically generates the necessary code sequences to write
18981 register X and/or read register Y when double-precision floating-point
18982 instructions are used.
18987 Write @var{src1} into the least significant half of X and @var{src2} into
18988 the most significant half of X.
18991 Write @var{src1} into Y.
18993 @item @samp{frdxhi}, @samp{frdxlo}
18994 Read the most or least (respectively) significant half of X and store it in
18998 Read the value of Y and store it into @var{dest}.
19001 Note that you can gain more local control over generation of Nios II custom
19002 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19003 and @code{target("no-custom-@var{insn}")} function attributes
19004 (@pxref{Function Attributes})
19005 or pragmas (@pxref{Function Specific Option Pragmas}).
19007 @item -mcustom-fpu-cfg=@var{name}
19008 @opindex mcustom-fpu-cfg
19010 This option enables a predefined, named set of custom instruction encodings
19011 (see @option{-mcustom-@var{insn}} above).
19012 Currently, the following sets are defined:
19014 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19015 @gccoptlist{-mcustom-fmuls=252 @gol
19016 -mcustom-fadds=253 @gol
19017 -mcustom-fsubs=254 @gol
19018 -fsingle-precision-constant}
19020 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19021 @gccoptlist{-mcustom-fmuls=252 @gol
19022 -mcustom-fadds=253 @gol
19023 -mcustom-fsubs=254 @gol
19024 -mcustom-fdivs=255 @gol
19025 -fsingle-precision-constant}
19027 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19028 @gccoptlist{-mcustom-floatus=243 @gol
19029 -mcustom-fixsi=244 @gol
19030 -mcustom-floatis=245 @gol
19031 -mcustom-fcmpgts=246 @gol
19032 -mcustom-fcmples=249 @gol
19033 -mcustom-fcmpeqs=250 @gol
19034 -mcustom-fcmpnes=251 @gol
19035 -mcustom-fmuls=252 @gol
19036 -mcustom-fadds=253 @gol
19037 -mcustom-fsubs=254 @gol
19038 -mcustom-fdivs=255 @gol
19039 -fsingle-precision-constant}
19041 Custom instruction assignments given by individual
19042 @option{-mcustom-@var{insn}=} options override those given by
19043 @option{-mcustom-fpu-cfg=}, regardless of the
19044 order of the options on the command line.
19046 Note that you can gain more local control over selection of a FPU
19047 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19048 function attribute (@pxref{Function Attributes})
19049 or pragma (@pxref{Function Specific Option Pragmas}).
19053 These additional @samp{-m} options are available for the Altera Nios II
19054 ELF (bare-metal) target:
19060 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19061 startup and termination code, and is typically used in conjunction with
19062 @option{-msys-crt0=} to specify the location of the alternate startup code
19063 provided by the HAL BSP.
19067 Link with a limited version of the C library, @option{-lsmallc}, rather than
19070 @item -msys-crt0=@var{startfile}
19072 @var{startfile} is the file name of the startfile (crt0) to use
19073 when linking. This option is only useful in conjunction with @option{-mhal}.
19075 @item -msys-lib=@var{systemlib}
19077 @var{systemlib} is the library name of the library that provides
19078 low-level system calls required by the C library,
19079 e.g. @code{read} and @code{write}.
19080 This option is typically used to link with a library provided by a HAL BSP.
19084 @node Nvidia PTX Options
19085 @subsection Nvidia PTX Options
19086 @cindex Nvidia PTX options
19087 @cindex nvptx options
19089 These options are defined for Nvidia PTX:
19097 Generate code for 32-bit or 64-bit ABI.
19100 @opindex mmainkernel
19101 Link in code for a __main kernel. This is for stand-alone instead of
19102 offloading execution.
19106 Apply partitioned execution optimizations. This is the default when any
19107 level of optimization is selected.
19111 @node PDP-11 Options
19112 @subsection PDP-11 Options
19113 @cindex PDP-11 Options
19115 These options are defined for the PDP-11:
19120 Use hardware FPP floating point. This is the default. (FIS floating
19121 point on the PDP-11/40 is not supported.)
19124 @opindex msoft-float
19125 Do not use hardware floating point.
19129 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19133 Return floating-point results in memory. This is the default.
19137 Generate code for a PDP-11/40.
19141 Generate code for a PDP-11/45. This is the default.
19145 Generate code for a PDP-11/10.
19147 @item -mbcopy-builtin
19148 @opindex mbcopy-builtin
19149 Use inline @code{movmemhi} patterns for copying memory. This is the
19154 Do not use inline @code{movmemhi} patterns for copying memory.
19160 Use 16-bit @code{int}. This is the default.
19166 Use 32-bit @code{int}.
19169 @itemx -mno-float32
19171 @opindex mno-float32
19172 Use 64-bit @code{float}. This is the default.
19175 @itemx -mno-float64
19177 @opindex mno-float64
19178 Use 32-bit @code{float}.
19182 Use @code{abshi2} pattern. This is the default.
19186 Do not use @code{abshi2} pattern.
19188 @item -mbranch-expensive
19189 @opindex mbranch-expensive
19190 Pretend that branches are expensive. This is for experimenting with
19191 code generation only.
19193 @item -mbranch-cheap
19194 @opindex mbranch-cheap
19195 Do not pretend that branches are expensive. This is the default.
19199 Use Unix assembler syntax. This is the default when configured for
19200 @samp{pdp11-*-bsd}.
19204 Use DEC assembler syntax. This is the default when configured for any
19205 PDP-11 target other than @samp{pdp11-*-bsd}.
19208 @node picoChip Options
19209 @subsection picoChip Options
19210 @cindex picoChip options
19212 These @samp{-m} options are defined for picoChip implementations:
19216 @item -mae=@var{ae_type}
19218 Set the instruction set, register set, and instruction scheduling
19219 parameters for array element type @var{ae_type}. Supported values
19220 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19222 @option{-mae=ANY} selects a completely generic AE type. Code
19223 generated with this option runs on any of the other AE types. The
19224 code is not as efficient as it would be if compiled for a specific
19225 AE type, and some types of operation (e.g., multiplication) do not
19226 work properly on all types of AE.
19228 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19229 for compiled code, and is the default.
19231 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19232 option may suffer from poor performance of byte (char) manipulation,
19233 since the DSP AE does not provide hardware support for byte load/stores.
19235 @item -msymbol-as-address
19236 Enable the compiler to directly use a symbol name as an address in a
19237 load/store instruction, without first loading it into a
19238 register. Typically, the use of this option generates larger
19239 programs, which run faster than when the option isn't used. However, the
19240 results vary from program to program, so it is left as a user option,
19241 rather than being permanently enabled.
19243 @item -mno-inefficient-warnings
19244 Disables warnings about the generation of inefficient code. These
19245 warnings can be generated, for example, when compiling code that
19246 performs byte-level memory operations on the MAC AE type. The MAC AE has
19247 no hardware support for byte-level memory operations, so all byte
19248 load/stores must be synthesized from word load/store operations. This is
19249 inefficient and a warning is generated to indicate
19250 that you should rewrite the code to avoid byte operations, or to target
19251 an AE type that has the necessary hardware support. This option disables
19256 @node PowerPC Options
19257 @subsection PowerPC Options
19258 @cindex PowerPC options
19260 These are listed under @xref{RS/6000 and PowerPC Options}.
19263 @subsection RL78 Options
19264 @cindex RL78 Options
19270 Links in additional target libraries to support operation within a
19279 Specifies the type of hardware multiplication and division support to
19280 be used. The simplest is @code{none}, which uses software for both
19281 multiplication and division. This is the default. The @code{g13}
19282 value is for the hardware multiply/divide peripheral found on the
19283 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19284 the multiplication and division instructions supported by the RL78/G14
19285 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19286 the value @code{mg10} is an alias for @code{none}.
19288 In addition a C preprocessor macro is defined, based upon the setting
19289 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19290 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19297 Specifies the RL78 core to target. The default is the G14 core, also
19298 known as an S3 core or just RL78. The G13 or S2 core does not have
19299 multiply or divide instructions, instead it uses a hardware peripheral
19300 for these operations. The G10 or S1 core does not have register
19301 banks, so it uses a different calling convention.
19303 If this option is set it also selects the type of hardware multiply
19304 support to use, unless this is overridden by an explicit
19305 @option{-mmul=none} option on the command line. Thus specifying
19306 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19307 peripheral and specifying @option{-mcpu=g10} disables the use of
19308 hardware multipications altogether.
19310 Note, although the RL78/G14 core is the default target, specifying
19311 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19312 change the behaviour of the toolchain since it also enables G14
19313 hardware multiply support. If these options are not specified on the
19314 command line then software multiplication routines will be used even
19315 though the code targets the RL78 core. This is for backwards
19316 compatibility with older toolchains which did not have hardware
19317 multiply and divide support.
19319 In addition a C preprocessor macro is defined, based upon the setting
19320 of this option. Possible values are: @code{__RL78_G10__},
19321 @code{__RL78_G13__} or @code{__RL78_G14__}.
19331 These are aliases for the corresponding @option{-mcpu=} option. They
19332 are provided for backwards compatibility.
19336 Allow the compiler to use all of the available registers. By default
19337 registers @code{r24..r31} are reserved for use in interrupt handlers.
19338 With this option enabled these registers can be used in ordinary
19341 @item -m64bit-doubles
19342 @itemx -m32bit-doubles
19343 @opindex m64bit-doubles
19344 @opindex m32bit-doubles
19345 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19346 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19347 @option{-m32bit-doubles}.
19351 @node RS/6000 and PowerPC Options
19352 @subsection IBM RS/6000 and PowerPC Options
19353 @cindex RS/6000 and PowerPC Options
19354 @cindex IBM RS/6000 and PowerPC Options
19356 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19358 @item -mpowerpc-gpopt
19359 @itemx -mno-powerpc-gpopt
19360 @itemx -mpowerpc-gfxopt
19361 @itemx -mno-powerpc-gfxopt
19364 @itemx -mno-powerpc64
19368 @itemx -mno-popcntb
19370 @itemx -mno-popcntd
19379 @itemx -mno-hard-dfp
19380 @opindex mpowerpc-gpopt
19381 @opindex mno-powerpc-gpopt
19382 @opindex mpowerpc-gfxopt
19383 @opindex mno-powerpc-gfxopt
19384 @opindex mpowerpc64
19385 @opindex mno-powerpc64
19389 @opindex mno-popcntb
19391 @opindex mno-popcntd
19397 @opindex mno-mfpgpr
19399 @opindex mno-hard-dfp
19400 You use these options to specify which instructions are available on the
19401 processor you are using. The default value of these options is
19402 determined when configuring GCC@. Specifying the
19403 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19404 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19405 rather than the options listed above.
19407 Specifying @option{-mpowerpc-gpopt} allows
19408 GCC to use the optional PowerPC architecture instructions in the
19409 General Purpose group, including floating-point square root. Specifying
19410 @option{-mpowerpc-gfxopt} allows GCC to
19411 use the optional PowerPC architecture instructions in the Graphics
19412 group, including floating-point select.
19414 The @option{-mmfcrf} option allows GCC to generate the move from
19415 condition register field instruction implemented on the POWER4
19416 processor and other processors that support the PowerPC V2.01
19418 The @option{-mpopcntb} option allows GCC to generate the popcount and
19419 double-precision FP reciprocal estimate instruction implemented on the
19420 POWER5 processor and other processors that support the PowerPC V2.02
19422 The @option{-mpopcntd} option allows GCC to generate the popcount
19423 instruction implemented on the POWER7 processor and other processors
19424 that support the PowerPC V2.06 architecture.
19425 The @option{-mfprnd} option allows GCC to generate the FP round to
19426 integer instructions implemented on the POWER5+ processor and other
19427 processors that support the PowerPC V2.03 architecture.
19428 The @option{-mcmpb} option allows GCC to generate the compare bytes
19429 instruction implemented on the POWER6 processor and other processors
19430 that support the PowerPC V2.05 architecture.
19431 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19432 general-purpose register instructions implemented on the POWER6X
19433 processor and other processors that support the extended PowerPC V2.05
19435 The @option{-mhard-dfp} option allows GCC to generate the decimal
19436 floating-point instructions implemented on some POWER processors.
19438 The @option{-mpowerpc64} option allows GCC to generate the additional
19439 64-bit instructions that are found in the full PowerPC64 architecture
19440 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19441 @option{-mno-powerpc64}.
19443 @item -mcpu=@var{cpu_type}
19445 Set architecture type, register usage, and
19446 instruction scheduling parameters for machine type @var{cpu_type}.
19447 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19448 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19449 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19450 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19451 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19452 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19453 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19454 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19455 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19456 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
19457 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
19460 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19461 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19462 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19463 architecture machine types, with an appropriate, generic processor
19464 model assumed for scheduling purposes.
19466 The other options specify a specific processor. Code generated under
19467 those options runs best on that processor, and may not run at all on
19470 The @option{-mcpu} options automatically enable or disable the
19473 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19474 -mpopcntb -mpopcntd -mpowerpc64 @gol
19475 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19476 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19477 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19478 -mquad-memory -mquad-memory-atomic -mmodulo -mfloat128 -mfloat128-hardware @gol
19479 -mpower9-fusion -mpower9-vector}
19481 The particular options set for any particular CPU varies between
19482 compiler versions, depending on what setting seems to produce optimal
19483 code for that CPU; it doesn't necessarily reflect the actual hardware's
19484 capabilities. If you wish to set an individual option to a particular
19485 value, you may specify it after the @option{-mcpu} option, like
19486 @option{-mcpu=970 -mno-altivec}.
19488 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19489 not enabled or disabled by the @option{-mcpu} option at present because
19490 AIX does not have full support for these options. You may still
19491 enable or disable them individually if you're sure it'll work in your
19494 @item -mtune=@var{cpu_type}
19496 Set the instruction scheduling parameters for machine type
19497 @var{cpu_type}, but do not set the architecture type or register usage,
19498 as @option{-mcpu=@var{cpu_type}} does. The same
19499 values for @var{cpu_type} are used for @option{-mtune} as for
19500 @option{-mcpu}. If both are specified, the code generated uses the
19501 architecture and registers set by @option{-mcpu}, but the
19502 scheduling parameters set by @option{-mtune}.
19504 @item -mcmodel=small
19505 @opindex mcmodel=small
19506 Generate PowerPC64 code for the small model: The TOC is limited to
19509 @item -mcmodel=medium
19510 @opindex mcmodel=medium
19511 Generate PowerPC64 code for the medium model: The TOC and other static
19512 data may be up to a total of 4G in size.
19514 @item -mcmodel=large
19515 @opindex mcmodel=large
19516 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19517 in size. Other data and code is only limited by the 64-bit address
19521 @itemx -mno-altivec
19523 @opindex mno-altivec
19524 Generate code that uses (does not use) AltiVec instructions, and also
19525 enable the use of built-in functions that allow more direct access to
19526 the AltiVec instruction set. You may also need to set
19527 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19530 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19531 @option{-maltivec=be}, the element order for Altivec intrinsics such
19532 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19533 match array element order corresponding to the endianness of the
19534 target. That is, element zero identifies the leftmost element in a
19535 vector register when targeting a big-endian platform, and identifies
19536 the rightmost element in a vector register when targeting a
19537 little-endian platform.
19540 @opindex maltivec=be
19541 Generate Altivec instructions using big-endian element order,
19542 regardless of whether the target is big- or little-endian. This is
19543 the default when targeting a big-endian platform.
19545 The element order is used to interpret element numbers in Altivec
19546 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19547 @code{vec_insert}. By default, these match array element order
19548 corresponding to the endianness for the target.
19551 @opindex maltivec=le
19552 Generate Altivec instructions using little-endian element order,
19553 regardless of whether the target is big- or little-endian. This is
19554 the default when targeting a little-endian platform. This option is
19555 currently ignored when targeting a big-endian platform.
19557 The element order is used to interpret element numbers in Altivec
19558 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19559 @code{vec_insert}. By default, these match array element order
19560 corresponding to the endianness for the target.
19565 @opindex mno-vrsave
19566 Generate VRSAVE instructions when generating AltiVec code.
19568 @item -mgen-cell-microcode
19569 @opindex mgen-cell-microcode
19570 Generate Cell microcode instructions.
19572 @item -mwarn-cell-microcode
19573 @opindex mwarn-cell-microcode
19574 Warn when a Cell microcode instruction is emitted. An example
19575 of a Cell microcode instruction is a variable shift.
19578 @opindex msecure-plt
19579 Generate code that allows @command{ld} and @command{ld.so}
19580 to build executables and shared
19581 libraries with non-executable @code{.plt} and @code{.got} sections.
19583 32-bit SYSV ABI option.
19587 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19589 requires @code{.plt} and @code{.got}
19590 sections that are both writable and executable.
19591 This is a PowerPC 32-bit SYSV ABI option.
19597 This switch enables or disables the generation of ISEL instructions.
19599 @item -misel=@var{yes/no}
19600 This switch has been deprecated. Use @option{-misel} and
19601 @option{-mno-isel} instead.
19607 This switch enables or disables the generation of SPE simd
19613 @opindex mno-paired
19614 This switch enables or disables the generation of PAIRED simd
19617 @item -mspe=@var{yes/no}
19618 This option has been deprecated. Use @option{-mspe} and
19619 @option{-mno-spe} instead.
19625 Generate code that uses (does not use) vector/scalar (VSX)
19626 instructions, and also enable the use of built-in functions that allow
19627 more direct access to the VSX instruction set.
19632 @opindex mno-crypto
19633 Enable the use (disable) of the built-in functions that allow direct
19634 access to the cryptographic instructions that were added in version
19635 2.07 of the PowerPC ISA.
19637 @item -mdirect-move
19638 @itemx -mno-direct-move
19639 @opindex mdirect-move
19640 @opindex mno-direct-move
19641 Generate code that uses (does not use) the instructions to move data
19642 between the general purpose registers and the vector/scalar (VSX)
19643 registers that were added in version 2.07 of the PowerPC ISA.
19645 @item -mpower8-fusion
19646 @itemx -mno-power8-fusion
19647 @opindex mpower8-fusion
19648 @opindex mno-power8-fusion
19649 Generate code that keeps (does not keeps) some integer operations
19650 adjacent so that the instructions can be fused together on power8 and
19653 @item -mpower8-vector
19654 @itemx -mno-power8-vector
19655 @opindex mpower8-vector
19656 @opindex mno-power8-vector
19657 Generate code that uses (does not use) the vector and scalar
19658 instructions that were added in version 2.07 of the PowerPC ISA. Also
19659 enable the use of built-in functions that allow more direct access to
19660 the vector instructions.
19662 @item -mquad-memory
19663 @itemx -mno-quad-memory
19664 @opindex mquad-memory
19665 @opindex mno-quad-memory
19666 Generate code that uses (does not use) the non-atomic quad word memory
19667 instructions. The @option{-mquad-memory} option requires use of
19670 @item -mquad-memory-atomic
19671 @itemx -mno-quad-memory-atomic
19672 @opindex mquad-memory-atomic
19673 @opindex mno-quad-memory-atomic
19674 Generate code that uses (does not use) the atomic quad word memory
19675 instructions. The @option{-mquad-memory-atomic} option requires use of
19678 @item -mupper-regs-df
19679 @itemx -mno-upper-regs-df
19680 @opindex mupper-regs-df
19681 @opindex mno-upper-regs-df
19682 Generate code that uses (does not use) the scalar double precision
19683 instructions that target all 64 registers in the vector/scalar
19684 floating point register set that were added in version 2.06 of the
19685 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19686 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19687 @option{-mvsx} options.
19689 @item -mupper-regs-sf
19690 @itemx -mno-upper-regs-sf
19691 @opindex mupper-regs-sf
19692 @opindex mno-upper-regs-sf
19693 Generate code that uses (does not use) the scalar single precision
19694 instructions that target all 64 registers in the vector/scalar
19695 floating point register set that were added in version 2.07 of the
19696 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19697 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19701 @itemx -mno-upper-regs
19702 @opindex mupper-regs
19703 @opindex mno-upper-regs
19704 Generate code that uses (does not use) the scalar
19705 instructions that target all 64 registers in the vector/scalar
19706 floating point register set, depending on the model of the machine.
19708 If the @option{-mno-upper-regs} option is used, it turns off both
19709 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19712 @itemx -mno-float128
19714 @opindex mno-float128
19715 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
19716 and use either software emulation for IEEE 128-bit floating point or
19717 hardware instructions.
19719 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
19720 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
19721 option. The @code{-mfloat128} option only works on PowerPC 64-bit
19724 @item -mfloat128-hardware
19725 @itemx -mno-float128-hardware
19726 @opindex mfloat128-hardware
19727 @opindex mno-float128-hardware
19728 Enable/disable using ISA 3.0 hardware instructions to support the
19729 @var{__float128} data type.
19734 @opindex mno-module
19735 Generate code that uses (does not use) the ISA 3.0 integer modulo
19736 instructions. The @option{-mmodulo} option is enabled by default
19737 with the @option{-mcpu=power9} option.
19739 @item -mpower9-fusion
19740 @itemx -mno-power9-fusion
19741 @opindex mpower9-fusion
19742 @opindex mno-power9-fusion
19743 Generate code that keeps (does not keeps) some operations adjacent so
19744 that the instructions can be fused together on power9 and later
19747 @item -mpower9-vector
19748 @itemx -mno-power9-vector
19749 @opindex mpower9-vector
19750 @opindex mno-power9-vector
19751 Generate code that uses (does not use) the vector and scalar
19752 instructions that were added in version 2.07 of the PowerPC ISA. Also
19753 enable the use of built-in functions that allow more direct access to
19754 the vector instructions.
19756 @item -mfloat-gprs=@var{yes/single/double/no}
19757 @itemx -mfloat-gprs
19758 @opindex mfloat-gprs
19759 This switch enables or disables the generation of floating-point
19760 operations on the general-purpose registers for architectures that
19763 The argument @samp{yes} or @samp{single} enables the use of
19764 single-precision floating-point operations.
19766 The argument @samp{double} enables the use of single and
19767 double-precision floating-point operations.
19769 The argument @samp{no} disables floating-point operations on the
19770 general-purpose registers.
19772 This option is currently only available on the MPC854x.
19778 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19779 targets (including GNU/Linux). The 32-bit environment sets int, long
19780 and pointer to 32 bits and generates code that runs on any PowerPC
19781 variant. The 64-bit environment sets int to 32 bits and long and
19782 pointer to 64 bits, and generates code for PowerPC64, as for
19783 @option{-mpowerpc64}.
19786 @itemx -mno-fp-in-toc
19787 @itemx -mno-sum-in-toc
19788 @itemx -mminimal-toc
19790 @opindex mno-fp-in-toc
19791 @opindex mno-sum-in-toc
19792 @opindex mminimal-toc
19793 Modify generation of the TOC (Table Of Contents), which is created for
19794 every executable file. The @option{-mfull-toc} option is selected by
19795 default. In that case, GCC allocates at least one TOC entry for
19796 each unique non-automatic variable reference in your program. GCC
19797 also places floating-point constants in the TOC@. However, only
19798 16,384 entries are available in the TOC@.
19800 If you receive a linker error message that saying you have overflowed
19801 the available TOC space, you can reduce the amount of TOC space used
19802 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19803 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19804 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19805 generate code to calculate the sum of an address and a constant at
19806 run time instead of putting that sum into the TOC@. You may specify one
19807 or both of these options. Each causes GCC to produce very slightly
19808 slower and larger code at the expense of conserving TOC space.
19810 If you still run out of space in the TOC even when you specify both of
19811 these options, specify @option{-mminimal-toc} instead. This option causes
19812 GCC to make only one TOC entry for every file. When you specify this
19813 option, GCC produces code that is slower and larger but which
19814 uses extremely little TOC space. You may wish to use this option
19815 only on files that contain less frequently-executed code.
19821 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19822 @code{long} type, and the infrastructure needed to support them.
19823 Specifying @option{-maix64} implies @option{-mpowerpc64},
19824 while @option{-maix32} disables the 64-bit ABI and
19825 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19828 @itemx -mno-xl-compat
19829 @opindex mxl-compat
19830 @opindex mno-xl-compat
19831 Produce code that conforms more closely to IBM XL compiler semantics
19832 when using AIX-compatible ABI@. Pass floating-point arguments to
19833 prototyped functions beyond the register save area (RSA) on the stack
19834 in addition to argument FPRs. Do not assume that most significant
19835 double in 128-bit long double value is properly rounded when comparing
19836 values and converting to double. Use XL symbol names for long double
19839 The AIX calling convention was extended but not initially documented to
19840 handle an obscure K&R C case of calling a function that takes the
19841 address of its arguments with fewer arguments than declared. IBM XL
19842 compilers access floating-point arguments that do not fit in the
19843 RSA from the stack when a subroutine is compiled without
19844 optimization. Because always storing floating-point arguments on the
19845 stack is inefficient and rarely needed, this option is not enabled by
19846 default and only is necessary when calling subroutines compiled by IBM
19847 XL compilers without optimization.
19851 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19852 application written to use message passing with special startup code to
19853 enable the application to run. The system must have PE installed in the
19854 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19855 must be overridden with the @option{-specs=} option to specify the
19856 appropriate directory location. The Parallel Environment does not
19857 support threads, so the @option{-mpe} option and the @option{-pthread}
19858 option are incompatible.
19860 @item -malign-natural
19861 @itemx -malign-power
19862 @opindex malign-natural
19863 @opindex malign-power
19864 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19865 @option{-malign-natural} overrides the ABI-defined alignment of larger
19866 types, such as floating-point doubles, on their natural size-based boundary.
19867 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19868 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19870 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19874 @itemx -mhard-float
19875 @opindex msoft-float
19876 @opindex mhard-float
19877 Generate code that does not use (uses) the floating-point register set.
19878 Software floating-point emulation is provided if you use the
19879 @option{-msoft-float} option, and pass the option to GCC when linking.
19881 @item -msingle-float
19882 @itemx -mdouble-float
19883 @opindex msingle-float
19884 @opindex mdouble-float
19885 Generate code for single- or double-precision floating-point operations.
19886 @option{-mdouble-float} implies @option{-msingle-float}.
19889 @opindex msimple-fpu
19890 Do not generate @code{sqrt} and @code{div} instructions for hardware
19891 floating-point unit.
19893 @item -mfpu=@var{name}
19895 Specify type of floating-point unit. Valid values for @var{name} are
19896 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19897 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19898 @samp{sp_full} (equivalent to @option{-msingle-float}),
19899 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19902 @opindex mxilinx-fpu
19903 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19906 @itemx -mno-multiple
19908 @opindex mno-multiple
19909 Generate code that uses (does not use) the load multiple word
19910 instructions and the store multiple word instructions. These
19911 instructions are generated by default on POWER systems, and not
19912 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19913 PowerPC systems, since those instructions do not work when the
19914 processor is in little-endian mode. The exceptions are PPC740 and
19915 PPC750 which permit these instructions in little-endian mode.
19920 @opindex mno-string
19921 Generate code that uses (does not use) the load string instructions
19922 and the store string word instructions to save multiple registers and
19923 do small block moves. These instructions are generated by default on
19924 POWER systems, and not generated on PowerPC systems. Do not use
19925 @option{-mstring} on little-endian PowerPC systems, since those
19926 instructions do not work when the processor is in little-endian mode.
19927 The exceptions are PPC740 and PPC750 which permit these instructions
19928 in little-endian mode.
19933 @opindex mno-update
19934 Generate code that uses (does not use) the load or store instructions
19935 that update the base register to the address of the calculated memory
19936 location. These instructions are generated by default. If you use
19937 @option{-mno-update}, there is a small window between the time that the
19938 stack pointer is updated and the address of the previous frame is
19939 stored, which means code that walks the stack frame across interrupts or
19940 signals may get corrupted data.
19942 @item -mavoid-indexed-addresses
19943 @itemx -mno-avoid-indexed-addresses
19944 @opindex mavoid-indexed-addresses
19945 @opindex mno-avoid-indexed-addresses
19946 Generate code that tries to avoid (not avoid) the use of indexed load
19947 or store instructions. These instructions can incur a performance
19948 penalty on Power6 processors in certain situations, such as when
19949 stepping through large arrays that cross a 16M boundary. This option
19950 is enabled by default when targeting Power6 and disabled otherwise.
19953 @itemx -mno-fused-madd
19954 @opindex mfused-madd
19955 @opindex mno-fused-madd
19956 Generate code that uses (does not use) the floating-point multiply and
19957 accumulate instructions. These instructions are generated by default
19958 if hardware floating point is used. The machine-dependent
19959 @option{-mfused-madd} option is now mapped to the machine-independent
19960 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19961 mapped to @option{-ffp-contract=off}.
19967 Generate code that uses (does not use) the half-word multiply and
19968 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19969 These instructions are generated by default when targeting those
19976 Generate code that uses (does not use) the string-search @samp{dlmzb}
19977 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19978 generated by default when targeting those processors.
19980 @item -mno-bit-align
19982 @opindex mno-bit-align
19983 @opindex mbit-align
19984 On System V.4 and embedded PowerPC systems do not (do) force structures
19985 and unions that contain bit-fields to be aligned to the base type of the
19988 For example, by default a structure containing nothing but 8
19989 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19990 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19991 the structure is aligned to a 1-byte boundary and is 1 byte in
19994 @item -mno-strict-align
19995 @itemx -mstrict-align
19996 @opindex mno-strict-align
19997 @opindex mstrict-align
19998 On System V.4 and embedded PowerPC systems do not (do) assume that
19999 unaligned memory references are handled by the system.
20001 @item -mrelocatable
20002 @itemx -mno-relocatable
20003 @opindex mrelocatable
20004 @opindex mno-relocatable
20005 Generate code that allows (does not allow) a static executable to be
20006 relocated to a different address at run time. A simple embedded
20007 PowerPC system loader should relocate the entire contents of
20008 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20009 a table of 32-bit addresses generated by this option. For this to
20010 work, all objects linked together must be compiled with
20011 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20012 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20014 @item -mrelocatable-lib
20015 @itemx -mno-relocatable-lib
20016 @opindex mrelocatable-lib
20017 @opindex mno-relocatable-lib
20018 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20019 @code{.fixup} section to allow static executables to be relocated at
20020 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20021 alignment of @option{-mrelocatable}. Objects compiled with
20022 @option{-mrelocatable-lib} may be linked with objects compiled with
20023 any combination of the @option{-mrelocatable} options.
20029 On System V.4 and embedded PowerPC systems do not (do) assume that
20030 register 2 contains a pointer to a global area pointing to the addresses
20031 used in the program.
20034 @itemx -mlittle-endian
20036 @opindex mlittle-endian
20037 On System V.4 and embedded PowerPC systems compile code for the
20038 processor in little-endian mode. The @option{-mlittle-endian} option is
20039 the same as @option{-mlittle}.
20042 @itemx -mbig-endian
20044 @opindex mbig-endian
20045 On System V.4 and embedded PowerPC systems compile code for the
20046 processor in big-endian mode. The @option{-mbig-endian} option is
20047 the same as @option{-mbig}.
20049 @item -mdynamic-no-pic
20050 @opindex mdynamic-no-pic
20051 On Darwin and Mac OS X systems, compile code so that it is not
20052 relocatable, but that its external references are relocatable. The
20053 resulting code is suitable for applications, but not shared
20056 @item -msingle-pic-base
20057 @opindex msingle-pic-base
20058 Treat the register used for PIC addressing as read-only, rather than
20059 loading it in the prologue for each function. The runtime system is
20060 responsible for initializing this register with an appropriate value
20061 before execution begins.
20063 @item -mprioritize-restricted-insns=@var{priority}
20064 @opindex mprioritize-restricted-insns
20065 This option controls the priority that is assigned to
20066 dispatch-slot restricted instructions during the second scheduling
20067 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20068 or @samp{2} to assign no, highest, or second-highest (respectively)
20069 priority to dispatch-slot restricted
20072 @item -msched-costly-dep=@var{dependence_type}
20073 @opindex msched-costly-dep
20074 This option controls which dependences are considered costly
20075 by the target during instruction scheduling. The argument
20076 @var{dependence_type} takes one of the following values:
20080 No dependence is costly.
20083 All dependences are costly.
20085 @item @samp{true_store_to_load}
20086 A true dependence from store to load is costly.
20088 @item @samp{store_to_load}
20089 Any dependence from store to load is costly.
20092 Any dependence for which the latency is greater than or equal to
20093 @var{number} is costly.
20096 @item -minsert-sched-nops=@var{scheme}
20097 @opindex minsert-sched-nops
20098 This option controls which NOP insertion scheme is used during
20099 the second scheduling pass. The argument @var{scheme} takes one of the
20107 Pad with NOPs any dispatch group that has vacant issue slots,
20108 according to the scheduler's grouping.
20110 @item @samp{regroup_exact}
20111 Insert NOPs to force costly dependent insns into
20112 separate groups. Insert exactly as many NOPs as needed to force an insn
20113 to a new group, according to the estimated processor grouping.
20116 Insert NOPs to force costly dependent insns into
20117 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20121 @opindex mcall-sysv
20122 On System V.4 and embedded PowerPC systems compile code using calling
20123 conventions that adhere to the March 1995 draft of the System V
20124 Application Binary Interface, PowerPC processor supplement. This is the
20125 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20127 @item -mcall-sysv-eabi
20129 @opindex mcall-sysv-eabi
20130 @opindex mcall-eabi
20131 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20133 @item -mcall-sysv-noeabi
20134 @opindex mcall-sysv-noeabi
20135 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20137 @item -mcall-aixdesc
20139 On System V.4 and embedded PowerPC systems compile code for the AIX
20143 @opindex mcall-linux
20144 On System V.4 and embedded PowerPC systems compile code for the
20145 Linux-based GNU system.
20147 @item -mcall-freebsd
20148 @opindex mcall-freebsd
20149 On System V.4 and embedded PowerPC systems compile code for the
20150 FreeBSD operating system.
20152 @item -mcall-netbsd
20153 @opindex mcall-netbsd
20154 On System V.4 and embedded PowerPC systems compile code for the
20155 NetBSD operating system.
20157 @item -mcall-openbsd
20158 @opindex mcall-netbsd
20159 On System V.4 and embedded PowerPC systems compile code for the
20160 OpenBSD operating system.
20162 @item -maix-struct-return
20163 @opindex maix-struct-return
20164 Return all structures in memory (as specified by the AIX ABI)@.
20166 @item -msvr4-struct-return
20167 @opindex msvr4-struct-return
20168 Return structures smaller than 8 bytes in registers (as specified by the
20171 @item -mabi=@var{abi-type}
20173 Extend the current ABI with a particular extension, or remove such extension.
20174 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20175 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20176 @samp{elfv1}, @samp{elfv2}@.
20180 Extend the current ABI with SPE ABI extensions. This does not change
20181 the default ABI, instead it adds the SPE ABI extensions to the current
20185 @opindex mabi=no-spe
20186 Disable Book-E SPE ABI extensions for the current ABI@.
20188 @item -mabi=ibmlongdouble
20189 @opindex mabi=ibmlongdouble
20190 Change the current ABI to use IBM extended-precision long double.
20191 This is a PowerPC 32-bit SYSV ABI option.
20193 @item -mabi=ieeelongdouble
20194 @opindex mabi=ieeelongdouble
20195 Change the current ABI to use IEEE extended-precision long double.
20196 This is a PowerPC 32-bit Linux ABI option.
20199 @opindex mabi=elfv1
20200 Change the current ABI to use the ELFv1 ABI.
20201 This is the default ABI for big-endian PowerPC 64-bit Linux.
20202 Overriding the default ABI requires special system support and is
20203 likely to fail in spectacular ways.
20206 @opindex mabi=elfv2
20207 Change the current ABI to use the ELFv2 ABI.
20208 This is the default ABI for little-endian PowerPC 64-bit Linux.
20209 Overriding the default ABI requires special system support and is
20210 likely to fail in spectacular ways.
20213 @itemx -mno-prototype
20214 @opindex mprototype
20215 @opindex mno-prototype
20216 On System V.4 and embedded PowerPC systems assume that all calls to
20217 variable argument functions are properly prototyped. Otherwise, the
20218 compiler must insert an instruction before every non-prototyped call to
20219 set or clear bit 6 of the condition code register (@code{CR}) to
20220 indicate whether floating-point values are passed in the floating-point
20221 registers in case the function takes variable arguments. With
20222 @option{-mprototype}, only calls to prototyped variable argument functions
20223 set or clear the bit.
20227 On embedded PowerPC systems, assume that the startup module is called
20228 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20229 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20234 On embedded PowerPC systems, assume that the startup module is called
20235 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20240 On embedded PowerPC systems, assume that the startup module is called
20241 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20244 @item -myellowknife
20245 @opindex myellowknife
20246 On embedded PowerPC systems, assume that the startup module is called
20247 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20252 On System V.4 and embedded PowerPC systems, specify that you are
20253 compiling for a VxWorks system.
20257 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20258 header to indicate that @samp{eabi} extended relocations are used.
20264 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20265 Embedded Applications Binary Interface (EABI), which is a set of
20266 modifications to the System V.4 specifications. Selecting @option{-meabi}
20267 means that the stack is aligned to an 8-byte boundary, a function
20268 @code{__eabi} is called from @code{main} to set up the EABI
20269 environment, and the @option{-msdata} option can use both @code{r2} and
20270 @code{r13} to point to two separate small data areas. Selecting
20271 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20272 no EABI initialization function is called from @code{main}, and the
20273 @option{-msdata} option only uses @code{r13} to point to a single
20274 small data area. The @option{-meabi} option is on by default if you
20275 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20278 @opindex msdata=eabi
20279 On System V.4 and embedded PowerPC systems, put small initialized
20280 @code{const} global and static data in the @code{.sdata2} section, which
20281 is pointed to by register @code{r2}. Put small initialized
20282 non-@code{const} global and static data in the @code{.sdata} section,
20283 which is pointed to by register @code{r13}. Put small uninitialized
20284 global and static data in the @code{.sbss} section, which is adjacent to
20285 the @code{.sdata} section. The @option{-msdata=eabi} option is
20286 incompatible with the @option{-mrelocatable} option. The
20287 @option{-msdata=eabi} option also sets the @option{-memb} option.
20290 @opindex msdata=sysv
20291 On System V.4 and embedded PowerPC systems, put small global and static
20292 data in the @code{.sdata} section, which is pointed to by register
20293 @code{r13}. Put small uninitialized global and static data in the
20294 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20295 The @option{-msdata=sysv} option is incompatible with the
20296 @option{-mrelocatable} option.
20298 @item -msdata=default
20300 @opindex msdata=default
20302 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20303 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20304 same as @option{-msdata=sysv}.
20307 @opindex msdata=data
20308 On System V.4 and embedded PowerPC systems, put small global
20309 data in the @code{.sdata} section. Put small uninitialized global
20310 data in the @code{.sbss} section. Do not use register @code{r13}
20311 to address small data however. This is the default behavior unless
20312 other @option{-msdata} options are used.
20316 @opindex msdata=none
20318 On embedded PowerPC systems, put all initialized global and static data
20319 in the @code{.data} section, and all uninitialized data in the
20320 @code{.bss} section.
20322 @item -mblock-move-inline-limit=@var{num}
20323 @opindex mblock-move-inline-limit
20324 Inline all block moves (such as calls to @code{memcpy} or structure
20325 copies) less than or equal to @var{num} bytes. The minimum value for
20326 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20327 targets. The default value is target-specific.
20331 @cindex smaller data references (PowerPC)
20332 @cindex .sdata/.sdata2 references (PowerPC)
20333 On embedded PowerPC systems, put global and static items less than or
20334 equal to @var{num} bytes into the small data or BSS sections instead of
20335 the normal data or BSS section. By default, @var{num} is 8. The
20336 @option{-G @var{num}} switch is also passed to the linker.
20337 All modules should be compiled with the same @option{-G @var{num}} value.
20340 @itemx -mno-regnames
20342 @opindex mno-regnames
20343 On System V.4 and embedded PowerPC systems do (do not) emit register
20344 names in the assembly language output using symbolic forms.
20347 @itemx -mno-longcall
20349 @opindex mno-longcall
20350 By default assume that all calls are far away so that a longer and more
20351 expensive calling sequence is required. This is required for calls
20352 farther than 32 megabytes (33,554,432 bytes) from the current location.
20353 A short call is generated if the compiler knows
20354 the call cannot be that far away. This setting can be overridden by
20355 the @code{shortcall} function attribute, or by @code{#pragma
20358 Some linkers are capable of detecting out-of-range calls and generating
20359 glue code on the fly. On these systems, long calls are unnecessary and
20360 generate slower code. As of this writing, the AIX linker can do this,
20361 as can the GNU linker for PowerPC/64. It is planned to add this feature
20362 to the GNU linker for 32-bit PowerPC systems as well.
20364 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20365 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20366 addresses represent the callee and the branch island. The
20367 Darwin/PPC linker prefers the first address and generates a @code{bl
20368 callee} if the PPC @code{bl} instruction reaches the callee directly;
20369 otherwise, the linker generates @code{bl L42} to call the branch
20370 island. The branch island is appended to the body of the
20371 calling function; it computes the full 32-bit address of the callee
20374 On Mach-O (Darwin) systems, this option directs the compiler emit to
20375 the glue for every direct call, and the Darwin linker decides whether
20376 to use or discard it.
20378 In the future, GCC may ignore all longcall specifications
20379 when the linker is known to generate glue.
20381 @item -mtls-markers
20382 @itemx -mno-tls-markers
20383 @opindex mtls-markers
20384 @opindex mno-tls-markers
20385 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20386 specifying the function argument. The relocation allows the linker to
20387 reliably associate function call with argument setup instructions for
20388 TLS optimization, which in turn allows GCC to better schedule the
20393 Adds support for multithreading with the @dfn{pthreads} library.
20394 This option sets flags for both the preprocessor and linker.
20399 This option enables use of the reciprocal estimate and
20400 reciprocal square root estimate instructions with additional
20401 Newton-Raphson steps to increase precision instead of doing a divide or
20402 square root and divide for floating-point arguments. You should use
20403 the @option{-ffast-math} option when using @option{-mrecip} (or at
20404 least @option{-funsafe-math-optimizations},
20405 @option{-ffinite-math-only}, @option{-freciprocal-math} and
20406 @option{-fno-trapping-math}). Note that while the throughput of the
20407 sequence is generally higher than the throughput of the non-reciprocal
20408 instruction, the precision of the sequence can be decreased by up to 2
20409 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20412 @item -mrecip=@var{opt}
20413 @opindex mrecip=opt
20414 This option controls which reciprocal estimate instructions
20415 may be used. @var{opt} is a comma-separated list of options, which may
20416 be preceded by a @code{!} to invert the option:
20421 Enable all estimate instructions.
20424 Enable the default instructions, equivalent to @option{-mrecip}.
20427 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20430 Enable the reciprocal approximation instructions for both
20431 single and double precision.
20434 Enable the single-precision reciprocal approximation instructions.
20437 Enable the double-precision reciprocal approximation instructions.
20440 Enable the reciprocal square root approximation instructions for both
20441 single and double precision.
20444 Enable the single-precision reciprocal square root approximation instructions.
20447 Enable the double-precision reciprocal square root approximation instructions.
20451 So, for example, @option{-mrecip=all,!rsqrtd} enables
20452 all of the reciprocal estimate instructions, except for the
20453 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20454 which handle the double-precision reciprocal square root calculations.
20456 @item -mrecip-precision
20457 @itemx -mno-recip-precision
20458 @opindex mrecip-precision
20459 Assume (do not assume) that the reciprocal estimate instructions
20460 provide higher-precision estimates than is mandated by the PowerPC
20461 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20462 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20463 The double-precision square root estimate instructions are not generated by
20464 default on low-precision machines, since they do not provide an
20465 estimate that converges after three steps.
20467 @item -mveclibabi=@var{type}
20468 @opindex mveclibabi
20469 Specifies the ABI type to use for vectorizing intrinsics using an
20470 external library. The only type supported at present is @samp{mass},
20471 which specifies to use IBM's Mathematical Acceleration Subsystem
20472 (MASS) libraries for vectorizing intrinsics using external libraries.
20473 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20474 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20475 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20476 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20477 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20478 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20479 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20480 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20481 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20482 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20483 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20484 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20485 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20486 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20487 for power7. Both @option{-ftree-vectorize} and
20488 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20489 libraries must be specified at link time.
20494 Generate (do not generate) the @code{friz} instruction when the
20495 @option{-funsafe-math-optimizations} option is used to optimize
20496 rounding of floating-point values to 64-bit integer and back to floating
20497 point. The @code{friz} instruction does not return the same value if
20498 the floating-point number is too large to fit in an integer.
20500 @item -mpointers-to-nested-functions
20501 @itemx -mno-pointers-to-nested-functions
20502 @opindex mpointers-to-nested-functions
20503 Generate (do not generate) code to load up the static chain register
20504 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20505 systems where a function pointer points to a 3-word descriptor giving
20506 the function address, TOC value to be loaded in register @code{r2}, and
20507 static chain value to be loaded in register @code{r11}. The
20508 @option{-mpointers-to-nested-functions} is on by default. You cannot
20509 call through pointers to nested functions or pointers
20510 to functions compiled in other languages that use the static chain if
20511 you use @option{-mno-pointers-to-nested-functions}.
20513 @item -msave-toc-indirect
20514 @itemx -mno-save-toc-indirect
20515 @opindex msave-toc-indirect
20516 Generate (do not generate) code to save the TOC value in the reserved
20517 stack location in the function prologue if the function calls through
20518 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20519 saved in the prologue, it is saved just before the call through the
20520 pointer. The @option{-mno-save-toc-indirect} option is the default.
20522 @item -mcompat-align-parm
20523 @itemx -mno-compat-align-parm
20524 @opindex mcompat-align-parm
20525 Generate (do not generate) code to pass structure parameters with a
20526 maximum alignment of 64 bits, for compatibility with older versions
20529 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20530 structure parameter on a 128-bit boundary when that structure contained
20531 a member requiring 128-bit alignment. This is corrected in more
20532 recent versions of GCC. This option may be used to generate code
20533 that is compatible with functions compiled with older versions of
20536 The @option{-mno-compat-align-parm} option is the default.
20540 @subsection RX Options
20543 These command-line options are defined for RX targets:
20546 @item -m64bit-doubles
20547 @itemx -m32bit-doubles
20548 @opindex m64bit-doubles
20549 @opindex m32bit-doubles
20550 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20551 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20552 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20553 works on 32-bit values, which is why the default is
20554 @option{-m32bit-doubles}.
20560 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20561 floating-point hardware. The default is enabled for the RX600
20562 series and disabled for the RX200 series.
20564 Floating-point instructions are only generated for 32-bit floating-point
20565 values, however, so the FPU hardware is not used for doubles if the
20566 @option{-m64bit-doubles} option is used.
20568 @emph{Note} If the @option{-fpu} option is enabled then
20569 @option{-funsafe-math-optimizations} is also enabled automatically.
20570 This is because the RX FPU instructions are themselves unsafe.
20572 @item -mcpu=@var{name}
20574 Selects the type of RX CPU to be targeted. Currently three types are
20575 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20576 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20578 The only difference between @samp{RX600} and @samp{RX610} is that the
20579 @samp{RX610} does not support the @code{MVTIPL} instruction.
20581 The @samp{RX200} series does not have a hardware floating-point unit
20582 and so @option{-nofpu} is enabled by default when this type is
20585 @item -mbig-endian-data
20586 @itemx -mlittle-endian-data
20587 @opindex mbig-endian-data
20588 @opindex mlittle-endian-data
20589 Store data (but not code) in the big-endian format. The default is
20590 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20593 @item -msmall-data-limit=@var{N}
20594 @opindex msmall-data-limit
20595 Specifies the maximum size in bytes of global and static variables
20596 which can be placed into the small data area. Using the small data
20597 area can lead to smaller and faster code, but the size of area is
20598 limited and it is up to the programmer to ensure that the area does
20599 not overflow. Also when the small data area is used one of the RX's
20600 registers (usually @code{r13}) is reserved for use pointing to this
20601 area, so it is no longer available for use by the compiler. This
20602 could result in slower and/or larger code if variables are pushed onto
20603 the stack instead of being held in this register.
20605 Note, common variables (variables that have not been initialized) and
20606 constants are not placed into the small data area as they are assigned
20607 to other sections in the output executable.
20609 The default value is zero, which disables this feature. Note, this
20610 feature is not enabled by default with higher optimization levels
20611 (@option{-O2} etc) because of the potentially detrimental effects of
20612 reserving a register. It is up to the programmer to experiment and
20613 discover whether this feature is of benefit to their program. See the
20614 description of the @option{-mpid} option for a description of how the
20615 actual register to hold the small data area pointer is chosen.
20621 Use the simulator runtime. The default is to use the libgloss
20622 board-specific runtime.
20624 @item -mas100-syntax
20625 @itemx -mno-as100-syntax
20626 @opindex mas100-syntax
20627 @opindex mno-as100-syntax
20628 When generating assembler output use a syntax that is compatible with
20629 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20630 assembler, but it has some restrictions so it is not generated by default.
20632 @item -mmax-constant-size=@var{N}
20633 @opindex mmax-constant-size
20634 Specifies the maximum size, in bytes, of a constant that can be used as
20635 an operand in a RX instruction. Although the RX instruction set does
20636 allow constants of up to 4 bytes in length to be used in instructions,
20637 a longer value equates to a longer instruction. Thus in some
20638 circumstances it can be beneficial to restrict the size of constants
20639 that are used in instructions. Constants that are too big are instead
20640 placed into a constant pool and referenced via register indirection.
20642 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20643 or 4 means that constants of any size are allowed.
20647 Enable linker relaxation. Linker relaxation is a process whereby the
20648 linker attempts to reduce the size of a program by finding shorter
20649 versions of various instructions. Disabled by default.
20651 @item -mint-register=@var{N}
20652 @opindex mint-register
20653 Specify the number of registers to reserve for fast interrupt handler
20654 functions. The value @var{N} can be between 0 and 4. A value of 1
20655 means that register @code{r13} is reserved for the exclusive use
20656 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20657 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20658 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20659 A value of 0, the default, does not reserve any registers.
20661 @item -msave-acc-in-interrupts
20662 @opindex msave-acc-in-interrupts
20663 Specifies that interrupt handler functions should preserve the
20664 accumulator register. This is only necessary if normal code might use
20665 the accumulator register, for example because it performs 64-bit
20666 multiplications. The default is to ignore the accumulator as this
20667 makes the interrupt handlers faster.
20673 Enables the generation of position independent data. When enabled any
20674 access to constant data is done via an offset from a base address
20675 held in a register. This allows the location of constant data to be
20676 determined at run time without requiring the executable to be
20677 relocated, which is a benefit to embedded applications with tight
20678 memory constraints. Data that can be modified is not affected by this
20681 Note, using this feature reserves a register, usually @code{r13}, for
20682 the constant data base address. This can result in slower and/or
20683 larger code, especially in complicated functions.
20685 The actual register chosen to hold the constant data base address
20686 depends upon whether the @option{-msmall-data-limit} and/or the
20687 @option{-mint-register} command-line options are enabled. Starting
20688 with register @code{r13} and proceeding downwards, registers are
20689 allocated first to satisfy the requirements of @option{-mint-register},
20690 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20691 is possible for the small data area register to be @code{r8} if both
20692 @option{-mint-register=4} and @option{-mpid} are specified on the
20695 By default this feature is not enabled. The default can be restored
20696 via the @option{-mno-pid} command-line option.
20698 @item -mno-warn-multiple-fast-interrupts
20699 @itemx -mwarn-multiple-fast-interrupts
20700 @opindex mno-warn-multiple-fast-interrupts
20701 @opindex mwarn-multiple-fast-interrupts
20702 Prevents GCC from issuing a warning message if it finds more than one
20703 fast interrupt handler when it is compiling a file. The default is to
20704 issue a warning for each extra fast interrupt handler found, as the RX
20705 only supports one such interrupt.
20707 @item -mallow-string-insns
20708 @itemx -mno-allow-string-insns
20709 @opindex mallow-string-insns
20710 @opindex mno-allow-string-insns
20711 Enables or disables the use of the string manipulation instructions
20712 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20713 @code{SWHILE} and also the @code{RMPA} instruction. These
20714 instructions may prefetch data, which is not safe to do if accessing
20715 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20716 for more information).
20718 The default is to allow these instructions, but it is not possible for
20719 GCC to reliably detect all circumstances where a string instruction
20720 might be used to access an I/O register, so their use cannot be
20721 disabled automatically. Instead it is reliant upon the programmer to
20722 use the @option{-mno-allow-string-insns} option if their program
20723 accesses I/O space.
20725 When the instructions are enabled GCC defines the C preprocessor
20726 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20727 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20733 Use only (or not only) @code{JSR} instructions to access functions.
20734 This option can be used when code size exceeds the range of @code{BSR}
20735 instructions. Note that @option{-mno-jsr} does not mean to not use
20736 @code{JSR} but instead means that any type of branch may be used.
20739 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20740 has special significance to the RX port when used with the
20741 @code{interrupt} function attribute. This attribute indicates a
20742 function intended to process fast interrupts. GCC ensures
20743 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20744 and/or @code{r13} and only provided that the normal use of the
20745 corresponding registers have been restricted via the
20746 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20749 @node S/390 and zSeries Options
20750 @subsection S/390 and zSeries Options
20751 @cindex S/390 and zSeries Options
20753 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20757 @itemx -msoft-float
20758 @opindex mhard-float
20759 @opindex msoft-float
20760 Use (do not use) the hardware floating-point instructions and registers
20761 for floating-point operations. When @option{-msoft-float} is specified,
20762 functions in @file{libgcc.a} are used to perform floating-point
20763 operations. When @option{-mhard-float} is specified, the compiler
20764 generates IEEE floating-point instructions. This is the default.
20767 @itemx -mno-hard-dfp
20769 @opindex mno-hard-dfp
20770 Use (do not use) the hardware decimal-floating-point instructions for
20771 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20772 specified, functions in @file{libgcc.a} are used to perform
20773 decimal-floating-point operations. When @option{-mhard-dfp} is
20774 specified, the compiler generates decimal-floating-point hardware
20775 instructions. This is the default for @option{-march=z9-ec} or higher.
20777 @item -mlong-double-64
20778 @itemx -mlong-double-128
20779 @opindex mlong-double-64
20780 @opindex mlong-double-128
20781 These switches control the size of @code{long double} type. A size
20782 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20783 type. This is the default.
20786 @itemx -mno-backchain
20787 @opindex mbackchain
20788 @opindex mno-backchain
20789 Store (do not store) the address of the caller's frame as backchain pointer
20790 into the callee's stack frame.
20791 A backchain may be needed to allow debugging using tools that do not understand
20792 DWARF 2 call frame information.
20793 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20794 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20795 the backchain is placed into the topmost word of the 96/160 byte register
20798 In general, code compiled with @option{-mbackchain} is call-compatible with
20799 code compiled with @option{-mmo-backchain}; however, use of the backchain
20800 for debugging purposes usually requires that the whole binary is built with
20801 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20802 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20803 to build a linux kernel use @option{-msoft-float}.
20805 The default is to not maintain the backchain.
20807 @item -mpacked-stack
20808 @itemx -mno-packed-stack
20809 @opindex mpacked-stack
20810 @opindex mno-packed-stack
20811 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20812 specified, the compiler uses the all fields of the 96/160 byte register save
20813 area only for their default purpose; unused fields still take up stack space.
20814 When @option{-mpacked-stack} is specified, register save slots are densely
20815 packed at the top of the register save area; unused space is reused for other
20816 purposes, allowing for more efficient use of the available stack space.
20817 However, when @option{-mbackchain} is also in effect, the topmost word of
20818 the save area is always used to store the backchain, and the return address
20819 register is always saved two words below the backchain.
20821 As long as the stack frame backchain is not used, code generated with
20822 @option{-mpacked-stack} is call-compatible with code generated with
20823 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20824 S/390 or zSeries generated code that uses the stack frame backchain at run
20825 time, not just for debugging purposes. Such code is not call-compatible
20826 with code compiled with @option{-mpacked-stack}. Also, note that the
20827 combination of @option{-mbackchain},
20828 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20829 to build a linux kernel use @option{-msoft-float}.
20831 The default is to not use the packed stack layout.
20834 @itemx -mno-small-exec
20835 @opindex msmall-exec
20836 @opindex mno-small-exec
20837 Generate (or do not generate) code using the @code{bras} instruction
20838 to do subroutine calls.
20839 This only works reliably if the total executable size does not
20840 exceed 64k. The default is to use the @code{basr} instruction instead,
20841 which does not have this limitation.
20847 When @option{-m31} is specified, generate code compliant to the
20848 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20849 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20850 particular to generate 64-bit instructions. For the @samp{s390}
20851 targets, the default is @option{-m31}, while the @samp{s390x}
20852 targets default to @option{-m64}.
20858 When @option{-mzarch} is specified, generate code using the
20859 instructions available on z/Architecture.
20860 When @option{-mesa} is specified, generate code using the
20861 instructions available on ESA/390. Note that @option{-mesa} is
20862 not possible with @option{-m64}.
20863 When generating code compliant to the GNU/Linux for S/390 ABI,
20864 the default is @option{-mesa}. When generating code compliant
20865 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20871 The @option{-mhtm} option enables a set of builtins making use of
20872 instructions available with the transactional execution facility
20873 introduced with the IBM zEnterprise EC12 machine generation
20874 @ref{S/390 System z Built-in Functions}.
20875 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
20881 When @option{-mvx} is specified, generate code using the instructions
20882 available with the vector extension facility introduced with the IBM
20883 z13 machine generation.
20884 This option changes the ABI for some vector type values with regard to
20885 alignment and calling conventions. In case vector type values are
20886 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
20887 command will be added to mark the resulting binary with the ABI used.
20888 @option{-mvx} is enabled by default when using @option{-march=z13}.
20891 @itemx -mno-zvector
20893 @opindex mno-zvector
20894 The @option{-mzvector} option enables vector language extensions and
20895 builtins using instructions available with the vector extension
20896 facility introduced with the IBM z13 machine generation.
20897 This option adds support for @samp{vector} to be used as a keyword to
20898 define vector type variables and arguments. @samp{vector} is only
20899 available when GNU extensions are enabled. It will not be expanded
20900 when requesting strict standard compliance e.g. with @option{-std=c99}.
20901 In addition to the GCC low-level builtins @option{-mzvector} enables
20902 a set of builtins added for compatibility with Altivec-style
20903 implementations like Power and Cell. In order to make use of these
20904 builtins the header file @file{vecintrin.h} needs to be included.
20905 @option{-mzvector} is disabled by default.
20911 Generate (or do not generate) code using the @code{mvcle} instruction
20912 to perform block moves. When @option{-mno-mvcle} is specified,
20913 use a @code{mvc} loop instead. This is the default unless optimizing for
20920 Print (or do not print) additional debug information when compiling.
20921 The default is to not print debug information.
20923 @item -march=@var{cpu-type}
20925 Generate code that runs on @var{cpu-type}, which is the name of a
20926 system representing a certain processor type. Possible values for
20927 @var{cpu-type} are @samp{z900}, @samp{z990}, @samp{z9-109},
20928 @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12}, and @samp{z13}.
20929 The default is @option{-march=z900}. @samp{g5} and @samp{g6} are
20930 deprecated and will be removed with future releases.
20932 @item -mtune=@var{cpu-type}
20934 Tune to @var{cpu-type} everything applicable about the generated code,
20935 except for the ABI and the set of available instructions.
20936 The list of @var{cpu-type} values is the same as for @option{-march}.
20937 The default is the value used for @option{-march}.
20940 @itemx -mno-tpf-trace
20941 @opindex mtpf-trace
20942 @opindex mno-tpf-trace
20943 Generate code that adds (does not add) in TPF OS specific branches to trace
20944 routines in the operating system. This option is off by default, even
20945 when compiling for the TPF OS@.
20948 @itemx -mno-fused-madd
20949 @opindex mfused-madd
20950 @opindex mno-fused-madd
20951 Generate code that uses (does not use) the floating-point multiply and
20952 accumulate instructions. These instructions are generated by default if
20953 hardware floating point is used.
20955 @item -mwarn-framesize=@var{framesize}
20956 @opindex mwarn-framesize
20957 Emit a warning if the current function exceeds the given frame size. Because
20958 this is a compile-time check it doesn't need to be a real problem when the program
20959 runs. It is intended to identify functions that most probably cause
20960 a stack overflow. It is useful to be used in an environment with limited stack
20961 size e.g.@: the linux kernel.
20963 @item -mwarn-dynamicstack
20964 @opindex mwarn-dynamicstack
20965 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20966 arrays. This is generally a bad idea with a limited stack size.
20968 @item -mstack-guard=@var{stack-guard}
20969 @itemx -mstack-size=@var{stack-size}
20970 @opindex mstack-guard
20971 @opindex mstack-size
20972 If these options are provided the S/390 back end emits additional instructions in
20973 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20974 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20975 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20976 the frame size of the compiled function is chosen.
20977 These options are intended to be used to help debugging stack overflow problems.
20978 The additionally emitted code causes only little overhead and hence can also be
20979 used in production-like systems without greater performance degradation. The given
20980 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20981 @var{stack-guard} without exceeding 64k.
20982 In order to be efficient the extra code makes the assumption that the stack starts
20983 at an address aligned to the value given by @var{stack-size}.
20984 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20986 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20988 If the hotpatch option is enabled, a ``hot-patching'' function
20989 prologue is generated for all functions in the compilation unit.
20990 The funtion label is prepended with the given number of two-byte
20991 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20992 the label, 2 * @var{post-halfwords} bytes are appended, using the
20993 largest NOP like instructions the architecture allows (maximum
20996 If both arguments are zero, hotpatching is disabled.
20998 This option can be overridden for individual functions with the
20999 @code{hotpatch} attribute.
21002 @node Score Options
21003 @subsection Score Options
21004 @cindex Score Options
21006 These options are defined for Score implementations:
21011 Compile code for big-endian mode. This is the default.
21015 Compile code for little-endian mode.
21019 Disable generation of @code{bcnz} instructions.
21023 Enable generation of unaligned load and store instructions.
21027 Enable the use of multiply-accumulate instructions. Disabled by default.
21031 Specify the SCORE5 as the target architecture.
21035 Specify the SCORE5U of the target architecture.
21039 Specify the SCORE7 as the target architecture. This is the default.
21043 Specify the SCORE7D as the target architecture.
21047 @subsection SH Options
21049 These @samp{-m} options are defined for the SH implementations:
21054 Generate code for the SH1.
21058 Generate code for the SH2.
21061 Generate code for the SH2e.
21065 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21066 that the floating-point unit is not used.
21068 @item -m2a-single-only
21069 @opindex m2a-single-only
21070 Generate code for the SH2a-FPU, in such a way that no double-precision
21071 floating-point operations are used.
21074 @opindex m2a-single
21075 Generate code for the SH2a-FPU assuming the floating-point unit is in
21076 single-precision mode by default.
21080 Generate code for the SH2a-FPU assuming the floating-point unit is in
21081 double-precision mode by default.
21085 Generate code for the SH3.
21089 Generate code for the SH3e.
21093 Generate code for the SH4 without a floating-point unit.
21095 @item -m4-single-only
21096 @opindex m4-single-only
21097 Generate code for the SH4 with a floating-point unit that only
21098 supports single-precision arithmetic.
21102 Generate code for the SH4 assuming the floating-point unit is in
21103 single-precision mode by default.
21107 Generate code for the SH4.
21111 Generate code for SH4-100.
21113 @item -m4-100-nofpu
21114 @opindex m4-100-nofpu
21115 Generate code for SH4-100 in such a way that the
21116 floating-point unit is not used.
21118 @item -m4-100-single
21119 @opindex m4-100-single
21120 Generate code for SH4-100 assuming the floating-point unit is in
21121 single-precision mode by default.
21123 @item -m4-100-single-only
21124 @opindex m4-100-single-only
21125 Generate code for SH4-100 in such a way that no double-precision
21126 floating-point operations are used.
21130 Generate code for SH4-200.
21132 @item -m4-200-nofpu
21133 @opindex m4-200-nofpu
21134 Generate code for SH4-200 without in such a way that the
21135 floating-point unit is not used.
21137 @item -m4-200-single
21138 @opindex m4-200-single
21139 Generate code for SH4-200 assuming the floating-point unit is in
21140 single-precision mode by default.
21142 @item -m4-200-single-only
21143 @opindex m4-200-single-only
21144 Generate code for SH4-200 in such a way that no double-precision
21145 floating-point operations are used.
21149 Generate code for SH4-300.
21151 @item -m4-300-nofpu
21152 @opindex m4-300-nofpu
21153 Generate code for SH4-300 without in such a way that the
21154 floating-point unit is not used.
21156 @item -m4-300-single
21157 @opindex m4-300-single
21158 Generate code for SH4-300 in such a way that no double-precision
21159 floating-point operations are used.
21161 @item -m4-300-single-only
21162 @opindex m4-300-single-only
21163 Generate code for SH4-300 in such a way that no double-precision
21164 floating-point operations are used.
21168 Generate code for SH4-340 (no MMU, no FPU).
21172 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21177 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21178 floating-point unit is not used.
21180 @item -m4a-single-only
21181 @opindex m4a-single-only
21182 Generate code for the SH4a, in such a way that no double-precision
21183 floating-point operations are used.
21186 @opindex m4a-single
21187 Generate code for the SH4a assuming the floating-point unit is in
21188 single-precision mode by default.
21192 Generate code for the SH4a.
21196 Same as @option{-m4a-nofpu}, except that it implicitly passes
21197 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21198 instructions at the moment.
21202 Compile code for the processor in big-endian mode.
21206 Compile code for the processor in little-endian mode.
21210 Align doubles at 64-bit boundaries. Note that this changes the calling
21211 conventions, and thus some functions from the standard C library do
21212 not work unless you recompile it first with @option{-mdalign}.
21216 Shorten some address references at link time, when possible; uses the
21217 linker option @option{-relax}.
21221 Use 32-bit offsets in @code{switch} tables. The default is to use
21226 Enable the use of bit manipulation instructions on SH2A.
21230 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21231 alignment constraints.
21235 Comply with the calling conventions defined by Renesas.
21238 @opindex mno-renesas
21239 Comply with the calling conventions defined for GCC before the Renesas
21240 conventions were available. This option is the default for all
21241 targets of the SH toolchain.
21244 @opindex mnomacsave
21245 Mark the @code{MAC} register as call-clobbered, even if
21246 @option{-mrenesas} is given.
21252 Control the IEEE compliance of floating-point comparisons, which affects the
21253 handling of cases where the result of a comparison is unordered. By default
21254 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21255 enabled @option{-mno-ieee} is implicitly set, which results in faster
21256 floating-point greater-equal and less-equal comparisons. The implcit settings
21257 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21259 @item -minline-ic_invalidate
21260 @opindex minline-ic_invalidate
21261 Inline code to invalidate instruction cache entries after setting up
21262 nested function trampolines.
21263 This option has no effect if @option{-musermode} is in effect and the selected
21264 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21266 If the selected code generation option does not allow the use of the @code{icbi}
21267 instruction, and @option{-musermode} is not in effect, the inlined code
21268 manipulates the instruction cache address array directly with an associative
21269 write. This not only requires privileged mode at run time, but it also
21270 fails if the cache line had been mapped via the TLB and has become unmapped.
21274 Dump instruction size and location in the assembly code.
21277 @opindex mpadstruct
21278 This option is deprecated. It pads structures to multiple of 4 bytes,
21279 which is incompatible with the SH ABI@.
21281 @item -matomic-model=@var{model}
21282 @opindex matomic-model=@var{model}
21283 Sets the model of atomic operations and additional parameters as a comma
21284 separated list. For details on the atomic built-in functions see
21285 @ref{__atomic Builtins}. The following models and parameters are supported:
21290 Disable compiler generated atomic sequences and emit library calls for atomic
21291 operations. This is the default if the target is not @code{sh*-*-linux*}.
21294 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21295 built-in functions. The generated atomic sequences require additional support
21296 from the interrupt/exception handling code of the system and are only suitable
21297 for SH3* and SH4* single-core systems. This option is enabled by default when
21298 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21299 this option also partially utilizes the hardware atomic instructions
21300 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21301 @samp{strict} is specified.
21304 Generate software atomic sequences that use a variable in the thread control
21305 block. This is a variation of the gUSA sequences which can also be used on
21306 SH1* and SH2* targets. The generated atomic sequences require additional
21307 support from the interrupt/exception handling code of the system and are only
21308 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21309 parameter has to be specified as well.
21312 Generate software atomic sequences that temporarily disable interrupts by
21313 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21314 in privileged mode and is only suitable for single-core systems. Additional
21315 support from the interrupt/exception handling code of the system is not
21316 required. This model is enabled by default when the target is
21317 @code{sh*-*-linux*} and SH1* or SH2*.
21320 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21321 instructions only. This is only available on SH4A and is suitable for
21322 multi-core systems. Since the hardware instructions support only 32 bit atomic
21323 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21324 Code compiled with this option is also compatible with other software
21325 atomic model interrupt/exception handling systems if executed on an SH4A
21326 system. Additional support from the interrupt/exception handling code of the
21327 system is not required for this model.
21330 This parameter specifies the offset in bytes of the variable in the thread
21331 control block structure that should be used by the generated atomic sequences
21332 when the @samp{soft-tcb} model has been selected. For other models this
21333 parameter is ignored. The specified value must be an integer multiple of four
21334 and in the range 0-1020.
21337 This parameter prevents mixed usage of multiple atomic models, even if they
21338 are compatible, and makes the compiler generate atomic sequences of the
21339 specified model only.
21345 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21346 Notice that depending on the particular hardware and software configuration
21347 this can degrade overall performance due to the operand cache line flushes
21348 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21349 processors the @code{tas.b} instruction must be used with caution since it
21350 can result in data corruption for certain cache configurations.
21353 @opindex mprefergot
21354 When generating position-independent code, emit function calls using
21355 the Global Offset Table instead of the Procedure Linkage Table.
21358 @itemx -mno-usermode
21360 @opindex mno-usermode
21361 Don't allow (allow) the compiler generating privileged mode code. Specifying
21362 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21363 inlined code would not work in user mode. @option{-musermode} is the default
21364 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21365 @option{-musermode} has no effect, since there is no user mode.
21367 @item -multcost=@var{number}
21368 @opindex multcost=@var{number}
21369 Set the cost to assume for a multiply insn.
21371 @item -mdiv=@var{strategy}
21372 @opindex mdiv=@var{strategy}
21373 Set the division strategy to be used for integer division operations.
21374 @var{strategy} can be one of:
21379 Calls a library function that uses the single-step division instruction
21380 @code{div1} to perform the operation. Division by zero calculates an
21381 unspecified result and does not trap. This is the default except for SH4,
21382 SH2A and SHcompact.
21385 Calls a library function that performs the operation in double precision
21386 floating point. Division by zero causes a floating-point exception. This is
21387 the default for SHcompact with FPU. Specifying this for targets that do not
21388 have a double precision FPU defaults to @code{call-div1}.
21391 Calls a library function that uses a lookup table for small divisors and
21392 the @code{div1} instruction with case distinction for larger divisors. Division
21393 by zero calculates an unspecified result and does not trap. This is the default
21394 for SH4. Specifying this for targets that do not have dynamic shift
21395 instructions defaults to @code{call-div1}.
21399 When a division strategy has not been specified the default strategy is
21400 selected based on the current target. For SH2A the default strategy is to
21401 use the @code{divs} and @code{divu} instructions instead of library function
21404 @item -maccumulate-outgoing-args
21405 @opindex maccumulate-outgoing-args
21406 Reserve space once for outgoing arguments in the function prologue rather
21407 than around each call. Generally beneficial for performance and size. Also
21408 needed for unwinding to avoid changing the stack frame around conditional code.
21410 @item -mdivsi3_libfunc=@var{name}
21411 @opindex mdivsi3_libfunc=@var{name}
21412 Set the name of the library function used for 32-bit signed division to
21414 This only affects the name used in the @samp{call} division strategies, and
21415 the compiler still expects the same sets of input/output/clobbered registers as
21416 if this option were not present.
21418 @item -mfixed-range=@var{register-range}
21419 @opindex mfixed-range
21420 Generate code treating the given register range as fixed registers.
21421 A fixed register is one that the register allocator can not use. This is
21422 useful when compiling kernel code. A register range is specified as
21423 two registers separated by a dash. Multiple register ranges can be
21424 specified separated by a comma.
21426 @item -mbranch-cost=@var{num}
21427 @opindex mbranch-cost=@var{num}
21428 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21429 make the compiler try to generate more branch-free code if possible.
21430 If not specified the value is selected depending on the processor type that
21431 is being compiled for.
21434 @itemx -mno-zdcbranch
21435 @opindex mzdcbranch
21436 @opindex mno-zdcbranch
21437 Assume (do not assume) that zero displacement conditional branch instructions
21438 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21439 compiler prefers zero displacement branch code sequences. This is
21440 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21441 disabled by specifying @option{-mno-zdcbranch}.
21443 @item -mcbranch-force-delay-slot
21444 @opindex mcbranch-force-delay-slot
21445 Force the usage of delay slots for conditional branches, which stuffs the delay
21446 slot with a @code{nop} if a suitable instruction can't be found. By default
21447 this option is disabled. It can be enabled to work around hardware bugs as
21448 found in the original SH7055.
21451 @itemx -mno-fused-madd
21452 @opindex mfused-madd
21453 @opindex mno-fused-madd
21454 Generate code that uses (does not use) the floating-point multiply and
21455 accumulate instructions. These instructions are generated by default
21456 if hardware floating point is used. The machine-dependent
21457 @option{-mfused-madd} option is now mapped to the machine-independent
21458 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21459 mapped to @option{-ffp-contract=off}.
21465 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21466 and cosine approximations. The option @option{-mfsca} must be used in
21467 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21468 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21469 approximations even if @option{-funsafe-math-optimizations} is in effect.
21475 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21476 reciprocal square root approximations. The option @option{-mfsrra} must be used
21477 in combination with @option{-funsafe-math-optimizations} and
21478 @option{-ffinite-math-only}. It is enabled by default when generating code for
21479 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21480 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21483 @item -mpretend-cmove
21484 @opindex mpretend-cmove
21485 Prefer zero-displacement conditional branches for conditional move instruction
21486 patterns. This can result in faster code on the SH4 processor.
21490 Generate code using the FDPIC ABI.
21494 @node Solaris 2 Options
21495 @subsection Solaris 2 Options
21496 @cindex Solaris 2 options
21498 These @samp{-m} options are supported on Solaris 2:
21501 @item -mclear-hwcap
21502 @opindex mclear-hwcap
21503 @option{-mclear-hwcap} tells the compiler to remove the hardware
21504 capabilities generated by the Solaris assembler. This is only necessary
21505 when object files use ISA extensions not supported by the current
21506 machine, but check at runtime whether or not to use them.
21508 @item -mimpure-text
21509 @opindex mimpure-text
21510 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21511 the compiler to not pass @option{-z text} to the linker when linking a
21512 shared object. Using this option, you can link position-dependent
21513 code into a shared object.
21515 @option{-mimpure-text} suppresses the ``relocations remain against
21516 allocatable but non-writable sections'' linker error message.
21517 However, the necessary relocations trigger copy-on-write, and the
21518 shared object is not actually shared across processes. Instead of
21519 using @option{-mimpure-text}, you should compile all source code with
21520 @option{-fpic} or @option{-fPIC}.
21524 These switches are supported in addition to the above on Solaris 2:
21529 Add support for multithreading using the POSIX threads library. This
21530 option sets flags for both the preprocessor and linker. This option does
21531 not affect the thread safety of object code produced by the compiler or
21532 that of libraries supplied with it.
21536 This is a synonym for @option{-pthreads}.
21539 @node SPARC Options
21540 @subsection SPARC Options
21541 @cindex SPARC options
21543 These @samp{-m} options are supported on the SPARC:
21546 @item -mno-app-regs
21548 @opindex mno-app-regs
21550 Specify @option{-mapp-regs} to generate output using the global registers
21551 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21552 global register 1, each global register 2 through 4 is then treated as an
21553 allocable register that is clobbered by function calls. This is the default.
21555 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21556 specify @option{-mno-app-regs}. You should compile libraries and system
21557 software with this option.
21563 With @option{-mflat}, the compiler does not generate save/restore instructions
21564 and uses a ``flat'' or single register window model. This model is compatible
21565 with the regular register window model. The local registers and the input
21566 registers (0--5) are still treated as ``call-saved'' registers and are
21567 saved on the stack as needed.
21569 With @option{-mno-flat} (the default), the compiler generates save/restore
21570 instructions (except for leaf functions). This is the normal operating mode.
21573 @itemx -mhard-float
21575 @opindex mhard-float
21576 Generate output containing floating-point instructions. This is the
21580 @itemx -msoft-float
21582 @opindex msoft-float
21583 Generate output containing library calls for floating point.
21584 @strong{Warning:} the requisite libraries are not available for all SPARC
21585 targets. Normally the facilities of the machine's usual C compiler are
21586 used, but this cannot be done directly in cross-compilation. You must make
21587 your own arrangements to provide suitable library functions for
21588 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21589 @samp{sparclite-*-*} do provide software floating-point support.
21591 @option{-msoft-float} changes the calling convention in the output file;
21592 therefore, it is only useful if you compile @emph{all} of a program with
21593 this option. In particular, you need to compile @file{libgcc.a}, the
21594 library that comes with GCC, with @option{-msoft-float} in order for
21597 @item -mhard-quad-float
21598 @opindex mhard-quad-float
21599 Generate output containing quad-word (long double) floating-point
21602 @item -msoft-quad-float
21603 @opindex msoft-quad-float
21604 Generate output containing library calls for quad-word (long double)
21605 floating-point instructions. The functions called are those specified
21606 in the SPARC ABI@. This is the default.
21608 As of this writing, there are no SPARC implementations that have hardware
21609 support for the quad-word floating-point instructions. They all invoke
21610 a trap handler for one of these instructions, and then the trap handler
21611 emulates the effect of the instruction. Because of the trap handler overhead,
21612 this is much slower than calling the ABI library routines. Thus the
21613 @option{-msoft-quad-float} option is the default.
21615 @item -mno-unaligned-doubles
21616 @itemx -munaligned-doubles
21617 @opindex mno-unaligned-doubles
21618 @opindex munaligned-doubles
21619 Assume that doubles have 8-byte alignment. This is the default.
21621 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21622 alignment only if they are contained in another type, or if they have an
21623 absolute address. Otherwise, it assumes they have 4-byte alignment.
21624 Specifying this option avoids some rare compatibility problems with code
21625 generated by other compilers. It is not the default because it results
21626 in a performance loss, especially for floating-point code.
21629 @itemx -mno-user-mode
21630 @opindex muser-mode
21631 @opindex mno-user-mode
21632 Do not generate code that can only run in supervisor mode. This is relevant
21633 only for the @code{casa} instruction emitted for the LEON3 processor. This
21636 @item -mfaster-structs
21637 @itemx -mno-faster-structs
21638 @opindex mfaster-structs
21639 @opindex mno-faster-structs
21640 With @option{-mfaster-structs}, the compiler assumes that structures
21641 should have 8-byte alignment. This enables the use of pairs of
21642 @code{ldd} and @code{std} instructions for copies in structure
21643 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21644 However, the use of this changed alignment directly violates the SPARC
21645 ABI@. Thus, it's intended only for use on targets where the developer
21646 acknowledges that their resulting code is not directly in line with
21647 the rules of the ABI@.
21649 @item -mstd-struct-return
21650 @itemx -mno-std-struct-return
21651 @opindex mstd-struct-return
21652 @opindex mno-std-struct-return
21653 With @option{-mstd-struct-return}, the compiler generates checking code
21654 in functions returning structures or unions to detect size mismatches
21655 between the two sides of function calls, as per the 32-bit ABI@.
21657 The default is @option{-mno-std-struct-return}. This option has no effect
21660 @item -mcpu=@var{cpu_type}
21662 Set the instruction set, register set, and instruction scheduling parameters
21663 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21664 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21665 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21666 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21667 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21668 @samp{niagara3} and @samp{niagara4}.
21670 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21671 which selects the best architecture option for the host processor.
21672 @option{-mcpu=native} has no effect if GCC does not recognize
21675 Default instruction scheduling parameters are used for values that select
21676 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21677 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21679 Here is a list of each supported architecture and their supported
21687 supersparc, hypersparc, leon, leon3
21690 f930, f934, sparclite86x
21696 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21699 By default (unless configured otherwise), GCC generates code for the V7
21700 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21701 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21702 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21703 SPARCStation 1, 2, IPX etc.
21705 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21706 architecture. The only difference from V7 code is that the compiler emits
21707 the integer multiply and integer divide instructions which exist in SPARC-V8
21708 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21709 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21712 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21713 the SPARC architecture. This adds the integer multiply, integer divide step
21714 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21715 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21716 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21717 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21718 MB86934 chip, which is the more recent SPARClite with FPU@.
21720 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21721 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21722 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21723 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21724 optimizes it for the TEMIC SPARClet chip.
21726 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21727 architecture. This adds 64-bit integer and floating-point move instructions,
21728 3 additional floating-point condition code registers and conditional move
21729 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21730 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21731 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21732 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21733 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21734 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21735 additionally optimizes it for Sun UltraSPARC T2 chips. With
21736 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21737 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21738 additionally optimizes it for Sun UltraSPARC T4 chips.
21740 @item -mtune=@var{cpu_type}
21742 Set the instruction scheduling parameters for machine type
21743 @var{cpu_type}, but do not set the instruction set or register set that the
21744 option @option{-mcpu=@var{cpu_type}} does.
21746 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21747 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21748 that select a particular CPU implementation. Those are @samp{cypress},
21749 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21750 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21751 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21752 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21753 toolchains, @samp{native} can also be used.
21758 @opindex mno-v8plus
21759 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21760 difference from the V8 ABI is that the global and out registers are
21761 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21762 mode for all SPARC-V9 processors.
21768 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21769 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21775 With @option{-mvis2}, GCC generates code that takes advantage of
21776 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21777 default is @option{-mvis2} when targeting a cpu that supports such
21778 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21779 also sets @option{-mvis}.
21785 With @option{-mvis3}, GCC generates code that takes advantage of
21786 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21787 default is @option{-mvis3} when targeting a cpu that supports such
21788 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21789 also sets @option{-mvis2} and @option{-mvis}.
21794 @opindex mno-cbcond
21795 With @option{-mcbcond}, GCC generates code that takes advantage of
21796 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21797 The default is @option{-mcbcond} when targeting a cpu that supports such
21798 instructions, such as niagara-4 and later.
21804 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21805 population count instruction. The default is @option{-mpopc}
21806 when targeting a cpu that supports such instructions, such as Niagara-2 and
21813 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21814 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21815 when targeting a cpu that supports such instructions, such as Niagara-3 and
21819 @opindex mfix-at697f
21820 Enable the documented workaround for the single erratum of the Atmel AT697F
21821 processor (which corresponds to erratum #13 of the AT697E processor).
21824 @opindex mfix-ut699
21825 Enable the documented workarounds for the floating-point errata and the data
21826 cache nullify errata of the UT699 processor.
21829 These @samp{-m} options are supported in addition to the above
21830 on SPARC-V9 processors in 64-bit environments:
21837 Generate code for a 32-bit or 64-bit environment.
21838 The 32-bit environment sets int, long and pointer to 32 bits.
21839 The 64-bit environment sets int to 32 bits and long and pointer
21842 @item -mcmodel=@var{which}
21844 Set the code model to one of
21848 The Medium/Low code model: 64-bit addresses, programs
21849 must be linked in the low 32 bits of memory. Programs can be statically
21850 or dynamically linked.
21853 The Medium/Middle code model: 64-bit addresses, programs
21854 must be linked in the low 44 bits of memory, the text and data segments must
21855 be less than 2GB in size and the data segment must be located within 2GB of
21859 The Medium/Anywhere code model: 64-bit addresses, programs
21860 may be linked anywhere in memory, the text and data segments must be less
21861 than 2GB in size and the data segment must be located within 2GB of the
21865 The Medium/Anywhere code model for embedded systems:
21866 64-bit addresses, the text and data segments must be less than 2GB in
21867 size, both starting anywhere in memory (determined at link time). The
21868 global register %g4 points to the base of the data segment. Programs
21869 are statically linked and PIC is not supported.
21872 @item -mmemory-model=@var{mem-model}
21873 @opindex mmemory-model
21874 Set the memory model in force on the processor to one of
21878 The default memory model for the processor and operating system.
21881 Relaxed Memory Order
21884 Partial Store Order
21890 Sequential Consistency
21893 These memory models are formally defined in Appendix D of the Sparc V9
21894 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21897 @itemx -mno-stack-bias
21898 @opindex mstack-bias
21899 @opindex mno-stack-bias
21900 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21901 frame pointer if present, are offset by @minus{}2047 which must be added back
21902 when making stack frame references. This is the default in 64-bit mode.
21903 Otherwise, assume no such offset is present.
21907 @subsection SPU Options
21908 @cindex SPU options
21910 These @samp{-m} options are supported on the SPU:
21914 @itemx -merror-reloc
21915 @opindex mwarn-reloc
21916 @opindex merror-reloc
21918 The loader for SPU does not handle dynamic relocations. By default, GCC
21919 gives an error when it generates code that requires a dynamic
21920 relocation. @option{-mno-error-reloc} disables the error,
21921 @option{-mwarn-reloc} generates a warning instead.
21924 @itemx -munsafe-dma
21926 @opindex munsafe-dma
21928 Instructions that initiate or test completion of DMA must not be
21929 reordered with respect to loads and stores of the memory that is being
21931 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21932 memory accesses, but that can lead to inefficient code in places where the
21933 memory is known to not change. Rather than mark the memory as volatile,
21934 you can use @option{-msafe-dma} to tell the compiler to treat
21935 the DMA instructions as potentially affecting all memory.
21937 @item -mbranch-hints
21938 @opindex mbranch-hints
21940 By default, GCC generates a branch hint instruction to avoid
21941 pipeline stalls for always-taken or probably-taken branches. A hint
21942 is not generated closer than 8 instructions away from its branch.
21943 There is little reason to disable them, except for debugging purposes,
21944 or to make an object a little bit smaller.
21948 @opindex msmall-mem
21949 @opindex mlarge-mem
21951 By default, GCC generates code assuming that addresses are never larger
21952 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21953 a full 32-bit address.
21958 By default, GCC links against startup code that assumes the SPU-style
21959 main function interface (which has an unconventional parameter list).
21960 With @option{-mstdmain}, GCC links your program against startup
21961 code that assumes a C99-style interface to @code{main}, including a
21962 local copy of @code{argv} strings.
21964 @item -mfixed-range=@var{register-range}
21965 @opindex mfixed-range
21966 Generate code treating the given register range as fixed registers.
21967 A fixed register is one that the register allocator cannot use. This is
21968 useful when compiling kernel code. A register range is specified as
21969 two registers separated by a dash. Multiple register ranges can be
21970 specified separated by a comma.
21976 Compile code assuming that pointers to the PPU address space accessed
21977 via the @code{__ea} named address space qualifier are either 32 or 64
21978 bits wide. The default is 32 bits. As this is an ABI-changing option,
21979 all object code in an executable must be compiled with the same setting.
21981 @item -maddress-space-conversion
21982 @itemx -mno-address-space-conversion
21983 @opindex maddress-space-conversion
21984 @opindex mno-address-space-conversion
21985 Allow/disallow treating the @code{__ea} address space as superset
21986 of the generic address space. This enables explicit type casts
21987 between @code{__ea} and generic pointer as well as implicit
21988 conversions of generic pointers to @code{__ea} pointers. The
21989 default is to allow address space pointer conversions.
21991 @item -mcache-size=@var{cache-size}
21992 @opindex mcache-size
21993 This option controls the version of libgcc that the compiler links to an
21994 executable and selects a software-managed cache for accessing variables
21995 in the @code{__ea} address space with a particular cache size. Possible
21996 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21997 and @samp{128}. The default cache size is 64KB.
21999 @item -matomic-updates
22000 @itemx -mno-atomic-updates
22001 @opindex matomic-updates
22002 @opindex mno-atomic-updates
22003 This option controls the version of libgcc that the compiler links to an
22004 executable and selects whether atomic updates to the software-managed
22005 cache of PPU-side variables are used. If you use atomic updates, changes
22006 to a PPU variable from SPU code using the @code{__ea} named address space
22007 qualifier do not interfere with changes to other PPU variables residing
22008 in the same cache line from PPU code. If you do not use atomic updates,
22009 such interference may occur; however, writing back cache lines is
22010 more efficient. The default behavior is to use atomic updates.
22013 @itemx -mdual-nops=@var{n}
22014 @opindex mdual-nops
22015 By default, GCC inserts nops to increase dual issue when it expects
22016 it to increase performance. @var{n} can be a value from 0 to 10. A
22017 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22018 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22020 @item -mhint-max-nops=@var{n}
22021 @opindex mhint-max-nops
22022 Maximum number of nops to insert for a branch hint. A branch hint must
22023 be at least 8 instructions away from the branch it is affecting. GCC
22024 inserts up to @var{n} nops to enforce this, otherwise it does not
22025 generate the branch hint.
22027 @item -mhint-max-distance=@var{n}
22028 @opindex mhint-max-distance
22029 The encoding of the branch hint instruction limits the hint to be within
22030 256 instructions of the branch it is affecting. By default, GCC makes
22031 sure it is within 125.
22034 @opindex msafe-hints
22035 Work around a hardware bug that causes the SPU to stall indefinitely.
22036 By default, GCC inserts the @code{hbrp} instruction to make sure
22037 this stall won't happen.
22041 @node System V Options
22042 @subsection Options for System V
22044 These additional options are available on System V Release 4 for
22045 compatibility with other compilers on those systems:
22050 Create a shared object.
22051 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22055 Identify the versions of each tool used by the compiler, in a
22056 @code{.ident} assembler directive in the output.
22060 Refrain from adding @code{.ident} directives to the output file (this is
22063 @item -YP,@var{dirs}
22065 Search the directories @var{dirs}, and no others, for libraries
22066 specified with @option{-l}.
22068 @item -Ym,@var{dir}
22070 Look in the directory @var{dir} to find the M4 preprocessor.
22071 The assembler uses this option.
22072 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22073 @c the generic assembler that comes with Solaris takes just -Ym.
22076 @node TILE-Gx Options
22077 @subsection TILE-Gx Options
22078 @cindex TILE-Gx options
22080 These @samp{-m} options are supported on the TILE-Gx:
22083 @item -mcmodel=small
22084 @opindex mcmodel=small
22085 Generate code for the small model. The distance for direct calls is
22086 limited to 500M in either direction. PC-relative addresses are 32
22087 bits. Absolute addresses support the full address range.
22089 @item -mcmodel=large
22090 @opindex mcmodel=large
22091 Generate code for the large model. There is no limitation on call
22092 distance, pc-relative addresses, or absolute addresses.
22094 @item -mcpu=@var{name}
22096 Selects the type of CPU to be targeted. Currently the only supported
22097 type is @samp{tilegx}.
22103 Generate code for a 32-bit or 64-bit environment. The 32-bit
22104 environment sets int, long, and pointer to 32 bits. The 64-bit
22105 environment sets int to 32 bits and long and pointer to 64 bits.
22108 @itemx -mlittle-endian
22109 @opindex mbig-endian
22110 @opindex mlittle-endian
22111 Generate code in big/little endian mode, respectively.
22114 @node TILEPro Options
22115 @subsection TILEPro Options
22116 @cindex TILEPro options
22118 These @samp{-m} options are supported on the TILEPro:
22121 @item -mcpu=@var{name}
22123 Selects the type of CPU to be targeted. Currently the only supported
22124 type is @samp{tilepro}.
22128 Generate code for a 32-bit environment, which sets int, long, and
22129 pointer to 32 bits. This is the only supported behavior so the flag
22130 is essentially ignored.
22134 @subsection V850 Options
22135 @cindex V850 Options
22137 These @samp{-m} options are defined for V850 implementations:
22141 @itemx -mno-long-calls
22142 @opindex mlong-calls
22143 @opindex mno-long-calls
22144 Treat all calls as being far away (near). If calls are assumed to be
22145 far away, the compiler always loads the function's address into a
22146 register, and calls indirect through the pointer.
22152 Do not optimize (do optimize) basic blocks that use the same index
22153 pointer 4 or more times to copy pointer into the @code{ep} register, and
22154 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22155 option is on by default if you optimize.
22157 @item -mno-prolog-function
22158 @itemx -mprolog-function
22159 @opindex mno-prolog-function
22160 @opindex mprolog-function
22161 Do not use (do use) external functions to save and restore registers
22162 at the prologue and epilogue of a function. The external functions
22163 are slower, but use less code space if more than one function saves
22164 the same number of registers. The @option{-mprolog-function} option
22165 is on by default if you optimize.
22169 Try to make the code as small as possible. At present, this just turns
22170 on the @option{-mep} and @option{-mprolog-function} options.
22172 @item -mtda=@var{n}
22174 Put static or global variables whose size is @var{n} bytes or less into
22175 the tiny data area that register @code{ep} points to. The tiny data
22176 area can hold up to 256 bytes in total (128 bytes for byte references).
22178 @item -msda=@var{n}
22180 Put static or global variables whose size is @var{n} bytes or less into
22181 the small data area that register @code{gp} points to. The small data
22182 area can hold up to 64 kilobytes.
22184 @item -mzda=@var{n}
22186 Put static or global variables whose size is @var{n} bytes or less into
22187 the first 32 kilobytes of memory.
22191 Specify that the target processor is the V850.
22195 Specify that the target processor is the V850E3V5. The preprocessor
22196 constant @code{__v850e3v5__} is defined if this option is used.
22200 Specify that the target processor is the V850E3V5. This is an alias for
22201 the @option{-mv850e3v5} option.
22205 Specify that the target processor is the V850E2V3. The preprocessor
22206 constant @code{__v850e2v3__} is defined if this option is used.
22210 Specify that the target processor is the V850E2. The preprocessor
22211 constant @code{__v850e2__} is defined if this option is used.
22215 Specify that the target processor is the V850E1. The preprocessor
22216 constants @code{__v850e1__} and @code{__v850e__} are defined if
22217 this option is used.
22221 Specify that the target processor is the V850ES. This is an alias for
22222 the @option{-mv850e1} option.
22226 Specify that the target processor is the V850E@. The preprocessor
22227 constant @code{__v850e__} is defined if this option is used.
22229 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22230 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22231 are defined then a default target processor is chosen and the
22232 relevant @samp{__v850*__} preprocessor constant is defined.
22234 The preprocessor constants @code{__v850} and @code{__v851__} are always
22235 defined, regardless of which processor variant is the target.
22237 @item -mdisable-callt
22238 @itemx -mno-disable-callt
22239 @opindex mdisable-callt
22240 @opindex mno-disable-callt
22241 This option suppresses generation of the @code{CALLT} instruction for the
22242 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22245 This option is enabled by default when the RH850 ABI is
22246 in use (see @option{-mrh850-abi}), and disabled by default when the
22247 GCC ABI is in use. If @code{CALLT} instructions are being generated
22248 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22254 Pass on (or do not pass on) the @option{-mrelax} command-line option
22258 @itemx -mno-long-jumps
22259 @opindex mlong-jumps
22260 @opindex mno-long-jumps
22261 Disable (or re-enable) the generation of PC-relative jump instructions.
22264 @itemx -mhard-float
22265 @opindex msoft-float
22266 @opindex mhard-float
22267 Disable (or re-enable) the generation of hardware floating point
22268 instructions. This option is only significant when the target
22269 architecture is @samp{V850E2V3} or higher. If hardware floating point
22270 instructions are being generated then the C preprocessor symbol
22271 @code{__FPU_OK__} is defined, otherwise the symbol
22272 @code{__NO_FPU__} is defined.
22276 Enables the use of the e3v5 LOOP instruction. The use of this
22277 instruction is not enabled by default when the e3v5 architecture is
22278 selected because its use is still experimental.
22282 @opindex mrh850-abi
22284 Enables support for the RH850 version of the V850 ABI. This is the
22285 default. With this version of the ABI the following rules apply:
22289 Integer sized structures and unions are returned via a memory pointer
22290 rather than a register.
22293 Large structures and unions (more than 8 bytes in size) are passed by
22297 Functions are aligned to 16-bit boundaries.
22300 The @option{-m8byte-align} command-line option is supported.
22303 The @option{-mdisable-callt} command-line option is enabled by
22304 default. The @option{-mno-disable-callt} command-line option is not
22308 When this version of the ABI is enabled the C preprocessor symbol
22309 @code{__V850_RH850_ABI__} is defined.
22313 Enables support for the old GCC version of the V850 ABI. With this
22314 version of the ABI the following rules apply:
22318 Integer sized structures and unions are returned in register @code{r10}.
22321 Large structures and unions (more than 8 bytes in size) are passed by
22325 Functions are aligned to 32-bit boundaries, unless optimizing for
22329 The @option{-m8byte-align} command-line option is not supported.
22332 The @option{-mdisable-callt} command-line option is supported but not
22333 enabled by default.
22336 When this version of the ABI is enabled the C preprocessor symbol
22337 @code{__V850_GCC_ABI__} is defined.
22339 @item -m8byte-align
22340 @itemx -mno-8byte-align
22341 @opindex m8byte-align
22342 @opindex mno-8byte-align
22343 Enables support for @code{double} and @code{long long} types to be
22344 aligned on 8-byte boundaries. The default is to restrict the
22345 alignment of all objects to at most 4-bytes. When
22346 @option{-m8byte-align} is in effect the C preprocessor symbol
22347 @code{__V850_8BYTE_ALIGN__} is defined.
22350 @opindex mbig-switch
22351 Generate code suitable for big switch tables. Use this option only if
22352 the assembler/linker complain about out of range branches within a switch
22357 This option causes r2 and r5 to be used in the code generated by
22358 the compiler. This setting is the default.
22360 @item -mno-app-regs
22361 @opindex mno-app-regs
22362 This option causes r2 and r5 to be treated as fixed registers.
22367 @subsection VAX Options
22368 @cindex VAX options
22370 These @samp{-m} options are defined for the VAX:
22375 Do not output certain jump instructions (@code{aobleq} and so on)
22376 that the Unix assembler for the VAX cannot handle across long
22381 Do output those jump instructions, on the assumption that the
22382 GNU assembler is being used.
22386 Output code for G-format floating-point numbers instead of D-format.
22389 @node Visium Options
22390 @subsection Visium Options
22391 @cindex Visium options
22397 A program which performs file I/O and is destined to run on an MCM target
22398 should be linked with this option. It causes the libraries libc.a and
22399 libdebug.a to be linked. The program should be run on the target under
22400 the control of the GDB remote debugging stub.
22404 A program which performs file I/O and is destined to run on the simulator
22405 should be linked with option. This causes libraries libc.a and libsim.a to
22409 @itemx -mhard-float
22411 @opindex mhard-float
22412 Generate code containing floating-point instructions. This is the
22416 @itemx -msoft-float
22418 @opindex msoft-float
22419 Generate code containing library calls for floating-point.
22421 @option{-msoft-float} changes the calling convention in the output file;
22422 therefore, it is only useful if you compile @emph{all} of a program with
22423 this option. In particular, you need to compile @file{libgcc.a}, the
22424 library that comes with GCC, with @option{-msoft-float} in order for
22427 @item -mcpu=@var{cpu_type}
22429 Set the instruction set, register set, and instruction scheduling parameters
22430 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22431 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22433 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22435 By default (unless configured otherwise), GCC generates code for the GR5
22436 variant of the Visium architecture.
22438 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22439 architecture. The only difference from GR5 code is that the compiler will
22440 generate block move instructions.
22442 @item -mtune=@var{cpu_type}
22444 Set the instruction scheduling parameters for machine type @var{cpu_type},
22445 but do not set the instruction set or register set that the option
22446 @option{-mcpu=@var{cpu_type}} would.
22450 Generate code for the supervisor mode, where there are no restrictions on
22451 the access to general registers. This is the default.
22454 @opindex muser-mode
22455 Generate code for the user mode, where the access to some general registers
22456 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22457 mode; on the GR6, only registers r29 to r31 are affected.
22461 @subsection VMS Options
22463 These @samp{-m} options are defined for the VMS implementations:
22466 @item -mvms-return-codes
22467 @opindex mvms-return-codes
22468 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22469 condition (e.g.@ error) codes.
22471 @item -mdebug-main=@var{prefix}
22472 @opindex mdebug-main=@var{prefix}
22473 Flag the first routine whose name starts with @var{prefix} as the main
22474 routine for the debugger.
22478 Default to 64-bit memory allocation routines.
22480 @item -mpointer-size=@var{size}
22481 @opindex mpointer-size=@var{size}
22482 Set the default size of pointers. Possible options for @var{size} are
22483 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22484 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22485 The later option disables @code{pragma pointer_size}.
22488 @node VxWorks Options
22489 @subsection VxWorks Options
22490 @cindex VxWorks Options
22492 The options in this section are defined for all VxWorks targets.
22493 Options specific to the target hardware are listed with the other
22494 options for that target.
22499 GCC can generate code for both VxWorks kernels and real time processes
22500 (RTPs). This option switches from the former to the latter. It also
22501 defines the preprocessor macro @code{__RTP__}.
22504 @opindex non-static
22505 Link an RTP executable against shared libraries rather than static
22506 libraries. The options @option{-static} and @option{-shared} can
22507 also be used for RTPs (@pxref{Link Options}); @option{-static}
22514 These options are passed down to the linker. They are defined for
22515 compatibility with Diab.
22518 @opindex Xbind-lazy
22519 Enable lazy binding of function calls. This option is equivalent to
22520 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22524 Disable lazy binding of function calls. This option is the default and
22525 is defined for compatibility with Diab.
22529 @subsection x86 Options
22530 @cindex x86 Options
22532 These @samp{-m} options are defined for the x86 family of computers.
22536 @item -march=@var{cpu-type}
22538 Generate instructions for the machine type @var{cpu-type}. In contrast to
22539 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22540 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22541 to generate code that may not run at all on processors other than the one
22542 indicated. Specifying @option{-march=@var{cpu-type}} implies
22543 @option{-mtune=@var{cpu-type}}.
22545 The choices for @var{cpu-type} are:
22549 This selects the CPU to generate code for at compilation time by determining
22550 the processor type of the compiling machine. Using @option{-march=native}
22551 enables all instruction subsets supported by the local machine (hence
22552 the result might not run on different machines). Using @option{-mtune=native}
22553 produces code optimized for the local machine under the constraints
22554 of the selected instruction set.
22557 Original Intel i386 CPU@.
22560 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22564 Intel Pentium CPU with no MMX support.
22567 Intel Lakemont MCU, based on Intel Pentium CPU.
22570 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22573 Intel Pentium Pro CPU@.
22576 When used with @option{-march}, the Pentium Pro
22577 instruction set is used, so the code runs on all i686 family chips.
22578 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22581 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22586 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22590 Intel Pentium M; low-power version of Intel Pentium III CPU
22591 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22595 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22598 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22602 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22603 SSE2 and SSE3 instruction set support.
22606 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22607 instruction set support.
22610 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22611 SSE4.1, SSE4.2 and POPCNT instruction set support.
22614 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22615 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22618 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22619 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22622 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22623 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22624 instruction set support.
22627 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22628 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22629 BMI, BMI2 and F16C instruction set support.
22632 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22633 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22634 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22637 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22638 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22639 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
22640 XSAVES instruction set support.
22643 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22644 instruction set support.
22647 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22648 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22651 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22652 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22653 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
22654 AVX512CD instruction set support.
22656 @item skylake-avx512
22657 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22658 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22659 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
22660 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
22663 AMD K6 CPU with MMX instruction set support.
22667 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
22670 @itemx athlon-tbird
22671 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22677 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22678 instruction set support.
22684 Processors based on the AMD K8 core with x86-64 instruction set support,
22685 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22686 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22687 instruction set extensions.)
22690 @itemx opteron-sse3
22691 @itemx athlon64-sse3
22692 Improved versions of AMD K8 cores with SSE3 instruction set support.
22696 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22697 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22698 instruction set extensions.)
22701 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22702 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22703 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22705 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22706 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22707 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22710 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22711 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22712 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22713 64-bit instruction set extensions.
22715 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22716 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22717 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22718 SSE4.2, ABM and 64-bit instruction set extensions.
22721 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
22722 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
22723 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
22724 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
22725 instruction set extensions.
22728 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22729 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22730 instruction set extensions.)
22733 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22734 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22735 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22738 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22742 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22743 instruction set support.
22746 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22747 implemented for this chip.)
22750 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22752 implemented for this chip.)
22755 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22758 @item -mtune=@var{cpu-type}
22760 Tune to @var{cpu-type} everything applicable about the generated code, except
22761 for the ABI and the set of available instructions.
22762 While picking a specific @var{cpu-type} schedules things appropriately
22763 for that particular chip, the compiler does not generate any code that
22764 cannot run on the default machine type unless you use a
22765 @option{-march=@var{cpu-type}} option.
22766 For example, if GCC is configured for i686-pc-linux-gnu
22767 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22768 but still runs on i686 machines.
22770 The choices for @var{cpu-type} are the same as for @option{-march}.
22771 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22775 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22776 If you know the CPU on which your code will run, then you should use
22777 the corresponding @option{-mtune} or @option{-march} option instead of
22778 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22779 of your application will have, then you should use this option.
22781 As new processors are deployed in the marketplace, the behavior of this
22782 option will change. Therefore, if you upgrade to a newer version of
22783 GCC, code generation controlled by this option will change to reflect
22785 that are most common at the time that version of GCC is released.
22787 There is no @option{-march=generic} option because @option{-march}
22788 indicates the instruction set the compiler can use, and there is no
22789 generic instruction set applicable to all processors. In contrast,
22790 @option{-mtune} indicates the processor (or, in this case, collection of
22791 processors) for which the code is optimized.
22794 Produce code optimized for the most current Intel processors, which are
22795 Haswell and Silvermont for this version of GCC. If you know the CPU
22796 on which your code will run, then you should use the corresponding
22797 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22798 But, if you want your application performs better on both Haswell and
22799 Silvermont, then you should use this option.
22801 As new Intel processors are deployed in the marketplace, the behavior of
22802 this option will change. Therefore, if you upgrade to a newer version of
22803 GCC, code generation controlled by this option will change to reflect
22804 the most current Intel processors at the time that version of GCC is
22807 There is no @option{-march=intel} option because @option{-march} indicates
22808 the instruction set the compiler can use, and there is no common
22809 instruction set applicable to all processors. In contrast,
22810 @option{-mtune} indicates the processor (or, in this case, collection of
22811 processors) for which the code is optimized.
22814 @item -mcpu=@var{cpu-type}
22816 A deprecated synonym for @option{-mtune}.
22818 @item -mfpmath=@var{unit}
22820 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22821 for @var{unit} are:
22825 Use the standard 387 floating-point coprocessor present on the majority of chips and
22826 emulated otherwise. Code compiled with this option runs almost everywhere.
22827 The temporary results are computed in 80-bit precision instead of the precision
22828 specified by the type, resulting in slightly different results compared to most
22829 of other chips. See @option{-ffloat-store} for more detailed description.
22831 This is the default choice for x86-32 targets.
22834 Use scalar floating-point instructions present in the SSE instruction set.
22835 This instruction set is supported by Pentium III and newer chips,
22836 and in the AMD line
22837 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22838 instruction set supports only single-precision arithmetic, thus the double and
22839 extended-precision arithmetic are still done using 387. A later version, present
22840 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22843 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22844 or @option{-msse2} switches to enable SSE extensions and make this option
22845 effective. For the x86-64 compiler, these extensions are enabled by default.
22847 The resulting code should be considerably faster in the majority of cases and avoid
22848 the numerical instability problems of 387 code, but may break some existing
22849 code that expects temporaries to be 80 bits.
22851 This is the default choice for the x86-64 compiler.
22856 Attempt to utilize both instruction sets at once. This effectively doubles the
22857 amount of available registers, and on chips with separate execution units for
22858 387 and SSE the execution resources too. Use this option with care, as it is
22859 still experimental, because the GCC register allocator does not model separate
22860 functional units well, resulting in unstable performance.
22863 @item -masm=@var{dialect}
22864 @opindex masm=@var{dialect}
22865 Output assembly instructions using selected @var{dialect}. Also affects
22866 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22867 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22868 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22869 not support @samp{intel}.
22872 @itemx -mno-ieee-fp
22874 @opindex mno-ieee-fp
22875 Control whether or not the compiler uses IEEE floating-point
22876 comparisons. These correctly handle the case where the result of a
22877 comparison is unordered.
22880 @opindex msoft-float
22881 Generate output containing library calls for floating point.
22883 @strong{Warning:} the requisite libraries are not part of GCC@.
22884 Normally the facilities of the machine's usual C compiler are used, but
22885 this can't be done directly in cross-compilation. You must make your
22886 own arrangements to provide suitable library functions for
22889 On machines where a function returns floating-point results in the 80387
22890 register stack, some floating-point opcodes may be emitted even if
22891 @option{-msoft-float} is used.
22893 @item -mno-fp-ret-in-387
22894 @opindex mno-fp-ret-in-387
22895 Do not use the FPU registers for return values of functions.
22897 The usual calling convention has functions return values of types
22898 @code{float} and @code{double} in an FPU register, even if there
22899 is no FPU@. The idea is that the operating system should emulate
22902 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22903 in ordinary CPU registers instead.
22905 @item -mno-fancy-math-387
22906 @opindex mno-fancy-math-387
22907 Some 387 emulators do not support the @code{sin}, @code{cos} and
22908 @code{sqrt} instructions for the 387. Specify this option to avoid
22909 generating those instructions. This option is the default on
22910 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22911 indicates that the target CPU always has an FPU and so the
22912 instruction does not need emulation. These
22913 instructions are not generated unless you also use the
22914 @option{-funsafe-math-optimizations} switch.
22916 @item -malign-double
22917 @itemx -mno-align-double
22918 @opindex malign-double
22919 @opindex mno-align-double
22920 Control whether GCC aligns @code{double}, @code{long double}, and
22921 @code{long long} variables on a two-word boundary or a one-word
22922 boundary. Aligning @code{double} variables on a two-word boundary
22923 produces code that runs somewhat faster on a Pentium at the
22924 expense of more memory.
22926 On x86-64, @option{-malign-double} is enabled by default.
22928 @strong{Warning:} if you use the @option{-malign-double} switch,
22929 structures containing the above types are aligned differently than
22930 the published application binary interface specifications for the x86-32
22931 and are not binary compatible with structures in code compiled
22932 without that switch.
22934 @item -m96bit-long-double
22935 @itemx -m128bit-long-double
22936 @opindex m96bit-long-double
22937 @opindex m128bit-long-double
22938 These switches control the size of @code{long double} type. The x86-32
22939 application binary interface specifies the size to be 96 bits,
22940 so @option{-m96bit-long-double} is the default in 32-bit mode.
22942 Modern architectures (Pentium and newer) prefer @code{long double}
22943 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22944 conforming to the ABI, this is not possible. So specifying
22945 @option{-m128bit-long-double} aligns @code{long double}
22946 to a 16-byte boundary by padding the @code{long double} with an additional
22949 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22950 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22952 Notice that neither of these options enable any extra precision over the x87
22953 standard of 80 bits for a @code{long double}.
22955 @strong{Warning:} if you override the default value for your target ABI, this
22956 changes the size of
22957 structures and arrays containing @code{long double} variables,
22958 as well as modifying the function calling convention for functions taking
22959 @code{long double}. Hence they are not binary-compatible
22960 with code compiled without that switch.
22962 @item -mlong-double-64
22963 @itemx -mlong-double-80
22964 @itemx -mlong-double-128
22965 @opindex mlong-double-64
22966 @opindex mlong-double-80
22967 @opindex mlong-double-128
22968 These switches control the size of @code{long double} type. A size
22969 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22970 type. This is the default for 32-bit Bionic C library. A size
22971 of 128 bits makes the @code{long double} type equivalent to the
22972 @code{__float128} type. This is the default for 64-bit Bionic C library.
22974 @strong{Warning:} if you override the default value for your target ABI, this
22975 changes the size of
22976 structures and arrays containing @code{long double} variables,
22977 as well as modifying the function calling convention for functions taking
22978 @code{long double}. Hence they are not binary-compatible
22979 with code compiled without that switch.
22981 @item -malign-data=@var{type}
22982 @opindex malign-data
22983 Control how GCC aligns variables. Supported values for @var{type} are
22984 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22985 and earlier, @samp{abi} uses alignment value as specified by the
22986 psABI, and @samp{cacheline} uses increased alignment value to match
22987 the cache line size. @samp{compat} is the default.
22989 @item -mlarge-data-threshold=@var{threshold}
22990 @opindex mlarge-data-threshold
22991 When @option{-mcmodel=medium} is specified, data objects larger than
22992 @var{threshold} are placed in the large data section. This value must be the
22993 same across all objects linked into the binary, and defaults to 65535.
22997 Use a different function-calling convention, in which functions that
22998 take a fixed number of arguments return with the @code{ret @var{num}}
22999 instruction, which pops their arguments while returning. This saves one
23000 instruction in the caller since there is no need to pop the arguments
23003 You can specify that an individual function is called with this calling
23004 sequence with the function attribute @code{stdcall}. You can also
23005 override the @option{-mrtd} option by using the function attribute
23006 @code{cdecl}. @xref{Function Attributes}.
23008 @strong{Warning:} this calling convention is incompatible with the one
23009 normally used on Unix, so you cannot use it if you need to call
23010 libraries compiled with the Unix compiler.
23012 Also, you must provide function prototypes for all functions that
23013 take variable numbers of arguments (including @code{printf});
23014 otherwise incorrect code is generated for calls to those
23017 In addition, seriously incorrect code results if you call a
23018 function with too many arguments. (Normally, extra arguments are
23019 harmlessly ignored.)
23021 @item -mregparm=@var{num}
23023 Control how many registers are used to pass integer arguments. By
23024 default, no registers are used to pass arguments, and at most 3
23025 registers can be used. You can control this behavior for a specific
23026 function by using the function attribute @code{regparm}.
23027 @xref{Function Attributes}.
23029 @strong{Warning:} if you use this switch, and
23030 @var{num} is nonzero, then you must build all modules with the same
23031 value, including any libraries. This includes the system libraries and
23035 @opindex msseregparm
23036 Use SSE register passing conventions for float and double arguments
23037 and return values. You can control this behavior for a specific
23038 function by using the function attribute @code{sseregparm}.
23039 @xref{Function Attributes}.
23041 @strong{Warning:} if you use this switch then you must build all
23042 modules with the same value, including any libraries. This includes
23043 the system libraries and startup modules.
23045 @item -mvect8-ret-in-mem
23046 @opindex mvect8-ret-in-mem
23047 Return 8-byte vectors in memory instead of MMX registers. This is the
23048 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
23049 Studio compilers until version 12. Later compiler versions (starting
23050 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
23051 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
23052 you need to remain compatible with existing code produced by those
23053 previous compiler versions or older versions of GCC@.
23062 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
23063 is specified, the significands of results of floating-point operations are
23064 rounded to 24 bits (single precision); @option{-mpc64} rounds the
23065 significands of results of floating-point operations to 53 bits (double
23066 precision) and @option{-mpc80} rounds the significands of results of
23067 floating-point operations to 64 bits (extended double precision), which is
23068 the default. When this option is used, floating-point operations in higher
23069 precisions are not available to the programmer without setting the FPU
23070 control word explicitly.
23072 Setting the rounding of floating-point operations to less than the default
23073 80 bits can speed some programs by 2% or more. Note that some mathematical
23074 libraries assume that extended-precision (80-bit) floating-point operations
23075 are enabled by default; routines in such libraries could suffer significant
23076 loss of accuracy, typically through so-called ``catastrophic cancellation'',
23077 when this option is used to set the precision to less than extended precision.
23079 @item -mstackrealign
23080 @opindex mstackrealign
23081 Realign the stack at entry. On the x86, the @option{-mstackrealign}
23082 option generates an alternate prologue and epilogue that realigns the
23083 run-time stack if necessary. This supports mixing legacy codes that keep
23084 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
23085 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
23086 applicable to individual functions.
23088 @item -mpreferred-stack-boundary=@var{num}
23089 @opindex mpreferred-stack-boundary
23090 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23091 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23092 the default is 4 (16 bytes or 128 bits).
23094 @strong{Warning:} When generating code for the x86-64 architecture with
23095 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
23096 used to keep the stack boundary aligned to 8 byte boundary. Since
23097 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
23098 intended to be used in controlled environment where stack space is
23099 important limitation. This option leads to wrong code when functions
23100 compiled with 16 byte stack alignment (such as functions from a standard
23101 library) are called with misaligned stack. In this case, SSE
23102 instructions may lead to misaligned memory access traps. In addition,
23103 variable arguments are handled incorrectly for 16 byte aligned
23104 objects (including x87 long double and __int128), leading to wrong
23105 results. You must build all modules with
23106 @option{-mpreferred-stack-boundary=3}, including any libraries. This
23107 includes the system libraries and startup modules.
23109 @item -mincoming-stack-boundary=@var{num}
23110 @opindex mincoming-stack-boundary
23111 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
23112 boundary. If @option{-mincoming-stack-boundary} is not specified,
23113 the one specified by @option{-mpreferred-stack-boundary} is used.
23115 On Pentium and Pentium Pro, @code{double} and @code{long double} values
23116 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
23117 suffer significant run time performance penalties. On Pentium III, the
23118 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
23119 properly if it is not 16-byte aligned.
23121 To ensure proper alignment of this values on the stack, the stack boundary
23122 must be as aligned as that required by any value stored on the stack.
23123 Further, every function must be generated such that it keeps the stack
23124 aligned. Thus calling a function compiled with a higher preferred
23125 stack boundary from a function compiled with a lower preferred stack
23126 boundary most likely misaligns the stack. It is recommended that
23127 libraries that use callbacks always use the default setting.
23129 This extra alignment does consume extra stack space, and generally
23130 increases code size. Code that is sensitive to stack space usage, such
23131 as embedded systems and operating system kernels, may want to reduce the
23132 preferred alignment to @option{-mpreferred-stack-boundary=2}.
23189 @itemx -mavx512ifma
23190 @opindex mavx512ifma
23192 @itemx -mavx512vbmi
23193 @opindex mavx512vbmi
23205 @opindex mclfushopt
23225 @itemx -mprefetchwt1
23226 @opindex mprefetchwt1
23282 These switches enable the use of instructions in the MMX, SSE,
23283 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
23284 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
23285 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
23286 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
23287 extended instruction sets. Each has a corresponding @option{-mno-} option
23288 to disable use of these instructions.
23290 These extensions are also available as built-in functions: see
23291 @ref{x86 Built-in Functions}, for details of the functions enabled and
23292 disabled by these switches.
23294 To generate SSE/SSE2 instructions automatically from floating-point
23295 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
23297 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
23298 generates new AVX instructions or AVX equivalence for all SSEx instructions
23301 These options enable GCC to use these extended instructions in
23302 generated code, even without @option{-mfpmath=sse}. Applications that
23303 perform run-time CPU detection must compile separate files for each
23304 supported architecture, using the appropriate flags. In particular,
23305 the file containing the CPU detection code should be compiled without
23308 @item -mdump-tune-features
23309 @opindex mdump-tune-features
23310 This option instructs GCC to dump the names of the x86 performance
23311 tuning features and default settings. The names can be used in
23312 @option{-mtune-ctrl=@var{feature-list}}.
23314 @item -mtune-ctrl=@var{feature-list}
23315 @opindex mtune-ctrl=@var{feature-list}
23316 This option is used to do fine grain control of x86 code generation features.
23317 @var{feature-list} is a comma separated list of @var{feature} names. See also
23318 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
23319 on if it is not preceded with @samp{^}, otherwise, it is turned off.
23320 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
23321 developers. Using it may lead to code paths not covered by testing and can
23322 potentially result in compiler ICEs or runtime errors.
23325 @opindex mno-default
23326 This option instructs GCC to turn off all tunable features. See also
23327 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23331 This option instructs GCC to emit a @code{cld} instruction in the prologue
23332 of functions that use string instructions. String instructions depend on
23333 the DF flag to select between autoincrement or autodecrement mode. While the
23334 ABI specifies the DF flag to be cleared on function entry, some operating
23335 systems violate this specification by not clearing the DF flag in their
23336 exception dispatchers. The exception handler can be invoked with the DF flag
23337 set, which leads to wrong direction mode when string instructions are used.
23338 This option can be enabled by default on 32-bit x86 targets by configuring
23339 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23340 instructions can be suppressed with the @option{-mno-cld} compiler option
23344 @opindex mvzeroupper
23345 This option instructs GCC to emit a @code{vzeroupper} instruction
23346 before a transfer of control flow out of the function to minimize
23347 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23350 @item -mprefer-avx128
23351 @opindex mprefer-avx128
23352 This option instructs GCC to use 128-bit AVX instructions instead of
23353 256-bit AVX instructions in the auto-vectorizer.
23357 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23358 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23359 (or oword) data types.
23360 This is useful for high-resolution counters that can be updated
23361 by multiple processors (or cores). This instruction is generated as part of
23362 atomic built-in functions: see @ref{__sync Builtins} or
23363 @ref{__atomic Builtins} for details.
23367 This option enables generation of @code{SAHF} instructions in 64-bit code.
23368 Early Intel Pentium 4 CPUs with Intel 64 support,
23369 prior to the introduction of Pentium 4 G1 step in December 2005,
23370 lacked the @code{LAHF} and @code{SAHF} instructions
23371 which are supported by AMD64.
23372 These are load and store instructions, respectively, for certain status flags.
23373 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23374 @code{drem}, and @code{remainder} built-in functions;
23375 see @ref{Other Builtins} for details.
23379 This option enables use of the @code{movbe} instruction to implement
23380 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23384 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23385 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23386 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23390 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23391 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23392 with an additional Newton-Raphson step
23393 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23394 (and their vectorized
23395 variants) for single-precision floating-point arguments. These instructions
23396 are generated only when @option{-funsafe-math-optimizations} is enabled
23397 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
23398 Note that while the throughput of the sequence is higher than the throughput
23399 of the non-reciprocal instruction, the precision of the sequence can be
23400 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23402 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23403 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23404 combination), and doesn't need @option{-mrecip}.
23406 Also note that GCC emits the above sequence with additional Newton-Raphson step
23407 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23408 already with @option{-ffast-math} (or the above option combination), and
23409 doesn't need @option{-mrecip}.
23411 @item -mrecip=@var{opt}
23412 @opindex mrecip=opt
23413 This option controls which reciprocal estimate instructions
23414 may be used. @var{opt} is a comma-separated list of options, which may
23415 be preceded by a @samp{!} to invert the option:
23419 Enable all estimate instructions.
23422 Enable the default instructions, equivalent to @option{-mrecip}.
23425 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23428 Enable the approximation for scalar division.
23431 Enable the approximation for vectorized division.
23434 Enable the approximation for scalar square root.
23437 Enable the approximation for vectorized square root.
23440 So, for example, @option{-mrecip=all,!sqrt} enables
23441 all of the reciprocal approximations, except for square root.
23443 @item -mveclibabi=@var{type}
23444 @opindex mveclibabi
23445 Specifies the ABI type to use for vectorizing intrinsics using an
23446 external library. Supported values for @var{type} are @samp{svml}
23447 for the Intel short
23448 vector math library and @samp{acml} for the AMD math core library.
23449 To use this option, both @option{-ftree-vectorize} and
23450 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23451 ABI-compatible library must be specified at link time.
23453 GCC currently emits calls to @code{vmldExp2},
23454 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23455 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23456 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23457 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23458 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23459 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23460 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23461 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23462 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23463 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23464 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23465 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23466 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23467 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23468 when @option{-mveclibabi=acml} is used.
23470 @item -mabi=@var{name}
23472 Generate code for the specified calling convention. Permissible values
23473 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23474 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23475 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23476 You can control this behavior for specific functions by
23477 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23478 @xref{Function Attributes}.
23480 @item -mtls-dialect=@var{type}
23481 @opindex mtls-dialect
23482 Generate code to access thread-local storage using the @samp{gnu} or
23483 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23484 @samp{gnu2} is more efficient, but it may add compile- and run-time
23485 requirements that cannot be satisfied on all systems.
23488 @itemx -mno-push-args
23489 @opindex mpush-args
23490 @opindex mno-push-args
23491 Use PUSH operations to store outgoing parameters. This method is shorter
23492 and usually equally fast as method using SUB/MOV operations and is enabled
23493 by default. In some cases disabling it may improve performance because of
23494 improved scheduling and reduced dependencies.
23496 @item -maccumulate-outgoing-args
23497 @opindex maccumulate-outgoing-args
23498 If enabled, the maximum amount of space required for outgoing arguments is
23499 computed in the function prologue. This is faster on most modern CPUs
23500 because of reduced dependencies, improved scheduling and reduced stack usage
23501 when the preferred stack boundary is not equal to 2. The drawback is a notable
23502 increase in code size. This switch implies @option{-mno-push-args}.
23506 Support thread-safe exception handling on MinGW. Programs that rely
23507 on thread-safe exception handling must compile and link all code with the
23508 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23509 @option{-D_MT}; when linking, it links in a special thread helper library
23510 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23512 @item -mms-bitfields
23513 @itemx -mno-ms-bitfields
23514 @opindex mms-bitfields
23515 @opindex mno-ms-bitfields
23517 Enable/disable bit-field layout compatible with the native Microsoft
23520 If @code{packed} is used on a structure, or if bit-fields are used,
23521 it may be that the Microsoft ABI lays out the structure differently
23522 than the way GCC normally does. Particularly when moving packed
23523 data between functions compiled with GCC and the native Microsoft compiler
23524 (either via function call or as data in a file), it may be necessary to access
23527 This option is enabled by default for Microsoft Windows
23528 targets. This behavior can also be controlled locally by use of variable
23529 or type attributes. For more information, see @ref{x86 Variable Attributes}
23530 and @ref{x86 Type Attributes}.
23532 The Microsoft structure layout algorithm is fairly simple with the exception
23533 of the bit-field packing.
23534 The padding and alignment of members of structures and whether a bit-field
23535 can straddle a storage-unit boundary are determine by these rules:
23538 @item Structure members are stored sequentially in the order in which they are
23539 declared: the first member has the lowest memory address and the last member
23542 @item Every data object has an alignment requirement. The alignment requirement
23543 for all data except structures, unions, and arrays is either the size of the
23544 object or the current packing size (specified with either the
23545 @code{aligned} attribute or the @code{pack} pragma),
23546 whichever is less. For structures, unions, and arrays,
23547 the alignment requirement is the largest alignment requirement of its members.
23548 Every object is allocated an offset so that:
23551 offset % alignment_requirement == 0
23554 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
23555 unit if the integral types are the same size and if the next bit-field fits
23556 into the current allocation unit without crossing the boundary imposed by the
23557 common alignment requirements of the bit-fields.
23560 MSVC interprets zero-length bit-fields in the following ways:
23563 @item If a zero-length bit-field is inserted between two bit-fields that
23564 are normally coalesced, the bit-fields are not coalesced.
23571 unsigned long bf_1 : 12;
23573 unsigned long bf_2 : 12;
23578 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
23579 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
23581 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
23582 alignment of the zero-length bit-field is greater than the member that follows it,
23583 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
23604 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
23605 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
23606 bit-field does not affect the alignment of @code{bar} or, as a result, the size
23609 Taking this into account, it is important to note the following:
23612 @item If a zero-length bit-field follows a normal bit-field, the type of the
23613 zero-length bit-field may affect the alignment of the structure as whole. For
23614 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
23615 normal bit-field, and is of type short.
23617 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
23618 still affect the alignment of the structure:
23629 Here, @code{t4} takes up 4 bytes.
23632 @item Zero-length bit-fields following non-bit-field members are ignored:
23644 Here, @code{t5} takes up 2 bytes.
23648 @item -mno-align-stringops
23649 @opindex mno-align-stringops
23650 Do not align the destination of inlined string operations. This switch reduces
23651 code size and improves performance in case the destination is already aligned,
23652 but GCC doesn't know about it.
23654 @item -minline-all-stringops
23655 @opindex minline-all-stringops
23656 By default GCC inlines string operations only when the destination is
23657 known to be aligned to least a 4-byte boundary.
23658 This enables more inlining and increases code
23659 size, but may improve performance of code that depends on fast
23660 @code{memcpy}, @code{strlen},
23661 and @code{memset} for short lengths.
23663 @item -minline-stringops-dynamically
23664 @opindex minline-stringops-dynamically
23665 For string operations of unknown size, use run-time checks with
23666 inline code for small blocks and a library call for large blocks.
23668 @item -mstringop-strategy=@var{alg}
23669 @opindex mstringop-strategy=@var{alg}
23670 Override the internal decision heuristic for the particular algorithm to use
23671 for inlining string operations. The allowed values for @var{alg} are:
23677 Expand using i386 @code{rep} prefix of the specified size.
23681 @itemx unrolled_loop
23682 Expand into an inline loop.
23685 Always use a library call.
23688 @item -mmemcpy-strategy=@var{strategy}
23689 @opindex mmemcpy-strategy=@var{strategy}
23690 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23691 should be inlined and what inline algorithm to use when the expected size
23692 of the copy operation is known. @var{strategy}
23693 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23694 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23695 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23696 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23697 in the list must be specified in increasing order. The minimal byte size for
23698 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23701 @item -mmemset-strategy=@var{strategy}
23702 @opindex mmemset-strategy=@var{strategy}
23703 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23704 @code{__builtin_memset} expansion.
23706 @item -momit-leaf-frame-pointer
23707 @opindex momit-leaf-frame-pointer
23708 Don't keep the frame pointer in a register for leaf functions. This
23709 avoids the instructions to save, set up, and restore frame pointers and
23710 makes an extra register available in leaf functions. The option
23711 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23712 which might make debugging harder.
23714 @item -mtls-direct-seg-refs
23715 @itemx -mno-tls-direct-seg-refs
23716 @opindex mtls-direct-seg-refs
23717 Controls whether TLS variables may be accessed with offsets from the
23718 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23719 or whether the thread base pointer must be added. Whether or not this
23720 is valid depends on the operating system, and whether it maps the
23721 segment to cover the entire TLS area.
23723 For systems that use the GNU C Library, the default is on.
23726 @itemx -mno-sse2avx
23728 Specify that the assembler should encode SSE instructions with VEX
23729 prefix. The option @option{-mavx} turns this on by default.
23734 If profiling is active (@option{-pg}), put the profiling
23735 counter call before the prologue.
23736 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23737 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23739 @item -mrecord-mcount
23740 @itemx -mno-record-mcount
23741 @opindex mrecord-mcount
23742 If profiling is active (@option{-pg}), generate a __mcount_loc section
23743 that contains pointers to each profiling call. This is useful for
23744 automatically patching and out calls.
23747 @itemx -mno-nop-mcount
23748 @opindex mnop-mcount
23749 If profiling is active (@option{-pg}), generate the calls to
23750 the profiling functions as nops. This is useful when they
23751 should be patched in later dynamically. This is likely only
23752 useful together with @option{-mrecord-mcount}.
23754 @item -mskip-rax-setup
23755 @itemx -mno-skip-rax-setup
23756 @opindex mskip-rax-setup
23757 When generating code for the x86-64 architecture with SSE extensions
23758 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23759 register when there are no variable arguments passed in vector registers.
23761 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23762 saving vector registers on stack when passing variable arguments, the
23763 impacts of this option are callees may waste some stack space,
23764 misbehave or jump to a random location. GCC 4.4 or newer don't have
23765 those issues, regardless the RAX register value.
23768 @itemx -mno-8bit-idiv
23769 @opindex m8bit-idiv
23770 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23771 much faster than 32-bit/64-bit integer divide. This option generates a
23772 run-time check. If both dividend and divisor are within range of 0
23773 to 255, 8-bit unsigned integer divide is used instead of
23774 32-bit/64-bit integer divide.
23776 @item -mavx256-split-unaligned-load
23777 @itemx -mavx256-split-unaligned-store
23778 @opindex mavx256-split-unaligned-load
23779 @opindex mavx256-split-unaligned-store
23780 Split 32-byte AVX unaligned load and store.
23782 @item -mstack-protector-guard=@var{guard}
23783 @opindex mstack-protector-guard=@var{guard}
23784 Generate stack protection code using canary at @var{guard}. Supported
23785 locations are @samp{global} for global canary or @samp{tls} for per-thread
23786 canary in the TLS block (the default). This option has effect only when
23787 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23789 @item -mmitigate-rop
23790 @opindex mmitigate-rop
23791 Try to avoid generating code sequences that contain unintended return
23792 opcodes, to mitigate against certain forms of attack. At the moment,
23793 this option is limited in what it can do and should not be relied
23794 on to provide serious protection.
23798 These @samp{-m} switches are supported in addition to the above
23799 on x86-64 processors in 64-bit environments.
23812 Generate code for a 16-bit, 32-bit or 64-bit environment.
23813 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23815 generates code that runs on any i386 system.
23817 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23818 types to 64 bits, and generates code for the x86-64 architecture.
23819 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23820 and @option{-mdynamic-no-pic} options.
23822 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23824 generates code for the x86-64 architecture.
23826 The @option{-m16} option is the same as @option{-m32}, except for that
23827 it outputs the @code{.code16gcc} assembly directive at the beginning of
23828 the assembly output so that the binary can run in 16-bit mode.
23830 The @option{-miamcu} option generates code which conforms to Intel MCU
23831 psABI. It requires the @option{-m32} option to be turned on.
23833 @item -mno-red-zone
23834 @opindex mno-red-zone
23835 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23836 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23837 stack pointer that is not modified by signal or interrupt handlers
23838 and therefore can be used for temporary data without adjusting the stack
23839 pointer. The flag @option{-mno-red-zone} disables this red zone.
23841 @item -mcmodel=small
23842 @opindex mcmodel=small
23843 Generate code for the small code model: the program and its symbols must
23844 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23845 Programs can be statically or dynamically linked. This is the default
23848 @item -mcmodel=kernel
23849 @opindex mcmodel=kernel
23850 Generate code for the kernel code model. The kernel runs in the
23851 negative 2 GB of the address space.
23852 This model has to be used for Linux kernel code.
23854 @item -mcmodel=medium
23855 @opindex mcmodel=medium
23856 Generate code for the medium model: the program is linked in the lower 2
23857 GB of the address space. Small symbols are also placed there. Symbols
23858 with sizes larger than @option{-mlarge-data-threshold} are put into
23859 large data or BSS sections and can be located above 2GB. Programs can
23860 be statically or dynamically linked.
23862 @item -mcmodel=large
23863 @opindex mcmodel=large
23864 Generate code for the large model. This model makes no assumptions
23865 about addresses and sizes of sections.
23867 @item -maddress-mode=long
23868 @opindex maddress-mode=long
23869 Generate code for long address mode. This is only supported for 64-bit
23870 and x32 environments. It is the default address mode for 64-bit
23873 @item -maddress-mode=short
23874 @opindex maddress-mode=short
23875 Generate code for short address mode. This is only supported for 32-bit
23876 and x32 environments. It is the default address mode for 32-bit and
23880 @node x86 Windows Options
23881 @subsection x86 Windows Options
23882 @cindex x86 Windows Options
23883 @cindex Windows Options for x86
23885 These additional options are available for Microsoft Windows targets:
23891 specifies that a console application is to be generated, by
23892 instructing the linker to set the PE header subsystem type
23893 required for console applications.
23894 This option is available for Cygwin and MinGW targets and is
23895 enabled by default on those targets.
23899 This option is available for Cygwin and MinGW targets. It
23900 specifies that a DLL---a dynamic link library---is to be
23901 generated, enabling the selection of the required runtime
23902 startup object and entry point.
23904 @item -mnop-fun-dllimport
23905 @opindex mnop-fun-dllimport
23906 This option is available for Cygwin and MinGW targets. It
23907 specifies that the @code{dllimport} attribute should be ignored.
23911 This option is available for MinGW targets. It specifies
23912 that MinGW-specific thread support is to be used.
23916 This option is available for MinGW-w64 targets. It causes
23917 the @code{UNICODE} preprocessor macro to be predefined, and
23918 chooses Unicode-capable runtime startup code.
23922 This option is available for Cygwin and MinGW targets. It
23923 specifies that the typical Microsoft Windows predefined macros are to
23924 be set in the pre-processor, but does not influence the choice
23925 of runtime library/startup code.
23929 This option is available for Cygwin and MinGW targets. It
23930 specifies that a GUI application is to be generated by
23931 instructing the linker to set the PE header subsystem type
23934 @item -fno-set-stack-executable
23935 @opindex fno-set-stack-executable
23936 This option is available for MinGW targets. It specifies that
23937 the executable flag for the stack used by nested functions isn't
23938 set. This is necessary for binaries running in kernel mode of
23939 Microsoft Windows, as there the User32 API, which is used to set executable
23940 privileges, isn't available.
23942 @item -fwritable-relocated-rdata
23943 @opindex fno-writable-relocated-rdata
23944 This option is available for MinGW and Cygwin targets. It specifies
23945 that relocated-data in read-only section is put into .data
23946 section. This is a necessary for older runtimes not supporting
23947 modification of .rdata sections for pseudo-relocation.
23949 @item -mpe-aligned-commons
23950 @opindex mpe-aligned-commons
23951 This option is available for Cygwin and MinGW targets. It
23952 specifies that the GNU extension to the PE file format that
23953 permits the correct alignment of COMMON variables should be
23954 used when generating code. It is enabled by default if
23955 GCC detects that the target assembler found during configuration
23956 supports the feature.
23959 See also under @ref{x86 Options} for standard options.
23961 @node Xstormy16 Options
23962 @subsection Xstormy16 Options
23963 @cindex Xstormy16 Options
23965 These options are defined for Xstormy16:
23970 Choose startup files and linker script suitable for the simulator.
23973 @node Xtensa Options
23974 @subsection Xtensa Options
23975 @cindex Xtensa Options
23977 These options are supported for Xtensa targets:
23981 @itemx -mno-const16
23983 @opindex mno-const16
23984 Enable or disable use of @code{CONST16} instructions for loading
23985 constant values. The @code{CONST16} instruction is currently not a
23986 standard option from Tensilica. When enabled, @code{CONST16}
23987 instructions are always used in place of the standard @code{L32R}
23988 instructions. The use of @code{CONST16} is enabled by default only if
23989 the @code{L32R} instruction is not available.
23992 @itemx -mno-fused-madd
23993 @opindex mfused-madd
23994 @opindex mno-fused-madd
23995 Enable or disable use of fused multiply/add and multiply/subtract
23996 instructions in the floating-point option. This has no effect if the
23997 floating-point option is not also enabled. Disabling fused multiply/add
23998 and multiply/subtract instructions forces the compiler to use separate
23999 instructions for the multiply and add/subtract operations. This may be
24000 desirable in some cases where strict IEEE 754-compliant results are
24001 required: the fused multiply add/subtract instructions do not round the
24002 intermediate result, thereby producing results with @emph{more} bits of
24003 precision than specified by the IEEE standard. Disabling fused multiply
24004 add/subtract instructions also ensures that the program output is not
24005 sensitive to the compiler's ability to combine multiply and add/subtract
24008 @item -mserialize-volatile
24009 @itemx -mno-serialize-volatile
24010 @opindex mserialize-volatile
24011 @opindex mno-serialize-volatile
24012 When this option is enabled, GCC inserts @code{MEMW} instructions before
24013 @code{volatile} memory references to guarantee sequential consistency.
24014 The default is @option{-mserialize-volatile}. Use
24015 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
24017 @item -mforce-no-pic
24018 @opindex mforce-no-pic
24019 For targets, like GNU/Linux, where all user-mode Xtensa code must be
24020 position-independent code (PIC), this option disables PIC for compiling
24023 @item -mtext-section-literals
24024 @itemx -mno-text-section-literals
24025 @opindex mtext-section-literals
24026 @opindex mno-text-section-literals
24027 These options control the treatment of literal pools. The default is
24028 @option{-mno-text-section-literals}, which places literals in a separate
24029 section in the output file. This allows the literal pool to be placed
24030 in a data RAM/ROM, and it also allows the linker to combine literal
24031 pools from separate object files to remove redundant literals and
24032 improve code size. With @option{-mtext-section-literals}, the literals
24033 are interspersed in the text section in order to keep them as close as
24034 possible to their references. This may be necessary for large assembly
24035 files. Literals for each function are placed right before that function.
24037 @item -mauto-litpools
24038 @itemx -mno-auto-litpools
24039 @opindex mauto-litpools
24040 @opindex mno-auto-litpools
24041 These options control the treatment of literal pools. The default is
24042 @option{-mno-auto-litpools}, which places literals in a separate
24043 section in the output file unless @option{-mtext-section-literals} is
24044 used. With @option{-mauto-litpools} the literals are interspersed in
24045 the text section by the assembler. Compiler does not produce explicit
24046 @code{.literal} directives and loads literals into registers with
24047 @code{MOVI} instructions instead of @code{L32R} to let the assembler
24048 do relaxation and place literals as necessary. This option allows
24049 assembler to create several literal pools per function and assemble
24050 very big functions, which may not be possible with
24051 @option{-mtext-section-literals}.
24053 @item -mtarget-align
24054 @itemx -mno-target-align
24055 @opindex mtarget-align
24056 @opindex mno-target-align
24057 When this option is enabled, GCC instructs the assembler to
24058 automatically align instructions to reduce branch penalties at the
24059 expense of some code density. The assembler attempts to widen density
24060 instructions to align branch targets and the instructions following call
24061 instructions. If there are not enough preceding safe density
24062 instructions to align a target, no widening is performed. The
24063 default is @option{-mtarget-align}. These options do not affect the
24064 treatment of auto-aligned instructions like @code{LOOP}, which the
24065 assembler always aligns, either by widening density instructions or
24066 by inserting NOP instructions.
24069 @itemx -mno-longcalls
24070 @opindex mlongcalls
24071 @opindex mno-longcalls
24072 When this option is enabled, GCC instructs the assembler to translate
24073 direct calls to indirect calls unless it can determine that the target
24074 of a direct call is in the range allowed by the call instruction. This
24075 translation typically occurs for calls to functions in other source
24076 files. Specifically, the assembler translates a direct @code{CALL}
24077 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
24078 The default is @option{-mno-longcalls}. This option should be used in
24079 programs where the call target can potentially be out of range. This
24080 option is implemented in the assembler, not the compiler, so the
24081 assembly code generated by GCC still shows direct call
24082 instructions---look at the disassembled object code to see the actual
24083 instructions. Note that the assembler uses an indirect call for
24084 every cross-file call, not just those that really are out of range.
24087 @node zSeries Options
24088 @subsection zSeries Options
24089 @cindex zSeries options
24091 These are listed under @xref{S/390 and zSeries Options}.
24093 @node Code Gen Options
24094 @section Options for Code Generation Conventions
24095 @cindex code generation conventions
24096 @cindex options, code generation
24097 @cindex run-time options
24099 These machine-independent options control the interface conventions
24100 used in code generation.
24102 Most of them have both positive and negative forms; the negative form
24103 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
24104 one of the forms is listed---the one that is not the default. You
24105 can figure out the other form by either removing @samp{no-} or adding
24109 @item -fbounds-check
24110 @opindex fbounds-check
24111 For front ends that support it, generate additional code to check that
24112 indices used to access arrays are within the declared range. This is
24113 currently only supported by the Java and Fortran front ends, where
24114 this option defaults to true and false respectively.
24116 @item -fstack-reuse=@var{reuse-level}
24117 @opindex fstack_reuse
24118 This option controls stack space reuse for user declared local/auto variables
24119 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
24120 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
24121 local variables and temporaries, @samp{named_vars} enables the reuse only for
24122 user defined local variables with names, and @samp{none} disables stack reuse
24123 completely. The default value is @samp{all}. The option is needed when the
24124 program extends the lifetime of a scoped local variable or a compiler generated
24125 temporary beyond the end point defined by the language. When a lifetime of
24126 a variable ends, and if the variable lives in memory, the optimizing compiler
24127 has the freedom to reuse its stack space with other temporaries or scoped
24128 local variables whose live range does not overlap with it. Legacy code extending
24129 local lifetime is likely to break with the stack reuse optimization.
24148 if (*p == 10) // out of scope use of local1
24159 A(int k) : i(k), j(k) @{ @}
24166 void foo(const A& ar)
24173 foo(A(10)); // temp object's lifetime ends when foo returns
24179 ap->i+= 10; // ap references out of scope temp whose space
24180 // is reused with a. What is the value of ap->i?
24185 The lifetime of a compiler generated temporary is well defined by the C++
24186 standard. When a lifetime of a temporary ends, and if the temporary lives
24187 in memory, the optimizing compiler has the freedom to reuse its stack
24188 space with other temporaries or scoped local variables whose live range
24189 does not overlap with it. However some of the legacy code relies on
24190 the behavior of older compilers in which temporaries' stack space is
24191 not reused, the aggressive stack reuse can lead to runtime errors. This
24192 option is used to control the temporary stack reuse optimization.
24196 This option generates traps for signed overflow on addition, subtraction,
24197 multiplication operations.
24198 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
24199 @option{-ftrapv} @option{-fwrapv} on the command-line results in
24200 @option{-fwrapv} being effective. Note that only active options override, so
24201 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
24202 results in @option{-ftrapv} being effective.
24206 This option instructs the compiler to assume that signed arithmetic
24207 overflow of addition, subtraction and multiplication wraps around
24208 using twos-complement representation. This flag enables some optimizations
24209 and disables others. This option is enabled by default for the Java
24210 front end, as required by the Java language specification.
24211 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
24212 @option{-ftrapv} @option{-fwrapv} on the command-line results in
24213 @option{-fwrapv} being effective. Note that only active options override, so
24214 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
24215 results in @option{-ftrapv} being effective.
24218 @opindex fexceptions
24219 Enable exception handling. Generates extra code needed to propagate
24220 exceptions. For some targets, this implies GCC generates frame
24221 unwind information for all functions, which can produce significant data
24222 size overhead, although it does not affect execution. If you do not
24223 specify this option, GCC enables it by default for languages like
24224 C++ that normally require exception handling, and disables it for
24225 languages like C that do not normally require it. However, you may need
24226 to enable this option when compiling C code that needs to interoperate
24227 properly with exception handlers written in C++. You may also wish to
24228 disable this option if you are compiling older C++ programs that don't
24229 use exception handling.
24231 @item -fnon-call-exceptions
24232 @opindex fnon-call-exceptions
24233 Generate code that allows trapping instructions to throw exceptions.
24234 Note that this requires platform-specific runtime support that does
24235 not exist everywhere. Moreover, it only allows @emph{trapping}
24236 instructions to throw exceptions, i.e.@: memory references or floating-point
24237 instructions. It does not allow exceptions to be thrown from
24238 arbitrary signal handlers such as @code{SIGALRM}.
24240 @item -fdelete-dead-exceptions
24241 @opindex fdelete-dead-exceptions
24242 Consider that instructions that may throw exceptions but don't otherwise
24243 contribute to the execution of the program can be optimized away.
24244 This option is enabled by default for the Ada front end, as permitted by
24245 the Ada language specification.
24246 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
24248 @item -funwind-tables
24249 @opindex funwind-tables
24250 Similar to @option{-fexceptions}, except that it just generates any needed
24251 static data, but does not affect the generated code in any other way.
24252 You normally do not need to enable this option; instead, a language processor
24253 that needs this handling enables it on your behalf.
24255 @item -fasynchronous-unwind-tables
24256 @opindex fasynchronous-unwind-tables
24257 Generate unwind table in DWARF 2 format, if supported by target machine. The
24258 table is exact at each instruction boundary, so it can be used for stack
24259 unwinding from asynchronous events (such as debugger or garbage collector).
24261 @item -fno-gnu-unique
24262 @opindex fno-gnu-unique
24263 On systems with recent GNU assembler and C library, the C++ compiler
24264 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
24265 of template static data members and static local variables in inline
24266 functions are unique even in the presence of @code{RTLD_LOCAL}; this
24267 is necessary to avoid problems with a library used by two different
24268 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
24269 therefore disagreeing with the other one about the binding of the
24270 symbol. But this causes @code{dlclose} to be ignored for affected
24271 DSOs; if your program relies on reinitialization of a DSO via
24272 @code{dlclose} and @code{dlopen}, you can use
24273 @option{-fno-gnu-unique}.
24275 @item -fpcc-struct-return
24276 @opindex fpcc-struct-return
24277 Return ``short'' @code{struct} and @code{union} values in memory like
24278 longer ones, rather than in registers. This convention is less
24279 efficient, but it has the advantage of allowing intercallability between
24280 GCC-compiled files and files compiled with other compilers, particularly
24281 the Portable C Compiler (pcc).
24283 The precise convention for returning structures in memory depends
24284 on the target configuration macros.
24286 Short structures and unions are those whose size and alignment match
24287 that of some integer type.
24289 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
24290 switch is not binary compatible with code compiled with the
24291 @option{-freg-struct-return} switch.
24292 Use it to conform to a non-default application binary interface.
24294 @item -freg-struct-return
24295 @opindex freg-struct-return
24296 Return @code{struct} and @code{union} values in registers when possible.
24297 This is more efficient for small structures than
24298 @option{-fpcc-struct-return}.
24300 If you specify neither @option{-fpcc-struct-return} nor
24301 @option{-freg-struct-return}, GCC defaults to whichever convention is
24302 standard for the target. If there is no standard convention, GCC
24303 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
24304 the principal compiler. In those cases, we can choose the standard, and
24305 we chose the more efficient register return alternative.
24307 @strong{Warning:} code compiled with the @option{-freg-struct-return}
24308 switch is not binary compatible with code compiled with the
24309 @option{-fpcc-struct-return} switch.
24310 Use it to conform to a non-default application binary interface.
24312 @item -fshort-enums
24313 @opindex fshort-enums
24314 Allocate to an @code{enum} type only as many bytes as it needs for the
24315 declared range of possible values. Specifically, the @code{enum} type
24316 is equivalent to the smallest integer type that has enough room.
24318 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
24319 code that is not binary compatible with code generated without that switch.
24320 Use it to conform to a non-default application binary interface.
24322 @item -fshort-double
24323 @opindex fshort-double
24324 Use the same size for @code{double} as for @code{float}.
24326 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
24327 code that is not binary compatible with code generated without that switch.
24328 Use it to conform to a non-default application binary interface.
24330 @item -fshort-wchar
24331 @opindex fshort-wchar
24332 Override the underlying type for @code{wchar_t} to be @code{short
24333 unsigned int} instead of the default for the target. This option is
24334 useful for building programs to run under WINE@.
24336 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
24337 code that is not binary compatible with code generated without that switch.
24338 Use it to conform to a non-default application binary interface.
24341 @opindex fno-common
24342 In C code, controls the placement of uninitialized global variables.
24343 Unix C compilers have traditionally permitted multiple definitions of
24344 such variables in different compilation units by placing the variables
24346 This is the behavior specified by @option{-fcommon}, and is the default
24347 for GCC on most targets.
24348 On the other hand, this behavior is not required by ISO C, and on some
24349 targets may carry a speed or code size penalty on variable references.
24350 The @option{-fno-common} option specifies that the compiler should place
24351 uninitialized global variables in the data section of the object file,
24352 rather than generating them as common blocks.
24353 This has the effect that if the same variable is declared
24354 (without @code{extern}) in two different compilations,
24355 you get a multiple-definition error when you link them.
24356 In this case, you must compile with @option{-fcommon} instead.
24357 Compiling with @option{-fno-common} is useful on targets for which
24358 it provides better performance, or if you wish to verify that the
24359 program will work on other systems that always treat uninitialized
24360 variable declarations this way.
24364 Ignore the @code{#ident} directive.
24366 @item -finhibit-size-directive
24367 @opindex finhibit-size-directive
24368 Don't output a @code{.size} assembler directive, or anything else that
24369 would cause trouble if the function is split in the middle, and the
24370 two halves are placed at locations far apart in memory. This option is
24371 used when compiling @file{crtstuff.c}; you should not need to use it
24374 @item -fverbose-asm
24375 @opindex fverbose-asm
24376 Put extra commentary information in the generated assembly code to
24377 make it more readable. This option is generally only of use to those
24378 who actually need to read the generated assembly code (perhaps while
24379 debugging the compiler itself).
24381 @option{-fno-verbose-asm}, the default, causes the
24382 extra information to be omitted and is useful when comparing two assembler
24385 @item -frecord-gcc-switches
24386 @opindex frecord-gcc-switches
24387 This switch causes the command line used to invoke the
24388 compiler to be recorded into the object file that is being created.
24389 This switch is only implemented on some targets and the exact format
24390 of the recording is target and binary file format dependent, but it
24391 usually takes the form of a section containing ASCII text. This
24392 switch is related to the @option{-fverbose-asm} switch, but that
24393 switch only records information in the assembler output file as
24394 comments, so it never reaches the object file.
24395 See also @option{-grecord-gcc-switches} for another
24396 way of storing compiler options into the object file.
24400 @cindex global offset table
24402 Generate position-independent code (PIC) suitable for use in a shared
24403 library, if supported for the target machine. Such code accesses all
24404 constant addresses through a global offset table (GOT)@. The dynamic
24405 loader resolves the GOT entries when the program starts (the dynamic
24406 loader is not part of GCC; it is part of the operating system). If
24407 the GOT size for the linked executable exceeds a machine-specific
24408 maximum size, you get an error message from the linker indicating that
24409 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
24410 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
24411 on the m68k and RS/6000. The x86 has no such limit.)
24413 Position-independent code requires special support, and therefore works
24414 only on certain machines. For the x86, GCC supports PIC for System V
24415 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
24416 position-independent.
24418 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
24423 If supported for the target machine, emit position-independent code,
24424 suitable for dynamic linking and avoiding any limit on the size of the
24425 global offset table. This option makes a difference on AArch64, m68k,
24426 PowerPC and SPARC@.
24428 Position-independent code requires special support, and therefore works
24429 only on certain machines.
24431 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
24438 These options are similar to @option{-fpic} and @option{-fPIC}, but
24439 generated position independent code can be only linked into executables.
24440 Usually these options are used when @option{-pie} GCC option is
24441 used during linking.
24443 @option{-fpie} and @option{-fPIE} both define the macros
24444 @code{__pie__} and @code{__PIE__}. The macros have the value 1
24445 for @option{-fpie} and 2 for @option{-fPIE}.
24449 Do not use the PLT for external function calls in position-independent code.
24450 Instead, load the callee address at call sites from the GOT and branch to it.
24451 This leads to more efficient code by eliminating PLT stubs and exposing
24452 GOT loads to optimizations. On architectures such as 32-bit x86 where
24453 PLT stubs expect the GOT pointer in a specific register, this gives more
24454 register allocation freedom to the compiler.
24455 Lazy binding requires use of the PLT;
24456 with @option{-fno-plt} all external symbols are resolved at load time.
24458 Alternatively, the function attribute @code{noplt} can be used to avoid calls
24459 through the PLT for specific external functions.
24461 In position-dependent code, a few targets also convert calls to
24462 functions that are marked to not use the PLT to use the GOT instead.
24464 @item -fno-jump-tables
24465 @opindex fno-jump-tables
24466 Do not use jump tables for switch statements even where it would be
24467 more efficient than other code generation strategies. This option is
24468 of use in conjunction with @option{-fpic} or @option{-fPIC} for
24469 building code that forms part of a dynamic linker and cannot
24470 reference the address of a jump table. On some targets, jump tables
24471 do not require a GOT and this option is not needed.
24473 @item -ffixed-@var{reg}
24475 Treat the register named @var{reg} as a fixed register; generated code
24476 should never refer to it (except perhaps as a stack pointer, frame
24477 pointer or in some other fixed role).
24479 @var{reg} must be the name of a register. The register names accepted
24480 are machine-specific and are defined in the @code{REGISTER_NAMES}
24481 macro in the machine description macro file.
24483 This flag does not have a negative form, because it specifies a
24486 @item -fcall-used-@var{reg}
24487 @opindex fcall-used
24488 Treat the register named @var{reg} as an allocable register that is
24489 clobbered by function calls. It may be allocated for temporaries or
24490 variables that do not live across a call. Functions compiled this way
24491 do not save and restore the register @var{reg}.
24493 It is an error to use this flag with the frame pointer or stack pointer.
24494 Use of this flag for other registers that have fixed pervasive roles in
24495 the machine's execution model produces disastrous results.
24497 This flag does not have a negative form, because it specifies a
24500 @item -fcall-saved-@var{reg}
24501 @opindex fcall-saved
24502 Treat the register named @var{reg} as an allocable register saved by
24503 functions. It may be allocated even for temporaries or variables that
24504 live across a call. Functions compiled this way save and restore
24505 the register @var{reg} if they use it.
24507 It is an error to use this flag with the frame pointer or stack pointer.
24508 Use of this flag for other registers that have fixed pervasive roles in
24509 the machine's execution model produces disastrous results.
24511 A different sort of disaster results from the use of this flag for
24512 a register in which function values may be returned.
24514 This flag does not have a negative form, because it specifies a
24517 @item -fpack-struct[=@var{n}]
24518 @opindex fpack-struct
24519 Without a value specified, pack all structure members together without
24520 holes. When a value is specified (which must be a small power of two), pack
24521 structure members according to this value, representing the maximum
24522 alignment (that is, objects with default alignment requirements larger than
24523 this are output potentially unaligned at the next fitting location.
24525 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
24526 code that is not binary compatible with code generated without that switch.
24527 Additionally, it makes the code suboptimal.
24528 Use it to conform to a non-default application binary interface.
24530 @item -finstrument-functions
24531 @opindex finstrument-functions
24532 Generate instrumentation calls for entry and exit to functions. Just
24533 after function entry and just before function exit, the following
24534 profiling functions are called with the address of the current
24535 function and its call site. (On some platforms,
24536 @code{__builtin_return_address} does not work beyond the current
24537 function, so the call site information may not be available to the
24538 profiling functions otherwise.)
24541 void __cyg_profile_func_enter (void *this_fn,
24543 void __cyg_profile_func_exit (void *this_fn,
24547 The first argument is the address of the start of the current function,
24548 which may be looked up exactly in the symbol table.
24550 This instrumentation is also done for functions expanded inline in other
24551 functions. The profiling calls indicate where, conceptually, the
24552 inline function is entered and exited. This means that addressable
24553 versions of such functions must be available. If all your uses of a
24554 function are expanded inline, this may mean an additional expansion of
24555 code size. If you use @code{extern inline} in your C code, an
24556 addressable version of such functions must be provided. (This is
24557 normally the case anyway, but if you get lucky and the optimizer always
24558 expands the functions inline, you might have gotten away without
24559 providing static copies.)
24561 A function may be given the attribute @code{no_instrument_function}, in
24562 which case this instrumentation is not done. This can be used, for
24563 example, for the profiling functions listed above, high-priority
24564 interrupt routines, and any functions from which the profiling functions
24565 cannot safely be called (perhaps signal handlers, if the profiling
24566 routines generate output or allocate memory).
24568 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
24569 @opindex finstrument-functions-exclude-file-list
24571 Set the list of functions that are excluded from instrumentation (see
24572 the description of @option{-finstrument-functions}). If the file that
24573 contains a function definition matches with one of @var{file}, then
24574 that function is not instrumented. The match is done on substrings:
24575 if the @var{file} parameter is a substring of the file name, it is
24576 considered to be a match.
24581 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
24585 excludes any inline function defined in files whose pathnames
24586 contain @file{/bits/stl} or @file{include/sys}.
24588 If, for some reason, you want to include letter @samp{,} in one of
24589 @var{sym}, write @samp{\,}. For example,
24590 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
24591 (note the single quote surrounding the option).
24593 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
24594 @opindex finstrument-functions-exclude-function-list
24596 This is similar to @option{-finstrument-functions-exclude-file-list},
24597 but this option sets the list of function names to be excluded from
24598 instrumentation. The function name to be matched is its user-visible
24599 name, such as @code{vector<int> blah(const vector<int> &)}, not the
24600 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
24601 match is done on substrings: if the @var{sym} parameter is a substring
24602 of the function name, it is considered to be a match. For C99 and C++
24603 extended identifiers, the function name must be given in UTF-8, not
24604 using universal character names.
24606 @item -fstack-check
24607 @opindex fstack-check
24608 Generate code to verify that you do not go beyond the boundary of the
24609 stack. You should specify this flag if you are running in an
24610 environment with multiple threads, but you only rarely need to specify it in
24611 a single-threaded environment since stack overflow is automatically
24612 detected on nearly all systems if there is only one stack.
24614 Note that this switch does not actually cause checking to be done; the
24615 operating system or the language runtime must do that. The switch causes
24616 generation of code to ensure that they see the stack being extended.
24618 You can additionally specify a string parameter: @samp{no} means no
24619 checking, @samp{generic} means force the use of old-style checking,
24620 @samp{specific} means use the best checking method and is equivalent
24621 to bare @option{-fstack-check}.
24623 Old-style checking is a generic mechanism that requires no specific
24624 target support in the compiler but comes with the following drawbacks:
24628 Modified allocation strategy for large objects: they are always
24629 allocated dynamically if their size exceeds a fixed threshold.
24632 Fixed limit on the size of the static frame of functions: when it is
24633 topped by a particular function, stack checking is not reliable and
24634 a warning is issued by the compiler.
24637 Inefficiency: because of both the modified allocation strategy and the
24638 generic implementation, code performance is hampered.
24641 Note that old-style stack checking is also the fallback method for
24642 @samp{specific} if no target support has been added in the compiler.
24644 @item -fstack-limit-register=@var{reg}
24645 @itemx -fstack-limit-symbol=@var{sym}
24646 @itemx -fno-stack-limit
24647 @opindex fstack-limit-register
24648 @opindex fstack-limit-symbol
24649 @opindex fno-stack-limit
24650 Generate code to ensure that the stack does not grow beyond a certain value,
24651 either the value of a register or the address of a symbol. If a larger
24652 stack is required, a signal is raised at run time. For most targets,
24653 the signal is raised before the stack overruns the boundary, so
24654 it is possible to catch the signal without taking special precautions.
24656 For instance, if the stack starts at absolute address @samp{0x80000000}
24657 and grows downwards, you can use the flags
24658 @option{-fstack-limit-symbol=__stack_limit} and
24659 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
24660 of 128KB@. Note that this may only work with the GNU linker.
24662 You can locally override stack limit checking by using the
24663 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
24665 @item -fsplit-stack
24666 @opindex fsplit-stack
24667 Generate code to automatically split the stack before it overflows.
24668 The resulting program has a discontiguous stack which can only
24669 overflow if the program is unable to allocate any more memory. This
24670 is most useful when running threaded programs, as it is no longer
24671 necessary to calculate a good stack size to use for each thread. This
24672 is currently only implemented for the x86 targets running
24675 When code compiled with @option{-fsplit-stack} calls code compiled
24676 without @option{-fsplit-stack}, there may not be much stack space
24677 available for the latter code to run. If compiling all code,
24678 including library code, with @option{-fsplit-stack} is not an option,
24679 then the linker can fix up these calls so that the code compiled
24680 without @option{-fsplit-stack} always has a large stack. Support for
24681 this is implemented in the gold linker in GNU binutils release 2.21
24684 @item -fleading-underscore
24685 @opindex fleading-underscore
24686 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
24687 change the way C symbols are represented in the object file. One use
24688 is to help link with legacy assembly code.
24690 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
24691 generate code that is not binary compatible with code generated without that
24692 switch. Use it to conform to a non-default application binary interface.
24693 Not all targets provide complete support for this switch.
24695 @item -ftls-model=@var{model}
24696 @opindex ftls-model
24697 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
24698 The @var{model} argument should be one of @samp{global-dynamic},
24699 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
24700 Note that the choice is subject to optimization: the compiler may use
24701 a more efficient model for symbols not visible outside of the translation
24702 unit, or if @option{-fpic} is not given on the command line.
24704 The default without @option{-fpic} is @samp{initial-exec}; with
24705 @option{-fpic} the default is @samp{global-dynamic}.
24707 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
24708 @opindex fvisibility
24709 Set the default ELF image symbol visibility to the specified option---all
24710 symbols are marked with this unless overridden within the code.
24711 Using this feature can very substantially improve linking and
24712 load times of shared object libraries, produce more optimized
24713 code, provide near-perfect API export and prevent symbol clashes.
24714 It is @strong{strongly} recommended that you use this in any shared objects
24717 Despite the nomenclature, @samp{default} always means public; i.e.,
24718 available to be linked against from outside the shared object.
24719 @samp{protected} and @samp{internal} are pretty useless in real-world
24720 usage so the only other commonly used option is @samp{hidden}.
24721 The default if @option{-fvisibility} isn't specified is
24722 @samp{default}, i.e., make every symbol public.
24724 A good explanation of the benefits offered by ensuring ELF
24725 symbols have the correct visibility is given by ``How To Write
24726 Shared Libraries'' by Ulrich Drepper (which can be found at
24727 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
24728 solution made possible by this option to marking things hidden when
24729 the default is public is to make the default hidden and mark things
24730 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
24731 and @code{__attribute__ ((visibility("default")))} instead of
24732 @code{__declspec(dllexport)} you get almost identical semantics with
24733 identical syntax. This is a great boon to those working with
24734 cross-platform projects.
24736 For those adding visibility support to existing code, you may find
24737 @code{#pragma GCC visibility} of use. This works by you enclosing
24738 the declarations you wish to set visibility for with (for example)
24739 @code{#pragma GCC visibility push(hidden)} and
24740 @code{#pragma GCC visibility pop}.
24741 Bear in mind that symbol visibility should be viewed @strong{as
24742 part of the API interface contract} and thus all new code should
24743 always specify visibility when it is not the default; i.e., declarations
24744 only for use within the local DSO should @strong{always} be marked explicitly
24745 as hidden as so to avoid PLT indirection overheads---making this
24746 abundantly clear also aids readability and self-documentation of the code.
24747 Note that due to ISO C++ specification requirements, @code{operator new} and
24748 @code{operator delete} must always be of default visibility.
24750 Be aware that headers from outside your project, in particular system
24751 headers and headers from any other library you use, may not be
24752 expecting to be compiled with visibility other than the default. You
24753 may need to explicitly say @code{#pragma GCC visibility push(default)}
24754 before including any such headers.
24756 @code{extern} declarations are not affected by @option{-fvisibility}, so
24757 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24758 no modifications. However, this means that calls to @code{extern}
24759 functions with no explicit visibility use the PLT, so it is more
24760 effective to use @code{__attribute ((visibility))} and/or
24761 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24762 declarations should be treated as hidden.
24764 Note that @option{-fvisibility} does affect C++ vague linkage
24765 entities. This means that, for instance, an exception class that is
24766 be thrown between DSOs must be explicitly marked with default
24767 visibility so that the @samp{type_info} nodes are unified between
24770 An overview of these techniques, their benefits and how to use them
24771 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24773 @item -fstrict-volatile-bitfields
24774 @opindex fstrict-volatile-bitfields
24775 This option should be used if accesses to volatile bit-fields (or other
24776 structure fields, although the compiler usually honors those types
24777 anyway) should use a single access of the width of the
24778 field's type, aligned to a natural alignment if possible. For
24779 example, targets with memory-mapped peripheral registers might require
24780 all such accesses to be 16 bits wide; with this flag you can
24781 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24782 is 16 bits on these targets) to force GCC to use 16-bit accesses
24783 instead of, perhaps, a more efficient 32-bit access.
24785 If this option is disabled, the compiler uses the most efficient
24786 instruction. In the previous example, that might be a 32-bit load
24787 instruction, even though that accesses bytes that do not contain
24788 any portion of the bit-field, or memory-mapped registers unrelated to
24789 the one being updated.
24791 In some cases, such as when the @code{packed} attribute is applied to a
24792 structure field, it may not be possible to access the field with a single
24793 read or write that is correctly aligned for the target machine. In this
24794 case GCC falls back to generating multiple accesses rather than code that
24795 will fault or truncate the result at run time.
24797 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24798 not allowed to touch non bit-field members. It is therefore recommended
24799 to define all bits of the field's type as bit-field members.
24801 The default value of this option is determined by the application binary
24802 interface for the target processor.
24804 @item -fsync-libcalls
24805 @opindex fsync-libcalls
24806 This option controls whether any out-of-line instance of the @code{__sync}
24807 family of functions may be used to implement the C++11 @code{__atomic}
24808 family of functions.
24810 The default value of this option is enabled, thus the only useful form
24811 of the option is @option{-fno-sync-libcalls}. This option is used in
24812 the implementation of the @file{libatomic} runtime library.
24818 @node Environment Variables
24819 @section Environment Variables Affecting GCC
24820 @cindex environment variables
24822 @c man begin ENVIRONMENT
24823 This section describes several environment variables that affect how GCC
24824 operates. Some of them work by specifying directories or prefixes to use
24825 when searching for various kinds of files. Some are used to specify other
24826 aspects of the compilation environment.
24828 Note that you can also specify places to search using options such as
24829 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24830 take precedence over places specified using environment variables, which
24831 in turn take precedence over those specified by the configuration of GCC@.
24832 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24833 GNU Compiler Collection (GCC) Internals}.
24838 @c @itemx LC_COLLATE
24840 @c @itemx LC_MONETARY
24841 @c @itemx LC_NUMERIC
24846 @c @findex LC_COLLATE
24847 @findex LC_MESSAGES
24848 @c @findex LC_MONETARY
24849 @c @findex LC_NUMERIC
24853 These environment variables control the way that GCC uses
24854 localization information which allows GCC to work with different
24855 national conventions. GCC inspects the locale categories
24856 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24857 so. These locale categories can be set to any value supported by your
24858 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24859 Kingdom encoded in UTF-8.
24861 The @env{LC_CTYPE} environment variable specifies character
24862 classification. GCC uses it to determine the character boundaries in
24863 a string; this is needed for some multibyte encodings that contain quote
24864 and escape characters that are otherwise interpreted as a string
24867 The @env{LC_MESSAGES} environment variable specifies the language to
24868 use in diagnostic messages.
24870 If the @env{LC_ALL} environment variable is set, it overrides the value
24871 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24872 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24873 environment variable. If none of these variables are set, GCC
24874 defaults to traditional C English behavior.
24878 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24879 files. GCC uses temporary files to hold the output of one stage of
24880 compilation which is to be used as input to the next stage: for example,
24881 the output of the preprocessor, which is the input to the compiler
24884 @item GCC_COMPARE_DEBUG
24885 @findex GCC_COMPARE_DEBUG
24886 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24887 @option{-fcompare-debug} to the compiler driver. See the documentation
24888 of this option for more details.
24890 @item GCC_EXEC_PREFIX
24891 @findex GCC_EXEC_PREFIX
24892 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24893 names of the subprograms executed by the compiler. No slash is added
24894 when this prefix is combined with the name of a subprogram, but you can
24895 specify a prefix that ends with a slash if you wish.
24897 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24898 an appropriate prefix to use based on the pathname it is invoked with.
24900 If GCC cannot find the subprogram using the specified prefix, it
24901 tries looking in the usual places for the subprogram.
24903 The default value of @env{GCC_EXEC_PREFIX} is
24904 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24905 the installed compiler. In many cases @var{prefix} is the value
24906 of @code{prefix} when you ran the @file{configure} script.
24908 Other prefixes specified with @option{-B} take precedence over this prefix.
24910 This prefix is also used for finding files such as @file{crt0.o} that are
24913 In addition, the prefix is used in an unusual way in finding the
24914 directories to search for header files. For each of the standard
24915 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24916 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24917 replacing that beginning with the specified prefix to produce an
24918 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24919 @file{foo/bar} just before it searches the standard directory
24920 @file{/usr/local/lib/bar}.
24921 If a standard directory begins with the configured
24922 @var{prefix} then the value of @var{prefix} is replaced by
24923 @env{GCC_EXEC_PREFIX} when looking for header files.
24925 @item COMPILER_PATH
24926 @findex COMPILER_PATH
24927 The value of @env{COMPILER_PATH} is a colon-separated list of
24928 directories, much like @env{PATH}. GCC tries the directories thus
24929 specified when searching for subprograms, if it can't find the
24930 subprograms using @env{GCC_EXEC_PREFIX}.
24933 @findex LIBRARY_PATH
24934 The value of @env{LIBRARY_PATH} is a colon-separated list of
24935 directories, much like @env{PATH}. When configured as a native compiler,
24936 GCC tries the directories thus specified when searching for special
24937 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24938 using GCC also uses these directories when searching for ordinary
24939 libraries for the @option{-l} option (but directories specified with
24940 @option{-L} come first).
24944 @cindex locale definition
24945 This variable is used to pass locale information to the compiler. One way in
24946 which this information is used is to determine the character set to be used
24947 when character literals, string literals and comments are parsed in C and C++.
24948 When the compiler is configured to allow multibyte characters,
24949 the following values for @env{LANG} are recognized:
24953 Recognize JIS characters.
24955 Recognize SJIS characters.
24957 Recognize EUCJP characters.
24960 If @env{LANG} is not defined, or if it has some other value, then the
24961 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24962 recognize and translate multibyte characters.
24966 Some additional environment variables affect the behavior of the
24969 @include cppenv.texi
24973 @node Precompiled Headers
24974 @section Using Precompiled Headers
24975 @cindex precompiled headers
24976 @cindex speed of compilation
24978 Often large projects have many header files that are included in every
24979 source file. The time the compiler takes to process these header files
24980 over and over again can account for nearly all of the time required to
24981 build the project. To make builds faster, GCC allows you to
24982 @dfn{precompile} a header file.
24984 To create a precompiled header file, simply compile it as you would any
24985 other file, if necessary using the @option{-x} option to make the driver
24986 treat it as a C or C++ header file. You may want to use a
24987 tool like @command{make} to keep the precompiled header up-to-date when
24988 the headers it contains change.
24990 A precompiled header file is searched for when @code{#include} is
24991 seen in the compilation. As it searches for the included file
24992 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24993 compiler looks for a precompiled header in each directory just before it
24994 looks for the include file in that directory. The name searched for is
24995 the name specified in the @code{#include} with @samp{.gch} appended. If
24996 the precompiled header file can't be used, it is ignored.
24998 For instance, if you have @code{#include "all.h"}, and you have
24999 @file{all.h.gch} in the same directory as @file{all.h}, then the
25000 precompiled header file is used if possible, and the original
25001 header is used otherwise.
25003 Alternatively, you might decide to put the precompiled header file in a
25004 directory and use @option{-I} to ensure that directory is searched
25005 before (or instead of) the directory containing the original header.
25006 Then, if you want to check that the precompiled header file is always
25007 used, you can put a file of the same name as the original header in this
25008 directory containing an @code{#error} command.
25010 This also works with @option{-include}. So yet another way to use
25011 precompiled headers, good for projects not designed with precompiled
25012 header files in mind, is to simply take most of the header files used by
25013 a project, include them from another header file, precompile that header
25014 file, and @option{-include} the precompiled header. If the header files
25015 have guards against multiple inclusion, they are skipped because
25016 they've already been included (in the precompiled header).
25018 If you need to precompile the same header file for different
25019 languages, targets, or compiler options, you can instead make a
25020 @emph{directory} named like @file{all.h.gch}, and put each precompiled
25021 header in the directory, perhaps using @option{-o}. It doesn't matter
25022 what you call the files in the directory; every precompiled header in
25023 the directory is considered. The first precompiled header
25024 encountered in the directory that is valid for this compilation is
25025 used; they're searched in no particular order.
25027 There are many other possibilities, limited only by your imagination,
25028 good sense, and the constraints of your build system.
25030 A precompiled header file can be used only when these conditions apply:
25034 Only one precompiled header can be used in a particular compilation.
25037 A precompiled header can't be used once the first C token is seen. You
25038 can have preprocessor directives before a precompiled header; you cannot
25039 include a precompiled header from inside another header.
25042 The precompiled header file must be produced for the same language as
25043 the current compilation. You can't use a C precompiled header for a C++
25047 The precompiled header file must have been produced by the same compiler
25048 binary as the current compilation is using.
25051 Any macros defined before the precompiled header is included must
25052 either be defined in the same way as when the precompiled header was
25053 generated, or must not affect the precompiled header, which usually
25054 means that they don't appear in the precompiled header at all.
25056 The @option{-D} option is one way to define a macro before a
25057 precompiled header is included; using a @code{#define} can also do it.
25058 There are also some options that define macros implicitly, like
25059 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
25062 @item If debugging information is output when using the precompiled
25063 header, using @option{-g} or similar, the same kind of debugging information
25064 must have been output when building the precompiled header. However,
25065 a precompiled header built using @option{-g} can be used in a compilation
25066 when no debugging information is being output.
25068 @item The same @option{-m} options must generally be used when building
25069 and using the precompiled header. @xref{Submodel Options},
25070 for any cases where this rule is relaxed.
25072 @item Each of the following options must be the same when building and using
25073 the precompiled header:
25075 @gccoptlist{-fexceptions}
25078 Some other command-line options starting with @option{-f},
25079 @option{-p}, or @option{-O} must be defined in the same way as when
25080 the precompiled header was generated. At present, it's not clear
25081 which options are safe to change and which are not; the safest choice
25082 is to use exactly the same options when generating and using the
25083 precompiled header. The following are known to be safe:
25085 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
25086 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
25087 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
25092 For all of these except the last, the compiler automatically
25093 ignores the precompiled header if the conditions aren't met. If you
25094 find an option combination that doesn't work and doesn't cause the
25095 precompiled header to be ignored, please consider filing a bug report,
25098 If you do use differing options when generating and using the
25099 precompiled header, the actual behavior is a mixture of the
25100 behavior for the options. For instance, if you use @option{-g} to
25101 generate the precompiled header but not when using it, you may or may
25102 not get debugging information for routines in the precompiled header.