1 @c Copyright (C) 1988-2015 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2015 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
76 Other options are passed on to one stage of processing. Some options
77 control the preprocessor and others the compiler itself. Yet other
78 options control the assembler and linker; most of these are not
79 documented here, since you rarely need to use any of them.
81 @cindex C compilation options
82 Most of the command-line options that you can use with GCC are useful
83 for C programs; when an option is only useful with another language
84 (usually C++), the explanation says so explicitly. If the description
85 for a particular option does not mention a source language, you can use
86 that option with all supported languages.
88 @cindex C++ compilation options
89 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
90 options for compiling C++ programs.
92 @cindex grouping options
93 @cindex options, grouping
94 The @command{gcc} program accepts options and file names as operands. Many
95 options have multi-letter names; therefore multiple single-letter options
96 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
99 @cindex order of options
100 @cindex options, order
101 You can mix options and other arguments. For the most part, the order
102 you use doesn't matter. Order does matter when you use several
103 options of the same kind; for example, if you specify @option{-L} more
104 than once, the directories are searched in the order specified. Also,
105 the placement of the @option{-l} option is significant.
107 Many options have long names starting with @samp{-f} or with
108 @samp{-W}---for example,
109 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
110 these have both positive and negative forms; the negative form of
111 @option{-ffoo} is @option{-fno-foo}. This manual documents
112 only one of these two forms, whichever one is not the default.
116 @xref{Option Index}, for an index to GCC's options.
119 * Option Summary:: Brief list of all options, without explanations.
120 * Overall Options:: Controlling the kind of output:
121 an executable, object files, assembler files,
122 or preprocessed source.
123 * Invoking G++:: Compiling C++ programs.
124 * C Dialect Options:: Controlling the variant of C language compiled.
125 * C++ Dialect Options:: Variations on C++.
126 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
128 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
130 * Warning Options:: How picky should the compiler be?
131 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
132 * Optimize Options:: How much optimization?
133 * Preprocessor Options:: Controlling header files and macro definitions.
134 Also, getting dependency information for Make.
135 * Assembler Options:: Passing options to the assembler.
136 * Link Options:: Specifying libraries and so on.
137 * Directory Options:: Where to find header files and libraries.
138 Where to find the compiler executable files.
139 * Spec Files:: How to pass switches to sub-processes.
140 * Target Options:: Running a cross-compiler, or an old version of GCC.
141 * Submodel Options:: Specifying minor hardware or convention variations,
142 such as 68010 vs 68020.
143 * Code Gen Options:: Specifying conventions for function calls, data layout
145 * Environment Variables:: Env vars that affect GCC.
146 * Precompiled Headers:: Compiling a header once, and using it many times.
152 @section Option Summary
154 Here is a summary of all the options, grouped by type. Explanations are
155 in the following sections.
158 @item Overall Options
159 @xref{Overall Options,,Options Controlling the Kind of Output}.
160 @gccoptlist{-c -S -E -o @var{file} -no-canonical-prefixes @gol
161 -pipe -pass-exit-codes @gol
162 -x @var{language} -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help @gol
163 --version -wrapper @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
164 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
166 @item C Language Options
167 @xref{C Dialect Options,,Options Controlling C Dialect}.
168 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
169 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
170 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
171 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
172 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness}
173 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
174 -fsigned-bitfields -fsigned-char @gol
175 -funsigned-bitfields -funsigned-char @gol
176 -trigraphs -traditional -traditional-cpp}
178 @item C++ Language Options
179 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
180 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
181 -fconstexpr-depth=@var{n} -ffriend-injection @gol
182 -fno-elide-constructors @gol
183 -fno-enforce-eh-specs @gol
184 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
185 -fno-implicit-templates @gol
186 -fno-implicit-inline-templates @gol
187 -fno-implement-inlines -fms-extensions @gol
188 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
189 -fno-optional-diags -fpermissive @gol
190 -fno-pretty-templates @gol
191 -frepo -fno-rtti -fsized-deallocation @gol
192 -fstats -ftemplate-backtrace-limit=@var{n} @gol
193 -ftemplate-depth=@var{n} @gol
194 -fno-threadsafe-statics -fuse-cxa-atexit @gol
195 -fno-weak -nostdinc++ @gol
196 -fvisibility-inlines-hidden @gol
197 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
198 -fvtv-counts -fvtv-debug @gol
199 -fvisibility-ms-compat @gol
200 -fext-numeric-literals @gol
201 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
202 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
203 -Wnamespaces -Wnarrowing @gol
204 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
205 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
206 -Wno-non-template-friend -Wold-style-cast @gol
207 -Woverloaded-virtual -Wno-pmf-conversions @gol
208 -Wsign-promo -Wvirtual-inheritance}
210 @item Objective-C and Objective-C++ Language Options
211 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
212 Objective-C and Objective-C++ Dialects}.
213 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
214 -fgnu-runtime -fnext-runtime @gol
215 -fno-nil-receivers @gol
216 -fobjc-abi-version=@var{n} @gol
217 -fobjc-call-cxx-cdtors @gol
218 -fobjc-direct-dispatch @gol
219 -fobjc-exceptions @gol
222 -fobjc-std=objc1 @gol
223 -fno-local-ivars @gol
224 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
225 -freplace-objc-classes @gol
228 -Wassign-intercept @gol
229 -Wno-protocol -Wselector @gol
230 -Wstrict-selector-match @gol
231 -Wundeclared-selector}
233 @item Diagnostic Message Formatting Options
234 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
235 @gccoptlist{-fmessage-length=@var{n} @gol
236 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
237 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
238 -fno-diagnostics-show-option -fno-diagnostics-show-caret}
240 @item Warning Options
241 @xref{Warning Options,,Options to Request or Suppress Warnings}.
242 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
243 -pedantic-errors @gol
244 -w -Wextra -Wall -Waddress -Waggregate-return @gol
245 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
246 -Wno-attributes -Wbool-compare -Wno-builtin-macro-redefined @gol
247 -Wc90-c99-compat -Wc99-c11-compat @gol
248 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
249 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
250 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdate-time -Wdelete-incomplete @gol
251 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
252 -Wdisabled-optimization @gol
253 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
254 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
255 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
256 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
257 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
258 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
259 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
260 -Wignored-qualifiers -Wincompatible-pointer-types @gol
261 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
262 -Winit-self -Winline -Wno-int-conversion @gol
263 -Wno-int-to-pointer-cast -Wno-invalid-offsetof @gol
264 -Winvalid-pch -Wlarger-than=@var{len} @gol
265 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
266 -Wmain -Wmaybe-uninitialized -Wmemset-transposed-args @gol
267 -Wmisleading-indentation -Wmissing-braces @gol
268 -Wmissing-field-initializers -Wmissing-include-dirs @gol
269 -Wno-multichar -Wnonnull -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
270 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
271 -Woverride-init-side-effects -Woverlength-strings @gol
272 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
273 -Wparentheses -Wno-pedantic-ms-format @gol
274 -Wplacement-new -Wpointer-arith -Wno-pointer-to-int-cast @gol
275 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
276 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
277 -Wshift-overflow -Wshift-overflow=@var{n} @gol
278 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
279 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
280 -Wno-scalar-storage-order @gol
281 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
282 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
283 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
284 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
285 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
286 -Wmissing-format-attribute -Wsubobject-linkage @gol
287 -Wswitch -Wswitch-default -Wswitch-enum -Wswitch-bool -Wsync-nand @gol
288 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
289 -Wtype-limits -Wundef @gol
290 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
291 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
292 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
293 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
294 -Wunused-const-variable @gol
295 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
296 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
297 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
298 -Wzero-as-null-pointer-constant}
300 @item C and Objective-C-only Warning Options
301 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
302 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
303 -Wold-style-declaration -Wold-style-definition @gol
304 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
305 -Wdeclaration-after-statement -Wpointer-sign}
307 @item Debugging Options
308 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
309 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
310 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
311 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
312 -fsanitize-undefined-trap-on-error @gol
313 -fcheck-pointer-bounds -fchecking -fchkp-check-incomplete-type @gol
314 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
315 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
316 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
317 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
318 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
319 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
320 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
321 -fchkp-use-wrappers @gol
322 -fdbg-cnt-list -fdbg-cnt=@var{counter-value-list} @gol
323 -fdisable-ipa-@var{pass_name} @gol
324 -fdisable-rtl-@var{pass_name} @gol
325 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
326 -fdisable-tree-@var{pass_name} @gol
327 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
328 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
329 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
330 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
331 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
333 -fdump-statistics @gol
335 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
336 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
337 -fdump-tree-cfg -fdump-tree-alias @gol
339 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
340 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
341 -fdump-tree-gimple@r{[}-raw@r{]} @gol
342 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
343 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
344 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
345 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
346 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
347 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
348 -fdump-tree-nrv -fdump-tree-vect @gol
349 -fdump-tree-sink @gol
350 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
351 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
352 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
353 -fdump-tree-vtable-verify @gol
354 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
355 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
356 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
357 -fdump-final-insns=@var{file} @gol
358 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
359 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
360 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
361 -fenable-@var{kind}-@var{pass} @gol
362 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
363 -fdebug-types-section -fmem-report-wpa @gol
364 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
366 -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
367 -frandom-seed=@var{number} -fsched-verbose=@var{n} @gol
368 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
369 -fstack-usage -ftest-coverage -ftime-report -fvar-tracking @gol
370 -fvar-tracking-assignments -fvar-tracking-assignments-toggle @gol
371 -g -g@var{level} -gtoggle -gcoff -gdwarf-@var{version} @gol
372 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
373 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
374 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
375 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
376 -fdebug-prefix-map=@var{old}=@var{new} @gol
377 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
378 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
379 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
380 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
381 -print-prog-name=@var{program} -print-search-dirs -Q @gol
382 -print-sysroot -print-sysroot-headers-suffix @gol
383 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
385 @item Optimization Options
386 @xref{Optimize Options,,Options that Control Optimization}.
387 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
388 -falign-jumps[=@var{n}] @gol
389 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
390 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
391 -fauto-inc-dec -fbranch-probabilities @gol
392 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
393 -fbtr-bb-exclusive -fcaller-saves @gol
394 -fcombine-stack-adjustments -fconserve-stack @gol
395 -fcompare-elim -fcprop-registers -fcrossjumping @gol
396 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
397 -fcx-limited-range @gol
398 -fdata-sections -fdce -fdelayed-branch @gol
399 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
400 -fdevirtualize-at-ltrans -fdse @gol
401 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
402 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
403 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
404 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
405 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
406 -fif-conversion2 -findirect-inlining @gol
407 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
408 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
409 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
410 -fira-algorithm=@var{algorithm} @gol
411 -fira-region=@var{region} -fira-hoist-pressure @gol
412 -fira-loop-pressure -fno-ira-share-save-slots @gol
413 -fno-ira-share-spill-slots -fira-verbose=@var{n} @gol
414 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
415 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
416 -fkeep-static-consts -flive-range-shrinkage @gol
417 -floop-block -floop-interchange -floop-strip-mine @gol
418 -floop-unroll-and-jam -floop-nest-optimize @gol
419 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
420 -flto-partition=@var{alg} -flto-report -flto-report-wpa -fmerge-all-constants @gol
421 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
422 -fmove-loop-invariants -fno-branch-count-reg @gol
423 -fno-defer-pop -fno-function-cse -fno-guess-branch-probability @gol
424 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
425 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
426 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
427 -fomit-frame-pointer -foptimize-sibling-calls @gol
428 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
429 -fprefetch-loop-arrays -fprofile-report @gol
430 -fprofile-correction -fprofile-dir=@var{path} -fprofile-generate @gol
431 -fprofile-generate=@var{path} @gol
432 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
433 -fprofile-reorder-functions @gol
434 -freciprocal-math -free -frename-registers -freorder-blocks @gol
435 -freorder-blocks-algorithm=@var{algorithm} @gol
436 -freorder-blocks-and-partition -freorder-functions @gol
437 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
438 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
439 -fsched-spec-load -fsched-spec-load-dangerous @gol
440 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
441 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
442 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
443 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
444 -fschedule-fusion @gol
445 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
446 -fselective-scheduling -fselective-scheduling2 @gol
447 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
448 -fsemantic-interposition -fshrink-wrap -fsignaling-nans @gol
449 -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
451 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
452 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
453 -fstack-protector-explicit -fstdarg-opt -fstrict-aliasing @gol
454 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
455 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
456 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
457 -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
458 -ftree-loop-if-convert-stores -ftree-loop-im @gol
459 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
460 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
461 -ftree-loop-vectorize @gol
462 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
463 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
464 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
465 -ftree-vectorize -ftree-vrp @gol
466 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
467 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
468 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
469 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
470 --param @var{name}=@var{value}
471 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
473 @item Preprocessor Options
474 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
475 @gccoptlist{-A@var{question}=@var{answer} @gol
476 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
477 -C -dD -dI -dM -dN @gol
478 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
479 -idirafter @var{dir} @gol
480 -include @var{file} -imacros @var{file} @gol
481 -iprefix @var{file} -iwithprefix @var{dir} @gol
482 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
483 -imultilib @var{dir} -isysroot @var{dir} @gol
484 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
485 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
486 -remap -trigraphs -undef -U@var{macro} @gol
487 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
489 @item Assembler Option
490 @xref{Assembler Options,,Passing Options to the Assembler}.
491 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
494 @xref{Link Options,,Options for Linking}.
495 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
496 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
497 -s -static -static-libgcc -static-libstdc++ @gol
498 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
499 -static-libmpx -static-libmpxwrappers @gol
500 -shared -shared-libgcc -symbolic @gol
501 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
502 -u @var{symbol} -z @var{keyword}}
504 @item Directory Options
505 @xref{Directory Options,,Options for Directory Search}.
506 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
507 -iquote@var{dir} -L@var{dir} -specs=@var{file} -I- @gol
508 --sysroot=@var{dir} --no-sysroot-suffix}
510 @item Machine Dependent Options
511 @xref{Submodel Options,,Hardware Models and Configurations}.
512 @c This list is ordered alphanumerically by subsection name.
513 @c Try and put the significant identifier (CPU or system) first,
514 @c so users have a clue at guessing where the ones they want will be.
516 @emph{AArch64 Options}
517 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
518 -mgeneral-regs-only @gol
519 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
521 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
522 -mtls-dialect=desc -mtls-dialect=traditional @gol
523 -mtls-size=@var{size} @gol
524 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
525 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
526 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
527 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
529 @emph{Adapteva Epiphany Options}
530 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
531 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
532 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
533 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
534 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
535 -msplit-vecmove-early -m1reg-@var{reg}}
538 @gccoptlist{-mbarrel-shifter @gol
539 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
540 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
541 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
542 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
543 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
544 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
545 -mlong-calls -mmedium-calls -msdata @gol
546 -mucb-mcount -mvolatile-cache @gol
547 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
548 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
549 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
550 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
551 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
552 -mtune=@var{cpu} -mmultcost=@var{num} @gol
553 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
554 -mdiv-rem -mcode-density}
557 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
558 -mabi=@var{name} @gol
559 -mapcs-stack-check -mno-apcs-stack-check @gol
560 -mapcs-float -mno-apcs-float @gol
561 -mapcs-reentrant -mno-apcs-reentrant @gol
562 -msched-prolog -mno-sched-prolog @gol
563 -mlittle-endian -mbig-endian @gol
564 -mfloat-abi=@var{name} @gol
565 -mfp16-format=@var{name}
566 -mthumb-interwork -mno-thumb-interwork @gol
567 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
568 -mtune=@var{name} -mprint-tune-info @gol
569 -mstructure-size-boundary=@var{n} @gol
570 -mabort-on-noreturn @gol
571 -mlong-calls -mno-long-calls @gol
572 -msingle-pic-base -mno-single-pic-base @gol
573 -mpic-register=@var{reg} @gol
574 -mnop-fun-dllimport @gol
575 -mpoke-function-name @gol
577 -mtpcs-frame -mtpcs-leaf-frame @gol
578 -mcaller-super-interworking -mcallee-super-interworking @gol
579 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
580 -mword-relocations @gol
581 -mfix-cortex-m3-ldrd @gol
582 -munaligned-access @gol
583 -mneon-for-64bits @gol
584 -mslow-flash-data @gol
585 -masm-syntax-unified @gol
589 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
590 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
591 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert}
593 @emph{Blackfin Options}
594 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
595 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
596 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
597 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
598 -mno-id-shared-library -mshared-library-id=@var{n} @gol
599 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
600 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
601 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
605 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
606 -msim -msdata=@var{sdata-type}}
609 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
610 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
611 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
612 -mstack-align -mdata-align -mconst-align @gol
613 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
614 -melf -maout -melinux -mlinux -sim -sim2 @gol
615 -mmul-bug-workaround -mno-mul-bug-workaround}
618 @gccoptlist{-mmac @gol
619 -mcr16cplus -mcr16c @gol
620 -msim -mint32 -mbit-ops
621 -mdata-model=@var{model}}
623 @emph{Darwin Options}
624 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
625 -arch_only -bind_at_load -bundle -bundle_loader @gol
626 -client_name -compatibility_version -current_version @gol
628 -dependency-file -dylib_file -dylinker_install_name @gol
629 -dynamic -dynamiclib -exported_symbols_list @gol
630 -filelist -flat_namespace -force_cpusubtype_ALL @gol
631 -force_flat_namespace -headerpad_max_install_names @gol
633 -image_base -init -install_name -keep_private_externs @gol
634 -multi_module -multiply_defined -multiply_defined_unused @gol
635 -noall_load -no_dead_strip_inits_and_terms @gol
636 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
637 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
638 -private_bundle -read_only_relocs -sectalign @gol
639 -sectobjectsymbols -whyload -seg1addr @gol
640 -sectcreate -sectobjectsymbols -sectorder @gol
641 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
642 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
643 -segprot -segs_read_only_addr -segs_read_write_addr @gol
644 -single_module -static -sub_library -sub_umbrella @gol
645 -twolevel_namespace -umbrella -undefined @gol
646 -unexported_symbols_list -weak_reference_mismatches @gol
647 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
648 -mkernel -mone-byte-bool}
650 @emph{DEC Alpha Options}
651 @gccoptlist{-mno-fp-regs -msoft-float @gol
652 -mieee -mieee-with-inexact -mieee-conformant @gol
653 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
654 -mtrap-precision=@var{mode} -mbuild-constants @gol
655 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
656 -mbwx -mmax -mfix -mcix @gol
657 -mfloat-vax -mfloat-ieee @gol
658 -mexplicit-relocs -msmall-data -mlarge-data @gol
659 -msmall-text -mlarge-text @gol
660 -mmemory-latency=@var{time}}
663 @gccoptlist{-msmall-model -mno-lsim}
666 @gccoptlist{-msim -mlra}
669 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
670 -mhard-float -msoft-float @gol
671 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
672 -mdouble -mno-double @gol
673 -mmedia -mno-media -mmuladd -mno-muladd @gol
674 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
675 -mlinked-fp -mlong-calls -malign-labels @gol
676 -mlibrary-pic -macc-4 -macc-8 @gol
677 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
678 -moptimize-membar -mno-optimize-membar @gol
679 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
680 -mvliw-branch -mno-vliw-branch @gol
681 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
682 -mno-nested-cond-exec -mtomcat-stats @gol
686 @emph{GNU/Linux Options}
687 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
688 -tno-android-cc -tno-android-ld}
690 @emph{H8/300 Options}
691 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
694 @gccoptlist{-march=@var{architecture-type} @gol
695 -mdisable-fpregs -mdisable-indexing @gol
696 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
697 -mfixed-range=@var{register-range} @gol
698 -mjump-in-delay -mlinker-opt -mlong-calls @gol
699 -mlong-load-store -mno-disable-fpregs @gol
700 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
701 -mno-jump-in-delay -mno-long-load-store @gol
702 -mno-portable-runtime -mno-soft-float @gol
703 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
704 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
705 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
706 -munix=@var{unix-std} -nolibdld -static -threads}
709 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
710 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
711 -mconstant-gp -mauto-pic -mfused-madd @gol
712 -minline-float-divide-min-latency @gol
713 -minline-float-divide-max-throughput @gol
714 -mno-inline-float-divide @gol
715 -minline-int-divide-min-latency @gol
716 -minline-int-divide-max-throughput @gol
717 -mno-inline-int-divide @gol
718 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
719 -mno-inline-sqrt @gol
720 -mdwarf2-asm -mearly-stop-bits @gol
721 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
722 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
723 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
724 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
725 -msched-spec-ldc -msched-spec-control-ldc @gol
726 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
727 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
728 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
729 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
732 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
733 -msign-extend-enabled -muser-enabled}
735 @emph{M32R/D Options}
736 @gccoptlist{-m32r2 -m32rx -m32r @gol
738 -malign-loops -mno-align-loops @gol
739 -missue-rate=@var{number} @gol
740 -mbranch-cost=@var{number} @gol
741 -mmodel=@var{code-size-model-type} @gol
742 -msdata=@var{sdata-type} @gol
743 -mno-flush-func -mflush-func=@var{name} @gol
744 -mno-flush-trap -mflush-trap=@var{number} @gol
748 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
750 @emph{M680x0 Options}
751 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
752 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
753 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
754 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
755 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
756 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
757 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
758 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
762 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
763 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
764 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
765 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
766 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
769 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
770 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
771 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
772 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
775 @emph{MicroBlaze Options}
776 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
777 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
778 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
779 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
780 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
783 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
784 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
785 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
786 -mips16 -mno-mips16 -mflip-mips16 @gol
787 -minterlink-compressed -mno-interlink-compressed @gol
788 -minterlink-mips16 -mno-interlink-mips16 @gol
789 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
790 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
791 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
792 -mno-float -msingle-float -mdouble-float @gol
793 -modd-spreg -mno-odd-spreg @gol
794 -mabs=@var{mode} -mnan=@var{encoding} @gol
795 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
798 -mvirt -mno-virt @gol
800 -mmicromips -mno-micromips @gol
801 -mfpu=@var{fpu-type} @gol
802 -msmartmips -mno-smartmips @gol
803 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
804 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
805 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
806 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
807 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
808 -membedded-data -mno-embedded-data @gol
809 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
810 -mcode-readable=@var{setting} @gol
811 -msplit-addresses -mno-split-addresses @gol
812 -mexplicit-relocs -mno-explicit-relocs @gol
813 -mcheck-zero-division -mno-check-zero-division @gol
814 -mdivide-traps -mdivide-breaks @gol
815 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
816 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
817 -mfix-24k -mno-fix-24k @gol
818 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
819 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
820 -mfix-vr4120 -mno-fix-vr4120 @gol
821 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
822 -mflush-func=@var{func} -mno-flush-func @gol
823 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
824 -mcompact-branches=@var{policy} @gol
825 -mfp-exceptions -mno-fp-exceptions @gol
826 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
827 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
828 -mframe-header-opt -mno-frame-header-opt}
831 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
832 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
833 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
834 -mno-base-addresses -msingle-exit -mno-single-exit}
836 @emph{MN10300 Options}
837 @gccoptlist{-mmult-bug -mno-mult-bug @gol
838 -mno-am33 -mam33 -mam33-2 -mam34 @gol
839 -mtune=@var{cpu-type} @gol
840 -mreturn-pointer-on-d0 @gol
841 -mno-crt0 -mrelax -mliw -msetlb}
844 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
846 @emph{MSP430 Options}
847 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
849 -mcode-region= -mdata-region= @gol
850 -msilicon-errata= -msilicon-errata-warn= @gol
854 @gccoptlist{-mbig-endian -mlittle-endian @gol
855 -mreduced-regs -mfull-regs @gol
856 -mcmov -mno-cmov @gol
857 -mperf-ext -mno-perf-ext @gol
858 -mv3push -mno-v3push @gol
859 -m16bit -mno-16bit @gol
860 -misr-vector-size=@var{num} @gol
861 -mcache-block-size=@var{num} @gol
862 -march=@var{arch} @gol
863 -mcmodel=@var{code-model} @gol
866 @emph{Nios II Options}
867 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
869 -mno-bypass-cache -mbypass-cache @gol
870 -mno-cache-volatile -mcache-volatile @gol
871 -mno-fast-sw-div -mfast-sw-div @gol
872 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
873 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
874 -mcustom-fpu-cfg=@var{name} @gol
875 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
876 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
878 @emph{Nvidia PTX Options}
879 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
881 @emph{PDP-11 Options}
882 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
883 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
884 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
885 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
886 -mbranch-expensive -mbranch-cheap @gol
887 -munix-asm -mdec-asm}
889 @emph{picoChip Options}
890 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
891 -msymbol-as-address -mno-inefficient-warnings}
893 @emph{PowerPC Options}
894 See RS/6000 and PowerPC Options.
897 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
898 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
899 -m64bit-doubles -m32bit-doubles}
901 @emph{RS/6000 and PowerPC Options}
902 @gccoptlist{-mcpu=@var{cpu-type} @gol
903 -mtune=@var{cpu-type} @gol
904 -mcmodel=@var{code-model} @gol
906 -maltivec -mno-altivec @gol
907 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
908 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
909 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
910 -mfprnd -mno-fprnd @gol
911 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
912 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
913 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
914 -malign-power -malign-natural @gol
915 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
916 -msingle-float -mdouble-float -msimple-fpu @gol
917 -mstring -mno-string -mupdate -mno-update @gol
918 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
919 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
920 -mstrict-align -mno-strict-align -mrelocatable @gol
921 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
922 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
923 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
924 -mprioritize-restricted-insns=@var{priority} @gol
925 -msched-costly-dep=@var{dependence_type} @gol
926 -minsert-sched-nops=@var{scheme} @gol
927 -mcall-sysv -mcall-netbsd @gol
928 -maix-struct-return -msvr4-struct-return @gol
929 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
930 -mblock-move-inline-limit=@var{num} @gol
931 -misel -mno-isel @gol
932 -misel=yes -misel=no @gol
934 -mspe=yes -mspe=no @gol
936 -mgen-cell-microcode -mwarn-cell-microcode @gol
937 -mvrsave -mno-vrsave @gol
938 -mmulhw -mno-mulhw @gol
939 -mdlmzb -mno-dlmzb @gol
940 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
941 -mprototype -mno-prototype @gol
942 -msim -mmvme -mads -myellowknife -memb -msdata @gol
943 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
944 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
945 -mno-recip-precision @gol
946 -mveclibabi=@var{type} -mfriz -mno-friz @gol
947 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
948 -msave-toc-indirect -mno-save-toc-indirect @gol
949 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
950 -mcrypto -mno-crypto -mdirect-move -mno-direct-move @gol
951 -mquad-memory -mno-quad-memory @gol
952 -mquad-memory-atomic -mno-quad-memory-atomic @gol
953 -mcompat-align-parm -mno-compat-align-parm @gol
954 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
955 -mupper-regs -mno-upper-regs -mmodulo -mno-modulo @gol
956 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
957 -mpower9-fusion -mno-mpower9-fusion -mpower9-vector -mno-power9-vector}
960 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
962 -mbig-endian-data -mlittle-endian-data @gol
965 -mas100-syntax -mno-as100-syntax@gol
967 -mmax-constant-size=@gol
970 -mallow-string-insns -mno-allow-string-insns@gol
972 -mno-warn-multiple-fast-interrupts@gol
973 -msave-acc-in-interrupts}
975 @emph{S/390 and zSeries Options}
976 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
977 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
978 -mlong-double-64 -mlong-double-128 @gol
979 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
980 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
981 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
982 -mhtm -mvx -mzvector @gol
983 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
984 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
985 -mhotpatch=@var{halfwords},@var{halfwords}}
988 @gccoptlist{-meb -mel @gol
992 -mscore5 -mscore5u -mscore7 -mscore7d}
995 @gccoptlist{-m1 -m2 -m2e @gol
996 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
998 -m4-nofpu -m4-single-only -m4-single -m4 @gol
999 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1000 -mb -ml -mdalign -mrelax @gol
1001 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1002 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1003 -mspace -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1004 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1005 -maccumulate-outgoing-args @gol
1006 -matomic-model=@var{atomic-model} @gol
1007 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1008 -mcbranch-force-delay-slot @gol
1009 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1010 -mpretend-cmove -mtas}
1012 @emph{Solaris 2 Options}
1013 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1016 @emph{SPARC Options}
1017 @gccoptlist{-mcpu=@var{cpu-type} @gol
1018 -mtune=@var{cpu-type} @gol
1019 -mcmodel=@var{code-model} @gol
1020 -mmemory-model=@var{mem-model} @gol
1021 -m32 -m64 -mapp-regs -mno-app-regs @gol
1022 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1023 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1024 -mhard-quad-float -msoft-quad-float @gol
1025 -mstack-bias -mno-stack-bias @gol
1026 -mstd-struct-return -mno-std-struct-return @gol
1027 -munaligned-doubles -mno-unaligned-doubles @gol
1028 -muser-mode -mno-user-mode @gol
1029 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1030 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1031 -mcbcond -mno-cbcond @gol
1032 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1033 -mfix-at697f -mfix-ut699}
1036 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1037 -msafe-dma -munsafe-dma @gol
1039 -msmall-mem -mlarge-mem -mstdmain @gol
1040 -mfixed-range=@var{register-range} @gol
1042 -maddress-space-conversion -mno-address-space-conversion @gol
1043 -mcache-size=@var{cache-size} @gol
1044 -matomic-updates -mno-atomic-updates}
1046 @emph{System V Options}
1047 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1049 @emph{TILE-Gx Options}
1050 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1051 -mcmodel=@var{code-model}}
1053 @emph{TILEPro Options}
1054 @gccoptlist{-mcpu=@var{cpu} -m32}
1057 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1058 -mprolog-function -mno-prolog-function -mspace @gol
1059 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1060 -mapp-regs -mno-app-regs @gol
1061 -mdisable-callt -mno-disable-callt @gol
1062 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1063 -mv850e -mv850 -mv850e3v5 @gol
1074 @gccoptlist{-mg -mgnu -munix}
1076 @emph{Visium Options}
1077 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1078 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1081 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1082 -mpointer-size=@var{size}}
1084 @emph{VxWorks Options}
1085 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1086 -Xbind-lazy -Xbind-now}
1089 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1090 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1091 -mfpmath=@var{unit} @gol
1092 -masm=@var{dialect} -mno-fancy-math-387 @gol
1093 -mno-fp-ret-in-387 -msoft-float @gol
1094 -mno-wide-multiply -mrtd -malign-double @gol
1095 -mpreferred-stack-boundary=@var{num} @gol
1096 -mincoming-stack-boundary=@var{num} @gol
1097 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1098 -mrecip -mrecip=@var{opt} @gol
1099 -mvzeroupper -mprefer-avx128 @gol
1100 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1101 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1102 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1103 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1104 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1105 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1106 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1107 -mpku -mthreads @gol
1108 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1109 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1110 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1111 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1112 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1113 -mregparm=@var{num} -msseregparm @gol
1114 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1115 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1116 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1117 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1118 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1119 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1120 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1121 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1124 @emph{x86 Windows Options}
1125 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1126 -mnop-fun-dllimport -mthread @gol
1127 -municode -mwin32 -mwindows -fno-set-stack-executable}
1129 @emph{Xstormy16 Options}
1132 @emph{Xtensa Options}
1133 @gccoptlist{-mconst16 -mno-const16 @gol
1134 -mfused-madd -mno-fused-madd @gol
1136 -mserialize-volatile -mno-serialize-volatile @gol
1137 -mtext-section-literals -mno-text-section-literals @gol
1138 -mauto-litpools -mno-auto-litpools @gol
1139 -mtarget-align -mno-target-align @gol
1140 -mlongcalls -mno-longcalls}
1142 @emph{zSeries Options}
1143 See S/390 and zSeries Options.
1145 @item Code Generation Options
1146 @xref{Code Gen Options,,Options for Code Generation Conventions}.
1147 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
1148 -ffixed-@var{reg} -fexceptions @gol
1149 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
1150 -fasynchronous-unwind-tables @gol
1151 -fno-gnu-unique @gol
1152 -finhibit-size-directive -finstrument-functions @gol
1153 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
1154 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{} @gol
1155 -fno-common -fno-ident @gol
1156 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
1157 -fno-jump-tables @gol
1158 -frecord-gcc-switches @gol
1159 -freg-struct-return -fshort-enums @gol
1160 -fshort-double -fshort-wchar @gol
1161 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
1162 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
1163 -fno-stack-limit -fsplit-stack @gol
1164 -fleading-underscore -ftls-model=@var{model} @gol
1165 -fstack-reuse=@var{reuse_level} @gol
1166 -ftrapv -fwrapv -fbounds-check @gol
1167 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
1168 -fstrict-volatile-bitfields -fsync-libcalls}
1172 @node Overall Options
1173 @section Options Controlling the Kind of Output
1175 Compilation can involve up to four stages: preprocessing, compilation
1176 proper, assembly and linking, always in that order. GCC is capable of
1177 preprocessing and compiling several files either into several
1178 assembler input files, or into one assembler input file; then each
1179 assembler input file produces an object file, and linking combines all
1180 the object files (those newly compiled, and those specified as input)
1181 into an executable file.
1183 @cindex file name suffix
1184 For any given input file, the file name suffix determines what kind of
1185 compilation is done:
1189 C source code that must be preprocessed.
1192 C source code that should not be preprocessed.
1195 C++ source code that should not be preprocessed.
1198 Objective-C source code. Note that you must link with the @file{libobjc}
1199 library to make an Objective-C program work.
1202 Objective-C source code that should not be preprocessed.
1206 Objective-C++ source code. Note that you must link with the @file{libobjc}
1207 library to make an Objective-C++ program work. Note that @samp{.M} refers
1208 to a literal capital M@.
1210 @item @var{file}.mii
1211 Objective-C++ source code that should not be preprocessed.
1214 C, C++, Objective-C or Objective-C++ header file to be turned into a
1215 precompiled header (default), or C, C++ header file to be turned into an
1216 Ada spec (via the @option{-fdump-ada-spec} switch).
1219 @itemx @var{file}.cp
1220 @itemx @var{file}.cxx
1221 @itemx @var{file}.cpp
1222 @itemx @var{file}.CPP
1223 @itemx @var{file}.c++
1225 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1226 the last two letters must both be literally @samp{x}. Likewise,
1227 @samp{.C} refers to a literal capital C@.
1231 Objective-C++ source code that must be preprocessed.
1233 @item @var{file}.mii
1234 Objective-C++ source code that should not be preprocessed.
1238 @itemx @var{file}.hp
1239 @itemx @var{file}.hxx
1240 @itemx @var{file}.hpp
1241 @itemx @var{file}.HPP
1242 @itemx @var{file}.h++
1243 @itemx @var{file}.tcc
1244 C++ header file to be turned into a precompiled header or Ada spec.
1247 @itemx @var{file}.for
1248 @itemx @var{file}.ftn
1249 Fixed form Fortran source code that should not be preprocessed.
1252 @itemx @var{file}.FOR
1253 @itemx @var{file}.fpp
1254 @itemx @var{file}.FPP
1255 @itemx @var{file}.FTN
1256 Fixed form Fortran source code that must be preprocessed (with the traditional
1259 @item @var{file}.f90
1260 @itemx @var{file}.f95
1261 @itemx @var{file}.f03
1262 @itemx @var{file}.f08
1263 Free form Fortran source code that should not be preprocessed.
1265 @item @var{file}.F90
1266 @itemx @var{file}.F95
1267 @itemx @var{file}.F03
1268 @itemx @var{file}.F08
1269 Free form Fortran source code that must be preprocessed (with the
1270 traditional preprocessor).
1275 @c FIXME: Descriptions of Java file types.
1281 @item @var{file}.ads
1282 Ada source code file that contains a library unit declaration (a
1283 declaration of a package, subprogram, or generic, or a generic
1284 instantiation), or a library unit renaming declaration (a package,
1285 generic, or subprogram renaming declaration). Such files are also
1288 @item @var{file}.adb
1289 Ada source code file containing a library unit body (a subprogram or
1290 package body). Such files are also called @dfn{bodies}.
1292 @c GCC also knows about some suffixes for languages not yet included:
1303 @itemx @var{file}.sx
1304 Assembler code that must be preprocessed.
1307 An object file to be fed straight into linking.
1308 Any file name with no recognized suffix is treated this way.
1312 You can specify the input language explicitly with the @option{-x} option:
1315 @item -x @var{language}
1316 Specify explicitly the @var{language} for the following input files
1317 (rather than letting the compiler choose a default based on the file
1318 name suffix). This option applies to all following input files until
1319 the next @option{-x} option. Possible values for @var{language} are:
1321 c c-header cpp-output
1322 c++ c++-header c++-cpp-output
1323 objective-c objective-c-header objective-c-cpp-output
1324 objective-c++ objective-c++-header objective-c++-cpp-output
1325 assembler assembler-with-cpp
1327 f77 f77-cpp-input f95 f95-cpp-input
1333 Turn off any specification of a language, so that subsequent files are
1334 handled according to their file name suffixes (as they are if @option{-x}
1335 has not been used at all).
1337 @item -pass-exit-codes
1338 @opindex pass-exit-codes
1339 Normally the @command{gcc} program exits with the code of 1 if any
1340 phase of the compiler returns a non-success return code. If you specify
1341 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1342 the numerically highest error produced by any phase returning an error
1343 indication. The C, C++, and Fortran front ends return 4 if an internal
1344 compiler error is encountered.
1347 If you only want some of the stages of compilation, you can use
1348 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1349 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1350 @command{gcc} is to stop. Note that some combinations (for example,
1351 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1356 Compile or assemble the source files, but do not link. The linking
1357 stage simply is not done. The ultimate output is in the form of an
1358 object file for each source file.
1360 By default, the object file name for a source file is made by replacing
1361 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1363 Unrecognized input files, not requiring compilation or assembly, are
1368 Stop after the stage of compilation proper; do not assemble. The output
1369 is in the form of an assembler code file for each non-assembler input
1372 By default, the assembler file name for a source file is made by
1373 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1375 Input files that don't require compilation are ignored.
1379 Stop after the preprocessing stage; do not run the compiler proper. The
1380 output is in the form of preprocessed source code, which is sent to the
1383 Input files that don't require preprocessing are ignored.
1385 @cindex output file option
1388 Place output in file @var{file}. This applies to whatever
1389 sort of output is being produced, whether it be an executable file,
1390 an object file, an assembler file or preprocessed C code.
1392 If @option{-o} is not specified, the default is to put an executable
1393 file in @file{a.out}, the object file for
1394 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1395 assembler file in @file{@var{source}.s}, a precompiled header file in
1396 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1401 Print (on standard error output) the commands executed to run the stages
1402 of compilation. Also print the version number of the compiler driver
1403 program and of the preprocessor and the compiler proper.
1407 Like @option{-v} except the commands are not executed and arguments
1408 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1409 This is useful for shell scripts to capture the driver-generated command lines.
1413 Use pipes rather than temporary files for communication between the
1414 various stages of compilation. This fails to work on some systems where
1415 the assembler is unable to read from a pipe; but the GNU assembler has
1420 Print (on the standard output) a description of the command-line options
1421 understood by @command{gcc}. If the @option{-v} option is also specified
1422 then @option{--help} is also passed on to the various processes
1423 invoked by @command{gcc}, so that they can display the command-line options
1424 they accept. If the @option{-Wextra} option has also been specified
1425 (prior to the @option{--help} option), then command-line options that
1426 have no documentation associated with them are also displayed.
1429 @opindex target-help
1430 Print (on the standard output) a description of target-specific command-line
1431 options for each tool. For some targets extra target-specific
1432 information may also be printed.
1434 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1435 Print (on the standard output) a description of the command-line
1436 options understood by the compiler that fit into all specified classes
1437 and qualifiers. These are the supported classes:
1440 @item @samp{optimizers}
1441 Display all of the optimization options supported by the
1444 @item @samp{warnings}
1445 Display all of the options controlling warning messages
1446 produced by the compiler.
1449 Display target-specific options. Unlike the
1450 @option{--target-help} option however, target-specific options of the
1451 linker and assembler are not displayed. This is because those
1452 tools do not currently support the extended @option{--help=} syntax.
1455 Display the values recognized by the @option{--param}
1458 @item @var{language}
1459 Display the options supported for @var{language}, where
1460 @var{language} is the name of one of the languages supported in this
1464 Display the options that are common to all languages.
1467 These are the supported qualifiers:
1470 @item @samp{undocumented}
1471 Display only those options that are undocumented.
1474 Display options taking an argument that appears after an equal
1475 sign in the same continuous piece of text, such as:
1476 @samp{--help=target}.
1478 @item @samp{separate}
1479 Display options taking an argument that appears as a separate word
1480 following the original option, such as: @samp{-o output-file}.
1483 Thus for example to display all the undocumented target-specific
1484 switches supported by the compiler, use:
1487 --help=target,undocumented
1490 The sense of a qualifier can be inverted by prefixing it with the
1491 @samp{^} character, so for example to display all binary warning
1492 options (i.e., ones that are either on or off and that do not take an
1493 argument) that have a description, use:
1496 --help=warnings,^joined,^undocumented
1499 The argument to @option{--help=} should not consist solely of inverted
1502 Combining several classes is possible, although this usually
1503 restricts the output so much that there is nothing to display. One
1504 case where it does work, however, is when one of the classes is
1505 @var{target}. For example, to display all the target-specific
1506 optimization options, use:
1509 --help=target,optimizers
1512 The @option{--help=} option can be repeated on the command line. Each
1513 successive use displays its requested class of options, skipping
1514 those that have already been displayed.
1516 If the @option{-Q} option appears on the command line before the
1517 @option{--help=} option, then the descriptive text displayed by
1518 @option{--help=} is changed. Instead of describing the displayed
1519 options, an indication is given as to whether the option is enabled,
1520 disabled or set to a specific value (assuming that the compiler
1521 knows this at the point where the @option{--help=} option is used).
1523 Here is a truncated example from the ARM port of @command{gcc}:
1526 % gcc -Q -mabi=2 --help=target -c
1527 The following options are target specific:
1529 -mabort-on-noreturn [disabled]
1533 The output is sensitive to the effects of previous command-line
1534 options, so for example it is possible to find out which optimizations
1535 are enabled at @option{-O2} by using:
1538 -Q -O2 --help=optimizers
1541 Alternatively you can discover which binary optimizations are enabled
1542 by @option{-O3} by using:
1545 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1546 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1547 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1550 @item -no-canonical-prefixes
1551 @opindex no-canonical-prefixes
1552 Do not expand any symbolic links, resolve references to @samp{/../}
1553 or @samp{/./}, or make the path absolute when generating a relative
1558 Display the version number and copyrights of the invoked GCC@.
1562 Invoke all subcommands under a wrapper program. The name of the
1563 wrapper program and its parameters are passed as a comma separated
1567 gcc -c t.c -wrapper gdb,--args
1571 This invokes all subprograms of @command{gcc} under
1572 @samp{gdb --args}, thus the invocation of @command{cc1} is
1573 @samp{gdb --args cc1 @dots{}}.
1575 @item -fplugin=@var{name}.so
1577 Load the plugin code in file @var{name}.so, assumed to be a
1578 shared object to be dlopen'd by the compiler. The base name of
1579 the shared object file is used to identify the plugin for the
1580 purposes of argument parsing (See
1581 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1582 Each plugin should define the callback functions specified in the
1585 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1586 @opindex fplugin-arg
1587 Define an argument called @var{key} with a value of @var{value}
1588 for the plugin called @var{name}.
1590 @item -fdump-ada-spec@r{[}-slim@r{]}
1591 @opindex fdump-ada-spec
1592 For C and C++ source and include files, generate corresponding Ada specs.
1593 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1594 GNAT User's Guide}, which provides detailed documentation on this feature.
1596 @item -fada-spec-parent=@var{unit}
1597 @opindex fada-spec-parent
1598 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1599 Ada specs as child units of parent @var{unit}.
1601 @item -fdump-go-spec=@var{file}
1602 @opindex fdump-go-spec
1603 For input files in any language, generate corresponding Go
1604 declarations in @var{file}. This generates Go @code{const},
1605 @code{type}, @code{var}, and @code{func} declarations which may be a
1606 useful way to start writing a Go interface to code written in some
1609 @include @value{srcdir}/../libiberty/at-file.texi
1613 @section Compiling C++ Programs
1615 @cindex suffixes for C++ source
1616 @cindex C++ source file suffixes
1617 C++ source files conventionally use one of the suffixes @samp{.C},
1618 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1619 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1620 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1621 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1622 files with these names and compiles them as C++ programs even if you
1623 call the compiler the same way as for compiling C programs (usually
1624 with the name @command{gcc}).
1628 However, the use of @command{gcc} does not add the C++ library.
1629 @command{g++} is a program that calls GCC and automatically specifies linking
1630 against the C++ library. It treats @samp{.c},
1631 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1632 files unless @option{-x} is used. This program is also useful when
1633 precompiling a C header file with a @samp{.h} extension for use in C++
1634 compilations. On many systems, @command{g++} is also installed with
1635 the name @command{c++}.
1637 @cindex invoking @command{g++}
1638 When you compile C++ programs, you may specify many of the same
1639 command-line options that you use for compiling programs in any
1640 language; or command-line options meaningful for C and related
1641 languages; or options that are meaningful only for C++ programs.
1642 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1643 explanations of options for languages related to C@.
1644 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1645 explanations of options that are meaningful only for C++ programs.
1647 @node C Dialect Options
1648 @section Options Controlling C Dialect
1649 @cindex dialect options
1650 @cindex language dialect options
1651 @cindex options, dialect
1653 The following options control the dialect of C (or languages derived
1654 from C, such as C++, Objective-C and Objective-C++) that the compiler
1658 @cindex ANSI support
1662 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1663 equivalent to @option{-std=c++98}.
1665 This turns off certain features of GCC that are incompatible with ISO
1666 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1667 such as the @code{asm} and @code{typeof} keywords, and
1668 predefined macros such as @code{unix} and @code{vax} that identify the
1669 type of system you are using. It also enables the undesirable and
1670 rarely used ISO trigraph feature. For the C compiler,
1671 it disables recognition of C++ style @samp{//} comments as well as
1672 the @code{inline} keyword.
1674 The alternate keywords @code{__asm__}, @code{__extension__},
1675 @code{__inline__} and @code{__typeof__} continue to work despite
1676 @option{-ansi}. You would not want to use them in an ISO C program, of
1677 course, but it is useful to put them in header files that might be included
1678 in compilations done with @option{-ansi}. Alternate predefined macros
1679 such as @code{__unix__} and @code{__vax__} are also available, with or
1680 without @option{-ansi}.
1682 The @option{-ansi} option does not cause non-ISO programs to be
1683 rejected gratuitously. For that, @option{-Wpedantic} is required in
1684 addition to @option{-ansi}. @xref{Warning Options}.
1686 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1687 option is used. Some header files may notice this macro and refrain
1688 from declaring certain functions or defining certain macros that the
1689 ISO standard doesn't call for; this is to avoid interfering with any
1690 programs that might use these names for other things.
1692 Functions that are normally built in but do not have semantics
1693 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1694 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1695 built-in functions provided by GCC}, for details of the functions
1700 Determine the language standard. @xref{Standards,,Language Standards
1701 Supported by GCC}, for details of these standard versions. This option
1702 is currently only supported when compiling C or C++.
1704 The compiler can accept several base standards, such as @samp{c90} or
1705 @samp{c++98}, and GNU dialects of those standards, such as
1706 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1707 compiler accepts all programs following that standard plus those
1708 using GNU extensions that do not contradict it. For example,
1709 @option{-std=c90} turns off certain features of GCC that are
1710 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1711 keywords, but not other GNU extensions that do not have a meaning in
1712 ISO C90, such as omitting the middle term of a @code{?:}
1713 expression. On the other hand, when a GNU dialect of a standard is
1714 specified, all features supported by the compiler are enabled, even when
1715 those features change the meaning of the base standard. As a result, some
1716 strict-conforming programs may be rejected. The particular standard
1717 is used by @option{-Wpedantic} to identify which features are GNU
1718 extensions given that version of the standard. For example
1719 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1720 comments, while @option{-std=gnu99 -Wpedantic} does not.
1722 A value for this option must be provided; possible values are
1728 Support all ISO C90 programs (certain GNU extensions that conflict
1729 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1731 @item iso9899:199409
1732 ISO C90 as modified in amendment 1.
1738 ISO C99. This standard is substantially completely supported, modulo
1739 bugs and floating-point issues
1740 (mainly but not entirely relating to optional C99 features from
1741 Annexes F and G). See
1742 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1743 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1748 ISO C11, the 2011 revision of the ISO C standard. This standard is
1749 substantially completely supported, modulo bugs, floating-point issues
1750 (mainly but not entirely relating to optional C11 features from
1751 Annexes F and G) and the optional Annexes K (Bounds-checking
1752 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1756 GNU dialect of ISO C90 (including some C99 features).
1760 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1764 GNU dialect of ISO C11. This is the default for C code.
1765 The name @samp{gnu1x} is deprecated.
1769 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1770 additional defect reports. Same as @option{-ansi} for C++ code.
1774 GNU dialect of @option{-std=c++98}. This is the default for
1779 The 2011 ISO C++ standard plus amendments.
1780 The name @samp{c++0x} is deprecated.
1784 GNU dialect of @option{-std=c++11}.
1785 The name @samp{gnu++0x} is deprecated.
1789 The 2014 ISO C++ standard plus amendments.
1790 The name @samp{c++1y} is deprecated.
1794 GNU dialect of @option{-std=c++14}.
1795 The name @samp{gnu++1y} is deprecated.
1798 The next revision of the ISO C++ standard, tentatively planned for
1799 2017. Support is highly experimental, and will almost certainly
1800 change in incompatible ways in future releases.
1803 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1804 and will almost certainly change in incompatible ways in future
1808 @item -fgnu89-inline
1809 @opindex fgnu89-inline
1810 The option @option{-fgnu89-inline} tells GCC to use the traditional
1811 GNU semantics for @code{inline} functions when in C99 mode.
1812 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1813 Using this option is roughly equivalent to adding the
1814 @code{gnu_inline} function attribute to all inline functions
1815 (@pxref{Function Attributes}).
1817 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1818 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1819 specifies the default behavior).
1820 This option is not supported in @option{-std=c90} or
1821 @option{-std=gnu90} mode.
1823 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1824 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1825 in effect for @code{inline} functions. @xref{Common Predefined
1826 Macros,,,cpp,The C Preprocessor}.
1828 @item -aux-info @var{filename}
1830 Output to the given filename prototyped declarations for all functions
1831 declared and/or defined in a translation unit, including those in header
1832 files. This option is silently ignored in any language other than C@.
1834 Besides declarations, the file indicates, in comments, the origin of
1835 each declaration (source file and line), whether the declaration was
1836 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1837 @samp{O} for old, respectively, in the first character after the line
1838 number and the colon), and whether it came from a declaration or a
1839 definition (@samp{C} or @samp{F}, respectively, in the following
1840 character). In the case of function definitions, a K&R-style list of
1841 arguments followed by their declarations is also provided, inside
1842 comments, after the declaration.
1844 @item -fallow-parameterless-variadic-functions
1845 @opindex fallow-parameterless-variadic-functions
1846 Accept variadic functions without named parameters.
1848 Although it is possible to define such a function, this is not very
1849 useful as it is not possible to read the arguments. This is only
1850 supported for C as this construct is allowed by C++.
1854 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1855 keyword, so that code can use these words as identifiers. You can use
1856 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1857 instead. @option{-ansi} implies @option{-fno-asm}.
1859 In C++, this switch only affects the @code{typeof} keyword, since
1860 @code{asm} and @code{inline} are standard keywords. You may want to
1861 use the @option{-fno-gnu-keywords} flag instead, which has the same
1862 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1863 switch only affects the @code{asm} and @code{typeof} keywords, since
1864 @code{inline} is a standard keyword in ISO C99.
1867 @itemx -fno-builtin-@var{function}
1868 @opindex fno-builtin
1869 @cindex built-in functions
1870 Don't recognize built-in functions that do not begin with
1871 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1872 functions provided by GCC}, for details of the functions affected,
1873 including those which are not built-in functions when @option{-ansi} or
1874 @option{-std} options for strict ISO C conformance are used because they
1875 do not have an ISO standard meaning.
1877 GCC normally generates special code to handle certain built-in functions
1878 more efficiently; for instance, calls to @code{alloca} may become single
1879 instructions which adjust the stack directly, and calls to @code{memcpy}
1880 may become inline copy loops. The resulting code is often both smaller
1881 and faster, but since the function calls no longer appear as such, you
1882 cannot set a breakpoint on those calls, nor can you change the behavior
1883 of the functions by linking with a different library. In addition,
1884 when a function is recognized as a built-in function, GCC may use
1885 information about that function to warn about problems with calls to
1886 that function, or to generate more efficient code, even if the
1887 resulting code still contains calls to that function. For example,
1888 warnings are given with @option{-Wformat} for bad calls to
1889 @code{printf} when @code{printf} is built in and @code{strlen} is
1890 known not to modify global memory.
1892 With the @option{-fno-builtin-@var{function}} option
1893 only the built-in function @var{function} is
1894 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1895 function is named that is not built-in in this version of GCC, this
1896 option is ignored. There is no corresponding
1897 @option{-fbuiltin-@var{function}} option; if you wish to enable
1898 built-in functions selectively when using @option{-fno-builtin} or
1899 @option{-ffreestanding}, you may define macros such as:
1902 #define abs(n) __builtin_abs ((n))
1903 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1908 @cindex hosted environment
1910 Assert that compilation targets a hosted environment. This implies
1911 @option{-fbuiltin}. A hosted environment is one in which the
1912 entire standard library is available, and in which @code{main} has a return
1913 type of @code{int}. Examples are nearly everything except a kernel.
1914 This is equivalent to @option{-fno-freestanding}.
1916 @item -ffreestanding
1917 @opindex ffreestanding
1918 @cindex hosted environment
1920 Assert that compilation targets a freestanding environment. This
1921 implies @option{-fno-builtin}. A freestanding environment
1922 is one in which the standard library may not exist, and program startup may
1923 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1924 This is equivalent to @option{-fno-hosted}.
1926 @xref{Standards,,Language Standards Supported by GCC}, for details of
1927 freestanding and hosted environments.
1931 @cindex OpenACC accelerator programming
1932 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1933 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1934 compiler generates accelerated code according to the OpenACC Application
1935 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1936 implies @option{-pthread}, and thus is only supported on targets that
1937 have support for @option{-pthread}.
1939 Note that this is an experimental feature, incomplete, and subject to
1940 change in future versions of GCC. See
1941 @w{@uref{https://gcc.gnu.org/wiki/OpenACC}} for more information.
1945 @cindex OpenMP parallel
1946 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1947 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1948 compiler generates parallel code according to the OpenMP Application
1949 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1950 implies @option{-pthread}, and thus is only supported on targets that
1951 have support for @option{-pthread}. @option{-fopenmp} implies
1952 @option{-fopenmp-simd}.
1955 @opindex fopenmp-simd
1958 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
1959 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
1964 @cindex Enable Cilk Plus
1965 Enable the usage of Cilk Plus language extension features for C/C++.
1966 When the option @option{-fcilkplus} is specified, enable the usage of
1967 the Cilk Plus Language extension features for C/C++. The present
1968 implementation follows ABI version 1.2. This is an experimental
1969 feature that is only partially complete, and whose interface may
1970 change in future versions of GCC as the official specification
1971 changes. Currently, all features but @code{_Cilk_for} have been
1976 When the option @option{-fgnu-tm} is specified, the compiler
1977 generates code for the Linux variant of Intel's current Transactional
1978 Memory ABI specification document (Revision 1.1, May 6 2009). This is
1979 an experimental feature whose interface may change in future versions
1980 of GCC, as the official specification changes. Please note that not
1981 all architectures are supported for this feature.
1983 For more information on GCC's support for transactional memory,
1984 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
1985 Transactional Memory Library}.
1987 Note that the transactional memory feature is not supported with
1988 non-call exceptions (@option{-fnon-call-exceptions}).
1990 @item -fms-extensions
1991 @opindex fms-extensions
1992 Accept some non-standard constructs used in Microsoft header files.
1994 In C++ code, this allows member names in structures to be similar
1995 to previous types declarations.
2004 Some cases of unnamed fields in structures and unions are only
2005 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2006 fields within structs/unions}, for details.
2008 Note that this option is off for all targets but x86
2009 targets using ms-abi.
2011 @item -fplan9-extensions
2012 @opindex fplan9-extensions
2013 Accept some non-standard constructs used in Plan 9 code.
2015 This enables @option{-fms-extensions}, permits passing pointers to
2016 structures with anonymous fields to functions that expect pointers to
2017 elements of the type of the field, and permits referring to anonymous
2018 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2019 struct/union fields within structs/unions}, for details. This is only
2020 supported for C, not C++.
2024 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2025 options for strict ISO C conformance) implies @option{-trigraphs}.
2027 @cindex traditional C language
2028 @cindex C language, traditional
2030 @itemx -traditional-cpp
2031 @opindex traditional-cpp
2032 @opindex traditional
2033 Formerly, these options caused GCC to attempt to emulate a pre-standard
2034 C compiler. They are now only supported with the @option{-E} switch.
2035 The preprocessor continues to support a pre-standard mode. See the GNU
2036 CPP manual for details.
2038 @item -fcond-mismatch
2039 @opindex fcond-mismatch
2040 Allow conditional expressions with mismatched types in the second and
2041 third arguments. The value of such an expression is void. This option
2042 is not supported for C++.
2044 @item -flax-vector-conversions
2045 @opindex flax-vector-conversions
2046 Allow implicit conversions between vectors with differing numbers of
2047 elements and/or incompatible element types. This option should not be
2050 @item -funsigned-char
2051 @opindex funsigned-char
2052 Let the type @code{char} be unsigned, like @code{unsigned char}.
2054 Each kind of machine has a default for what @code{char} should
2055 be. It is either like @code{unsigned char} by default or like
2056 @code{signed char} by default.
2058 Ideally, a portable program should always use @code{signed char} or
2059 @code{unsigned char} when it depends on the signedness of an object.
2060 But many programs have been written to use plain @code{char} and
2061 expect it to be signed, or expect it to be unsigned, depending on the
2062 machines they were written for. This option, and its inverse, let you
2063 make such a program work with the opposite default.
2065 The type @code{char} is always a distinct type from each of
2066 @code{signed char} or @code{unsigned char}, even though its behavior
2067 is always just like one of those two.
2070 @opindex fsigned-char
2071 Let the type @code{char} be signed, like @code{signed char}.
2073 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2074 the negative form of @option{-funsigned-char}. Likewise, the option
2075 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2077 @item -fsigned-bitfields
2078 @itemx -funsigned-bitfields
2079 @itemx -fno-signed-bitfields
2080 @itemx -fno-unsigned-bitfields
2081 @opindex fsigned-bitfields
2082 @opindex funsigned-bitfields
2083 @opindex fno-signed-bitfields
2084 @opindex fno-unsigned-bitfields
2085 These options control whether a bit-field is signed or unsigned, when the
2086 declaration does not use either @code{signed} or @code{unsigned}. By
2087 default, such a bit-field is signed, because this is consistent: the
2088 basic integer types such as @code{int} are signed types.
2090 @item -fsso-struct=@var{endianness}
2091 @opindex fsso-struct
2092 Set the default scalar storage order of structures and unions to the
2093 specified endianness. The accepted values are @samp{big-endian} and
2094 @samp{little-endian}. If the option is not passed, the compiler uses
2095 the native endianness of the target. This option is not supported for C++.
2097 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2098 code that is not binary compatible with code generated without it if the
2099 specified endianness is not the native endianness of the target.
2102 @node C++ Dialect Options
2103 @section Options Controlling C++ Dialect
2105 @cindex compiler options, C++
2106 @cindex C++ options, command-line
2107 @cindex options, C++
2108 This section describes the command-line options that are only meaningful
2109 for C++ programs. You can also use most of the GNU compiler options
2110 regardless of what language your program is in. For example, you
2111 might compile a file @file{firstClass.C} like this:
2114 g++ -g -fstrict-enums -O -c firstClass.C
2118 In this example, only @option{-fstrict-enums} is an option meant
2119 only for C++ programs; you can use the other options with any
2120 language supported by GCC@.
2122 Here is a list of options that are @emph{only} for compiling C++ programs:
2126 @item -fabi-version=@var{n}
2127 @opindex fabi-version
2128 Use version @var{n} of the C++ ABI@. The default is version 0.
2130 Version 0 refers to the version conforming most closely to
2131 the C++ ABI specification. Therefore, the ABI obtained using version 0
2132 will change in different versions of G++ as ABI bugs are fixed.
2134 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2136 Version 2 is the version of the C++ ABI that first appeared in G++
2137 3.4, and was the default through G++ 4.9.
2139 Version 3 corrects an error in mangling a constant address as a
2142 Version 4, which first appeared in G++ 4.5, implements a standard
2143 mangling for vector types.
2145 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2146 attribute const/volatile on function pointer types, decltype of a
2147 plain decl, and use of a function parameter in the declaration of
2150 Version 6, which first appeared in G++ 4.7, corrects the promotion
2151 behavior of C++11 scoped enums and the mangling of template argument
2152 packs, const/static_cast, prefix ++ and --, and a class scope function
2153 used as a template argument.
2155 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2156 builtin type and corrects the mangling of lambdas in default argument
2159 Version 8, which first appeared in G++ 4.9, corrects the substitution
2160 behavior of function types with function-cv-qualifiers.
2162 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2165 Version 10, which first appeared in G++ 6.1, adds mangling of
2166 attributes that affect type identity, such as ia32 calling convention
2167 attributes (e.g. @samp{stdcall}).
2169 See also @option{-Wabi}.
2171 @item -fabi-compat-version=@var{n}
2172 @opindex fabi-compat-version
2173 On targets that support strong aliases, G++
2174 works around mangling changes by creating an alias with the correct
2175 mangled name when defining a symbol with an incorrect mangled name.
2176 This switch specifies which ABI version to use for the alias.
2178 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2179 compatibility). If another ABI version is explicitly selected, this
2180 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2181 use @option{-fabi-compat-version=2}.
2183 If this option is not provided but @option{-Wabi=@var{n}} is, that
2184 version is used for compatibility aliases. If this option is provided
2185 along with @option{-Wabi} (without the version), the version from this
2186 option is used for the warning.
2188 @item -fno-access-control
2189 @opindex fno-access-control
2190 Turn off all access checking. This switch is mainly useful for working
2191 around bugs in the access control code.
2195 Check that the pointer returned by @code{operator new} is non-null
2196 before attempting to modify the storage allocated. This check is
2197 normally unnecessary because the C++ standard specifies that
2198 @code{operator new} only returns @code{0} if it is declared
2199 @code{throw()}, in which case the compiler always checks the
2200 return value even without this option. In all other cases, when
2201 @code{operator new} has a non-empty exception specification, memory
2202 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2203 @samp{new (nothrow)}.
2205 @item -fconstexpr-depth=@var{n}
2206 @opindex fconstexpr-depth
2207 Set the maximum nested evaluation depth for C++11 constexpr functions
2208 to @var{n}. A limit is needed to detect endless recursion during
2209 constant expression evaluation. The minimum specified by the standard
2212 @item -fdeduce-init-list
2213 @opindex fdeduce-init-list
2214 Enable deduction of a template type parameter as
2215 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2218 template <class T> auto forward(T t) -> decltype (realfn (t))
2225 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2229 This deduction was implemented as a possible extension to the
2230 originally proposed semantics for the C++11 standard, but was not part
2231 of the final standard, so it is disabled by default. This option is
2232 deprecated, and may be removed in a future version of G++.
2234 @item -ffriend-injection
2235 @opindex ffriend-injection
2236 Inject friend functions into the enclosing namespace, so that they are
2237 visible outside the scope of the class in which they are declared.
2238 Friend functions were documented to work this way in the old Annotated
2239 C++ Reference Manual.
2240 However, in ISO C++ a friend function that is not declared
2241 in an enclosing scope can only be found using argument dependent
2242 lookup. GCC defaults to the standard behavior.
2244 This option is for compatibility, and may be removed in a future
2247 @item -fno-elide-constructors
2248 @opindex fno-elide-constructors
2249 The C++ standard allows an implementation to omit creating a temporary
2250 that is only used to initialize another object of the same type.
2251 Specifying this option disables that optimization, and forces G++ to
2252 call the copy constructor in all cases.
2254 @item -fno-enforce-eh-specs
2255 @opindex fno-enforce-eh-specs
2256 Don't generate code to check for violation of exception specifications
2257 at run time. This option violates the C++ standard, but may be useful
2258 for reducing code size in production builds, much like defining
2259 @code{NDEBUG}. This does not give user code permission to throw
2260 exceptions in violation of the exception specifications; the compiler
2261 still optimizes based on the specifications, so throwing an
2262 unexpected exception results in undefined behavior at run time.
2264 @item -fextern-tls-init
2265 @itemx -fno-extern-tls-init
2266 @opindex fextern-tls-init
2267 @opindex fno-extern-tls-init
2268 The C++11 and OpenMP standards allow @code{thread_local} and
2269 @code{threadprivate} variables to have dynamic (runtime)
2270 initialization. To support this, any use of such a variable goes
2271 through a wrapper function that performs any necessary initialization.
2272 When the use and definition of the variable are in the same
2273 translation unit, this overhead can be optimized away, but when the
2274 use is in a different translation unit there is significant overhead
2275 even if the variable doesn't actually need dynamic initialization. If
2276 the programmer can be sure that no use of the variable in a
2277 non-defining TU needs to trigger dynamic initialization (either
2278 because the variable is statically initialized, or a use of the
2279 variable in the defining TU will be executed before any uses in
2280 another TU), they can avoid this overhead with the
2281 @option{-fno-extern-tls-init} option.
2283 On targets that support symbol aliases, the default is
2284 @option{-fextern-tls-init}. On targets that do not support symbol
2285 aliases, the default is @option{-fno-extern-tls-init}.
2288 @itemx -fno-for-scope
2290 @opindex fno-for-scope
2291 If @option{-ffor-scope} is specified, the scope of variables declared in
2292 a @i{for-init-statement} is limited to the @code{for} loop itself,
2293 as specified by the C++ standard.
2294 If @option{-fno-for-scope} is specified, the scope of variables declared in
2295 a @i{for-init-statement} extends to the end of the enclosing scope,
2296 as was the case in old versions of G++, and other (traditional)
2297 implementations of C++.
2299 If neither flag is given, the default is to follow the standard,
2300 but to allow and give a warning for old-style code that would
2301 otherwise be invalid, or have different behavior.
2303 @item -fno-gnu-keywords
2304 @opindex fno-gnu-keywords
2305 Do not recognize @code{typeof} as a keyword, so that code can use this
2306 word as an identifier. You can use the keyword @code{__typeof__} instead.
2307 @option{-ansi} implies @option{-fno-gnu-keywords}.
2309 @item -fno-implicit-templates
2310 @opindex fno-implicit-templates
2311 Never emit code for non-inline templates that are instantiated
2312 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2313 @xref{Template Instantiation}, for more information.
2315 @item -fno-implicit-inline-templates
2316 @opindex fno-implicit-inline-templates
2317 Don't emit code for implicit instantiations of inline templates, either.
2318 The default is to handle inlines differently so that compiles with and
2319 without optimization need the same set of explicit instantiations.
2321 @item -fno-implement-inlines
2322 @opindex fno-implement-inlines
2323 To save space, do not emit out-of-line copies of inline functions
2324 controlled by @code{#pragma implementation}. This causes linker
2325 errors if these functions are not inlined everywhere they are called.
2327 @item -fms-extensions
2328 @opindex fms-extensions
2329 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2330 int and getting a pointer to member function via non-standard syntax.
2332 @item -fno-nonansi-builtins
2333 @opindex fno-nonansi-builtins
2334 Disable built-in declarations of functions that are not mandated by
2335 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2336 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2339 @opindex fnothrow-opt
2340 Treat a @code{throw()} exception specification as if it were a
2341 @code{noexcept} specification to reduce or eliminate the text size
2342 overhead relative to a function with no exception specification. If
2343 the function has local variables of types with non-trivial
2344 destructors, the exception specification actually makes the
2345 function smaller because the EH cleanups for those variables can be
2346 optimized away. The semantic effect is that an exception thrown out of
2347 a function with such an exception specification results in a call
2348 to @code{terminate} rather than @code{unexpected}.
2350 @item -fno-operator-names
2351 @opindex fno-operator-names
2352 Do not treat the operator name keywords @code{and}, @code{bitand},
2353 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2354 synonyms as keywords.
2356 @item -fno-optional-diags
2357 @opindex fno-optional-diags
2358 Disable diagnostics that the standard says a compiler does not need to
2359 issue. Currently, the only such diagnostic issued by G++ is the one for
2360 a name having multiple meanings within a class.
2363 @opindex fpermissive
2364 Downgrade some diagnostics about nonconformant code from errors to
2365 warnings. Thus, using @option{-fpermissive} allows some
2366 nonconforming code to compile.
2368 @item -fno-pretty-templates
2369 @opindex fno-pretty-templates
2370 When an error message refers to a specialization of a function
2371 template, the compiler normally prints the signature of the
2372 template followed by the template arguments and any typedefs or
2373 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2374 rather than @code{void f(int)}) so that it's clear which template is
2375 involved. When an error message refers to a specialization of a class
2376 template, the compiler omits any template arguments that match
2377 the default template arguments for that template. If either of these
2378 behaviors make it harder to understand the error message rather than
2379 easier, you can use @option{-fno-pretty-templates} to disable them.
2383 Enable automatic template instantiation at link time. This option also
2384 implies @option{-fno-implicit-templates}. @xref{Template
2385 Instantiation}, for more information.
2389 Disable generation of information about every class with virtual
2390 functions for use by the C++ run-time type identification features
2391 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2392 of the language, you can save some space by using this flag. Note that
2393 exception handling uses the same information, but G++ generates it as
2394 needed. The @code{dynamic_cast} operator can still be used for casts that
2395 do not require run-time type information, i.e.@: casts to @code{void *} or to
2396 unambiguous base classes.
2398 @item -fsized-deallocation
2399 @opindex fsized-deallocation
2400 Enable the built-in global declarations
2402 void operator delete (void *, std::size_t) noexcept;
2403 void operator delete[] (void *, std::size_t) noexcept;
2405 as introduced in C++14. This is useful for user-defined replacement
2406 deallocation functions that, for example, use the size of the object
2407 to make deallocation faster. Enabled by default under
2408 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2409 warns about places that might want to add a definition.
2413 Emit statistics about front-end processing at the end of the compilation.
2414 This information is generally only useful to the G++ development team.
2416 @item -fstrict-enums
2417 @opindex fstrict-enums
2418 Allow the compiler to optimize using the assumption that a value of
2419 enumerated type can only be one of the values of the enumeration (as
2420 defined in the C++ standard; basically, a value that can be
2421 represented in the minimum number of bits needed to represent all the
2422 enumerators). This assumption may not be valid if the program uses a
2423 cast to convert an arbitrary integer value to the enumerated type.
2425 @item -ftemplate-backtrace-limit=@var{n}
2426 @opindex ftemplate-backtrace-limit
2427 Set the maximum number of template instantiation notes for a single
2428 warning or error to @var{n}. The default value is 10.
2430 @item -ftemplate-depth=@var{n}
2431 @opindex ftemplate-depth
2432 Set the maximum instantiation depth for template classes to @var{n}.
2433 A limit on the template instantiation depth is needed to detect
2434 endless recursions during template class instantiation. ANSI/ISO C++
2435 conforming programs must not rely on a maximum depth greater than 17
2436 (changed to 1024 in C++11). The default value is 900, as the compiler
2437 can run out of stack space before hitting 1024 in some situations.
2439 @item -fno-threadsafe-statics
2440 @opindex fno-threadsafe-statics
2441 Do not emit the extra code to use the routines specified in the C++
2442 ABI for thread-safe initialization of local statics. You can use this
2443 option to reduce code size slightly in code that doesn't need to be
2446 @item -fuse-cxa-atexit
2447 @opindex fuse-cxa-atexit
2448 Register destructors for objects with static storage duration with the
2449 @code{__cxa_atexit} function rather than the @code{atexit} function.
2450 This option is required for fully standards-compliant handling of static
2451 destructors, but only works if your C library supports
2452 @code{__cxa_atexit}.
2454 @item -fno-use-cxa-get-exception-ptr
2455 @opindex fno-use-cxa-get-exception-ptr
2456 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2457 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2458 if the runtime routine is not available.
2460 @item -fvisibility-inlines-hidden
2461 @opindex fvisibility-inlines-hidden
2462 This switch declares that the user does not attempt to compare
2463 pointers to inline functions or methods where the addresses of the two functions
2464 are taken in different shared objects.
2466 The effect of this is that GCC may, effectively, mark inline methods with
2467 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2468 appear in the export table of a DSO and do not require a PLT indirection
2469 when used within the DSO@. Enabling this option can have a dramatic effect
2470 on load and link times of a DSO as it massively reduces the size of the
2471 dynamic export table when the library makes heavy use of templates.
2473 The behavior of this switch is not quite the same as marking the
2474 methods as hidden directly, because it does not affect static variables
2475 local to the function or cause the compiler to deduce that
2476 the function is defined in only one shared object.
2478 You may mark a method as having a visibility explicitly to negate the
2479 effect of the switch for that method. For example, if you do want to
2480 compare pointers to a particular inline method, you might mark it as
2481 having default visibility. Marking the enclosing class with explicit
2482 visibility has no effect.
2484 Explicitly instantiated inline methods are unaffected by this option
2485 as their linkage might otherwise cross a shared library boundary.
2486 @xref{Template Instantiation}.
2488 @item -fvisibility-ms-compat
2489 @opindex fvisibility-ms-compat
2490 This flag attempts to use visibility settings to make GCC's C++
2491 linkage model compatible with that of Microsoft Visual Studio.
2493 The flag makes these changes to GCC's linkage model:
2497 It sets the default visibility to @code{hidden}, like
2498 @option{-fvisibility=hidden}.
2501 Types, but not their members, are not hidden by default.
2504 The One Definition Rule is relaxed for types without explicit
2505 visibility specifications that are defined in more than one
2506 shared object: those declarations are permitted if they are
2507 permitted when this option is not used.
2510 In new code it is better to use @option{-fvisibility=hidden} and
2511 export those classes that are intended to be externally visible.
2512 Unfortunately it is possible for code to rely, perhaps accidentally,
2513 on the Visual Studio behavior.
2515 Among the consequences of these changes are that static data members
2516 of the same type with the same name but defined in different shared
2517 objects are different, so changing one does not change the other;
2518 and that pointers to function members defined in different shared
2519 objects may not compare equal. When this flag is given, it is a
2520 violation of the ODR to define types with the same name differently.
2522 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
2523 @opindex fvtable-verify
2524 Turn on (or off, if using @option{-fvtable-verify=none}) the security
2525 feature that verifies at run time, for every virtual call, that
2526 the vtable pointer through which the call is made is valid for the type of
2527 the object, and has not been corrupted or overwritten. If an invalid vtable
2528 pointer is detected at run time, an error is reported and execution of the
2529 program is immediately halted.
2531 This option causes run-time data structures to be built at program startup,
2532 which are used for verifying the vtable pointers.
2533 The options @samp{std} and @samp{preinit}
2534 control the timing of when these data structures are built. In both cases the
2535 data structures are built before execution reaches @code{main}. Using
2536 @option{-fvtable-verify=std} causes the data structures to be built after
2537 shared libraries have been loaded and initialized.
2538 @option{-fvtable-verify=preinit} causes them to be built before shared
2539 libraries have been loaded and initialized.
2541 If this option appears multiple times in the command line with different
2542 values specified, @samp{none} takes highest priority over both @samp{std} and
2543 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
2547 When used in conjunction with @option{-fvtable-verify=std} or
2548 @option{-fvtable-verify=preinit}, causes debug versions of the
2549 runtime functions for the vtable verification feature to be called.
2550 This flag also causes the compiler to log information about which
2551 vtable pointers it finds for each class.
2552 This information is written to a file named @file{vtv_set_ptr_data.log}
2553 in the directory named by the environment variable @env{VTV_LOGS_DIR}
2554 if that is defined or the current working directory otherwise.
2556 Note: This feature @emph{appends} data to the log file. If you want a fresh log
2557 file, be sure to delete any existing one.
2560 @opindex fvtv-counts
2561 This is a debugging flag. When used in conjunction with
2562 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
2563 causes the compiler to keep track of the total number of virtual calls
2564 it encounters and the number of verifications it inserts. It also
2565 counts the number of calls to certain run-time library functions
2566 that it inserts and logs this information for each compilation unit.
2567 The compiler writes this information to a file named
2568 @file{vtv_count_data.log} in the directory named by the environment
2569 variable @env{VTV_LOGS_DIR} if that is defined or the current working
2570 directory otherwise. It also counts the size of the vtable pointer sets
2571 for each class, and writes this information to @file{vtv_class_set_sizes.log}
2572 in the same directory.
2574 Note: This feature @emph{appends} data to the log files. To get fresh log
2575 files, be sure to delete any existing ones.
2579 Do not use weak symbol support, even if it is provided by the linker.
2580 By default, G++ uses weak symbols if they are available. This
2581 option exists only for testing, and should not be used by end-users;
2582 it results in inferior code and has no benefits. This option may
2583 be removed in a future release of G++.
2587 Do not search for header files in the standard directories specific to
2588 C++, but do still search the other standard directories. (This option
2589 is used when building the C++ library.)
2592 In addition, these optimization, warning, and code generation options
2593 have meanings only for C++ programs:
2596 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2599 Warn when G++ it generates code that is probably not compatible with
2600 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2601 ABI with each major release, normally @option{-Wabi} will warn only if
2602 there is a check added later in a release series for an ABI issue
2603 discovered since the initial release. @option{-Wabi} will warn about
2604 more things if an older ABI version is selected (with
2605 @option{-fabi-version=@var{n}}).
2607 @option{-Wabi} can also be used with an explicit version number to
2608 warn about compatibility with a particular @option{-fabi-version}
2609 level, e.g. @option{-Wabi=2} to warn about changes relative to
2610 @option{-fabi-version=2}.
2612 If an explicit version number is provided and
2613 @option{-fabi-compat-version} is not specified, the version number
2614 from this option is used for compatibility aliases. If no explicit
2615 version number is provided with this option, but
2616 @option{-fabi-compat-version} is specified, that version number is
2617 used for ABI warnings.
2619 Although an effort has been made to warn about
2620 all such cases, there are probably some cases that are not warned about,
2621 even though G++ is generating incompatible code. There may also be
2622 cases where warnings are emitted even though the code that is generated
2625 You should rewrite your code to avoid these warnings if you are
2626 concerned about the fact that code generated by G++ may not be binary
2627 compatible with code generated by other compilers.
2629 Known incompatibilities in @option{-fabi-version=2} (which was the
2630 default from GCC 3.4 to 4.9) include:
2635 A template with a non-type template parameter of reference type was
2636 mangled incorrectly:
2639 template <int &> struct S @{@};
2643 This was fixed in @option{-fabi-version=3}.
2646 SIMD vector types declared using @code{__attribute ((vector_size))} were
2647 mangled in a non-standard way that does not allow for overloading of
2648 functions taking vectors of different sizes.
2650 The mangling was changed in @option{-fabi-version=4}.
2653 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2654 qualifiers, and @code{decltype} of a plain declaration was folded away.
2656 These mangling issues were fixed in @option{-fabi-version=5}.
2659 Scoped enumerators passed as arguments to a variadic function are
2660 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2661 On most targets this does not actually affect the parameter passing
2662 ABI, as there is no way to pass an argument smaller than @code{int}.
2664 Also, the ABI changed the mangling of template argument packs,
2665 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2666 a class scope function used as a template argument.
2668 These issues were corrected in @option{-fabi-version=6}.
2671 Lambdas in default argument scope were mangled incorrectly, and the
2672 ABI changed the mangling of @code{nullptr_t}.
2674 These issues were corrected in @option{-fabi-version=7}.
2677 When mangling a function type with function-cv-qualifiers, the
2678 un-qualified function type was incorrectly treated as a substitution
2681 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2684 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2685 unaligned accesses. Note that this did not affect the ABI of a
2686 function with a @code{nullptr_t} parameter, as parameters have a
2689 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2692 Target-specific attributes that affect the identity of a type, such as
2693 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2694 did not affect the mangled name, leading to name collisions when
2695 function pointers were used as template arguments.
2697 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2701 It also warns about psABI-related changes. The known psABI changes at this
2707 For SysV/x86-64, unions with @code{long double} members are
2708 passed in memory as specified in psABI. For example:
2718 @code{union U} is always passed in memory.
2722 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2725 Warn when a type with an ABI tag is used in a context that does not
2726 have that ABI tag. See @ref{C++ Attributes} for more information
2729 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2730 @opindex Wctor-dtor-privacy
2731 @opindex Wno-ctor-dtor-privacy
2732 Warn when a class seems unusable because all the constructors or
2733 destructors in that class are private, and it has neither friends nor
2734 public static member functions. Also warn if there are no non-private
2735 methods, and there's at least one private member function that isn't
2736 a constructor or destructor.
2738 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2739 @opindex Wdelete-non-virtual-dtor
2740 @opindex Wno-delete-non-virtual-dtor
2741 Warn when @code{delete} is used to destroy an instance of a class that
2742 has virtual functions and non-virtual destructor. It is unsafe to delete
2743 an instance of a derived class through a pointer to a base class if the
2744 base class does not have a virtual destructor. This warning is enabled
2747 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2748 @opindex Wliteral-suffix
2749 @opindex Wno-literal-suffix
2750 Warn when a string or character literal is followed by a ud-suffix which does
2751 not begin with an underscore. As a conforming extension, GCC treats such
2752 suffixes as separate preprocessing tokens in order to maintain backwards
2753 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2757 #define __STDC_FORMAT_MACROS
2758 #include <inttypes.h>
2763 printf("My int64: %" PRId64"\n", i64);
2767 In this case, @code{PRId64} is treated as a separate preprocessing token.
2769 This warning is enabled by default.
2771 @item -Wlto-type-mismatch
2772 @opindex Wlto-type-mismatch
2773 @opindex Wno-lto-type-mistmach
2775 During the link-time optimization warn about type mismatches in between
2776 global declarations from different compilation units.
2777 Requires @option{-flto} to be enabled. Enabled by default.
2779 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2781 @opindex Wno-narrowing
2782 Warn when a narrowing conversion prohibited by C++11 occurs within
2786 int i = @{ 2.2 @}; // error: narrowing from double to int
2789 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2791 With @option{-std=c++11}, @option{-Wno-narrowing} suppresses the diagnostic
2792 required by the standard. Note that this does not affect the meaning
2793 of well-formed code; narrowing conversions are still considered
2794 ill-formed in SFINAE context.
2796 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2798 @opindex Wno-noexcept
2799 Warn when a noexcept-expression evaluates to false because of a call
2800 to a function that does not have a non-throwing exception
2801 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2802 the compiler to never throw an exception.
2804 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2805 @opindex Wnon-virtual-dtor
2806 @opindex Wno-non-virtual-dtor
2807 Warn when a class has virtual functions and an accessible non-virtual
2808 destructor itself or in an accessible polymorphic base class, in which
2809 case it is possible but unsafe to delete an instance of a derived
2810 class through a pointer to the class itself or base class. This
2811 warning is automatically enabled if @option{-Weffc++} is specified.
2813 @item -Wreorder @r{(C++ and Objective-C++ only)}
2815 @opindex Wno-reorder
2816 @cindex reordering, warning
2817 @cindex warning for reordering of member initializers
2818 Warn when the order of member initializers given in the code does not
2819 match the order in which they must be executed. For instance:
2825 A(): j (0), i (1) @{ @}
2830 The compiler rearranges the member initializers for @code{i}
2831 and @code{j} to match the declaration order of the members, emitting
2832 a warning to that effect. This warning is enabled by @option{-Wall}.
2834 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2835 @opindex fext-numeric-literals
2836 @opindex fno-ext-numeric-literals
2837 Accept imaginary, fixed-point, or machine-defined
2838 literal number suffixes as GNU extensions.
2839 When this option is turned off these suffixes are treated
2840 as C++11 user-defined literal numeric suffixes.
2841 This is on by default for all pre-C++11 dialects and all GNU dialects:
2842 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2843 @option{-std=gnu++14}.
2844 This option is off by default
2845 for ISO C++11 onwards (@option{-std=c++11}, ...).
2848 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2851 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2854 Warn about violations of the following style guidelines from Scott Meyers'
2855 @cite{Effective C++} series of books:
2859 Define a copy constructor and an assignment operator for classes
2860 with dynamically-allocated memory.
2863 Prefer initialization to assignment in constructors.
2866 Have @code{operator=} return a reference to @code{*this}.
2869 Don't try to return a reference when you must return an object.
2872 Distinguish between prefix and postfix forms of increment and
2873 decrement operators.
2876 Never overload @code{&&}, @code{||}, or @code{,}.
2880 This option also enables @option{-Wnon-virtual-dtor}, which is also
2881 one of the effective C++ recommendations. However, the check is
2882 extended to warn about the lack of virtual destructor in accessible
2883 non-polymorphic bases classes too.
2885 When selecting this option, be aware that the standard library
2886 headers do not obey all of these guidelines; use @samp{grep -v}
2887 to filter out those warnings.
2889 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2890 @opindex Wstrict-null-sentinel
2891 @opindex Wno-strict-null-sentinel
2892 Warn about the use of an uncasted @code{NULL} as sentinel. When
2893 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2894 to @code{__null}. Although it is a null pointer constant rather than a
2895 null pointer, it is guaranteed to be of the same size as a pointer.
2896 But this use is not portable across different compilers.
2898 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2899 @opindex Wno-non-template-friend
2900 @opindex Wnon-template-friend
2901 Disable warnings when non-templatized friend functions are declared
2902 within a template. Since the advent of explicit template specification
2903 support in G++, if the name of the friend is an unqualified-id (i.e.,
2904 @samp{friend foo(int)}), the C++ language specification demands that the
2905 friend declare or define an ordinary, nontemplate function. (Section
2906 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2907 could be interpreted as a particular specialization of a templatized
2908 function. Because this non-conforming behavior is no longer the default
2909 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2910 check existing code for potential trouble spots and is on by default.
2911 This new compiler behavior can be turned off with
2912 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2913 but disables the helpful warning.
2915 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2916 @opindex Wold-style-cast
2917 @opindex Wno-old-style-cast
2918 Warn if an old-style (C-style) cast to a non-void type is used within
2919 a C++ program. The new-style casts (@code{dynamic_cast},
2920 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2921 less vulnerable to unintended effects and much easier to search for.
2923 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2924 @opindex Woverloaded-virtual
2925 @opindex Wno-overloaded-virtual
2926 @cindex overloaded virtual function, warning
2927 @cindex warning for overloaded virtual function
2928 Warn when a function declaration hides virtual functions from a
2929 base class. For example, in:
2936 struct B: public A @{
2941 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2952 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2953 @opindex Wno-pmf-conversions
2954 @opindex Wpmf-conversions
2955 Disable the diagnostic for converting a bound pointer to member function
2958 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2959 @opindex Wsign-promo
2960 @opindex Wno-sign-promo
2961 Warn when overload resolution chooses a promotion from unsigned or
2962 enumerated type to a signed type, over a conversion to an unsigned type of
2963 the same size. Previous versions of G++ tried to preserve
2964 unsignedness, but the standard mandates the current behavior.
2966 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2968 Warn when a primary template declaration is encountered. Some coding
2969 rules disallow templates, and this may be used to enforce that rule.
2970 The warning is inactive inside a system header file, such as the STL, so
2971 one can still use the STL. One may also instantiate or specialize
2974 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2975 @opindex Wmultiple-inheritance
2976 Warn when a class is defined with multiple direct base classes. Some
2977 coding rules disallow multiple inheritance, and this may be used to
2978 enforce that rule. The warning is inactive inside a system header file,
2979 such as the STL, so one can still use the STL. One may also define
2980 classes that indirectly use multiple inheritance.
2982 @item -Wvirtual-inheritance
2983 @opindex Wvirtual-inheritance
2984 Warn when a class is defined with a virtual direct base classe. Some
2985 coding rules disallow multiple inheritance, and this may be used to
2986 enforce that rule. The warning is inactive inside a system header file,
2987 such as the STL, so one can still use the STL. One may also define
2988 classes that indirectly use virtual inheritance.
2991 @opindex Wnamespaces
2992 Warn when a namespace definition is opened. Some coding rules disallow
2993 namespaces, and this may be used to enforce that rule. The warning is
2994 inactive inside a system header file, such as the STL, so one can still
2995 use the STL. One may also use using directives and qualified names.
2997 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
2999 @opindex Wno-terminate
3000 Disable the warning about a throw-expression that will immediately
3001 result in a call to @code{terminate}.
3004 @node Objective-C and Objective-C++ Dialect Options
3005 @section Options Controlling Objective-C and Objective-C++ Dialects
3007 @cindex compiler options, Objective-C and Objective-C++
3008 @cindex Objective-C and Objective-C++ options, command-line
3009 @cindex options, Objective-C and Objective-C++
3010 (NOTE: This manual does not describe the Objective-C and Objective-C++
3011 languages themselves. @xref{Standards,,Language Standards
3012 Supported by GCC}, for references.)
3014 This section describes the command-line options that are only meaningful
3015 for Objective-C and Objective-C++ programs. You can also use most of
3016 the language-independent GNU compiler options.
3017 For example, you might compile a file @file{some_class.m} like this:
3020 gcc -g -fgnu-runtime -O -c some_class.m
3024 In this example, @option{-fgnu-runtime} is an option meant only for
3025 Objective-C and Objective-C++ programs; you can use the other options with
3026 any language supported by GCC@.
3028 Note that since Objective-C is an extension of the C language, Objective-C
3029 compilations may also use options specific to the C front-end (e.g.,
3030 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3031 C++-specific options (e.g., @option{-Wabi}).
3033 Here is a list of options that are @emph{only} for compiling Objective-C
3034 and Objective-C++ programs:
3037 @item -fconstant-string-class=@var{class-name}
3038 @opindex fconstant-string-class
3039 Use @var{class-name} as the name of the class to instantiate for each
3040 literal string specified with the syntax @code{@@"@dots{}"}. The default
3041 class name is @code{NXConstantString} if the GNU runtime is being used, and
3042 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3043 @option{-fconstant-cfstrings} option, if also present, overrides the
3044 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3045 to be laid out as constant CoreFoundation strings.
3048 @opindex fgnu-runtime
3049 Generate object code compatible with the standard GNU Objective-C
3050 runtime. This is the default for most types of systems.
3052 @item -fnext-runtime
3053 @opindex fnext-runtime
3054 Generate output compatible with the NeXT runtime. This is the default
3055 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3056 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3059 @item -fno-nil-receivers
3060 @opindex fno-nil-receivers
3061 Assume that all Objective-C message dispatches (@code{[receiver
3062 message:arg]}) in this translation unit ensure that the receiver is
3063 not @code{nil}. This allows for more efficient entry points in the
3064 runtime to be used. This option is only available in conjunction with
3065 the NeXT runtime and ABI version 0 or 1.
3067 @item -fobjc-abi-version=@var{n}
3068 @opindex fobjc-abi-version
3069 Use version @var{n} of the Objective-C ABI for the selected runtime.
3070 This option is currently supported only for the NeXT runtime. In that
3071 case, Version 0 is the traditional (32-bit) ABI without support for
3072 properties and other Objective-C 2.0 additions. Version 1 is the
3073 traditional (32-bit) ABI with support for properties and other
3074 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3075 nothing is specified, the default is Version 0 on 32-bit target
3076 machines, and Version 2 on 64-bit target machines.
3078 @item -fobjc-call-cxx-cdtors
3079 @opindex fobjc-call-cxx-cdtors
3080 For each Objective-C class, check if any of its instance variables is a
3081 C++ object with a non-trivial default constructor. If so, synthesize a
3082 special @code{- (id) .cxx_construct} instance method which runs
3083 non-trivial default constructors on any such instance variables, in order,
3084 and then return @code{self}. Similarly, check if any instance variable
3085 is a C++ object with a non-trivial destructor, and if so, synthesize a
3086 special @code{- (void) .cxx_destruct} method which runs
3087 all such default destructors, in reverse order.
3089 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3090 methods thusly generated only operate on instance variables
3091 declared in the current Objective-C class, and not those inherited
3092 from superclasses. It is the responsibility of the Objective-C
3093 runtime to invoke all such methods in an object's inheritance
3094 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3095 by the runtime immediately after a new object instance is allocated;
3096 the @code{- (void) .cxx_destruct} methods are invoked immediately
3097 before the runtime deallocates an object instance.
3099 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3100 support for invoking the @code{- (id) .cxx_construct} and
3101 @code{- (void) .cxx_destruct} methods.
3103 @item -fobjc-direct-dispatch
3104 @opindex fobjc-direct-dispatch
3105 Allow fast jumps to the message dispatcher. On Darwin this is
3106 accomplished via the comm page.
3108 @item -fobjc-exceptions
3109 @opindex fobjc-exceptions
3110 Enable syntactic support for structured exception handling in
3111 Objective-C, similar to what is offered by C++ and Java. This option
3112 is required to use the Objective-C keywords @code{@@try},
3113 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3114 @code{@@synchronized}. This option is available with both the GNU
3115 runtime and the NeXT runtime (but not available in conjunction with
3116 the NeXT runtime on Mac OS X 10.2 and earlier).
3120 Enable garbage collection (GC) in Objective-C and Objective-C++
3121 programs. This option is only available with the NeXT runtime; the
3122 GNU runtime has a different garbage collection implementation that
3123 does not require special compiler flags.
3125 @item -fobjc-nilcheck
3126 @opindex fobjc-nilcheck
3127 For the NeXT runtime with version 2 of the ABI, check for a nil
3128 receiver in method invocations before doing the actual method call.
3129 This is the default and can be disabled using
3130 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3131 checked for nil in this way no matter what this flag is set to.
3132 Currently this flag does nothing when the GNU runtime, or an older
3133 version of the NeXT runtime ABI, is used.
3135 @item -fobjc-std=objc1
3137 Conform to the language syntax of Objective-C 1.0, the language
3138 recognized by GCC 4.0. This only affects the Objective-C additions to
3139 the C/C++ language; it does not affect conformance to C/C++ standards,
3140 which is controlled by the separate C/C++ dialect option flags. When
3141 this option is used with the Objective-C or Objective-C++ compiler,
3142 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3143 This is useful if you need to make sure that your Objective-C code can
3144 be compiled with older versions of GCC@.
3146 @item -freplace-objc-classes
3147 @opindex freplace-objc-classes
3148 Emit a special marker instructing @command{ld(1)} not to statically link in
3149 the resulting object file, and allow @command{dyld(1)} to load it in at
3150 run time instead. This is used in conjunction with the Fix-and-Continue
3151 debugging mode, where the object file in question may be recompiled and
3152 dynamically reloaded in the course of program execution, without the need
3153 to restart the program itself. Currently, Fix-and-Continue functionality
3154 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3159 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3160 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3161 compile time) with static class references that get initialized at load time,
3162 which improves run-time performance. Specifying the @option{-fzero-link} flag
3163 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3164 to be retained. This is useful in Zero-Link debugging mode, since it allows
3165 for individual class implementations to be modified during program execution.
3166 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3167 regardless of command-line options.
3169 @item -fno-local-ivars
3170 @opindex fno-local-ivars
3171 @opindex flocal-ivars
3172 By default instance variables in Objective-C can be accessed as if
3173 they were local variables from within the methods of the class they're
3174 declared in. This can lead to shadowing between instance variables
3175 and other variables declared either locally inside a class method or
3176 globally with the same name. Specifying the @option{-fno-local-ivars}
3177 flag disables this behavior thus avoiding variable shadowing issues.
3179 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3180 @opindex fivar-visibility
3181 Set the default instance variable visibility to the specified option
3182 so that instance variables declared outside the scope of any access
3183 modifier directives default to the specified visibility.
3187 Dump interface declarations for all classes seen in the source file to a
3188 file named @file{@var{sourcename}.decl}.
3190 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3191 @opindex Wassign-intercept
3192 @opindex Wno-assign-intercept
3193 Warn whenever an Objective-C assignment is being intercepted by the
3196 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3197 @opindex Wno-protocol
3199 If a class is declared to implement a protocol, a warning is issued for
3200 every method in the protocol that is not implemented by the class. The
3201 default behavior is to issue a warning for every method not explicitly
3202 implemented in the class, even if a method implementation is inherited
3203 from the superclass. If you use the @option{-Wno-protocol} option, then
3204 methods inherited from the superclass are considered to be implemented,
3205 and no warning is issued for them.
3207 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3209 @opindex Wno-selector
3210 Warn if multiple methods of different types for the same selector are
3211 found during compilation. The check is performed on the list of methods
3212 in the final stage of compilation. Additionally, a check is performed
3213 for each selector appearing in a @code{@@selector(@dots{})}
3214 expression, and a corresponding method for that selector has been found
3215 during compilation. Because these checks scan the method table only at
3216 the end of compilation, these warnings are not produced if the final
3217 stage of compilation is not reached, for example because an error is
3218 found during compilation, or because the @option{-fsyntax-only} option is
3221 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3222 @opindex Wstrict-selector-match
3223 @opindex Wno-strict-selector-match
3224 Warn if multiple methods with differing argument and/or return types are
3225 found for a given selector when attempting to send a message using this
3226 selector to a receiver of type @code{id} or @code{Class}. When this flag
3227 is off (which is the default behavior), the compiler omits such warnings
3228 if any differences found are confined to types that share the same size
3231 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3232 @opindex Wundeclared-selector
3233 @opindex Wno-undeclared-selector
3234 Warn if a @code{@@selector(@dots{})} expression referring to an
3235 undeclared selector is found. A selector is considered undeclared if no
3236 method with that name has been declared before the
3237 @code{@@selector(@dots{})} expression, either explicitly in an
3238 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3239 an @code{@@implementation} section. This option always performs its
3240 checks as soon as a @code{@@selector(@dots{})} expression is found,
3241 while @option{-Wselector} only performs its checks in the final stage of
3242 compilation. This also enforces the coding style convention
3243 that methods and selectors must be declared before being used.
3245 @item -print-objc-runtime-info
3246 @opindex print-objc-runtime-info
3247 Generate C header describing the largest structure that is passed by
3252 @node Diagnostic Message Formatting Options
3253 @section Options to Control Diagnostic Messages Formatting
3254 @cindex options to control diagnostics formatting
3255 @cindex diagnostic messages
3256 @cindex message formatting
3258 Traditionally, diagnostic messages have been formatted irrespective of
3259 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3260 options described below
3261 to control the formatting algorithm for diagnostic messages,
3262 e.g.@: how many characters per line, how often source location
3263 information should be reported. Note that some language front ends may not
3264 honor these options.
3267 @item -fmessage-length=@var{n}
3268 @opindex fmessage-length
3269 Try to format error messages so that they fit on lines of about
3270 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3271 done; each error message appears on a single line. This is the
3272 default for all front ends.
3274 @item -fdiagnostics-show-location=once
3275 @opindex fdiagnostics-show-location
3276 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3277 reporter to emit source location information @emph{once}; that is, in
3278 case the message is too long to fit on a single physical line and has to
3279 be wrapped, the source location won't be emitted (as prefix) again,
3280 over and over, in subsequent continuation lines. This is the default
3283 @item -fdiagnostics-show-location=every-line
3284 Only meaningful in line-wrapping mode. Instructs the diagnostic
3285 messages reporter to emit the same source location information (as
3286 prefix) for physical lines that result from the process of breaking
3287 a message which is too long to fit on a single line.
3289 @item -fdiagnostics-color[=@var{WHEN}]
3290 @itemx -fno-diagnostics-color
3291 @opindex fdiagnostics-color
3292 @cindex highlight, color, colour
3293 @vindex GCC_COLORS @r{environment variable}
3294 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3295 or @samp{auto}. The default depends on how the compiler has been configured,
3296 it can be any of the above @var{WHEN} options or also @samp{never}
3297 if @env{GCC_COLORS} environment variable isn't present in the environment,
3298 and @samp{auto} otherwise.
3299 @samp{auto} means to use color only when the standard error is a terminal.
3300 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3301 aliases for @option{-fdiagnostics-color=always} and
3302 @option{-fdiagnostics-color=never}, respectively.
3304 The colors are defined by the environment variable @env{GCC_COLORS}.
3305 Its value is a colon-separated list of capabilities and Select Graphic
3306 Rendition (SGR) substrings. SGR commands are interpreted by the
3307 terminal or terminal emulator. (See the section in the documentation
3308 of your text terminal for permitted values and their meanings as
3309 character attributes.) These substring values are integers in decimal
3310 representation and can be concatenated with semicolons.
3311 Common values to concatenate include
3313 @samp{4} for underline,
3315 @samp{7} for inverse,
3316 @samp{39} for default foreground color,
3317 @samp{30} to @samp{37} for foreground colors,
3318 @samp{90} to @samp{97} for 16-color mode foreground colors,
3319 @samp{38;5;0} to @samp{38;5;255}
3320 for 88-color and 256-color modes foreground colors,
3321 @samp{49} for default background color,
3322 @samp{40} to @samp{47} for background colors,
3323 @samp{100} to @samp{107} for 16-color mode background colors,
3324 and @samp{48;5;0} to @samp{48;5;255}
3325 for 88-color and 256-color modes background colors.
3327 The default @env{GCC_COLORS} is
3329 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3332 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3333 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3334 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3335 string disables colors.
3336 Supported capabilities are as follows.
3340 @vindex error GCC_COLORS @r{capability}
3341 SGR substring for error: markers.
3344 @vindex warning GCC_COLORS @r{capability}
3345 SGR substring for warning: markers.
3348 @vindex note GCC_COLORS @r{capability}
3349 SGR substring for note: markers.
3352 @vindex caret GCC_COLORS @r{capability}
3353 SGR substring for caret line.
3356 @vindex locus GCC_COLORS @r{capability}
3357 SGR substring for location information, @samp{file:line} or
3358 @samp{file:line:column} etc.
3361 @vindex quote GCC_COLORS @r{capability}
3362 SGR substring for information printed within quotes.
3365 @item -fno-diagnostics-show-option
3366 @opindex fno-diagnostics-show-option
3367 @opindex fdiagnostics-show-option
3368 By default, each diagnostic emitted includes text indicating the
3369 command-line option that directly controls the diagnostic (if such an
3370 option is known to the diagnostic machinery). Specifying the
3371 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3373 @item -fno-diagnostics-show-caret
3374 @opindex fno-diagnostics-show-caret
3375 @opindex fdiagnostics-show-caret
3376 By default, each diagnostic emitted includes the original source line
3377 and a caret '^' indicating the column. This option suppresses this
3378 information. The source line is truncated to @var{n} characters, if
3379 the @option{-fmessage-length=n} option is given. When the output is done
3380 to the terminal, the width is limited to the width given by the
3381 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3385 @node Warning Options
3386 @section Options to Request or Suppress Warnings
3387 @cindex options to control warnings
3388 @cindex warning messages
3389 @cindex messages, warning
3390 @cindex suppressing warnings
3392 Warnings are diagnostic messages that report constructions that
3393 are not inherently erroneous but that are risky or suggest there
3394 may have been an error.
3396 The following language-independent options do not enable specific
3397 warnings but control the kinds of diagnostics produced by GCC@.
3400 @cindex syntax checking
3402 @opindex fsyntax-only
3403 Check the code for syntax errors, but don't do anything beyond that.
3405 @item -fmax-errors=@var{n}
3406 @opindex fmax-errors
3407 Limits the maximum number of error messages to @var{n}, at which point
3408 GCC bails out rather than attempting to continue processing the source
3409 code. If @var{n} is 0 (the default), there is no limit on the number
3410 of error messages produced. If @option{-Wfatal-errors} is also
3411 specified, then @option{-Wfatal-errors} takes precedence over this
3416 Inhibit all warning messages.
3421 Make all warnings into errors.
3426 Make the specified warning into an error. The specifier for a warning
3427 is appended; for example @option{-Werror=switch} turns the warnings
3428 controlled by @option{-Wswitch} into errors. This switch takes a
3429 negative form, to be used to negate @option{-Werror} for specific
3430 warnings; for example @option{-Wno-error=switch} makes
3431 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3434 The warning message for each controllable warning includes the
3435 option that controls the warning. That option can then be used with
3436 @option{-Werror=} and @option{-Wno-error=} as described above.
3437 (Printing of the option in the warning message can be disabled using the
3438 @option{-fno-diagnostics-show-option} flag.)
3440 Note that specifying @option{-Werror=}@var{foo} automatically implies
3441 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3444 @item -Wfatal-errors
3445 @opindex Wfatal-errors
3446 @opindex Wno-fatal-errors
3447 This option causes the compiler to abort compilation on the first error
3448 occurred rather than trying to keep going and printing further error
3453 You can request many specific warnings with options beginning with
3454 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3455 implicit declarations. Each of these specific warning options also
3456 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3457 example, @option{-Wno-implicit}. This manual lists only one of the
3458 two forms, whichever is not the default. For further
3459 language-specific options also refer to @ref{C++ Dialect Options} and
3460 @ref{Objective-C and Objective-C++ Dialect Options}.
3462 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3463 options, such as @option{-Wunused}, which may turn on further options,
3464 such as @option{-Wunused-value}. The combined effect of positive and
3465 negative forms is that more specific options have priority over less
3466 specific ones, independently of their position in the command-line. For
3467 options of the same specificity, the last one takes effect. Options
3468 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3469 as if they appeared at the end of the command-line.
3471 When an unrecognized warning option is requested (e.g.,
3472 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3473 that the option is not recognized. However, if the @option{-Wno-} form
3474 is used, the behavior is slightly different: no diagnostic is
3475 produced for @option{-Wno-unknown-warning} unless other diagnostics
3476 are being produced. This allows the use of new @option{-Wno-} options
3477 with old compilers, but if something goes wrong, the compiler
3478 warns that an unrecognized option is present.
3485 Issue all the warnings demanded by strict ISO C and ISO C++;
3486 reject all programs that use forbidden extensions, and some other
3487 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3488 version of the ISO C standard specified by any @option{-std} option used.
3490 Valid ISO C and ISO C++ programs should compile properly with or without
3491 this option (though a rare few require @option{-ansi} or a
3492 @option{-std} option specifying the required version of ISO C)@. However,
3493 without this option, certain GNU extensions and traditional C and C++
3494 features are supported as well. With this option, they are rejected.
3496 @option{-Wpedantic} does not cause warning messages for use of the
3497 alternate keywords whose names begin and end with @samp{__}. Pedantic
3498 warnings are also disabled in the expression that follows
3499 @code{__extension__}. However, only system header files should use
3500 these escape routes; application programs should avoid them.
3501 @xref{Alternate Keywords}.
3503 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3504 C conformance. They soon find that it does not do quite what they want:
3505 it finds some non-ISO practices, but not all---only those for which
3506 ISO C @emph{requires} a diagnostic, and some others for which
3507 diagnostics have been added.
3509 A feature to report any failure to conform to ISO C might be useful in
3510 some instances, but would require considerable additional work and would
3511 be quite different from @option{-Wpedantic}. We don't have plans to
3512 support such a feature in the near future.
3514 Where the standard specified with @option{-std} represents a GNU
3515 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3516 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3517 extended dialect is based. Warnings from @option{-Wpedantic} are given
3518 where they are required by the base standard. (It does not make sense
3519 for such warnings to be given only for features not in the specified GNU
3520 C dialect, since by definition the GNU dialects of C include all
3521 features the compiler supports with the given option, and there would be
3522 nothing to warn about.)
3524 @item -pedantic-errors
3525 @opindex pedantic-errors
3526 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3527 requires a diagnostic, in some cases where there is undefined behavior
3528 at compile-time and in some other cases that do not prevent compilation
3529 of programs that are valid according to the standard. This is not
3530 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3531 by this option and not enabled by the latter and vice versa.
3536 This enables all the warnings about constructions that some users
3537 consider questionable, and that are easy to avoid (or modify to
3538 prevent the warning), even in conjunction with macros. This also
3539 enables some language-specific warnings described in @ref{C++ Dialect
3540 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3542 @option{-Wall} turns on the following warning flags:
3544 @gccoptlist{-Waddress @gol
3545 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3547 -Wc++11-compat -Wc++14-compat@gol
3548 -Wchar-subscripts @gol
3550 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3552 -Wimplicit @r{(C and Objective-C only)} @gol
3553 -Wimplicit-int @r{(C and Objective-C only)} @gol
3554 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3555 -Winit-self @r{(only for C++)} @gol
3556 -Wlogical-not-parentheses
3557 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3558 -Wmaybe-uninitialized @gol
3559 -Wmemset-transposed-args @gol
3560 -Wmisleading-indentation @r{(only for C/C++)} @gol
3561 -Wmissing-braces @r{(only for C/ObjC)} @gol
3562 -Wnarrowing @r{(only for C++)} @gol
3569 -Wsequence-point @gol
3570 -Wsign-compare @r{(only in C++)} @gol
3571 -Wsizeof-pointer-memaccess @gol
3572 -Wstrict-aliasing @gol
3573 -Wstrict-overflow=1 @gol
3575 -Wtautological-compare @gol
3577 -Wuninitialized @gol
3578 -Wunknown-pragmas @gol
3579 -Wunused-function @gol
3582 -Wunused-variable @gol
3583 -Wvolatile-register-var @gol
3586 Note that some warning flags are not implied by @option{-Wall}. Some of
3587 them warn about constructions that users generally do not consider
3588 questionable, but which occasionally you might wish to check for;
3589 others warn about constructions that are necessary or hard to avoid in
3590 some cases, and there is no simple way to modify the code to suppress
3591 the warning. Some of them are enabled by @option{-Wextra} but many of
3592 them must be enabled individually.
3598 This enables some extra warning flags that are not enabled by
3599 @option{-Wall}. (This option used to be called @option{-W}. The older
3600 name is still supported, but the newer name is more descriptive.)
3602 @gccoptlist{-Wclobbered @gol
3604 -Wignored-qualifiers @gol
3605 -Wmissing-field-initializers @gol
3606 -Wmissing-parameter-type @r{(C only)} @gol
3607 -Wold-style-declaration @r{(C only)} @gol
3608 -Woverride-init @gol
3609 -Wsign-compare @r{(C only)} @gol
3611 -Wuninitialized @gol
3612 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3613 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3614 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3617 The option @option{-Wextra} also prints warning messages for the
3623 A pointer is compared against integer zero with @code{<}, @code{<=},
3624 @code{>}, or @code{>=}.
3627 (C++ only) An enumerator and a non-enumerator both appear in a
3628 conditional expression.
3631 (C++ only) Ambiguous virtual bases.
3634 (C++ only) Subscripting an array that has been declared @code{register}.
3637 (C++ only) Taking the address of a variable that has been declared
3641 (C++ only) A base class is not initialized in a derived class's copy
3646 @item -Wchar-subscripts
3647 @opindex Wchar-subscripts
3648 @opindex Wno-char-subscripts
3649 Warn if an array subscript has type @code{char}. This is a common cause
3650 of error, as programmers often forget that this type is signed on some
3652 This warning is enabled by @option{-Wall}.
3656 @opindex Wno-comment
3657 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3658 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3659 This warning is enabled by @option{-Wall}.
3661 @item -Wno-coverage-mismatch
3662 @opindex Wno-coverage-mismatch
3663 Warn if feedback profiles do not match when using the
3664 @option{-fprofile-use} option.
3665 If a source file is changed between compiling with @option{-fprofile-gen} and
3666 with @option{-fprofile-use}, the files with the profile feedback can fail
3667 to match the source file and GCC cannot use the profile feedback
3668 information. By default, this warning is enabled and is treated as an
3669 error. @option{-Wno-coverage-mismatch} can be used to disable the
3670 warning or @option{-Wno-error=coverage-mismatch} can be used to
3671 disable the error. Disabling the error for this warning can result in
3672 poorly optimized code and is useful only in the
3673 case of very minor changes such as bug fixes to an existing code-base.
3674 Completely disabling the warning is not recommended.
3677 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3679 Suppress warning messages emitted by @code{#warning} directives.
3681 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3682 @opindex Wdouble-promotion
3683 @opindex Wno-double-promotion
3684 Give a warning when a value of type @code{float} is implicitly
3685 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3686 floating-point unit implement @code{float} in hardware, but emulate
3687 @code{double} in software. On such a machine, doing computations
3688 using @code{double} values is much more expensive because of the
3689 overhead required for software emulation.
3691 It is easy to accidentally do computations with @code{double} because
3692 floating-point literals are implicitly of type @code{double}. For
3696 float area(float radius)
3698 return 3.14159 * radius * radius;
3702 the compiler performs the entire computation with @code{double}
3703 because the floating-point literal is a @code{double}.
3706 @itemx -Wformat=@var{n}
3709 @opindex ffreestanding
3710 @opindex fno-builtin
3712 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3713 the arguments supplied have types appropriate to the format string
3714 specified, and that the conversions specified in the format string make
3715 sense. This includes standard functions, and others specified by format
3716 attributes (@pxref{Function Attributes}), in the @code{printf},
3717 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3718 not in the C standard) families (or other target-specific families).
3719 Which functions are checked without format attributes having been
3720 specified depends on the standard version selected, and such checks of
3721 functions without the attribute specified are disabled by
3722 @option{-ffreestanding} or @option{-fno-builtin}.
3724 The formats are checked against the format features supported by GNU
3725 libc version 2.2. These include all ISO C90 and C99 features, as well
3726 as features from the Single Unix Specification and some BSD and GNU
3727 extensions. Other library implementations may not support all these
3728 features; GCC does not support warning about features that go beyond a
3729 particular library's limitations. However, if @option{-Wpedantic} is used
3730 with @option{-Wformat}, warnings are given about format features not
3731 in the selected standard version (but not for @code{strfmon} formats,
3732 since those are not in any version of the C standard). @xref{C Dialect
3733 Options,,Options Controlling C Dialect}.
3740 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3741 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3742 @option{-Wformat} also checks for null format arguments for several
3743 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3744 aspects of this level of format checking can be disabled by the
3745 options: @option{-Wno-format-contains-nul},
3746 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3747 @option{-Wformat} is enabled by @option{-Wall}.
3749 @item -Wno-format-contains-nul
3750 @opindex Wno-format-contains-nul
3751 @opindex Wformat-contains-nul
3752 If @option{-Wformat} is specified, do not warn about format strings that
3755 @item -Wno-format-extra-args
3756 @opindex Wno-format-extra-args
3757 @opindex Wformat-extra-args
3758 If @option{-Wformat} is specified, do not warn about excess arguments to a
3759 @code{printf} or @code{scanf} format function. The C standard specifies
3760 that such arguments are ignored.
3762 Where the unused arguments lie between used arguments that are
3763 specified with @samp{$} operand number specifications, normally
3764 warnings are still given, since the implementation could not know what
3765 type to pass to @code{va_arg} to skip the unused arguments. However,
3766 in the case of @code{scanf} formats, this option suppresses the
3767 warning if the unused arguments are all pointers, since the Single
3768 Unix Specification says that such unused arguments are allowed.
3770 @item -Wno-format-zero-length
3771 @opindex Wno-format-zero-length
3772 @opindex Wformat-zero-length
3773 If @option{-Wformat} is specified, do not warn about zero-length formats.
3774 The C standard specifies that zero-length formats are allowed.
3779 Enable @option{-Wformat} plus additional format checks. Currently
3780 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3783 @item -Wformat-nonliteral
3784 @opindex Wformat-nonliteral
3785 @opindex Wno-format-nonliteral
3786 If @option{-Wformat} is specified, also warn if the format string is not a
3787 string literal and so cannot be checked, unless the format function
3788 takes its format arguments as a @code{va_list}.
3790 @item -Wformat-security
3791 @opindex Wformat-security
3792 @opindex Wno-format-security
3793 If @option{-Wformat} is specified, also warn about uses of format
3794 functions that represent possible security problems. At present, this
3795 warns about calls to @code{printf} and @code{scanf} functions where the
3796 format string is not a string literal and there are no format arguments,
3797 as in @code{printf (foo);}. This may be a security hole if the format
3798 string came from untrusted input and contains @samp{%n}. (This is
3799 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3800 in future warnings may be added to @option{-Wformat-security} that are not
3801 included in @option{-Wformat-nonliteral}.)
3803 @item -Wformat-signedness
3804 @opindex Wformat-signedness
3805 @opindex Wno-format-signedness
3806 If @option{-Wformat} is specified, also warn if the format string
3807 requires an unsigned argument and the argument is signed and vice versa.
3810 @opindex Wformat-y2k
3811 @opindex Wno-format-y2k
3812 If @option{-Wformat} is specified, also warn about @code{strftime}
3813 formats that may yield only a two-digit year.
3818 @opindex Wno-nonnull
3819 Warn about passing a null pointer for arguments marked as
3820 requiring a non-null value by the @code{nonnull} function attribute.
3822 Also warns when comparing an argument marked with the @code{nonnull}
3823 function attribute against null inside the function.
3825 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3826 can be disabled with the @option{-Wno-nonnull} option.
3828 @item -Wnull-dereference
3829 @opindex Wnull-dereference
3830 @opindex Wno-null-dereference
3831 Warn if the compiler detects paths that trigger erroneous or
3832 undefined behavior due to dereferencing a null pointer. This option
3833 is only active when @option{-fdelete-null-pointer-checks} is active,
3834 which is enabled by optimizations in most targets. The precision of
3835 the warnings depends on the optimization options used.
3837 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3839 @opindex Wno-init-self
3840 Warn about uninitialized variables that are initialized with themselves.
3841 Note this option can only be used with the @option{-Wuninitialized} option.
3843 For example, GCC warns about @code{i} being uninitialized in the
3844 following snippet only when @option{-Winit-self} has been specified:
3855 This warning is enabled by @option{-Wall} in C++.
3857 @item -Wimplicit-int @r{(C and Objective-C only)}
3858 @opindex Wimplicit-int
3859 @opindex Wno-implicit-int
3860 Warn when a declaration does not specify a type.
3861 This warning is enabled by @option{-Wall}.
3863 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3864 @opindex Wimplicit-function-declaration
3865 @opindex Wno-implicit-function-declaration
3866 Give a warning whenever a function is used before being declared. In
3867 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3868 enabled by default and it is made into an error by
3869 @option{-pedantic-errors}. This warning is also enabled by
3872 @item -Wimplicit @r{(C and Objective-C only)}
3874 @opindex Wno-implicit
3875 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3876 This warning is enabled by @option{-Wall}.
3878 @item -Wignored-qualifiers @r{(C and C++ only)}
3879 @opindex Wignored-qualifiers
3880 @opindex Wno-ignored-qualifiers
3881 Warn if the return type of a function has a type qualifier
3882 such as @code{const}. For ISO C such a type qualifier has no effect,
3883 since the value returned by a function is not an lvalue.
3884 For C++, the warning is only emitted for scalar types or @code{void}.
3885 ISO C prohibits qualified @code{void} return types on function
3886 definitions, so such return types always receive a warning
3887 even without this option.
3889 This warning is also enabled by @option{-Wextra}.
3894 Warn if the type of @code{main} is suspicious. @code{main} should be
3895 a function with external linkage, returning int, taking either zero
3896 arguments, two, or three arguments of appropriate types. This warning
3897 is enabled by default in C++ and is enabled by either @option{-Wall}
3898 or @option{-Wpedantic}.
3900 @item -Wmisleading-indentation @r{(C and C++ only)}
3901 @opindex Wmisleading-indentation
3902 @opindex Wno-misleading-indentation
3903 Warn when the indentation of the code does not reflect the block structure.
3904 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3905 @code{for} clauses with a guarded statement that does not use braces,
3906 followed by an unguarded statement with the same indentation.
3908 In the following example, the call to ``bar'' is misleadingly indented as
3909 if it were guarded by the ``if'' conditional.
3912 if (some_condition ())
3914 bar (); /* Gotcha: this is not guarded by the "if". */
3917 In the case of mixed tabs and spaces, the warning uses the
3918 @option{-ftabstop=} option to determine if the statements line up
3921 The warning is not issued for code involving multiline preprocessor logic
3922 such as the following example.
3927 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3933 The warning is not issued after a @code{#line} directive, since this
3934 typically indicates autogenerated code, and no assumptions can be made
3935 about the layout of the file that the directive references.
3937 This warning is enabled by @option{-Wall} in C and C++.
3939 @item -Wmissing-braces
3940 @opindex Wmissing-braces
3941 @opindex Wno-missing-braces
3942 Warn if an aggregate or union initializer is not fully bracketed. In
3943 the following example, the initializer for @code{a} is not fully
3944 bracketed, but that for @code{b} is fully bracketed. This warning is
3945 enabled by @option{-Wall} in C.
3948 int a[2][2] = @{ 0, 1, 2, 3 @};
3949 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
3952 This warning is enabled by @option{-Wall}.
3954 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
3955 @opindex Wmissing-include-dirs
3956 @opindex Wno-missing-include-dirs
3957 Warn if a user-supplied include directory does not exist.
3960 @opindex Wparentheses
3961 @opindex Wno-parentheses
3962 Warn if parentheses are omitted in certain contexts, such
3963 as when there is an assignment in a context where a truth value
3964 is expected, or when operators are nested whose precedence people
3965 often get confused about.
3967 Also warn if a comparison like @code{x<=y<=z} appears; this is
3968 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
3969 interpretation from that of ordinary mathematical notation.
3971 Also warn about constructions where there may be confusion to which
3972 @code{if} statement an @code{else} branch belongs. Here is an example of
3987 In C/C++, every @code{else} branch belongs to the innermost possible
3988 @code{if} statement, which in this example is @code{if (b)}. This is
3989 often not what the programmer expected, as illustrated in the above
3990 example by indentation the programmer chose. When there is the
3991 potential for this confusion, GCC issues a warning when this flag
3992 is specified. To eliminate the warning, add explicit braces around
3993 the innermost @code{if} statement so there is no way the @code{else}
3994 can belong to the enclosing @code{if}. The resulting code
4011 Also warn for dangerous uses of the GNU extension to
4012 @code{?:} with omitted middle operand. When the condition
4013 in the @code{?}: operator is a boolean expression, the omitted value is
4014 always 1. Often programmers expect it to be a value computed
4015 inside the conditional expression instead.
4017 This warning is enabled by @option{-Wall}.
4019 @item -Wsequence-point
4020 @opindex Wsequence-point
4021 @opindex Wno-sequence-point
4022 Warn about code that may have undefined semantics because of violations
4023 of sequence point rules in the C and C++ standards.
4025 The C and C++ standards define the order in which expressions in a C/C++
4026 program are evaluated in terms of @dfn{sequence points}, which represent
4027 a partial ordering between the execution of parts of the program: those
4028 executed before the sequence point, and those executed after it. These
4029 occur after the evaluation of a full expression (one which is not part
4030 of a larger expression), after the evaluation of the first operand of a
4031 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4032 function is called (but after the evaluation of its arguments and the
4033 expression denoting the called function), and in certain other places.
4034 Other than as expressed by the sequence point rules, the order of
4035 evaluation of subexpressions of an expression is not specified. All
4036 these rules describe only a partial order rather than a total order,
4037 since, for example, if two functions are called within one expression
4038 with no sequence point between them, the order in which the functions
4039 are called is not specified. However, the standards committee have
4040 ruled that function calls do not overlap.
4042 It is not specified when between sequence points modifications to the
4043 values of objects take effect. Programs whose behavior depends on this
4044 have undefined behavior; the C and C++ standards specify that ``Between
4045 the previous and next sequence point an object shall have its stored
4046 value modified at most once by the evaluation of an expression.
4047 Furthermore, the prior value shall be read only to determine the value
4048 to be stored.''. If a program breaks these rules, the results on any
4049 particular implementation are entirely unpredictable.
4051 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4052 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4053 diagnosed by this option, and it may give an occasional false positive
4054 result, but in general it has been found fairly effective at detecting
4055 this sort of problem in programs.
4057 The standard is worded confusingly, therefore there is some debate
4058 over the precise meaning of the sequence point rules in subtle cases.
4059 Links to discussions of the problem, including proposed formal
4060 definitions, may be found on the GCC readings page, at
4061 @uref{http://gcc.gnu.org/@/readings.html}.
4063 This warning is enabled by @option{-Wall} for C and C++.
4065 @item -Wno-return-local-addr
4066 @opindex Wno-return-local-addr
4067 @opindex Wreturn-local-addr
4068 Do not warn about returning a pointer (or in C++, a reference) to a
4069 variable that goes out of scope after the function returns.
4072 @opindex Wreturn-type
4073 @opindex Wno-return-type
4074 Warn whenever a function is defined with a return type that defaults
4075 to @code{int}. Also warn about any @code{return} statement with no
4076 return value in a function whose return type is not @code{void}
4077 (falling off the end of the function body is considered returning
4078 without a value), and about a @code{return} statement with an
4079 expression in a function whose return type is @code{void}.
4081 For C++, a function without return type always produces a diagnostic
4082 message, even when @option{-Wno-return-type} is specified. The only
4083 exceptions are @code{main} and functions defined in system headers.
4085 This warning is enabled by @option{-Wall}.
4087 @item -Wshift-count-negative
4088 @opindex Wshift-count-negative
4089 @opindex Wno-shift-count-negative
4090 Warn if shift count is negative. This warning is enabled by default.
4092 @item -Wshift-count-overflow
4093 @opindex Wshift-count-overflow
4094 @opindex Wno-shift-count-overflow
4095 Warn if shift count >= width of type. This warning is enabled by default.
4097 @item -Wshift-negative-value
4098 @opindex Wshift-negative-value
4099 @opindex Wno-shift-negative-value
4100 Warn if left shifting a negative value. This warning is enabled by
4101 @option{-Wextra} in C99 and C++11 modes (and newer).
4103 @item -Wshift-overflow
4104 @itemx -Wshift-overflow=@var{n}
4105 @opindex Wshift-overflow
4106 @opindex Wno-shift-overflow
4107 Warn about left shift overflows. This warning is enabled by
4108 default in C99 and C++11 modes (and newer).
4111 @item -Wshift-overflow=1
4112 This is the warning level of @option{-Wshift-overflow} and is enabled
4113 by default in C99 and C++11 modes (and newer). This warning level does
4114 not warn about left-shifting 1 into the sign bit. (However, in C, such
4115 an overflow is still rejected in contexts where an integer constant expression
4118 @item -Wshift-overflow=2
4119 This warning level also warns about left-shifting 1 into the sign bit,
4120 unless C++14 mode is active.
4126 Warn whenever a @code{switch} statement has an index of enumerated type
4127 and lacks a @code{case} for one or more of the named codes of that
4128 enumeration. (The presence of a @code{default} label prevents this
4129 warning.) @code{case} labels outside the enumeration range also
4130 provoke warnings when this option is used (even if there is a
4131 @code{default} label).
4132 This warning is enabled by @option{-Wall}.
4134 @item -Wswitch-default
4135 @opindex Wswitch-default
4136 @opindex Wno-switch-default
4137 Warn whenever a @code{switch} statement does not have a @code{default}
4141 @opindex Wswitch-enum
4142 @opindex Wno-switch-enum
4143 Warn whenever a @code{switch} statement has an index of enumerated type
4144 and lacks a @code{case} for one or more of the named codes of that
4145 enumeration. @code{case} labels outside the enumeration range also
4146 provoke warnings when this option is used. The only difference
4147 between @option{-Wswitch} and this option is that this option gives a
4148 warning about an omitted enumeration code even if there is a
4149 @code{default} label.
4152 @opindex Wswitch-bool
4153 @opindex Wno-switch-bool
4154 Warn whenever a @code{switch} statement has an index of boolean type
4155 and the case values are outside the range of a boolean type.
4156 It is possible to suppress this warning by casting the controlling
4157 expression to a type other than @code{bool}. For example:
4160 switch ((int) (a == 4))
4166 This warning is enabled by default for C and C++ programs.
4168 @item -Wsync-nand @r{(C and C++ only)}
4170 @opindex Wno-sync-nand
4171 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4172 built-in functions are used. These functions changed semantics in GCC 4.4.
4176 @opindex Wno-trigraphs
4177 Warn if any trigraphs are encountered that might change the meaning of
4178 the program (trigraphs within comments are not warned about).
4179 This warning is enabled by @option{-Wall}.
4181 @item -Wunused-but-set-parameter
4182 @opindex Wunused-but-set-parameter
4183 @opindex Wno-unused-but-set-parameter
4184 Warn whenever a function parameter is assigned to, but otherwise unused
4185 (aside from its declaration).
4187 To suppress this warning use the @code{unused} attribute
4188 (@pxref{Variable Attributes}).
4190 This warning is also enabled by @option{-Wunused} together with
4193 @item -Wunused-but-set-variable
4194 @opindex Wunused-but-set-variable
4195 @opindex Wno-unused-but-set-variable
4196 Warn whenever a local variable is assigned to, but otherwise unused
4197 (aside from its declaration).
4198 This warning is enabled by @option{-Wall}.
4200 To suppress this warning use the @code{unused} attribute
4201 (@pxref{Variable Attributes}).
4203 This warning is also enabled by @option{-Wunused}, which is enabled
4206 @item -Wunused-function
4207 @opindex Wunused-function
4208 @opindex Wno-unused-function
4209 Warn whenever a static function is declared but not defined or a
4210 non-inline static function is unused.
4211 This warning is enabled by @option{-Wall}.
4213 @item -Wunused-label
4214 @opindex Wunused-label
4215 @opindex Wno-unused-label
4216 Warn whenever a label is declared but not used.
4217 This warning is enabled by @option{-Wall}.
4219 To suppress this warning use the @code{unused} attribute
4220 (@pxref{Variable Attributes}).
4222 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4223 @opindex Wunused-local-typedefs
4224 Warn when a typedef locally defined in a function is not used.
4225 This warning is enabled by @option{-Wall}.
4227 @item -Wunused-parameter
4228 @opindex Wunused-parameter
4229 @opindex Wno-unused-parameter
4230 Warn whenever a function parameter is unused aside from its declaration.
4232 To suppress this warning use the @code{unused} attribute
4233 (@pxref{Variable Attributes}).
4235 @item -Wno-unused-result
4236 @opindex Wunused-result
4237 @opindex Wno-unused-result
4238 Do not warn if a caller of a function marked with attribute
4239 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4240 its return value. The default is @option{-Wunused-result}.
4242 @item -Wunused-variable
4243 @opindex Wunused-variable
4244 @opindex Wno-unused-variable
4245 Warn whenever a local or static variable is unused aside from its
4246 declaration. This option implies @option{-Wunused-const-variable} for C,
4247 but not for C++. This warning is enabled by @option{-Wall}.
4249 To suppress this warning use the @code{unused} attribute
4250 (@pxref{Variable Attributes}).
4252 @item -Wunused-const-variable
4253 @opindex Wunused-const-variable
4254 @opindex Wno-unused-const-variable
4255 Warn whenever a constant static variable is unused aside from its declaration.
4256 This warning is enabled by @option{-Wunused-variable} for C, but not for C++.
4257 In C++ this is normally not an error since const variables take the place of
4258 @code{#define}s in C++.
4260 To suppress this warning use the @code{unused} attribute
4261 (@pxref{Variable Attributes}).
4263 @item -Wunused-value
4264 @opindex Wunused-value
4265 @opindex Wno-unused-value
4266 Warn whenever a statement computes a result that is explicitly not
4267 used. To suppress this warning cast the unused expression to
4268 @code{void}. This includes an expression-statement or the left-hand
4269 side of a comma expression that contains no side effects. For example,
4270 an expression such as @code{x[i,j]} causes a warning, while
4271 @code{x[(void)i,j]} does not.
4273 This warning is enabled by @option{-Wall}.
4278 All the above @option{-Wunused} options combined.
4280 In order to get a warning about an unused function parameter, you must
4281 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4282 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4284 @item -Wuninitialized
4285 @opindex Wuninitialized
4286 @opindex Wno-uninitialized
4287 Warn if an automatic variable is used without first being initialized
4288 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4289 warn if a non-static reference or non-static @code{const} member
4290 appears in a class without constructors.
4292 If you want to warn about code that uses the uninitialized value of the
4293 variable in its own initializer, use the @option{-Winit-self} option.
4295 These warnings occur for individual uninitialized or clobbered
4296 elements of structure, union or array variables as well as for
4297 variables that are uninitialized or clobbered as a whole. They do
4298 not occur for variables or elements declared @code{volatile}. Because
4299 these warnings depend on optimization, the exact variables or elements
4300 for which there are warnings depends on the precise optimization
4301 options and version of GCC used.
4303 Note that there may be no warning about a variable that is used only
4304 to compute a value that itself is never used, because such
4305 computations may be deleted by data flow analysis before the warnings
4308 @item -Wmaybe-uninitialized
4309 @opindex Wmaybe-uninitialized
4310 @opindex Wno-maybe-uninitialized
4311 For an automatic variable, if there exists a path from the function
4312 entry to a use of the variable that is initialized, but there exist
4313 some other paths for which the variable is not initialized, the compiler
4314 emits a warning if it cannot prove the uninitialized paths are not
4315 executed at run time. These warnings are made optional because GCC is
4316 not smart enough to see all the reasons why the code might be correct
4317 in spite of appearing to have an error. Here is one example of how
4338 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4339 always initialized, but GCC doesn't know this. To suppress the
4340 warning, you need to provide a default case with assert(0) or
4343 @cindex @code{longjmp} warnings
4344 This option also warns when a non-volatile automatic variable might be
4345 changed by a call to @code{longjmp}. These warnings as well are possible
4346 only in optimizing compilation.
4348 The compiler sees only the calls to @code{setjmp}. It cannot know
4349 where @code{longjmp} will be called; in fact, a signal handler could
4350 call it at any point in the code. As a result, you may get a warning
4351 even when there is in fact no problem because @code{longjmp} cannot
4352 in fact be called at the place that would cause a problem.
4354 Some spurious warnings can be avoided if you declare all the functions
4355 you use that never return as @code{noreturn}. @xref{Function
4358 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4360 @item -Wunknown-pragmas
4361 @opindex Wunknown-pragmas
4362 @opindex Wno-unknown-pragmas
4363 @cindex warning for unknown pragmas
4364 @cindex unknown pragmas, warning
4365 @cindex pragmas, warning of unknown
4366 Warn when a @code{#pragma} directive is encountered that is not understood by
4367 GCC@. If this command-line option is used, warnings are even issued
4368 for unknown pragmas in system header files. This is not the case if
4369 the warnings are only enabled by the @option{-Wall} command-line option.
4372 @opindex Wno-pragmas
4374 Do not warn about misuses of pragmas, such as incorrect parameters,
4375 invalid syntax, or conflicts between pragmas. See also
4376 @option{-Wunknown-pragmas}.
4378 @item -Wstrict-aliasing
4379 @opindex Wstrict-aliasing
4380 @opindex Wno-strict-aliasing
4381 This option is only active when @option{-fstrict-aliasing} is active.
4382 It warns about code that might break the strict aliasing rules that the
4383 compiler is using for optimization. The warning does not catch all
4384 cases, but does attempt to catch the more common pitfalls. It is
4385 included in @option{-Wall}.
4386 It is equivalent to @option{-Wstrict-aliasing=3}
4388 @item -Wstrict-aliasing=n
4389 @opindex Wstrict-aliasing=n
4390 This option is only active when @option{-fstrict-aliasing} is active.
4391 It warns about code that might break the strict aliasing rules that the
4392 compiler is using for optimization.
4393 Higher levels correspond to higher accuracy (fewer false positives).
4394 Higher levels also correspond to more effort, similar to the way @option{-O}
4396 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4398 Level 1: Most aggressive, quick, least accurate.
4399 Possibly useful when higher levels
4400 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4401 false negatives. However, it has many false positives.
4402 Warns for all pointer conversions between possibly incompatible types,
4403 even if never dereferenced. Runs in the front end only.
4405 Level 2: Aggressive, quick, not too precise.
4406 May still have many false positives (not as many as level 1 though),
4407 and few false negatives (but possibly more than level 1).
4408 Unlike level 1, it only warns when an address is taken. Warns about
4409 incomplete types. Runs in the front end only.
4411 Level 3 (default for @option{-Wstrict-aliasing}):
4412 Should have very few false positives and few false
4413 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4414 Takes care of the common pun+dereference pattern in the front end:
4415 @code{*(int*)&some_float}.
4416 If optimization is enabled, it also runs in the back end, where it deals
4417 with multiple statement cases using flow-sensitive points-to information.
4418 Only warns when the converted pointer is dereferenced.
4419 Does not warn about incomplete types.
4421 @item -Wstrict-overflow
4422 @itemx -Wstrict-overflow=@var{n}
4423 @opindex Wstrict-overflow
4424 @opindex Wno-strict-overflow
4425 This option is only active when @option{-fstrict-overflow} is active.
4426 It warns about cases where the compiler optimizes based on the
4427 assumption that signed overflow does not occur. Note that it does not
4428 warn about all cases where the code might overflow: it only warns
4429 about cases where the compiler implements some optimization. Thus
4430 this warning depends on the optimization level.
4432 An optimization that assumes that signed overflow does not occur is
4433 perfectly safe if the values of the variables involved are such that
4434 overflow never does, in fact, occur. Therefore this warning can
4435 easily give a false positive: a warning about code that is not
4436 actually a problem. To help focus on important issues, several
4437 warning levels are defined. No warnings are issued for the use of
4438 undefined signed overflow when estimating how many iterations a loop
4439 requires, in particular when determining whether a loop will be
4443 @item -Wstrict-overflow=1
4444 Warn about cases that are both questionable and easy to avoid. For
4445 example, with @option{-fstrict-overflow}, the compiler simplifies
4446 @code{x + 1 > x} to @code{1}. This level of
4447 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4448 are not, and must be explicitly requested.
4450 @item -Wstrict-overflow=2
4451 Also warn about other cases where a comparison is simplified to a
4452 constant. For example: @code{abs (x) >= 0}. This can only be
4453 simplified when @option{-fstrict-overflow} is in effect, because
4454 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4455 zero. @option{-Wstrict-overflow} (with no level) is the same as
4456 @option{-Wstrict-overflow=2}.
4458 @item -Wstrict-overflow=3
4459 Also warn about other cases where a comparison is simplified. For
4460 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4462 @item -Wstrict-overflow=4
4463 Also warn about other simplifications not covered by the above cases.
4464 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4466 @item -Wstrict-overflow=5
4467 Also warn about cases where the compiler reduces the magnitude of a
4468 constant involved in a comparison. For example: @code{x + 2 > y} is
4469 simplified to @code{x + 1 >= y}. This is reported only at the
4470 highest warning level because this simplification applies to many
4471 comparisons, so this warning level gives a very large number of
4475 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4476 @opindex Wsuggest-attribute=
4477 @opindex Wno-suggest-attribute=
4478 Warn for cases where adding an attribute may be beneficial. The
4479 attributes currently supported are listed below.
4482 @item -Wsuggest-attribute=pure
4483 @itemx -Wsuggest-attribute=const
4484 @itemx -Wsuggest-attribute=noreturn
4485 @opindex Wsuggest-attribute=pure
4486 @opindex Wno-suggest-attribute=pure
4487 @opindex Wsuggest-attribute=const
4488 @opindex Wno-suggest-attribute=const
4489 @opindex Wsuggest-attribute=noreturn
4490 @opindex Wno-suggest-attribute=noreturn
4492 Warn about functions that might be candidates for attributes
4493 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4494 functions visible in other compilation units or (in the case of @code{pure} and
4495 @code{const}) if it cannot prove that the function returns normally. A function
4496 returns normally if it doesn't contain an infinite loop or return abnormally
4497 by throwing, calling @code{abort} or trapping. This analysis requires option
4498 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4499 higher. Higher optimization levels improve the accuracy of the analysis.
4501 @item -Wsuggest-attribute=format
4502 @itemx -Wmissing-format-attribute
4503 @opindex Wsuggest-attribute=format
4504 @opindex Wmissing-format-attribute
4505 @opindex Wno-suggest-attribute=format
4506 @opindex Wno-missing-format-attribute
4510 Warn about function pointers that might be candidates for @code{format}
4511 attributes. Note these are only possible candidates, not absolute ones.
4512 GCC guesses that function pointers with @code{format} attributes that
4513 are used in assignment, initialization, parameter passing or return
4514 statements should have a corresponding @code{format} attribute in the
4515 resulting type. I.e.@: the left-hand side of the assignment or
4516 initialization, the type of the parameter variable, or the return type
4517 of the containing function respectively should also have a @code{format}
4518 attribute to avoid the warning.
4520 GCC also warns about function definitions that might be
4521 candidates for @code{format} attributes. Again, these are only
4522 possible candidates. GCC guesses that @code{format} attributes
4523 might be appropriate for any function that calls a function like
4524 @code{vprintf} or @code{vscanf}, but this might not always be the
4525 case, and some functions for which @code{format} attributes are
4526 appropriate may not be detected.
4529 @item -Wsuggest-final-types
4530 @opindex Wno-suggest-final-types
4531 @opindex Wsuggest-final-types
4532 Warn about types with virtual methods where code quality would be improved
4533 if the type were declared with the C++11 @code{final} specifier,
4535 declared in an anonymous namespace. This allows GCC to more aggressively
4536 devirtualize the polymorphic calls. This warning is more effective with link
4537 time optimization, where the information about the class hierarchy graph is
4540 @item -Wsuggest-final-methods
4541 @opindex Wno-suggest-final-methods
4542 @opindex Wsuggest-final-methods
4543 Warn about virtual methods where code quality would be improved if the method
4544 were declared with the C++11 @code{final} specifier,
4545 or, if possible, its type were
4546 declared in an anonymous namespace or with the @code{final} specifier.
4548 more effective with link time optimization, where the information about the
4549 class hierarchy graph is more complete. It is recommended to first consider
4550 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4553 @item -Wsuggest-override
4554 Warn about overriding virtual functions that are not marked with the override
4557 @item -Warray-bounds
4558 @itemx -Warray-bounds=@var{n}
4559 @opindex Wno-array-bounds
4560 @opindex Warray-bounds
4561 This option is only active when @option{-ftree-vrp} is active
4562 (default for @option{-O2} and above). It warns about subscripts to arrays
4563 that are always out of bounds. This warning is enabled by @option{-Wall}.
4566 @item -Warray-bounds=1
4567 This is the warning level of @option{-Warray-bounds} and is enabled
4568 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4570 @item -Warray-bounds=2
4571 This warning level also warns about out of bounds access for
4572 arrays at the end of a struct and for arrays accessed through
4573 pointers. This warning level may give a larger number of
4574 false positives and is deactivated by default.
4577 @item -Wbool-compare
4578 @opindex Wno-bool-compare
4579 @opindex Wbool-compare
4580 Warn about boolean expression compared with an integer value different from
4581 @code{true}/@code{false}. For instance, the following comparison is
4586 if ((n > 1) == 2) @{ @dots{} @}
4588 This warning is enabled by @option{-Wall}.
4590 @item -Wduplicated-cond
4591 @opindex Wno-duplicated-cond
4592 @opindex Wduplicated-cond
4593 Warn about duplicated conditions in an if-else-if chain. For instance,
4594 warn for the following code:
4596 if (p->q != NULL) @{ @dots{} @}
4597 else if (p->q != NULL) @{ @dots{} @}
4600 @item -Wframe-address
4601 @opindex Wno-frame-address
4602 @opindex Wframe-address
4603 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4604 is called with an argument greater than 0. Such calls may return indeterminate
4605 values or crash the program. The warning is included in @option{-Wall}.
4607 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4608 @opindex Wno-discarded-qualifiers
4609 @opindex Wdiscarded-qualifiers
4610 Do not warn if type qualifiers on pointers are being discarded.
4611 Typically, the compiler warns if a @code{const char *} variable is
4612 passed to a function that takes a @code{char *} parameter. This option
4613 can be used to suppress such a warning.
4615 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4616 @opindex Wno-discarded-array-qualifiers
4617 @opindex Wdiscarded-array-qualifiers
4618 Do not warn if type qualifiers on arrays which are pointer targets
4619 are being discarded. Typically, the compiler warns if a
4620 @code{const int (*)[]} variable is passed to a function that
4621 takes a @code{int (*)[]} parameter. This option can be used to
4622 suppress such a warning.
4624 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4625 @opindex Wno-incompatible-pointer-types
4626 @opindex Wincompatible-pointer-types
4627 Do not warn when there is a conversion between pointers that have incompatible
4628 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4629 which warns for pointer argument passing or assignment with different
4632 @item -Wno-int-conversion @r{(C and Objective-C only)}
4633 @opindex Wno-int-conversion
4634 @opindex Wint-conversion
4635 Do not warn about incompatible integer to pointer and pointer to integer
4636 conversions. This warning is about implicit conversions; for explicit
4637 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4638 @option{-Wno-pointer-to-int-cast} may be used.
4640 @item -Wno-div-by-zero
4641 @opindex Wno-div-by-zero
4642 @opindex Wdiv-by-zero
4643 Do not warn about compile-time integer division by zero. Floating-point
4644 division by zero is not warned about, as it can be a legitimate way of
4645 obtaining infinities and NaNs.
4647 @item -Wsystem-headers
4648 @opindex Wsystem-headers
4649 @opindex Wno-system-headers
4650 @cindex warnings from system headers
4651 @cindex system headers, warnings from
4652 Print warning messages for constructs found in system header files.
4653 Warnings from system headers are normally suppressed, on the assumption
4654 that they usually do not indicate real problems and would only make the
4655 compiler output harder to read. Using this command-line option tells
4656 GCC to emit warnings from system headers as if they occurred in user
4657 code. However, note that using @option{-Wall} in conjunction with this
4658 option does @emph{not} warn about unknown pragmas in system
4659 headers---for that, @option{-Wunknown-pragmas} must also be used.
4661 @item -Wtautological-compare
4662 @opindex Wtautological-compare
4663 @opindex Wno-tautological-compare
4664 Warn if a self-comparison always evaluates to true or false. This
4665 warning detects various mistakes such as:
4669 if (i > i) @{ @dots{} @}
4671 This warning is enabled by @option{-Wall}.
4674 @opindex Wtrampolines
4675 @opindex Wno-trampolines
4676 Warn about trampolines generated for pointers to nested functions.
4677 A trampoline is a small piece of data or code that is created at run
4678 time on the stack when the address of a nested function is taken, and is
4679 used to call the nested function indirectly. For some targets, it is
4680 made up of data only and thus requires no special treatment. But, for
4681 most targets, it is made up of code and thus requires the stack to be
4682 made executable in order for the program to work properly.
4685 @opindex Wfloat-equal
4686 @opindex Wno-float-equal
4687 Warn if floating-point values are used in equality comparisons.
4689 The idea behind this is that sometimes it is convenient (for the
4690 programmer) to consider floating-point values as approximations to
4691 infinitely precise real numbers. If you are doing this, then you need
4692 to compute (by analyzing the code, or in some other way) the maximum or
4693 likely maximum error that the computation introduces, and allow for it
4694 when performing comparisons (and when producing output, but that's a
4695 different problem). In particular, instead of testing for equality, you
4696 should check to see whether the two values have ranges that overlap; and
4697 this is done with the relational operators, so equality comparisons are
4700 @item -Wtraditional @r{(C and Objective-C only)}
4701 @opindex Wtraditional
4702 @opindex Wno-traditional
4703 Warn about certain constructs that behave differently in traditional and
4704 ISO C@. Also warn about ISO C constructs that have no traditional C
4705 equivalent, and/or problematic constructs that should be avoided.
4709 Macro parameters that appear within string literals in the macro body.
4710 In traditional C macro replacement takes place within string literals,
4711 but in ISO C it does not.
4714 In traditional C, some preprocessor directives did not exist.
4715 Traditional preprocessors only considered a line to be a directive
4716 if the @samp{#} appeared in column 1 on the line. Therefore
4717 @option{-Wtraditional} warns about directives that traditional C
4718 understands but ignores because the @samp{#} does not appear as the
4719 first character on the line. It also suggests you hide directives like
4720 @code{#pragma} not understood by traditional C by indenting them. Some
4721 traditional implementations do not recognize @code{#elif}, so this option
4722 suggests avoiding it altogether.
4725 A function-like macro that appears without arguments.
4728 The unary plus operator.
4731 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4732 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4733 constants.) Note, these suffixes appear in macros defined in the system
4734 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4735 Use of these macros in user code might normally lead to spurious
4736 warnings, however GCC's integrated preprocessor has enough context to
4737 avoid warning in these cases.
4740 A function declared external in one block and then used after the end of
4744 A @code{switch} statement has an operand of type @code{long}.
4747 A non-@code{static} function declaration follows a @code{static} one.
4748 This construct is not accepted by some traditional C compilers.
4751 The ISO type of an integer constant has a different width or
4752 signedness from its traditional type. This warning is only issued if
4753 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4754 typically represent bit patterns, are not warned about.
4757 Usage of ISO string concatenation is detected.
4760 Initialization of automatic aggregates.
4763 Identifier conflicts with labels. Traditional C lacks a separate
4764 namespace for labels.
4767 Initialization of unions. If the initializer is zero, the warning is
4768 omitted. This is done under the assumption that the zero initializer in
4769 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4770 initializer warnings and relies on default initialization to zero in the
4774 Conversions by prototypes between fixed/floating-point values and vice
4775 versa. The absence of these prototypes when compiling with traditional
4776 C causes serious problems. This is a subset of the possible
4777 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4780 Use of ISO C style function definitions. This warning intentionally is
4781 @emph{not} issued for prototype declarations or variadic functions
4782 because these ISO C features appear in your code when using
4783 libiberty's traditional C compatibility macros, @code{PARAMS} and
4784 @code{VPARAMS}. This warning is also bypassed for nested functions
4785 because that feature is already a GCC extension and thus not relevant to
4786 traditional C compatibility.
4789 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4790 @opindex Wtraditional-conversion
4791 @opindex Wno-traditional-conversion
4792 Warn if a prototype causes a type conversion that is different from what
4793 would happen to the same argument in the absence of a prototype. This
4794 includes conversions of fixed point to floating and vice versa, and
4795 conversions changing the width or signedness of a fixed-point argument
4796 except when the same as the default promotion.
4798 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4799 @opindex Wdeclaration-after-statement
4800 @opindex Wno-declaration-after-statement
4801 Warn when a declaration is found after a statement in a block. This
4802 construct, known from C++, was introduced with ISO C99 and is by default
4803 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4808 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4810 @item -Wno-endif-labels
4811 @opindex Wno-endif-labels
4812 @opindex Wendif-labels
4813 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4818 Warn whenever a local variable or type declaration shadows another
4819 variable, parameter, type, class member (in C++), or instance variable
4820 (in Objective-C) or whenever a built-in function is shadowed. Note
4821 that in C++, the compiler warns if a local variable shadows an
4822 explicit typedef, but not if it shadows a struct/class/enum.
4824 @item -Wno-shadow-ivar @r{(Objective-C only)}
4825 @opindex Wno-shadow-ivar
4826 @opindex Wshadow-ivar
4827 Do not warn whenever a local variable shadows an instance variable in an
4830 @item -Wlarger-than=@var{len}
4831 @opindex Wlarger-than=@var{len}
4832 @opindex Wlarger-than-@var{len}
4833 Warn whenever an object of larger than @var{len} bytes is defined.
4835 @item -Wframe-larger-than=@var{len}
4836 @opindex Wframe-larger-than
4837 Warn if the size of a function frame is larger than @var{len} bytes.
4838 The computation done to determine the stack frame size is approximate
4839 and not conservative.
4840 The actual requirements may be somewhat greater than @var{len}
4841 even if you do not get a warning. In addition, any space allocated
4842 via @code{alloca}, variable-length arrays, or related constructs
4843 is not included by the compiler when determining
4844 whether or not to issue a warning.
4846 @item -Wno-free-nonheap-object
4847 @opindex Wno-free-nonheap-object
4848 @opindex Wfree-nonheap-object
4849 Do not warn when attempting to free an object that was not allocated
4852 @item -Wstack-usage=@var{len}
4853 @opindex Wstack-usage
4854 Warn if the stack usage of a function might be larger than @var{len} bytes.
4855 The computation done to determine the stack usage is conservative.
4856 Any space allocated via @code{alloca}, variable-length arrays, or related
4857 constructs is included by the compiler when determining whether or not to
4860 The message is in keeping with the output of @option{-fstack-usage}.
4864 If the stack usage is fully static but exceeds the specified amount, it's:
4867 warning: stack usage is 1120 bytes
4870 If the stack usage is (partly) dynamic but bounded, it's:
4873 warning: stack usage might be 1648 bytes
4876 If the stack usage is (partly) dynamic and not bounded, it's:
4879 warning: stack usage might be unbounded
4883 @item -Wunsafe-loop-optimizations
4884 @opindex Wunsafe-loop-optimizations
4885 @opindex Wno-unsafe-loop-optimizations
4886 Warn if the loop cannot be optimized because the compiler cannot
4887 assume anything on the bounds of the loop indices. With
4888 @option{-funsafe-loop-optimizations} warn if the compiler makes
4891 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4892 @opindex Wno-pedantic-ms-format
4893 @opindex Wpedantic-ms-format
4894 When used in combination with @option{-Wformat}
4895 and @option{-pedantic} without GNU extensions, this option
4896 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4897 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4898 which depend on the MS runtime.
4900 @item -Wplacement-new
4901 @opindex Wplacement-new
4902 @opindex Wno-placement-new
4903 Warn about placement new expressions with undefined behavior, such as
4904 constructing an object in a buffer that is smaller than the type of
4905 the object. For example, the placement new expression below is diagnosed
4906 because it attempts to construct an array of 64 integers in a buffer only
4912 This warning is enabled by default.
4914 @item -Wpointer-arith
4915 @opindex Wpointer-arith
4916 @opindex Wno-pointer-arith
4917 Warn about anything that depends on the ``size of'' a function type or
4918 of @code{void}. GNU C assigns these types a size of 1, for
4919 convenience in calculations with @code{void *} pointers and pointers
4920 to functions. In C++, warn also when an arithmetic operation involves
4921 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
4924 @opindex Wtype-limits
4925 @opindex Wno-type-limits
4926 Warn if a comparison is always true or always false due to the limited
4927 range of the data type, but do not warn for constant expressions. For
4928 example, warn if an unsigned variable is compared against zero with
4929 @code{<} or @code{>=}. This warning is also enabled by
4932 @item -Wbad-function-cast @r{(C and Objective-C only)}
4933 @opindex Wbad-function-cast
4934 @opindex Wno-bad-function-cast
4935 Warn when a function call is cast to a non-matching type.
4936 For example, warn if a call to a function returning an integer type
4937 is cast to a pointer type.
4939 @item -Wc90-c99-compat @r{(C and Objective-C only)}
4940 @opindex Wc90-c99-compat
4941 @opindex Wno-c90-c99-compat
4942 Warn about features not present in ISO C90, but present in ISO C99.
4943 For instance, warn about use of variable length arrays, @code{long long}
4944 type, @code{bool} type, compound literals, designated initializers, and so
4945 on. This option is independent of the standards mode. Warnings are disabled
4946 in the expression that follows @code{__extension__}.
4948 @item -Wc99-c11-compat @r{(C and Objective-C only)}
4949 @opindex Wc99-c11-compat
4950 @opindex Wno-c99-c11-compat
4951 Warn about features not present in ISO C99, but present in ISO C11.
4952 For instance, warn about use of anonymous structures and unions,
4953 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
4954 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
4955 and so on. This option is independent of the standards mode. Warnings are
4956 disabled in the expression that follows @code{__extension__}.
4958 @item -Wc++-compat @r{(C and Objective-C only)}
4959 @opindex Wc++-compat
4960 Warn about ISO C constructs that are outside of the common subset of
4961 ISO C and ISO C++, e.g.@: request for implicit conversion from
4962 @code{void *} to a pointer to non-@code{void} type.
4964 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
4965 @opindex Wc++11-compat
4966 Warn about C++ constructs whose meaning differs between ISO C++ 1998
4967 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
4968 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
4969 enabled by @option{-Wall}.
4971 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
4972 @opindex Wc++14-compat
4973 Warn about C++ constructs whose meaning differs between ISO C++ 2011
4974 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
4978 @opindex Wno-cast-qual
4979 Warn whenever a pointer is cast so as to remove a type qualifier from
4980 the target type. For example, warn if a @code{const char *} is cast
4981 to an ordinary @code{char *}.
4983 Also warn when making a cast that introduces a type qualifier in an
4984 unsafe way. For example, casting @code{char **} to @code{const char **}
4985 is unsafe, as in this example:
4988 /* p is char ** value. */
4989 const char **q = (const char **) p;
4990 /* Assignment of readonly string to const char * is OK. */
4992 /* Now char** pointer points to read-only memory. */
4997 @opindex Wcast-align
4998 @opindex Wno-cast-align
4999 Warn whenever a pointer is cast such that the required alignment of the
5000 target is increased. For example, warn if a @code{char *} is cast to
5001 an @code{int *} on machines where integers can only be accessed at
5002 two- or four-byte boundaries.
5004 @item -Wwrite-strings
5005 @opindex Wwrite-strings
5006 @opindex Wno-write-strings
5007 When compiling C, give string constants the type @code{const
5008 char[@var{length}]} so that copying the address of one into a
5009 non-@code{const} @code{char *} pointer produces a warning. These
5010 warnings help you find at compile time code that can try to write
5011 into a string constant, but only if you have been very careful about
5012 using @code{const} in declarations and prototypes. Otherwise, it is
5013 just a nuisance. This is why we did not make @option{-Wall} request
5016 When compiling C++, warn about the deprecated conversion from string
5017 literals to @code{char *}. This warning is enabled by default for C++
5022 @opindex Wno-clobbered
5023 Warn for variables that might be changed by @code{longjmp} or
5024 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5026 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5027 @opindex Wconditionally-supported
5028 @opindex Wno-conditionally-supported
5029 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5032 @opindex Wconversion
5033 @opindex Wno-conversion
5034 Warn for implicit conversions that may alter a value. This includes
5035 conversions between real and integer, like @code{abs (x)} when
5036 @code{x} is @code{double}; conversions between signed and unsigned,
5037 like @code{unsigned ui = -1}; and conversions to smaller types, like
5038 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5039 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5040 changed by the conversion like in @code{abs (2.0)}. Warnings about
5041 conversions between signed and unsigned integers can be disabled by
5042 using @option{-Wno-sign-conversion}.
5044 For C++, also warn for confusing overload resolution for user-defined
5045 conversions; and conversions that never use a type conversion
5046 operator: conversions to @code{void}, the same type, a base class or a
5047 reference to them. Warnings about conversions between signed and
5048 unsigned integers are disabled by default in C++ unless
5049 @option{-Wsign-conversion} is explicitly enabled.
5051 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5052 @opindex Wconversion-null
5053 @opindex Wno-conversion-null
5054 Do not warn for conversions between @code{NULL} and non-pointer
5055 types. @option{-Wconversion-null} is enabled by default.
5057 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5058 @opindex Wzero-as-null-pointer-constant
5059 @opindex Wno-zero-as-null-pointer-constant
5060 Warn when a literal '0' is used as null pointer constant. This can
5061 be useful to facilitate the conversion to @code{nullptr} in C++11.
5063 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5064 @opindex Wsubobject-linkage
5065 @opindex Wno-subobject-linkage
5066 Warn if a class type has a base or a field whose type uses the anonymous
5067 namespace or depends on a type with no linkage. If a type A depends on
5068 a type B with no or internal linkage, defining it in multiple
5069 translation units would be an ODR violation because the meaning of B
5070 is different in each translation unit. If A only appears in a single
5071 translation unit, the best way to silence the warning is to give it
5072 internal linkage by putting it in an anonymous namespace as well. The
5073 compiler doesn't give this warning for types defined in the main .C
5074 file, as those are unlikely to have multiple definitions.
5075 @option{-Wsubobject-linkage} is enabled by default.
5079 @opindex Wno-date-time
5080 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5081 are encountered as they might prevent bit-wise-identical reproducible
5084 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5085 @opindex Wdelete-incomplete
5086 @opindex Wno-delete-incomplete
5087 Warn when deleting a pointer to incomplete type, which may cause
5088 undefined behavior at runtime. This warning is enabled by default.
5090 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5091 @opindex Wuseless-cast
5092 @opindex Wno-useless-cast
5093 Warn when an expression is casted to its own type.
5096 @opindex Wempty-body
5097 @opindex Wno-empty-body
5098 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5099 while} statement. This warning is also enabled by @option{-Wextra}.
5101 @item -Wenum-compare
5102 @opindex Wenum-compare
5103 @opindex Wno-enum-compare
5104 Warn about a comparison between values of different enumerated types.
5105 In C++ enumeral mismatches in conditional expressions are also
5106 diagnosed and the warning is enabled by default. In C this warning is
5107 enabled by @option{-Wall}.
5109 @item -Wjump-misses-init @r{(C, Objective-C only)}
5110 @opindex Wjump-misses-init
5111 @opindex Wno-jump-misses-init
5112 Warn if a @code{goto} statement or a @code{switch} statement jumps
5113 forward across the initialization of a variable, or jumps backward to a
5114 label after the variable has been initialized. This only warns about
5115 variables that are initialized when they are declared. This warning is
5116 only supported for C and Objective-C; in C++ this sort of branch is an
5119 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5120 can be disabled with the @option{-Wno-jump-misses-init} option.
5122 @item -Wsign-compare
5123 @opindex Wsign-compare
5124 @opindex Wno-sign-compare
5125 @cindex warning for comparison of signed and unsigned values
5126 @cindex comparison of signed and unsigned values, warning
5127 @cindex signed and unsigned values, comparison warning
5128 Warn when a comparison between signed and unsigned values could produce
5129 an incorrect result when the signed value is converted to unsigned.
5130 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5131 also enabled by @option{-Wextra}.
5133 @item -Wsign-conversion
5134 @opindex Wsign-conversion
5135 @opindex Wno-sign-conversion
5136 Warn for implicit conversions that may change the sign of an integer
5137 value, like assigning a signed integer expression to an unsigned
5138 integer variable. An explicit cast silences the warning. In C, this
5139 option is enabled also by @option{-Wconversion}.
5141 @item -Wfloat-conversion
5142 @opindex Wfloat-conversion
5143 @opindex Wno-float-conversion
5144 Warn for implicit conversions that reduce the precision of a real value.
5145 This includes conversions from real to integer, and from higher precision
5146 real to lower precision real values. This option is also enabled by
5147 @option{-Wconversion}.
5149 @item -Wno-scalar-storage-order
5150 @opindex -Wno-scalar-storage-order
5151 @opindex -Wscalar-storage-order
5152 Do not warn on suspicious constructs involving reverse scalar storage order.
5154 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5155 @opindex Wsized-deallocation
5156 @opindex Wno-sized-deallocation
5157 Warn about a definition of an unsized deallocation function
5159 void operator delete (void *) noexcept;
5160 void operator delete[] (void *) noexcept;
5162 without a definition of the corresponding sized deallocation function
5164 void operator delete (void *, std::size_t) noexcept;
5165 void operator delete[] (void *, std::size_t) noexcept;
5167 or vice versa. Enabled by @option{-Wextra} along with
5168 @option{-fsized-deallocation}.
5170 @item -Wsizeof-pointer-memaccess
5171 @opindex Wsizeof-pointer-memaccess
5172 @opindex Wno-sizeof-pointer-memaccess
5173 Warn for suspicious length parameters to certain string and memory built-in
5174 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5175 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5176 but a pointer, and suggests a possible fix, or about
5177 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5180 @item -Wsizeof-array-argument
5181 @opindex Wsizeof-array-argument
5182 @opindex Wno-sizeof-array-argument
5183 Warn when the @code{sizeof} operator is applied to a parameter that is
5184 declared as an array in a function definition. This warning is enabled by
5185 default for C and C++ programs.
5187 @item -Wmemset-transposed-args
5188 @opindex Wmemset-transposed-args
5189 @opindex Wno-memset-transposed-args
5190 Warn for suspicious calls to the @code{memset} built-in function, if the
5191 second argument is not zero and the third argument is zero. This warns e.g.@
5192 about @code{memset (buf, sizeof buf, 0)} where most probably
5193 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5194 is only emitted if the third argument is literal zero. If it is some
5195 expression that is folded to zero, a cast of zero to some type, etc.,
5196 it is far less likely that the user has mistakenly exchanged the arguments
5197 and no warning is emitted. This warning is enabled by @option{-Wall}.
5201 @opindex Wno-address
5202 Warn about suspicious uses of memory addresses. These include using
5203 the address of a function in a conditional expression, such as
5204 @code{void func(void); if (func)}, and comparisons against the memory
5205 address of a string literal, such as @code{if (x == "abc")}. Such
5206 uses typically indicate a programmer error: the address of a function
5207 always evaluates to true, so their use in a conditional usually
5208 indicate that the programmer forgot the parentheses in a function
5209 call; and comparisons against string literals result in unspecified
5210 behavior and are not portable in C, so they usually indicate that the
5211 programmer intended to use @code{strcmp}. This warning is enabled by
5215 @opindex Wlogical-op
5216 @opindex Wno-logical-op
5217 Warn about suspicious uses of logical operators in expressions.
5218 This includes using logical operators in contexts where a
5219 bit-wise operator is likely to be expected. Also warns when
5220 the operands of a logical operator are the same:
5223 if (a < 0 && a < 0) @{ @dots{} @}
5226 @item -Wlogical-not-parentheses
5227 @opindex Wlogical-not-parentheses
5228 @opindex Wno-logical-not-parentheses
5229 Warn about logical not used on the left hand side operand of a comparison.
5230 This option does not warn if the RHS operand is of a boolean type. Its
5231 purpose is to detect suspicious code like the following:
5235 if (!a > 1) @{ @dots{} @}
5238 It is possible to suppress the warning by wrapping the LHS into
5241 if ((!a) > 1) @{ @dots{} @}
5244 This warning is enabled by @option{-Wall}.
5246 @item -Waggregate-return
5247 @opindex Waggregate-return
5248 @opindex Wno-aggregate-return
5249 Warn if any functions that return structures or unions are defined or
5250 called. (In languages where you can return an array, this also elicits
5253 @item -Wno-aggressive-loop-optimizations
5254 @opindex Wno-aggressive-loop-optimizations
5255 @opindex Waggressive-loop-optimizations
5256 Warn if in a loop with constant number of iterations the compiler detects
5257 undefined behavior in some statement during one or more of the iterations.
5259 @item -Wno-attributes
5260 @opindex Wno-attributes
5261 @opindex Wattributes
5262 Do not warn if an unexpected @code{__attribute__} is used, such as
5263 unrecognized attributes, function attributes applied to variables,
5264 etc. This does not stop errors for incorrect use of supported
5267 @item -Wno-builtin-macro-redefined
5268 @opindex Wno-builtin-macro-redefined
5269 @opindex Wbuiltin-macro-redefined
5270 Do not warn if certain built-in macros are redefined. This suppresses
5271 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5272 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5274 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5275 @opindex Wstrict-prototypes
5276 @opindex Wno-strict-prototypes
5277 Warn if a function is declared or defined without specifying the
5278 argument types. (An old-style function definition is permitted without
5279 a warning if preceded by a declaration that specifies the argument
5282 @item -Wold-style-declaration @r{(C and Objective-C only)}
5283 @opindex Wold-style-declaration
5284 @opindex Wno-old-style-declaration
5285 Warn for obsolescent usages, according to the C Standard, in a
5286 declaration. For example, warn if storage-class specifiers like
5287 @code{static} are not the first things in a declaration. This warning
5288 is also enabled by @option{-Wextra}.
5290 @item -Wold-style-definition @r{(C and Objective-C only)}
5291 @opindex Wold-style-definition
5292 @opindex Wno-old-style-definition
5293 Warn if an old-style function definition is used. A warning is given
5294 even if there is a previous prototype.
5296 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5297 @opindex Wmissing-parameter-type
5298 @opindex Wno-missing-parameter-type
5299 A function parameter is declared without a type specifier in K&R-style
5306 This warning is also enabled by @option{-Wextra}.
5308 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5309 @opindex Wmissing-prototypes
5310 @opindex Wno-missing-prototypes
5311 Warn if a global function is defined without a previous prototype
5312 declaration. This warning is issued even if the definition itself
5313 provides a prototype. Use this option to detect global functions
5314 that do not have a matching prototype declaration in a header file.
5315 This option is not valid for C++ because all function declarations
5316 provide prototypes and a non-matching declaration declares an
5317 overload rather than conflict with an earlier declaration.
5318 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5320 @item -Wmissing-declarations
5321 @opindex Wmissing-declarations
5322 @opindex Wno-missing-declarations
5323 Warn if a global function is defined without a previous declaration.
5324 Do so even if the definition itself provides a prototype.
5325 Use this option to detect global functions that are not declared in
5326 header files. In C, no warnings are issued for functions with previous
5327 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5328 missing prototypes. In C++, no warnings are issued for function templates,
5329 or for inline functions, or for functions in anonymous namespaces.
5331 @item -Wmissing-field-initializers
5332 @opindex Wmissing-field-initializers
5333 @opindex Wno-missing-field-initializers
5337 Warn if a structure's initializer has some fields missing. For
5338 example, the following code causes such a warning, because
5339 @code{x.h} is implicitly zero:
5342 struct s @{ int f, g, h; @};
5343 struct s x = @{ 3, 4 @};
5346 This option does not warn about designated initializers, so the following
5347 modification does not trigger a warning:
5350 struct s @{ int f, g, h; @};
5351 struct s x = @{ .f = 3, .g = 4 @};
5354 In C++ this option does not warn either about the empty @{ @}
5355 initializer, for example:
5358 struct s @{ int f, g, h; @};
5362 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5363 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5365 @item -Wno-multichar
5366 @opindex Wno-multichar
5368 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5369 Usually they indicate a typo in the user's code, as they have
5370 implementation-defined values, and should not be used in portable code.
5372 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5373 @opindex Wnormalized=
5374 @opindex Wnormalized
5375 @opindex Wno-normalized
5378 @cindex character set, input normalization
5379 In ISO C and ISO C++, two identifiers are different if they are
5380 different sequences of characters. However, sometimes when characters
5381 outside the basic ASCII character set are used, you can have two
5382 different character sequences that look the same. To avoid confusion,
5383 the ISO 10646 standard sets out some @dfn{normalization rules} which
5384 when applied ensure that two sequences that look the same are turned into
5385 the same sequence. GCC can warn you if you are using identifiers that
5386 have not been normalized; this option controls that warning.
5388 There are four levels of warning supported by GCC@. The default is
5389 @option{-Wnormalized=nfc}, which warns about any identifier that is
5390 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5391 recommended form for most uses. It is equivalent to
5392 @option{-Wnormalized}.
5394 Unfortunately, there are some characters allowed in identifiers by
5395 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5396 identifiers. That is, there's no way to use these symbols in portable
5397 ISO C or C++ and have all your identifiers in NFC@.
5398 @option{-Wnormalized=id} suppresses the warning for these characters.
5399 It is hoped that future versions of the standards involved will correct
5400 this, which is why this option is not the default.
5402 You can switch the warning off for all characters by writing
5403 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5404 only do this if you are using some other normalization scheme (like
5405 ``D''), because otherwise you can easily create bugs that are
5406 literally impossible to see.
5408 Some characters in ISO 10646 have distinct meanings but look identical
5409 in some fonts or display methodologies, especially once formatting has
5410 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5411 LETTER N'', displays just like a regular @code{n} that has been
5412 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5413 normalization scheme to convert all these into a standard form as
5414 well, and GCC warns if your code is not in NFKC if you use
5415 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5416 about every identifier that contains the letter O because it might be
5417 confused with the digit 0, and so is not the default, but may be
5418 useful as a local coding convention if the programming environment
5419 cannot be fixed to display these characters distinctly.
5421 @item -Wno-deprecated
5422 @opindex Wno-deprecated
5423 @opindex Wdeprecated
5424 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5426 @item -Wno-deprecated-declarations
5427 @opindex Wno-deprecated-declarations
5428 @opindex Wdeprecated-declarations
5429 Do not warn about uses of functions (@pxref{Function Attributes}),
5430 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5431 Attributes}) marked as deprecated by using the @code{deprecated}
5435 @opindex Wno-overflow
5437 Do not warn about compile-time overflow in constant expressions.
5442 Warn about One Definition Rule violations during link-time optimization.
5443 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5446 @opindex Wopenm-simd
5447 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5448 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5449 option can be used to relax the cost model.
5451 @item -Woverride-init @r{(C and Objective-C only)}
5452 @opindex Woverride-init
5453 @opindex Wno-override-init
5457 Warn if an initialized field without side effects is overridden when
5458 using designated initializers (@pxref{Designated Inits, , Designated
5461 This warning is included in @option{-Wextra}. To get other
5462 @option{-Wextra} warnings without this one, use @option{-Wextra
5463 -Wno-override-init}.
5465 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5466 @opindex Woverride-init-side-effects
5467 @opindex Wno-override-init-side-effects
5468 Warn if an initialized field with side effects is overridden when
5469 using designated initializers (@pxref{Designated Inits, , Designated
5470 Initializers}). This warning is enabled by default.
5475 Warn if a structure is given the packed attribute, but the packed
5476 attribute has no effect on the layout or size of the structure.
5477 Such structures may be mis-aligned for little benefit. For
5478 instance, in this code, the variable @code{f.x} in @code{struct bar}
5479 is misaligned even though @code{struct bar} does not itself
5480 have the packed attribute:
5487 @} __attribute__((packed));
5495 @item -Wpacked-bitfield-compat
5496 @opindex Wpacked-bitfield-compat
5497 @opindex Wno-packed-bitfield-compat
5498 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5499 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5500 the change can lead to differences in the structure layout. GCC
5501 informs you when the offset of such a field has changed in GCC 4.4.
5502 For example there is no longer a 4-bit padding between field @code{a}
5503 and @code{b} in this structure:
5510 @} __attribute__ ((packed));
5513 This warning is enabled by default. Use
5514 @option{-Wno-packed-bitfield-compat} to disable this warning.
5519 Warn if padding is included in a structure, either to align an element
5520 of the structure or to align the whole structure. Sometimes when this
5521 happens it is possible to rearrange the fields of the structure to
5522 reduce the padding and so make the structure smaller.
5524 @item -Wredundant-decls
5525 @opindex Wredundant-decls
5526 @opindex Wno-redundant-decls
5527 Warn if anything is declared more than once in the same scope, even in
5528 cases where multiple declaration is valid and changes nothing.
5530 @item -Wnested-externs @r{(C and Objective-C only)}
5531 @opindex Wnested-externs
5532 @opindex Wno-nested-externs
5533 Warn if an @code{extern} declaration is encountered within a function.
5535 @item -Wno-inherited-variadic-ctor
5536 @opindex Winherited-variadic-ctor
5537 @opindex Wno-inherited-variadic-ctor
5538 Suppress warnings about use of C++11 inheriting constructors when the
5539 base class inherited from has a C variadic constructor; the warning is
5540 on by default because the ellipsis is not inherited.
5545 Warn if a function that is declared as inline cannot be inlined.
5546 Even with this option, the compiler does not warn about failures to
5547 inline functions declared in system headers.
5549 The compiler uses a variety of heuristics to determine whether or not
5550 to inline a function. For example, the compiler takes into account
5551 the size of the function being inlined and the amount of inlining
5552 that has already been done in the current function. Therefore,
5553 seemingly insignificant changes in the source program can cause the
5554 warnings produced by @option{-Winline} to appear or disappear.
5556 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5557 @opindex Wno-invalid-offsetof
5558 @opindex Winvalid-offsetof
5559 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5560 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5561 to a non-standard-layout type is undefined. In existing C++ implementations,
5562 however, @code{offsetof} typically gives meaningful results.
5563 This flag is for users who are aware that they are
5564 writing nonportable code and who have deliberately chosen to ignore the
5567 The restrictions on @code{offsetof} may be relaxed in a future version
5568 of the C++ standard.
5570 @item -Wno-int-to-pointer-cast
5571 @opindex Wno-int-to-pointer-cast
5572 @opindex Wint-to-pointer-cast
5573 Suppress warnings from casts to pointer type of an integer of a
5574 different size. In C++, casting to a pointer type of smaller size is
5575 an error. @option{Wint-to-pointer-cast} is enabled by default.
5578 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5579 @opindex Wno-pointer-to-int-cast
5580 @opindex Wpointer-to-int-cast
5581 Suppress warnings from casts from a pointer to an integer type of a
5585 @opindex Winvalid-pch
5586 @opindex Wno-invalid-pch
5587 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5588 the search path but can't be used.
5592 @opindex Wno-long-long
5593 Warn if @code{long long} type is used. This is enabled by either
5594 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5595 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5597 @item -Wvariadic-macros
5598 @opindex Wvariadic-macros
5599 @opindex Wno-variadic-macros
5600 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5601 alternate syntax is used in ISO C99 mode. This is enabled by either
5602 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5603 messages, use @option{-Wno-variadic-macros}.
5607 @opindex Wno-varargs
5608 Warn upon questionable usage of the macros used to handle variable
5609 arguments like @code{va_start}. This is default. To inhibit the
5610 warning messages, use @option{-Wno-varargs}.
5612 @item -Wvector-operation-performance
5613 @opindex Wvector-operation-performance
5614 @opindex Wno-vector-operation-performance
5615 Warn if vector operation is not implemented via SIMD capabilities of the
5616 architecture. Mainly useful for the performance tuning.
5617 Vector operation can be implemented @code{piecewise}, which means that the
5618 scalar operation is performed on every vector element;
5619 @code{in parallel}, which means that the vector operation is implemented
5620 using scalars of wider type, which normally is more performance efficient;
5621 and @code{as a single scalar}, which means that vector fits into a
5624 @item -Wno-virtual-move-assign
5625 @opindex Wvirtual-move-assign
5626 @opindex Wno-virtual-move-assign
5627 Suppress warnings about inheriting from a virtual base with a
5628 non-trivial C++11 move assignment operator. This is dangerous because
5629 if the virtual base is reachable along more than one path, it is
5630 moved multiple times, which can mean both objects end up in the
5631 moved-from state. If the move assignment operator is written to avoid
5632 moving from a moved-from object, this warning can be disabled.
5637 Warn if variable length array is used in the code.
5638 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5639 the variable length array.
5641 @item -Wvolatile-register-var
5642 @opindex Wvolatile-register-var
5643 @opindex Wno-volatile-register-var
5644 Warn if a register variable is declared volatile. The volatile
5645 modifier does not inhibit all optimizations that may eliminate reads
5646 and/or writes to register variables. This warning is enabled by
5649 @item -Wdisabled-optimization
5650 @opindex Wdisabled-optimization
5651 @opindex Wno-disabled-optimization
5652 Warn if a requested optimization pass is disabled. This warning does
5653 not generally indicate that there is anything wrong with your code; it
5654 merely indicates that GCC's optimizers are unable to handle the code
5655 effectively. Often, the problem is that your code is too big or too
5656 complex; GCC refuses to optimize programs when the optimization
5657 itself is likely to take inordinate amounts of time.
5659 @item -Wpointer-sign @r{(C and Objective-C only)}
5660 @opindex Wpointer-sign
5661 @opindex Wno-pointer-sign
5662 Warn for pointer argument passing or assignment with different signedness.
5663 This option is only supported for C and Objective-C@. It is implied by
5664 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5665 @option{-Wno-pointer-sign}.
5667 @item -Wstack-protector
5668 @opindex Wstack-protector
5669 @opindex Wno-stack-protector
5670 This option is only active when @option{-fstack-protector} is active. It
5671 warns about functions that are not protected against stack smashing.
5673 @item -Woverlength-strings
5674 @opindex Woverlength-strings
5675 @opindex Wno-overlength-strings
5676 Warn about string constants that are longer than the ``minimum
5677 maximum'' length specified in the C standard. Modern compilers
5678 generally allow string constants that are much longer than the
5679 standard's minimum limit, but very portable programs should avoid
5680 using longer strings.
5682 The limit applies @emph{after} string constant concatenation, and does
5683 not count the trailing NUL@. In C90, the limit was 509 characters; in
5684 C99, it was raised to 4095. C++98 does not specify a normative
5685 minimum maximum, so we do not diagnose overlength strings in C++@.
5687 This option is implied by @option{-Wpedantic}, and can be disabled with
5688 @option{-Wno-overlength-strings}.
5690 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5691 @opindex Wunsuffixed-float-constants
5693 Issue a warning for any floating constant that does not have
5694 a suffix. When used together with @option{-Wsystem-headers} it
5695 warns about such constants in system header files. This can be useful
5696 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5697 from the decimal floating-point extension to C99.
5699 @item -Wno-designated-init @r{(C and Objective-C only)}
5700 Suppress warnings when a positional initializer is used to initialize
5701 a structure that has been marked with the @code{designated_init}
5706 @node Debugging Options
5707 @section Options for Debugging Your Program or GCC
5708 @cindex options, debugging
5709 @cindex debugging information options
5711 GCC has various special options that are used for debugging
5712 either your program or GCC:
5717 Produce debugging information in the operating system's native format
5718 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
5721 On most systems that use stabs format, @option{-g} enables use of extra
5722 debugging information that only GDB can use; this extra information
5723 makes debugging work better in GDB but probably makes other debuggers
5725 refuse to read the program. If you want to control for certain whether
5726 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5727 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5729 GCC allows you to use @option{-g} with
5730 @option{-O}. The shortcuts taken by optimized code may occasionally
5731 produce surprising results: some variables you declared may not exist
5732 at all; flow of control may briefly move where you did not expect it;
5733 some statements may not be executed because they compute constant
5734 results or their values are already at hand; some statements may
5735 execute in different places because they have been moved out of loops.
5737 Nevertheless it proves possible to debug optimized output. This makes
5738 it reasonable to use the optimizer for programs that might have bugs.
5740 The following options are useful when GCC is generated with the
5741 capability for more than one debugging format.
5744 @opindex gsplit-dwarf
5745 Separate as much dwarf debugging information as possible into a
5746 separate output file with the extension .dwo. This option allows
5747 the build system to avoid linking files with debug information. To
5748 be useful, this option requires a debugger capable of reading .dwo
5753 Produce debugging information for use by GDB@. This means to use the
5754 most expressive format available (DWARF 2, stabs, or the native format
5755 if neither of those are supported), including GDB extensions if at all
5760 Generate dwarf .debug_pubnames and .debug_pubtypes sections.
5762 @item -ggnu-pubnames
5763 @opindex ggnu-pubnames
5764 Generate .debug_pubnames and .debug_pubtypes sections in a format
5765 suitable for conversion into a GDB@ index. This option is only useful
5766 with a linker that can produce GDB@ index version 7.
5770 Produce debugging information in stabs format (if that is supported),
5771 without GDB extensions. This is the format used by DBX on most BSD
5772 systems. On MIPS, Alpha and System V Release 4 systems this option
5773 produces stabs debugging output that is not understood by DBX or SDB@.
5774 On System V Release 4 systems this option requires the GNU assembler.
5776 @item -feliminate-unused-debug-symbols
5777 @opindex feliminate-unused-debug-symbols
5778 Produce debugging information in stabs format (if that is supported),
5779 for only symbols that are actually used.
5781 @item -femit-class-debug-always
5782 @opindex femit-class-debug-always
5783 Instead of emitting debugging information for a C++ class in only one
5784 object file, emit it in all object files using the class. This option
5785 should be used only with debuggers that are unable to handle the way GCC
5786 normally emits debugging information for classes because using this
5787 option increases the size of debugging information by as much as a
5790 @item -fdebug-types-section
5791 @opindex fdebug-types-section
5792 @opindex fno-debug-types-section
5793 When using DWARF Version 4 or higher, type DIEs can be put into
5794 their own @code{.debug_types} section instead of making them part of the
5795 @code{.debug_info} section. It is more efficient to put them in a separate
5796 comdat sections since the linker can then remove duplicates.
5797 But not all DWARF consumers support @code{.debug_types} sections yet
5798 and on some objects @code{.debug_types} produces larger instead of smaller
5799 debugging information.
5803 Produce debugging information in stabs format (if that is supported),
5804 using GNU extensions understood only by the GNU debugger (GDB)@. The
5805 use of these extensions is likely to make other debuggers crash or
5806 refuse to read the program.
5810 Produce debugging information in COFF format (if that is supported).
5811 This is the format used by SDB on most System V systems prior to
5816 Produce debugging information in XCOFF format (if that is supported).
5817 This is the format used by the DBX debugger on IBM RS/6000 systems.
5821 Produce debugging information in XCOFF format (if that is supported),
5822 using GNU extensions understood only by the GNU debugger (GDB)@. The
5823 use of these extensions is likely to make other debuggers crash or
5824 refuse to read the program, and may cause assemblers other than the GNU
5825 assembler (GAS) to fail with an error.
5827 @item -gdwarf-@var{version}
5828 @opindex gdwarf-@var{version}
5829 Produce debugging information in DWARF format (if that is supported).
5830 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5831 for most targets is 4. DWARF Version 5 is only experimental.
5833 Note that with DWARF Version 2, some ports require and always
5834 use some non-conflicting DWARF 3 extensions in the unwind tables.
5836 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5837 for maximum benefit.
5839 @item -grecord-gcc-switches
5840 @opindex grecord-gcc-switches
5841 This switch causes the command-line options used to invoke the
5842 compiler that may affect code generation to be appended to the
5843 DW_AT_producer attribute in DWARF debugging information. The options
5844 are concatenated with spaces separating them from each other and from
5845 the compiler version. See also @option{-frecord-gcc-switches} for another
5846 way of storing compiler options into the object file. This is the default.
5848 @item -gno-record-gcc-switches
5849 @opindex gno-record-gcc-switches
5850 Disallow appending command-line options to the DW_AT_producer attribute
5851 in DWARF debugging information.
5853 @item -gstrict-dwarf
5854 @opindex gstrict-dwarf
5855 Disallow using extensions of later DWARF standard version than selected
5856 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
5857 DWARF extensions from later standard versions is allowed.
5859 @item -gno-strict-dwarf
5860 @opindex gno-strict-dwarf
5861 Allow using extensions of later DWARF standard version than selected with
5862 @option{-gdwarf-@var{version}}.
5864 @item -gz@r{[}=@var{type}@r{]}
5866 Produce compressed debug sections in DWARF format, if that is supported.
5867 If @var{type} is not given, the default type depends on the capabilities
5868 of the assembler and linker used. @var{type} may be one of
5869 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
5870 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
5871 compression in traditional GNU format). If the linker doesn't support
5872 writing compressed debug sections, the option is rejected. Otherwise,
5873 if the assembler does not support them, @option{-gz} is silently ignored
5874 when producing object files.
5878 Produce debugging information in Alpha/VMS debug format (if that is
5879 supported). This is the format used by DEBUG on Alpha/VMS systems.
5882 @itemx -ggdb@var{level}
5883 @itemx -gstabs@var{level}
5884 @itemx -gcoff@var{level}
5885 @itemx -gxcoff@var{level}
5886 @itemx -gvms@var{level}
5887 Request debugging information and also use @var{level} to specify how
5888 much information. The default level is 2.
5890 Level 0 produces no debug information at all. Thus, @option{-g0} negates
5893 Level 1 produces minimal information, enough for making backtraces in
5894 parts of the program that you don't plan to debug. This includes
5895 descriptions of functions and external variables, and line number
5896 tables, but no information about local variables.
5898 Level 3 includes extra information, such as all the macro definitions
5899 present in the program. Some debuggers support macro expansion when
5900 you use @option{-g3}.
5902 @option{-gdwarf-2} does not accept a concatenated debug level, because
5903 GCC used to support an option @option{-gdwarf} that meant to generate
5904 debug information in version 1 of the DWARF format (which is very
5905 different from version 2), and it would have been too confusing. That
5906 debug format is long obsolete, but the option cannot be changed now.
5907 Instead use an additional @option{-g@var{level}} option to change the
5908 debug level for DWARF.
5912 Turn off generation of debug info, if leaving out this option
5913 generates it, or turn it on at level 2 otherwise. The position of this
5914 argument in the command line does not matter; it takes effect after all
5915 other options are processed, and it does so only once, no matter how
5916 many times it is given. This is mainly intended to be used with
5917 @option{-fcompare-debug}.
5919 @item -fsanitize=address
5920 @opindex fsanitize=address
5921 Enable AddressSanitizer, a fast memory error detector.
5922 Memory access instructions are instrumented to detect
5923 out-of-bounds and use-after-free bugs.
5924 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
5925 more details. The run-time behavior can be influenced using the
5926 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
5927 the available options are shown at startup of the instrumended program. See
5928 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
5929 for a list of supported options.
5931 @item -fsanitize=kernel-address
5932 @opindex fsanitize=kernel-address
5933 Enable AddressSanitizer for Linux kernel.
5934 See @uref{https://github.com/google/kasan/wiki} for more details.
5936 @item -fsanitize=thread
5937 @opindex fsanitize=thread
5938 Enable ThreadSanitizer, a fast data race detector.
5939 Memory access instructions are instrumented to detect
5940 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
5941 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
5942 environment variable; see
5943 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
5946 @item -fsanitize=leak
5947 @opindex fsanitize=leak
5948 Enable LeakSanitizer, a memory leak detector.
5949 This option only matters for linking of executables and if neither
5950 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
5951 case the executable is linked against a library that overrides @code{malloc}
5952 and other allocator functions. See
5953 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
5954 details. The run-time behavior can be influenced using the
5955 @env{LSAN_OPTIONS} environment variable.
5957 @item -fsanitize=undefined
5958 @opindex fsanitize=undefined
5959 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
5960 Various computations are instrumented to detect undefined behavior
5961 at runtime. Current suboptions are:
5965 @item -fsanitize=shift
5966 @opindex fsanitize=shift
5967 This option enables checking that the result of a shift operation is
5968 not undefined. Note that what exactly is considered undefined differs
5969 slightly between C and C++, as well as between ISO C90 and C99, etc.
5971 @item -fsanitize=integer-divide-by-zero
5972 @opindex fsanitize=integer-divide-by-zero
5973 Detect integer division by zero as well as @code{INT_MIN / -1} division.
5975 @item -fsanitize=unreachable
5976 @opindex fsanitize=unreachable
5977 With this option, the compiler turns the @code{__builtin_unreachable}
5978 call into a diagnostics message call instead. When reaching the
5979 @code{__builtin_unreachable} call, the behavior is undefined.
5981 @item -fsanitize=vla-bound
5982 @opindex fsanitize=vla-bound
5983 This option instructs the compiler to check that the size of a variable
5984 length array is positive.
5986 @item -fsanitize=null
5987 @opindex fsanitize=null
5988 This option enables pointer checking. Particularly, the application
5989 built with this option turned on will issue an error message when it
5990 tries to dereference a NULL pointer, or if a reference (possibly an
5991 rvalue reference) is bound to a NULL pointer, or if a method is invoked
5992 on an object pointed by a NULL pointer.
5994 @item -fsanitize=return
5995 @opindex fsanitize=return
5996 This option enables return statement checking. Programs
5997 built with this option turned on will issue an error message
5998 when the end of a non-void function is reached without actually
5999 returning a value. This option works in C++ only.
6001 @item -fsanitize=signed-integer-overflow
6002 @opindex fsanitize=signed-integer-overflow
6003 This option enables signed integer overflow checking. We check that
6004 the result of @code{+}, @code{*}, and both unary and binary @code{-}
6005 does not overflow in the signed arithmetics. Note, integer promotion
6006 rules must be taken into account. That is, the following is not an
6009 signed char a = SCHAR_MAX;
6013 @item -fsanitize=bounds
6014 @opindex fsanitize=bounds
6015 This option enables instrumentation of array bounds. Various out of bounds
6016 accesses are detected. Flexible array members, flexible array member-like
6017 arrays, and initializers of variables with static storage are not instrumented.
6019 @item -fsanitize=bounds-strict
6020 @opindex fsanitize=bounds-strict
6021 This option enables strict instrumentation of array bounds. Most out of bounds
6022 accesses are detected, including flexible array members and flexible array
6023 member-like arrays. Initializers of variables with static storage are not
6026 @item -fsanitize=alignment
6027 @opindex fsanitize=alignment
6029 This option enables checking of alignment of pointers when they are
6030 dereferenced, or when a reference is bound to insufficiently aligned target,
6031 or when a method or constructor is invoked on insufficiently aligned object.
6033 @item -fsanitize=object-size
6034 @opindex fsanitize=object-size
6035 This option enables instrumentation of memory references using the
6036 @code{__builtin_object_size} function. Various out of bounds pointer
6037 accesses are detected.
6039 @item -fsanitize=float-divide-by-zero
6040 @opindex fsanitize=float-divide-by-zero
6041 Detect floating-point division by zero. Unlike other similar options,
6042 @option{-fsanitize=float-divide-by-zero} is not enabled by
6043 @option{-fsanitize=undefined}, since floating-point division by zero can
6044 be a legitimate way of obtaining infinities and NaNs.
6046 @item -fsanitize=float-cast-overflow
6047 @opindex fsanitize=float-cast-overflow
6048 This option enables floating-point type to integer conversion checking.
6049 We check that the result of the conversion does not overflow.
6050 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
6051 not enabled by @option{-fsanitize=undefined}.
6052 This option does not work well with @code{FE_INVALID} exceptions enabled.
6054 @item -fsanitize=nonnull-attribute
6055 @opindex fsanitize=nonnull-attribute
6057 This option enables instrumentation of calls, checking whether null values
6058 are not passed to arguments marked as requiring a non-null value by the
6059 @code{nonnull} function attribute.
6061 @item -fsanitize=returns-nonnull-attribute
6062 @opindex fsanitize=returns-nonnull-attribute
6064 This option enables instrumentation of return statements in functions
6065 marked with @code{returns_nonnull} function attribute, to detect returning
6066 of null values from such functions.
6068 @item -fsanitize=bool
6069 @opindex fsanitize=bool
6071 This option enables instrumentation of loads from bool. If a value other
6072 than 0/1 is loaded, a run-time error is issued.
6074 @item -fsanitize=enum
6075 @opindex fsanitize=enum
6077 This option enables instrumentation of loads from an enum type. If
6078 a value outside the range of values for the enum type is loaded,
6079 a run-time error is issued.
6081 @item -fsanitize=vptr
6082 @opindex fsanitize=vptr
6084 This option enables instrumentation of C++ member function calls, member
6085 accesses and some conversions between pointers to base and derived classes,
6086 to verify the referenced object has the correct dynamic type.
6090 While @option{-ftrapv} causes traps for signed overflows to be emitted,
6091 @option{-fsanitize=undefined} gives a diagnostic message.
6092 This currently works only for the C family of languages.
6094 @item -fno-sanitize=all
6095 @opindex fno-sanitize=all
6097 This option disables all previously enabled sanitizers.
6098 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
6101 @item -fasan-shadow-offset=@var{number}
6102 @opindex fasan-shadow-offset
6103 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
6104 It is useful for experimenting with different shadow memory layouts in
6105 Kernel AddressSanitizer.
6107 @item -fsanitize-sections=@var{s1},@var{s2},...
6108 @opindex fsanitize-sections
6109 Sanitize global variables in selected user-defined sections. @var{si} may
6112 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
6113 @opindex fsanitize-recover
6114 @opindex fno-sanitize-recover
6115 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
6116 mentioned in comma-separated list of @var{opts}. Enabling this option
6117 for a sanitizer component causes it to attempt to continue
6118 running the program as if no error happened. This means multiple
6119 runtime errors can be reported in a single program run, and the exit
6120 code of the program may indicate success even when errors
6121 have been reported. The @option{-fno-sanitize-recover=} option
6122 can be used to alter
6123 this behavior: only the first detected error is reported
6124 and program then exits with a non-zero exit code.
6126 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
6127 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
6128 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
6129 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
6130 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
6131 for which this feature is experimental.
6132 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
6133 accepted, the former enables recovery for all sanitizers that support it,
6134 the latter disables recovery for all sanitizers that support it.
6136 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
6138 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6141 Similarly @option{-fno-sanitize-recover} is equivalent to
6143 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
6146 @item -fsanitize-undefined-trap-on-error
6147 @opindex fsanitize-undefined-trap-on-error
6148 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
6149 report undefined behavior using @code{__builtin_trap} rather than
6150 a @code{libubsan} library routine. The advantage of this is that the
6151 @code{libubsan} library is not needed and is not linked in, so this
6152 is usable even in freestanding environments.
6154 @item -fsanitize-coverage=trace-pc
6155 @opindex fsanitize-coverage=trace-pc
6156 Enable coverage-guided fuzzing code instrumentation.
6157 Inserts call to __sanitizer_cov_trace_pc into every basic block.
6159 @item -fcheck-pointer-bounds
6160 @opindex fcheck-pointer-bounds
6161 @opindex fno-check-pointer-bounds
6162 @cindex Pointer Bounds Checker options
6163 Enable Pointer Bounds Checker instrumentation. Each memory reference
6164 is instrumented with checks of the pointer used for memory access against
6165 bounds associated with that pointer.
6168 is only an implementation for Intel MPX available, thus x86 target
6169 and @option{-mmpx} are required to enable this feature.
6170 MPX-based instrumentation requires
6171 a runtime library to enable MPX in hardware and handle bounds
6172 violation signals. By default when @option{-fcheck-pointer-bounds}
6173 and @option{-mmpx} options are used to link a program, the GCC driver
6174 links against the @file{libmpx} runtime library and @file{libmpxwrappers}
6175 library. It also passes '-z bndplt' to a linker in case it supports this
6176 option (which is checked on libmpx configuration). Note that old versions
6177 of linker may ignore option. Gold linker doesn't support '-z bndplt'
6178 option. With no '-z bndplt' support in linker all calls to dynamic libraries
6179 lose passed bounds reducing overall protection level. It's highly
6180 recommended to use linker with '-z bndplt' support. In case such linker
6181 is not available it is adviced to always use @option{-static-libmpxwrappers}
6182 for better protection level or use @option{-static} to completely avoid
6183 external calls to dynamic libraries. MPX-based instrumentation
6184 may be used for debugging and also may be included in production code
6185 to increase program security. Depending on usage, you may
6186 have different requirements for the runtime library. The current version
6187 of the MPX runtime library is more oriented for use as a debugging
6188 tool. MPX runtime library usage implies @option{-lpthread}. See
6189 also @option{-static-libmpx}. The runtime library behavior can be
6190 influenced using various @env{CHKP_RT_*} environment variables. See
6191 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
6194 Generated instrumentation may be controlled by various
6195 @option{-fchkp-*} options and by the @code{bnd_variable_size}
6196 structure field attribute (@pxref{Type Attributes}) and
6197 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
6198 (@pxref{Function Attributes}). GCC also provides a number of built-in
6199 functions for controlling the Pointer Bounds Checker. @xref{Pointer
6200 Bounds Checker builtins}, for more information.
6204 @opindex fno-checking
6205 Enable internal consistency checking. The default depends on
6206 the compiler configuration.
6208 @item -fchkp-check-incomplete-type
6209 @opindex fchkp-check-incomplete-type
6210 @opindex fno-chkp-check-incomplete-type
6211 Generate pointer bounds checks for variables with incomplete type.
6214 @item -fchkp-narrow-bounds
6215 @opindex fchkp-narrow-bounds
6216 @opindex fno-chkp-narrow-bounds
6217 Controls bounds used by Pointer Bounds Checker for pointers to object
6218 fields. If narrowing is enabled then field bounds are used. Otherwise
6219 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
6220 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
6222 @item -fchkp-first-field-has-own-bounds
6223 @opindex fchkp-first-field-has-own-bounds
6224 @opindex fno-chkp-first-field-has-own-bounds
6225 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
6226 first field in the structure. By default a pointer to the first field has
6227 the same bounds as a pointer to the whole structure.
6229 @item -fchkp-narrow-to-innermost-array
6230 @opindex fchkp-narrow-to-innermost-array
6231 @opindex fno-chkp-narrow-to-innermost-array
6232 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
6233 case of nested static array access. By default this option is disabled and
6234 bounds of the outermost array are used.
6236 @item -fchkp-optimize
6237 @opindex fchkp-optimize
6238 @opindex fno-chkp-optimize
6239 Enables Pointer Bounds Checker optimizations. Enabled by default at
6240 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
6242 @item -fchkp-use-fast-string-functions
6243 @opindex fchkp-use-fast-string-functions
6244 @opindex fno-chkp-use-fast-string-functions
6245 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
6246 by Pointer Bounds Checker. Disabled by default.
6248 @item -fchkp-use-nochk-string-functions
6249 @opindex fchkp-use-nochk-string-functions
6250 @opindex fno-chkp-use-nochk-string-functions
6251 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
6252 by Pointer Bounds Checker. Disabled by default.
6254 @item -fchkp-use-static-bounds
6255 @opindex fchkp-use-static-bounds
6256 @opindex fno-chkp-use-static-bounds
6257 Allow Pointer Bounds Checker to generate static bounds holding
6258 bounds of static variables. Enabled by default.
6260 @item -fchkp-use-static-const-bounds
6261 @opindex fchkp-use-static-const-bounds
6262 @opindex fno-chkp-use-static-const-bounds
6263 Use statically-initialized bounds for constant bounds instead of
6264 generating them each time they are required. By default enabled when
6265 @option{-fchkp-use-static-bounds} is enabled.
6267 @item -fchkp-treat-zero-dynamic-size-as-infinite
6268 @opindex fchkp-treat-zero-dynamic-size-as-infinite
6269 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
6270 With this option, objects with incomplete type whose
6271 dynamically-obtained size is zero are treated as having infinite size
6272 instead by Pointer Bounds
6273 Checker. This option may be helpful if a program is linked with a library
6274 missing size information for some symbols. Disabled by default.
6276 @item -fchkp-check-read
6277 @opindex fchkp-check-read
6278 @opindex fno-chkp-check-read
6279 Instructs Pointer Bounds Checker to generate checks for all read
6280 accesses to memory. Enabled by default.
6282 @item -fchkp-check-write
6283 @opindex fchkp-check-write
6284 @opindex fno-chkp-check-write
6285 Instructs Pointer Bounds Checker to generate checks for all write
6286 accesses to memory. Enabled by default.
6288 @item -fchkp-store-bounds
6289 @opindex fchkp-store-bounds
6290 @opindex fno-chkp-store-bounds
6291 Instructs Pointer Bounds Checker to generate bounds stores for
6292 pointer writes. Enabled by default.
6294 @item -fchkp-instrument-calls
6295 @opindex fchkp-instrument-calls
6296 @opindex fno-chkp-instrument-calls
6297 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
6300 @item -fchkp-instrument-marked-only
6301 @opindex fchkp-instrument-marked-only
6302 @opindex fno-chkp-instrument-marked-only
6303 Instructs Pointer Bounds Checker to instrument only functions
6304 marked with the @code{bnd_instrument} attribute
6305 (@pxref{Function Attributes}). Disabled by default.
6307 @item -fchkp-use-wrappers
6308 @opindex fchkp-use-wrappers
6309 @opindex fno-chkp-use-wrappers
6310 Allows Pointer Bounds Checker to replace calls to built-in functions
6311 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
6312 is used to link a program, the GCC driver automatically links
6313 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
6316 @item -fdump-final-insns@r{[}=@var{file}@r{]}
6317 @opindex fdump-final-insns
6318 Dump the final internal representation (RTL) to @var{file}. If the
6319 optional argument is omitted (or if @var{file} is @code{.}), the name
6320 of the dump file is determined by appending @code{.gkd} to the
6321 compilation output file name.
6323 @item -fcompare-debug@r{[}=@var{opts}@r{]}
6324 @opindex fcompare-debug
6325 @opindex fno-compare-debug
6326 If no error occurs during compilation, run the compiler a second time,
6327 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
6328 passed to the second compilation. Dump the final internal
6329 representation in both compilations, and print an error if they differ.
6331 If the equal sign is omitted, the default @option{-gtoggle} is used.
6333 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
6334 and nonzero, implicitly enables @option{-fcompare-debug}. If
6335 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
6336 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
6339 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
6340 is equivalent to @option{-fno-compare-debug}, which disables the dumping
6341 of the final representation and the second compilation, preventing even
6342 @env{GCC_COMPARE_DEBUG} from taking effect.
6344 To verify full coverage during @option{-fcompare-debug} testing, set
6345 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
6346 which GCC rejects as an invalid option in any actual compilation
6347 (rather than preprocessing, assembly or linking). To get just a
6348 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
6349 not overridden} will do.
6351 @item -fcompare-debug-second
6352 @opindex fcompare-debug-second
6353 This option is implicitly passed to the compiler for the second
6354 compilation requested by @option{-fcompare-debug}, along with options to
6355 silence warnings, and omitting other options that would cause
6356 side-effect compiler outputs to files or to the standard output. Dump
6357 files and preserved temporary files are renamed so as to contain the
6358 @code{.gk} additional extension during the second compilation, to avoid
6359 overwriting those generated by the first.
6361 When this option is passed to the compiler driver, it causes the
6362 @emph{first} compilation to be skipped, which makes it useful for little
6363 other than debugging the compiler proper.
6365 @item -feliminate-dwarf2-dups
6366 @opindex feliminate-dwarf2-dups
6367 Compress DWARF 2 debugging information by eliminating duplicated
6368 information about each symbol. This option only makes sense when
6369 generating DWARF 2 debugging information with @option{-gdwarf-2}.
6371 @item -femit-struct-debug-baseonly
6372 @opindex femit-struct-debug-baseonly
6373 Emit debug information for struct-like types
6374 only when the base name of the compilation source file
6375 matches the base name of file in which the struct is defined.
6377 This option substantially reduces the size of debugging information,
6378 but at significant potential loss in type information to the debugger.
6379 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6380 See @option{-femit-struct-debug-detailed} for more detailed control.
6382 This option works only with DWARF 2.
6384 @item -femit-struct-debug-reduced
6385 @opindex femit-struct-debug-reduced
6386 Emit debug information for struct-like types
6387 only when the base name of the compilation source file
6388 matches the base name of file in which the type is defined,
6389 unless the struct is a template or defined in a system header.
6391 This option significantly reduces the size of debugging information,
6392 with some potential loss in type information to the debugger.
6393 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6394 See @option{-femit-struct-debug-detailed} for more detailed control.
6396 This option works only with DWARF 2.
6398 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6399 @opindex femit-struct-debug-detailed
6400 Specify the struct-like types
6401 for which the compiler generates debug information.
6402 The intent is to reduce duplicate struct debug information
6403 between different object files within the same program.
6405 This option is a detailed version of
6406 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6407 which serves for most needs.
6409 A specification has the syntax@*
6410 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6412 The optional first word limits the specification to
6413 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6414 A struct type is used directly when it is the type of a variable, member.
6415 Indirect uses arise through pointers to structs.
6416 That is, when use of an incomplete struct is valid, the use is indirect.
6418 @samp{struct one direct; struct two * indirect;}.
6420 The optional second word limits the specification to
6421 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6422 Generic structs are a bit complicated to explain.
6423 For C++, these are non-explicit specializations of template classes,
6424 or non-template classes within the above.
6425 Other programming languages have generics,
6426 but @option{-femit-struct-debug-detailed} does not yet implement them.
6428 The third word specifies the source files for those
6429 structs for which the compiler should emit debug information.
6430 The values @samp{none} and @samp{any} have the normal meaning.
6431 The value @samp{base} means that
6432 the base of name of the file in which the type declaration appears
6433 must match the base of the name of the main compilation file.
6434 In practice, this means that when compiling @file{foo.c}, debug information
6435 is generated for types declared in that file and @file{foo.h},
6436 but not other header files.
6437 The value @samp{sys} means those types satisfying @samp{base}
6438 or declared in system or compiler headers.
6440 You may need to experiment to determine the best settings for your application.
6442 The default is @option{-femit-struct-debug-detailed=all}.
6444 This option works only with DWARF 2.
6446 @item -fno-merge-debug-strings
6447 @opindex fmerge-debug-strings
6448 @opindex fno-merge-debug-strings
6449 Direct the linker to not merge together strings in the debugging
6450 information that are identical in different object files. Merging is
6451 not supported by all assemblers or linkers. Merging decreases the size
6452 of the debug information in the output file at the cost of increasing
6453 link processing time. Merging is enabled by default.
6455 @item -fdebug-prefix-map=@var{old}=@var{new}
6456 @opindex fdebug-prefix-map
6457 When compiling files in directory @file{@var{old}}, record debugging
6458 information describing them as in @file{@var{new}} instead.
6460 @item -fno-dwarf2-cfi-asm
6461 @opindex fdwarf2-cfi-asm
6462 @opindex fno-dwarf2-cfi-asm
6463 Emit DWARF 2 unwind info as compiler generated @code{.eh_frame} section
6464 instead of using GAS @code{.cfi_*} directives.
6466 @cindex @command{prof}
6469 Generate extra code to write profile information suitable for the
6470 analysis program @command{prof}. You must use this option when compiling
6471 the source files you want data about, and you must also use it when
6474 @cindex @command{gprof}
6477 Generate extra code to write profile information suitable for the
6478 analysis program @command{gprof}. You must use this option when compiling
6479 the source files you want data about, and you must also use it when
6484 Makes the compiler print out each function name as it is compiled, and
6485 print some statistics about each pass when it finishes.
6488 @opindex ftime-report
6489 Makes the compiler print some statistics about the time consumed by each
6490 pass when it finishes.
6493 @opindex fmem-report
6494 Makes the compiler print some statistics about permanent memory
6495 allocation when it finishes.
6497 @item -fmem-report-wpa
6498 @opindex fmem-report-wpa
6499 Makes the compiler print some statistics about permanent memory
6500 allocation for the WPA phase only.
6502 @item -fpre-ipa-mem-report
6503 @opindex fpre-ipa-mem-report
6504 @item -fpost-ipa-mem-report
6505 @opindex fpost-ipa-mem-report
6506 Makes the compiler print some statistics about permanent memory
6507 allocation before or after interprocedural optimization.
6509 @item -fprofile-report
6510 @opindex fprofile-report
6511 Makes the compiler print some statistics about consistency of the
6512 (estimated) profile and effect of individual passes.
6515 @opindex fstack-usage
6516 Makes the compiler output stack usage information for the program, on a
6517 per-function basis. The filename for the dump is made by appending
6518 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
6519 the output file, if explicitly specified and it is not an executable,
6520 otherwise it is the basename of the source file. An entry is made up
6525 The name of the function.
6529 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
6532 The qualifier @code{static} means that the function manipulates the stack
6533 statically: a fixed number of bytes are allocated for the frame on function
6534 entry and released on function exit; no stack adjustments are otherwise made
6535 in the function. The second field is this fixed number of bytes.
6537 The qualifier @code{dynamic} means that the function manipulates the stack
6538 dynamically: in addition to the static allocation described above, stack
6539 adjustments are made in the body of the function, for example to push/pop
6540 arguments around function calls. If the qualifier @code{bounded} is also
6541 present, the amount of these adjustments is bounded at compile time and
6542 the second field is an upper bound of the total amount of stack used by
6543 the function. If it is not present, the amount of these adjustments is
6544 not bounded at compile time and the second field only represents the
6547 @item -fprofile-arcs
6548 @opindex fprofile-arcs
6549 Add code so that program flow @dfn{arcs} are instrumented. During
6550 execution the program records how many times each branch and call is
6551 executed and how many times it is taken or returns. When the compiled
6552 program exits it saves this data to a file called
6553 @file{@var{auxname}.gcda} for each source file. The data may be used for
6554 profile-directed optimizations (@option{-fbranch-probabilities}), or for
6555 test coverage analysis (@option{-ftest-coverage}). Each object file's
6556 @var{auxname} is generated from the name of the output file, if
6557 explicitly specified and it is not the final executable, otherwise it is
6558 the basename of the source file. In both cases any suffix is removed
6559 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
6560 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
6561 @xref{Cross-profiling}.
6563 @cindex @command{gcov}
6567 This option is used to compile and link code instrumented for coverage
6568 analysis. The option is a synonym for @option{-fprofile-arcs}
6569 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
6570 linking). See the documentation for those options for more details.
6575 Compile the source files with @option{-fprofile-arcs} plus optimization
6576 and code generation options. For test coverage analysis, use the
6577 additional @option{-ftest-coverage} option. You do not need to profile
6578 every source file in a program.
6581 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
6582 (the latter implies the former).
6585 Run the program on a representative workload to generate the arc profile
6586 information. This may be repeated any number of times. You can run
6587 concurrent instances of your program, and provided that the file system
6588 supports locking, the data files will be correctly updated. Also
6589 @code{fork} calls are detected and correctly handled (double counting
6593 For profile-directed optimizations, compile the source files again with
6594 the same optimization and code generation options plus
6595 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
6596 Control Optimization}).
6599 For test coverage analysis, use @command{gcov} to produce human readable
6600 information from the @file{.gcno} and @file{.gcda} files. Refer to the
6601 @command{gcov} documentation for further information.
6605 With @option{-fprofile-arcs}, for each function of your program GCC
6606 creates a program flow graph, then finds a spanning tree for the graph.
6607 Only arcs that are not on the spanning tree have to be instrumented: the
6608 compiler adds code to count the number of times that these arcs are
6609 executed. When an arc is the only exit or only entrance to a block, the
6610 instrumentation code can be added to the block; otherwise, a new basic
6611 block must be created to hold the instrumentation code.
6614 @item -ftest-coverage
6615 @opindex ftest-coverage
6616 Produce a notes file that the @command{gcov} code-coverage utility
6617 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
6618 show program coverage. Each source file's note file is called
6619 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
6620 above for a description of @var{auxname} and instructions on how to
6621 generate test coverage data. Coverage data matches the source files
6622 more closely if you do not optimize.
6624 @item -fdbg-cnt-list
6625 @opindex fdbg-cnt-list
6626 Print the name and the counter upper bound for all debug counters.
6629 @item -fdbg-cnt=@var{counter-value-list}
6631 Set the internal debug counter upper bound. @var{counter-value-list}
6632 is a comma-separated list of @var{name}:@var{value} pairs
6633 which sets the upper bound of each debug counter @var{name} to @var{value}.
6634 All debug counters have the initial upper bound of @code{UINT_MAX};
6635 thus @code{dbg_cnt} returns true always unless the upper bound
6636 is set by this option.
6637 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
6638 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
6640 @item -fenable-@var{kind}-@var{pass}
6641 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
6645 This is a set of options that are used to explicitly disable/enable
6646 optimization passes. These options are intended for use for debugging GCC.
6647 Compiler users should use regular options for enabling/disabling
6652 @item -fdisable-ipa-@var{pass}
6653 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6654 statically invoked in the compiler multiple times, the pass name should be
6655 appended with a sequential number starting from 1.
6657 @item -fdisable-rtl-@var{pass}
6658 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
6659 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
6660 statically invoked in the compiler multiple times, the pass name should be
6661 appended with a sequential number starting from 1. @var{range-list} is a
6662 comma-separated list of function ranges or assembler names. Each range is a number
6663 pair separated by a colon. The range is inclusive in both ends. If the range
6664 is trivial, the number pair can be simplified as a single number. If the
6665 function's call graph node's @var{uid} falls within one of the specified ranges,
6666 the @var{pass} is disabled for that function. The @var{uid} is shown in the
6667 function header of a dump file, and the pass names can be dumped by using
6668 option @option{-fdump-passes}.
6670 @item -fdisable-tree-@var{pass}
6671 @itemx -fdisable-tree-@var{pass}=@var{range-list}
6672 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
6675 @item -fenable-ipa-@var{pass}
6676 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
6677 statically invoked in the compiler multiple times, the pass name should be
6678 appended with a sequential number starting from 1.
6680 @item -fenable-rtl-@var{pass}
6681 @itemx -fenable-rtl-@var{pass}=@var{range-list}
6682 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
6683 description and examples.
6685 @item -fenable-tree-@var{pass}
6686 @itemx -fenable-tree-@var{pass}=@var{range-list}
6687 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
6688 of option arguments.
6692 Here are some examples showing uses of these options.
6696 # disable ccp1 for all functions
6698 # disable complete unroll for function whose cgraph node uid is 1
6699 -fenable-tree-cunroll=1
6700 # disable gcse2 for functions at the following ranges [1,1],
6701 # [300,400], and [400,1000]
6702 # disable gcse2 for functions foo and foo2
6703 -fdisable-rtl-gcse2=foo,foo2
6704 # disable early inlining
6705 -fdisable-tree-einline
6706 # disable ipa inlining
6707 -fdisable-ipa-inline
6708 # enable tree full unroll
6709 -fenable-tree-unroll
6713 @item -d@var{letters}
6714 @itemx -fdump-rtl-@var{pass}
6715 @itemx -fdump-rtl-@var{pass}=@var{filename}
6717 @opindex fdump-rtl-@var{pass}
6718 Says to make debugging dumps during compilation at times specified by
6719 @var{letters}. This is used for debugging the RTL-based passes of the
6720 compiler. The file names for most of the dumps are made by appending
6721 a pass number and a word to the @var{dumpname}, and the files are
6722 created in the directory of the output file. In case of
6723 @option{=@var{filename}} option, the dump is output on the given file
6724 instead of the pass numbered dump files. Note that the pass number is
6725 assigned as passes are registered into the pass manager. Most passes
6726 are registered in the order that they will execute and for these passes
6727 the number corresponds to the pass execution order. However, passes
6728 registered by plugins, passes specific to compilation targets, or
6729 passes that are otherwise registered after all the other passes are
6730 numbered higher than a pass named "final", even if they are executed
6731 earlier. @var{dumpname} is generated from the name of the output
6732 file if explicitly specified and not an executable, otherwise it is
6733 the basename of the source file. These switches may have different
6734 effects when @option{-E} is used for preprocessing.
6736 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
6737 @option{-d} option @var{letters}. Here are the possible
6738 letters for use in @var{pass} and @var{letters}, and their meanings:
6742 @item -fdump-rtl-alignments
6743 @opindex fdump-rtl-alignments
6744 Dump after branch alignments have been computed.
6746 @item -fdump-rtl-asmcons
6747 @opindex fdump-rtl-asmcons
6748 Dump after fixing rtl statements that have unsatisfied in/out constraints.
6750 @item -fdump-rtl-auto_inc_dec
6751 @opindex fdump-rtl-auto_inc_dec
6752 Dump after auto-inc-dec discovery. This pass is only run on
6753 architectures that have auto inc or auto dec instructions.
6755 @item -fdump-rtl-barriers
6756 @opindex fdump-rtl-barriers
6757 Dump after cleaning up the barrier instructions.
6759 @item -fdump-rtl-bbpart
6760 @opindex fdump-rtl-bbpart
6761 Dump after partitioning hot and cold basic blocks.
6763 @item -fdump-rtl-bbro
6764 @opindex fdump-rtl-bbro
6765 Dump after block reordering.
6767 @item -fdump-rtl-btl1
6768 @itemx -fdump-rtl-btl2
6769 @opindex fdump-rtl-btl2
6770 @opindex fdump-rtl-btl2
6771 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
6772 after the two branch
6773 target load optimization passes.
6775 @item -fdump-rtl-bypass
6776 @opindex fdump-rtl-bypass
6777 Dump after jump bypassing and control flow optimizations.
6779 @item -fdump-rtl-combine
6780 @opindex fdump-rtl-combine
6781 Dump after the RTL instruction combination pass.
6783 @item -fdump-rtl-compgotos
6784 @opindex fdump-rtl-compgotos
6785 Dump after duplicating the computed gotos.
6787 @item -fdump-rtl-ce1
6788 @itemx -fdump-rtl-ce2
6789 @itemx -fdump-rtl-ce3
6790 @opindex fdump-rtl-ce1
6791 @opindex fdump-rtl-ce2
6792 @opindex fdump-rtl-ce3
6793 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
6794 @option{-fdump-rtl-ce3} enable dumping after the three
6795 if conversion passes.
6797 @item -fdump-rtl-cprop_hardreg
6798 @opindex fdump-rtl-cprop_hardreg
6799 Dump after hard register copy propagation.
6801 @item -fdump-rtl-csa
6802 @opindex fdump-rtl-csa
6803 Dump after combining stack adjustments.
6805 @item -fdump-rtl-cse1
6806 @itemx -fdump-rtl-cse2
6807 @opindex fdump-rtl-cse1
6808 @opindex fdump-rtl-cse2
6809 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
6810 the two common subexpression elimination passes.
6812 @item -fdump-rtl-dce
6813 @opindex fdump-rtl-dce
6814 Dump after the standalone dead code elimination passes.
6816 @item -fdump-rtl-dbr
6817 @opindex fdump-rtl-dbr
6818 Dump after delayed branch scheduling.
6820 @item -fdump-rtl-dce1
6821 @itemx -fdump-rtl-dce2
6822 @opindex fdump-rtl-dce1
6823 @opindex fdump-rtl-dce2
6824 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
6825 the two dead store elimination passes.
6828 @opindex fdump-rtl-eh
6829 Dump after finalization of EH handling code.
6831 @item -fdump-rtl-eh_ranges
6832 @opindex fdump-rtl-eh_ranges
6833 Dump after conversion of EH handling range regions.
6835 @item -fdump-rtl-expand
6836 @opindex fdump-rtl-expand
6837 Dump after RTL generation.
6839 @item -fdump-rtl-fwprop1
6840 @itemx -fdump-rtl-fwprop2
6841 @opindex fdump-rtl-fwprop1
6842 @opindex fdump-rtl-fwprop2
6843 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
6844 dumping after the two forward propagation passes.
6846 @item -fdump-rtl-gcse1
6847 @itemx -fdump-rtl-gcse2
6848 @opindex fdump-rtl-gcse1
6849 @opindex fdump-rtl-gcse2
6850 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
6851 after global common subexpression elimination.
6853 @item -fdump-rtl-init-regs
6854 @opindex fdump-rtl-init-regs
6855 Dump after the initialization of the registers.
6857 @item -fdump-rtl-initvals
6858 @opindex fdump-rtl-initvals
6859 Dump after the computation of the initial value sets.
6861 @item -fdump-rtl-into_cfglayout
6862 @opindex fdump-rtl-into_cfglayout
6863 Dump after converting to cfglayout mode.
6865 @item -fdump-rtl-ira
6866 @opindex fdump-rtl-ira
6867 Dump after iterated register allocation.
6869 @item -fdump-rtl-jump
6870 @opindex fdump-rtl-jump
6871 Dump after the second jump optimization.
6873 @item -fdump-rtl-loop2
6874 @opindex fdump-rtl-loop2
6875 @option{-fdump-rtl-loop2} enables dumping after the rtl
6876 loop optimization passes.
6878 @item -fdump-rtl-mach
6879 @opindex fdump-rtl-mach
6880 Dump after performing the machine dependent reorganization pass, if that
6883 @item -fdump-rtl-mode_sw
6884 @opindex fdump-rtl-mode_sw
6885 Dump after removing redundant mode switches.
6887 @item -fdump-rtl-rnreg
6888 @opindex fdump-rtl-rnreg
6889 Dump after register renumbering.
6891 @item -fdump-rtl-outof_cfglayout
6892 @opindex fdump-rtl-outof_cfglayout
6893 Dump after converting from cfglayout mode.
6895 @item -fdump-rtl-peephole2
6896 @opindex fdump-rtl-peephole2
6897 Dump after the peephole pass.
6899 @item -fdump-rtl-postreload
6900 @opindex fdump-rtl-postreload
6901 Dump after post-reload optimizations.
6903 @item -fdump-rtl-pro_and_epilogue
6904 @opindex fdump-rtl-pro_and_epilogue
6905 Dump after generating the function prologues and epilogues.
6907 @item -fdump-rtl-sched1
6908 @itemx -fdump-rtl-sched2
6909 @opindex fdump-rtl-sched1
6910 @opindex fdump-rtl-sched2
6911 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
6912 after the basic block scheduling passes.
6914 @item -fdump-rtl-ree
6915 @opindex fdump-rtl-ree
6916 Dump after sign/zero extension elimination.
6918 @item -fdump-rtl-seqabstr
6919 @opindex fdump-rtl-seqabstr
6920 Dump after common sequence discovery.
6922 @item -fdump-rtl-shorten
6923 @opindex fdump-rtl-shorten
6924 Dump after shortening branches.
6926 @item -fdump-rtl-sibling
6927 @opindex fdump-rtl-sibling
6928 Dump after sibling call optimizations.
6930 @item -fdump-rtl-split1
6931 @itemx -fdump-rtl-split2
6932 @itemx -fdump-rtl-split3
6933 @itemx -fdump-rtl-split4
6934 @itemx -fdump-rtl-split5
6935 @opindex fdump-rtl-split1
6936 @opindex fdump-rtl-split2
6937 @opindex fdump-rtl-split3
6938 @opindex fdump-rtl-split4
6939 @opindex fdump-rtl-split5
6940 These options enable dumping after five rounds of
6941 instruction splitting.
6943 @item -fdump-rtl-sms
6944 @opindex fdump-rtl-sms
6945 Dump after modulo scheduling. This pass is only run on some
6948 @item -fdump-rtl-stack
6949 @opindex fdump-rtl-stack
6950 Dump after conversion from GCC's ``flat register file'' registers to the
6951 x87's stack-like registers. This pass is only run on x86 variants.
6953 @item -fdump-rtl-subreg1
6954 @itemx -fdump-rtl-subreg2
6955 @opindex fdump-rtl-subreg1
6956 @opindex fdump-rtl-subreg2
6957 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
6958 the two subreg expansion passes.
6960 @item -fdump-rtl-unshare
6961 @opindex fdump-rtl-unshare
6962 Dump after all rtl has been unshared.
6964 @item -fdump-rtl-vartrack
6965 @opindex fdump-rtl-vartrack
6966 Dump after variable tracking.
6968 @item -fdump-rtl-vregs
6969 @opindex fdump-rtl-vregs
6970 Dump after converting virtual registers to hard registers.
6972 @item -fdump-rtl-web
6973 @opindex fdump-rtl-web
6974 Dump after live range splitting.
6976 @item -fdump-rtl-regclass
6977 @itemx -fdump-rtl-subregs_of_mode_init
6978 @itemx -fdump-rtl-subregs_of_mode_finish
6979 @itemx -fdump-rtl-dfinit
6980 @itemx -fdump-rtl-dfinish
6981 @opindex fdump-rtl-regclass
6982 @opindex fdump-rtl-subregs_of_mode_init
6983 @opindex fdump-rtl-subregs_of_mode_finish
6984 @opindex fdump-rtl-dfinit
6985 @opindex fdump-rtl-dfinish
6986 These dumps are defined but always produce empty files.
6989 @itemx -fdump-rtl-all
6991 @opindex fdump-rtl-all
6992 Produce all the dumps listed above.
6996 Annotate the assembler output with miscellaneous debugging information.
7000 Dump all macro definitions, at the end of preprocessing, in addition to
7005 Produce a core dump whenever an error occurs.
7009 Annotate the assembler output with a comment indicating which
7010 pattern and alternative is used. The length of each instruction is
7015 Dump the RTL in the assembler output as a comment before each instruction.
7016 Also turns on @option{-dp} annotation.
7020 Just generate RTL for a function instead of compiling it. Usually used
7021 with @option{-fdump-rtl-expand}.
7025 @opindex fdump-noaddr
7026 When doing debugging dumps, suppress address output. This makes it more
7027 feasible to use diff on debugging dumps for compiler invocations with
7028 different compiler binaries and/or different
7029 text / bss / data / heap / stack / dso start locations.
7032 @opindex freport-bug
7033 Collect and dump debug information into temporary file if ICE in C/C++
7036 @item -fdump-unnumbered
7037 @opindex fdump-unnumbered
7038 When doing debugging dumps, suppress instruction numbers and address output.
7039 This makes it more feasible to use diff on debugging dumps for compiler
7040 invocations with different options, in particular with and without
7043 @item -fdump-unnumbered-links
7044 @opindex fdump-unnumbered-links
7045 When doing debugging dumps (see @option{-d} option above), suppress
7046 instruction numbers for the links to the previous and next instructions
7049 @item -fdump-translation-unit @r{(C++ only)}
7050 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
7051 @opindex fdump-translation-unit
7052 Dump a representation of the tree structure for the entire translation
7053 unit to a file. The file name is made by appending @file{.tu} to the
7054 source file name, and the file is created in the same directory as the
7055 output file. If the @samp{-@var{options}} form is used, @var{options}
7056 controls the details of the dump as described for the
7057 @option{-fdump-tree} options.
7059 @item -fdump-class-hierarchy @r{(C++ only)}
7060 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
7061 @opindex fdump-class-hierarchy
7062 Dump a representation of each class's hierarchy and virtual function
7063 table layout to a file. The file name is made by appending
7064 @file{.class} to the source file name, and the file is created in the
7065 same directory as the output file. If the @samp{-@var{options}} form
7066 is used, @var{options} controls the details of the dump as described
7067 for the @option{-fdump-tree} options.
7069 @item -fdump-ipa-@var{switch}
7071 Control the dumping at various stages of inter-procedural analysis
7072 language tree to a file. The file name is generated by appending a
7073 switch specific suffix to the source file name, and the file is created
7074 in the same directory as the output file. The following dumps are
7079 Enables all inter-procedural analysis dumps.
7082 Dumps information about call-graph optimization, unused function removal,
7083 and inlining decisions.
7086 Dump after function inlining.
7091 @opindex fdump-passes
7092 Dump the list of optimization passes that are turned on and off by
7093 the current command-line options.
7095 @item -fdump-statistics-@var{option}
7096 @opindex fdump-statistics
7097 Enable and control dumping of pass statistics in a separate file. The
7098 file name is generated by appending a suffix ending in
7099 @samp{.statistics} to the source file name, and the file is created in
7100 the same directory as the output file. If the @samp{-@var{option}}
7101 form is used, @samp{-stats} causes counters to be summed over the
7102 whole compilation unit while @samp{-details} dumps every event as
7103 the passes generate them. The default with no option is to sum
7104 counters for each function compiled.
7106 @item -fdump-tree-@var{switch}
7107 @itemx -fdump-tree-@var{switch}-@var{options}
7108 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
7110 Control the dumping at various stages of processing the intermediate
7111 language tree to a file. The file name is generated by appending a
7112 switch-specific suffix to the source file name, and the file is
7113 created in the same directory as the output file. In case of
7114 @option{=@var{filename}} option, the dump is output on the given file
7115 instead of the auto named dump files. If the @samp{-@var{options}}
7116 form is used, @var{options} is a list of @samp{-} separated options
7117 which control the details of the dump. Not all options are applicable
7118 to all dumps; those that are not meaningful are ignored. The
7119 following options are available
7123 Print the address of each node. Usually this is not meaningful as it
7124 changes according to the environment and source file. Its primary use
7125 is for tying up a dump file with a debug environment.
7127 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
7128 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
7129 use working backward from mangled names in the assembly file.
7131 When dumping front-end intermediate representations, inhibit dumping
7132 of members of a scope or body of a function merely because that scope
7133 has been reached. Only dump such items when they are directly reachable
7136 When dumping pretty-printed trees, this option inhibits dumping the
7137 bodies of control structures.
7139 When dumping RTL, print the RTL in slim (condensed) form instead of
7140 the default LISP-like representation.
7142 Print a raw representation of the tree. By default, trees are
7143 pretty-printed into a C-like representation.
7145 Enable more detailed dumps (not honored by every dump option). Also
7146 include information from the optimization passes.
7148 Enable dumping various statistics about the pass (not honored by every dump
7151 Enable showing basic block boundaries (disabled in raw dumps).
7153 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
7154 dump a representation of the control flow graph suitable for viewing with
7155 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
7156 the file is pretty-printed as a subgraph, so that GraphViz can render them
7157 all in a single plot.
7159 This option currently only works for RTL dumps, and the RTL is always
7160 dumped in slim form.
7162 Enable showing virtual operands for every statement.
7164 Enable showing line numbers for statements.
7166 Enable showing the unique ID (@code{DECL_UID}) for each variable.
7168 Enable showing the tree dump for each statement.
7170 Enable showing the EH region number holding each statement.
7172 Enable showing scalar evolution analysis details.
7174 Enable showing optimization information (only available in certain
7177 Enable showing missed optimization information (only available in certain
7180 Enable other detailed optimization information (only available in
7182 @item =@var{filename}
7183 Instead of an auto named dump file, output into the given file
7184 name. The file names @file{stdout} and @file{stderr} are treated
7185 specially and are considered already open standard streams. For
7189 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
7190 -fdump-tree-pre=stderr file.c
7193 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
7194 output on to @file{stderr}. If two conflicting dump filenames are
7195 given for the same pass, then the latter option overrides the earlier
7199 @opindex fdump-tree-split-paths
7200 Dump each function after splitting paths to loop backedges. The file
7201 name is made by appending @file{.split-paths} to the source file name.
7204 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
7205 and @option{lineno}.
7208 Turn on all optimization options, i.e., @option{optimized},
7209 @option{missed}, and @option{note}.
7212 The following tree dumps are possible:
7216 @opindex fdump-tree-original
7217 Dump before any tree based optimization, to @file{@var{file}.original}.
7220 @opindex fdump-tree-optimized
7221 Dump after all tree based optimization, to @file{@var{file}.optimized}.
7224 @opindex fdump-tree-gimple
7225 Dump each function before and after the gimplification pass to a file. The
7226 file name is made by appending @file{.gimple} to the source file name.
7229 @opindex fdump-tree-cfg
7230 Dump the control flow graph of each function to a file. The file name is
7231 made by appending @file{.cfg} to the source file name.
7234 @opindex fdump-tree-ch
7235 Dump each function after copying loop headers. The file name is made by
7236 appending @file{.ch} to the source file name.
7239 @opindex fdump-tree-ssa
7240 Dump SSA related information to a file. The file name is made by appending
7241 @file{.ssa} to the source file name.
7244 @opindex fdump-tree-alias
7245 Dump aliasing information for each function. The file name is made by
7246 appending @file{.alias} to the source file name.
7249 @opindex fdump-tree-ccp
7250 Dump each function after CCP@. The file name is made by appending
7251 @file{.ccp} to the source file name.
7254 @opindex fdump-tree-storeccp
7255 Dump each function after STORE-CCP@. The file name is made by appending
7256 @file{.storeccp} to the source file name.
7259 @opindex fdump-tree-pre
7260 Dump trees after partial redundancy elimination. The file name is made
7261 by appending @file{.pre} to the source file name.
7264 @opindex fdump-tree-fre
7265 Dump trees after full redundancy elimination. The file name is made
7266 by appending @file{.fre} to the source file name.
7269 @opindex fdump-tree-copyprop
7270 Dump trees after copy propagation. The file name is made
7271 by appending @file{.copyprop} to the source file name.
7273 @item store_copyprop
7274 @opindex fdump-tree-store_copyprop
7275 Dump trees after store copy-propagation. The file name is made
7276 by appending @file{.store_copyprop} to the source file name.
7279 @opindex fdump-tree-dce
7280 Dump each function after dead code elimination. The file name is made by
7281 appending @file{.dce} to the source file name.
7284 @opindex fdump-tree-sra
7285 Dump each function after performing scalar replacement of aggregates. The
7286 file name is made by appending @file{.sra} to the source file name.
7289 @opindex fdump-tree-sink
7290 Dump each function after performing code sinking. The file name is made
7291 by appending @file{.sink} to the source file name.
7294 @opindex fdump-tree-dom
7295 Dump each function after applying dominator tree optimizations. The file
7296 name is made by appending @file{.dom} to the source file name.
7299 @opindex fdump-tree-dse
7300 Dump each function after applying dead store elimination. The file
7301 name is made by appending @file{.dse} to the source file name.
7304 @opindex fdump-tree-phiopt
7305 Dump each function after optimizing PHI nodes into straightline code. The file
7306 name is made by appending @file{.phiopt} to the source file name.
7309 @opindex fdump-tree-backprop
7310 Dump each function after back-propagating use information up the definition
7311 chain. The file name is made by appending @file{.backprop} to the
7315 @opindex fdump-tree-forwprop
7316 Dump each function after forward propagating single use variables. The file
7317 name is made by appending @file{.forwprop} to the source file name.
7320 @opindex fdump-tree-nrv
7321 Dump each function after applying the named return value optimization on
7322 generic trees. The file name is made by appending @file{.nrv} to the source
7326 @opindex fdump-tree-vect
7327 Dump each function after applying vectorization of loops. The file name is
7328 made by appending @file{.vect} to the source file name.
7331 @opindex fdump-tree-slp
7332 Dump each function after applying vectorization of basic blocks. The file name
7333 is made by appending @file{.slp} to the source file name.
7336 @opindex fdump-tree-vrp
7337 Dump each function after Value Range Propagation (VRP). The file name
7338 is made by appending @file{.vrp} to the source file name.
7341 @opindex fdump-tree-oaccdevlow
7342 Dump each function after applying device-specific OpenACC transformations.
7343 The file name is made by appending @file{.oaccdevlow} to the source file name.
7346 @opindex fdump-tree-all
7347 Enable all the available tree dumps with the flags provided in this option.
7351 @itemx -fopt-info-@var{options}
7352 @itemx -fopt-info-@var{options}=@var{filename}
7354 Controls optimization dumps from various optimization passes. If the
7355 @samp{-@var{options}} form is used, @var{options} is a list of
7356 @samp{-} separated option keywords to select the dump details and
7359 The @var{options} can be divided into two groups: options describing the
7360 verbosity of the dump, and options describing which optimizations
7361 should be included. The options from both the groups can be freely
7362 mixed as they are non-overlapping. However, in case of any conflicts,
7363 the later options override the earlier options on the command
7366 The following options control the dump verbosity:
7370 Print information when an optimization is successfully applied. It is
7371 up to a pass to decide which information is relevant. For example, the
7372 vectorizer passes print the source location of loops which are
7373 successfully vectorized.
7375 Print information about missed optimizations. Individual passes
7376 control which information to include in the output.
7378 Print verbose information about optimizations, such as certain
7379 transformations, more detailed messages about decisions etc.
7381 Print detailed optimization information. This includes
7382 @samp{optimized}, @samp{missed}, and @samp{note}.
7385 One or more of the following option keywords can be used to describe a
7386 group of optimizations:
7390 Enable dumps from all interprocedural optimizations.
7392 Enable dumps from all loop optimizations.
7394 Enable dumps from all inlining optimizations.
7396 Enable dumps from all vectorization optimizations.
7398 Enable dumps from all optimizations. This is a superset of
7399 the optimization groups listed above.
7403 omitted, it defaults to @samp{optimized-optall}, which means to dump all
7404 info about successful optimizations from all the passes.
7406 If the @var{filename} is provided, then the dumps from all the
7407 applicable optimizations are concatenated into the @var{filename}.
7408 Otherwise the dump is output onto @file{stderr}. Though multiple
7409 @option{-fopt-info} options are accepted, only one of them can include
7410 a @var{filename}. If other filenames are provided then all but the
7411 first such option are ignored.
7413 Note that the output @var{filename} is overwritten
7414 in case of multiple translation units. If a combined output from
7415 multiple translation units is desired, @file{stderr} should be used
7418 In the following example, the optimization info is output to
7427 gcc -O3 -fopt-info-missed=missed.all
7431 outputs missed optimization report from all the passes into
7432 @file{missed.all}, and this one:
7435 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
7439 prints information about missed optimization opportunities from
7440 vectorization passes on @file{stderr}.
7441 Note that @option{-fopt-info-vec-missed} is equivalent to
7442 @option{-fopt-info-missed-vec}.
7446 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
7450 outputs information about missed optimizations as well as
7451 optimized locations from all the inlining passes into
7457 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
7461 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
7462 in conflict since only one output file is allowed. In this case, only
7463 the first option takes effect and the subsequent options are
7464 ignored. Thus only @file{vec.miss} is produced which contains
7465 dumps from the vectorizer about missed opportunities.
7467 @item -frandom-seed=@var{number}
7468 @opindex frandom-seed
7469 This option provides a seed that GCC uses in place of
7470 random numbers in generating certain symbol names
7471 that have to be different in every compiled file. It is also used to
7472 place unique stamps in coverage data files and the object files that
7473 produce them. You can use the @option{-frandom-seed} option to produce
7474 reproducibly identical object files.
7476 The @var{number} should be different for every file you compile.
7478 @item -fsched-verbose=@var{n}
7479 @opindex fsched-verbose
7480 On targets that use instruction scheduling, this option controls the
7481 amount of debugging output the scheduler prints to the dump files.
7483 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
7484 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
7485 For @var{n} greater than one, it also output basic block probabilities,
7486 detailed ready list information and unit/insn info. For @var{n} greater
7487 than two, it includes RTL at abort point, control-flow and regions info.
7488 And for @var{n} over four, @option{-fsched-verbose} also includes
7492 @itemx -save-temps=cwd
7494 Store the usual ``temporary'' intermediate files permanently; place them
7495 in the current directory and name them based on the source file. Thus,
7496 compiling @file{foo.c} with @option{-c -save-temps} produces files
7497 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
7498 preprocessed @file{foo.i} output file even though the compiler now
7499 normally uses an integrated preprocessor.
7501 When used in combination with the @option{-x} command-line option,
7502 @option{-save-temps} is sensible enough to avoid over writing an
7503 input source file with the same extension as an intermediate file.
7504 The corresponding intermediate file may be obtained by renaming the
7505 source file before using @option{-save-temps}.
7507 If you invoke GCC in parallel, compiling several different source
7508 files that share a common base name in different subdirectories or the
7509 same source file compiled for multiple output destinations, it is
7510 likely that the different parallel compilers will interfere with each
7511 other, and overwrite the temporary files. For instance:
7514 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
7515 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
7518 may result in @file{foo.i} and @file{foo.o} being written to
7519 simultaneously by both compilers.
7521 @item -save-temps=obj
7522 @opindex save-temps=obj
7523 Store the usual ``temporary'' intermediate files permanently. If the
7524 @option{-o} option is used, the temporary files are based on the
7525 object file. If the @option{-o} option is not used, the
7526 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
7531 gcc -save-temps=obj -c foo.c
7532 gcc -save-temps=obj -c bar.c -o dir/xbar.o
7533 gcc -save-temps=obj foobar.c -o dir2/yfoobar
7537 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
7538 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
7539 @file{dir2/yfoobar.o}.
7541 @item -time@r{[}=@var{file}@r{]}
7543 Report the CPU time taken by each subprocess in the compilation
7544 sequence. For C source files, this is the compiler proper and assembler
7545 (plus the linker if linking is done).
7547 Without the specification of an output file, the output looks like this:
7554 The first number on each line is the ``user time'', that is time spent
7555 executing the program itself. The second number is ``system time'',
7556 time spent executing operating system routines on behalf of the program.
7557 Both numbers are in seconds.
7559 With the specification of an output file, the output is appended to the
7560 named file, and it looks like this:
7563 0.12 0.01 cc1 @var{options}
7564 0.00 0.01 as @var{options}
7567 The ``user time'' and the ``system time'' are moved before the program
7568 name, and the options passed to the program are displayed, so that one
7569 can later tell what file was being compiled, and with which options.
7571 @item -fvar-tracking
7572 @opindex fvar-tracking
7573 Run variable tracking pass. It computes where variables are stored at each
7574 position in code. Better debugging information is then generated
7575 (if the debugging information format supports this information).
7577 It is enabled by default when compiling with optimization (@option{-Os},
7578 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
7579 the debug info format supports it.
7581 @item -fvar-tracking-assignments
7582 @opindex fvar-tracking-assignments
7583 @opindex fno-var-tracking-assignments
7584 Annotate assignments to user variables early in the compilation and
7585 attempt to carry the annotations over throughout the compilation all the
7586 way to the end, in an attempt to improve debug information while
7587 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
7589 It can be enabled even if var-tracking is disabled, in which case
7590 annotations are created and maintained, but discarded at the end.
7591 By default, this flag is enabled together with @option{-fvar-tracking},
7592 except when selective scheduling is enabled.
7594 @item -fvar-tracking-assignments-toggle
7595 @opindex fvar-tracking-assignments-toggle
7596 @opindex fno-var-tracking-assignments-toggle
7597 Toggle @option{-fvar-tracking-assignments}, in the same way that
7598 @option{-gtoggle} toggles @option{-g}.
7600 @item -print-file-name=@var{library}
7601 @opindex print-file-name
7602 Print the full absolute name of the library file @var{library} that
7603 would be used when linking---and don't do anything else. With this
7604 option, GCC does not compile or link anything; it just prints the
7607 @item -print-multi-directory
7608 @opindex print-multi-directory
7609 Print the directory name corresponding to the multilib selected by any
7610 other switches present in the command line. This directory is supposed
7611 to exist in @env{GCC_EXEC_PREFIX}.
7613 @item -print-multi-lib
7614 @opindex print-multi-lib
7615 Print the mapping from multilib directory names to compiler switches
7616 that enable them. The directory name is separated from the switches by
7617 @samp{;}, and each switch starts with an @samp{@@} instead of the
7618 @samp{-}, without spaces between multiple switches. This is supposed to
7619 ease shell processing.
7621 @item -print-multi-os-directory
7622 @opindex print-multi-os-directory
7623 Print the path to OS libraries for the selected
7624 multilib, relative to some @file{lib} subdirectory. If OS libraries are
7625 present in the @file{lib} subdirectory and no multilibs are used, this is
7626 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
7627 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
7628 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
7629 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
7631 @item -print-multiarch
7632 @opindex print-multiarch
7633 Print the path to OS libraries for the selected multiarch,
7634 relative to some @file{lib} subdirectory.
7636 @item -print-prog-name=@var{program}
7637 @opindex print-prog-name
7638 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
7640 @item -print-libgcc-file-name
7641 @opindex print-libgcc-file-name
7642 Same as @option{-print-file-name=libgcc.a}.
7644 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
7645 but you do want to link with @file{libgcc.a}. You can do:
7648 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
7651 @item -print-search-dirs
7652 @opindex print-search-dirs
7653 Print the name of the configured installation directory and a list of
7654 program and library directories @command{gcc} searches---and don't do anything else.
7656 This is useful when @command{gcc} prints the error message
7657 @samp{installation problem, cannot exec cpp0: No such file or directory}.
7658 To resolve this you either need to put @file{cpp0} and the other compiler
7659 components where @command{gcc} expects to find them, or you can set the environment
7660 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
7661 Don't forget the trailing @samp{/}.
7662 @xref{Environment Variables}.
7664 @item -print-sysroot
7665 @opindex print-sysroot
7666 Print the target sysroot directory that is used during
7667 compilation. This is the target sysroot specified either at configure
7668 time or using the @option{--sysroot} option, possibly with an extra
7669 suffix that depends on compilation options. If no target sysroot is
7670 specified, the option prints nothing.
7672 @item -print-sysroot-headers-suffix
7673 @opindex print-sysroot-headers-suffix
7674 Print the suffix added to the target sysroot when searching for
7675 headers, or give an error if the compiler is not configured with such
7676 a suffix---and don't do anything else.
7679 @opindex dumpmachine
7680 Print the compiler's target machine (for example,
7681 @samp{i686-pc-linux-gnu})---and don't do anything else.
7684 @opindex dumpversion
7685 Print the compiler version (for example, @code{3.0})---and don't do
7690 Print the compiler's built-in specs---and don't do anything else. (This
7691 is used when GCC itself is being built.) @xref{Spec Files}.
7693 @item -fno-eliminate-unused-debug-types
7694 @opindex feliminate-unused-debug-types
7695 @opindex fno-eliminate-unused-debug-types
7696 Normally, when producing DWARF 2 output, GCC avoids producing debug symbol
7697 output for types that are nowhere used in the source file being compiled.
7698 Sometimes it is useful to have GCC emit debugging
7699 information for all types declared in a compilation
7700 unit, regardless of whether or not they are actually used
7701 in that compilation unit, for example
7702 if, in the debugger, you want to cast a value to a type that is
7703 not actually used in your program (but is declared). More often,
7704 however, this results in a significant amount of wasted space.
7707 @node Optimize Options
7708 @section Options That Control Optimization
7709 @cindex optimize options
7710 @cindex options, optimization
7712 These options control various sorts of optimizations.
7714 Without any optimization option, the compiler's goal is to reduce the
7715 cost of compilation and to make debugging produce the expected
7716 results. Statements are independent: if you stop the program with a
7717 breakpoint between statements, you can then assign a new value to any
7718 variable or change the program counter to any other statement in the
7719 function and get exactly the results you expect from the source
7722 Turning on optimization flags makes the compiler attempt to improve
7723 the performance and/or code size at the expense of compilation time
7724 and possibly the ability to debug the program.
7726 The compiler performs optimization based on the knowledge it has of the
7727 program. Compiling multiple files at once to a single output file mode allows
7728 the compiler to use information gained from all of the files when compiling
7731 Not all optimizations are controlled directly by a flag. Only
7732 optimizations that have a flag are listed in this section.
7734 Most optimizations are only enabled if an @option{-O} level is set on
7735 the command line. Otherwise they are disabled, even if individual
7736 optimization flags are specified.
7738 Depending on the target and how GCC was configured, a slightly different
7739 set of optimizations may be enabled at each @option{-O} level than
7740 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
7741 to find out the exact set of optimizations that are enabled at each level.
7742 @xref{Overall Options}, for examples.
7749 Optimize. Optimizing compilation takes somewhat more time, and a lot
7750 more memory for a large function.
7752 With @option{-O}, the compiler tries to reduce code size and execution
7753 time, without performing any optimizations that take a great deal of
7756 @option{-O} turns on the following optimization flags:
7759 -fbranch-count-reg @gol
7760 -fcombine-stack-adjustments @gol
7762 -fcprop-registers @gol
7765 -fdelayed-branch @gol
7767 -fforward-propagate @gol
7768 -fguess-branch-probability @gol
7769 -fif-conversion2 @gol
7770 -fif-conversion @gol
7771 -finline-functions-called-once @gol
7772 -fipa-pure-const @gol
7774 -fipa-reference @gol
7775 -fmerge-constants @gol
7776 -fmove-loop-invariants @gol
7777 -freorder-blocks @gol
7779 -fsplit-wide-types @gol
7785 -ftree-coalesce-vars @gol
7786 -ftree-copy-prop @gol
7788 -ftree-dominator-opts @gol
7790 -ftree-forwprop @gol
7800 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
7801 where doing so does not interfere with debugging.
7805 Optimize even more. GCC performs nearly all supported optimizations
7806 that do not involve a space-speed tradeoff.
7807 As compared to @option{-O}, this option increases both compilation time
7808 and the performance of the generated code.
7810 @option{-O2} turns on all optimization flags specified by @option{-O}. It
7811 also turns on the following optimization flags:
7812 @gccoptlist{-fthread-jumps @gol
7813 -falign-functions -falign-jumps @gol
7814 -falign-loops -falign-labels @gol
7817 -fcse-follow-jumps -fcse-skip-blocks @gol
7818 -fdelete-null-pointer-checks @gol
7819 -fdevirtualize -fdevirtualize-speculatively @gol
7820 -fexpensive-optimizations @gol
7821 -fgcse -fgcse-lm @gol
7822 -fhoist-adjacent-loads @gol
7823 -finline-small-functions @gol
7824 -findirect-inlining @gol
7826 -fipa-cp-alignment @gol
7829 -fisolate-erroneous-paths-dereference @gol
7831 -foptimize-sibling-calls @gol
7832 -foptimize-strlen @gol
7833 -fpartial-inlining @gol
7835 -freorder-blocks-algorithm=stc @gol
7836 -freorder-blocks-and-partition -freorder-functions @gol
7837 -frerun-cse-after-loop @gol
7838 -fsched-interblock -fsched-spec @gol
7839 -fschedule-insns -fschedule-insns2 @gol
7840 -fstrict-aliasing -fstrict-overflow @gol
7841 -ftree-builtin-call-dce @gol
7842 -ftree-switch-conversion -ftree-tail-merge @gol
7847 Please note the warning under @option{-fgcse} about
7848 invoking @option{-O2} on programs that use computed gotos.
7852 Optimize yet more. @option{-O3} turns on all optimizations specified
7853 by @option{-O2} and also turns on the @option{-finline-functions},
7854 @option{-funswitch-loops}, @option{-fpredictive-commoning},
7855 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
7856 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
7857 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
7858 @option{-ftree-partial-pre} and @option{-fipa-cp-clone} options.
7862 Reduce compilation time and make debugging produce the expected
7863 results. This is the default.
7867 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
7868 do not typically increase code size. It also performs further
7869 optimizations designed to reduce code size.
7871 @option{-Os} disables the following optimization flags:
7872 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
7873 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
7874 -freorder-blocks-and-partition -fprefetch-loop-arrays}
7878 Disregard strict standards compliance. @option{-Ofast} enables all
7879 @option{-O3} optimizations. It also enables optimizations that are not
7880 valid for all standard-compliant programs.
7881 It turns on @option{-ffast-math} and the Fortran-specific
7882 @option{-fno-protect-parens} and @option{-fstack-arrays}.
7886 Optimize debugging experience. @option{-Og} enables optimizations
7887 that do not interfere with debugging. It should be the optimization
7888 level of choice for the standard edit-compile-debug cycle, offering
7889 a reasonable level of optimization while maintaining fast compilation
7890 and a good debugging experience.
7893 If you use multiple @option{-O} options, with or without level numbers,
7894 the last such option is the one that is effective.
7896 Options of the form @option{-f@var{flag}} specify machine-independent
7897 flags. Most flags have both positive and negative forms; the negative
7898 form of @option{-ffoo} is @option{-fno-foo}. In the table
7899 below, only one of the forms is listed---the one you typically
7900 use. You can figure out the other form by either removing @samp{no-}
7903 The following options control specific optimizations. They are either
7904 activated by @option{-O} options or are related to ones that are. You
7905 can use the following flags in the rare cases when ``fine-tuning'' of
7906 optimizations to be performed is desired.
7909 @item -fno-defer-pop
7910 @opindex fno-defer-pop
7911 Always pop the arguments to each function call as soon as that function
7912 returns. For machines that must pop arguments after a function call,
7913 the compiler normally lets arguments accumulate on the stack for several
7914 function calls and pops them all at once.
7916 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7918 @item -fforward-propagate
7919 @opindex fforward-propagate
7920 Perform a forward propagation pass on RTL@. The pass tries to combine two
7921 instructions and checks if the result can be simplified. If loop unrolling
7922 is active, two passes are performed and the second is scheduled after
7925 This option is enabled by default at optimization levels @option{-O},
7926 @option{-O2}, @option{-O3}, @option{-Os}.
7928 @item -ffp-contract=@var{style}
7929 @opindex ffp-contract
7930 @option{-ffp-contract=off} disables floating-point expression contraction.
7931 @option{-ffp-contract=fast} enables floating-point expression contraction
7932 such as forming of fused multiply-add operations if the target has
7933 native support for them.
7934 @option{-ffp-contract=on} enables floating-point expression contraction
7935 if allowed by the language standard. This is currently not implemented
7936 and treated equal to @option{-ffp-contract=off}.
7938 The default is @option{-ffp-contract=fast}.
7940 @item -fomit-frame-pointer
7941 @opindex fomit-frame-pointer
7942 Don't keep the frame pointer in a register for functions that
7943 don't need one. This avoids the instructions to save, set up and
7944 restore frame pointers; it also makes an extra register available
7945 in many functions. @strong{It also makes debugging impossible on
7948 On some machines, such as the VAX, this flag has no effect, because
7949 the standard calling sequence automatically handles the frame pointer
7950 and nothing is saved by pretending it doesn't exist. The
7951 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
7952 whether a target machine supports this flag. @xref{Registers,,Register
7953 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
7955 The default setting (when not optimizing for
7956 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
7957 @option{-fomit-frame-pointer}. You can configure GCC with the
7958 @option{--enable-frame-pointer} configure option to change the default.
7960 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7962 @item -foptimize-sibling-calls
7963 @opindex foptimize-sibling-calls
7964 Optimize sibling and tail recursive calls.
7966 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7968 @item -foptimize-strlen
7969 @opindex foptimize-strlen
7970 Optimize various standard C string functions (e.g. @code{strlen},
7971 @code{strchr} or @code{strcpy}) and
7972 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
7974 Enabled at levels @option{-O2}, @option{-O3}.
7978 Do not expand any functions inline apart from those marked with
7979 the @code{always_inline} attribute. This is the default when not
7982 Single functions can be exempted from inlining by marking them
7983 with the @code{noinline} attribute.
7985 @item -finline-small-functions
7986 @opindex finline-small-functions
7987 Integrate functions into their callers when their body is smaller than expected
7988 function call code (so overall size of program gets smaller). The compiler
7989 heuristically decides which functions are simple enough to be worth integrating
7990 in this way. This inlining applies to all functions, even those not declared
7993 Enabled at level @option{-O2}.
7995 @item -findirect-inlining
7996 @opindex findirect-inlining
7997 Inline also indirect calls that are discovered to be known at compile
7998 time thanks to previous inlining. This option has any effect only
7999 when inlining itself is turned on by the @option{-finline-functions}
8000 or @option{-finline-small-functions} options.
8002 Enabled at level @option{-O2}.
8004 @item -finline-functions
8005 @opindex finline-functions
8006 Consider all functions for inlining, even if they are not declared inline.
8007 The compiler heuristically decides which functions are worth integrating
8010 If all calls to a given function are integrated, and the function is
8011 declared @code{static}, then the function is normally not output as
8012 assembler code in its own right.
8014 Enabled at level @option{-O3}.
8016 @item -finline-functions-called-once
8017 @opindex finline-functions-called-once
8018 Consider all @code{static} functions called once for inlining into their
8019 caller even if they are not marked @code{inline}. If a call to a given
8020 function is integrated, then the function is not output as assembler code
8023 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
8025 @item -fearly-inlining
8026 @opindex fearly-inlining
8027 Inline functions marked by @code{always_inline} and functions whose body seems
8028 smaller than the function call overhead early before doing
8029 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
8030 makes profiling significantly cheaper and usually inlining faster on programs
8031 having large chains of nested wrapper functions.
8037 Perform interprocedural scalar replacement of aggregates, removal of
8038 unused parameters and replacement of parameters passed by reference
8039 by parameters passed by value.
8041 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
8043 @item -finline-limit=@var{n}
8044 @opindex finline-limit
8045 By default, GCC limits the size of functions that can be inlined. This flag
8046 allows coarse control of this limit. @var{n} is the size of functions that
8047 can be inlined in number of pseudo instructions.
8049 Inlining is actually controlled by a number of parameters, which may be
8050 specified individually by using @option{--param @var{name}=@var{value}}.
8051 The @option{-finline-limit=@var{n}} option sets some of these parameters
8055 @item max-inline-insns-single
8056 is set to @var{n}/2.
8057 @item max-inline-insns-auto
8058 is set to @var{n}/2.
8061 See below for a documentation of the individual
8062 parameters controlling inlining and for the defaults of these parameters.
8064 @emph{Note:} there may be no value to @option{-finline-limit} that results
8065 in default behavior.
8067 @emph{Note:} pseudo instruction represents, in this particular context, an
8068 abstract measurement of function's size. In no way does it represent a count
8069 of assembly instructions and as such its exact meaning might change from one
8070 release to an another.
8072 @item -fno-keep-inline-dllexport
8073 @opindex fno-keep-inline-dllexport
8074 This is a more fine-grained version of @option{-fkeep-inline-functions},
8075 which applies only to functions that are declared using the @code{dllexport}
8076 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
8079 @item -fkeep-inline-functions
8080 @opindex fkeep-inline-functions
8081 In C, emit @code{static} functions that are declared @code{inline}
8082 into the object file, even if the function has been inlined into all
8083 of its callers. This switch does not affect functions using the
8084 @code{extern inline} extension in GNU C90@. In C++, emit any and all
8085 inline functions into the object file.
8087 @item -fkeep-static-functions
8088 @opindex fkeep-static-functions
8089 Emit @code{static} functions into the object file, even if the function
8092 @item -fkeep-static-consts
8093 @opindex fkeep-static-consts
8094 Emit variables declared @code{static const} when optimization isn't turned
8095 on, even if the variables aren't referenced.
8097 GCC enables this option by default. If you want to force the compiler to
8098 check if a variable is referenced, regardless of whether or not
8099 optimization is turned on, use the @option{-fno-keep-static-consts} option.
8101 @item -fmerge-constants
8102 @opindex fmerge-constants
8103 Attempt to merge identical constants (string constants and floating-point
8104 constants) across compilation units.
8106 This option is the default for optimized compilation if the assembler and
8107 linker support it. Use @option{-fno-merge-constants} to inhibit this
8110 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8112 @item -fmerge-all-constants
8113 @opindex fmerge-all-constants
8114 Attempt to merge identical constants and identical variables.
8116 This option implies @option{-fmerge-constants}. In addition to
8117 @option{-fmerge-constants} this considers e.g.@: even constant initialized
8118 arrays or initialized constant variables with integral or floating-point
8119 types. Languages like C or C++ require each variable, including multiple
8120 instances of the same variable in recursive calls, to have distinct locations,
8121 so using this option results in non-conforming
8124 @item -fmodulo-sched
8125 @opindex fmodulo-sched
8126 Perform swing modulo scheduling immediately before the first scheduling
8127 pass. This pass looks at innermost loops and reorders their
8128 instructions by overlapping different iterations.
8130 @item -fmodulo-sched-allow-regmoves
8131 @opindex fmodulo-sched-allow-regmoves
8132 Perform more aggressive SMS-based modulo scheduling with register moves
8133 allowed. By setting this flag certain anti-dependences edges are
8134 deleted, which triggers the generation of reg-moves based on the
8135 life-range analysis. This option is effective only with
8136 @option{-fmodulo-sched} enabled.
8138 @item -fno-branch-count-reg
8139 @opindex fno-branch-count-reg
8140 Do not use ``decrement and branch'' instructions on a count register,
8141 but instead generate a sequence of instructions that decrement a
8142 register, compare it against zero, then branch based upon the result.
8143 This option is only meaningful on architectures that support such
8144 instructions, which include x86, PowerPC, IA-64 and S/390.
8146 Enabled by default at @option{-O1} and higher.
8148 The default is @option{-fbranch-count-reg}.
8150 @item -fno-function-cse
8151 @opindex fno-function-cse
8152 Do not put function addresses in registers; make each instruction that
8153 calls a constant function contain the function's address explicitly.
8155 This option results in less efficient code, but some strange hacks
8156 that alter the assembler output may be confused by the optimizations
8157 performed when this option is not used.
8159 The default is @option{-ffunction-cse}
8161 @item -fno-zero-initialized-in-bss
8162 @opindex fno-zero-initialized-in-bss
8163 If the target supports a BSS section, GCC by default puts variables that
8164 are initialized to zero into BSS@. This can save space in the resulting
8167 This option turns off this behavior because some programs explicitly
8168 rely on variables going to the data section---e.g., so that the
8169 resulting executable can find the beginning of that section and/or make
8170 assumptions based on that.
8172 The default is @option{-fzero-initialized-in-bss}.
8174 @item -fthread-jumps
8175 @opindex fthread-jumps
8176 Perform optimizations that check to see if a jump branches to a
8177 location where another comparison subsumed by the first is found. If
8178 so, the first branch is redirected to either the destination of the
8179 second branch or a point immediately following it, depending on whether
8180 the condition is known to be true or false.
8182 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8184 @item -fsplit-wide-types
8185 @opindex fsplit-wide-types
8186 When using a type that occupies multiple registers, such as @code{long
8187 long} on a 32-bit system, split the registers apart and allocate them
8188 independently. This normally generates better code for those types,
8189 but may make debugging more difficult.
8191 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
8194 @item -fcse-follow-jumps
8195 @opindex fcse-follow-jumps
8196 In common subexpression elimination (CSE), scan through jump instructions
8197 when the target of the jump is not reached by any other path. For
8198 example, when CSE encounters an @code{if} statement with an
8199 @code{else} clause, CSE follows the jump when the condition
8202 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8204 @item -fcse-skip-blocks
8205 @opindex fcse-skip-blocks
8206 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
8207 follow jumps that conditionally skip over blocks. When CSE
8208 encounters a simple @code{if} statement with no else clause,
8209 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
8210 body of the @code{if}.
8212 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8214 @item -frerun-cse-after-loop
8215 @opindex frerun-cse-after-loop
8216 Re-run common subexpression elimination after loop optimizations are
8219 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8223 Perform a global common subexpression elimination pass.
8224 This pass also performs global constant and copy propagation.
8226 @emph{Note:} When compiling a program using computed gotos, a GCC
8227 extension, you may get better run-time performance if you disable
8228 the global common subexpression elimination pass by adding
8229 @option{-fno-gcse} to the command line.
8231 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8235 When @option{-fgcse-lm} is enabled, global common subexpression elimination
8236 attempts to move loads that are only killed by stores into themselves. This
8237 allows a loop containing a load/store sequence to be changed to a load outside
8238 the loop, and a copy/store within the loop.
8240 Enabled by default when @option{-fgcse} is enabled.
8244 When @option{-fgcse-sm} is enabled, a store motion pass is run after
8245 global common subexpression elimination. This pass attempts to move
8246 stores out of loops. When used in conjunction with @option{-fgcse-lm},
8247 loops containing a load/store sequence can be changed to a load before
8248 the loop and a store after the loop.
8250 Not enabled at any optimization level.
8254 When @option{-fgcse-las} is enabled, the global common subexpression
8255 elimination pass eliminates redundant loads that come after stores to the
8256 same memory location (both partial and full redundancies).
8258 Not enabled at any optimization level.
8260 @item -fgcse-after-reload
8261 @opindex fgcse-after-reload
8262 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
8263 pass is performed after reload. The purpose of this pass is to clean up
8266 @item -faggressive-loop-optimizations
8267 @opindex faggressive-loop-optimizations
8268 This option tells the loop optimizer to use language constraints to
8269 derive bounds for the number of iterations of a loop. This assumes that
8270 loop code does not invoke undefined behavior by for example causing signed
8271 integer overflows or out-of-bound array accesses. The bounds for the
8272 number of iterations of a loop are used to guide loop unrolling and peeling
8273 and loop exit test optimizations.
8274 This option is enabled by default.
8276 @item -funsafe-loop-optimizations
8277 @opindex funsafe-loop-optimizations
8278 This option tells the loop optimizer to assume that loop indices do not
8279 overflow, and that loops with nontrivial exit condition are not
8280 infinite. This enables a wider range of loop optimizations even if
8281 the loop optimizer itself cannot prove that these assumptions are valid.
8282 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
8283 if it finds this kind of loop.
8285 @item -fcrossjumping
8286 @opindex fcrossjumping
8287 Perform cross-jumping transformation.
8288 This transformation unifies equivalent code and saves code size. The
8289 resulting code may or may not perform better than without cross-jumping.
8291 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8293 @item -fauto-inc-dec
8294 @opindex fauto-inc-dec
8295 Combine increments or decrements of addresses with memory accesses.
8296 This pass is always skipped on architectures that do not have
8297 instructions to support this. Enabled by default at @option{-O} and
8298 higher on architectures that support this.
8302 Perform dead code elimination (DCE) on RTL@.
8303 Enabled by default at @option{-O} and higher.
8307 Perform dead store elimination (DSE) on RTL@.
8308 Enabled by default at @option{-O} and higher.
8310 @item -fif-conversion
8311 @opindex fif-conversion
8312 Attempt to transform conditional jumps into branch-less equivalents. This
8313 includes use of conditional moves, min, max, set flags and abs instructions, and
8314 some tricks doable by standard arithmetics. The use of conditional execution
8315 on chips where it is available is controlled by @option{-fif-conversion2}.
8317 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8319 @item -fif-conversion2
8320 @opindex fif-conversion2
8321 Use conditional execution (where available) to transform conditional jumps into
8322 branch-less equivalents.
8324 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8326 @item -fdeclone-ctor-dtor
8327 @opindex fdeclone-ctor-dtor
8328 The C++ ABI requires multiple entry points for constructors and
8329 destructors: one for a base subobject, one for a complete object, and
8330 one for a virtual destructor that calls operator delete afterwards.
8331 For a hierarchy with virtual bases, the base and complete variants are
8332 clones, which means two copies of the function. With this option, the
8333 base and complete variants are changed to be thunks that call a common
8336 Enabled by @option{-Os}.
8338 @item -fdelete-null-pointer-checks
8339 @opindex fdelete-null-pointer-checks
8340 Assume that programs cannot safely dereference null pointers, and that
8341 no code or data element resides at address zero.
8342 This option enables simple constant
8343 folding optimizations at all optimization levels. In addition, other
8344 optimization passes in GCC use this flag to control global dataflow
8345 analyses that eliminate useless checks for null pointers; these assume
8346 that a memory access to address zero always results in a trap, so
8347 that if a pointer is checked after it has already been dereferenced,
8350 Note however that in some environments this assumption is not true.
8351 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
8352 for programs that depend on that behavior.
8354 This option is enabled by default on most targets. On Nios II ELF, it
8355 defaults to off. On AVR and CR16, this option is completely disabled.
8357 Passes that use the dataflow information
8358 are enabled independently at different optimization levels.
8360 @item -fdevirtualize
8361 @opindex fdevirtualize
8362 Attempt to convert calls to virtual functions to direct calls. This
8363 is done both within a procedure and interprocedurally as part of
8364 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
8365 propagation (@option{-fipa-cp}).
8366 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8368 @item -fdevirtualize-speculatively
8369 @opindex fdevirtualize-speculatively
8370 Attempt to convert calls to virtual functions to speculative direct calls.
8371 Based on the analysis of the type inheritance graph, determine for a given call
8372 the set of likely targets. If the set is small, preferably of size 1, change
8373 the call into a conditional deciding between direct and indirect calls. The
8374 speculative calls enable more optimizations, such as inlining. When they seem
8375 useless after further optimization, they are converted back into original form.
8377 @item -fdevirtualize-at-ltrans
8378 @opindex fdevirtualize-at-ltrans
8379 Stream extra information needed for aggressive devirtualization when running
8380 the link-time optimizer in local transformation mode.
8381 This option enables more devirtualization but
8382 significantly increases the size of streamed data. For this reason it is
8383 disabled by default.
8385 @item -fexpensive-optimizations
8386 @opindex fexpensive-optimizations
8387 Perform a number of minor optimizations that are relatively expensive.
8389 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8393 Attempt to remove redundant extension instructions. This is especially
8394 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
8395 registers after writing to their lower 32-bit half.
8397 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
8398 @option{-O3}, @option{-Os}.
8400 @item -fno-lifetime-dse
8401 @opindex fno-lifetime-dse
8402 In C++ the value of an object is only affected by changes within its
8403 lifetime: when the constructor begins, the object has an indeterminate
8404 value, and any changes during the lifetime of the object are dead when
8405 the object is destroyed. Normally dead store elimination will take
8406 advantage of this; if your code relies on the value of the object
8407 storage persisting beyond the lifetime of the object, you can use this
8408 flag to disable this optimization.
8410 @item -flive-range-shrinkage
8411 @opindex flive-range-shrinkage
8412 Attempt to decrease register pressure through register live range
8413 shrinkage. This is helpful for fast processors with small or moderate
8416 @item -fira-algorithm=@var{algorithm}
8417 @opindex fira-algorithm
8418 Use the specified coloring algorithm for the integrated register
8419 allocator. The @var{algorithm} argument can be @samp{priority}, which
8420 specifies Chow's priority coloring, or @samp{CB}, which specifies
8421 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
8422 for all architectures, but for those targets that do support it, it is
8423 the default because it generates better code.
8425 @item -fira-region=@var{region}
8426 @opindex fira-region
8427 Use specified regions for the integrated register allocator. The
8428 @var{region} argument should be one of the following:
8433 Use all loops as register allocation regions.
8434 This can give the best results for machines with a small and/or
8435 irregular register set.
8438 Use all loops except for loops with small register pressure
8439 as the regions. This value usually gives
8440 the best results in most cases and for most architectures,
8441 and is enabled by default when compiling with optimization for speed
8442 (@option{-O}, @option{-O2}, @dots{}).
8445 Use all functions as a single region.
8446 This typically results in the smallest code size, and is enabled by default for
8447 @option{-Os} or @option{-O0}.
8451 @item -fira-hoist-pressure
8452 @opindex fira-hoist-pressure
8453 Use IRA to evaluate register pressure in the code hoisting pass for
8454 decisions to hoist expressions. This option usually results in smaller
8455 code, but it can slow the compiler down.
8457 This option is enabled at level @option{-Os} for all targets.
8459 @item -fira-loop-pressure
8460 @opindex fira-loop-pressure
8461 Use IRA to evaluate register pressure in loops for decisions to move
8462 loop invariants. This option usually results in generation
8463 of faster and smaller code on machines with large register files (>= 32
8464 registers), but it can slow the compiler down.
8466 This option is enabled at level @option{-O3} for some targets.
8468 @item -fno-ira-share-save-slots
8469 @opindex fno-ira-share-save-slots
8470 Disable sharing of stack slots used for saving call-used hard
8471 registers living through a call. Each hard register gets a
8472 separate stack slot, and as a result function stack frames are
8475 @item -fno-ira-share-spill-slots
8476 @opindex fno-ira-share-spill-slots
8477 Disable sharing of stack slots allocated for pseudo-registers. Each
8478 pseudo-register that does not get a hard register gets a separate
8479 stack slot, and as a result function stack frames are larger.
8481 @item -fira-verbose=@var{n}
8482 @opindex fira-verbose
8483 Control the verbosity of the dump file for the integrated register allocator.
8484 The default value is 5. If the value @var{n} is greater or equal to 10,
8485 the dump output is sent to stderr using the same format as @var{n} minus 10.
8489 Enable CFG-sensitive rematerialization in LRA. Instead of loading
8490 values of spilled pseudos, LRA tries to rematerialize (recalculate)
8491 values if it is profitable.
8493 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8495 @item -fdelayed-branch
8496 @opindex fdelayed-branch
8497 If supported for the target machine, attempt to reorder instructions
8498 to exploit instruction slots available after delayed branch
8501 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8503 @item -fschedule-insns
8504 @opindex fschedule-insns
8505 If supported for the target machine, attempt to reorder instructions to
8506 eliminate execution stalls due to required data being unavailable. This
8507 helps machines that have slow floating point or memory load instructions
8508 by allowing other instructions to be issued until the result of the load
8509 or floating-point instruction is required.
8511 Enabled at levels @option{-O2}, @option{-O3}.
8513 @item -fschedule-insns2
8514 @opindex fschedule-insns2
8515 Similar to @option{-fschedule-insns}, but requests an additional pass of
8516 instruction scheduling after register allocation has been done. This is
8517 especially useful on machines with a relatively small number of
8518 registers and where memory load instructions take more than one cycle.
8520 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8522 @item -fno-sched-interblock
8523 @opindex fno-sched-interblock
8524 Don't schedule instructions across basic blocks. This is normally
8525 enabled by default when scheduling before register allocation, i.e.@:
8526 with @option{-fschedule-insns} or at @option{-O2} or higher.
8528 @item -fno-sched-spec
8529 @opindex fno-sched-spec
8530 Don't allow speculative motion of non-load instructions. This is normally
8531 enabled by default when scheduling before register allocation, i.e.@:
8532 with @option{-fschedule-insns} or at @option{-O2} or higher.
8534 @item -fsched-pressure
8535 @opindex fsched-pressure
8536 Enable register pressure sensitive insn scheduling before register
8537 allocation. This only makes sense when scheduling before register
8538 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
8539 @option{-O2} or higher. Usage of this option can improve the
8540 generated code and decrease its size by preventing register pressure
8541 increase above the number of available hard registers and subsequent
8542 spills in register allocation.
8544 @item -fsched-spec-load
8545 @opindex fsched-spec-load
8546 Allow speculative motion of some load instructions. This only makes
8547 sense when scheduling before register allocation, i.e.@: with
8548 @option{-fschedule-insns} or at @option{-O2} or higher.
8550 @item -fsched-spec-load-dangerous
8551 @opindex fsched-spec-load-dangerous
8552 Allow speculative motion of more load instructions. This only makes
8553 sense when scheduling before register allocation, i.e.@: with
8554 @option{-fschedule-insns} or at @option{-O2} or higher.
8556 @item -fsched-stalled-insns
8557 @itemx -fsched-stalled-insns=@var{n}
8558 @opindex fsched-stalled-insns
8559 Define how many insns (if any) can be moved prematurely from the queue
8560 of stalled insns into the ready list during the second scheduling pass.
8561 @option{-fno-sched-stalled-insns} means that no insns are moved
8562 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
8563 on how many queued insns can be moved prematurely.
8564 @option{-fsched-stalled-insns} without a value is equivalent to
8565 @option{-fsched-stalled-insns=1}.
8567 @item -fsched-stalled-insns-dep
8568 @itemx -fsched-stalled-insns-dep=@var{n}
8569 @opindex fsched-stalled-insns-dep
8570 Define how many insn groups (cycles) are examined for a dependency
8571 on a stalled insn that is a candidate for premature removal from the queue
8572 of stalled insns. This has an effect only during the second scheduling pass,
8573 and only if @option{-fsched-stalled-insns} is used.
8574 @option{-fno-sched-stalled-insns-dep} is equivalent to
8575 @option{-fsched-stalled-insns-dep=0}.
8576 @option{-fsched-stalled-insns-dep} without a value is equivalent to
8577 @option{-fsched-stalled-insns-dep=1}.
8579 @item -fsched2-use-superblocks
8580 @opindex fsched2-use-superblocks
8581 When scheduling after register allocation, use superblock scheduling.
8582 This allows motion across basic block boundaries,
8583 resulting in faster schedules. This option is experimental, as not all machine
8584 descriptions used by GCC model the CPU closely enough to avoid unreliable
8585 results from the algorithm.
8587 This only makes sense when scheduling after register allocation, i.e.@: with
8588 @option{-fschedule-insns2} or at @option{-O2} or higher.
8590 @item -fsched-group-heuristic
8591 @opindex fsched-group-heuristic
8592 Enable the group heuristic in the scheduler. This heuristic favors
8593 the instruction that belongs to a schedule group. This is enabled
8594 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8595 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8597 @item -fsched-critical-path-heuristic
8598 @opindex fsched-critical-path-heuristic
8599 Enable the critical-path heuristic in the scheduler. This heuristic favors
8600 instructions on the critical path. This is enabled by default when
8601 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
8602 or @option{-fschedule-insns2} or at @option{-O2} or higher.
8604 @item -fsched-spec-insn-heuristic
8605 @opindex fsched-spec-insn-heuristic
8606 Enable the speculative instruction heuristic in the scheduler. This
8607 heuristic favors speculative instructions with greater dependency weakness.
8608 This is enabled by default when scheduling is enabled, i.e.@:
8609 with @option{-fschedule-insns} or @option{-fschedule-insns2}
8610 or at @option{-O2} or higher.
8612 @item -fsched-rank-heuristic
8613 @opindex fsched-rank-heuristic
8614 Enable the rank heuristic in the scheduler. This heuristic favors
8615 the instruction belonging to a basic block with greater size or frequency.
8616 This is enabled by default when scheduling is enabled, i.e.@:
8617 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8618 at @option{-O2} or higher.
8620 @item -fsched-last-insn-heuristic
8621 @opindex fsched-last-insn-heuristic
8622 Enable the last-instruction heuristic in the scheduler. This heuristic
8623 favors the instruction that is less dependent on the last instruction
8624 scheduled. This is enabled by default when scheduling is enabled,
8625 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8626 at @option{-O2} or higher.
8628 @item -fsched-dep-count-heuristic
8629 @opindex fsched-dep-count-heuristic
8630 Enable the dependent-count heuristic in the scheduler. This heuristic
8631 favors the instruction that has more instructions depending on it.
8632 This is enabled by default when scheduling is enabled, i.e.@:
8633 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
8634 at @option{-O2} or higher.
8636 @item -freschedule-modulo-scheduled-loops
8637 @opindex freschedule-modulo-scheduled-loops
8638 Modulo scheduling is performed before traditional scheduling. If a loop
8639 is modulo scheduled, later scheduling passes may change its schedule.
8640 Use this option to control that behavior.
8642 @item -fselective-scheduling
8643 @opindex fselective-scheduling
8644 Schedule instructions using selective scheduling algorithm. Selective
8645 scheduling runs instead of the first scheduler pass.
8647 @item -fselective-scheduling2
8648 @opindex fselective-scheduling2
8649 Schedule instructions using selective scheduling algorithm. Selective
8650 scheduling runs instead of the second scheduler pass.
8652 @item -fsel-sched-pipelining
8653 @opindex fsel-sched-pipelining
8654 Enable software pipelining of innermost loops during selective scheduling.
8655 This option has no effect unless one of @option{-fselective-scheduling} or
8656 @option{-fselective-scheduling2} is turned on.
8658 @item -fsel-sched-pipelining-outer-loops
8659 @opindex fsel-sched-pipelining-outer-loops
8660 When pipelining loops during selective scheduling, also pipeline outer loops.
8661 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
8663 @item -fsemantic-interposition
8664 @opindex fsemantic-interposition
8665 Some object formats, like ELF, allow interposing of symbols by the
8667 This means that for symbols exported from the DSO, the compiler cannot perform
8668 interprocedural propagation, inlining and other optimizations in anticipation
8669 that the function or variable in question may change. While this feature is
8670 useful, for example, to rewrite memory allocation functions by a debugging
8671 implementation, it is expensive in the terms of code quality.
8672 With @option{-fno-semantic-interposition} the compiler assumes that
8673 if interposition happens for functions the overwriting function will have
8674 precisely the same semantics (and side effects).
8675 Similarly if interposition happens
8676 for variables, the constructor of the variable will be the same. The flag
8677 has no effect for functions explicitly declared inline
8678 (where it is never allowed for interposition to change semantics)
8679 and for symbols explicitly declared weak.
8682 @opindex fshrink-wrap
8683 Emit function prologues only before parts of the function that need it,
8684 rather than at the top of the function. This flag is enabled by default at
8685 @option{-O} and higher.
8687 @item -fcaller-saves
8688 @opindex fcaller-saves
8689 Enable allocation of values to registers that are clobbered by
8690 function calls, by emitting extra instructions to save and restore the
8691 registers around such calls. Such allocation is done only when it
8692 seems to result in better code.
8694 This option is always enabled by default on certain machines, usually
8695 those which have no call-preserved registers to use instead.
8697 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8699 @item -fcombine-stack-adjustments
8700 @opindex fcombine-stack-adjustments
8701 Tracks stack adjustments (pushes and pops) and stack memory references
8702 and then tries to find ways to combine them.
8704 Enabled by default at @option{-O1} and higher.
8708 Use caller save registers for allocation if those registers are not used by
8709 any called function. In that case it is not necessary to save and restore
8710 them around calls. This is only possible if called functions are part of
8711 same compilation unit as current function and they are compiled before it.
8713 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8715 @item -fconserve-stack
8716 @opindex fconserve-stack
8717 Attempt to minimize stack usage. The compiler attempts to use less
8718 stack space, even if that makes the program slower. This option
8719 implies setting the @option{large-stack-frame} parameter to 100
8720 and the @option{large-stack-frame-growth} parameter to 400.
8722 @item -ftree-reassoc
8723 @opindex ftree-reassoc
8724 Perform reassociation on trees. This flag is enabled by default
8725 at @option{-O} and higher.
8729 Perform partial redundancy elimination (PRE) on trees. This flag is
8730 enabled by default at @option{-O2} and @option{-O3}.
8732 @item -ftree-partial-pre
8733 @opindex ftree-partial-pre
8734 Make partial redundancy elimination (PRE) more aggressive. This flag is
8735 enabled by default at @option{-O3}.
8737 @item -ftree-forwprop
8738 @opindex ftree-forwprop
8739 Perform forward propagation on trees. This flag is enabled by default
8740 at @option{-O} and higher.
8744 Perform full redundancy elimination (FRE) on trees. The difference
8745 between FRE and PRE is that FRE only considers expressions
8746 that are computed on all paths leading to the redundant computation.
8747 This analysis is faster than PRE, though it exposes fewer redundancies.
8748 This flag is enabled by default at @option{-O} and higher.
8750 @item -ftree-phiprop
8751 @opindex ftree-phiprop
8752 Perform hoisting of loads from conditional pointers on trees. This
8753 pass is enabled by default at @option{-O} and higher.
8755 @item -fhoist-adjacent-loads
8756 @opindex fhoist-adjacent-loads
8757 Speculatively hoist loads from both branches of an if-then-else if the
8758 loads are from adjacent locations in the same structure and the target
8759 architecture has a conditional move instruction. This flag is enabled
8760 by default at @option{-O2} and higher.
8762 @item -ftree-copy-prop
8763 @opindex ftree-copy-prop
8764 Perform copy propagation on trees. This pass eliminates unnecessary
8765 copy operations. This flag is enabled by default at @option{-O} and
8768 @item -fipa-pure-const
8769 @opindex fipa-pure-const
8770 Discover which functions are pure or constant.
8771 Enabled by default at @option{-O} and higher.
8773 @item -fipa-reference
8774 @opindex fipa-reference
8775 Discover which static variables do not escape the
8777 Enabled by default at @option{-O} and higher.
8781 Perform interprocedural pointer analysis and interprocedural modification
8782 and reference analysis. This option can cause excessive memory and
8783 compile-time usage on large compilation units. It is not enabled by
8784 default at any optimization level.
8787 @opindex fipa-profile
8788 Perform interprocedural profile propagation. The functions called only from
8789 cold functions are marked as cold. Also functions executed once (such as
8790 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
8791 functions and loop less parts of functions executed once are then optimized for
8793 Enabled by default at @option{-O} and higher.
8797 Perform interprocedural constant propagation.
8798 This optimization analyzes the program to determine when values passed
8799 to functions are constants and then optimizes accordingly.
8800 This optimization can substantially increase performance
8801 if the application has constants passed to functions.
8802 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
8804 @item -fipa-cp-clone
8805 @opindex fipa-cp-clone
8806 Perform function cloning to make interprocedural constant propagation stronger.
8807 When enabled, interprocedural constant propagation performs function cloning
8808 when externally visible function can be called with constant arguments.
8809 Because this optimization can create multiple copies of functions,
8810 it may significantly increase code size
8811 (see @option{--param ipcp-unit-growth=@var{value}}).
8812 This flag is enabled by default at @option{-O3}.
8814 @item -fipa-cp-alignment
8815 @opindex -fipa-cp-alignment
8816 When enabled, this optimization propagates alignment of function
8817 parameters to support better vectorization and string operations.
8819 This flag is enabled by default at @option{-O2} and @option{-Os}. It
8820 requires that @option{-fipa-cp} is enabled.
8824 Perform Identical Code Folding for functions and read-only variables.
8825 The optimization reduces code size and may disturb unwind stacks by replacing
8826 a function by equivalent one with a different name. The optimization works
8827 more effectively with link time optimization enabled.
8829 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
8830 works on different levels and thus the optimizations are not same - there are
8831 equivalences that are found only by GCC and equivalences found only by Gold.
8833 This flag is enabled by default at @option{-O2} and @option{-Os}.
8835 @item -fisolate-erroneous-paths-dereference
8836 @opindex fisolate-erroneous-paths-dereference
8837 Detect paths that trigger erroneous or undefined behavior due to
8838 dereferencing a null pointer. Isolate those paths from the main control
8839 flow and turn the statement with erroneous or undefined behavior into a trap.
8840 This flag is enabled by default at @option{-O2} and higher and depends on
8841 @option{-fdelete-null-pointer-checks} also being enabled.
8843 @item -fisolate-erroneous-paths-attribute
8844 @opindex fisolate-erroneous-paths-attribute
8845 Detect paths that trigger erroneous or undefined behavior due a null value
8846 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
8847 attribute. Isolate those paths from the main control flow and turn the
8848 statement with erroneous or undefined behavior into a trap. This is not
8849 currently enabled, but may be enabled by @option{-O2} in the future.
8853 Perform forward store motion on trees. This flag is
8854 enabled by default at @option{-O} and higher.
8856 @item -ftree-bit-ccp
8857 @opindex ftree-bit-ccp
8858 Perform sparse conditional bit constant propagation on trees and propagate
8859 pointer alignment information.
8860 This pass only operates on local scalar variables and is enabled by default
8861 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
8865 Perform sparse conditional constant propagation (CCP) on trees. This
8866 pass only operates on local scalar variables and is enabled by default
8867 at @option{-O} and higher.
8869 @item -fssa-backprop
8870 @opindex fssa-backprop
8871 Propagate information about uses of a value up the definition chain
8872 in order to simplify the definitions. For example, this pass strips
8873 sign operations if the sign of a value never matters. The flag is
8874 enabled by default at @option{-O} and higher.
8877 @opindex fssa-phiopt
8878 Perform pattern matching on SSA PHI nodes to optimize conditional
8879 code. This pass is enabled by default at @option{-O} and higher.
8881 @item -ftree-switch-conversion
8882 @opindex ftree-switch-conversion
8883 Perform conversion of simple initializations in a switch to
8884 initializations from a scalar array. This flag is enabled by default
8885 at @option{-O2} and higher.
8887 @item -ftree-tail-merge
8888 @opindex ftree-tail-merge
8889 Look for identical code sequences. When found, replace one with a jump to the
8890 other. This optimization is known as tail merging or cross jumping. This flag
8891 is enabled by default at @option{-O2} and higher. The compilation time
8893 be limited using @option{max-tail-merge-comparisons} parameter and
8894 @option{max-tail-merge-iterations} parameter.
8898 Perform dead code elimination (DCE) on trees. This flag is enabled by
8899 default at @option{-O} and higher.
8901 @item -ftree-builtin-call-dce
8902 @opindex ftree-builtin-call-dce
8903 Perform conditional dead code elimination (DCE) for calls to built-in functions
8904 that may set @code{errno} but are otherwise side-effect free. This flag is
8905 enabled by default at @option{-O2} and higher if @option{-Os} is not also
8908 @item -ftree-dominator-opts
8909 @opindex ftree-dominator-opts
8910 Perform a variety of simple scalar cleanups (constant/copy
8911 propagation, redundancy elimination, range propagation and expression
8912 simplification) based on a dominator tree traversal. This also
8913 performs jump threading (to reduce jumps to jumps). This flag is
8914 enabled by default at @option{-O} and higher.
8918 Perform dead store elimination (DSE) on trees. A dead store is a store into
8919 a memory location that is later overwritten by another store without
8920 any intervening loads. In this case the earlier store can be deleted. This
8921 flag is enabled by default at @option{-O} and higher.
8925 Perform loop header copying on trees. This is beneficial since it increases
8926 effectiveness of code motion optimizations. It also saves one jump. This flag
8927 is enabled by default at @option{-O} and higher. It is not enabled
8928 for @option{-Os}, since it usually increases code size.
8930 @item -ftree-loop-optimize
8931 @opindex ftree-loop-optimize
8932 Perform loop optimizations on trees. This flag is enabled by default
8933 at @option{-O} and higher.
8935 @item -ftree-loop-linear
8936 @itemx -floop-interchange
8937 @itemx -floop-strip-mine
8939 @itemx -floop-unroll-and-jam
8940 @opindex ftree-loop-linear
8941 @opindex floop-interchange
8942 @opindex floop-strip-mine
8943 @opindex floop-block
8944 @opindex floop-unroll-and-jam
8945 Perform loop nest optimizations. Same as
8946 @option{-floop-nest-optimize}. To use this code transformation, GCC has
8947 to be configured with @option{--with-isl} to enable the Graphite loop
8948 transformation infrastructure.
8950 @item -fgraphite-identity
8951 @opindex fgraphite-identity
8952 Enable the identity transformation for graphite. For every SCoP we generate
8953 the polyhedral representation and transform it back to gimple. Using
8954 @option{-fgraphite-identity} we can check the costs or benefits of the
8955 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
8956 are also performed by the code generator isl, like index splitting and
8957 dead code elimination in loops.
8959 @item -floop-nest-optimize
8960 @opindex floop-nest-optimize
8961 Enable the isl based loop nest optimizer. This is a generic loop nest
8962 optimizer based on the Pluto optimization algorithms. It calculates a loop
8963 structure optimized for data-locality and parallelism. This option
8966 @item -floop-parallelize-all
8967 @opindex floop-parallelize-all
8968 Use the Graphite data dependence analysis to identify loops that can
8969 be parallelized. Parallelize all the loops that can be analyzed to
8970 not contain loop carried dependences without checking that it is
8971 profitable to parallelize the loops.
8973 @item -ftree-coalesce-vars
8974 @opindex ftree-coalesce-vars
8975 While transforming the program out of the SSA representation, attempt to
8976 reduce copying by coalescing versions of different user-defined
8977 variables, instead of just compiler temporaries. This may severely
8978 limit the ability to debug an optimized program compiled with
8979 @option{-fno-var-tracking-assignments}. In the negated form, this flag
8980 prevents SSA coalescing of user variables. This option is enabled by
8981 default if optimization is enabled, and it does very little otherwise.
8983 @item -ftree-loop-if-convert
8984 @opindex ftree-loop-if-convert
8985 Attempt to transform conditional jumps in the innermost loops to
8986 branch-less equivalents. The intent is to remove control-flow from
8987 the innermost loops in order to improve the ability of the
8988 vectorization pass to handle these loops. This is enabled by default
8989 if vectorization is enabled.
8991 @item -ftree-loop-if-convert-stores
8992 @opindex ftree-loop-if-convert-stores
8993 Attempt to also if-convert conditional jumps containing memory writes.
8994 This transformation can be unsafe for multi-threaded programs as it
8995 transforms conditional memory writes into unconditional memory writes.
8998 for (i = 0; i < N; i++)
9004 for (i = 0; i < N; i++)
9005 A[i] = cond ? expr : A[i];
9007 potentially producing data races.
9009 @item -ftree-loop-distribution
9010 @opindex ftree-loop-distribution
9011 Perform loop distribution. This flag can improve cache performance on
9012 big loop bodies and allow further loop optimizations, like
9013 parallelization or vectorization, to take place. For example, the loop
9030 @item -ftree-loop-distribute-patterns
9031 @opindex ftree-loop-distribute-patterns
9032 Perform loop distribution of patterns that can be code generated with
9033 calls to a library. This flag is enabled by default at @option{-O3}.
9035 This pass distributes the initialization loops and generates a call to
9036 memset zero. For example, the loop
9052 and the initialization loop is transformed into a call to memset zero.
9054 @item -ftree-loop-im
9055 @opindex ftree-loop-im
9056 Perform loop invariant motion on trees. This pass moves only invariants that
9057 are hard to handle at RTL level (function calls, operations that expand to
9058 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
9059 operands of conditions that are invariant out of the loop, so that we can use
9060 just trivial invariantness analysis in loop unswitching. The pass also includes
9063 @item -ftree-loop-ivcanon
9064 @opindex ftree-loop-ivcanon
9065 Create a canonical counter for number of iterations in loops for which
9066 determining number of iterations requires complicated analysis. Later
9067 optimizations then may determine the number easily. Useful especially
9068 in connection with unrolling.
9072 Perform induction variable optimizations (strength reduction, induction
9073 variable merging and induction variable elimination) on trees.
9075 @item -ftree-parallelize-loops=n
9076 @opindex ftree-parallelize-loops
9077 Parallelize loops, i.e., split their iteration space to run in n threads.
9078 This is only possible for loops whose iterations are independent
9079 and can be arbitrarily reordered. The optimization is only
9080 profitable on multiprocessor machines, for loops that are CPU-intensive,
9081 rather than constrained e.g.@: by memory bandwidth. This option
9082 implies @option{-pthread}, and thus is only supported on targets
9083 that have support for @option{-pthread}.
9087 Perform function-local points-to analysis on trees. This flag is
9088 enabled by default at @option{-O} and higher.
9092 Perform scalar replacement of aggregates. This pass replaces structure
9093 references with scalars to prevent committing structures to memory too
9094 early. This flag is enabled by default at @option{-O} and higher.
9098 Perform temporary expression replacement during the SSA->normal phase. Single
9099 use/single def temporaries are replaced at their use location with their
9100 defining expression. This results in non-GIMPLE code, but gives the expanders
9101 much more complex trees to work on resulting in better RTL generation. This is
9102 enabled by default at @option{-O} and higher.
9106 Perform straight-line strength reduction on trees. This recognizes related
9107 expressions involving multiplications and replaces them by less expensive
9108 calculations when possible. This is enabled by default at @option{-O} and
9111 @item -ftree-vectorize
9112 @opindex ftree-vectorize
9113 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
9114 and @option{-ftree-slp-vectorize} if not explicitly specified.
9116 @item -ftree-loop-vectorize
9117 @opindex ftree-loop-vectorize
9118 Perform loop vectorization on trees. This flag is enabled by default at
9119 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9121 @item -ftree-slp-vectorize
9122 @opindex ftree-slp-vectorize
9123 Perform basic block vectorization on trees. This flag is enabled by default at
9124 @option{-O3} and when @option{-ftree-vectorize} is enabled.
9126 @item -fvect-cost-model=@var{model}
9127 @opindex fvect-cost-model
9128 Alter the cost model used for vectorization. The @var{model} argument
9129 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
9130 With the @samp{unlimited} model the vectorized code-path is assumed
9131 to be profitable while with the @samp{dynamic} model a runtime check
9132 guards the vectorized code-path to enable it only for iteration
9133 counts that will likely execute faster than when executing the original
9134 scalar loop. The @samp{cheap} model disables vectorization of
9135 loops where doing so would be cost prohibitive for example due to
9136 required runtime checks for data dependence or alignment but otherwise
9137 is equal to the @samp{dynamic} model.
9138 The default cost model depends on other optimization flags and is
9139 either @samp{dynamic} or @samp{cheap}.
9141 @item -fsimd-cost-model=@var{model}
9142 @opindex fsimd-cost-model
9143 Alter the cost model used for vectorization of loops marked with the OpenMP
9144 or Cilk Plus simd directive. The @var{model} argument should be one of
9145 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
9146 have the same meaning as described in @option{-fvect-cost-model} and by
9147 default a cost model defined with @option{-fvect-cost-model} is used.
9151 Perform Value Range Propagation on trees. This is similar to the
9152 constant propagation pass, but instead of values, ranges of values are
9153 propagated. This allows the optimizers to remove unnecessary range
9154 checks like array bound checks and null pointer checks. This is
9155 enabled by default at @option{-O2} and higher. Null pointer check
9156 elimination is only done if @option{-fdelete-null-pointer-checks} is
9160 @opindex fsplit-paths
9161 Split paths leading to loop backedges. This can improve dead code
9162 elimination and common subexpression elimination. This is enabled by
9163 default at @option{-O2} and above.
9165 @item -fsplit-ivs-in-unroller
9166 @opindex fsplit-ivs-in-unroller
9167 Enables expression of values of induction variables in later iterations
9168 of the unrolled loop using the value in the first iteration. This breaks
9169 long dependency chains, thus improving efficiency of the scheduling passes.
9171 A combination of @option{-fweb} and CSE is often sufficient to obtain the
9172 same effect. However, that is not reliable in cases where the loop body
9173 is more complicated than a single basic block. It also does not work at all
9174 on some architectures due to restrictions in the CSE pass.
9176 This optimization is enabled by default.
9178 @item -fvariable-expansion-in-unroller
9179 @opindex fvariable-expansion-in-unroller
9180 With this option, the compiler creates multiple copies of some
9181 local variables when unrolling a loop, which can result in superior code.
9183 @item -fpartial-inlining
9184 @opindex fpartial-inlining
9185 Inline parts of functions. This option has any effect only
9186 when inlining itself is turned on by the @option{-finline-functions}
9187 or @option{-finline-small-functions} options.
9189 Enabled at level @option{-O2}.
9191 @item -fpredictive-commoning
9192 @opindex fpredictive-commoning
9193 Perform predictive commoning optimization, i.e., reusing computations
9194 (especially memory loads and stores) performed in previous
9195 iterations of loops.
9197 This option is enabled at level @option{-O3}.
9199 @item -fprefetch-loop-arrays
9200 @opindex fprefetch-loop-arrays
9201 If supported by the target machine, generate instructions to prefetch
9202 memory to improve the performance of loops that access large arrays.
9204 This option may generate better or worse code; results are highly
9205 dependent on the structure of loops within the source code.
9207 Disabled at level @option{-Os}.
9210 @itemx -fno-peephole2
9211 @opindex fno-peephole
9212 @opindex fno-peephole2
9213 Disable any machine-specific peephole optimizations. The difference
9214 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
9215 are implemented in the compiler; some targets use one, some use the
9216 other, a few use both.
9218 @option{-fpeephole} is enabled by default.
9219 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9221 @item -fno-guess-branch-probability
9222 @opindex fno-guess-branch-probability
9223 Do not guess branch probabilities using heuristics.
9225 GCC uses heuristics to guess branch probabilities if they are
9226 not provided by profiling feedback (@option{-fprofile-arcs}). These
9227 heuristics are based on the control flow graph. If some branch probabilities
9228 are specified by @code{__builtin_expect}, then the heuristics are
9229 used to guess branch probabilities for the rest of the control flow graph,
9230 taking the @code{__builtin_expect} info into account. The interactions
9231 between the heuristics and @code{__builtin_expect} can be complex, and in
9232 some cases, it may be useful to disable the heuristics so that the effects
9233 of @code{__builtin_expect} are easier to understand.
9235 The default is @option{-fguess-branch-probability} at levels
9236 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9238 @item -freorder-blocks
9239 @opindex freorder-blocks
9240 Reorder basic blocks in the compiled function in order to reduce number of
9241 taken branches and improve code locality.
9243 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9245 @item -freorder-blocks-algorithm=@var{algorithm}
9246 @opindex freorder-blocks-algorithm
9247 Use the specified algorithm for basic block reordering. The
9248 @var{algorithm} argument can be @samp{simple}, which does not increase
9249 code size (except sometimes due to secondary effects like alignment),
9250 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
9251 put all often executed code together, minimizing the number of branches
9252 executed by making extra copies of code.
9254 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
9255 @samp{stc} at levels @option{-O2}, @option{-O3}.
9257 @item -freorder-blocks-and-partition
9258 @opindex freorder-blocks-and-partition
9259 In addition to reordering basic blocks in the compiled function, in order
9260 to reduce number of taken branches, partitions hot and cold basic blocks
9261 into separate sections of the assembly and .o files, to improve
9262 paging and cache locality performance.
9264 This optimization is automatically turned off in the presence of
9265 exception handling, for linkonce sections, for functions with a user-defined
9266 section attribute and on any architecture that does not support named
9269 Enabled for x86 at levels @option{-O2}, @option{-O3}.
9271 @item -freorder-functions
9272 @opindex freorder-functions
9273 Reorder functions in the object file in order to
9274 improve code locality. This is implemented by using special
9275 subsections @code{.text.hot} for most frequently executed functions and
9276 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
9277 the linker so object file format must support named sections and linker must
9278 place them in a reasonable way.
9280 Also profile feedback must be available to make this option effective. See
9281 @option{-fprofile-arcs} for details.
9283 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
9285 @item -fstrict-aliasing
9286 @opindex fstrict-aliasing
9287 Allow the compiler to assume the strictest aliasing rules applicable to
9288 the language being compiled. For C (and C++), this activates
9289 optimizations based on the type of expressions. In particular, an
9290 object of one type is assumed never to reside at the same address as an
9291 object of a different type, unless the types are almost the same. For
9292 example, an @code{unsigned int} can alias an @code{int}, but not a
9293 @code{void*} or a @code{double}. A character type may alias any other
9296 @anchor{Type-punning}Pay special attention to code like this:
9309 The practice of reading from a different union member than the one most
9310 recently written to (called ``type-punning'') is common. Even with
9311 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
9312 is accessed through the union type. So, the code above works as
9313 expected. @xref{Structures unions enumerations and bit-fields
9314 implementation}. However, this code might not:
9325 Similarly, access by taking the address, casting the resulting pointer
9326 and dereferencing the result has undefined behavior, even if the cast
9327 uses a union type, e.g.:
9331 return ((union a_union *) &d)->i;
9335 The @option{-fstrict-aliasing} option is enabled at levels
9336 @option{-O2}, @option{-O3}, @option{-Os}.
9338 @item -fstrict-overflow
9339 @opindex fstrict-overflow
9340 Allow the compiler to assume strict signed overflow rules, depending
9341 on the language being compiled. For C (and C++) this means that
9342 overflow when doing arithmetic with signed numbers is undefined, which
9343 means that the compiler may assume that it does not happen. This
9344 permits various optimizations. For example, the compiler assumes
9345 that an expression like @code{i + 10 > i} is always true for
9346 signed @code{i}. This assumption is only valid if signed overflow is
9347 undefined, as the expression is false if @code{i + 10} overflows when
9348 using twos complement arithmetic. When this option is in effect any
9349 attempt to determine whether an operation on signed numbers
9350 overflows must be written carefully to not actually involve overflow.
9352 This option also allows the compiler to assume strict pointer
9353 semantics: given a pointer to an object, if adding an offset to that
9354 pointer does not produce a pointer to the same object, the addition is
9355 undefined. This permits the compiler to conclude that @code{p + u >
9356 p} is always true for a pointer @code{p} and unsigned integer
9357 @code{u}. This assumption is only valid because pointer wraparound is
9358 undefined, as the expression is false if @code{p + u} overflows using
9359 twos complement arithmetic.
9361 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
9362 that integer signed overflow is fully defined: it wraps. When
9363 @option{-fwrapv} is used, there is no difference between
9364 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
9365 integers. With @option{-fwrapv} certain types of overflow are
9366 permitted. For example, if the compiler gets an overflow when doing
9367 arithmetic on constants, the overflowed value can still be used with
9368 @option{-fwrapv}, but not otherwise.
9370 The @option{-fstrict-overflow} option is enabled at levels
9371 @option{-O2}, @option{-O3}, @option{-Os}.
9373 @item -falign-functions
9374 @itemx -falign-functions=@var{n}
9375 @opindex falign-functions
9376 Align the start of functions to the next power-of-two greater than
9377 @var{n}, skipping up to @var{n} bytes. For instance,
9378 @option{-falign-functions=32} aligns functions to the next 32-byte
9379 boundary, but @option{-falign-functions=24} aligns to the next
9380 32-byte boundary only if this can be done by skipping 23 bytes or less.
9382 @option{-fno-align-functions} and @option{-falign-functions=1} are
9383 equivalent and mean that functions are not aligned.
9385 Some assemblers only support this flag when @var{n} is a power of two;
9386 in that case, it is rounded up.
9388 If @var{n} is not specified or is zero, use a machine-dependent default.
9390 Enabled at levels @option{-O2}, @option{-O3}.
9392 @item -falign-labels
9393 @itemx -falign-labels=@var{n}
9394 @opindex falign-labels
9395 Align all branch targets to a power-of-two boundary, skipping up to
9396 @var{n} bytes like @option{-falign-functions}. This option can easily
9397 make code slower, because it must insert dummy operations for when the
9398 branch target is reached in the usual flow of the code.
9400 @option{-fno-align-labels} and @option{-falign-labels=1} are
9401 equivalent and mean that labels are not aligned.
9403 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
9404 are greater than this value, then their values are used instead.
9406 If @var{n} is not specified or is zero, use a machine-dependent default
9407 which is very likely to be @samp{1}, meaning no alignment.
9409 Enabled at levels @option{-O2}, @option{-O3}.
9412 @itemx -falign-loops=@var{n}
9413 @opindex falign-loops
9414 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
9415 like @option{-falign-functions}. If the loops are
9416 executed many times, this makes up for any execution of the dummy
9419 @option{-fno-align-loops} and @option{-falign-loops=1} are
9420 equivalent and mean that loops are not aligned.
9422 If @var{n} is not specified or is zero, use a machine-dependent default.
9424 Enabled at levels @option{-O2}, @option{-O3}.
9427 @itemx -falign-jumps=@var{n}
9428 @opindex falign-jumps
9429 Align branch targets to a power-of-two boundary, for branch targets
9430 where the targets can only be reached by jumping, skipping up to @var{n}
9431 bytes like @option{-falign-functions}. In this case, no dummy operations
9434 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
9435 equivalent and mean that loops are not aligned.
9437 If @var{n} is not specified or is zero, use a machine-dependent default.
9439 Enabled at levels @option{-O2}, @option{-O3}.
9441 @item -funit-at-a-time
9442 @opindex funit-at-a-time
9443 This option is left for compatibility reasons. @option{-funit-at-a-time}
9444 has no effect, while @option{-fno-unit-at-a-time} implies
9445 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
9449 @item -fno-toplevel-reorder
9450 @opindex fno-toplevel-reorder
9451 Do not reorder top-level functions, variables, and @code{asm}
9452 statements. Output them in the same order that they appear in the
9453 input file. When this option is used, unreferenced static variables
9454 are not removed. This option is intended to support existing code
9455 that relies on a particular ordering. For new code, it is better to
9456 use attributes when possible.
9458 Enabled at level @option{-O0}. When disabled explicitly, it also implies
9459 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
9464 Constructs webs as commonly used for register allocation purposes and assign
9465 each web individual pseudo register. This allows the register allocation pass
9466 to operate on pseudos directly, but also strengthens several other optimization
9467 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
9468 however, make debugging impossible, since variables no longer stay in a
9471 Enabled by default with @option{-funroll-loops}.
9473 @item -fwhole-program
9474 @opindex fwhole-program
9475 Assume that the current compilation unit represents the whole program being
9476 compiled. All public functions and variables with the exception of @code{main}
9477 and those merged by attribute @code{externally_visible} become static functions
9478 and in effect are optimized more aggressively by interprocedural optimizers.
9480 This option should not be used in combination with @option{-flto}.
9481 Instead relying on a linker plugin should provide safer and more precise
9484 @item -flto[=@var{n}]
9486 This option runs the standard link-time optimizer. When invoked
9487 with source code, it generates GIMPLE (one of GCC's internal
9488 representations) and writes it to special ELF sections in the object
9489 file. When the object files are linked together, all the function
9490 bodies are read from these ELF sections and instantiated as if they
9491 had been part of the same translation unit.
9493 To use the link-time optimizer, @option{-flto} and optimization
9494 options should be specified at compile time and during the final link.
9498 gcc -c -O2 -flto foo.c
9499 gcc -c -O2 -flto bar.c
9500 gcc -o myprog -flto -O2 foo.o bar.o
9503 The first two invocations to GCC save a bytecode representation
9504 of GIMPLE into special ELF sections inside @file{foo.o} and
9505 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
9506 @file{foo.o} and @file{bar.o}, merges the two files into a single
9507 internal image, and compiles the result as usual. Since both
9508 @file{foo.o} and @file{bar.o} are merged into a single image, this
9509 causes all the interprocedural analyses and optimizations in GCC to
9510 work across the two files as if they were a single one. This means,
9511 for example, that the inliner is able to inline functions in
9512 @file{bar.o} into functions in @file{foo.o} and vice-versa.
9514 Another (simpler) way to enable link-time optimization is:
9517 gcc -o myprog -flto -O2 foo.c bar.c
9520 The above generates bytecode for @file{foo.c} and @file{bar.c},
9521 merges them together into a single GIMPLE representation and optimizes
9522 them as usual to produce @file{myprog}.
9524 The only important thing to keep in mind is that to enable link-time
9525 optimizations you need to use the GCC driver to perform the link-step.
9526 GCC then automatically performs link-time optimization if any of the
9527 objects involved were compiled with the @option{-flto} command-line option.
9529 should specify the optimization options to be used for link-time
9530 optimization though GCC tries to be clever at guessing an
9531 optimization level to use from the options used at compile-time
9532 if you fail to specify one at link-time. You can always override
9533 the automatic decision to do link-time optimization at link-time
9534 by passing @option{-fno-lto} to the link command.
9536 To make whole program optimization effective, it is necessary to make
9537 certain whole program assumptions. The compiler needs to know
9538 what functions and variables can be accessed by libraries and runtime
9539 outside of the link-time optimized unit. When supported by the linker,
9540 the linker plugin (see @option{-fuse-linker-plugin}) passes information
9541 to the compiler about used and externally visible symbols. When
9542 the linker plugin is not available, @option{-fwhole-program} should be
9543 used to allow the compiler to make these assumptions, which leads
9544 to more aggressive optimization decisions.
9546 When @option{-fuse-linker-plugin} is not enabled then, when a file is
9547 compiled with @option{-flto}, the generated object file is larger than
9548 a regular object file because it contains GIMPLE bytecodes and the usual
9549 final code (see @option{-ffat-lto-objects}. This means that
9550 object files with LTO information can be linked as normal object
9551 files; if @option{-fno-lto} is passed to the linker, no
9552 interprocedural optimizations are applied. Note that when
9553 @option{-fno-fat-lto-objects} is enabled the compile-stage is faster
9554 but you cannot perform a regular, non-LTO link on them.
9556 Additionally, the optimization flags used to compile individual files
9557 are not necessarily related to those used at link time. For instance,
9560 gcc -c -O0 -ffat-lto-objects -flto foo.c
9561 gcc -c -O0 -ffat-lto-objects -flto bar.c
9562 gcc -o myprog -O3 foo.o bar.o
9565 This produces individual object files with unoptimized assembler
9566 code, but the resulting binary @file{myprog} is optimized at
9567 @option{-O3}. If, instead, the final binary is generated with
9568 @option{-fno-lto}, then @file{myprog} is not optimized.
9570 When producing the final binary, GCC only
9571 applies link-time optimizations to those files that contain bytecode.
9572 Therefore, you can mix and match object files and libraries with
9573 GIMPLE bytecodes and final object code. GCC automatically selects
9574 which files to optimize in LTO mode and which files to link without
9577 There are some code generation flags preserved by GCC when
9578 generating bytecodes, as they need to be used during the final link
9579 stage. Generally options specified at link-time override those
9580 specified at compile-time.
9582 If you do not specify an optimization level option @option{-O} at
9583 link-time then GCC computes one based on the optimization levels
9584 used when compiling the object files. The highest optimization
9587 Currently, the following options and their setting are take from
9588 the first object file that explicitely specified it:
9589 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
9590 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
9591 and all the @option{-m} target flags.
9593 Certain ABI changing flags are required to match in all compilation-units
9594 and trying to override this at link-time with a conflicting value
9595 is ignored. This includes options such as @option{-freg-struct-return}
9596 and @option{-fpcc-struct-return}.
9598 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
9599 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
9600 are passed through to the link stage and merged conservatively for
9601 conflicting translation units. Specifically
9602 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
9603 precedence and for example @option{-ffp-contract=off} takes precedence
9604 over @option{-ffp-contract=fast}. You can override them at linke-time.
9606 It is recommended that you compile all the files participating in the
9607 same link with the same options and also specify those options at
9610 If LTO encounters objects with C linkage declared with incompatible
9611 types in separate translation units to be linked together (undefined
9612 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
9613 issued. The behavior is still undefined at run time. Similar
9614 diagnostics may be raised for other languages.
9616 Another feature of LTO is that it is possible to apply interprocedural
9617 optimizations on files written in different languages:
9622 gfortran -c -flto baz.f90
9623 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
9626 Notice that the final link is done with @command{g++} to get the C++
9627 runtime libraries and @option{-lgfortran} is added to get the Fortran
9628 runtime libraries. In general, when mixing languages in LTO mode, you
9629 should use the same link command options as when mixing languages in a
9630 regular (non-LTO) compilation.
9632 If object files containing GIMPLE bytecode are stored in a library archive, say
9633 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
9634 are using a linker with plugin support. To create static libraries suitable
9635 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
9636 and @command{ranlib};
9637 to show the symbols of object files with GIMPLE bytecode, use
9638 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
9639 and @command{nm} have been compiled with plugin support. At link time, use the the
9640 flag @option{-fuse-linker-plugin} to ensure that the library participates in
9641 the LTO optimization process:
9644 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
9647 With the linker plugin enabled, the linker extracts the needed
9648 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
9649 to make them part of the aggregated GIMPLE image to be optimized.
9651 If you are not using a linker with plugin support and/or do not
9652 enable the linker plugin, then the objects inside @file{libfoo.a}
9653 are extracted and linked as usual, but they do not participate
9654 in the LTO optimization process. In order to make a static library suitable
9655 for both LTO optimization and usual linkage, compile its object files with
9656 @option{-flto} @option{-ffat-lto-objects}.
9658 Link-time optimizations do not require the presence of the whole program to
9659 operate. If the program does not require any symbols to be exported, it is
9660 possible to combine @option{-flto} and @option{-fwhole-program} to allow
9661 the interprocedural optimizers to use more aggressive assumptions which may
9662 lead to improved optimization opportunities.
9663 Use of @option{-fwhole-program} is not needed when linker plugin is
9664 active (see @option{-fuse-linker-plugin}).
9666 The current implementation of LTO makes no
9667 attempt to generate bytecode that is portable between different
9668 types of hosts. The bytecode files are versioned and there is a
9669 strict version check, so bytecode files generated in one version of
9670 GCC do not work with an older or newer version of GCC.
9672 Link-time optimization does not work well with generation of debugging
9673 information. Combining @option{-flto} with
9674 @option{-g} is currently experimental and expected to produce unexpected
9677 If you specify the optional @var{n}, the optimization and code
9678 generation done at link time is executed in parallel using @var{n}
9679 parallel jobs by utilizing an installed @command{make} program. The
9680 environment variable @env{MAKE} may be used to override the program
9681 used. The default value for @var{n} is 1.
9683 You can also specify @option{-flto=jobserver} to use GNU make's
9684 job server mode to determine the number of parallel jobs. This
9685 is useful when the Makefile calling GCC is already executing in parallel.
9686 You must prepend a @samp{+} to the command recipe in the parent Makefile
9687 for this to work. This option likely only works if @env{MAKE} is
9690 @item -flto-partition=@var{alg}
9691 @opindex flto-partition
9692 Specify the partitioning algorithm used by the link-time optimizer.
9693 The value is either @samp{1to1} to specify a partitioning mirroring
9694 the original source files or @samp{balanced} to specify partitioning
9695 into equally sized chunks (whenever possible) or @samp{max} to create
9696 new partition for every symbol where possible. Specifying @samp{none}
9697 as an algorithm disables partitioning and streaming completely.
9698 The default value is @samp{balanced}. While @samp{1to1} can be used
9699 as an workaround for various code ordering issues, the @samp{max}
9700 partitioning is intended for internal testing only.
9701 The value @samp{one} specifies that exactly one partition should be
9702 used while the value @samp{none} bypasses partitioning and executes
9703 the link-time optimization step directly from the WPA phase.
9705 @item -flto-odr-type-merging
9706 @opindex flto-odr-type-merging
9707 Enable streaming of mangled types names of C++ types and their unification
9708 at linktime. This increases size of LTO object files, but enable
9709 diagnostics about One Definition Rule violations.
9711 @item -flto-compression-level=@var{n}
9712 @opindex flto-compression-level
9713 This option specifies the level of compression used for intermediate
9714 language written to LTO object files, and is only meaningful in
9715 conjunction with LTO mode (@option{-flto}). Valid
9716 values are 0 (no compression) to 9 (maximum compression). Values
9717 outside this range are clamped to either 0 or 9. If the option is not
9718 given, a default balanced compression setting is used.
9721 @opindex flto-report
9722 Prints a report with internal details on the workings of the link-time
9723 optimizer. The contents of this report vary from version to version.
9724 It is meant to be useful to GCC developers when processing object
9725 files in LTO mode (via @option{-flto}).
9727 Disabled by default.
9729 @item -flto-report-wpa
9730 @opindex flto-report-wpa
9731 Like @option{-flto-report}, but only print for the WPA phase of Link
9734 @item -fuse-linker-plugin
9735 @opindex fuse-linker-plugin
9736 Enables the use of a linker plugin during link-time optimization. This
9737 option relies on plugin support in the linker, which is available in gold
9738 or in GNU ld 2.21 or newer.
9740 This option enables the extraction of object files with GIMPLE bytecode out
9741 of library archives. This improves the quality of optimization by exposing
9742 more code to the link-time optimizer. This information specifies what
9743 symbols can be accessed externally (by non-LTO object or during dynamic
9744 linking). Resulting code quality improvements on binaries (and shared
9745 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
9746 See @option{-flto} for a description of the effect of this flag and how to
9749 This option is enabled by default when LTO support in GCC is enabled
9750 and GCC was configured for use with
9751 a linker supporting plugins (GNU ld 2.21 or newer or gold).
9753 @item -ffat-lto-objects
9754 @opindex ffat-lto-objects
9755 Fat LTO objects are object files that contain both the intermediate language
9756 and the object code. This makes them usable for both LTO linking and normal
9757 linking. This option is effective only when compiling with @option{-flto}
9758 and is ignored at link time.
9760 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
9761 requires the complete toolchain to be aware of LTO. It requires a linker with
9762 linker plugin support for basic functionality. Additionally,
9763 @command{nm}, @command{ar} and @command{ranlib}
9764 need to support linker plugins to allow a full-featured build environment
9765 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
9766 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
9767 to these tools. With non fat LTO makefiles need to be modified to use them.
9769 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
9772 @item -fcompare-elim
9773 @opindex fcompare-elim
9774 After register allocation and post-register allocation instruction splitting,
9775 identify arithmetic instructions that compute processor flags similar to a
9776 comparison operation based on that arithmetic. If possible, eliminate the
9777 explicit comparison operation.
9779 This pass only applies to certain targets that cannot explicitly represent
9780 the comparison operation before register allocation is complete.
9782 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9784 @item -fcprop-registers
9785 @opindex fcprop-registers
9786 After register allocation and post-register allocation instruction splitting,
9787 perform a copy-propagation pass to try to reduce scheduling dependencies
9788 and occasionally eliminate the copy.
9790 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
9792 @item -fprofile-correction
9793 @opindex fprofile-correction
9794 Profiles collected using an instrumented binary for multi-threaded programs may
9795 be inconsistent due to missed counter updates. When this option is specified,
9796 GCC uses heuristics to correct or smooth out such inconsistencies. By
9797 default, GCC emits an error message when an inconsistent profile is detected.
9799 @item -fprofile-dir=@var{path}
9800 @opindex fprofile-dir
9802 Set the directory to search for the profile data files in to @var{path}.
9803 This option affects only the profile data generated by
9804 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9805 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9806 and its related options. Both absolute and relative paths can be used.
9807 By default, GCC uses the current directory as @var{path}, thus the
9808 profile data file appears in the same directory as the object file.
9810 @item -fprofile-generate
9811 @itemx -fprofile-generate=@var{path}
9812 @opindex fprofile-generate
9814 Enable options usually used for instrumenting application to produce
9815 profile useful for later recompilation with profile feedback based
9816 optimization. You must use @option{-fprofile-generate} both when
9817 compiling and when linking your program.
9819 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9821 If @var{path} is specified, GCC looks at the @var{path} to find
9822 the profile feedback data files. See @option{-fprofile-dir}.
9825 @itemx -fprofile-use=@var{path}
9826 @opindex fprofile-use
9827 Enable profile feedback-directed optimizations,
9828 and the following optimizations
9829 which are generally profitable only with profile feedback available:
9830 @option{-fbranch-probabilities}, @option{-fvpt},
9831 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9832 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
9834 By default, GCC emits an error message if the feedback profiles do not
9835 match the source code. This error can be turned into a warning by using
9836 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
9839 If @var{path} is specified, GCC looks at the @var{path} to find
9840 the profile feedback data files. See @option{-fprofile-dir}.
9842 @item -fauto-profile
9843 @itemx -fauto-profile=@var{path}
9844 @opindex fauto-profile
9845 Enable sampling-based feedback-directed optimizations,
9846 and the following optimizations
9847 which are generally profitable only with profile feedback available:
9848 @option{-fbranch-probabilities}, @option{-fvpt},
9849 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
9850 @option{-ftree-vectorize},
9851 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
9852 @option{-fpredictive-commoning}, @option{-funswitch-loops},
9853 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
9855 @var{path} is the name of a file containing AutoFDO profile information.
9856 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
9858 Producing an AutoFDO profile data file requires running your program
9859 with the @command{perf} utility on a supported GNU/Linux target system.
9860 For more information, see @uref{https://perf.wiki.kernel.org/}.
9864 perf record -e br_inst_retired:near_taken -b -o perf.data \
9868 Then use the @command{create_gcov} tool to convert the raw profile data
9869 to a format that can be used by GCC.@ You must also supply the
9870 unstripped binary for your program to this tool.
9871 See @uref{https://github.com/google/autofdo}.
9875 create_gcov --binary=your_program.unstripped --profile=perf.data \
9880 The following options control compiler behavior regarding floating-point
9881 arithmetic. These options trade off between speed and
9882 correctness. All must be specifically enabled.
9886 @opindex ffloat-store
9887 Do not store floating-point variables in registers, and inhibit other
9888 options that might change whether a floating-point value is taken from a
9891 @cindex floating-point precision
9892 This option prevents undesirable excess precision on machines such as
9893 the 68000 where the floating registers (of the 68881) keep more
9894 precision than a @code{double} is supposed to have. Similarly for the
9895 x86 architecture. For most programs, the excess precision does only
9896 good, but a few programs rely on the precise definition of IEEE floating
9897 point. Use @option{-ffloat-store} for such programs, after modifying
9898 them to store all pertinent intermediate computations into variables.
9900 @item -fexcess-precision=@var{style}
9901 @opindex fexcess-precision
9902 This option allows further control over excess precision on machines
9903 where floating-point registers have more precision than the IEEE
9904 @code{float} and @code{double} types and the processor does not
9905 support operations rounding to those types. By default,
9906 @option{-fexcess-precision=fast} is in effect; this means that
9907 operations are carried out in the precision of the registers and that
9908 it is unpredictable when rounding to the types specified in the source
9909 code takes place. When compiling C, if
9910 @option{-fexcess-precision=standard} is specified then excess
9911 precision follows the rules specified in ISO C99; in particular,
9912 both casts and assignments cause values to be rounded to their
9913 semantic types (whereas @option{-ffloat-store} only affects
9914 assignments). This option is enabled by default for C if a strict
9915 conformance option such as @option{-std=c99} is used.
9918 @option{-fexcess-precision=standard} is not implemented for languages
9919 other than C, and has no effect if
9920 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
9921 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
9922 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
9923 semantics apply without excess precision, and in the latter, rounding
9928 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
9929 @option{-ffinite-math-only}, @option{-fno-rounding-math},
9930 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
9932 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
9934 This option is not turned on by any @option{-O} option besides
9935 @option{-Ofast} since it can result in incorrect output for programs
9936 that depend on an exact implementation of IEEE or ISO rules/specifications
9937 for math functions. It may, however, yield faster code for programs
9938 that do not require the guarantees of these specifications.
9940 @item -fno-math-errno
9941 @opindex fno-math-errno
9942 Do not set @code{errno} after calling math functions that are executed
9943 with a single instruction, e.g., @code{sqrt}. A program that relies on
9944 IEEE exceptions for math error handling may want to use this flag
9945 for speed while maintaining IEEE arithmetic compatibility.
9947 This option is not turned on by any @option{-O} option since
9948 it can result in incorrect output for programs that depend on
9949 an exact implementation of IEEE or ISO rules/specifications for
9950 math functions. It may, however, yield faster code for programs
9951 that do not require the guarantees of these specifications.
9953 The default is @option{-fmath-errno}.
9955 On Darwin systems, the math library never sets @code{errno}. There is
9956 therefore no reason for the compiler to consider the possibility that
9957 it might, and @option{-fno-math-errno} is the default.
9959 @item -funsafe-math-optimizations
9960 @opindex funsafe-math-optimizations
9962 Allow optimizations for floating-point arithmetic that (a) assume
9963 that arguments and results are valid and (b) may violate IEEE or
9964 ANSI standards. When used at link-time, it may include libraries
9965 or startup files that change the default FPU control word or other
9966 similar optimizations.
9968 This option is not turned on by any @option{-O} option since
9969 it can result in incorrect output for programs that depend on
9970 an exact implementation of IEEE or ISO rules/specifications for
9971 math functions. It may, however, yield faster code for programs
9972 that do not require the guarantees of these specifications.
9973 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
9974 @option{-fassociative-math} and @option{-freciprocal-math}.
9976 The default is @option{-fno-unsafe-math-optimizations}.
9978 @item -fassociative-math
9979 @opindex fassociative-math
9981 Allow re-association of operands in series of floating-point operations.
9982 This violates the ISO C and C++ language standard by possibly changing
9983 computation result. NOTE: re-ordering may change the sign of zero as
9984 well as ignore NaNs and inhibit or create underflow or overflow (and
9985 thus cannot be used on code that relies on rounding behavior like
9986 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
9987 and thus may not be used when ordered comparisons are required.
9988 This option requires that both @option{-fno-signed-zeros} and
9989 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
9990 much sense with @option{-frounding-math}. For Fortran the option
9991 is automatically enabled when both @option{-fno-signed-zeros} and
9992 @option{-fno-trapping-math} are in effect.
9994 The default is @option{-fno-associative-math}.
9996 @item -freciprocal-math
9997 @opindex freciprocal-math
9999 Allow the reciprocal of a value to be used instead of dividing by
10000 the value if this enables optimizations. For example @code{x / y}
10001 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
10002 is subject to common subexpression elimination. Note that this loses
10003 precision and increases the number of flops operating on the value.
10005 The default is @option{-fno-reciprocal-math}.
10007 @item -ffinite-math-only
10008 @opindex ffinite-math-only
10009 Allow optimizations for floating-point arithmetic that assume
10010 that arguments and results are not NaNs or +-Infs.
10012 This option is not turned on by any @option{-O} option since
10013 it can result in incorrect output for programs that depend on
10014 an exact implementation of IEEE or ISO rules/specifications for
10015 math functions. It may, however, yield faster code for programs
10016 that do not require the guarantees of these specifications.
10018 The default is @option{-fno-finite-math-only}.
10020 @item -fno-signed-zeros
10021 @opindex fno-signed-zeros
10022 Allow optimizations for floating-point arithmetic that ignore the
10023 signedness of zero. IEEE arithmetic specifies the behavior of
10024 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
10025 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
10026 This option implies that the sign of a zero result isn't significant.
10028 The default is @option{-fsigned-zeros}.
10030 @item -fno-trapping-math
10031 @opindex fno-trapping-math
10032 Compile code assuming that floating-point operations cannot generate
10033 user-visible traps. These traps include division by zero, overflow,
10034 underflow, inexact result and invalid operation. This option requires
10035 that @option{-fno-signaling-nans} be in effect. Setting this option may
10036 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
10038 This option should never be turned on by any @option{-O} option since
10039 it can result in incorrect output for programs that depend on
10040 an exact implementation of IEEE or ISO rules/specifications for
10043 The default is @option{-ftrapping-math}.
10045 @item -frounding-math
10046 @opindex frounding-math
10047 Disable transformations and optimizations that assume default floating-point
10048 rounding behavior. This is round-to-zero for all floating point
10049 to integer conversions, and round-to-nearest for all other arithmetic
10050 truncations. This option should be specified for programs that change
10051 the FP rounding mode dynamically, or that may be executed with a
10052 non-default rounding mode. This option disables constant folding of
10053 floating-point expressions at compile time (which may be affected by
10054 rounding mode) and arithmetic transformations that are unsafe in the
10055 presence of sign-dependent rounding modes.
10057 The default is @option{-fno-rounding-math}.
10059 This option is experimental and does not currently guarantee to
10060 disable all GCC optimizations that are affected by rounding mode.
10061 Future versions of GCC may provide finer control of this setting
10062 using C99's @code{FENV_ACCESS} pragma. This command-line option
10063 will be used to specify the default state for @code{FENV_ACCESS}.
10065 @item -fsignaling-nans
10066 @opindex fsignaling-nans
10067 Compile code assuming that IEEE signaling NaNs may generate user-visible
10068 traps during floating-point operations. Setting this option disables
10069 optimizations that may change the number of exceptions visible with
10070 signaling NaNs. This option implies @option{-ftrapping-math}.
10072 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
10075 The default is @option{-fno-signaling-nans}.
10077 This option is experimental and does not currently guarantee to
10078 disable all GCC optimizations that affect signaling NaN behavior.
10080 @item -fsingle-precision-constant
10081 @opindex fsingle-precision-constant
10082 Treat floating-point constants as single precision instead of
10083 implicitly converting them to double-precision constants.
10085 @item -fcx-limited-range
10086 @opindex fcx-limited-range
10087 When enabled, this option states that a range reduction step is not
10088 needed when performing complex division. Also, there is no checking
10089 whether the result of a complex multiplication or division is @code{NaN
10090 + I*NaN}, with an attempt to rescue the situation in that case. The
10091 default is @option{-fno-cx-limited-range}, but is enabled by
10092 @option{-ffast-math}.
10094 This option controls the default setting of the ISO C99
10095 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
10098 @item -fcx-fortran-rules
10099 @opindex fcx-fortran-rules
10100 Complex multiplication and division follow Fortran rules. Range
10101 reduction is done as part of complex division, but there is no checking
10102 whether the result of a complex multiplication or division is @code{NaN
10103 + I*NaN}, with an attempt to rescue the situation in that case.
10105 The default is @option{-fno-cx-fortran-rules}.
10109 The following options control optimizations that may improve
10110 performance, but are not enabled by any @option{-O} options. This
10111 section includes experimental options that may produce broken code.
10114 @item -fbranch-probabilities
10115 @opindex fbranch-probabilities
10116 After running a program compiled with @option{-fprofile-arcs}
10117 (@pxref{Debugging Options,, Options for Debugging Your Program or
10118 @command{gcc}}), you can compile it a second time using
10119 @option{-fbranch-probabilities}, to improve optimizations based on
10120 the number of times each branch was taken. When a program
10121 compiled with @option{-fprofile-arcs} exits, it saves arc execution
10122 counts to a file called @file{@var{sourcename}.gcda} for each source
10123 file. The information in this data file is very dependent on the
10124 structure of the generated code, so you must use the same source code
10125 and the same optimization options for both compilations.
10127 With @option{-fbranch-probabilities}, GCC puts a
10128 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
10129 These can be used to improve optimization. Currently, they are only
10130 used in one place: in @file{reorg.c}, instead of guessing which path a
10131 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
10132 exactly determine which path is taken more often.
10134 @item -fprofile-values
10135 @opindex fprofile-values
10136 If combined with @option{-fprofile-arcs}, it adds code so that some
10137 data about values of expressions in the program is gathered.
10139 With @option{-fbranch-probabilities}, it reads back the data gathered
10140 from profiling values of expressions for usage in optimizations.
10142 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
10144 @item -fprofile-reorder-functions
10145 @opindex fprofile-reorder-functions
10146 Function reordering based on profile instrumentation collects
10147 first time of execution of a function and orders these functions
10148 in ascending order.
10150 Enabled with @option{-fprofile-use}.
10154 If combined with @option{-fprofile-arcs}, this option instructs the compiler
10155 to add code to gather information about values of expressions.
10157 With @option{-fbranch-probabilities}, it reads back the data gathered
10158 and actually performs the optimizations based on them.
10159 Currently the optimizations include specialization of division operations
10160 using the knowledge about the value of the denominator.
10162 @item -frename-registers
10163 @opindex frename-registers
10164 Attempt to avoid false dependencies in scheduled code by making use
10165 of registers left over after register allocation. This optimization
10166 most benefits processors with lots of registers. Depending on the
10167 debug information format adopted by the target, however, it can
10168 make debugging impossible, since variables no longer stay in
10169 a ``home register''.
10171 Enabled by default with @option{-funroll-loops} and @option{-fpeel-loops}.
10173 @item -fschedule-fusion
10174 @opindex fschedule-fusion
10175 Performs a target dependent pass over the instruction stream to schedule
10176 instructions of same type together because target machine can execute them
10177 more efficiently if they are adjacent to each other in the instruction flow.
10179 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
10183 Perform tail duplication to enlarge superblock size. This transformation
10184 simplifies the control flow of the function allowing other optimizations to do
10187 Enabled with @option{-fprofile-use}.
10189 @item -funroll-loops
10190 @opindex funroll-loops
10191 Unroll loops whose number of iterations can be determined at compile time or
10192 upon entry to the loop. @option{-funroll-loops} implies
10193 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
10194 It also turns on complete loop peeling (i.e.@: complete removal of loops with
10195 a small constant number of iterations). This option makes code larger, and may
10196 or may not make it run faster.
10198 Enabled with @option{-fprofile-use}.
10200 @item -funroll-all-loops
10201 @opindex funroll-all-loops
10202 Unroll all loops, even if their number of iterations is uncertain when
10203 the loop is entered. This usually makes programs run more slowly.
10204 @option{-funroll-all-loops} implies the same options as
10205 @option{-funroll-loops}.
10208 @opindex fpeel-loops
10209 Peels loops for which there is enough information that they do not
10210 roll much (from profile feedback). It also turns on complete loop peeling
10211 (i.e.@: complete removal of loops with small constant number of iterations).
10213 Enabled with @option{-fprofile-use}.
10215 @item -fmove-loop-invariants
10216 @opindex fmove-loop-invariants
10217 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
10218 at level @option{-O1}
10220 @item -funswitch-loops
10221 @opindex funswitch-loops
10222 Move branches with loop invariant conditions out of the loop, with duplicates
10223 of the loop on both branches (modified according to result of the condition).
10225 @item -ffunction-sections
10226 @itemx -fdata-sections
10227 @opindex ffunction-sections
10228 @opindex fdata-sections
10229 Place each function or data item into its own section in the output
10230 file if the target supports arbitrary sections. The name of the
10231 function or the name of the data item determines the section's name
10232 in the output file.
10234 Use these options on systems where the linker can perform optimizations
10235 to improve locality of reference in the instruction space. Most systems
10236 using the ELF object format and SPARC processors running Solaris 2 have
10237 linkers with such optimizations. AIX may have these optimizations in
10240 Only use these options when there are significant benefits from doing
10241 so. When you specify these options, the assembler and linker
10242 create larger object and executable files and are also slower.
10243 You cannot use @command{gprof} on all systems if you
10244 specify this option, and you may have problems with debugging if
10245 you specify both this option and @option{-g}.
10247 @item -fbranch-target-load-optimize
10248 @opindex fbranch-target-load-optimize
10249 Perform branch target register load optimization before prologue / epilogue
10251 The use of target registers can typically be exposed only during reload,
10252 thus hoisting loads out of loops and doing inter-block scheduling needs
10253 a separate optimization pass.
10255 @item -fbranch-target-load-optimize2
10256 @opindex fbranch-target-load-optimize2
10257 Perform branch target register load optimization after prologue / epilogue
10260 @item -fbtr-bb-exclusive
10261 @opindex fbtr-bb-exclusive
10262 When performing branch target register load optimization, don't reuse
10263 branch target registers within any basic block.
10265 @item -fstack-protector
10266 @opindex fstack-protector
10267 Emit extra code to check for buffer overflows, such as stack smashing
10268 attacks. This is done by adding a guard variable to functions with
10269 vulnerable objects. This includes functions that call @code{alloca}, and
10270 functions with buffers larger than 8 bytes. The guards are initialized
10271 when a function is entered and then checked when the function exits.
10272 If a guard check fails, an error message is printed and the program exits.
10274 @item -fstack-protector-all
10275 @opindex fstack-protector-all
10276 Like @option{-fstack-protector} except that all functions are protected.
10278 @item -fstack-protector-strong
10279 @opindex fstack-protector-strong
10280 Like @option{-fstack-protector} but includes additional functions to
10281 be protected --- those that have local array definitions, or have
10282 references to local frame addresses.
10284 @item -fstack-protector-explicit
10285 @opindex fstack-protector-explicit
10286 Like @option{-fstack-protector} but only protects those functions which
10287 have the @code{stack_protect} attribute.
10290 @opindex fstdarg-opt
10291 Optimize the prologue of variadic argument functions with respect to usage of
10294 @item -fsection-anchors
10295 @opindex fsection-anchors
10296 Try to reduce the number of symbolic address calculations by using
10297 shared ``anchor'' symbols to address nearby objects. This transformation
10298 can help to reduce the number of GOT entries and GOT accesses on some
10301 For example, the implementation of the following function @code{foo}:
10304 static int a, b, c;
10305 int foo (void) @{ return a + b + c; @}
10309 usually calculates the addresses of all three variables, but if you
10310 compile it with @option{-fsection-anchors}, it accesses the variables
10311 from a common anchor point instead. The effect is similar to the
10312 following pseudocode (which isn't valid C):
10317 register int *xr = &x;
10318 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
10322 Not all targets support this option.
10324 @item --param @var{name}=@var{value}
10326 In some places, GCC uses various constants to control the amount of
10327 optimization that is done. For example, GCC does not inline functions
10328 that contain more than a certain number of instructions. You can
10329 control some of these constants on the command line using the
10330 @option{--param} option.
10332 The names of specific parameters, and the meaning of the values, are
10333 tied to the internals of the compiler, and are subject to change
10334 without notice in future releases.
10336 In each case, the @var{value} is an integer. The allowable choices for
10340 @item predictable-branch-outcome
10341 When branch is predicted to be taken with probability lower than this threshold
10342 (in percent), then it is considered well predictable. The default is 10.
10344 @item max-crossjump-edges
10345 The maximum number of incoming edges to consider for cross-jumping.
10346 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
10347 the number of edges incoming to each block. Increasing values mean
10348 more aggressive optimization, making the compilation time increase with
10349 probably small improvement in executable size.
10351 @item min-crossjump-insns
10352 The minimum number of instructions that must be matched at the end
10353 of two blocks before cross-jumping is performed on them. This
10354 value is ignored in the case where all instructions in the block being
10355 cross-jumped from are matched. The default value is 5.
10357 @item max-grow-copy-bb-insns
10358 The maximum code size expansion factor when copying basic blocks
10359 instead of jumping. The expansion is relative to a jump instruction.
10360 The default value is 8.
10362 @item max-goto-duplication-insns
10363 The maximum number of instructions to duplicate to a block that jumps
10364 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
10365 passes, GCC factors computed gotos early in the compilation process,
10366 and unfactors them as late as possible. Only computed jumps at the
10367 end of a basic blocks with no more than max-goto-duplication-insns are
10368 unfactored. The default value is 8.
10370 @item max-delay-slot-insn-search
10371 The maximum number of instructions to consider when looking for an
10372 instruction to fill a delay slot. If more than this arbitrary number of
10373 instructions are searched, the time savings from filling the delay slot
10374 are minimal, so stop searching. Increasing values mean more
10375 aggressive optimization, making the compilation time increase with probably
10376 small improvement in execution time.
10378 @item max-delay-slot-live-search
10379 When trying to fill delay slots, the maximum number of instructions to
10380 consider when searching for a block with valid live register
10381 information. Increasing this arbitrarily chosen value means more
10382 aggressive optimization, increasing the compilation time. This parameter
10383 should be removed when the delay slot code is rewritten to maintain the
10384 control-flow graph.
10386 @item max-gcse-memory
10387 The approximate maximum amount of memory that can be allocated in
10388 order to perform the global common subexpression elimination
10389 optimization. If more memory than specified is required, the
10390 optimization is not done.
10392 @item max-gcse-insertion-ratio
10393 If the ratio of expression insertions to deletions is larger than this value
10394 for any expression, then RTL PRE inserts or removes the expression and thus
10395 leaves partially redundant computations in the instruction stream. The default value is 20.
10397 @item max-pending-list-length
10398 The maximum number of pending dependencies scheduling allows
10399 before flushing the current state and starting over. Large functions
10400 with few branches or calls can create excessively large lists which
10401 needlessly consume memory and resources.
10403 @item max-modulo-backtrack-attempts
10404 The maximum number of backtrack attempts the scheduler should make
10405 when modulo scheduling a loop. Larger values can exponentially increase
10408 @item max-inline-insns-single
10409 Several parameters control the tree inliner used in GCC@.
10410 This number sets the maximum number of instructions (counted in GCC's
10411 internal representation) in a single function that the tree inliner
10412 considers for inlining. This only affects functions declared
10413 inline and methods implemented in a class declaration (C++).
10414 The default value is 400.
10416 @item max-inline-insns-auto
10417 When you use @option{-finline-functions} (included in @option{-O3}),
10418 a lot of functions that would otherwise not be considered for inlining
10419 by the compiler are investigated. To those functions, a different
10420 (more restrictive) limit compared to functions declared inline can
10422 The default value is 40.
10424 @item inline-min-speedup
10425 When estimated performance improvement of caller + callee runtime exceeds this
10426 threshold (in precent), the function can be inlined regardless the limit on
10427 @option{--param max-inline-insns-single} and @option{--param
10428 max-inline-insns-auto}.
10430 @item large-function-insns
10431 The limit specifying really large functions. For functions larger than this
10432 limit after inlining, inlining is constrained by
10433 @option{--param large-function-growth}. This parameter is useful primarily
10434 to avoid extreme compilation time caused by non-linear algorithms used by the
10436 The default value is 2700.
10438 @item large-function-growth
10439 Specifies maximal growth of large function caused by inlining in percents.
10440 The default value is 100 which limits large function growth to 2.0 times
10443 @item large-unit-insns
10444 The limit specifying large translation unit. Growth caused by inlining of
10445 units larger than this limit is limited by @option{--param inline-unit-growth}.
10446 For small units this might be too tight.
10447 For example, consider a unit consisting of function A
10448 that is inline and B that just calls A three times. If B is small relative to
10449 A, the growth of unit is 300\% and yet such inlining is very sane. For very
10450 large units consisting of small inlineable functions, however, the overall unit
10451 growth limit is needed to avoid exponential explosion of code size. Thus for
10452 smaller units, the size is increased to @option{--param large-unit-insns}
10453 before applying @option{--param inline-unit-growth}. The default is 10000.
10455 @item inline-unit-growth
10456 Specifies maximal overall growth of the compilation unit caused by inlining.
10457 The default value is 20 which limits unit growth to 1.2 times the original
10458 size. Cold functions (either marked cold via an attribute or by profile
10459 feedback) are not accounted into the unit size.
10461 @item ipcp-unit-growth
10462 Specifies maximal overall growth of the compilation unit caused by
10463 interprocedural constant propagation. The default value is 10 which limits
10464 unit growth to 1.1 times the original size.
10466 @item large-stack-frame
10467 The limit specifying large stack frames. While inlining the algorithm is trying
10468 to not grow past this limit too much. The default value is 256 bytes.
10470 @item large-stack-frame-growth
10471 Specifies maximal growth of large stack frames caused by inlining in percents.
10472 The default value is 1000 which limits large stack frame growth to 11 times
10475 @item max-inline-insns-recursive
10476 @itemx max-inline-insns-recursive-auto
10477 Specifies the maximum number of instructions an out-of-line copy of a
10478 self-recursive inline
10479 function can grow into by performing recursive inlining.
10481 @option{--param max-inline-insns-recursive} applies to functions
10483 For functions not declared inline, recursive inlining
10484 happens only when @option{-finline-functions} (included in @option{-O3}) is
10485 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
10486 default value is 450.
10488 @item max-inline-recursive-depth
10489 @itemx max-inline-recursive-depth-auto
10490 Specifies the maximum recursion depth used for recursive inlining.
10492 @option{--param max-inline-recursive-depth} applies to functions
10493 declared inline. For functions not declared inline, recursive inlining
10494 happens only when @option{-finline-functions} (included in @option{-O3}) is
10495 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
10496 default value is 8.
10498 @item min-inline-recursive-probability
10499 Recursive inlining is profitable only for function having deep recursion
10500 in average and can hurt for function having little recursion depth by
10501 increasing the prologue size or complexity of function body to other
10504 When profile feedback is available (see @option{-fprofile-generate}) the actual
10505 recursion depth can be guessed from probability that function recurses via a
10506 given call expression. This parameter limits inlining only to call expressions
10507 whose probability exceeds the given threshold (in percents).
10508 The default value is 10.
10510 @item early-inlining-insns
10511 Specify growth that the early inliner can make. In effect it increases
10512 the amount of inlining for code having a large abstraction penalty.
10513 The default value is 14.
10515 @item max-early-inliner-iterations
10516 Limit of iterations of the early inliner. This basically bounds
10517 the number of nested indirect calls the early inliner can resolve.
10518 Deeper chains are still handled by late inlining.
10520 @item comdat-sharing-probability
10521 Probability (in percent) that C++ inline function with comdat visibility
10522 are shared across multiple compilation units. The default value is 20.
10524 @item profile-func-internal-id
10525 A parameter to control whether to use function internal id in profile
10526 database lookup. If the value is 0, the compiler uses an id that
10527 is based on function assembler name and filename, which makes old profile
10528 data more tolerant to source changes such as function reordering etc.
10529 The default value is 0.
10531 @item min-vect-loop-bound
10532 The minimum number of iterations under which loops are not vectorized
10533 when @option{-ftree-vectorize} is used. The number of iterations after
10534 vectorization needs to be greater than the value specified by this option
10535 to allow vectorization. The default value is 0.
10537 @item gcse-cost-distance-ratio
10538 Scaling factor in calculation of maximum distance an expression
10539 can be moved by GCSE optimizations. This is currently supported only in the
10540 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
10541 is with simple expressions, i.e., the expressions that have cost
10542 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
10543 hoisting of simple expressions. The default value is 10.
10545 @item gcse-unrestricted-cost
10546 Cost, roughly measured as the cost of a single typical machine
10547 instruction, at which GCSE optimizations do not constrain
10548 the distance an expression can travel. This is currently
10549 supported only in the code hoisting pass. The lesser the cost,
10550 the more aggressive code hoisting is. Specifying 0
10551 allows all expressions to travel unrestricted distances.
10552 The default value is 3.
10554 @item max-hoist-depth
10555 The depth of search in the dominator tree for expressions to hoist.
10556 This is used to avoid quadratic behavior in hoisting algorithm.
10557 The value of 0 does not limit on the search, but may slow down compilation
10558 of huge functions. The default value is 30.
10560 @item max-tail-merge-comparisons
10561 The maximum amount of similar bbs to compare a bb with. This is used to
10562 avoid quadratic behavior in tree tail merging. The default value is 10.
10564 @item max-tail-merge-iterations
10565 The maximum amount of iterations of the pass over the function. This is used to
10566 limit compilation time in tree tail merging. The default value is 2.
10568 @item max-unrolled-insns
10569 The maximum number of instructions that a loop may have to be unrolled.
10570 If a loop is unrolled, this parameter also determines how many times
10571 the loop code is unrolled.
10573 @item max-average-unrolled-insns
10574 The maximum number of instructions biased by probabilities of their execution
10575 that a loop may have to be unrolled. If a loop is unrolled,
10576 this parameter also determines how many times the loop code is unrolled.
10578 @item max-unroll-times
10579 The maximum number of unrollings of a single loop.
10581 @item max-peeled-insns
10582 The maximum number of instructions that a loop may have to be peeled.
10583 If a loop is peeled, this parameter also determines how many times
10584 the loop code is peeled.
10586 @item max-peel-times
10587 The maximum number of peelings of a single loop.
10589 @item max-peel-branches
10590 The maximum number of branches on the hot path through the peeled sequence.
10592 @item max-completely-peeled-insns
10593 The maximum number of insns of a completely peeled loop.
10595 @item max-completely-peel-times
10596 The maximum number of iterations of a loop to be suitable for complete peeling.
10598 @item max-completely-peel-loop-nest-depth
10599 The maximum depth of a loop nest suitable for complete peeling.
10601 @item max-unswitch-insns
10602 The maximum number of insns of an unswitched loop.
10604 @item max-unswitch-level
10605 The maximum number of branches unswitched in a single loop.
10607 @item lim-expensive
10608 The minimum cost of an expensive expression in the loop invariant motion.
10610 @item iv-consider-all-candidates-bound
10611 Bound on number of candidates for induction variables, below which
10612 all candidates are considered for each use in induction variable
10613 optimizations. If there are more candidates than this,
10614 only the most relevant ones are considered to avoid quadratic time complexity.
10616 @item iv-max-considered-uses
10617 The induction variable optimizations give up on loops that contain more
10618 induction variable uses.
10620 @item iv-always-prune-cand-set-bound
10621 If the number of candidates in the set is smaller than this value,
10622 always try to remove unnecessary ivs from the set
10623 when adding a new one.
10625 @item scev-max-expr-size
10626 Bound on size of expressions used in the scalar evolutions analyzer.
10627 Large expressions slow the analyzer.
10629 @item scev-max-expr-complexity
10630 Bound on the complexity of the expressions in the scalar evolutions analyzer.
10631 Complex expressions slow the analyzer.
10633 @item vect-max-version-for-alignment-checks
10634 The maximum number of run-time checks that can be performed when
10635 doing loop versioning for alignment in the vectorizer.
10637 @item vect-max-version-for-alias-checks
10638 The maximum number of run-time checks that can be performed when
10639 doing loop versioning for alias in the vectorizer.
10641 @item vect-max-peeling-for-alignment
10642 The maximum number of loop peels to enhance access alignment
10643 for vectorizer. Value -1 means 'no limit'.
10645 @item max-iterations-to-track
10646 The maximum number of iterations of a loop the brute-force algorithm
10647 for analysis of the number of iterations of the loop tries to evaluate.
10649 @item hot-bb-count-ws-permille
10650 A basic block profile count is considered hot if it contributes to
10651 the given permillage (i.e. 0...1000) of the entire profiled execution.
10653 @item hot-bb-frequency-fraction
10654 Select fraction of the entry block frequency of executions of basic block in
10655 function given basic block needs to have to be considered hot.
10657 @item max-predicted-iterations
10658 The maximum number of loop iterations we predict statically. This is useful
10659 in cases where a function contains a single loop with known bound and
10660 another loop with unknown bound.
10661 The known number of iterations is predicted correctly, while
10662 the unknown number of iterations average to roughly 10. This means that the
10663 loop without bounds appears artificially cold relative to the other one.
10665 @item builtin-expect-probability
10666 Control the probability of the expression having the specified value. This
10667 parameter takes a percentage (i.e. 0 ... 100) as input.
10668 The default probability of 90 is obtained empirically.
10670 @item align-threshold
10672 Select fraction of the maximal frequency of executions of a basic block in
10673 a function to align the basic block.
10675 @item align-loop-iterations
10677 A loop expected to iterate at least the selected number of iterations is
10680 @item tracer-dynamic-coverage
10681 @itemx tracer-dynamic-coverage-feedback
10683 This value is used to limit superblock formation once the given percentage of
10684 executed instructions is covered. This limits unnecessary code size
10687 The @option{tracer-dynamic-coverage-feedback} parameter
10688 is used only when profile
10689 feedback is available. The real profiles (as opposed to statically estimated
10690 ones) are much less balanced allowing the threshold to be larger value.
10692 @item tracer-max-code-growth
10693 Stop tail duplication once code growth has reached given percentage. This is
10694 a rather artificial limit, as most of the duplicates are eliminated later in
10695 cross jumping, so it may be set to much higher values than is the desired code
10698 @item tracer-min-branch-ratio
10700 Stop reverse growth when the reverse probability of best edge is less than this
10701 threshold (in percent).
10703 @item tracer-min-branch-ratio
10704 @itemx tracer-min-branch-ratio-feedback
10706 Stop forward growth if the best edge has probability lower than this
10709 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
10710 compilation for profile feedback and one for compilation without. The value
10711 for compilation with profile feedback needs to be more conservative (higher) in
10712 order to make tracer effective.
10714 @item max-cse-path-length
10716 The maximum number of basic blocks on path that CSE considers.
10719 @item max-cse-insns
10720 The maximum number of instructions CSE processes before flushing.
10721 The default is 1000.
10723 @item ggc-min-expand
10725 GCC uses a garbage collector to manage its own memory allocation. This
10726 parameter specifies the minimum percentage by which the garbage
10727 collector's heap should be allowed to expand between collections.
10728 Tuning this may improve compilation speed; it has no effect on code
10731 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
10732 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
10733 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
10734 GCC is not able to calculate RAM on a particular platform, the lower
10735 bound of 30% is used. Setting this parameter and
10736 @option{ggc-min-heapsize} to zero causes a full collection to occur at
10737 every opportunity. This is extremely slow, but can be useful for
10740 @item ggc-min-heapsize
10742 Minimum size of the garbage collector's heap before it begins bothering
10743 to collect garbage. The first collection occurs after the heap expands
10744 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
10745 tuning this may improve compilation speed, and has no effect on code
10748 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
10749 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
10750 with a lower bound of 4096 (four megabytes) and an upper bound of
10751 131072 (128 megabytes). If GCC is not able to calculate RAM on a
10752 particular platform, the lower bound is used. Setting this parameter
10753 very large effectively disables garbage collection. Setting this
10754 parameter and @option{ggc-min-expand} to zero causes a full collection
10755 to occur at every opportunity.
10757 @item max-reload-search-insns
10758 The maximum number of instruction reload should look backward for equivalent
10759 register. Increasing values mean more aggressive optimization, making the
10760 compilation time increase with probably slightly better performance.
10761 The default value is 100.
10763 @item max-cselib-memory-locations
10764 The maximum number of memory locations cselib should take into account.
10765 Increasing values mean more aggressive optimization, making the compilation time
10766 increase with probably slightly better performance. The default value is 500.
10768 @item reorder-blocks-duplicate
10769 @itemx reorder-blocks-duplicate-feedback
10771 Used by the basic block reordering pass to decide whether to use unconditional
10772 branch or duplicate the code on its destination. Code is duplicated when its
10773 estimated size is smaller than this value multiplied by the estimated size of
10774 unconditional jump in the hot spots of the program.
10776 The @option{reorder-block-duplicate-feedback} parameter
10777 is used only when profile
10778 feedback is available. It may be set to higher values than
10779 @option{reorder-block-duplicate} since information about the hot spots is more
10782 @item max-sched-ready-insns
10783 The maximum number of instructions ready to be issued the scheduler should
10784 consider at any given time during the first scheduling pass. Increasing
10785 values mean more thorough searches, making the compilation time increase
10786 with probably little benefit. The default value is 100.
10788 @item max-sched-region-blocks
10789 The maximum number of blocks in a region to be considered for
10790 interblock scheduling. The default value is 10.
10792 @item max-pipeline-region-blocks
10793 The maximum number of blocks in a region to be considered for
10794 pipelining in the selective scheduler. The default value is 15.
10796 @item max-sched-region-insns
10797 The maximum number of insns in a region to be considered for
10798 interblock scheduling. The default value is 100.
10800 @item max-pipeline-region-insns
10801 The maximum number of insns in a region to be considered for
10802 pipelining in the selective scheduler. The default value is 200.
10804 @item min-spec-prob
10805 The minimum probability (in percents) of reaching a source block
10806 for interblock speculative scheduling. The default value is 40.
10808 @item max-sched-extend-regions-iters
10809 The maximum number of iterations through CFG to extend regions.
10810 A value of 0 (the default) disables region extensions.
10812 @item max-sched-insn-conflict-delay
10813 The maximum conflict delay for an insn to be considered for speculative motion.
10814 The default value is 3.
10816 @item sched-spec-prob-cutoff
10817 The minimal probability of speculation success (in percents), so that
10818 speculative insns are scheduled.
10819 The default value is 40.
10821 @item sched-spec-state-edge-prob-cutoff
10822 The minimum probability an edge must have for the scheduler to save its
10824 The default value is 10.
10826 @item sched-mem-true-dep-cost
10827 Minimal distance (in CPU cycles) between store and load targeting same
10828 memory locations. The default value is 1.
10830 @item selsched-max-lookahead
10831 The maximum size of the lookahead window of selective scheduling. It is a
10832 depth of search for available instructions.
10833 The default value is 50.
10835 @item selsched-max-sched-times
10836 The maximum number of times that an instruction is scheduled during
10837 selective scheduling. This is the limit on the number of iterations
10838 through which the instruction may be pipelined. The default value is 2.
10840 @item selsched-max-insns-to-rename
10841 The maximum number of best instructions in the ready list that are considered
10842 for renaming in the selective scheduler. The default value is 2.
10845 The minimum value of stage count that swing modulo scheduler
10846 generates. The default value is 2.
10848 @item max-last-value-rtl
10849 The maximum size measured as number of RTLs that can be recorded in an expression
10850 in combiner for a pseudo register as last known value of that register. The default
10853 @item max-combine-insns
10854 The maximum number of instructions the RTL combiner tries to combine.
10855 The default value is 2 at @option{-Og} and 4 otherwise.
10857 @item integer-share-limit
10858 Small integer constants can use a shared data structure, reducing the
10859 compiler's memory usage and increasing its speed. This sets the maximum
10860 value of a shared integer constant. The default value is 256.
10862 @item ssp-buffer-size
10863 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
10864 protection when @option{-fstack-protection} is used.
10866 @item min-size-for-stack-sharing
10867 The minimum size of variables taking part in stack slot sharing when not
10868 optimizing. The default value is 32.
10870 @item max-jump-thread-duplication-stmts
10871 Maximum number of statements allowed in a block that needs to be
10872 duplicated when threading jumps.
10874 @item max-fields-for-field-sensitive
10875 Maximum number of fields in a structure treated in
10876 a field sensitive manner during pointer analysis. The default is zero
10877 for @option{-O0} and @option{-O1},
10878 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
10880 @item prefetch-latency
10881 Estimate on average number of instructions that are executed before
10882 prefetch finishes. The distance prefetched ahead is proportional
10883 to this constant. Increasing this number may also lead to less
10884 streams being prefetched (see @option{simultaneous-prefetches}).
10886 @item simultaneous-prefetches
10887 Maximum number of prefetches that can run at the same time.
10889 @item l1-cache-line-size
10890 The size of cache line in L1 cache, in bytes.
10892 @item l1-cache-size
10893 The size of L1 cache, in kilobytes.
10895 @item l2-cache-size
10896 The size of L2 cache, in kilobytes.
10898 @item min-insn-to-prefetch-ratio
10899 The minimum ratio between the number of instructions and the
10900 number of prefetches to enable prefetching in a loop.
10902 @item prefetch-min-insn-to-mem-ratio
10903 The minimum ratio between the number of instructions and the
10904 number of memory references to enable prefetching in a loop.
10906 @item use-canonical-types
10907 Whether the compiler should use the ``canonical'' type system. By
10908 default, this should always be 1, which uses a more efficient internal
10909 mechanism for comparing types in C++ and Objective-C++. However, if
10910 bugs in the canonical type system are causing compilation failures,
10911 set this value to 0 to disable canonical types.
10913 @item switch-conversion-max-branch-ratio
10914 Switch initialization conversion refuses to create arrays that are
10915 bigger than @option{switch-conversion-max-branch-ratio} times the number of
10916 branches in the switch.
10918 @item max-partial-antic-length
10919 Maximum length of the partial antic set computed during the tree
10920 partial redundancy elimination optimization (@option{-ftree-pre}) when
10921 optimizing at @option{-O3} and above. For some sorts of source code
10922 the enhanced partial redundancy elimination optimization can run away,
10923 consuming all of the memory available on the host machine. This
10924 parameter sets a limit on the length of the sets that are computed,
10925 which prevents the runaway behavior. Setting a value of 0 for
10926 this parameter allows an unlimited set length.
10928 @item sccvn-max-scc-size
10929 Maximum size of a strongly connected component (SCC) during SCCVN
10930 processing. If this limit is hit, SCCVN processing for the whole
10931 function is not done and optimizations depending on it are
10932 disabled. The default maximum SCC size is 10000.
10934 @item sccvn-max-alias-queries-per-access
10935 Maximum number of alias-oracle queries we perform when looking for
10936 redundancies for loads and stores. If this limit is hit the search
10937 is aborted and the load or store is not considered redundant. The
10938 number of queries is algorithmically limited to the number of
10939 stores on all paths from the load to the function entry.
10940 The default maxmimum number of queries is 1000.
10942 @item ira-max-loops-num
10943 IRA uses regional register allocation by default. If a function
10944 contains more loops than the number given by this parameter, only at most
10945 the given number of the most frequently-executed loops form regions
10946 for regional register allocation. The default value of the
10949 @item ira-max-conflict-table-size
10950 Although IRA uses a sophisticated algorithm to compress the conflict
10951 table, the table can still require excessive amounts of memory for
10952 huge functions. If the conflict table for a function could be more
10953 than the size in MB given by this parameter, the register allocator
10954 instead uses a faster, simpler, and lower-quality
10955 algorithm that does not require building a pseudo-register conflict table.
10956 The default value of the parameter is 2000.
10958 @item ira-loop-reserved-regs
10959 IRA can be used to evaluate more accurate register pressure in loops
10960 for decisions to move loop invariants (see @option{-O3}). The number
10961 of available registers reserved for some other purposes is given
10962 by this parameter. The default value of the parameter is 2, which is
10963 the minimal number of registers needed by typical instructions.
10964 This value is the best found from numerous experiments.
10966 @item lra-inheritance-ebb-probability-cutoff
10967 LRA tries to reuse values reloaded in registers in subsequent insns.
10968 This optimization is called inheritance. EBB is used as a region to
10969 do this optimization. The parameter defines a minimal fall-through
10970 edge probability in percentage used to add BB to inheritance EBB in
10971 LRA. The default value of the parameter is 40. The value was chosen
10972 from numerous runs of SPEC2000 on x86-64.
10974 @item loop-invariant-max-bbs-in-loop
10975 Loop invariant motion can be very expensive, both in compilation time and
10976 in amount of needed compile-time memory, with very large loops. Loops
10977 with more basic blocks than this parameter won't have loop invariant
10978 motion optimization performed on them. The default value of the
10979 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
10981 @item loop-max-datarefs-for-datadeps
10982 Building data dapendencies is expensive for very large loops. This
10983 parameter limits the number of data references in loops that are
10984 considered for data dependence analysis. These large loops are no
10985 handled by the optimizations using loop data dependencies.
10986 The default value is 1000.
10988 @item max-vartrack-size
10989 Sets a maximum number of hash table slots to use during variable
10990 tracking dataflow analysis of any function. If this limit is exceeded
10991 with variable tracking at assignments enabled, analysis for that
10992 function is retried without it, after removing all debug insns from
10993 the function. If the limit is exceeded even without debug insns, var
10994 tracking analysis is completely disabled for the function. Setting
10995 the parameter to zero makes it unlimited.
10997 @item max-vartrack-expr-depth
10998 Sets a maximum number of recursion levels when attempting to map
10999 variable names or debug temporaries to value expressions. This trades
11000 compilation time for more complete debug information. If this is set too
11001 low, value expressions that are available and could be represented in
11002 debug information may end up not being used; setting this higher may
11003 enable the compiler to find more complex debug expressions, but compile
11004 time and memory use may grow. The default is 12.
11006 @item min-nondebug-insn-uid
11007 Use uids starting at this parameter for nondebug insns. The range below
11008 the parameter is reserved exclusively for debug insns created by
11009 @option{-fvar-tracking-assignments}, but debug insns may get
11010 (non-overlapping) uids above it if the reserved range is exhausted.
11012 @item ipa-sra-ptr-growth-factor
11013 IPA-SRA replaces a pointer to an aggregate with one or more new
11014 parameters only when their cumulative size is less or equal to
11015 @option{ipa-sra-ptr-growth-factor} times the size of the original
11018 @item sra-max-scalarization-size-Ospeed
11019 @item sra-max-scalarization-size-Osize
11020 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
11021 replace scalar parts of aggregates with uses of independent scalar
11022 variables. These parameters control the maximum size, in storage units,
11023 of aggregate which is considered for replacement when compiling for
11025 (@option{sra-max-scalarization-size-Ospeed}) or size
11026 (@option{sra-max-scalarization-size-Osize}) respectively.
11028 @item tm-max-aggregate-size
11029 When making copies of thread-local variables in a transaction, this
11030 parameter specifies the size in bytes after which variables are
11031 saved with the logging functions as opposed to save/restore code
11032 sequence pairs. This option only applies when using
11035 @item graphite-max-nb-scop-params
11036 To avoid exponential effects in the Graphite loop transforms, the
11037 number of parameters in a Static Control Part (SCoP) is bounded. The
11038 default value is 10 parameters. A variable whose value is unknown at
11039 compilation time and defined outside a SCoP is a parameter of the SCoP.
11041 @item graphite-max-bbs-per-function
11042 To avoid exponential effects in the detection of SCoPs, the size of
11043 the functions analyzed by Graphite is bounded. The default value is
11046 @item loop-block-tile-size
11047 Loop blocking or strip mining transforms, enabled with
11048 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
11049 loop in the loop nest by a given number of iterations. The strip
11050 length can be changed using the @option{loop-block-tile-size}
11051 parameter. The default value is 51 iterations.
11053 @item loop-unroll-jam-size
11054 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
11055 default value is 4.
11057 @item loop-unroll-jam-depth
11058 Specify the dimension to be unrolled (counting from the most inner loop)
11059 for the @option{-floop-unroll-and-jam}. The default value is 2.
11061 @item ipa-cp-value-list-size
11062 IPA-CP attempts to track all possible values and types passed to a function's
11063 parameter in order to propagate them and perform devirtualization.
11064 @option{ipa-cp-value-list-size} is the maximum number of values and types it
11065 stores per one formal parameter of a function.
11067 @item ipa-cp-eval-threshold
11068 IPA-CP calculates its own score of cloning profitability heuristics
11069 and performs those cloning opportunities with scores that exceed
11070 @option{ipa-cp-eval-threshold}.
11072 @item ipa-cp-recursion-penalty
11073 Percentage penalty the recursive functions will receive when they
11074 are evaluated for cloning.
11076 @item ipa-cp-single-call-penalty
11077 Percentage penalty functions containg a single call to another
11078 function will receive when they are evaluated for cloning.
11081 @item ipa-max-agg-items
11082 IPA-CP is also capable to propagate a number of scalar values passed
11083 in an aggregate. @option{ipa-max-agg-items} controls the maximum
11084 number of such values per one parameter.
11086 @item ipa-cp-loop-hint-bonus
11087 When IPA-CP determines that a cloning candidate would make the number
11088 of iterations of a loop known, it adds a bonus of
11089 @option{ipa-cp-loop-hint-bonus} to the profitability score of
11092 @item ipa-cp-array-index-hint-bonus
11093 When IPA-CP determines that a cloning candidate would make the index of
11094 an array access known, it adds a bonus of
11095 @option{ipa-cp-array-index-hint-bonus} to the profitability
11096 score of the candidate.
11098 @item ipa-max-aa-steps
11099 During its analysis of function bodies, IPA-CP employs alias analysis
11100 in order to track values pointed to by function parameters. In order
11101 not spend too much time analyzing huge functions, it gives up and
11102 consider all memory clobbered after examining
11103 @option{ipa-max-aa-steps} statements modifying memory.
11105 @item lto-partitions
11106 Specify desired number of partitions produced during WHOPR compilation.
11107 The number of partitions should exceed the number of CPUs used for compilation.
11108 The default value is 32.
11110 @item lto-minpartition
11111 Size of minimal partition for WHOPR (in estimated instructions).
11112 This prevents expenses of splitting very small programs into too many
11115 @item cxx-max-namespaces-for-diagnostic-help
11116 The maximum number of namespaces to consult for suggestions when C++
11117 name lookup fails for an identifier. The default is 1000.
11119 @item sink-frequency-threshold
11120 The maximum relative execution frequency (in percents) of the target block
11121 relative to a statement's original block to allow statement sinking of a
11122 statement. Larger numbers result in more aggressive statement sinking.
11123 The default value is 75. A small positive adjustment is applied for
11124 statements with memory operands as those are even more profitable so sink.
11126 @item max-stores-to-sink
11127 The maximum number of conditional stores paires that can be sunk. Set to 0
11128 if either vectorization (@option{-ftree-vectorize}) or if-conversion
11129 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
11131 @item allow-store-data-races
11132 Allow optimizers to introduce new data races on stores.
11133 Set to 1 to allow, otherwise to 0. This option is enabled by default
11134 at optimization level @option{-Ofast}.
11136 @item case-values-threshold
11137 The smallest number of different values for which it is best to use a
11138 jump-table instead of a tree of conditional branches. If the value is
11139 0, use the default for the machine. The default is 0.
11141 @item tree-reassoc-width
11142 Set the maximum number of instructions executed in parallel in
11143 reassociated tree. This parameter overrides target dependent
11144 heuristics used by default if has non zero value.
11146 @item sched-pressure-algorithm
11147 Choose between the two available implementations of
11148 @option{-fsched-pressure}. Algorithm 1 is the original implementation
11149 and is the more likely to prevent instructions from being reordered.
11150 Algorithm 2 was designed to be a compromise between the relatively
11151 conservative approach taken by algorithm 1 and the rather aggressive
11152 approach taken by the default scheduler. It relies more heavily on
11153 having a regular register file and accurate register pressure classes.
11154 See @file{haifa-sched.c} in the GCC sources for more details.
11156 The default choice depends on the target.
11158 @item max-slsr-cand-scan
11159 Set the maximum number of existing candidates that are considered when
11160 seeking a basis for a new straight-line strength reduction candidate.
11163 Enable buffer overflow detection for global objects. This kind
11164 of protection is enabled by default if you are using
11165 @option{-fsanitize=address} option.
11166 To disable global objects protection use @option{--param asan-globals=0}.
11169 Enable buffer overflow detection for stack objects. This kind of
11170 protection is enabled by default when using @option{-fsanitize=address}.
11171 To disable stack protection use @option{--param asan-stack=0} option.
11173 @item asan-instrument-reads
11174 Enable buffer overflow detection for memory reads. This kind of
11175 protection is enabled by default when using @option{-fsanitize=address}.
11176 To disable memory reads protection use
11177 @option{--param asan-instrument-reads=0}.
11179 @item asan-instrument-writes
11180 Enable buffer overflow detection for memory writes. This kind of
11181 protection is enabled by default when using @option{-fsanitize=address}.
11182 To disable memory writes protection use
11183 @option{--param asan-instrument-writes=0} option.
11185 @item asan-memintrin
11186 Enable detection for built-in functions. This kind of protection
11187 is enabled by default when using @option{-fsanitize=address}.
11188 To disable built-in functions protection use
11189 @option{--param asan-memintrin=0}.
11191 @item asan-use-after-return
11192 Enable detection of use-after-return. This kind of protection
11193 is enabled by default when using @option{-fsanitize=address} option.
11194 To disable use-after-return detection use
11195 @option{--param asan-use-after-return=0}.
11197 @item asan-instrumentation-with-call-threshold
11198 If number of memory accesses in function being instrumented
11199 is greater or equal to this number, use callbacks instead of inline checks.
11200 E.g. to disable inline code use
11201 @option{--param asan-instrumentation-with-call-threshold=0}.
11203 @item chkp-max-ctor-size
11204 Static constructors generated by Pointer Bounds Checker may become very
11205 large and significantly increase compile time at optimization level
11206 @option{-O1} and higher. This parameter is a maximum nubmer of statements
11207 in a single generated constructor. Default value is 5000.
11209 @item max-fsm-thread-path-insns
11210 Maximum number of instructions to copy when duplicating blocks on a
11211 finite state automaton jump thread path. The default is 100.
11213 @item max-fsm-thread-length
11214 Maximum number of basic blocks on a finite state automaton jump thread
11215 path. The default is 10.
11217 @item max-fsm-thread-paths
11218 Maximum number of new jump thread paths to create for a finite state
11219 automaton. The default is 50.
11221 @item parloops-chunk-size
11222 Chunk size of omp schedule for loops parallelized by parloops. The default
11225 @item parloops-schedule
11226 Schedule type of omp schedule for loops parallelized by parloops (static,
11227 dynamic, guided, auto, runtime). The default is static.
11229 @item max-ssa-name-query-depth
11230 Maximum depth of recursion when querying properties of SSA names in things
11231 like fold routines. One level of recursion corresponds to following a
11236 @node Preprocessor Options
11237 @section Options Controlling the Preprocessor
11238 @cindex preprocessor options
11239 @cindex options, preprocessor
11241 These options control the C preprocessor, which is run on each C source
11242 file before actual compilation.
11244 If you use the @option{-E} option, nothing is done except preprocessing.
11245 Some of these options make sense only together with @option{-E} because
11246 they cause the preprocessor output to be unsuitable for actual
11250 @item -Wp,@var{option}
11252 You can use @option{-Wp,@var{option}} to bypass the compiler driver
11253 and pass @var{option} directly through to the preprocessor. If
11254 @var{option} contains commas, it is split into multiple options at the
11255 commas. However, many options are modified, translated or interpreted
11256 by the compiler driver before being passed to the preprocessor, and
11257 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
11258 interface is undocumented and subject to change, so whenever possible
11259 you should avoid using @option{-Wp} and let the driver handle the
11262 @item -Xpreprocessor @var{option}
11263 @opindex Xpreprocessor
11264 Pass @var{option} as an option to the preprocessor. You can use this to
11265 supply system-specific preprocessor options that GCC does not
11268 If you want to pass an option that takes an argument, you must use
11269 @option{-Xpreprocessor} twice, once for the option and once for the argument.
11271 @item -no-integrated-cpp
11272 @opindex no-integrated-cpp
11273 Perform preprocessing as a separate pass before compilation.
11274 By default, GCC performs preprocessing as an integrated part of
11275 input tokenization and parsing.
11276 If this option is provided, the appropriate language front end
11277 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
11278 and Objective-C, respectively) is instead invoked twice,
11279 once for preprocessing only and once for actual compilation
11280 of the preprocessed input.
11281 This option may be useful in conjunction with the @option{-B} or
11282 @option{-wrapper} options to specify an alternate preprocessor or
11283 perform additional processing of the program source between
11284 normal preprocessing and compilation.
11287 @include cppopts.texi
11289 @node Assembler Options
11290 @section Passing Options to the Assembler
11292 @c prevent bad page break with this line
11293 You can pass options to the assembler.
11296 @item -Wa,@var{option}
11298 Pass @var{option} as an option to the assembler. If @var{option}
11299 contains commas, it is split into multiple options at the commas.
11301 @item -Xassembler @var{option}
11302 @opindex Xassembler
11303 Pass @var{option} as an option to the assembler. You can use this to
11304 supply system-specific assembler options that GCC does not
11307 If you want to pass an option that takes an argument, you must use
11308 @option{-Xassembler} twice, once for the option and once for the argument.
11313 @section Options for Linking
11314 @cindex link options
11315 @cindex options, linking
11317 These options come into play when the compiler links object files into
11318 an executable output file. They are meaningless if the compiler is
11319 not doing a link step.
11323 @item @var{object-file-name}
11324 A file name that does not end in a special recognized suffix is
11325 considered to name an object file or library. (Object files are
11326 distinguished from libraries by the linker according to the file
11327 contents.) If linking is done, these object files are used as input
11336 If any of these options is used, then the linker is not run, and
11337 object file names should not be used as arguments. @xref{Overall
11341 @opindex fuse-ld=bfd
11342 Use the @command{bfd} linker instead of the default linker.
11344 @item -fuse-ld=gold
11345 @opindex fuse-ld=gold
11346 Use the @command{gold} linker instead of the default linker.
11349 @item -l@var{library}
11350 @itemx -l @var{library}
11352 Search the library named @var{library} when linking. (The second
11353 alternative with the library as a separate argument is only for
11354 POSIX compliance and is not recommended.)
11356 It makes a difference where in the command you write this option; the
11357 linker searches and processes libraries and object files in the order they
11358 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
11359 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
11360 to functions in @samp{z}, those functions may not be loaded.
11362 The linker searches a standard list of directories for the library,
11363 which is actually a file named @file{lib@var{library}.a}. The linker
11364 then uses this file as if it had been specified precisely by name.
11366 The directories searched include several standard system directories
11367 plus any that you specify with @option{-L}.
11369 Normally the files found this way are library files---archive files
11370 whose members are object files. The linker handles an archive file by
11371 scanning through it for members which define symbols that have so far
11372 been referenced but not defined. But if the file that is found is an
11373 ordinary object file, it is linked in the usual fashion. The only
11374 difference between using an @option{-l} option and specifying a file name
11375 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
11376 and searches several directories.
11380 You need this special case of the @option{-l} option in order to
11381 link an Objective-C or Objective-C++ program.
11383 @item -nostartfiles
11384 @opindex nostartfiles
11385 Do not use the standard system startup files when linking.
11386 The standard system libraries are used normally, unless @option{-nostdlib}
11387 or @option{-nodefaultlibs} is used.
11389 @item -nodefaultlibs
11390 @opindex nodefaultlibs
11391 Do not use the standard system libraries when linking.
11392 Only the libraries you specify are passed to the linker, and options
11393 specifying linkage of the system libraries, such as @option{-static-libgcc}
11394 or @option{-shared-libgcc}, are ignored.
11395 The standard startup files are used normally, unless @option{-nostartfiles}
11398 The compiler may generate calls to @code{memcmp},
11399 @code{memset}, @code{memcpy} and @code{memmove}.
11400 These entries are usually resolved by entries in
11401 libc. These entry points should be supplied through some other
11402 mechanism when this option is specified.
11406 Do not use the standard system startup files or libraries when linking.
11407 No startup files and only the libraries you specify are passed to
11408 the linker, and options specifying linkage of the system libraries, such as
11409 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
11411 The compiler may generate calls to @code{memcmp}, @code{memset},
11412 @code{memcpy} and @code{memmove}.
11413 These entries are usually resolved by entries in
11414 libc. These entry points should be supplied through some other
11415 mechanism when this option is specified.
11417 @cindex @option{-lgcc}, use with @option{-nostdlib}
11418 @cindex @option{-nostdlib} and unresolved references
11419 @cindex unresolved references and @option{-nostdlib}
11420 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
11421 @cindex @option{-nodefaultlibs} and unresolved references
11422 @cindex unresolved references and @option{-nodefaultlibs}
11423 One of the standard libraries bypassed by @option{-nostdlib} and
11424 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
11425 which GCC uses to overcome shortcomings of particular machines, or special
11426 needs for some languages.
11427 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
11428 Collection (GCC) Internals},
11429 for more discussion of @file{libgcc.a}.)
11430 In most cases, you need @file{libgcc.a} even when you want to avoid
11431 other standard libraries. In other words, when you specify @option{-nostdlib}
11432 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
11433 This ensures that you have no unresolved references to internal GCC
11434 library subroutines.
11435 (An example of such an internal subroutine is @code{__main}, used to ensure C++
11436 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
11437 GNU Compiler Collection (GCC) Internals}.)
11441 Produce a position independent executable on targets that support it.
11442 For predictable results, you must also specify the same set of options
11443 used for compilation (@option{-fpie}, @option{-fPIE},
11444 or model suboptions) when you specify this linker option.
11448 Don't produce a position independent executable.
11452 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
11453 that support it. This instructs the linker to add all symbols, not
11454 only used ones, to the dynamic symbol table. This option is needed
11455 for some uses of @code{dlopen} or to allow obtaining backtraces
11456 from within a program.
11460 Remove all symbol table and relocation information from the executable.
11464 On systems that support dynamic linking, this prevents linking with the shared
11465 libraries. On other systems, this option has no effect.
11469 Produce a shared object which can then be linked with other objects to
11470 form an executable. Not all systems support this option. For predictable
11471 results, you must also specify the same set of options used for compilation
11472 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
11473 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
11474 needs to build supplementary stub code for constructors to work. On
11475 multi-libbed systems, @samp{gcc -shared} must select the correct support
11476 libraries to link against. Failing to supply the correct flags may lead
11477 to subtle defects. Supplying them in cases where they are not necessary
11480 @item -shared-libgcc
11481 @itemx -static-libgcc
11482 @opindex shared-libgcc
11483 @opindex static-libgcc
11484 On systems that provide @file{libgcc} as a shared library, these options
11485 force the use of either the shared or static version, respectively.
11486 If no shared version of @file{libgcc} was built when the compiler was
11487 configured, these options have no effect.
11489 There are several situations in which an application should use the
11490 shared @file{libgcc} instead of the static version. The most common
11491 of these is when the application wishes to throw and catch exceptions
11492 across different shared libraries. In that case, each of the libraries
11493 as well as the application itself should use the shared @file{libgcc}.
11495 Therefore, the G++ and GCJ drivers automatically add
11496 @option{-shared-libgcc} whenever you build a shared library or a main
11497 executable, because C++ and Java programs typically use exceptions, so
11498 this is the right thing to do.
11500 If, instead, you use the GCC driver to create shared libraries, you may
11501 find that they are not always linked with the shared @file{libgcc}.
11502 If GCC finds, at its configuration time, that you have a non-GNU linker
11503 or a GNU linker that does not support option @option{--eh-frame-hdr},
11504 it links the shared version of @file{libgcc} into shared libraries
11505 by default. Otherwise, it takes advantage of the linker and optimizes
11506 away the linking with the shared version of @file{libgcc}, linking with
11507 the static version of libgcc by default. This allows exceptions to
11508 propagate through such shared libraries, without incurring relocation
11509 costs at library load time.
11511 However, if a library or main executable is supposed to throw or catch
11512 exceptions, you must link it using the G++ or GCJ driver, as appropriate
11513 for the languages used in the program, or using the option
11514 @option{-shared-libgcc}, such that it is linked with the shared
11517 @item -static-libasan
11518 @opindex static-libasan
11519 When the @option{-fsanitize=address} option is used to link a program,
11520 the GCC driver automatically links against @option{libasan}. If
11521 @file{libasan} is available as a shared library, and the @option{-static}
11522 option is not used, then this links against the shared version of
11523 @file{libasan}. The @option{-static-libasan} option directs the GCC
11524 driver to link @file{libasan} statically, without necessarily linking
11525 other libraries statically.
11527 @item -static-libtsan
11528 @opindex static-libtsan
11529 When the @option{-fsanitize=thread} option is used to link a program,
11530 the GCC driver automatically links against @option{libtsan}. If
11531 @file{libtsan} is available as a shared library, and the @option{-static}
11532 option is not used, then this links against the shared version of
11533 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
11534 driver to link @file{libtsan} statically, without necessarily linking
11535 other libraries statically.
11537 @item -static-liblsan
11538 @opindex static-liblsan
11539 When the @option{-fsanitize=leak} option is used to link a program,
11540 the GCC driver automatically links against @option{liblsan}. If
11541 @file{liblsan} is available as a shared library, and the @option{-static}
11542 option is not used, then this links against the shared version of
11543 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
11544 driver to link @file{liblsan} statically, without necessarily linking
11545 other libraries statically.
11547 @item -static-libubsan
11548 @opindex static-libubsan
11549 When the @option{-fsanitize=undefined} option is used to link a program,
11550 the GCC driver automatically links against @option{libubsan}. If
11551 @file{libubsan} is available as a shared library, and the @option{-static}
11552 option is not used, then this links against the shared version of
11553 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
11554 driver to link @file{libubsan} statically, without necessarily linking
11555 other libraries statically.
11557 @item -static-libmpx
11558 @opindex static-libmpx
11559 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
11560 used to link a program, the GCC driver automatically links against
11561 @file{libmpx}. If @file{libmpx} is available as a shared library,
11562 and the @option{-static} option is not used, then this links against
11563 the shared version of @file{libmpx}. The @option{-static-libmpx}
11564 option directs the GCC driver to link @file{libmpx} statically,
11565 without necessarily linking other libraries statically.
11567 @item -static-libmpxwrappers
11568 @opindex static-libmpxwrappers
11569 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
11570 to link a program without also using @option{-fno-chkp-use-wrappers}, the
11571 GCC driver automatically links against @file{libmpxwrappers}. If
11572 @file{libmpxwrappers} is available as a shared library, and the
11573 @option{-static} option is not used, then this links against the shared
11574 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
11575 option directs the GCC driver to link @file{libmpxwrappers} statically,
11576 without necessarily linking other libraries statically.
11578 @item -static-libstdc++
11579 @opindex static-libstdc++
11580 When the @command{g++} program is used to link a C++ program, it
11581 normally automatically links against @option{libstdc++}. If
11582 @file{libstdc++} is available as a shared library, and the
11583 @option{-static} option is not used, then this links against the
11584 shared version of @file{libstdc++}. That is normally fine. However, it
11585 is sometimes useful to freeze the version of @file{libstdc++} used by
11586 the program without going all the way to a fully static link. The
11587 @option{-static-libstdc++} option directs the @command{g++} driver to
11588 link @file{libstdc++} statically, without necessarily linking other
11589 libraries statically.
11593 Bind references to global symbols when building a shared object. Warn
11594 about any unresolved references (unless overridden by the link editor
11595 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
11598 @item -T @var{script}
11600 @cindex linker script
11601 Use @var{script} as the linker script. This option is supported by most
11602 systems using the GNU linker. On some targets, such as bare-board
11603 targets without an operating system, the @option{-T} option may be required
11604 when linking to avoid references to undefined symbols.
11606 @item -Xlinker @var{option}
11608 Pass @var{option} as an option to the linker. You can use this to
11609 supply system-specific linker options that GCC does not recognize.
11611 If you want to pass an option that takes a separate argument, you must use
11612 @option{-Xlinker} twice, once for the option and once for the argument.
11613 For example, to pass @option{-assert definitions}, you must write
11614 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
11615 @option{-Xlinker "-assert definitions"}, because this passes the entire
11616 string as a single argument, which is not what the linker expects.
11618 When using the GNU linker, it is usually more convenient to pass
11619 arguments to linker options using the @option{@var{option}=@var{value}}
11620 syntax than as separate arguments. For example, you can specify
11621 @option{-Xlinker -Map=output.map} rather than
11622 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
11623 this syntax for command-line options.
11625 @item -Wl,@var{option}
11627 Pass @var{option} as an option to the linker. If @var{option} contains
11628 commas, it is split into multiple options at the commas. You can use this
11629 syntax to pass an argument to the option.
11630 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
11631 linker. When using the GNU linker, you can also get the same effect with
11632 @option{-Wl,-Map=output.map}.
11634 @item -u @var{symbol}
11636 Pretend the symbol @var{symbol} is undefined, to force linking of
11637 library modules to define it. You can use @option{-u} multiple times with
11638 different symbols to force loading of additional library modules.
11640 @item -z @var{keyword}
11642 @option{-z} is passed directly on to the linker along with the keyword
11643 @var{keyword}. See the section in the documentation of your linker for
11644 permitted values and their meanings.
11647 @node Directory Options
11648 @section Options for Directory Search
11649 @cindex directory options
11650 @cindex options, directory search
11651 @cindex search path
11653 These options specify directories to search for header files, for
11654 libraries and for parts of the compiler:
11659 Add the directory @var{dir} to the head of the list of directories to be
11660 searched for header files. This can be used to override a system header
11661 file, substituting your own version, since these directories are
11662 searched before the system header file directories. However, you should
11663 not use this option to add directories that contain vendor-supplied
11664 system header files (use @option{-isystem} for that). If you use more than
11665 one @option{-I} option, the directories are scanned in left-to-right
11666 order; the standard system directories come after.
11668 If a standard system include directory, or a directory specified with
11669 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
11670 option is ignored. The directory is still searched but as a
11671 system directory at its normal position in the system include chain.
11672 This is to ensure that GCC's procedure to fix buggy system headers and
11673 the ordering for the @code{include_next} directive are not inadvertently changed.
11674 If you really need to change the search order for system directories,
11675 use the @option{-nostdinc} and/or @option{-isystem} options.
11677 @item -iplugindir=@var{dir}
11678 @opindex iplugindir=
11679 Set the directory to search for plugins that are passed
11680 by @option{-fplugin=@var{name}} instead of
11681 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
11682 to be used by the user, but only passed by the driver.
11684 @item -iquote@var{dir}
11686 Add the directory @var{dir} to the head of the list of directories to
11687 be searched for header files only for the case of @code{#include
11688 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
11689 otherwise just like @option{-I}.
11693 Add directory @var{dir} to the list of directories to be searched
11696 @item -B@var{prefix}
11698 This option specifies where to find the executables, libraries,
11699 include files, and data files of the compiler itself.
11701 The compiler driver program runs one or more of the subprograms
11702 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11703 @var{prefix} as a prefix for each program it tries to run, both with and
11704 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
11706 For each subprogram to be run, the compiler driver first tries the
11707 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11708 is not specified, the driver tries two standard prefixes,
11709 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11710 those results in a file name that is found, the unmodified program
11711 name is searched for using the directories specified in your
11712 @env{PATH} environment variable.
11714 The compiler checks to see if the path provided by @option{-B}
11715 refers to a directory, and if necessary it adds a directory
11716 separator character at the end of the path.
11718 @option{-B} prefixes that effectively specify directory names also apply
11719 to libraries in the linker, because the compiler translates these
11720 options into @option{-L} options for the linker. They also apply to
11721 include files in the preprocessor, because the compiler translates these
11722 options into @option{-isystem} options for the preprocessor. In this case,
11723 the compiler appends @samp{include} to the prefix.
11725 The runtime support file @file{libgcc.a} can also be searched for using
11726 the @option{-B} prefix, if needed. If it is not found there, the two
11727 standard prefixes above are tried, and that is all. The file is left
11728 out of the link if it is not found by those means.
11730 Another way to specify a prefix much like the @option{-B} prefix is to use
11731 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11734 As a special kludge, if the path provided by @option{-B} is
11735 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11736 9, then it is replaced by @file{[dir/]include}. This is to help
11737 with boot-strapping the compiler.
11739 @item -specs=@var{file}
11741 Process @var{file} after the compiler reads in the standard @file{specs}
11742 file, in order to override the defaults which the @command{gcc} driver
11743 program uses when determining what switches to pass to @command{cc1},
11744 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
11745 @option{-specs=@var{file}} can be specified on the command line, and they
11746 are processed in order, from left to right.
11748 @item --sysroot=@var{dir}
11750 Use @var{dir} as the logical root directory for headers and libraries.
11751 For example, if the compiler normally searches for headers in
11752 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11753 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11755 If you use both this option and the @option{-isysroot} option, then
11756 the @option{--sysroot} option applies to libraries, but the
11757 @option{-isysroot} option applies to header files.
11759 The GNU linker (beginning with version 2.16) has the necessary support
11760 for this option. If your linker does not support this option, the
11761 header file aspect of @option{--sysroot} still works, but the
11762 library aspect does not.
11764 @item --no-sysroot-suffix
11765 @opindex no-sysroot-suffix
11766 For some targets, a suffix is added to the root directory specified
11767 with @option{--sysroot}, depending on the other options used, so that
11768 headers may for example be found in
11769 @file{@var{dir}/@var{suffix}/usr/include} instead of
11770 @file{@var{dir}/usr/include}. This option disables the addition of
11775 This option has been deprecated. Please use @option{-iquote} instead for
11776 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11778 Any directories you specify with @option{-I} options before the @option{-I-}
11779 option are searched only for the case of @code{#include "@var{file}"};
11780 they are not searched for @code{#include <@var{file}>}.
11782 If additional directories are specified with @option{-I} options after
11783 the @option{-I-} option, these directories are searched for all @code{#include}
11784 directives. (Ordinarily @emph{all} @option{-I} directories are used
11787 In addition, the @option{-I-} option inhibits the use of the current
11788 directory (where the current input file came from) as the first search
11789 directory for @code{#include "@var{file}"}. There is no way to
11790 override this effect of @option{-I-}. With @option{-I.} you can specify
11791 searching the directory that is current when the compiler is
11792 invoked. That is not exactly the same as what the preprocessor does
11793 by default, but it is often satisfactory.
11795 @option{-I-} does not inhibit the use of the standard system directories
11796 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11803 @section Specifying Subprocesses and the Switches to Pass to Them
11806 @command{gcc} is a driver program. It performs its job by invoking a
11807 sequence of other programs to do the work of compiling, assembling and
11808 linking. GCC interprets its command-line parameters and uses these to
11809 deduce which programs it should invoke, and which command-line options
11810 it ought to place on their command lines. This behavior is controlled
11811 by @dfn{spec strings}. In most cases there is one spec string for each
11812 program that GCC can invoke, but a few programs have multiple spec
11813 strings to control their behavior. The spec strings built into GCC can
11814 be overridden by using the @option{-specs=} command-line switch to specify
11817 @dfn{Spec files} are plaintext files that are used to construct spec
11818 strings. They consist of a sequence of directives separated by blank
11819 lines. The type of directive is determined by the first non-whitespace
11820 character on the line, which can be one of the following:
11823 @item %@var{command}
11824 Issues a @var{command} to the spec file processor. The commands that can
11828 @item %include <@var{file}>
11829 @cindex @code{%include}
11830 Search for @var{file} and insert its text at the current point in the
11833 @item %include_noerr <@var{file}>
11834 @cindex @code{%include_noerr}
11835 Just like @samp{%include}, but do not generate an error message if the include
11836 file cannot be found.
11838 @item %rename @var{old_name} @var{new_name}
11839 @cindex @code{%rename}
11840 Rename the spec string @var{old_name} to @var{new_name}.
11844 @item *[@var{spec_name}]:
11845 This tells the compiler to create, override or delete the named spec
11846 string. All lines after this directive up to the next directive or
11847 blank line are considered to be the text for the spec string. If this
11848 results in an empty string then the spec is deleted. (Or, if the
11849 spec did not exist, then nothing happens.) Otherwise, if the spec
11850 does not currently exist a new spec is created. If the spec does
11851 exist then its contents are overridden by the text of this
11852 directive, unless the first character of that text is the @samp{+}
11853 character, in which case the text is appended to the spec.
11855 @item [@var{suffix}]:
11856 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
11857 and up to the next directive or blank line are considered to make up the
11858 spec string for the indicated suffix. When the compiler encounters an
11859 input file with the named suffix, it processes the spec string in
11860 order to work out how to compile that file. For example:
11864 z-compile -input %i
11867 This says that any input file whose name ends in @samp{.ZZ} should be
11868 passed to the program @samp{z-compile}, which should be invoked with the
11869 command-line switch @option{-input} and with the result of performing the
11870 @samp{%i} substitution. (See below.)
11872 As an alternative to providing a spec string, the text following a
11873 suffix directive can be one of the following:
11876 @item @@@var{language}
11877 This says that the suffix is an alias for a known @var{language}. This is
11878 similar to using the @option{-x} command-line switch to GCC to specify a
11879 language explicitly. For example:
11886 Says that .ZZ files are, in fact, C++ source files.
11889 This causes an error messages saying:
11892 @var{name} compiler not installed on this system.
11896 GCC already has an extensive list of suffixes built into it.
11897 This directive adds an entry to the end of the list of suffixes, but
11898 since the list is searched from the end backwards, it is effectively
11899 possible to override earlier entries using this technique.
11903 GCC has the following spec strings built into it. Spec files can
11904 override these strings or create their own. Note that individual
11905 targets can also add their own spec strings to this list.
11908 asm Options to pass to the assembler
11909 asm_final Options to pass to the assembler post-processor
11910 cpp Options to pass to the C preprocessor
11911 cc1 Options to pass to the C compiler
11912 cc1plus Options to pass to the C++ compiler
11913 endfile Object files to include at the end of the link
11914 link Options to pass to the linker
11915 lib Libraries to include on the command line to the linker
11916 libgcc Decides which GCC support library to pass to the linker
11917 linker Sets the name of the linker
11918 predefines Defines to be passed to the C preprocessor
11919 signed_char Defines to pass to CPP to say whether @code{char} is signed
11921 startfile Object files to include at the start of the link
11924 Here is a small example of a spec file:
11927 %rename lib old_lib
11930 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
11933 This example renames the spec called @samp{lib} to @samp{old_lib} and
11934 then overrides the previous definition of @samp{lib} with a new one.
11935 The new definition adds in some extra command-line options before
11936 including the text of the old definition.
11938 @dfn{Spec strings} are a list of command-line options to be passed to their
11939 corresponding program. In addition, the spec strings can contain
11940 @samp{%}-prefixed sequences to substitute variable text or to
11941 conditionally insert text into the command line. Using these constructs
11942 it is possible to generate quite complex command lines.
11944 Here is a table of all defined @samp{%}-sequences for spec
11945 strings. Note that spaces are not generated automatically around the
11946 results of expanding these sequences. Therefore you can concatenate them
11947 together or combine them with constant text in a single argument.
11951 Substitute one @samp{%} into the program name or argument.
11954 Substitute the name of the input file being processed.
11957 Substitute the basename of the input file being processed.
11958 This is the substring up to (and not including) the last period
11959 and not including the directory.
11962 This is the same as @samp{%b}, but include the file suffix (text after
11966 Marks the argument containing or following the @samp{%d} as a
11967 temporary file name, so that that file is deleted if GCC exits
11968 successfully. Unlike @samp{%g}, this contributes no text to the
11971 @item %g@var{suffix}
11972 Substitute a file name that has suffix @var{suffix} and is chosen
11973 once per compilation, and mark the argument in the same way as
11974 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
11975 name is now chosen in a way that is hard to predict even when previously
11976 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
11977 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
11978 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
11979 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
11980 was simply substituted with a file name chosen once per compilation,
11981 without regard to any appended suffix (which was therefore treated
11982 just like ordinary text), making such attacks more likely to succeed.
11984 @item %u@var{suffix}
11985 Like @samp{%g}, but generates a new temporary file name
11986 each time it appears instead of once per compilation.
11988 @item %U@var{suffix}
11989 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
11990 new one if there is no such last file name. In the absence of any
11991 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
11992 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
11993 involves the generation of two distinct file names, one
11994 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
11995 simply substituted with a file name chosen for the previous @samp{%u},
11996 without regard to any appended suffix.
11998 @item %j@var{suffix}
11999 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
12000 writable, and if @option{-save-temps} is not used;
12001 otherwise, substitute the name
12002 of a temporary file, just like @samp{%u}. This temporary file is not
12003 meant for communication between processes, but rather as a junk
12004 disposal mechanism.
12006 @item %|@var{suffix}
12007 @itemx %m@var{suffix}
12008 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
12009 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
12010 all. These are the two most common ways to instruct a program that it
12011 should read from standard input or write to standard output. If you
12012 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
12013 construct: see for example @file{f/lang-specs.h}.
12015 @item %.@var{SUFFIX}
12016 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
12017 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
12018 terminated by the next space or %.
12021 Marks the argument containing or following the @samp{%w} as the
12022 designated output file of this compilation. This puts the argument
12023 into the sequence of arguments that @samp{%o} substitutes.
12026 Substitutes the names of all the output files, with spaces
12027 automatically placed around them. You should write spaces
12028 around the @samp{%o} as well or the results are undefined.
12029 @samp{%o} is for use in the specs for running the linker.
12030 Input files whose names have no recognized suffix are not compiled
12031 at all, but they are included among the output files, so they are
12035 Substitutes the suffix for object files. Note that this is
12036 handled specially when it immediately follows @samp{%g, %u, or %U},
12037 because of the need for those to form complete file names. The
12038 handling is such that @samp{%O} is treated exactly as if it had already
12039 been substituted, except that @samp{%g, %u, and %U} do not currently
12040 support additional @var{suffix} characters following @samp{%O} as they do
12041 following, for example, @samp{.o}.
12044 Substitutes the standard macro predefinitions for the
12045 current target machine. Use this when running @command{cpp}.
12048 Like @samp{%p}, but puts @samp{__} before and after the name of each
12049 predefined macro, except for macros that start with @samp{__} or with
12050 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
12054 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
12055 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
12056 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
12057 and @option{-imultilib} as necessary.
12060 Current argument is the name of a library or startup file of some sort.
12061 Search for that file in a standard list of directories and substitute
12062 the full name found. The current working directory is included in the
12063 list of directories scanned.
12066 Current argument is the name of a linker script. Search for that file
12067 in the current list of directories to scan for libraries. If the file
12068 is located insert a @option{--script} option into the command line
12069 followed by the full path name found. If the file is not found then
12070 generate an error message. Note: the current working directory is not
12074 Print @var{str} as an error message. @var{str} is terminated by a newline.
12075 Use this when inconsistent options are detected.
12077 @item %(@var{name})
12078 Substitute the contents of spec string @var{name} at this point.
12080 @item %x@{@var{option}@}
12081 Accumulate an option for @samp{%X}.
12084 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
12088 Output the accumulated assembler options specified by @option{-Wa}.
12091 Output the accumulated preprocessor options specified by @option{-Wp}.
12094 Process the @code{asm} spec. This is used to compute the
12095 switches to be passed to the assembler.
12098 Process the @code{asm_final} spec. This is a spec string for
12099 passing switches to an assembler post-processor, if such a program is
12103 Process the @code{link} spec. This is the spec for computing the
12104 command line passed to the linker. Typically it makes use of the
12105 @samp{%L %G %S %D and %E} sequences.
12108 Dump out a @option{-L} option for each directory that GCC believes might
12109 contain startup files. If the target supports multilibs then the
12110 current multilib directory is prepended to each of these paths.
12113 Process the @code{lib} spec. This is a spec string for deciding which
12114 libraries are included on the command line to the linker.
12117 Process the @code{libgcc} spec. This is a spec string for deciding
12118 which GCC support library is included on the command line to the linker.
12121 Process the @code{startfile} spec. This is a spec for deciding which
12122 object files are the first ones passed to the linker. Typically
12123 this might be a file named @file{crt0.o}.
12126 Process the @code{endfile} spec. This is a spec string that specifies
12127 the last object files that are passed to the linker.
12130 Process the @code{cpp} spec. This is used to construct the arguments
12131 to be passed to the C preprocessor.
12134 Process the @code{cc1} spec. This is used to construct the options to be
12135 passed to the actual C compiler (@command{cc1}).
12138 Process the @code{cc1plus} spec. This is used to construct the options to be
12139 passed to the actual C++ compiler (@command{cc1plus}).
12142 Substitute the variable part of a matched option. See below.
12143 Note that each comma in the substituted string is replaced by
12147 Remove all occurrences of @code{-S} from the command line. Note---this
12148 command is position dependent. @samp{%} commands in the spec string
12149 before this one see @code{-S}, @samp{%} commands in the spec string
12150 after this one do not.
12152 @item %:@var{function}(@var{args})
12153 Call the named function @var{function}, passing it @var{args}.
12154 @var{args} is first processed as a nested spec string, then split
12155 into an argument vector in the usual fashion. The function returns
12156 a string which is processed as if it had appeared literally as part
12157 of the current spec.
12159 The following built-in spec functions are provided:
12162 @item @code{getenv}
12163 The @code{getenv} spec function takes two arguments: an environment
12164 variable name and a string. If the environment variable is not
12165 defined, a fatal error is issued. Otherwise, the return value is the
12166 value of the environment variable concatenated with the string. For
12167 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
12170 %:getenv(TOPDIR /include)
12173 expands to @file{/path/to/top/include}.
12175 @item @code{if-exists}
12176 The @code{if-exists} spec function takes one argument, an absolute
12177 pathname to a file. If the file exists, @code{if-exists} returns the
12178 pathname. Here is a small example of its usage:
12182 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
12185 @item @code{if-exists-else}
12186 The @code{if-exists-else} spec function is similar to the @code{if-exists}
12187 spec function, except that it takes two arguments. The first argument is
12188 an absolute pathname to a file. If the file exists, @code{if-exists-else}
12189 returns the pathname. If it does not exist, it returns the second argument.
12190 This way, @code{if-exists-else} can be used to select one file or another,
12191 based on the existence of the first. Here is a small example of its usage:
12195 crt0%O%s %:if-exists(crti%O%s) \
12196 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
12199 @item @code{replace-outfile}
12200 The @code{replace-outfile} spec function takes two arguments. It looks for the
12201 first argument in the outfiles array and replaces it with the second argument. Here
12202 is a small example of its usage:
12205 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
12208 @item @code{remove-outfile}
12209 The @code{remove-outfile} spec function takes one argument. It looks for the
12210 first argument in the outfiles array and removes it. Here is a small example
12214 %:remove-outfile(-lm)
12217 @item @code{pass-through-libs}
12218 The @code{pass-through-libs} spec function takes any number of arguments. It
12219 finds any @option{-l} options and any non-options ending in @file{.a} (which it
12220 assumes are the names of linker input library archive files) and returns a
12221 result containing all the found arguments each prepended by
12222 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
12223 intended to be passed to the LTO linker plugin.
12226 %:pass-through-libs(%G %L %G)
12229 @item @code{print-asm-header}
12230 The @code{print-asm-header} function takes no arguments and simply
12231 prints a banner like:
12237 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
12240 It is used to separate compiler options from assembler options
12241 in the @option{--target-help} output.
12244 @item %@{@code{S}@}
12245 Substitutes the @code{-S} switch, if that switch is given to GCC@.
12246 If that switch is not specified, this substitutes nothing. Note that
12247 the leading dash is omitted when specifying this option, and it is
12248 automatically inserted if the substitution is performed. Thus the spec
12249 string @samp{%@{foo@}} matches the command-line option @option{-foo}
12250 and outputs the command-line option @option{-foo}.
12252 @item %W@{@code{S}@}
12253 Like %@{@code{S}@} but mark last argument supplied within as a file to be
12254 deleted on failure.
12256 @item %@{@code{S}*@}
12257 Substitutes all the switches specified to GCC whose names start
12258 with @code{-S}, but which also take an argument. This is used for
12259 switches like @option{-o}, @option{-D}, @option{-I}, etc.
12260 GCC considers @option{-o foo} as being
12261 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
12262 text, including the space. Thus two arguments are generated.
12264 @item %@{@code{S}*&@code{T}*@}
12265 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
12266 (the order of @code{S} and @code{T} in the spec is not significant).
12267 There can be any number of ampersand-separated variables; for each the
12268 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
12270 @item %@{@code{S}:@code{X}@}
12271 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
12273 @item %@{!@code{S}:@code{X}@}
12274 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
12276 @item %@{@code{S}*:@code{X}@}
12277 Substitutes @code{X} if one or more switches whose names start with
12278 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
12279 once, no matter how many such switches appeared. However, if @code{%*}
12280 appears somewhere in @code{X}, then @code{X} is substituted once
12281 for each matching switch, with the @code{%*} replaced by the part of
12282 that switch matching the @code{*}.
12284 If @code{%*} appears as the last part of a spec sequence then a space
12285 is added after the end of the last substitution. If there is more
12286 text in the sequence, however, then a space is not generated. This
12287 allows the @code{%*} substitution to be used as part of a larger
12288 string. For example, a spec string like this:
12291 %@{mcu=*:--script=%*/memory.ld@}
12295 when matching an option like @option{-mcu=newchip} produces:
12298 --script=newchip/memory.ld
12301 @item %@{.@code{S}:@code{X}@}
12302 Substitutes @code{X}, if processing a file with suffix @code{S}.
12304 @item %@{!.@code{S}:@code{X}@}
12305 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
12307 @item %@{,@code{S}:@code{X}@}
12308 Substitutes @code{X}, if processing a file for language @code{S}.
12310 @item %@{!,@code{S}:@code{X}@}
12311 Substitutes @code{X}, if not processing a file for language @code{S}.
12313 @item %@{@code{S}|@code{P}:@code{X}@}
12314 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
12315 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
12316 @code{*} sequences as well, although they have a stronger binding than
12317 the @samp{|}. If @code{%*} appears in @code{X}, all of the
12318 alternatives must be starred, and only the first matching alternative
12321 For example, a spec string like this:
12324 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
12328 outputs the following command-line options from the following input
12329 command-line options:
12334 -d fred.c -foo -baz -boggle
12335 -d jim.d -bar -baz -boggle
12338 @item %@{S:X; T:Y; :D@}
12340 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
12341 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
12342 be as many clauses as you need. This may be combined with @code{.},
12343 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
12348 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
12349 construct may contain other nested @samp{%} constructs or spaces, or
12350 even newlines. They are processed as usual, as described above.
12351 Trailing white space in @code{X} is ignored. White space may also
12352 appear anywhere on the left side of the colon in these constructs,
12353 except between @code{.} or @code{*} and the corresponding word.
12355 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
12356 handled specifically in these constructs. If another value of
12357 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
12358 @option{-W} switch is found later in the command line, the earlier
12359 switch value is ignored, except with @{@code{S}*@} where @code{S} is
12360 just one letter, which passes all matching options.
12362 The character @samp{|} at the beginning of the predicate text is used to
12363 indicate that a command should be piped to the following command, but
12364 only if @option{-pipe} is specified.
12366 It is built into GCC which switches take arguments and which do not.
12367 (You might think it would be useful to generalize this to allow each
12368 compiler's spec to say which switches take arguments. But this cannot
12369 be done in a consistent fashion. GCC cannot even decide which input
12370 files have been specified without knowing which switches take arguments,
12371 and it must know which input files to compile in order to tell which
12374 GCC also knows implicitly that arguments starting in @option{-l} are to be
12375 treated as compiler output files, and passed to the linker in their
12376 proper position among the other output files.
12378 @c man begin OPTIONS
12380 @node Target Options
12381 @section Specifying Target Machine and Compiler Version
12382 @cindex target options
12383 @cindex cross compiling
12384 @cindex specifying machine version
12385 @cindex specifying compiler version and target machine
12386 @cindex compiler version, specifying
12387 @cindex target machine, specifying
12389 The usual way to run GCC is to run the executable called @command{gcc}, or
12390 @command{@var{machine}-gcc} when cross-compiling, or
12391 @command{@var{machine}-gcc-@var{version}} to run a version other than the
12392 one that was installed last.
12394 @node Submodel Options
12395 @section Hardware Models and Configurations
12396 @cindex submodel options
12397 @cindex specifying hardware config
12398 @cindex hardware models and configurations, specifying
12399 @cindex machine dependent options
12401 Each target machine types can have its own
12402 special options, starting with @samp{-m}, to choose among various
12403 hardware models or configurations---for example, 68010 vs 68020,
12404 floating coprocessor or none. A single installed version of the
12405 compiler can compile for any model or configuration, according to the
12408 Some configurations of the compiler also support additional special
12409 options, usually for compatibility with other compilers on the same
12412 @c This list is ordered alphanumerically by subsection name.
12413 @c It should be the same order and spelling as these options are listed
12414 @c in Machine Dependent Options
12417 * AArch64 Options::
12418 * Adapteva Epiphany Options::
12422 * Blackfin Options::
12427 * DEC Alpha Options::
12431 * GNU/Linux Options::
12441 * MicroBlaze Options::
12444 * MN10300 Options::
12448 * Nios II Options::
12449 * Nvidia PTX Options::
12451 * picoChip Options::
12452 * PowerPC Options::
12454 * RS/6000 and PowerPC Options::
12456 * S/390 and zSeries Options::
12459 * Solaris 2 Options::
12462 * System V Options::
12463 * TILE-Gx Options::
12464 * TILEPro Options::
12469 * VxWorks Options::
12471 * x86 Windows Options::
12472 * Xstormy16 Options::
12474 * zSeries Options::
12477 @node AArch64 Options
12478 @subsection AArch64 Options
12479 @cindex AArch64 Options
12481 These options are defined for AArch64 implementations:
12485 @item -mabi=@var{name}
12487 Generate code for the specified data model. Permissible values
12488 are @samp{ilp32} for SysV-like data model where int, long int and pointer
12489 are 32-bit, and @samp{lp64} for SysV-like data model where int is 32-bit,
12490 but long int and pointer are 64-bit.
12492 The default depends on the specific target configuration. Note that
12493 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12494 entire program with the same ABI, and link with a compatible set of libraries.
12497 @opindex mbig-endian
12498 Generate big-endian code. This is the default when GCC is configured for an
12499 @samp{aarch64_be-*-*} target.
12501 @item -mgeneral-regs-only
12502 @opindex mgeneral-regs-only
12503 Generate code which uses only the general-purpose registers. This will prevent
12504 the compiler from using floating-point and Advanced SIMD registers but will not
12505 impose any restrictions on the assembler.
12507 @item -mlittle-endian
12508 @opindex mlittle-endian
12509 Generate little-endian code. This is the default when GCC is configured for an
12510 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
12512 @item -mcmodel=tiny
12513 @opindex mcmodel=tiny
12514 Generate code for the tiny code model. The program and its statically defined
12515 symbols must be within 1GB of each other. Pointers are 64 bits. Programs can
12516 be statically or dynamically linked. This model is not fully implemented and
12517 mostly treated as @samp{small}.
12519 @item -mcmodel=small
12520 @opindex mcmodel=small
12521 Generate code for the small code model. The program and its statically defined
12522 symbols must be within 4GB of each other. Pointers are 64 bits. Programs can
12523 be statically or dynamically linked. This is the default code model.
12525 @item -mcmodel=large
12526 @opindex mcmodel=large
12527 Generate code for the large code model. This makes no assumptions about
12528 addresses and sizes of sections. Pointers are 64 bits. Programs can be
12529 statically linked only.
12531 @item -mstrict-align
12532 @opindex mstrict-align
12533 Do not assume that unaligned memory references are handled by the system.
12535 @item -momit-leaf-frame-pointer
12536 @itemx -mno-omit-leaf-frame-pointer
12537 @opindex momit-leaf-frame-pointer
12538 @opindex mno-omit-leaf-frame-pointer
12539 Omit or keep the frame pointer in leaf functions. The former behaviour is the
12542 @item -mtls-dialect=desc
12543 @opindex mtls-dialect=desc
12544 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
12545 of TLS variables. This is the default.
12547 @item -mtls-dialect=traditional
12548 @opindex mtls-dialect=traditional
12549 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
12552 @item -mtls-size=@var{size}
12554 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
12555 This option depends on binutils higher than 2.25.
12557 @item -mfix-cortex-a53-835769
12558 @itemx -mno-fix-cortex-a53-835769
12559 @opindex mfix-cortex-a53-835769
12560 @opindex mno-fix-cortex-a53-835769
12561 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
12562 This involves inserting a NOP instruction between memory instructions and
12563 64-bit integer multiply-accumulate instructions.
12565 @item -mfix-cortex-a53-843419
12566 @itemx -mno-fix-cortex-a53-843419
12567 @opindex mfix-cortex-a53-843419
12568 @opindex mno-fix-cortex-a53-843419
12569 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
12570 This erratum workaround is made at link time and this will only pass the
12571 corresponding flag to the linker.
12573 @item -mlow-precision-recip-sqrt
12574 @item -mno-low-precision-recip-sqrt
12575 @opindex -mlow-precision-recip-sqrt
12576 @opindex -mno-low-precision-recip-sqrt
12577 The square root estimate uses two steps instead of three for double-precision,
12578 and one step instead of two for single-precision.
12579 Thus reducing latency and precision.
12580 This is only relevant if @option{-ffast-math} activates
12581 reciprocal square root estimate instructions.
12582 Which in turn depends on the target processor.
12584 @item -march=@var{name}
12586 Specify the name of the target architecture and, optionally, one or
12587 more feature modifiers. This option has the form
12588 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
12590 The permissible values for @var{arch} are @samp{armv8-a},
12591 @samp{armv8.1-a} or @var{native}.
12593 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
12594 support for the ARMv8.1 architecture extension. In particular, it
12595 enables the @samp{+crc} and @samp{+lse} features.
12597 The value @samp{native} is available on native AArch64 GNU/Linux and
12598 causes the compiler to pick the architecture of the host system. This
12599 option has no effect if the compiler is unable to recognize the
12600 architecture of the host system,
12602 The permissible values for @var{feature} are listed in the sub-section
12603 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12604 Feature Modifiers}. Where conflicting feature modifiers are
12605 specified, the right-most feature is used.
12607 GCC uses @var{name} to determine what kind of instructions it can emit
12608 when generating assembly code. If @option{-march} is specified
12609 without either of @option{-mtune} or @option{-mcpu} also being
12610 specified, the code is tuned to perform well across a range of target
12611 processors implementing the target architecture.
12613 @item -mtune=@var{name}
12615 Specify the name of the target processor for which GCC should tune the
12616 performance of the code. Permissible values for this option are:
12617 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
12618 @samp{cortex-a72}, @samp{exynos-m1}, @samp{qdf24xx}, @samp{thunderx},
12621 Additionally, this option can specify that GCC should tune the performance
12622 of the code for a big.LITTLE system. Permissible values for this
12623 option are: @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
12625 Additionally on native AArch64 GNU/Linux systems the value
12626 @samp{native} is available. This option causes the compiler to pick
12627 the architecture of and tune the performance of the code for the
12628 processor of the host system. This option has no effect if the
12629 compiler is unable to recognize the architecture of the host system.
12631 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
12632 are specified, the code is tuned to perform well across a range
12633 of target processors.
12635 This option cannot be suffixed by feature modifiers.
12637 @item -mcpu=@var{name}
12639 Specify the name of the target processor, optionally suffixed by one
12640 or more feature modifiers. This option has the form
12641 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
12642 the permissible values for @var{cpu} are the same as those available
12643 for @option{-mtune}. The permissible values for @var{feature} are
12644 documented in the sub-section on
12645 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
12646 Feature Modifiers}. Where conflicting feature modifiers are
12647 specified, the right-most feature is used.
12649 Additionally on native AArch64 GNU/Linux systems the value
12650 @samp{native} is available. This option causes the compiler to tune
12651 the performance of the code for the processor of the host system.
12652 This option has no effect if the compiler is unable to recognize the
12653 architecture of the host system.
12655 GCC uses @var{name} to determine what kind of instructions it can emit when
12656 generating assembly code (as if by @option{-march}) and to determine
12657 the target processor for which to tune for performance (as if
12658 by @option{-mtune}). Where this option is used in conjunction
12659 with @option{-march} or @option{-mtune}, those options take precedence
12660 over the appropriate part of this option.
12662 @item -moverride=@var{string}
12664 Override tuning decisions made by the back-end in response to a
12665 @option{-mtune=} switch. The syntax, semantics, and accepted values
12666 for @var{string} in this option are not guaranteed to be consistent
12669 This option is only intended to be useful when developing GCC.
12671 @item -mpc-relative-literal-loads
12672 @opindex mpcrelativeliteralloads
12673 Enable PC relative literal loads. If this option is used, literal
12674 pools are assumed to have a range of up to 1MiB and an appropriate
12675 instruction sequence is used. This option has no impact when used
12676 with @option{-mcmodel=tiny}.
12680 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
12681 @anchor{aarch64-feature-modifiers}
12682 @cindex @option{-march} feature modifiers
12683 @cindex @option{-mcpu} feature modifiers
12684 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
12685 the following and their inverses @option{no@var{feature}}:
12689 Enable CRC extension. This is on by default for
12690 @option{-march=armv8.1-a}.
12692 Enable Crypto extension. This also enables Advanced SIMD and floating-point
12695 Enable floating-point instructions. This is on by default for all possible
12696 values for options @option{-march} and @option{-mcpu}.
12698 Enable Advanced SIMD instructions. This also enables floating-point
12699 instructions. This is on by default for all possible values for options
12700 @option{-march} and @option{-mcpu}.
12702 Enable Large System Extension instructions. This is on by default for
12703 @option{-march=armv8.1-a}.
12707 That is, @option{crypto} implies @option{simd} implies @option{fp}.
12708 Conversely, @option{nofp} (or equivalently, @option{-mgeneral-regs-only})
12709 implies @option{nosimd} implies @option{nocrypto}.
12711 @node Adapteva Epiphany Options
12712 @subsection Adapteva Epiphany Options
12714 These @samp{-m} options are defined for Adapteva Epiphany:
12717 @item -mhalf-reg-file
12718 @opindex mhalf-reg-file
12719 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
12720 That allows code to run on hardware variants that lack these registers.
12722 @item -mprefer-short-insn-regs
12723 @opindex mprefer-short-insn-regs
12724 Preferrentially allocate registers that allow short instruction generation.
12725 This can result in increased instruction count, so this may either reduce or
12726 increase overall code size.
12728 @item -mbranch-cost=@var{num}
12729 @opindex mbranch-cost
12730 Set the cost of branches to roughly @var{num} ``simple'' instructions.
12731 This cost is only a heuristic and is not guaranteed to produce
12732 consistent results across releases.
12736 Enable the generation of conditional moves.
12738 @item -mnops=@var{num}
12740 Emit @var{num} NOPs before every other generated instruction.
12742 @item -mno-soft-cmpsf
12743 @opindex mno-soft-cmpsf
12744 For single-precision floating-point comparisons, emit an @code{fsub} instruction
12745 and test the flags. This is faster than a software comparison, but can
12746 get incorrect results in the presence of NaNs, or when two different small
12747 numbers are compared such that their difference is calculated as zero.
12748 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
12749 software comparisons.
12751 @item -mstack-offset=@var{num}
12752 @opindex mstack-offset
12753 Set the offset between the top of the stack and the stack pointer.
12754 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
12755 can be used by leaf functions without stack allocation.
12756 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
12757 Note also that this option changes the ABI; compiling a program with a
12758 different stack offset than the libraries have been compiled with
12759 generally does not work.
12760 This option can be useful if you want to evaluate if a different stack
12761 offset would give you better code, but to actually use a different stack
12762 offset to build working programs, it is recommended to configure the
12763 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
12765 @item -mno-round-nearest
12766 @opindex mno-round-nearest
12767 Make the scheduler assume that the rounding mode has been set to
12768 truncating. The default is @option{-mround-nearest}.
12771 @opindex mlong-calls
12772 If not otherwise specified by an attribute, assume all calls might be beyond
12773 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
12774 function address into a register before performing a (otherwise direct) call.
12775 This is the default.
12777 @item -mshort-calls
12778 @opindex short-calls
12779 If not otherwise specified by an attribute, assume all direct calls are
12780 in the range of the @code{b} / @code{bl} instructions, so use these instructions
12781 for direct calls. The default is @option{-mlong-calls}.
12785 Assume addresses can be loaded as 16-bit unsigned values. This does not
12786 apply to function addresses for which @option{-mlong-calls} semantics
12789 @item -mfp-mode=@var{mode}
12791 Set the prevailing mode of the floating-point unit.
12792 This determines the floating-point mode that is provided and expected
12793 at function call and return time. Making this mode match the mode you
12794 predominantly need at function start can make your programs smaller and
12795 faster by avoiding unnecessary mode switches.
12797 @var{mode} can be set to one the following values:
12801 Any mode at function entry is valid, and retained or restored when
12802 the function returns, and when it calls other functions.
12803 This mode is useful for compiling libraries or other compilation units
12804 you might want to incorporate into different programs with different
12805 prevailing FPU modes, and the convenience of being able to use a single
12806 object file outweighs the size and speed overhead for any extra
12807 mode switching that might be needed, compared with what would be needed
12808 with a more specific choice of prevailing FPU mode.
12811 This is the mode used for floating-point calculations with
12812 truncating (i.e.@: round towards zero) rounding mode. That includes
12813 conversion from floating point to integer.
12815 @item round-nearest
12816 This is the mode used for floating-point calculations with
12817 round-to-nearest-or-even rounding mode.
12820 This is the mode used to perform integer calculations in the FPU, e.g.@:
12821 integer multiply, or integer multiply-and-accumulate.
12824 The default is @option{-mfp-mode=caller}
12826 @item -mnosplit-lohi
12827 @itemx -mno-postinc
12828 @itemx -mno-postmodify
12829 @opindex mnosplit-lohi
12830 @opindex mno-postinc
12831 @opindex mno-postmodify
12832 Code generation tweaks that disable, respectively, splitting of 32-bit
12833 loads, generation of post-increment addresses, and generation of
12834 post-modify addresses. The defaults are @option{msplit-lohi},
12835 @option{-mpost-inc}, and @option{-mpost-modify}.
12837 @item -mnovect-double
12838 @opindex mno-vect-double
12839 Change the preferred SIMD mode to SImode. The default is
12840 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
12842 @item -max-vect-align=@var{num}
12843 @opindex max-vect-align
12844 The maximum alignment for SIMD vector mode types.
12845 @var{num} may be 4 or 8. The default is 8.
12846 Note that this is an ABI change, even though many library function
12847 interfaces are unaffected if they don't use SIMD vector modes
12848 in places that affect size and/or alignment of relevant types.
12850 @item -msplit-vecmove-early
12851 @opindex msplit-vecmove-early
12852 Split vector moves into single word moves before reload. In theory this
12853 can give better register allocation, but so far the reverse seems to be
12854 generally the case.
12856 @item -m1reg-@var{reg}
12858 Specify a register to hold the constant @minus{}1, which makes loading small negative
12859 constants and certain bitmasks faster.
12860 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
12861 which specify use of that register as a fixed register,
12862 and @samp{none}, which means that no register is used for this
12863 purpose. The default is @option{-m1reg-none}.
12868 @subsection ARC Options
12869 @cindex ARC options
12871 The following options control the architecture variant for which code
12874 @c architecture variants
12877 @item -mbarrel-shifter
12878 @opindex mbarrel-shifter
12879 Generate instructions supported by barrel shifter. This is the default
12880 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
12882 @item -mcpu=@var{cpu}
12884 Set architecture type, register usage, and instruction scheduling
12885 parameters for @var{cpu}. There are also shortcut alias options
12886 available for backward compatibility and convenience. Supported
12887 values for @var{cpu} are
12894 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
12899 Compile for ARC601. Alias: @option{-mARC601}.
12905 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
12906 This is the default when configured with @option{--with-cpu=arc700}@.
12910 Compile for ARC EM.
12914 Compile for ARC HS.
12919 @itemx -mdpfp-compact
12920 @opindex mdpfp-compact
12921 FPX: Generate Double Precision FPX instructions, tuned for the compact
12925 @opindex mdpfp-fast
12926 FPX: Generate Double Precision FPX instructions, tuned for the fast
12929 @item -mno-dpfp-lrsr
12930 @opindex mno-dpfp-lrsr
12931 Disable LR and SR instructions from using FPX extension aux registers.
12935 Generate Extended arithmetic instructions. Currently only
12936 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
12937 supported. This is always enabled for @option{-mcpu=ARC700}.
12941 Do not generate mpy instructions for ARC700.
12945 Generate 32x16 bit multiply and mac instructions.
12949 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
12953 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
12958 @itemx -mspfp-compact
12959 @opindex mspfp-compact
12960 FPX: Generate Single Precision FPX instructions, tuned for the compact
12964 @opindex mspfp-fast
12965 FPX: Generate Single Precision FPX instructions, tuned for the fast
12970 Enable generation of ARC SIMD instructions via target-specific
12971 builtins. Only valid for @option{-mcpu=ARC700}.
12974 @opindex msoft-float
12975 This option ignored; it is provided for compatibility purposes only.
12976 Software floating point code is emitted by default, and this default
12977 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
12978 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
12979 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
12983 Generate swap instructions.
12987 This enables Locked Load/Store Conditional extension to implement
12988 atomic memopry built-in functions. Not available for ARC 6xx or ARC
12993 Enable DIV/REM instructions for ARCv2 cores.
12995 @item -mcode-density
12996 @opindex mcode-density
12997 Enable code density instructions for ARC EM, default on for ARC HS.
12999 @item -mmpy-option=@var{multo}
13000 @opindex mmpy-option
13001 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
13002 the default value. The recognized values for @var{multo} are:
13006 No multiplier available.
13010 The multiply option is set to w: 16x16 multiplier, fully pipelined.
13011 The following instructions are enabled: MPYW, and MPYUW.
13015 The multiply option is set to wlh1: 32x32 multiplier, fully
13016 pipelined (1 stage). The following instructions are additionaly
13017 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
13021 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
13022 (2 stages). The following instructions are additionaly enabled: MPY,
13023 MPYU, MPYM, MPYMU, and MPY_S.
13027 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
13028 sequential. The following instructions are additionaly enabled: MPY,
13029 MPYU, MPYM, MPYMU, and MPY_S.
13033 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
13034 sequential. The following instructions are additionaly enabled: MPY,
13035 MPYU, MPYM, MPYMU, and MPY_S.
13039 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
13040 sequential. The following instructions are additionaly enabled: MPY,
13041 MPYU, MPYM, MPYMU, and MPY_S.
13045 This option is only available for ARCv2 cores@.
13049 The following options are passed through to the assembler, and also
13050 define preprocessor macro symbols.
13052 @c Flags used by the assembler, but for which we define preprocessor
13053 @c macro symbols as well.
13056 @opindex mdsp-packa
13057 Passed down to the assembler to enable the DSP Pack A extensions.
13058 Also sets the preprocessor symbol @code{__Xdsp_packa}.
13062 Passed down to the assembler to enable the dual viterbi butterfly
13063 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
13065 @c ARC700 4.10 extension instruction
13068 Passed down to the assembler to enable the Locked Load/Store
13069 Conditional extension. Also sets the preprocessor symbol
13074 Passed down to the assembler. Also sets the preprocessor symbol
13075 @code{__Xxmac_d16}.
13079 Passed down to the assembler. Also sets the preprocessor symbol
13082 @c ARC700 4.10 extension instruction
13085 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
13086 extension instruction. Also sets the preprocessor symbol
13089 @c ARC700 4.10 extension instruction
13092 Passed down to the assembler to enable the swap byte ordering
13093 extension instruction. Also sets the preprocessor symbol
13097 @opindex mtelephony
13098 Passed down to the assembler to enable dual and single operand
13099 instructions for telephony. Also sets the preprocessor symbol
13100 @code{__Xtelephony}.
13104 Passed down to the assembler to enable the XY Memory extension. Also
13105 sets the preprocessor symbol @code{__Xxy}.
13109 The following options control how the assembly code is annotated:
13111 @c Assembly annotation options
13115 Annotate assembler instructions with estimated addresses.
13117 @item -mannotate-align
13118 @opindex mannotate-align
13119 Explain what alignment considerations lead to the decision to make an
13120 instruction short or long.
13124 The following options are passed through to the linker:
13126 @c options passed through to the linker
13130 Passed through to the linker, to specify use of the @code{arclinux} emulation.
13131 This option is enabled by default in tool chains built for
13132 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
13133 when profiling is not requested.
13135 @item -marclinux_prof
13136 @opindex marclinux_prof
13137 Passed through to the linker, to specify use of the
13138 @code{arclinux_prof} emulation. This option is enabled by default in
13139 tool chains built for @w{@code{arc-linux-uclibc}} and
13140 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
13144 The following options control the semantics of generated code:
13146 @c semantically relevant code generation options
13149 @opindex mlong-calls
13150 Generate call insns as register indirect calls, thus providing access
13151 to the full 32-bit address range.
13153 @item -mmedium-calls
13154 @opindex mmedium-calls
13155 Don't use less than 25 bit addressing range for calls, which is the
13156 offset available for an unconditional branch-and-link
13157 instruction. Conditional execution of function calls is suppressed, to
13158 allow use of the 25-bit range, rather than the 21-bit range with
13159 conditional branch-and-link. This is the default for tool chains built
13160 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13164 Do not generate sdata references. This is the default for tool chains
13165 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13169 @opindex mucb-mcount
13170 Instrument with mcount calls as used in UCB code. I.e. do the
13171 counting in the callee, not the caller. By default ARC instrumentation
13172 counts in the caller.
13174 @item -mvolatile-cache
13175 @opindex mvolatile-cache
13176 Use ordinarily cached memory accesses for volatile references. This is the
13179 @item -mno-volatile-cache
13180 @opindex mno-volatile-cache
13181 Enable cache bypass for volatile references.
13185 The following options fine tune code generation:
13186 @c code generation tuning options
13189 @opindex malign-call
13190 Do alignment optimizations for call instructions.
13192 @item -mauto-modify-reg
13193 @opindex mauto-modify-reg
13194 Enable the use of pre/post modify with register displacement.
13196 @item -mbbit-peephole
13197 @opindex mbbit-peephole
13198 Enable bbit peephole2.
13202 This option disables a target-specific pass in @file{arc_reorg} to
13203 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13204 generation driven by the combiner pass.
13206 @item -mcase-vector-pcrel
13207 @opindex mcase-vector-pcrel
13208 Use pc-relative switch case tables - this enables case table shortening.
13209 This is the default for @option{-Os}.
13211 @item -mcompact-casesi
13212 @opindex mcompact-casesi
13213 Enable compact casesi pattern.
13214 This is the default for @option{-Os}.
13216 @item -mno-cond-exec
13217 @opindex mno-cond-exec
13218 Disable ARCompact specific pass to generate conditional execution instructions.
13219 Due to delay slot scheduling and interactions between operand numbers,
13220 literal sizes, instruction lengths, and the support for conditional execution,
13221 the target-independent pass to generate conditional execution is often lacking,
13222 so the ARC port has kept a special pass around that tries to find more
13223 conditional execution generating opportunities after register allocation,
13224 branch shortening, and delay slot scheduling have been done. This pass
13225 generally, but not always, improves performance and code size, at the cost of
13226 extra compilation time, which is why there is an option to switch it off.
13227 If you have a problem with call instructions exceeding their allowable
13228 offset range because they are conditionalized, you should consider using
13229 @option{-mmedium-calls} instead.
13231 @item -mearly-cbranchsi
13232 @opindex mearly-cbranchsi
13233 Enable pre-reload use of the cbranchsi pattern.
13235 @item -mexpand-adddi
13236 @opindex mexpand-adddi
13237 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13238 @code{add.f}, @code{adc} etc.
13240 @item -mindexed-loads
13241 @opindex mindexed-loads
13242 Enable the use of indexed loads. This can be problematic because some
13243 optimizers then assume that indexed stores exist, which is not
13248 Enable Local Register Allocation. This is still experimental for ARC,
13249 so by default the compiler uses standard reload
13250 (i.e. @option{-mno-lra}).
13252 @item -mlra-priority-none
13253 @opindex mlra-priority-none
13254 Don't indicate any priority for target registers.
13256 @item -mlra-priority-compact
13257 @opindex mlra-priority-compact
13258 Indicate target register priority for r0..r3 / r12..r15.
13260 @item -mlra-priority-noncompact
13261 @opindex mlra-priority-noncompact
13262 Reduce target regsiter priority for r0..r3 / r12..r15.
13264 @item -mno-millicode
13265 @opindex mno-millicode
13266 When optimizing for size (using @option{-Os}), prologues and epilogues
13267 that have to save or restore a large number of registers are often
13268 shortened by using call to a special function in libgcc; this is
13269 referred to as a @emph{millicode} call. As these calls can pose
13270 performance issues, and/or cause linking issues when linking in a
13271 nonstandard way, this option is provided to turn off millicode call
13275 @opindex mmixed-code
13276 Tweak register allocation to help 16-bit instruction generation.
13277 This generally has the effect of decreasing the average instruction size
13278 while increasing the instruction count.
13282 Enable 'q' instruction alternatives.
13283 This is the default for @option{-Os}.
13287 Enable Rcq constraint handling - most short code generation depends on this.
13288 This is the default.
13292 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13293 This is the default.
13295 @item -msize-level=@var{level}
13296 @opindex msize-level
13297 Fine-tune size optimization with regards to instruction lengths and alignment.
13298 The recognized values for @var{level} are:
13301 No size optimization. This level is deprecated and treated like @samp{1}.
13304 Short instructions are used opportunistically.
13307 In addition, alignment of loops and of code after barriers are dropped.
13310 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13314 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13315 the behavior when this is not set is equivalent to level @samp{1}.
13317 @item -mtune=@var{cpu}
13319 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13320 by @option{-mcpu=}.
13322 Supported values for @var{cpu} are
13326 Tune for ARC600 cpu.
13329 Tune for ARC601 cpu.
13332 Tune for ARC700 cpu with standard multiplier block.
13335 Tune for ARC700 cpu with XMAC block.
13338 Tune for ARC725D cpu.
13341 Tune for ARC750D cpu.
13345 @item -mmultcost=@var{num}
13347 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13348 normal instruction.
13350 @item -munalign-prob-threshold=@var{probability}
13351 @opindex munalign-prob-threshold
13352 Set probability threshold for unaligning branches.
13353 When tuning for @samp{ARC700} and optimizing for speed, branches without
13354 filled delay slot are preferably emitted unaligned and long, unless
13355 profiling indicates that the probability for the branch to be taken
13356 is below @var{probability}. @xref{Cross-profiling}.
13357 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13361 The following options are maintained for backward compatibility, but
13362 are now deprecated and will be removed in a future release:
13364 @c Deprecated options
13372 @opindex mbig-endian
13375 Compile code for big endian targets. Use of these options is now
13376 deprecated. Users wanting big-endian code, should use the
13377 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13378 building the tool chain, for which big-endian is the default.
13380 @item -mlittle-endian
13381 @opindex mlittle-endian
13384 Compile code for little endian targets. Use of these options is now
13385 deprecated. Users wanting little-endian code should use the
13386 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13387 building the tool chain, for which little-endian is the default.
13389 @item -mbarrel_shifter
13390 @opindex mbarrel_shifter
13391 Replaced by @option{-mbarrel-shifter}.
13393 @item -mdpfp_compact
13394 @opindex mdpfp_compact
13395 Replaced by @option{-mdpfp-compact}.
13398 @opindex mdpfp_fast
13399 Replaced by @option{-mdpfp-fast}.
13402 @opindex mdsp_packa
13403 Replaced by @option{-mdsp-packa}.
13407 Replaced by @option{-mea}.
13411 Replaced by @option{-mmac-24}.
13415 Replaced by @option{-mmac-d16}.
13417 @item -mspfp_compact
13418 @opindex mspfp_compact
13419 Replaced by @option{-mspfp-compact}.
13422 @opindex mspfp_fast
13423 Replaced by @option{-mspfp-fast}.
13425 @item -mtune=@var{cpu}
13427 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
13428 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
13429 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
13431 @item -multcost=@var{num}
13433 Replaced by @option{-mmultcost}.
13438 @subsection ARM Options
13439 @cindex ARM options
13441 These @samp{-m} options are defined for the ARM port:
13444 @item -mabi=@var{name}
13446 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
13447 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
13450 @opindex mapcs-frame
13451 Generate a stack frame that is compliant with the ARM Procedure Call
13452 Standard for all functions, even if this is not strictly necessary for
13453 correct execution of the code. Specifying @option{-fomit-frame-pointer}
13454 with this option causes the stack frames not to be generated for
13455 leaf functions. The default is @option{-mno-apcs-frame}.
13456 This option is deprecated.
13460 This is a synonym for @option{-mapcs-frame} and is deprecated.
13463 @c not currently implemented
13464 @item -mapcs-stack-check
13465 @opindex mapcs-stack-check
13466 Generate code to check the amount of stack space available upon entry to
13467 every function (that actually uses some stack space). If there is
13468 insufficient space available then either the function
13469 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
13470 called, depending upon the amount of stack space required. The runtime
13471 system is required to provide these functions. The default is
13472 @option{-mno-apcs-stack-check}, since this produces smaller code.
13474 @c not currently implemented
13476 @opindex mapcs-float
13477 Pass floating-point arguments using the floating-point registers. This is
13478 one of the variants of the APCS@. This option is recommended if the
13479 target hardware has a floating-point unit or if a lot of floating-point
13480 arithmetic is going to be performed by the code. The default is
13481 @option{-mno-apcs-float}, since the size of integer-only code is
13482 slightly increased if @option{-mapcs-float} is used.
13484 @c not currently implemented
13485 @item -mapcs-reentrant
13486 @opindex mapcs-reentrant
13487 Generate reentrant, position-independent code. The default is
13488 @option{-mno-apcs-reentrant}.
13491 @item -mthumb-interwork
13492 @opindex mthumb-interwork
13493 Generate code that supports calling between the ARM and Thumb
13494 instruction sets. Without this option, on pre-v5 architectures, the
13495 two instruction sets cannot be reliably used inside one program. The
13496 default is @option{-mno-thumb-interwork}, since slightly larger code
13497 is generated when @option{-mthumb-interwork} is specified. In AAPCS
13498 configurations this option is meaningless.
13500 @item -mno-sched-prolog
13501 @opindex mno-sched-prolog
13502 Prevent the reordering of instructions in the function prologue, or the
13503 merging of those instruction with the instructions in the function's
13504 body. This means that all functions start with a recognizable set
13505 of instructions (or in fact one of a choice from a small set of
13506 different function prologues), and this information can be used to
13507 locate the start of functions inside an executable piece of code. The
13508 default is @option{-msched-prolog}.
13510 @item -mfloat-abi=@var{name}
13511 @opindex mfloat-abi
13512 Specifies which floating-point ABI to use. Permissible values
13513 are: @samp{soft}, @samp{softfp} and @samp{hard}.
13515 Specifying @samp{soft} causes GCC to generate output containing
13516 library calls for floating-point operations.
13517 @samp{softfp} allows the generation of code using hardware floating-point
13518 instructions, but still uses the soft-float calling conventions.
13519 @samp{hard} allows generation of floating-point instructions
13520 and uses FPU-specific calling conventions.
13522 The default depends on the specific target configuration. Note that
13523 the hard-float and soft-float ABIs are not link-compatible; you must
13524 compile your entire program with the same ABI, and link with a
13525 compatible set of libraries.
13527 @item -mlittle-endian
13528 @opindex mlittle-endian
13529 Generate code for a processor running in little-endian mode. This is
13530 the default for all standard configurations.
13533 @opindex mbig-endian
13534 Generate code for a processor running in big-endian mode; the default is
13535 to compile code for a little-endian processor.
13537 @item -march=@var{name}
13539 This specifies the name of the target ARM architecture. GCC uses this
13540 name to determine what kind of instructions it can emit when generating
13541 assembly code. This option can be used in conjunction with or instead
13542 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
13543 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
13544 @samp{armv5}, @samp{armv5t}, @samp{armv5e}, @samp{armv5te},
13545 @samp{armv6}, @samp{armv6j},
13546 @samp{armv6t2}, @samp{armv6z}, @samp{armv6kz}, @samp{armv6-m},
13547 @samp{armv7}, @samp{armv7-a}, @samp{armv7-r}, @samp{armv7-m}, @samp{armv7e-m},
13548 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
13549 @samp{armv8.1-a+crc}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312}.
13551 @option{-march=armv7ve} is the armv7-a architecture with virtualization
13554 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
13555 architecture together with the optional CRC32 extensions.
13557 @option{-march=native} causes the compiler to auto-detect the architecture
13558 of the build computer. At present, this feature is only supported on
13559 GNU/Linux, and not all architectures are recognized. If the auto-detect
13560 is unsuccessful the option has no effect.
13562 @item -mtune=@var{name}
13564 This option specifies the name of the target ARM processor for
13565 which GCC should tune the performance of the code.
13566 For some ARM implementations better performance can be obtained by using
13568 Permissible names are: @samp{arm2}, @samp{arm250},
13569 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
13570 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
13571 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
13572 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
13574 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
13575 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
13576 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
13577 @samp{strongarm1110},
13578 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
13579 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
13580 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
13581 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
13582 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
13583 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
13584 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
13585 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
13586 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
13587 @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57}, @samp{cortex-a72},
13589 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-m7},
13594 @samp{cortex-m0plus},
13595 @samp{cortex-m1.small-multiply},
13596 @samp{cortex-m0.small-multiply},
13597 @samp{cortex-m0plus.small-multiply},
13600 @samp{marvell-pj4},
13601 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
13602 @samp{fa526}, @samp{fa626},
13603 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
13606 Additionally, this option can specify that GCC should tune the performance
13607 of the code for a big.LITTLE system. Permissible names are:
13608 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
13609 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53}.
13611 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
13612 performance for a blend of processors within architecture @var{arch}.
13613 The aim is to generate code that run well on the current most popular
13614 processors, balancing between optimizations that benefit some CPUs in the
13615 range, and avoiding performance pitfalls of other CPUs. The effects of
13616 this option may change in future GCC versions as CPU models come and go.
13618 @option{-mtune=native} causes the compiler to auto-detect the CPU
13619 of the build computer. At present, this feature is only supported on
13620 GNU/Linux, and not all architectures are recognized. If the auto-detect is
13621 unsuccessful the option has no effect.
13623 @item -mcpu=@var{name}
13625 This specifies the name of the target ARM processor. GCC uses this name
13626 to derive the name of the target ARM architecture (as if specified
13627 by @option{-march}) and the ARM processor type for which to tune for
13628 performance (as if specified by @option{-mtune}). Where this option
13629 is used in conjunction with @option{-march} or @option{-mtune},
13630 those options take precedence over the appropriate part of this option.
13632 Permissible names for this option are the same as those for
13635 @option{-mcpu=generic-@var{arch}} is also permissible, and is
13636 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
13637 See @option{-mtune} for more information.
13639 @option{-mcpu=native} causes the compiler to auto-detect the CPU
13640 of the build computer. At present, this feature is only supported on
13641 GNU/Linux, and not all architectures are recognized. If the auto-detect
13642 is unsuccessful the option has no effect.
13644 @item -mfpu=@var{name}
13646 This specifies what floating-point hardware (or hardware emulation) is
13647 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
13648 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
13649 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
13650 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
13651 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
13652 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
13654 If @option{-msoft-float} is specified this specifies the format of
13655 floating-point values.
13657 If the selected floating-point hardware includes the NEON extension
13658 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
13659 operations are not generated by GCC's auto-vectorization pass unless
13660 @option{-funsafe-math-optimizations} is also specified. This is
13661 because NEON hardware does not fully implement the IEEE 754 standard for
13662 floating-point arithmetic (in particular denormal values are treated as
13663 zero), so the use of NEON instructions may lead to a loss of precision.
13665 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}).
13667 @item -mfp16-format=@var{name}
13668 @opindex mfp16-format
13669 Specify the format of the @code{__fp16} half-precision floating-point type.
13670 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
13671 the default is @samp{none}, in which case the @code{__fp16} type is not
13672 defined. @xref{Half-Precision}, for more information.
13674 @item -mstructure-size-boundary=@var{n}
13675 @opindex mstructure-size-boundary
13676 The sizes of all structures and unions are rounded up to a multiple
13677 of the number of bits set by this option. Permissible values are 8, 32
13678 and 64. The default value varies for different toolchains. For the COFF
13679 targeted toolchain the default value is 8. A value of 64 is only allowed
13680 if the underlying ABI supports it.
13682 Specifying a larger number can produce faster, more efficient code, but
13683 can also increase the size of the program. Different values are potentially
13684 incompatible. Code compiled with one value cannot necessarily expect to
13685 work with code or libraries compiled with another value, if they exchange
13686 information using structures or unions.
13688 @item -mabort-on-noreturn
13689 @opindex mabort-on-noreturn
13690 Generate a call to the function @code{abort} at the end of a
13691 @code{noreturn} function. It is executed if the function tries to
13695 @itemx -mno-long-calls
13696 @opindex mlong-calls
13697 @opindex mno-long-calls
13698 Tells the compiler to perform function calls by first loading the
13699 address of the function into a register and then performing a subroutine
13700 call on this register. This switch is needed if the target function
13701 lies outside of the 64-megabyte addressing range of the offset-based
13702 version of subroutine call instruction.
13704 Even if this switch is enabled, not all function calls are turned
13705 into long calls. The heuristic is that static functions, functions
13706 that have the @code{short_call} attribute, functions that are inside
13707 the scope of a @code{#pragma no_long_calls} directive, and functions whose
13708 definitions have already been compiled within the current compilation
13709 unit are not turned into long calls. The exceptions to this rule are
13710 that weak function definitions, functions with the @code{long_call}
13711 attribute or the @code{section} attribute, and functions that are within
13712 the scope of a @code{#pragma long_calls} directive are always
13713 turned into long calls.
13715 This feature is not enabled by default. Specifying
13716 @option{-mno-long-calls} restores the default behavior, as does
13717 placing the function calls within the scope of a @code{#pragma
13718 long_calls_off} directive. Note these switches have no effect on how
13719 the compiler generates code to handle function calls via function
13722 @item -msingle-pic-base
13723 @opindex msingle-pic-base
13724 Treat the register used for PIC addressing as read-only, rather than
13725 loading it in the prologue for each function. The runtime system is
13726 responsible for initializing this register with an appropriate value
13727 before execution begins.
13729 @item -mpic-register=@var{reg}
13730 @opindex mpic-register
13731 Specify the register to be used for PIC addressing.
13732 For standard PIC base case, the default is any suitable register
13733 determined by compiler. For single PIC base case, the default is
13734 @samp{R9} if target is EABI based or stack-checking is enabled,
13735 otherwise the default is @samp{R10}.
13737 @item -mpic-data-is-text-relative
13738 @opindex mpic-data-is-text-relative
13739 Assume that each data segments are relative to text segment at load time.
13740 Therefore, it permits addressing data using PC-relative operations.
13741 This option is on by default for targets other than VxWorks RTP.
13743 @item -mpoke-function-name
13744 @opindex mpoke-function-name
13745 Write the name of each function into the text section, directly
13746 preceding the function prologue. The generated code is similar to this:
13750 .ascii "arm_poke_function_name", 0
13753 .word 0xff000000 + (t1 - t0)
13754 arm_poke_function_name
13756 stmfd sp!, @{fp, ip, lr, pc@}
13760 When performing a stack backtrace, code can inspect the value of
13761 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
13762 location @code{pc - 12} and the top 8 bits are set, then we know that
13763 there is a function name embedded immediately preceding this location
13764 and has length @code{((pc[-3]) & 0xff000000)}.
13771 Select between generating code that executes in ARM and Thumb
13772 states. The default for most configurations is to generate code
13773 that executes in ARM state, but the default can be changed by
13774 configuring GCC with the @option{--with-mode=}@var{state}
13777 You can also override the ARM and Thumb mode for each function
13778 by using the @code{target("thumb")} and @code{target("arm")} function attributes
13779 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
13782 @opindex mtpcs-frame
13783 Generate a stack frame that is compliant with the Thumb Procedure Call
13784 Standard for all non-leaf functions. (A leaf function is one that does
13785 not call any other functions.) The default is @option{-mno-tpcs-frame}.
13787 @item -mtpcs-leaf-frame
13788 @opindex mtpcs-leaf-frame
13789 Generate a stack frame that is compliant with the Thumb Procedure Call
13790 Standard for all leaf functions. (A leaf function is one that does
13791 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
13793 @item -mcallee-super-interworking
13794 @opindex mcallee-super-interworking
13795 Gives all externally visible functions in the file being compiled an ARM
13796 instruction set header which switches to Thumb mode before executing the
13797 rest of the function. This allows these functions to be called from
13798 non-interworking code. This option is not valid in AAPCS configurations
13799 because interworking is enabled by default.
13801 @item -mcaller-super-interworking
13802 @opindex mcaller-super-interworking
13803 Allows calls via function pointers (including virtual functions) to
13804 execute correctly regardless of whether the target code has been
13805 compiled for interworking or not. There is a small overhead in the cost
13806 of executing a function pointer if this option is enabled. This option
13807 is not valid in AAPCS configurations because interworking is enabled
13810 @item -mtp=@var{name}
13812 Specify the access model for the thread local storage pointer. The valid
13813 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
13814 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
13815 (supported in the arm6k architecture), and @samp{auto}, which uses the
13816 best available method for the selected processor. The default setting is
13819 @item -mtls-dialect=@var{dialect}
13820 @opindex mtls-dialect
13821 Specify the dialect to use for accessing thread local storage. Two
13822 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
13823 @samp{gnu} dialect selects the original GNU scheme for supporting
13824 local and global dynamic TLS models. The @samp{gnu2} dialect
13825 selects the GNU descriptor scheme, which provides better performance
13826 for shared libraries. The GNU descriptor scheme is compatible with
13827 the original scheme, but does require new assembler, linker and
13828 library support. Initial and local exec TLS models are unaffected by
13829 this option and always use the original scheme.
13831 @item -mword-relocations
13832 @opindex mword-relocations
13833 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
13834 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
13835 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
13838 @item -mfix-cortex-m3-ldrd
13839 @opindex mfix-cortex-m3-ldrd
13840 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
13841 with overlapping destination and base registers are used. This option avoids
13842 generating these instructions. This option is enabled by default when
13843 @option{-mcpu=cortex-m3} is specified.
13845 @item -munaligned-access
13846 @itemx -mno-unaligned-access
13847 @opindex munaligned-access
13848 @opindex mno-unaligned-access
13849 Enables (or disables) reading and writing of 16- and 32- bit values
13850 from addresses that are not 16- or 32- bit aligned. By default
13851 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
13852 architectures, and enabled for all other architectures. If unaligned
13853 access is not enabled then words in packed data structures are
13854 accessed a byte at a time.
13856 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
13857 generated object file to either true or false, depending upon the
13858 setting of this option. If unaligned access is enabled then the
13859 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
13862 @item -mneon-for-64bits
13863 @opindex mneon-for-64bits
13864 Enables using Neon to handle scalar 64-bits operations. This is
13865 disabled by default since the cost of moving data from core registers
13868 @item -mslow-flash-data
13869 @opindex mslow-flash-data
13870 Assume loading data from flash is slower than fetching instruction.
13871 Therefore literal load is minimized for better performance.
13872 This option is only supported when compiling for ARMv7 M-profile and
13875 @item -masm-syntax-unified
13876 @opindex masm-syntax-unified
13877 Assume inline assembler is using unified asm syntax. The default is
13878 currently off which implies divided syntax. This option has no impact
13879 on Thumb2. However, this may change in future releases of GCC.
13880 Divided syntax should be considered deprecated.
13882 @item -mrestrict-it
13883 @opindex mrestrict-it
13884 Restricts generation of IT blocks to conform to the rules of ARMv8.
13885 IT blocks can only contain a single 16-bit instruction from a select
13886 set of instructions. This option is on by default for ARMv8 Thumb mode.
13888 @item -mprint-tune-info
13889 @opindex mprint-tune-info
13890 Print CPU tuning information as comment in assembler file. This is
13891 an option used only for regression testing of the compiler and not
13892 intended for ordinary use in compiling code. This option is disabled
13897 @subsection AVR Options
13898 @cindex AVR Options
13900 These options are defined for AVR implementations:
13903 @item -mmcu=@var{mcu}
13905 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
13907 The default for this option is@tie{}@samp{avr2}.
13909 GCC supports the following AVR devices and ISAs:
13911 @include avr-mmcu.texi
13913 @item -maccumulate-args
13914 @opindex maccumulate-args
13915 Accumulate outgoing function arguments and acquire/release the needed
13916 stack space for outgoing function arguments once in function
13917 prologue/epilogue. Without this option, outgoing arguments are pushed
13918 before calling a function and popped afterwards.
13920 Popping the arguments after the function call can be expensive on
13921 AVR so that accumulating the stack space might lead to smaller
13922 executables because arguments need not to be removed from the
13923 stack after such a function call.
13925 This option can lead to reduced code size for functions that perform
13926 several calls to functions that get their arguments on the stack like
13927 calls to printf-like functions.
13929 @item -mbranch-cost=@var{cost}
13930 @opindex mbranch-cost
13931 Set the branch costs for conditional branch instructions to
13932 @var{cost}. Reasonable values for @var{cost} are small, non-negative
13933 integers. The default branch cost is 0.
13935 @item -mcall-prologues
13936 @opindex mcall-prologues
13937 Functions prologues/epilogues are expanded as calls to appropriate
13938 subroutines. Code size is smaller.
13942 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
13943 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
13944 and @code{long long} is 4 bytes. Please note that this option does not
13945 conform to the C standards, but it results in smaller code
13948 @item -mn-flash=@var{num}
13950 Assume that the flash memory has a size of
13951 @var{num} times 64@tie{}KiB.
13953 @item -mno-interrupts
13954 @opindex mno-interrupts
13955 Generated code is not compatible with hardware interrupts.
13956 Code size is smaller.
13960 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
13961 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
13962 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
13963 the assembler's command line and the @option{--relax} option to the
13964 linker's command line.
13966 Jump relaxing is performed by the linker because jump offsets are not
13967 known before code is located. Therefore, the assembler code generated by the
13968 compiler is the same, but the instructions in the executable may
13969 differ from instructions in the assembler code.
13971 Relaxing must be turned on if linker stubs are needed, see the
13972 section on @code{EIND} and linker stubs below.
13976 Assume that the device supports the Read-Modify-Write
13977 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
13981 Treat the stack pointer register as an 8-bit register,
13982 i.e.@: assume the high byte of the stack pointer is zero.
13983 In general, you don't need to set this option by hand.
13985 This option is used internally by the compiler to select and
13986 build multilibs for architectures @code{avr2} and @code{avr25}.
13987 These architectures mix devices with and without @code{SPH}.
13988 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
13989 the compiler driver adds or removes this option from the compiler
13990 proper's command line, because the compiler then knows if the device
13991 or architecture has an 8-bit stack pointer and thus no @code{SPH}
13996 Use address register @code{X} in a way proposed by the hardware. This means
13997 that @code{X} is only used in indirect, post-increment or
13998 pre-decrement addressing.
14000 Without this option, the @code{X} register may be used in the same way
14001 as @code{Y} or @code{Z} which then is emulated by additional
14003 For example, loading a value with @code{X+const} addressing with a
14004 small non-negative @code{const < 64} to a register @var{Rn} is
14008 adiw r26, const ; X += const
14009 ld @var{Rn}, X ; @var{Rn} = *X
14010 sbiw r26, const ; X -= const
14014 @opindex mtiny-stack
14015 Only change the lower 8@tie{}bits of the stack pointer.
14018 @opindex nodevicelib
14019 Don't link against AVR-LibC's device specific library @code{libdev.a}.
14021 @item -Waddr-space-convert
14022 @opindex Waddr-space-convert
14023 Warn about conversions between address spaces in the case where the
14024 resulting address space is not contained in the incoming address space.
14027 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
14028 @cindex @code{EIND}
14029 Pointers in the implementation are 16@tie{}bits wide.
14030 The address of a function or label is represented as word address so
14031 that indirect jumps and calls can target any code address in the
14032 range of 64@tie{}Ki words.
14034 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
14035 bytes of program memory space, there is a special function register called
14036 @code{EIND} that serves as most significant part of the target address
14037 when @code{EICALL} or @code{EIJMP} instructions are used.
14039 Indirect jumps and calls on these devices are handled as follows by
14040 the compiler and are subject to some limitations:
14045 The compiler never sets @code{EIND}.
14048 The compiler uses @code{EIND} implicitely in @code{EICALL}/@code{EIJMP}
14049 instructions or might read @code{EIND} directly in order to emulate an
14050 indirect call/jump by means of a @code{RET} instruction.
14053 The compiler assumes that @code{EIND} never changes during the startup
14054 code or during the application. In particular, @code{EIND} is not
14055 saved/restored in function or interrupt service routine
14059 For indirect calls to functions and computed goto, the linker
14060 generates @emph{stubs}. Stubs are jump pads sometimes also called
14061 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
14062 The stub contains a direct jump to the desired address.
14065 Linker relaxation must be turned on so that the linker generates
14066 the stubs correctly in all situations. See the compiler option
14067 @option{-mrelax} and the linker option @option{--relax}.
14068 There are corner cases where the linker is supposed to generate stubs
14069 but aborts without relaxation and without a helpful error message.
14072 The default linker script is arranged for code with @code{EIND = 0}.
14073 If code is supposed to work for a setup with @code{EIND != 0}, a custom
14074 linker script has to be used in order to place the sections whose
14075 name start with @code{.trampolines} into the segment where @code{EIND}
14079 The startup code from libgcc never sets @code{EIND}.
14080 Notice that startup code is a blend of code from libgcc and AVR-LibC.
14081 For the impact of AVR-LibC on @code{EIND}, see the
14082 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
14085 It is legitimate for user-specific startup code to set up @code{EIND}
14086 early, for example by means of initialization code located in
14087 section @code{.init3}. Such code runs prior to general startup code
14088 that initializes RAM and calls constructors, but after the bit
14089 of startup code from AVR-LibC that sets @code{EIND} to the segment
14090 where the vector table is located.
14092 #include <avr/io.h>
14095 __attribute__((section(".init3"),naked,used,no_instrument_function))
14096 init3_set_eind (void)
14098 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
14099 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
14104 The @code{__trampolines_start} symbol is defined in the linker script.
14107 Stubs are generated automatically by the linker if
14108 the following two conditions are met:
14111 @item The address of a label is taken by means of the @code{gs} modifier
14112 (short for @emph{generate stubs}) like so:
14114 LDI r24, lo8(gs(@var{func}))
14115 LDI r25, hi8(gs(@var{func}))
14117 @item The final location of that label is in a code segment
14118 @emph{outside} the segment where the stubs are located.
14122 The compiler emits such @code{gs} modifiers for code labels in the
14123 following situations:
14125 @item Taking address of a function or code label.
14126 @item Computed goto.
14127 @item If prologue-save function is used, see @option{-mcall-prologues}
14128 command-line option.
14129 @item Switch/case dispatch tables. If you do not want such dispatch
14130 tables you can specify the @option{-fno-jump-tables} command-line option.
14131 @item C and C++ constructors/destructors called during startup/shutdown.
14132 @item If the tools hit a @code{gs()} modifier explained above.
14136 Jumping to non-symbolic addresses like so is @emph{not} supported:
14141 /* Call function at word address 0x2 */
14142 return ((int(*)(void)) 0x2)();
14146 Instead, a stub has to be set up, i.e.@: the function has to be called
14147 through a symbol (@code{func_4} in the example):
14152 extern int func_4 (void);
14154 /* Call function at byte address 0x4 */
14159 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
14160 Alternatively, @code{func_4} can be defined in the linker script.
14163 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14164 @cindex @code{RAMPD}
14165 @cindex @code{RAMPX}
14166 @cindex @code{RAMPY}
14167 @cindex @code{RAMPZ}
14168 Some AVR devices support memories larger than the 64@tie{}KiB range
14169 that can be accessed with 16-bit pointers. To access memory locations
14170 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14171 register is used as high part of the address:
14172 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14173 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14174 register, respectively, to get a wide address. Similarly,
14175 @code{RAMPD} is used together with direct addressing.
14179 The startup code initializes the @code{RAMP} special function
14180 registers with zero.
14183 If a @ref{AVR Named Address Spaces,named address space} other than
14184 generic or @code{__flash} is used, then @code{RAMPZ} is set
14185 as needed before the operation.
14188 If the device supports RAM larger than 64@tie{}KiB and the compiler
14189 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14190 is reset to zero after the operation.
14193 If the device comes with a specific @code{RAMP} register, the ISR
14194 prologue/epilogue saves/restores that SFR and initializes it with
14195 zero in case the ISR code might (implicitly) use it.
14198 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14199 If you use inline assembler to read from locations outside the
14200 16-bit address range and change one of the @code{RAMP} registers,
14201 you must reset it to zero after the access.
14205 @subsubsection AVR Built-in Macros
14207 GCC defines several built-in macros so that the user code can test
14208 for the presence or absence of features. Almost any of the following
14209 built-in macros are deduced from device capabilities and thus
14210 triggered by the @option{-mmcu=} command-line option.
14212 For even more AVR-specific built-in macros see
14213 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14218 Build-in macro that resolves to a decimal number that identifies the
14219 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14220 Possible values are:
14222 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14223 @code{4}, @code{5}, @code{51}, @code{6}
14225 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14226 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14230 @code{100}, @code{102}, @code{104},
14231 @code{105}, @code{106}, @code{107}
14233 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14234 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14235 If @var{mcu} specifies a device, this built-in macro is set
14236 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14237 defined to @code{4}.
14239 @item __AVR_@var{Device}__
14240 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14241 the device's name. For example, @option{-mmcu=atmega8} defines the
14242 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14243 @code{__AVR_ATtiny261A__}, etc.
14245 The built-in macros' names follow
14246 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14247 the device name as from the AVR user manual. The difference between
14248 @var{Device} in the built-in macro and @var{device} in
14249 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14251 If @var{device} is not a device but only a core architecture like
14252 @samp{avr51}, this macro is not defined.
14254 @item __AVR_DEVICE_NAME__
14255 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14256 the device's name. For example, with @option{-mmcu=atmega8} the macro
14257 is defined to @code{atmega8}.
14259 If @var{device} is not a device but only a core architecture like
14260 @samp{avr51}, this macro is not defined.
14262 @item __AVR_XMEGA__
14263 The device / architecture belongs to the XMEGA family of devices.
14265 @item __AVR_HAVE_ELPM__
14266 The device has the @code{ELPM} instruction.
14268 @item __AVR_HAVE_ELPMX__
14269 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14270 R@var{n},Z+} instructions.
14272 @item __AVR_HAVE_MOVW__
14273 The device has the @code{MOVW} instruction to perform 16-bit
14274 register-register moves.
14276 @item __AVR_HAVE_LPMX__
14277 The device has the @code{LPM R@var{n},Z} and
14278 @code{LPM R@var{n},Z+} instructions.
14280 @item __AVR_HAVE_MUL__
14281 The device has a hardware multiplier.
14283 @item __AVR_HAVE_JMP_CALL__
14284 The device has the @code{JMP} and @code{CALL} instructions.
14285 This is the case for devices with at least 16@tie{}KiB of program
14288 @item __AVR_HAVE_EIJMP_EICALL__
14289 @itemx __AVR_3_BYTE_PC__
14290 The device has the @code{EIJMP} and @code{EICALL} instructions.
14291 This is the case for devices with more than 128@tie{}KiB of program memory.
14292 This also means that the program counter
14293 (PC) is 3@tie{}bytes wide.
14295 @item __AVR_2_BYTE_PC__
14296 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14297 with up to 128@tie{}KiB of program memory.
14299 @item __AVR_HAVE_8BIT_SP__
14300 @itemx __AVR_HAVE_16BIT_SP__
14301 The stack pointer (SP) register is treated as 8-bit respectively
14302 16-bit register by the compiler.
14303 The definition of these macros is affected by @option{-mtiny-stack}.
14305 @item __AVR_HAVE_SPH__
14307 The device has the SPH (high part of stack pointer) special function
14308 register or has an 8-bit stack pointer, respectively.
14309 The definition of these macros is affected by @option{-mmcu=} and
14310 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14313 @item __AVR_HAVE_RAMPD__
14314 @itemx __AVR_HAVE_RAMPX__
14315 @itemx __AVR_HAVE_RAMPY__
14316 @itemx __AVR_HAVE_RAMPZ__
14317 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14318 @code{RAMPZ} special function register, respectively.
14320 @item __NO_INTERRUPTS__
14321 This macro reflects the @option{-mno-interrupts} command-line option.
14323 @item __AVR_ERRATA_SKIP__
14324 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14325 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14326 instructions because of a hardware erratum. Skip instructions are
14327 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14328 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14331 @item __AVR_ISA_RMW__
14332 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14334 @item __AVR_SFR_OFFSET__=@var{offset}
14335 Instructions that can address I/O special function registers directly
14336 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14337 address as if addressed by an instruction to access RAM like @code{LD}
14338 or @code{STS}. This offset depends on the device architecture and has
14339 to be subtracted from the RAM address in order to get the
14340 respective I/O@tie{}address.
14342 @item __WITH_AVRLIBC__
14343 The compiler is configured to be used together with AVR-Libc.
14344 See the @option{--with-avrlibc} configure option.
14348 @node Blackfin Options
14349 @subsection Blackfin Options
14350 @cindex Blackfin Options
14353 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14355 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14356 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14357 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14358 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14359 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14360 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14361 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14362 @samp{bf561}, @samp{bf592}.
14364 The optional @var{sirevision} specifies the silicon revision of the target
14365 Blackfin processor. Any workarounds available for the targeted silicon revision
14366 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14367 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14368 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14369 hexadecimal digits representing the major and minor numbers in the silicon
14370 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14371 is not defined. If @var{sirevision} is @samp{any}, the
14372 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14373 If this optional @var{sirevision} is not used, GCC assumes the latest known
14374 silicon revision of the targeted Blackfin processor.
14376 GCC defines a preprocessor macro for the specified @var{cpu}.
14377 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14378 provided by libgloss to be linked in if @option{-msim} is not given.
14380 Without this option, @samp{bf532} is used as the processor by default.
14382 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14383 only the preprocessor macro is defined.
14387 Specifies that the program will be run on the simulator. This causes
14388 the simulator BSP provided by libgloss to be linked in. This option
14389 has effect only for @samp{bfin-elf} toolchain.
14390 Certain other options, such as @option{-mid-shared-library} and
14391 @option{-mfdpic}, imply @option{-msim}.
14393 @item -momit-leaf-frame-pointer
14394 @opindex momit-leaf-frame-pointer
14395 Don't keep the frame pointer in a register for leaf functions. This
14396 avoids the instructions to save, set up and restore frame pointers and
14397 makes an extra register available in leaf functions. The option
14398 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
14399 which might make debugging harder.
14401 @item -mspecld-anomaly
14402 @opindex mspecld-anomaly
14403 When enabled, the compiler ensures that the generated code does not
14404 contain speculative loads after jump instructions. If this option is used,
14405 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
14407 @item -mno-specld-anomaly
14408 @opindex mno-specld-anomaly
14409 Don't generate extra code to prevent speculative loads from occurring.
14411 @item -mcsync-anomaly
14412 @opindex mcsync-anomaly
14413 When enabled, the compiler ensures that the generated code does not
14414 contain CSYNC or SSYNC instructions too soon after conditional branches.
14415 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
14417 @item -mno-csync-anomaly
14418 @opindex mno-csync-anomaly
14419 Don't generate extra code to prevent CSYNC or SSYNC instructions from
14420 occurring too soon after a conditional branch.
14424 When enabled, the compiler is free to take advantage of the knowledge that
14425 the entire program fits into the low 64k of memory.
14428 @opindex mno-low-64k
14429 Assume that the program is arbitrarily large. This is the default.
14431 @item -mstack-check-l1
14432 @opindex mstack-check-l1
14433 Do stack checking using information placed into L1 scratchpad memory by the
14436 @item -mid-shared-library
14437 @opindex mid-shared-library
14438 Generate code that supports shared libraries via the library ID method.
14439 This allows for execute in place and shared libraries in an environment
14440 without virtual memory management. This option implies @option{-fPIC}.
14441 With a @samp{bfin-elf} target, this option implies @option{-msim}.
14443 @item -mno-id-shared-library
14444 @opindex mno-id-shared-library
14445 Generate code that doesn't assume ID-based shared libraries are being used.
14446 This is the default.
14448 @item -mleaf-id-shared-library
14449 @opindex mleaf-id-shared-library
14450 Generate code that supports shared libraries via the library ID method,
14451 but assumes that this library or executable won't link against any other
14452 ID shared libraries. That allows the compiler to use faster code for jumps
14455 @item -mno-leaf-id-shared-library
14456 @opindex mno-leaf-id-shared-library
14457 Do not assume that the code being compiled won't link against any ID shared
14458 libraries. Slower code is generated for jump and call insns.
14460 @item -mshared-library-id=n
14461 @opindex mshared-library-id
14462 Specifies the identification number of the ID-based shared library being
14463 compiled. Specifying a value of 0 generates more compact code; specifying
14464 other values forces the allocation of that number to the current
14465 library but is no more space- or time-efficient than omitting this option.
14469 Generate code that allows the data segment to be located in a different
14470 area of memory from the text segment. This allows for execute in place in
14471 an environment without virtual memory management by eliminating relocations
14472 against the text section.
14474 @item -mno-sep-data
14475 @opindex mno-sep-data
14476 Generate code that assumes that the data segment follows the text segment.
14477 This is the default.
14480 @itemx -mno-long-calls
14481 @opindex mlong-calls
14482 @opindex mno-long-calls
14483 Tells the compiler to perform function calls by first loading the
14484 address of the function into a register and then performing a subroutine
14485 call on this register. This switch is needed if the target function
14486 lies outside of the 24-bit addressing range of the offset-based
14487 version of subroutine call instruction.
14489 This feature is not enabled by default. Specifying
14490 @option{-mno-long-calls} restores the default behavior. Note these
14491 switches have no effect on how the compiler generates code to handle
14492 function calls via function pointers.
14496 Link with the fast floating-point library. This library relaxes some of
14497 the IEEE floating-point standard's rules for checking inputs against
14498 Not-a-Number (NAN), in the interest of performance.
14501 @opindex minline-plt
14502 Enable inlining of PLT entries in function calls to functions that are
14503 not known to bind locally. It has no effect without @option{-mfdpic}.
14506 @opindex mmulticore
14507 Build a standalone application for multicore Blackfin processors.
14508 This option causes proper start files and link scripts supporting
14509 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
14510 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
14512 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
14513 selects the one-application-per-core programming model. Without
14514 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
14515 programming model is used. In this model, the main function of Core B
14516 should be named as @code{coreb_main}.
14518 If this option is not used, the single-core application programming
14523 Build a standalone application for Core A of BF561 when using
14524 the one-application-per-core programming model. Proper start files
14525 and link scripts are used to support Core A, and the macro
14526 @code{__BFIN_COREA} is defined.
14527 This option can only be used in conjunction with @option{-mmulticore}.
14531 Build a standalone application for Core B of BF561 when using
14532 the one-application-per-core programming model. Proper start files
14533 and link scripts are used to support Core B, and the macro
14534 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
14535 should be used instead of @code{main}.
14536 This option can only be used in conjunction with @option{-mmulticore}.
14540 Build a standalone application for SDRAM. Proper start files and
14541 link scripts are used to put the application into SDRAM, and the macro
14542 @code{__BFIN_SDRAM} is defined.
14543 The loader should initialize SDRAM before loading the application.
14547 Assume that ICPLBs are enabled at run time. This has an effect on certain
14548 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
14549 are enabled; for standalone applications the default is off.
14553 @subsection C6X Options
14554 @cindex C6X Options
14557 @item -march=@var{name}
14559 This specifies the name of the target architecture. GCC uses this
14560 name to determine what kind of instructions it can emit when generating
14561 assembly code. Permissible names are: @samp{c62x},
14562 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
14565 @opindex mbig-endian
14566 Generate code for a big-endian target.
14568 @item -mlittle-endian
14569 @opindex mlittle-endian
14570 Generate code for a little-endian target. This is the default.
14574 Choose startup files and linker script suitable for the simulator.
14576 @item -msdata=default
14577 @opindex msdata=default
14578 Put small global and static data in the @code{.neardata} section,
14579 which is pointed to by register @code{B14}. Put small uninitialized
14580 global and static data in the @code{.bss} section, which is adjacent
14581 to the @code{.neardata} section. Put small read-only data into the
14582 @code{.rodata} section. The corresponding sections used for large
14583 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
14586 @opindex msdata=all
14587 Put all data, not just small objects, into the sections reserved for
14588 small data, and use addressing relative to the @code{B14} register to
14592 @opindex msdata=none
14593 Make no use of the sections reserved for small data, and use absolute
14594 addresses to access all data. Put all initialized global and static
14595 data in the @code{.fardata} section, and all uninitialized data in the
14596 @code{.far} section. Put all constant data into the @code{.const}
14601 @subsection CRIS Options
14602 @cindex CRIS Options
14604 These options are defined specifically for the CRIS ports.
14607 @item -march=@var{architecture-type}
14608 @itemx -mcpu=@var{architecture-type}
14611 Generate code for the specified architecture. The choices for
14612 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
14613 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
14614 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
14617 @item -mtune=@var{architecture-type}
14619 Tune to @var{architecture-type} everything applicable about the generated
14620 code, except for the ABI and the set of available instructions. The
14621 choices for @var{architecture-type} are the same as for
14622 @option{-march=@var{architecture-type}}.
14624 @item -mmax-stack-frame=@var{n}
14625 @opindex mmax-stack-frame
14626 Warn when the stack frame of a function exceeds @var{n} bytes.
14632 The options @option{-metrax4} and @option{-metrax100} are synonyms for
14633 @option{-march=v3} and @option{-march=v8} respectively.
14635 @item -mmul-bug-workaround
14636 @itemx -mno-mul-bug-workaround
14637 @opindex mmul-bug-workaround
14638 @opindex mno-mul-bug-workaround
14639 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
14640 models where it applies. This option is active by default.
14644 Enable CRIS-specific verbose debug-related information in the assembly
14645 code. This option also has the effect of turning off the @samp{#NO_APP}
14646 formatted-code indicator to the assembler at the beginning of the
14651 Do not use condition-code results from previous instruction; always emit
14652 compare and test instructions before use of condition codes.
14654 @item -mno-side-effects
14655 @opindex mno-side-effects
14656 Do not emit instructions with side effects in addressing modes other than
14659 @item -mstack-align
14660 @itemx -mno-stack-align
14661 @itemx -mdata-align
14662 @itemx -mno-data-align
14663 @itemx -mconst-align
14664 @itemx -mno-const-align
14665 @opindex mstack-align
14666 @opindex mno-stack-align
14667 @opindex mdata-align
14668 @opindex mno-data-align
14669 @opindex mconst-align
14670 @opindex mno-const-align
14671 These options (@samp{no-} options) arrange (eliminate arrangements) for the
14672 stack frame, individual data and constants to be aligned for the maximum
14673 single data access size for the chosen CPU model. The default is to
14674 arrange for 32-bit alignment. ABI details such as structure layout are
14675 not affected by these options.
14683 Similar to the stack- data- and const-align options above, these options
14684 arrange for stack frame, writable data and constants to all be 32-bit,
14685 16-bit or 8-bit aligned. The default is 32-bit alignment.
14687 @item -mno-prologue-epilogue
14688 @itemx -mprologue-epilogue
14689 @opindex mno-prologue-epilogue
14690 @opindex mprologue-epilogue
14691 With @option{-mno-prologue-epilogue}, the normal function prologue and
14692 epilogue which set up the stack frame are omitted and no return
14693 instructions or return sequences are generated in the code. Use this
14694 option only together with visual inspection of the compiled code: no
14695 warnings or errors are generated when call-saved registers must be saved,
14696 or storage for local variables needs to be allocated.
14700 @opindex mno-gotplt
14702 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
14703 instruction sequences that load addresses for functions from the PLT part
14704 of the GOT rather than (traditional on other architectures) calls to the
14705 PLT@. The default is @option{-mgotplt}.
14709 Legacy no-op option only recognized with the cris-axis-elf and
14710 cris-axis-linux-gnu targets.
14714 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
14718 This option, recognized for the cris-axis-elf, arranges
14719 to link with input-output functions from a simulator library. Code,
14720 initialized data and zero-initialized data are allocated consecutively.
14724 Like @option{-sim}, but pass linker options to locate initialized data at
14725 0x40000000 and zero-initialized data at 0x80000000.
14729 @subsection CR16 Options
14730 @cindex CR16 Options
14732 These options are defined specifically for the CR16 ports.
14738 Enable the use of multiply-accumulate instructions. Disabled by default.
14742 @opindex mcr16cplus
14744 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
14749 Links the library libsim.a which is in compatible with simulator. Applicable
14750 to ELF compiler only.
14754 Choose integer type as 32-bit wide.
14758 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
14760 @item -mdata-model=@var{model}
14761 @opindex mdata-model
14762 Choose a data model. The choices for @var{model} are @samp{near},
14763 @samp{far} or @samp{medium}. @samp{medium} is default.
14764 However, @samp{far} is not valid with @option{-mcr16c}, as the
14765 CR16C architecture does not support the far data model.
14768 @node Darwin Options
14769 @subsection Darwin Options
14770 @cindex Darwin options
14772 These options are defined for all architectures running the Darwin operating
14775 FSF GCC on Darwin does not create ``fat'' object files; it creates
14776 an object file for the single architecture that GCC was built to
14777 target. Apple's GCC on Darwin does create ``fat'' files if multiple
14778 @option{-arch} options are used; it does so by running the compiler or
14779 linker multiple times and joining the results together with
14782 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
14783 @samp{i686}) is determined by the flags that specify the ISA
14784 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
14785 @option{-force_cpusubtype_ALL} option can be used to override this.
14787 The Darwin tools vary in their behavior when presented with an ISA
14788 mismatch. The assembler, @file{as}, only permits instructions to
14789 be used that are valid for the subtype of the file it is generating,
14790 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
14791 The linker for shared libraries, @file{/usr/bin/libtool}, fails
14792 and prints an error if asked to create a shared library with a less
14793 restrictive subtype than its input files (for instance, trying to put
14794 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
14795 for executables, @command{ld}, quietly gives the executable the most
14796 restrictive subtype of any of its input files.
14801 Add the framework directory @var{dir} to the head of the list of
14802 directories to be searched for header files. These directories are
14803 interleaved with those specified by @option{-I} options and are
14804 scanned in a left-to-right order.
14806 A framework directory is a directory with frameworks in it. A
14807 framework is a directory with a @file{Headers} and/or
14808 @file{PrivateHeaders} directory contained directly in it that ends
14809 in @file{.framework}. The name of a framework is the name of this
14810 directory excluding the @file{.framework}. Headers associated with
14811 the framework are found in one of those two directories, with
14812 @file{Headers} being searched first. A subframework is a framework
14813 directory that is in a framework's @file{Frameworks} directory.
14814 Includes of subframework headers can only appear in a header of a
14815 framework that contains the subframework, or in a sibling subframework
14816 header. Two subframeworks are siblings if they occur in the same
14817 framework. A subframework should not have the same name as a
14818 framework; a warning is issued if this is violated. Currently a
14819 subframework cannot have subframeworks; in the future, the mechanism
14820 may be extended to support this. The standard frameworks can be found
14821 in @file{/System/Library/Frameworks} and
14822 @file{/Library/Frameworks}. An example include looks like
14823 @code{#include <Framework/header.h>}, where @file{Framework} denotes
14824 the name of the framework and @file{header.h} is found in the
14825 @file{PrivateHeaders} or @file{Headers} directory.
14827 @item -iframework@var{dir}
14828 @opindex iframework
14829 Like @option{-F} except the directory is a treated as a system
14830 directory. The main difference between this @option{-iframework} and
14831 @option{-F} is that with @option{-iframework} the compiler does not
14832 warn about constructs contained within header files found via
14833 @var{dir}. This option is valid only for the C family of languages.
14837 Emit debugging information for symbols that are used. For stabs
14838 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
14839 This is by default ON@.
14843 Emit debugging information for all symbols and types.
14845 @item -mmacosx-version-min=@var{version}
14846 The earliest version of MacOS X that this executable will run on
14847 is @var{version}. Typical values of @var{version} include @code{10.1},
14848 @code{10.2}, and @code{10.3.9}.
14850 If the compiler was built to use the system's headers by default,
14851 then the default for this option is the system version on which the
14852 compiler is running, otherwise the default is to make choices that
14853 are compatible with as many systems and code bases as possible.
14857 Enable kernel development mode. The @option{-mkernel} option sets
14858 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
14859 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
14860 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
14861 applicable. This mode also sets @option{-mno-altivec},
14862 @option{-msoft-float}, @option{-fno-builtin} and
14863 @option{-mlong-branch} for PowerPC targets.
14865 @item -mone-byte-bool
14866 @opindex mone-byte-bool
14867 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
14868 By default @code{sizeof(bool)} is @code{4} when compiling for
14869 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
14870 option has no effect on x86.
14872 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
14873 to generate code that is not binary compatible with code generated
14874 without that switch. Using this switch may require recompiling all
14875 other modules in a program, including system libraries. Use this
14876 switch to conform to a non-default data model.
14878 @item -mfix-and-continue
14879 @itemx -ffix-and-continue
14880 @itemx -findirect-data
14881 @opindex mfix-and-continue
14882 @opindex ffix-and-continue
14883 @opindex findirect-data
14884 Generate code suitable for fast turnaround development, such as to
14885 allow GDB to dynamically load @file{.o} files into already-running
14886 programs. @option{-findirect-data} and @option{-ffix-and-continue}
14887 are provided for backwards compatibility.
14891 Loads all members of static archive libraries.
14892 See man ld(1) for more information.
14894 @item -arch_errors_fatal
14895 @opindex arch_errors_fatal
14896 Cause the errors having to do with files that have the wrong architecture
14899 @item -bind_at_load
14900 @opindex bind_at_load
14901 Causes the output file to be marked such that the dynamic linker will
14902 bind all undefined references when the file is loaded or launched.
14906 Produce a Mach-o bundle format file.
14907 See man ld(1) for more information.
14909 @item -bundle_loader @var{executable}
14910 @opindex bundle_loader
14911 This option specifies the @var{executable} that will load the build
14912 output file being linked. See man ld(1) for more information.
14915 @opindex dynamiclib
14916 When passed this option, GCC produces a dynamic library instead of
14917 an executable when linking, using the Darwin @file{libtool} command.
14919 @item -force_cpusubtype_ALL
14920 @opindex force_cpusubtype_ALL
14921 This causes GCC's output file to have the @samp{ALL} subtype, instead of
14922 one controlled by the @option{-mcpu} or @option{-march} option.
14924 @item -allowable_client @var{client_name}
14925 @itemx -client_name
14926 @itemx -compatibility_version
14927 @itemx -current_version
14929 @itemx -dependency-file
14931 @itemx -dylinker_install_name
14933 @itemx -exported_symbols_list
14936 @itemx -flat_namespace
14937 @itemx -force_flat_namespace
14938 @itemx -headerpad_max_install_names
14941 @itemx -install_name
14942 @itemx -keep_private_externs
14943 @itemx -multi_module
14944 @itemx -multiply_defined
14945 @itemx -multiply_defined_unused
14948 @itemx -no_dead_strip_inits_and_terms
14949 @itemx -nofixprebinding
14950 @itemx -nomultidefs
14952 @itemx -noseglinkedit
14953 @itemx -pagezero_size
14955 @itemx -prebind_all_twolevel_modules
14956 @itemx -private_bundle
14958 @itemx -read_only_relocs
14960 @itemx -sectobjectsymbols
14964 @itemx -sectobjectsymbols
14967 @itemx -segs_read_only_addr
14969 @itemx -segs_read_write_addr
14970 @itemx -seg_addr_table
14971 @itemx -seg_addr_table_filename
14972 @itemx -seglinkedit
14974 @itemx -segs_read_only_addr
14975 @itemx -segs_read_write_addr
14976 @itemx -single_module
14978 @itemx -sub_library
14980 @itemx -sub_umbrella
14981 @itemx -twolevel_namespace
14984 @itemx -unexported_symbols_list
14985 @itemx -weak_reference_mismatches
14986 @itemx -whatsloaded
14987 @opindex allowable_client
14988 @opindex client_name
14989 @opindex compatibility_version
14990 @opindex current_version
14991 @opindex dead_strip
14992 @opindex dependency-file
14993 @opindex dylib_file
14994 @opindex dylinker_install_name
14996 @opindex exported_symbols_list
14998 @opindex flat_namespace
14999 @opindex force_flat_namespace
15000 @opindex headerpad_max_install_names
15001 @opindex image_base
15003 @opindex install_name
15004 @opindex keep_private_externs
15005 @opindex multi_module
15006 @opindex multiply_defined
15007 @opindex multiply_defined_unused
15008 @opindex noall_load
15009 @opindex no_dead_strip_inits_and_terms
15010 @opindex nofixprebinding
15011 @opindex nomultidefs
15013 @opindex noseglinkedit
15014 @opindex pagezero_size
15016 @opindex prebind_all_twolevel_modules
15017 @opindex private_bundle
15018 @opindex read_only_relocs
15020 @opindex sectobjectsymbols
15023 @opindex sectcreate
15024 @opindex sectobjectsymbols
15027 @opindex segs_read_only_addr
15028 @opindex segs_read_write_addr
15029 @opindex seg_addr_table
15030 @opindex seg_addr_table_filename
15031 @opindex seglinkedit
15033 @opindex segs_read_only_addr
15034 @opindex segs_read_write_addr
15035 @opindex single_module
15037 @opindex sub_library
15038 @opindex sub_umbrella
15039 @opindex twolevel_namespace
15042 @opindex unexported_symbols_list
15043 @opindex weak_reference_mismatches
15044 @opindex whatsloaded
15045 These options are passed to the Darwin linker. The Darwin linker man page
15046 describes them in detail.
15049 @node DEC Alpha Options
15050 @subsection DEC Alpha Options
15052 These @samp{-m} options are defined for the DEC Alpha implementations:
15055 @item -mno-soft-float
15056 @itemx -msoft-float
15057 @opindex mno-soft-float
15058 @opindex msoft-float
15059 Use (do not use) the hardware floating-point instructions for
15060 floating-point operations. When @option{-msoft-float} is specified,
15061 functions in @file{libgcc.a} are used to perform floating-point
15062 operations. Unless they are replaced by routines that emulate the
15063 floating-point operations, or compiled in such a way as to call such
15064 emulations routines, these routines issue floating-point
15065 operations. If you are compiling for an Alpha without floating-point
15066 operations, you must ensure that the library is built so as not to call
15069 Note that Alpha implementations without floating-point operations are
15070 required to have floating-point registers.
15073 @itemx -mno-fp-regs
15075 @opindex mno-fp-regs
15076 Generate code that uses (does not use) the floating-point register set.
15077 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
15078 register set is not used, floating-point operands are passed in integer
15079 registers as if they were integers and floating-point results are passed
15080 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
15081 so any function with a floating-point argument or return value called by code
15082 compiled with @option{-mno-fp-regs} must also be compiled with that
15085 A typical use of this option is building a kernel that does not use,
15086 and hence need not save and restore, any floating-point registers.
15090 The Alpha architecture implements floating-point hardware optimized for
15091 maximum performance. It is mostly compliant with the IEEE floating-point
15092 standard. However, for full compliance, software assistance is
15093 required. This option generates code fully IEEE-compliant code
15094 @emph{except} that the @var{inexact-flag} is not maintained (see below).
15095 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
15096 defined during compilation. The resulting code is less efficient but is
15097 able to correctly support denormalized numbers and exceptional IEEE
15098 values such as not-a-number and plus/minus infinity. Other Alpha
15099 compilers call this option @option{-ieee_with_no_inexact}.
15101 @item -mieee-with-inexact
15102 @opindex mieee-with-inexact
15103 This is like @option{-mieee} except the generated code also maintains
15104 the IEEE @var{inexact-flag}. Turning on this option causes the
15105 generated code to implement fully-compliant IEEE math. In addition to
15106 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
15107 macro. On some Alpha implementations the resulting code may execute
15108 significantly slower than the code generated by default. Since there is
15109 very little code that depends on the @var{inexact-flag}, you should
15110 normally not specify this option. Other Alpha compilers call this
15111 option @option{-ieee_with_inexact}.
15113 @item -mfp-trap-mode=@var{trap-mode}
15114 @opindex mfp-trap-mode
15115 This option controls what floating-point related traps are enabled.
15116 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
15117 The trap mode can be set to one of four values:
15121 This is the default (normal) setting. The only traps that are enabled
15122 are the ones that cannot be disabled in software (e.g., division by zero
15126 In addition to the traps enabled by @samp{n}, underflow traps are enabled
15130 Like @samp{u}, but the instructions are marked to be safe for software
15131 completion (see Alpha architecture manual for details).
15134 Like @samp{su}, but inexact traps are enabled as well.
15137 @item -mfp-rounding-mode=@var{rounding-mode}
15138 @opindex mfp-rounding-mode
15139 Selects the IEEE rounding mode. Other Alpha compilers call this option
15140 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
15145 Normal IEEE rounding mode. Floating-point numbers are rounded towards
15146 the nearest machine number or towards the even machine number in case
15150 Round towards minus infinity.
15153 Chopped rounding mode. Floating-point numbers are rounded towards zero.
15156 Dynamic rounding mode. A field in the floating-point control register
15157 (@var{fpcr}, see Alpha architecture reference manual) controls the
15158 rounding mode in effect. The C library initializes this register for
15159 rounding towards plus infinity. Thus, unless your program modifies the
15160 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15163 @item -mtrap-precision=@var{trap-precision}
15164 @opindex mtrap-precision
15165 In the Alpha architecture, floating-point traps are imprecise. This
15166 means without software assistance it is impossible to recover from a
15167 floating trap and program execution normally needs to be terminated.
15168 GCC can generate code that can assist operating system trap handlers
15169 in determining the exact location that caused a floating-point trap.
15170 Depending on the requirements of an application, different levels of
15171 precisions can be selected:
15175 Program precision. This option is the default and means a trap handler
15176 can only identify which program caused a floating-point exception.
15179 Function precision. The trap handler can determine the function that
15180 caused a floating-point exception.
15183 Instruction precision. The trap handler can determine the exact
15184 instruction that caused a floating-point exception.
15187 Other Alpha compilers provide the equivalent options called
15188 @option{-scope_safe} and @option{-resumption_safe}.
15190 @item -mieee-conformant
15191 @opindex mieee-conformant
15192 This option marks the generated code as IEEE conformant. You must not
15193 use this option unless you also specify @option{-mtrap-precision=i} and either
15194 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15195 is to emit the line @samp{.eflag 48} in the function prologue of the
15196 generated assembly file.
15198 @item -mbuild-constants
15199 @opindex mbuild-constants
15200 Normally GCC examines a 32- or 64-bit integer constant to
15201 see if it can construct it from smaller constants in two or three
15202 instructions. If it cannot, it outputs the constant as a literal and
15203 generates code to load it from the data segment at run time.
15205 Use this option to require GCC to construct @emph{all} integer constants
15206 using code, even if it takes more instructions (the maximum is six).
15208 You typically use this option to build a shared library dynamic
15209 loader. Itself a shared library, it must relocate itself in memory
15210 before it can find the variables and constants in its own data segment.
15228 Indicate whether GCC should generate code to use the optional BWX,
15229 CIX, FIX and MAX instruction sets. The default is to use the instruction
15230 sets supported by the CPU type specified via @option{-mcpu=} option or that
15231 of the CPU on which GCC was built if none is specified.
15234 @itemx -mfloat-ieee
15235 @opindex mfloat-vax
15236 @opindex mfloat-ieee
15237 Generate code that uses (does not use) VAX F and G floating-point
15238 arithmetic instead of IEEE single and double precision.
15240 @item -mexplicit-relocs
15241 @itemx -mno-explicit-relocs
15242 @opindex mexplicit-relocs
15243 @opindex mno-explicit-relocs
15244 Older Alpha assemblers provided no way to generate symbol relocations
15245 except via assembler macros. Use of these macros does not allow
15246 optimal instruction scheduling. GNU binutils as of version 2.12
15247 supports a new syntax that allows the compiler to explicitly mark
15248 which relocations should apply to which instructions. This option
15249 is mostly useful for debugging, as GCC detects the capabilities of
15250 the assembler when it is built and sets the default accordingly.
15253 @itemx -mlarge-data
15254 @opindex msmall-data
15255 @opindex mlarge-data
15256 When @option{-mexplicit-relocs} is in effect, static data is
15257 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15258 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15259 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15260 16-bit relocations off of the @code{$gp} register. This limits the
15261 size of the small data area to 64KB, but allows the variables to be
15262 directly accessed via a single instruction.
15264 The default is @option{-mlarge-data}. With this option the data area
15265 is limited to just below 2GB@. Programs that require more than 2GB of
15266 data must use @code{malloc} or @code{mmap} to allocate the data in the
15267 heap instead of in the program's data segment.
15269 When generating code for shared libraries, @option{-fpic} implies
15270 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15273 @itemx -mlarge-text
15274 @opindex msmall-text
15275 @opindex mlarge-text
15276 When @option{-msmall-text} is used, the compiler assumes that the
15277 code of the entire program (or shared library) fits in 4MB, and is
15278 thus reachable with a branch instruction. When @option{-msmall-data}
15279 is used, the compiler can assume that all local symbols share the
15280 same @code{$gp} value, and thus reduce the number of instructions
15281 required for a function call from 4 to 1.
15283 The default is @option{-mlarge-text}.
15285 @item -mcpu=@var{cpu_type}
15287 Set the instruction set and instruction scheduling parameters for
15288 machine type @var{cpu_type}. You can specify either the @samp{EV}
15289 style name or the corresponding chip number. GCC supports scheduling
15290 parameters for the EV4, EV5 and EV6 family of processors and
15291 chooses the default values for the instruction set from the processor
15292 you specify. If you do not specify a processor type, GCC defaults
15293 to the processor on which the compiler was built.
15295 Supported values for @var{cpu_type} are
15301 Schedules as an EV4 and has no instruction set extensions.
15305 Schedules as an EV5 and has no instruction set extensions.
15309 Schedules as an EV5 and supports the BWX extension.
15314 Schedules as an EV5 and supports the BWX and MAX extensions.
15318 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15322 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15325 Native toolchains also support the value @samp{native},
15326 which selects the best architecture option for the host processor.
15327 @option{-mcpu=native} has no effect if GCC does not recognize
15330 @item -mtune=@var{cpu_type}
15332 Set only the instruction scheduling parameters for machine type
15333 @var{cpu_type}. The instruction set is not changed.
15335 Native toolchains also support the value @samp{native},
15336 which selects the best architecture option for the host processor.
15337 @option{-mtune=native} has no effect if GCC does not recognize
15340 @item -mmemory-latency=@var{time}
15341 @opindex mmemory-latency
15342 Sets the latency the scheduler should assume for typical memory
15343 references as seen by the application. This number is highly
15344 dependent on the memory access patterns used by the application
15345 and the size of the external cache on the machine.
15347 Valid options for @var{time} are
15351 A decimal number representing clock cycles.
15357 The compiler contains estimates of the number of clock cycles for
15358 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15359 (also called Dcache, Scache, and Bcache), as well as to main memory.
15360 Note that L3 is only valid for EV5.
15366 @subsection FR30 Options
15367 @cindex FR30 Options
15369 These options are defined specifically for the FR30 port.
15373 @item -msmall-model
15374 @opindex msmall-model
15375 Use the small address space model. This can produce smaller code, but
15376 it does assume that all symbolic values and addresses fit into a
15381 Assume that runtime support has been provided and so there is no need
15382 to include the simulator library (@file{libsim.a}) on the linker
15388 @subsection FT32 Options
15389 @cindex FT32 Options
15391 These options are defined specifically for the FT32 port.
15397 Specifies that the program will be run on the simulator. This causes
15398 an alternate runtime startup and library to be linked.
15399 You must not use this option when generating programs that will run on
15400 real hardware; you must provide your own runtime library for whatever
15401 I/O functions are needed.
15405 Enable Local Register Allocation. This is still experimental for FT32,
15406 so by default the compiler uses standard reload.
15411 @subsection FRV Options
15412 @cindex FRV Options
15418 Only use the first 32 general-purpose registers.
15423 Use all 64 general-purpose registers.
15428 Use only the first 32 floating-point registers.
15433 Use all 64 floating-point registers.
15436 @opindex mhard-float
15438 Use hardware instructions for floating-point operations.
15441 @opindex msoft-float
15443 Use library routines for floating-point operations.
15448 Dynamically allocate condition code registers.
15453 Do not try to dynamically allocate condition code registers, only
15454 use @code{icc0} and @code{fcc0}.
15459 Change ABI to use double word insns.
15464 Do not use double word instructions.
15469 Use floating-point double instructions.
15472 @opindex mno-double
15474 Do not use floating-point double instructions.
15479 Use media instructions.
15484 Do not use media instructions.
15489 Use multiply and add/subtract instructions.
15492 @opindex mno-muladd
15494 Do not use multiply and add/subtract instructions.
15499 Select the FDPIC ABI, which uses function descriptors to represent
15500 pointers to functions. Without any PIC/PIE-related options, it
15501 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
15502 assumes GOT entries and small data are within a 12-bit range from the
15503 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
15504 are computed with 32 bits.
15505 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15508 @opindex minline-plt
15510 Enable inlining of PLT entries in function calls to functions that are
15511 not known to bind locally. It has no effect without @option{-mfdpic}.
15512 It's enabled by default if optimizing for speed and compiling for
15513 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
15514 optimization option such as @option{-O3} or above is present in the
15520 Assume a large TLS segment when generating thread-local code.
15525 Do not assume a large TLS segment when generating thread-local code.
15530 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
15531 that is known to be in read-only sections. It's enabled by default,
15532 except for @option{-fpic} or @option{-fpie}: even though it may help
15533 make the global offset table smaller, it trades 1 instruction for 4.
15534 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
15535 one of which may be shared by multiple symbols, and it avoids the need
15536 for a GOT entry for the referenced symbol, so it's more likely to be a
15537 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
15539 @item -multilib-library-pic
15540 @opindex multilib-library-pic
15542 Link with the (library, not FD) pic libraries. It's implied by
15543 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
15544 @option{-fpic} without @option{-mfdpic}. You should never have to use
15548 @opindex mlinked-fp
15550 Follow the EABI requirement of always creating a frame pointer whenever
15551 a stack frame is allocated. This option is enabled by default and can
15552 be disabled with @option{-mno-linked-fp}.
15555 @opindex mlong-calls
15557 Use indirect addressing to call functions outside the current
15558 compilation unit. This allows the functions to be placed anywhere
15559 within the 32-bit address space.
15561 @item -malign-labels
15562 @opindex malign-labels
15564 Try to align labels to an 8-byte boundary by inserting NOPs into the
15565 previous packet. This option only has an effect when VLIW packing
15566 is enabled. It doesn't create new packets; it merely adds NOPs to
15569 @item -mlibrary-pic
15570 @opindex mlibrary-pic
15572 Generate position-independent EABI code.
15577 Use only the first four media accumulator registers.
15582 Use all eight media accumulator registers.
15587 Pack VLIW instructions.
15592 Do not pack VLIW instructions.
15595 @opindex mno-eflags
15597 Do not mark ABI switches in e_flags.
15600 @opindex mcond-move
15602 Enable the use of conditional-move instructions (default).
15604 This switch is mainly for debugging the compiler and will likely be removed
15605 in a future version.
15607 @item -mno-cond-move
15608 @opindex mno-cond-move
15610 Disable the use of conditional-move instructions.
15612 This switch is mainly for debugging the compiler and will likely be removed
15613 in a future version.
15618 Enable the use of conditional set instructions (default).
15620 This switch is mainly for debugging the compiler and will likely be removed
15621 in a future version.
15626 Disable the use of conditional set instructions.
15628 This switch is mainly for debugging the compiler and will likely be removed
15629 in a future version.
15632 @opindex mcond-exec
15634 Enable the use of conditional execution (default).
15636 This switch is mainly for debugging the compiler and will likely be removed
15637 in a future version.
15639 @item -mno-cond-exec
15640 @opindex mno-cond-exec
15642 Disable the use of conditional execution.
15644 This switch is mainly for debugging the compiler and will likely be removed
15645 in a future version.
15647 @item -mvliw-branch
15648 @opindex mvliw-branch
15650 Run a pass to pack branches into VLIW instructions (default).
15652 This switch is mainly for debugging the compiler and will likely be removed
15653 in a future version.
15655 @item -mno-vliw-branch
15656 @opindex mno-vliw-branch
15658 Do not run a pass to pack branches into VLIW instructions.
15660 This switch is mainly for debugging the compiler and will likely be removed
15661 in a future version.
15663 @item -mmulti-cond-exec
15664 @opindex mmulti-cond-exec
15666 Enable optimization of @code{&&} and @code{||} in conditional execution
15669 This switch is mainly for debugging the compiler and will likely be removed
15670 in a future version.
15672 @item -mno-multi-cond-exec
15673 @opindex mno-multi-cond-exec
15675 Disable optimization of @code{&&} and @code{||} in conditional execution.
15677 This switch is mainly for debugging the compiler and will likely be removed
15678 in a future version.
15680 @item -mnested-cond-exec
15681 @opindex mnested-cond-exec
15683 Enable nested conditional execution optimizations (default).
15685 This switch is mainly for debugging the compiler and will likely be removed
15686 in a future version.
15688 @item -mno-nested-cond-exec
15689 @opindex mno-nested-cond-exec
15691 Disable nested conditional execution optimizations.
15693 This switch is mainly for debugging the compiler and will likely be removed
15694 in a future version.
15696 @item -moptimize-membar
15697 @opindex moptimize-membar
15699 This switch removes redundant @code{membar} instructions from the
15700 compiler-generated code. It is enabled by default.
15702 @item -mno-optimize-membar
15703 @opindex mno-optimize-membar
15705 This switch disables the automatic removal of redundant @code{membar}
15706 instructions from the generated code.
15708 @item -mtomcat-stats
15709 @opindex mtomcat-stats
15711 Cause gas to print out tomcat statistics.
15713 @item -mcpu=@var{cpu}
15716 Select the processor type for which to generate code. Possible values are
15717 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
15718 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
15722 @node GNU/Linux Options
15723 @subsection GNU/Linux Options
15725 These @samp{-m} options are defined for GNU/Linux targets:
15730 Use the GNU C library. This is the default except
15731 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
15732 @samp{*-*-linux-*android*} targets.
15736 Use uClibc C library. This is the default on
15737 @samp{*-*-linux-*uclibc*} targets.
15741 Use the musl C library. This is the default on
15742 @samp{*-*-linux-*musl*} targets.
15746 Use Bionic C library. This is the default on
15747 @samp{*-*-linux-*android*} targets.
15751 Compile code compatible with Android platform. This is the default on
15752 @samp{*-*-linux-*android*} targets.
15754 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
15755 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
15756 this option makes the GCC driver pass Android-specific options to the linker.
15757 Finally, this option causes the preprocessor macro @code{__ANDROID__}
15760 @item -tno-android-cc
15761 @opindex tno-android-cc
15762 Disable compilation effects of @option{-mandroid}, i.e., do not enable
15763 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
15764 @option{-fno-rtti} by default.
15766 @item -tno-android-ld
15767 @opindex tno-android-ld
15768 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
15769 linking options to the linker.
15773 @node H8/300 Options
15774 @subsection H8/300 Options
15776 These @samp{-m} options are defined for the H8/300 implementations:
15781 Shorten some address references at link time, when possible; uses the
15782 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
15783 ld, Using ld}, for a fuller description.
15787 Generate code for the H8/300H@.
15791 Generate code for the H8S@.
15795 Generate code for the H8S and H8/300H in the normal mode. This switch
15796 must be used either with @option{-mh} or @option{-ms}.
15800 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
15804 Extended registers are stored on stack before execution of function
15805 with monitor attribute. Default option is @option{-mexr}.
15806 This option is valid only for H8S targets.
15810 Extended registers are not stored on stack before execution of function
15811 with monitor attribute. Default option is @option{-mno-exr}.
15812 This option is valid only for H8S targets.
15816 Make @code{int} data 32 bits by default.
15819 @opindex malign-300
15820 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
15821 The default for the H8/300H and H8S is to align longs and floats on
15823 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
15824 This option has no effect on the H8/300.
15828 @subsection HPPA Options
15829 @cindex HPPA Options
15831 These @samp{-m} options are defined for the HPPA family of computers:
15834 @item -march=@var{architecture-type}
15836 Generate code for the specified architecture. The choices for
15837 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
15838 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
15839 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
15840 architecture option for your machine. Code compiled for lower numbered
15841 architectures runs on higher numbered architectures, but not the
15844 @item -mpa-risc-1-0
15845 @itemx -mpa-risc-1-1
15846 @itemx -mpa-risc-2-0
15847 @opindex mpa-risc-1-0
15848 @opindex mpa-risc-1-1
15849 @opindex mpa-risc-2-0
15850 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
15852 @item -mjump-in-delay
15853 @opindex mjump-in-delay
15854 This option is ignored and provided for compatibility purposes only.
15856 @item -mdisable-fpregs
15857 @opindex mdisable-fpregs
15858 Prevent floating-point registers from being used in any manner. This is
15859 necessary for compiling kernels that perform lazy context switching of
15860 floating-point registers. If you use this option and attempt to perform
15861 floating-point operations, the compiler aborts.
15863 @item -mdisable-indexing
15864 @opindex mdisable-indexing
15865 Prevent the compiler from using indexing address modes. This avoids some
15866 rather obscure problems when compiling MIG generated code under MACH@.
15868 @item -mno-space-regs
15869 @opindex mno-space-regs
15870 Generate code that assumes the target has no space registers. This allows
15871 GCC to generate faster indirect calls and use unscaled index address modes.
15873 Such code is suitable for level 0 PA systems and kernels.
15875 @item -mfast-indirect-calls
15876 @opindex mfast-indirect-calls
15877 Generate code that assumes calls never cross space boundaries. This
15878 allows GCC to emit code that performs faster indirect calls.
15880 This option does not work in the presence of shared libraries or nested
15883 @item -mfixed-range=@var{register-range}
15884 @opindex mfixed-range
15885 Generate code treating the given register range as fixed registers.
15886 A fixed register is one that the register allocator cannot use. This is
15887 useful when compiling kernel code. A register range is specified as
15888 two registers separated by a dash. Multiple register ranges can be
15889 specified separated by a comma.
15891 @item -mlong-load-store
15892 @opindex mlong-load-store
15893 Generate 3-instruction load and store sequences as sometimes required by
15894 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
15897 @item -mportable-runtime
15898 @opindex mportable-runtime
15899 Use the portable calling conventions proposed by HP for ELF systems.
15903 Enable the use of assembler directives only GAS understands.
15905 @item -mschedule=@var{cpu-type}
15907 Schedule code according to the constraints for the machine type
15908 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
15909 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
15910 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
15911 proper scheduling option for your machine. The default scheduling is
15915 @opindex mlinker-opt
15916 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
15917 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
15918 linkers in which they give bogus error messages when linking some programs.
15921 @opindex msoft-float
15922 Generate output containing library calls for floating point.
15923 @strong{Warning:} the requisite libraries are not available for all HPPA
15924 targets. Normally the facilities of the machine's usual C compiler are
15925 used, but this cannot be done directly in cross-compilation. You must make
15926 your own arrangements to provide suitable library functions for
15929 @option{-msoft-float} changes the calling convention in the output file;
15930 therefore, it is only useful if you compile @emph{all} of a program with
15931 this option. In particular, you need to compile @file{libgcc.a}, the
15932 library that comes with GCC, with @option{-msoft-float} in order for
15937 Generate the predefine, @code{_SIO}, for server IO@. The default is
15938 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
15939 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
15940 options are available under HP-UX and HI-UX@.
15944 Use options specific to GNU @command{ld}.
15945 This passes @option{-shared} to @command{ld} when
15946 building a shared library. It is the default when GCC is configured,
15947 explicitly or implicitly, with the GNU linker. This option does not
15948 affect which @command{ld} is called; it only changes what parameters
15949 are passed to that @command{ld}.
15950 The @command{ld} that is called is determined by the
15951 @option{--with-ld} configure option, GCC's program search path, and
15952 finally by the user's @env{PATH}. The linker used by GCC can be printed
15953 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
15954 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15958 Use options specific to HP @command{ld}.
15959 This passes @option{-b} to @command{ld} when building
15960 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
15961 links. It is the default when GCC is configured, explicitly or
15962 implicitly, with the HP linker. This option does not affect
15963 which @command{ld} is called; it only changes what parameters are passed to that
15965 The @command{ld} that is called is determined by the @option{--with-ld}
15966 configure option, GCC's program search path, and finally by the user's
15967 @env{PATH}. The linker used by GCC can be printed using @samp{which
15968 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
15969 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
15972 @opindex mno-long-calls
15973 Generate code that uses long call sequences. This ensures that a call
15974 is always able to reach linker generated stubs. The default is to generate
15975 long calls only when the distance from the call site to the beginning
15976 of the function or translation unit, as the case may be, exceeds a
15977 predefined limit set by the branch type being used. The limits for
15978 normal calls are 7,600,000 and 240,000 bytes, respectively for the
15979 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
15982 Distances are measured from the beginning of functions when using the
15983 @option{-ffunction-sections} option, or when using the @option{-mgas}
15984 and @option{-mno-portable-runtime} options together under HP-UX with
15987 It is normally not desirable to use this option as it degrades
15988 performance. However, it may be useful in large applications,
15989 particularly when partial linking is used to build the application.
15991 The types of long calls used depends on the capabilities of the
15992 assembler and linker, and the type of code being generated. The
15993 impact on systems that support long absolute calls, and long pic
15994 symbol-difference or pc-relative calls should be relatively small.
15995 However, an indirect call is used on 32-bit ELF systems in pic code
15996 and it is quite long.
15998 @item -munix=@var{unix-std}
16000 Generate compiler predefines and select a startfile for the specified
16001 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
16002 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
16003 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
16004 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
16005 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
16008 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
16009 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
16010 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
16011 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
16012 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
16013 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
16015 It is @emph{important} to note that this option changes the interfaces
16016 for various library routines. It also affects the operational behavior
16017 of the C library. Thus, @emph{extreme} care is needed in using this
16020 Library code that is intended to operate with more than one UNIX
16021 standard must test, set and restore the variable @code{__xpg4_extended_mask}
16022 as appropriate. Most GNU software doesn't provide this capability.
16026 Suppress the generation of link options to search libdld.sl when the
16027 @option{-static} option is specified on HP-UX 10 and later.
16031 The HP-UX implementation of setlocale in libc has a dependency on
16032 libdld.sl. There isn't an archive version of libdld.sl. Thus,
16033 when the @option{-static} option is specified, special link options
16034 are needed to resolve this dependency.
16036 On HP-UX 10 and later, the GCC driver adds the necessary options to
16037 link with libdld.sl when the @option{-static} option is specified.
16038 This causes the resulting binary to be dynamic. On the 64-bit port,
16039 the linkers generate dynamic binaries by default in any case. The
16040 @option{-nolibdld} option can be used to prevent the GCC driver from
16041 adding these link options.
16045 Add support for multithreading with the @dfn{dce thread} library
16046 under HP-UX@. This option sets flags for both the preprocessor and
16050 @node IA-64 Options
16051 @subsection IA-64 Options
16052 @cindex IA-64 Options
16054 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16058 @opindex mbig-endian
16059 Generate code for a big-endian target. This is the default for HP-UX@.
16061 @item -mlittle-endian
16062 @opindex mlittle-endian
16063 Generate code for a little-endian target. This is the default for AIX5
16069 @opindex mno-gnu-as
16070 Generate (or don't) code for the GNU assembler. This is the default.
16071 @c Also, this is the default if the configure option @option{--with-gnu-as}
16077 @opindex mno-gnu-ld
16078 Generate (or don't) code for the GNU linker. This is the default.
16079 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16084 Generate code that does not use a global pointer register. The result
16085 is not position independent code, and violates the IA-64 ABI@.
16087 @item -mvolatile-asm-stop
16088 @itemx -mno-volatile-asm-stop
16089 @opindex mvolatile-asm-stop
16090 @opindex mno-volatile-asm-stop
16091 Generate (or don't) a stop bit immediately before and after volatile asm
16094 @item -mregister-names
16095 @itemx -mno-register-names
16096 @opindex mregister-names
16097 @opindex mno-register-names
16098 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16099 the stacked registers. This may make assembler output more readable.
16105 Disable (or enable) optimizations that use the small data section. This may
16106 be useful for working around optimizer bugs.
16108 @item -mconstant-gp
16109 @opindex mconstant-gp
16110 Generate code that uses a single constant global pointer value. This is
16111 useful when compiling kernel code.
16115 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16116 This is useful when compiling firmware code.
16118 @item -minline-float-divide-min-latency
16119 @opindex minline-float-divide-min-latency
16120 Generate code for inline divides of floating-point values
16121 using the minimum latency algorithm.
16123 @item -minline-float-divide-max-throughput
16124 @opindex minline-float-divide-max-throughput
16125 Generate code for inline divides of floating-point values
16126 using the maximum throughput algorithm.
16128 @item -mno-inline-float-divide
16129 @opindex mno-inline-float-divide
16130 Do not generate inline code for divides of floating-point values.
16132 @item -minline-int-divide-min-latency
16133 @opindex minline-int-divide-min-latency
16134 Generate code for inline divides of integer values
16135 using the minimum latency algorithm.
16137 @item -minline-int-divide-max-throughput
16138 @opindex minline-int-divide-max-throughput
16139 Generate code for inline divides of integer values
16140 using the maximum throughput algorithm.
16142 @item -mno-inline-int-divide
16143 @opindex mno-inline-int-divide
16144 Do not generate inline code for divides of integer values.
16146 @item -minline-sqrt-min-latency
16147 @opindex minline-sqrt-min-latency
16148 Generate code for inline square roots
16149 using the minimum latency algorithm.
16151 @item -minline-sqrt-max-throughput
16152 @opindex minline-sqrt-max-throughput
16153 Generate code for inline square roots
16154 using the maximum throughput algorithm.
16156 @item -mno-inline-sqrt
16157 @opindex mno-inline-sqrt
16158 Do not generate inline code for @code{sqrt}.
16161 @itemx -mno-fused-madd
16162 @opindex mfused-madd
16163 @opindex mno-fused-madd
16164 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16165 instructions. The default is to use these instructions.
16167 @item -mno-dwarf2-asm
16168 @itemx -mdwarf2-asm
16169 @opindex mno-dwarf2-asm
16170 @opindex mdwarf2-asm
16171 Don't (or do) generate assembler code for the DWARF 2 line number debugging
16172 info. This may be useful when not using the GNU assembler.
16174 @item -mearly-stop-bits
16175 @itemx -mno-early-stop-bits
16176 @opindex mearly-stop-bits
16177 @opindex mno-early-stop-bits
16178 Allow stop bits to be placed earlier than immediately preceding the
16179 instruction that triggered the stop bit. This can improve instruction
16180 scheduling, but does not always do so.
16182 @item -mfixed-range=@var{register-range}
16183 @opindex mfixed-range
16184 Generate code treating the given register range as fixed registers.
16185 A fixed register is one that the register allocator cannot use. This is
16186 useful when compiling kernel code. A register range is specified as
16187 two registers separated by a dash. Multiple register ranges can be
16188 specified separated by a comma.
16190 @item -mtls-size=@var{tls-size}
16192 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16195 @item -mtune=@var{cpu-type}
16197 Tune the instruction scheduling for a particular CPU, Valid values are
16198 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16199 and @samp{mckinley}.
16205 Generate code for a 32-bit or 64-bit environment.
16206 The 32-bit environment sets int, long and pointer to 32 bits.
16207 The 64-bit environment sets int to 32 bits and long and pointer
16208 to 64 bits. These are HP-UX specific flags.
16210 @item -mno-sched-br-data-spec
16211 @itemx -msched-br-data-spec
16212 @opindex mno-sched-br-data-spec
16213 @opindex msched-br-data-spec
16214 (Dis/En)able data speculative scheduling before reload.
16215 This results in generation of @code{ld.a} instructions and
16216 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16217 The default is 'disable'.
16219 @item -msched-ar-data-spec
16220 @itemx -mno-sched-ar-data-spec
16221 @opindex msched-ar-data-spec
16222 @opindex mno-sched-ar-data-spec
16223 (En/Dis)able data speculative scheduling after reload.
16224 This results in generation of @code{ld.a} instructions and
16225 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16226 The default is 'enable'.
16228 @item -mno-sched-control-spec
16229 @itemx -msched-control-spec
16230 @opindex mno-sched-control-spec
16231 @opindex msched-control-spec
16232 (Dis/En)able control speculative scheduling. This feature is
16233 available only during region scheduling (i.e.@: before reload).
16234 This results in generation of the @code{ld.s} instructions and
16235 the corresponding check instructions @code{chk.s}.
16236 The default is 'disable'.
16238 @item -msched-br-in-data-spec
16239 @itemx -mno-sched-br-in-data-spec
16240 @opindex msched-br-in-data-spec
16241 @opindex mno-sched-br-in-data-spec
16242 (En/Dis)able speculative scheduling of the instructions that
16243 are dependent on the data speculative loads before reload.
16244 This is effective only with @option{-msched-br-data-spec} enabled.
16245 The default is 'enable'.
16247 @item -msched-ar-in-data-spec
16248 @itemx -mno-sched-ar-in-data-spec
16249 @opindex msched-ar-in-data-spec
16250 @opindex mno-sched-ar-in-data-spec
16251 (En/Dis)able speculative scheduling of the instructions that
16252 are dependent on the data speculative loads after reload.
16253 This is effective only with @option{-msched-ar-data-spec} enabled.
16254 The default is 'enable'.
16256 @item -msched-in-control-spec
16257 @itemx -mno-sched-in-control-spec
16258 @opindex msched-in-control-spec
16259 @opindex mno-sched-in-control-spec
16260 (En/Dis)able speculative scheduling of the instructions that
16261 are dependent on the control speculative loads.
16262 This is effective only with @option{-msched-control-spec} enabled.
16263 The default is 'enable'.
16265 @item -mno-sched-prefer-non-data-spec-insns
16266 @itemx -msched-prefer-non-data-spec-insns
16267 @opindex mno-sched-prefer-non-data-spec-insns
16268 @opindex msched-prefer-non-data-spec-insns
16269 If enabled, data-speculative instructions are chosen for schedule
16270 only if there are no other choices at the moment. This makes
16271 the use of the data speculation much more conservative.
16272 The default is 'disable'.
16274 @item -mno-sched-prefer-non-control-spec-insns
16275 @itemx -msched-prefer-non-control-spec-insns
16276 @opindex mno-sched-prefer-non-control-spec-insns
16277 @opindex msched-prefer-non-control-spec-insns
16278 If enabled, control-speculative instructions are chosen for schedule
16279 only if there are no other choices at the moment. This makes
16280 the use of the control speculation much more conservative.
16281 The default is 'disable'.
16283 @item -mno-sched-count-spec-in-critical-path
16284 @itemx -msched-count-spec-in-critical-path
16285 @opindex mno-sched-count-spec-in-critical-path
16286 @opindex msched-count-spec-in-critical-path
16287 If enabled, speculative dependencies are considered during
16288 computation of the instructions priorities. This makes the use of the
16289 speculation a bit more conservative.
16290 The default is 'disable'.
16292 @item -msched-spec-ldc
16293 @opindex msched-spec-ldc
16294 Use a simple data speculation check. This option is on by default.
16296 @item -msched-control-spec-ldc
16297 @opindex msched-spec-ldc
16298 Use a simple check for control speculation. This option is on by default.
16300 @item -msched-stop-bits-after-every-cycle
16301 @opindex msched-stop-bits-after-every-cycle
16302 Place a stop bit after every cycle when scheduling. This option is on
16305 @item -msched-fp-mem-deps-zero-cost
16306 @opindex msched-fp-mem-deps-zero-cost
16307 Assume that floating-point stores and loads are not likely to cause a conflict
16308 when placed into the same instruction group. This option is disabled by
16311 @item -msel-sched-dont-check-control-spec
16312 @opindex msel-sched-dont-check-control-spec
16313 Generate checks for control speculation in selective scheduling.
16314 This flag is disabled by default.
16316 @item -msched-max-memory-insns=@var{max-insns}
16317 @opindex msched-max-memory-insns
16318 Limit on the number of memory insns per instruction group, giving lower
16319 priority to subsequent memory insns attempting to schedule in the same
16320 instruction group. Frequently useful to prevent cache bank conflicts.
16321 The default value is 1.
16323 @item -msched-max-memory-insns-hard-limit
16324 @opindex msched-max-memory-insns-hard-limit
16325 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16326 disallowing more than that number in an instruction group.
16327 Otherwise, the limit is ``soft'', meaning that non-memory operations
16328 are preferred when the limit is reached, but memory operations may still
16334 @subsection LM32 Options
16335 @cindex LM32 options
16337 These @option{-m} options are defined for the LatticeMico32 architecture:
16340 @item -mbarrel-shift-enabled
16341 @opindex mbarrel-shift-enabled
16342 Enable barrel-shift instructions.
16344 @item -mdivide-enabled
16345 @opindex mdivide-enabled
16346 Enable divide and modulus instructions.
16348 @item -mmultiply-enabled
16349 @opindex multiply-enabled
16350 Enable multiply instructions.
16352 @item -msign-extend-enabled
16353 @opindex msign-extend-enabled
16354 Enable sign extend instructions.
16356 @item -muser-enabled
16357 @opindex muser-enabled
16358 Enable user-defined instructions.
16363 @subsection M32C Options
16364 @cindex M32C options
16367 @item -mcpu=@var{name}
16369 Select the CPU for which code is generated. @var{name} may be one of
16370 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16371 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16372 the M32C/80 series.
16376 Specifies that the program will be run on the simulator. This causes
16377 an alternate runtime library to be linked in which supports, for
16378 example, file I/O@. You must not use this option when generating
16379 programs that will run on real hardware; you must provide your own
16380 runtime library for whatever I/O functions are needed.
16382 @item -memregs=@var{number}
16384 Specifies the number of memory-based pseudo-registers GCC uses
16385 during code generation. These pseudo-registers are used like real
16386 registers, so there is a tradeoff between GCC's ability to fit the
16387 code into available registers, and the performance penalty of using
16388 memory instead of registers. Note that all modules in a program must
16389 be compiled with the same value for this option. Because of that, you
16390 must not use this option with GCC's default runtime libraries.
16394 @node M32R/D Options
16395 @subsection M32R/D Options
16396 @cindex M32R/D options
16398 These @option{-m} options are defined for Renesas M32R/D architectures:
16403 Generate code for the M32R/2@.
16407 Generate code for the M32R/X@.
16411 Generate code for the M32R@. This is the default.
16413 @item -mmodel=small
16414 @opindex mmodel=small
16415 Assume all objects live in the lower 16MB of memory (so that their addresses
16416 can be loaded with the @code{ld24} instruction), and assume all subroutines
16417 are reachable with the @code{bl} instruction.
16418 This is the default.
16420 The addressability of a particular object can be set with the
16421 @code{model} attribute.
16423 @item -mmodel=medium
16424 @opindex mmodel=medium
16425 Assume objects may be anywhere in the 32-bit address space (the compiler
16426 generates @code{seth/add3} instructions to load their addresses), and
16427 assume all subroutines are reachable with the @code{bl} instruction.
16429 @item -mmodel=large
16430 @opindex mmodel=large
16431 Assume objects may be anywhere in the 32-bit address space (the compiler
16432 generates @code{seth/add3} instructions to load their addresses), and
16433 assume subroutines may not be reachable with the @code{bl} instruction
16434 (the compiler generates the much slower @code{seth/add3/jl}
16435 instruction sequence).
16438 @opindex msdata=none
16439 Disable use of the small data area. Variables are put into
16440 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
16441 @code{section} attribute has been specified).
16442 This is the default.
16444 The small data area consists of sections @code{.sdata} and @code{.sbss}.
16445 Objects may be explicitly put in the small data area with the
16446 @code{section} attribute using one of these sections.
16448 @item -msdata=sdata
16449 @opindex msdata=sdata
16450 Put small global and static data in the small data area, but do not
16451 generate special code to reference them.
16454 @opindex msdata=use
16455 Put small global and static data in the small data area, and generate
16456 special instructions to reference them.
16460 @cindex smaller data references
16461 Put global and static objects less than or equal to @var{num} bytes
16462 into the small data or BSS sections instead of the normal data or BSS
16463 sections. The default value of @var{num} is 8.
16464 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
16465 for this option to have any effect.
16467 All modules should be compiled with the same @option{-G @var{num}} value.
16468 Compiling with different values of @var{num} may or may not work; if it
16469 doesn't the linker gives an error message---incorrect code is not
16474 Makes the M32R-specific code in the compiler display some statistics
16475 that might help in debugging programs.
16477 @item -malign-loops
16478 @opindex malign-loops
16479 Align all loops to a 32-byte boundary.
16481 @item -mno-align-loops
16482 @opindex mno-align-loops
16483 Do not enforce a 32-byte alignment for loops. This is the default.
16485 @item -missue-rate=@var{number}
16486 @opindex missue-rate=@var{number}
16487 Issue @var{number} instructions per cycle. @var{number} can only be 1
16490 @item -mbranch-cost=@var{number}
16491 @opindex mbranch-cost=@var{number}
16492 @var{number} can only be 1 or 2. If it is 1 then branches are
16493 preferred over conditional code, if it is 2, then the opposite applies.
16495 @item -mflush-trap=@var{number}
16496 @opindex mflush-trap=@var{number}
16497 Specifies the trap number to use to flush the cache. The default is
16498 12. Valid numbers are between 0 and 15 inclusive.
16500 @item -mno-flush-trap
16501 @opindex mno-flush-trap
16502 Specifies that the cache cannot be flushed by using a trap.
16504 @item -mflush-func=@var{name}
16505 @opindex mflush-func=@var{name}
16506 Specifies the name of the operating system function to call to flush
16507 the cache. The default is @samp{_flush_cache}, but a function call
16508 is only used if a trap is not available.
16510 @item -mno-flush-func
16511 @opindex mno-flush-func
16512 Indicates that there is no OS function for flushing the cache.
16516 @node M680x0 Options
16517 @subsection M680x0 Options
16518 @cindex M680x0 options
16520 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
16521 The default settings depend on which architecture was selected when
16522 the compiler was configured; the defaults for the most common choices
16526 @item -march=@var{arch}
16528 Generate code for a specific M680x0 or ColdFire instruction set
16529 architecture. Permissible values of @var{arch} for M680x0
16530 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
16531 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
16532 architectures are selected according to Freescale's ISA classification
16533 and the permissible values are: @samp{isaa}, @samp{isaaplus},
16534 @samp{isab} and @samp{isac}.
16536 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
16537 code for a ColdFire target. The @var{arch} in this macro is one of the
16538 @option{-march} arguments given above.
16540 When used together, @option{-march} and @option{-mtune} select code
16541 that runs on a family of similar processors but that is optimized
16542 for a particular microarchitecture.
16544 @item -mcpu=@var{cpu}
16546 Generate code for a specific M680x0 or ColdFire processor.
16547 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
16548 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
16549 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
16550 below, which also classifies the CPUs into families:
16552 @multitable @columnfractions 0.20 0.80
16553 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
16554 @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}
16555 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
16556 @item @samp{5206e} @tab @samp{5206e}
16557 @item @samp{5208} @tab @samp{5207} @samp{5208}
16558 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
16559 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
16560 @item @samp{5216} @tab @samp{5214} @samp{5216}
16561 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
16562 @item @samp{5225} @tab @samp{5224} @samp{5225}
16563 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
16564 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
16565 @item @samp{5249} @tab @samp{5249}
16566 @item @samp{5250} @tab @samp{5250}
16567 @item @samp{5271} @tab @samp{5270} @samp{5271}
16568 @item @samp{5272} @tab @samp{5272}
16569 @item @samp{5275} @tab @samp{5274} @samp{5275}
16570 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
16571 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
16572 @item @samp{5307} @tab @samp{5307}
16573 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
16574 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
16575 @item @samp{5407} @tab @samp{5407}
16576 @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}
16579 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
16580 @var{arch} is compatible with @var{cpu}. Other combinations of
16581 @option{-mcpu} and @option{-march} are rejected.
16583 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
16584 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
16585 where the value of @var{family} is given by the table above.
16587 @item -mtune=@var{tune}
16589 Tune the code for a particular microarchitecture within the
16590 constraints set by @option{-march} and @option{-mcpu}.
16591 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
16592 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
16593 and @samp{cpu32}. The ColdFire microarchitectures
16594 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
16596 You can also use @option{-mtune=68020-40} for code that needs
16597 to run relatively well on 68020, 68030 and 68040 targets.
16598 @option{-mtune=68020-60} is similar but includes 68060 targets
16599 as well. These two options select the same tuning decisions as
16600 @option{-m68020-40} and @option{-m68020-60} respectively.
16602 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
16603 when tuning for 680x0 architecture @var{arch}. It also defines
16604 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
16605 option is used. If GCC is tuning for a range of architectures,
16606 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
16607 it defines the macros for every architecture in the range.
16609 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
16610 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
16611 of the arguments given above.
16617 Generate output for a 68000. This is the default
16618 when the compiler is configured for 68000-based systems.
16619 It is equivalent to @option{-march=68000}.
16621 Use this option for microcontrollers with a 68000 or EC000 core,
16622 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
16626 Generate output for a 68010. This is the default
16627 when the compiler is configured for 68010-based systems.
16628 It is equivalent to @option{-march=68010}.
16634 Generate output for a 68020. This is the default
16635 when the compiler is configured for 68020-based systems.
16636 It is equivalent to @option{-march=68020}.
16640 Generate output for a 68030. This is the default when the compiler is
16641 configured for 68030-based systems. It is equivalent to
16642 @option{-march=68030}.
16646 Generate output for a 68040. This is the default when the compiler is
16647 configured for 68040-based systems. It is equivalent to
16648 @option{-march=68040}.
16650 This option inhibits the use of 68881/68882 instructions that have to be
16651 emulated by software on the 68040. Use this option if your 68040 does not
16652 have code to emulate those instructions.
16656 Generate output for a 68060. This is the default when the compiler is
16657 configured for 68060-based systems. It is equivalent to
16658 @option{-march=68060}.
16660 This option inhibits the use of 68020 and 68881/68882 instructions that
16661 have to be emulated by software on the 68060. Use this option if your 68060
16662 does not have code to emulate those instructions.
16666 Generate output for a CPU32. This is the default
16667 when the compiler is configured for CPU32-based systems.
16668 It is equivalent to @option{-march=cpu32}.
16670 Use this option for microcontrollers with a
16671 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
16672 68336, 68340, 68341, 68349 and 68360.
16676 Generate output for a 520X ColdFire CPU@. This is the default
16677 when the compiler is configured for 520X-based systems.
16678 It is equivalent to @option{-mcpu=5206}, and is now deprecated
16679 in favor of that option.
16681 Use this option for microcontroller with a 5200 core, including
16682 the MCF5202, MCF5203, MCF5204 and MCF5206.
16686 Generate output for a 5206e ColdFire CPU@. The option is now
16687 deprecated in favor of the equivalent @option{-mcpu=5206e}.
16691 Generate output for a member of the ColdFire 528X family.
16692 The option is now deprecated in favor of the equivalent
16693 @option{-mcpu=528x}.
16697 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
16698 in favor of the equivalent @option{-mcpu=5307}.
16702 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
16703 in favor of the equivalent @option{-mcpu=5407}.
16707 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
16708 This includes use of hardware floating-point instructions.
16709 The option is equivalent to @option{-mcpu=547x}, and is now
16710 deprecated in favor of that option.
16714 Generate output for a 68040, without using any of the new instructions.
16715 This results in code that can run relatively efficiently on either a
16716 68020/68881 or a 68030 or a 68040. The generated code does use the
16717 68881 instructions that are emulated on the 68040.
16719 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
16723 Generate output for a 68060, without using any of the new instructions.
16724 This results in code that can run relatively efficiently on either a
16725 68020/68881 or a 68030 or a 68040. The generated code does use the
16726 68881 instructions that are emulated on the 68060.
16728 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
16732 @opindex mhard-float
16734 Generate floating-point instructions. This is the default for 68020
16735 and above, and for ColdFire devices that have an FPU@. It defines the
16736 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
16737 on ColdFire targets.
16740 @opindex msoft-float
16741 Do not generate floating-point instructions; use library calls instead.
16742 This is the default for 68000, 68010, and 68832 targets. It is also
16743 the default for ColdFire devices that have no FPU.
16749 Generate (do not generate) ColdFire hardware divide and remainder
16750 instructions. If @option{-march} is used without @option{-mcpu},
16751 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
16752 architectures. Otherwise, the default is taken from the target CPU
16753 (either the default CPU, or the one specified by @option{-mcpu}). For
16754 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
16755 @option{-mcpu=5206e}.
16757 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
16761 Consider type @code{int} to be 16 bits wide, like @code{short int}.
16762 Additionally, parameters passed on the stack are also aligned to a
16763 16-bit boundary even on targets whose API mandates promotion to 32-bit.
16767 Do not consider type @code{int} to be 16 bits wide. This is the default.
16770 @itemx -mno-bitfield
16771 @opindex mnobitfield
16772 @opindex mno-bitfield
16773 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
16774 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
16778 Do use the bit-field instructions. The @option{-m68020} option implies
16779 @option{-mbitfield}. This is the default if you use a configuration
16780 designed for a 68020.
16784 Use a different function-calling convention, in which functions
16785 that take a fixed number of arguments return with the @code{rtd}
16786 instruction, which pops their arguments while returning. This
16787 saves one instruction in the caller since there is no need to pop
16788 the arguments there.
16790 This calling convention is incompatible with the one normally
16791 used on Unix, so you cannot use it if you need to call libraries
16792 compiled with the Unix compiler.
16794 Also, you must provide function prototypes for all functions that
16795 take variable numbers of arguments (including @code{printf});
16796 otherwise incorrect code is generated for calls to those
16799 In addition, seriously incorrect code results if you call a
16800 function with too many arguments. (Normally, extra arguments are
16801 harmlessly ignored.)
16803 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
16804 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
16808 Do not use the calling conventions selected by @option{-mrtd}.
16809 This is the default.
16812 @itemx -mno-align-int
16813 @opindex malign-int
16814 @opindex mno-align-int
16815 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
16816 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
16817 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
16818 Aligning variables on 32-bit boundaries produces code that runs somewhat
16819 faster on processors with 32-bit busses at the expense of more memory.
16821 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
16822 aligns structures containing the above types differently than
16823 most published application binary interface specifications for the m68k.
16827 Use the pc-relative addressing mode of the 68000 directly, instead of
16828 using a global offset table. At present, this option implies @option{-fpic},
16829 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
16830 not presently supported with @option{-mpcrel}, though this could be supported for
16831 68020 and higher processors.
16833 @item -mno-strict-align
16834 @itemx -mstrict-align
16835 @opindex mno-strict-align
16836 @opindex mstrict-align
16837 Do not (do) assume that unaligned memory references are handled by
16841 Generate code that allows the data segment to be located in a different
16842 area of memory from the text segment. This allows for execute-in-place in
16843 an environment without virtual memory management. This option implies
16846 @item -mno-sep-data
16847 Generate code that assumes that the data segment follows the text segment.
16848 This is the default.
16850 @item -mid-shared-library
16851 Generate code that supports shared libraries via the library ID method.
16852 This allows for execute-in-place and shared libraries in an environment
16853 without virtual memory management. This option implies @option{-fPIC}.
16855 @item -mno-id-shared-library
16856 Generate code that doesn't assume ID-based shared libraries are being used.
16857 This is the default.
16859 @item -mshared-library-id=n
16860 Specifies the identification number of the ID-based shared library being
16861 compiled. Specifying a value of 0 generates more compact code; specifying
16862 other values forces the allocation of that number to the current
16863 library, but is no more space- or time-efficient than omitting this option.
16869 When generating position-independent code for ColdFire, generate code
16870 that works if the GOT has more than 8192 entries. This code is
16871 larger and slower than code generated without this option. On M680x0
16872 processors, this option is not needed; @option{-fPIC} suffices.
16874 GCC normally uses a single instruction to load values from the GOT@.
16875 While this is relatively efficient, it only works if the GOT
16876 is smaller than about 64k. Anything larger causes the linker
16877 to report an error such as:
16879 @cindex relocation truncated to fit (ColdFire)
16881 relocation truncated to fit: R_68K_GOT16O foobar
16884 If this happens, you should recompile your code with @option{-mxgot}.
16885 It should then work with very large GOTs. However, code generated with
16886 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
16887 the value of a global symbol.
16889 Note that some linkers, including newer versions of the GNU linker,
16890 can create multiple GOTs and sort GOT entries. If you have such a linker,
16891 you should only need to use @option{-mxgot} when compiling a single
16892 object file that accesses more than 8192 GOT entries. Very few do.
16894 These options have no effect unless GCC is generating
16895 position-independent code.
16899 @node MCore Options
16900 @subsection MCore Options
16901 @cindex MCore options
16903 These are the @samp{-m} options defined for the Motorola M*Core
16909 @itemx -mno-hardlit
16911 @opindex mno-hardlit
16912 Inline constants into the code stream if it can be done in two
16913 instructions or less.
16919 Use the divide instruction. (Enabled by default).
16921 @item -mrelax-immediate
16922 @itemx -mno-relax-immediate
16923 @opindex mrelax-immediate
16924 @opindex mno-relax-immediate
16925 Allow arbitrary-sized immediates in bit operations.
16927 @item -mwide-bitfields
16928 @itemx -mno-wide-bitfields
16929 @opindex mwide-bitfields
16930 @opindex mno-wide-bitfields
16931 Always treat bit-fields as @code{int}-sized.
16933 @item -m4byte-functions
16934 @itemx -mno-4byte-functions
16935 @opindex m4byte-functions
16936 @opindex mno-4byte-functions
16937 Force all functions to be aligned to a 4-byte boundary.
16939 @item -mcallgraph-data
16940 @itemx -mno-callgraph-data
16941 @opindex mcallgraph-data
16942 @opindex mno-callgraph-data
16943 Emit callgraph information.
16946 @itemx -mno-slow-bytes
16947 @opindex mslow-bytes
16948 @opindex mno-slow-bytes
16949 Prefer word access when reading byte quantities.
16951 @item -mlittle-endian
16952 @itemx -mbig-endian
16953 @opindex mlittle-endian
16954 @opindex mbig-endian
16955 Generate code for a little-endian target.
16961 Generate code for the 210 processor.
16965 Assume that runtime support has been provided and so omit the
16966 simulator library (@file{libsim.a)} from the linker command line.
16968 @item -mstack-increment=@var{size}
16969 @opindex mstack-increment
16970 Set the maximum amount for a single stack increment operation. Large
16971 values can increase the speed of programs that contain functions
16972 that need a large amount of stack space, but they can also trigger a
16973 segmentation fault if the stack is extended too much. The default
16979 @subsection MeP Options
16980 @cindex MeP options
16986 Enables the @code{abs} instruction, which is the absolute difference
16987 between two registers.
16991 Enables all the optional instructions---average, multiply, divide, bit
16992 operations, leading zero, absolute difference, min/max, clip, and
16998 Enables the @code{ave} instruction, which computes the average of two
17001 @item -mbased=@var{n}
17003 Variables of size @var{n} bytes or smaller are placed in the
17004 @code{.based} section by default. Based variables use the @code{$tp}
17005 register as a base register, and there is a 128-byte limit to the
17006 @code{.based} section.
17010 Enables the bit operation instructions---bit test (@code{btstm}), set
17011 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17012 test-and-set (@code{tas}).
17014 @item -mc=@var{name}
17016 Selects which section constant data is placed in. @var{name} may
17017 be @samp{tiny}, @samp{near}, or @samp{far}.
17021 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17022 useful unless you also provide @option{-mminmax}.
17024 @item -mconfig=@var{name}
17026 Selects one of the built-in core configurations. Each MeP chip has
17027 one or more modules in it; each module has a core CPU and a variety of
17028 coprocessors, optional instructions, and peripherals. The
17029 @code{MeP-Integrator} tool, not part of GCC, provides these
17030 configurations through this option; using this option is the same as
17031 using all the corresponding command-line options. The default
17032 configuration is @samp{default}.
17036 Enables the coprocessor instructions. By default, this is a 32-bit
17037 coprocessor. Note that the coprocessor is normally enabled via the
17038 @option{-mconfig=} option.
17042 Enables the 32-bit coprocessor's instructions.
17046 Enables the 64-bit coprocessor's instructions.
17050 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17054 Causes constant variables to be placed in the @code{.near} section.
17058 Enables the @code{div} and @code{divu} instructions.
17062 Generate big-endian code.
17066 Generate little-endian code.
17068 @item -mio-volatile
17069 @opindex mio-volatile
17070 Tells the compiler that any variable marked with the @code{io}
17071 attribute is to be considered volatile.
17075 Causes variables to be assigned to the @code{.far} section by default.
17079 Enables the @code{leadz} (leading zero) instruction.
17083 Causes variables to be assigned to the @code{.near} section by default.
17087 Enables the @code{min} and @code{max} instructions.
17091 Enables the multiplication and multiply-accumulate instructions.
17095 Disables all the optional instructions enabled by @option{-mall-opts}.
17099 Enables the @code{repeat} and @code{erepeat} instructions, used for
17100 low-overhead looping.
17104 Causes all variables to default to the @code{.tiny} section. Note
17105 that there is a 65536-byte limit to this section. Accesses to these
17106 variables use the @code{%gp} base register.
17110 Enables the saturation instructions. Note that the compiler does not
17111 currently generate these itself, but this option is included for
17112 compatibility with other tools, like @code{as}.
17116 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17120 Link the simulator run-time libraries.
17124 Link the simulator runtime libraries, excluding built-in support
17125 for reset and exception vectors and tables.
17129 Causes all functions to default to the @code{.far} section. Without
17130 this option, functions default to the @code{.near} section.
17132 @item -mtiny=@var{n}
17134 Variables that are @var{n} bytes or smaller are allocated to the
17135 @code{.tiny} section. These variables use the @code{$gp} base
17136 register. The default for this option is 4, but note that there's a
17137 65536-byte limit to the @code{.tiny} section.
17141 @node MicroBlaze Options
17142 @subsection MicroBlaze Options
17143 @cindex MicroBlaze Options
17148 @opindex msoft-float
17149 Use software emulation for floating point (default).
17152 @opindex mhard-float
17153 Use hardware floating-point instructions.
17157 Do not optimize block moves, use @code{memcpy}.
17159 @item -mno-clearbss
17160 @opindex mno-clearbss
17161 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17163 @item -mcpu=@var{cpu-type}
17165 Use features of, and schedule code for, the given CPU.
17166 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17167 where @var{X} is a major version, @var{YY} is the minor version, and
17168 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17169 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17171 @item -mxl-soft-mul
17172 @opindex mxl-soft-mul
17173 Use software multiply emulation (default).
17175 @item -mxl-soft-div
17176 @opindex mxl-soft-div
17177 Use software emulation for divides (default).
17179 @item -mxl-barrel-shift
17180 @opindex mxl-barrel-shift
17181 Use the hardware barrel shifter.
17183 @item -mxl-pattern-compare
17184 @opindex mxl-pattern-compare
17185 Use pattern compare instructions.
17187 @item -msmall-divides
17188 @opindex msmall-divides
17189 Use table lookup optimization for small signed integer divisions.
17191 @item -mxl-stack-check
17192 @opindex mxl-stack-check
17193 This option is deprecated. Use @option{-fstack-check} instead.
17196 @opindex mxl-gp-opt
17197 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17199 @item -mxl-multiply-high
17200 @opindex mxl-multiply-high
17201 Use multiply high instructions for high part of 32x32 multiply.
17203 @item -mxl-float-convert
17204 @opindex mxl-float-convert
17205 Use hardware floating-point conversion instructions.
17207 @item -mxl-float-sqrt
17208 @opindex mxl-float-sqrt
17209 Use hardware floating-point square root instruction.
17212 @opindex mbig-endian
17213 Generate code for a big-endian target.
17215 @item -mlittle-endian
17216 @opindex mlittle-endian
17217 Generate code for a little-endian target.
17220 @opindex mxl-reorder
17221 Use reorder instructions (swap and byte reversed load/store).
17223 @item -mxl-mode-@var{app-model}
17224 Select application model @var{app-model}. Valid models are
17227 normal executable (default), uses startup code @file{crt0.o}.
17230 for use with Xilinx Microprocessor Debugger (XMD) based
17231 software intrusive debug agent called xmdstub. This uses startup file
17232 @file{crt1.o} and sets the start address of the program to 0x800.
17235 for applications that are loaded using a bootloader.
17236 This model uses startup file @file{crt2.o} which does not contain a processor
17237 reset vector handler. This is suitable for transferring control on a
17238 processor reset to the bootloader rather than the application.
17241 for applications that do not require any of the
17242 MicroBlaze vectors. This option may be useful for applications running
17243 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17246 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17247 @option{-mxl-mode-@var{app-model}}.
17252 @subsection MIPS Options
17253 @cindex MIPS options
17259 Generate big-endian code.
17263 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17266 @item -march=@var{arch}
17268 Generate code that runs on @var{arch}, which can be the name of a
17269 generic MIPS ISA, or the name of a particular processor.
17271 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17272 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17273 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17274 @samp{mips64r5} and @samp{mips64r6}.
17275 The processor names are:
17276 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17277 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17278 @samp{5kc}, @samp{5kf},
17280 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17281 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17282 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17283 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17284 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17287 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17289 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17290 @samp{m5100}, @samp{m5101},
17291 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17294 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17295 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17296 @samp{rm7000}, @samp{rm9000},
17297 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17300 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17301 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17302 @samp{xlr} and @samp{xlp}.
17303 The special value @samp{from-abi} selects the
17304 most compatible architecture for the selected ABI (that is,
17305 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17307 The native Linux/GNU toolchain also supports the value @samp{native},
17308 which selects the best architecture option for the host processor.
17309 @option{-march=native} has no effect if GCC does not recognize
17312 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17313 (for example, @option{-march=r2k}). Prefixes are optional, and
17314 @samp{vr} may be written @samp{r}.
17316 Names of the form @samp{@var{n}f2_1} refer to processors with
17317 FPUs clocked at half the rate of the core, names of the form
17318 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17319 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17320 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17321 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17322 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17323 accepted as synonyms for @samp{@var{n}f1_1}.
17325 GCC defines two macros based on the value of this option. The first
17326 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17327 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17328 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17329 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17330 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17332 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17333 above. In other words, it has the full prefix and does not
17334 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17335 the macro names the resolved architecture (either @code{"mips1"} or
17336 @code{"mips3"}). It names the default architecture when no
17337 @option{-march} option is given.
17339 @item -mtune=@var{arch}
17341 Optimize for @var{arch}. Among other things, this option controls
17342 the way instructions are scheduled, and the perceived cost of arithmetic
17343 operations. The list of @var{arch} values is the same as for
17346 When this option is not used, GCC optimizes for the processor
17347 specified by @option{-march}. By using @option{-march} and
17348 @option{-mtune} together, it is possible to generate code that
17349 runs on a family of processors, but optimize the code for one
17350 particular member of that family.
17352 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17353 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17354 @option{-march} ones described above.
17358 Equivalent to @option{-march=mips1}.
17362 Equivalent to @option{-march=mips2}.
17366 Equivalent to @option{-march=mips3}.
17370 Equivalent to @option{-march=mips4}.
17374 Equivalent to @option{-march=mips32}.
17378 Equivalent to @option{-march=mips32r3}.
17382 Equivalent to @option{-march=mips32r5}.
17386 Equivalent to @option{-march=mips32r6}.
17390 Equivalent to @option{-march=mips64}.
17394 Equivalent to @option{-march=mips64r2}.
17398 Equivalent to @option{-march=mips64r3}.
17402 Equivalent to @option{-march=mips64r5}.
17406 Equivalent to @option{-march=mips64r6}.
17411 @opindex mno-mips16
17412 Generate (do not generate) MIPS16 code. If GCC is targeting a
17413 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
17415 MIPS16 code generation can also be controlled on a per-function basis
17416 by means of @code{mips16} and @code{nomips16} attributes.
17417 @xref{Function Attributes}, for more information.
17419 @item -mflip-mips16
17420 @opindex mflip-mips16
17421 Generate MIPS16 code on alternating functions. This option is provided
17422 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
17423 not intended for ordinary use in compiling user code.
17425 @item -minterlink-compressed
17426 @item -mno-interlink-compressed
17427 @opindex minterlink-compressed
17428 @opindex mno-interlink-compressed
17429 Require (do not require) that code using the standard (uncompressed) MIPS ISA
17430 be link-compatible with MIPS16 and microMIPS code, and vice versa.
17432 For example, code using the standard ISA encoding cannot jump directly
17433 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
17434 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
17435 knows that the target of the jump is not compressed.
17437 @item -minterlink-mips16
17438 @itemx -mno-interlink-mips16
17439 @opindex minterlink-mips16
17440 @opindex mno-interlink-mips16
17441 Aliases of @option{-minterlink-compressed} and
17442 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
17443 and are retained for backwards compatibility.
17455 Generate code for the given ABI@.
17457 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
17458 generates 64-bit code when you select a 64-bit architecture, but you
17459 can use @option{-mgp32} to get 32-bit code instead.
17461 For information about the O64 ABI, see
17462 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
17464 GCC supports a variant of the o32 ABI in which floating-point registers
17465 are 64 rather than 32 bits wide. You can select this combination with
17466 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
17467 and @code{mfhc1} instructions and is therefore only supported for
17468 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
17470 The register assignments for arguments and return values remain the
17471 same, but each scalar value is passed in a single 64-bit register
17472 rather than a pair of 32-bit registers. For example, scalar
17473 floating-point values are returned in @samp{$f0} only, not a
17474 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
17475 remains the same in that the even-numbered double-precision registers
17478 Two additional variants of the o32 ABI are supported to enable
17479 a transition from 32-bit to 64-bit registers. These are FPXX
17480 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
17481 The FPXX extension mandates that all code must execute correctly
17482 when run using 32-bit or 64-bit registers. The code can be interlinked
17483 with either FP32 or FP64, but not both.
17484 The FP64A extension is similar to the FP64 extension but forbids the
17485 use of odd-numbered single-precision registers. This can be used
17486 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
17487 processors and allows both FP32 and FP64A code to interlink and
17488 run in the same process without changing FPU modes.
17491 @itemx -mno-abicalls
17493 @opindex mno-abicalls
17494 Generate (do not generate) code that is suitable for SVR4-style
17495 dynamic objects. @option{-mabicalls} is the default for SVR4-based
17500 Generate (do not generate) code that is fully position-independent,
17501 and that can therefore be linked into shared libraries. This option
17502 only affects @option{-mabicalls}.
17504 All @option{-mabicalls} code has traditionally been position-independent,
17505 regardless of options like @option{-fPIC} and @option{-fpic}. However,
17506 as an extension, the GNU toolchain allows executables to use absolute
17507 accesses for locally-binding symbols. It can also use shorter GP
17508 initialization sequences and generate direct calls to locally-defined
17509 functions. This mode is selected by @option{-mno-shared}.
17511 @option{-mno-shared} depends on binutils 2.16 or higher and generates
17512 objects that can only be linked by the GNU linker. However, the option
17513 does not affect the ABI of the final executable; it only affects the ABI
17514 of relocatable objects. Using @option{-mno-shared} generally makes
17515 executables both smaller and quicker.
17517 @option{-mshared} is the default.
17523 Assume (do not assume) that the static and dynamic linkers
17524 support PLTs and copy relocations. This option only affects
17525 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
17526 has no effect without @option{-msym32}.
17528 You can make @option{-mplt} the default by configuring
17529 GCC with @option{--with-mips-plt}. The default is
17530 @option{-mno-plt} otherwise.
17536 Lift (do not lift) the usual restrictions on the size of the global
17539 GCC normally uses a single instruction to load values from the GOT@.
17540 While this is relatively efficient, it only works if the GOT
17541 is smaller than about 64k. Anything larger causes the linker
17542 to report an error such as:
17544 @cindex relocation truncated to fit (MIPS)
17546 relocation truncated to fit: R_MIPS_GOT16 foobar
17549 If this happens, you should recompile your code with @option{-mxgot}.
17550 This works with very large GOTs, although the code is also
17551 less efficient, since it takes three instructions to fetch the
17552 value of a global symbol.
17554 Note that some linkers can create multiple GOTs. If you have such a
17555 linker, you should only need to use @option{-mxgot} when a single object
17556 file accesses more than 64k's worth of GOT entries. Very few do.
17558 These options have no effect unless GCC is generating position
17563 Assume that general-purpose registers are 32 bits wide.
17567 Assume that general-purpose registers are 64 bits wide.
17571 Assume that floating-point registers are 32 bits wide.
17575 Assume that floating-point registers are 64 bits wide.
17579 Do not assume the width of floating-point registers.
17582 @opindex mhard-float
17583 Use floating-point coprocessor instructions.
17586 @opindex msoft-float
17587 Do not use floating-point coprocessor instructions. Implement
17588 floating-point calculations using library calls instead.
17592 Equivalent to @option{-msoft-float}, but additionally asserts that the
17593 program being compiled does not perform any floating-point operations.
17594 This option is presently supported only by some bare-metal MIPS
17595 configurations, where it may select a special set of libraries
17596 that lack all floating-point support (including, for example, the
17597 floating-point @code{printf} formats).
17598 If code compiled with @option{-mno-float} accidentally contains
17599 floating-point operations, it is likely to suffer a link-time
17600 or run-time failure.
17602 @item -msingle-float
17603 @opindex msingle-float
17604 Assume that the floating-point coprocessor only supports single-precision
17607 @item -mdouble-float
17608 @opindex mdouble-float
17609 Assume that the floating-point coprocessor supports double-precision
17610 operations. This is the default.
17613 @itemx -mno-odd-spreg
17614 @opindex modd-spreg
17615 @opindex mno-odd-spreg
17616 Enable the use of odd-numbered single-precision floating-point registers
17617 for the o32 ABI. This is the default for processors that are known to
17618 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
17622 @itemx -mabs=legacy
17624 @opindex mabs=legacy
17625 These options control the treatment of the special not-a-number (NaN)
17626 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
17627 @code{neg.@i{fmt}} machine instructions.
17629 By default or when @option{-mabs=legacy} is used the legacy
17630 treatment is selected. In this case these instructions are considered
17631 arithmetic and avoided where correct operation is required and the
17632 input operand might be a NaN. A longer sequence of instructions that
17633 manipulate the sign bit of floating-point datum manually is used
17634 instead unless the @option{-ffinite-math-only} option has also been
17637 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
17638 this case these instructions are considered non-arithmetic and therefore
17639 operating correctly in all cases, including in particular where the
17640 input operand is a NaN. These instructions are therefore always used
17641 for the respective operations.
17644 @itemx -mnan=legacy
17646 @opindex mnan=legacy
17647 These options control the encoding of the special not-a-number (NaN)
17648 IEEE 754 floating-point data.
17650 The @option{-mnan=legacy} option selects the legacy encoding. In this
17651 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
17652 significand field being 0, whereas signalling NaNs (sNaNs) are denoted
17653 by the first bit of their trailing significand field being 1.
17655 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
17656 this case qNaNs are denoted by the first bit of their trailing
17657 significand field being 1, whereas sNaNs are denoted by the first bit of
17658 their trailing significand field being 0.
17660 The default is @option{-mnan=legacy} unless GCC has been configured with
17661 @option{--with-nan=2008}.
17667 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
17668 implement atomic memory built-in functions. When neither option is
17669 specified, GCC uses the instructions if the target architecture
17672 @option{-mllsc} is useful if the runtime environment can emulate the
17673 instructions and @option{-mno-llsc} can be useful when compiling for
17674 nonstandard ISAs. You can make either option the default by
17675 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
17676 respectively. @option{--with-llsc} is the default for some
17677 configurations; see the installation documentation for details.
17683 Use (do not use) revision 1 of the MIPS DSP ASE@.
17684 @xref{MIPS DSP Built-in Functions}. This option defines the
17685 preprocessor macro @code{__mips_dsp}. It also defines
17686 @code{__mips_dsp_rev} to 1.
17692 Use (do not use) revision 2 of the MIPS DSP ASE@.
17693 @xref{MIPS DSP Built-in Functions}. This option defines the
17694 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
17695 It also defines @code{__mips_dsp_rev} to 2.
17698 @itemx -mno-smartmips
17699 @opindex msmartmips
17700 @opindex mno-smartmips
17701 Use (do not use) the MIPS SmartMIPS ASE.
17703 @item -mpaired-single
17704 @itemx -mno-paired-single
17705 @opindex mpaired-single
17706 @opindex mno-paired-single
17707 Use (do not use) paired-single floating-point instructions.
17708 @xref{MIPS Paired-Single Support}. This option requires
17709 hardware floating-point support to be enabled.
17715 Use (do not use) MIPS Digital Media Extension instructions.
17716 This option can only be used when generating 64-bit code and requires
17717 hardware floating-point support to be enabled.
17722 @opindex mno-mips3d
17723 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
17724 The option @option{-mips3d} implies @option{-mpaired-single}.
17727 @itemx -mno-micromips
17728 @opindex mmicromips
17729 @opindex mno-mmicromips
17730 Generate (do not generate) microMIPS code.
17732 MicroMIPS code generation can also be controlled on a per-function basis
17733 by means of @code{micromips} and @code{nomicromips} attributes.
17734 @xref{Function Attributes}, for more information.
17740 Use (do not use) MT Multithreading instructions.
17746 Use (do not use) the MIPS MCU ASE instructions.
17752 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
17758 Use (do not use) the MIPS Virtualization Application Specific instructions.
17764 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
17768 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
17769 an explanation of the default and the way that the pointer size is
17774 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
17776 The default size of @code{int}s, @code{long}s and pointers depends on
17777 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
17778 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
17779 32-bit @code{long}s. Pointers are the same size as @code{long}s,
17780 or the same size as integer registers, whichever is smaller.
17786 Assume (do not assume) that all symbols have 32-bit values, regardless
17787 of the selected ABI@. This option is useful in combination with
17788 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
17789 to generate shorter and faster references to symbolic addresses.
17793 Put definitions of externally-visible data in a small data section
17794 if that data is no bigger than @var{num} bytes. GCC can then generate
17795 more efficient accesses to the data; see @option{-mgpopt} for details.
17797 The default @option{-G} option depends on the configuration.
17799 @item -mlocal-sdata
17800 @itemx -mno-local-sdata
17801 @opindex mlocal-sdata
17802 @opindex mno-local-sdata
17803 Extend (do not extend) the @option{-G} behavior to local data too,
17804 such as to static variables in C@. @option{-mlocal-sdata} is the
17805 default for all configurations.
17807 If the linker complains that an application is using too much small data,
17808 you might want to try rebuilding the less performance-critical parts with
17809 @option{-mno-local-sdata}. You might also want to build large
17810 libraries with @option{-mno-local-sdata}, so that the libraries leave
17811 more room for the main program.
17813 @item -mextern-sdata
17814 @itemx -mno-extern-sdata
17815 @opindex mextern-sdata
17816 @opindex mno-extern-sdata
17817 Assume (do not assume) that externally-defined data is in
17818 a small data section if the size of that data is within the @option{-G} limit.
17819 @option{-mextern-sdata} is the default for all configurations.
17821 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
17822 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
17823 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
17824 is placed in a small data section. If @var{Var} is defined by another
17825 module, you must either compile that module with a high-enough
17826 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
17827 definition. If @var{Var} is common, you must link the application
17828 with a high-enough @option{-G} setting.
17830 The easiest way of satisfying these restrictions is to compile
17831 and link every module with the same @option{-G} option. However,
17832 you may wish to build a library that supports several different
17833 small data limits. You can do this by compiling the library with
17834 the highest supported @option{-G} setting and additionally using
17835 @option{-mno-extern-sdata} to stop the library from making assumptions
17836 about externally-defined data.
17842 Use (do not use) GP-relative accesses for symbols that are known to be
17843 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
17844 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
17847 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
17848 might not hold the value of @code{_gp}. For example, if the code is
17849 part of a library that might be used in a boot monitor, programs that
17850 call boot monitor routines pass an unknown value in @code{$gp}.
17851 (In such situations, the boot monitor itself is usually compiled
17852 with @option{-G0}.)
17854 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
17855 @option{-mno-extern-sdata}.
17857 @item -membedded-data
17858 @itemx -mno-embedded-data
17859 @opindex membedded-data
17860 @opindex mno-embedded-data
17861 Allocate variables to the read-only data section first if possible, then
17862 next in the small data section if possible, otherwise in data. This gives
17863 slightly slower code than the default, but reduces the amount of RAM required
17864 when executing, and thus may be preferred for some embedded systems.
17866 @item -muninit-const-in-rodata
17867 @itemx -mno-uninit-const-in-rodata
17868 @opindex muninit-const-in-rodata
17869 @opindex mno-uninit-const-in-rodata
17870 Put uninitialized @code{const} variables in the read-only data section.
17871 This option is only meaningful in conjunction with @option{-membedded-data}.
17873 @item -mcode-readable=@var{setting}
17874 @opindex mcode-readable
17875 Specify whether GCC may generate code that reads from executable sections.
17876 There are three possible settings:
17879 @item -mcode-readable=yes
17880 Instructions may freely access executable sections. This is the
17883 @item -mcode-readable=pcrel
17884 MIPS16 PC-relative load instructions can access executable sections,
17885 but other instructions must not do so. This option is useful on 4KSc
17886 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
17887 It is also useful on processors that can be configured to have a dual
17888 instruction/data SRAM interface and that, like the M4K, automatically
17889 redirect PC-relative loads to the instruction RAM.
17891 @item -mcode-readable=no
17892 Instructions must not access executable sections. This option can be
17893 useful on targets that are configured to have a dual instruction/data
17894 SRAM interface but that (unlike the M4K) do not automatically redirect
17895 PC-relative loads to the instruction RAM.
17898 @item -msplit-addresses
17899 @itemx -mno-split-addresses
17900 @opindex msplit-addresses
17901 @opindex mno-split-addresses
17902 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
17903 relocation operators. This option has been superseded by
17904 @option{-mexplicit-relocs} but is retained for backwards compatibility.
17906 @item -mexplicit-relocs
17907 @itemx -mno-explicit-relocs
17908 @opindex mexplicit-relocs
17909 @opindex mno-explicit-relocs
17910 Use (do not use) assembler relocation operators when dealing with symbolic
17911 addresses. The alternative, selected by @option{-mno-explicit-relocs},
17912 is to use assembler macros instead.
17914 @option{-mexplicit-relocs} is the default if GCC was configured
17915 to use an assembler that supports relocation operators.
17917 @item -mcheck-zero-division
17918 @itemx -mno-check-zero-division
17919 @opindex mcheck-zero-division
17920 @opindex mno-check-zero-division
17921 Trap (do not trap) on integer division by zero.
17923 The default is @option{-mcheck-zero-division}.
17925 @item -mdivide-traps
17926 @itemx -mdivide-breaks
17927 @opindex mdivide-traps
17928 @opindex mdivide-breaks
17929 MIPS systems check for division by zero by generating either a
17930 conditional trap or a break instruction. Using traps results in
17931 smaller code, but is only supported on MIPS II and later. Also, some
17932 versions of the Linux kernel have a bug that prevents trap from
17933 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
17934 allow conditional traps on architectures that support them and
17935 @option{-mdivide-breaks} to force the use of breaks.
17937 The default is usually @option{-mdivide-traps}, but this can be
17938 overridden at configure time using @option{--with-divide=breaks}.
17939 Divide-by-zero checks can be completely disabled using
17940 @option{-mno-check-zero-division}.
17945 @opindex mno-memcpy
17946 Force (do not force) the use of @code{memcpy} for non-trivial block
17947 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
17948 most constant-sized copies.
17951 @itemx -mno-long-calls
17952 @opindex mlong-calls
17953 @opindex mno-long-calls
17954 Disable (do not disable) use of the @code{jal} instruction. Calling
17955 functions using @code{jal} is more efficient but requires the caller
17956 and callee to be in the same 256 megabyte segment.
17958 This option has no effect on abicalls code. The default is
17959 @option{-mno-long-calls}.
17965 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
17966 instructions, as provided by the R4650 ISA@.
17972 Enable (disable) use of the @code{madd} and @code{msub} integer
17973 instructions. The default is @option{-mimadd} on architectures
17974 that support @code{madd} and @code{msub} except for the 74k
17975 architecture where it was found to generate slower code.
17978 @itemx -mno-fused-madd
17979 @opindex mfused-madd
17980 @opindex mno-fused-madd
17981 Enable (disable) use of the floating-point multiply-accumulate
17982 instructions, when they are available. The default is
17983 @option{-mfused-madd}.
17985 On the R8000 CPU when multiply-accumulate instructions are used,
17986 the intermediate product is calculated to infinite precision
17987 and is not subject to the FCSR Flush to Zero bit. This may be
17988 undesirable in some circumstances. On other processors the result
17989 is numerically identical to the equivalent computation using
17990 separate multiply, add, subtract and negate instructions.
17994 Tell the MIPS assembler to not run its preprocessor over user
17995 assembler files (with a @samp{.s} suffix) when assembling them.
18000 @opindex mno-fix-24k
18001 Work around the 24K E48 (lost data on stores during refill) errata.
18002 The workarounds are implemented by the assembler rather than by GCC@.
18005 @itemx -mno-fix-r4000
18006 @opindex mfix-r4000
18007 @opindex mno-fix-r4000
18008 Work around certain R4000 CPU errata:
18011 A double-word or a variable shift may give an incorrect result if executed
18012 immediately after starting an integer division.
18014 A double-word or a variable shift may give an incorrect result if executed
18015 while an integer multiplication is in progress.
18017 An integer division may give an incorrect result if started in a delay slot
18018 of a taken branch or a jump.
18022 @itemx -mno-fix-r4400
18023 @opindex mfix-r4400
18024 @opindex mno-fix-r4400
18025 Work around certain R4400 CPU errata:
18028 A double-word or a variable shift may give an incorrect result if executed
18029 immediately after starting an integer division.
18033 @itemx -mno-fix-r10000
18034 @opindex mfix-r10000
18035 @opindex mno-fix-r10000
18036 Work around certain R10000 errata:
18039 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18040 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18043 This option can only be used if the target architecture supports
18044 branch-likely instructions. @option{-mfix-r10000} is the default when
18045 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18049 @itemx -mno-fix-rm7000
18050 @opindex mfix-rm7000
18051 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18052 workarounds are implemented by the assembler rather than by GCC@.
18055 @itemx -mno-fix-vr4120
18056 @opindex mfix-vr4120
18057 Work around certain VR4120 errata:
18060 @code{dmultu} does not always produce the correct result.
18062 @code{div} and @code{ddiv} do not always produce the correct result if one
18063 of the operands is negative.
18065 The workarounds for the division errata rely on special functions in
18066 @file{libgcc.a}. At present, these functions are only provided by
18067 the @code{mips64vr*-elf} configurations.
18069 Other VR4120 errata require a NOP to be inserted between certain pairs of
18070 instructions. These errata are handled by the assembler, not by GCC itself.
18073 @opindex mfix-vr4130
18074 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18075 workarounds are implemented by the assembler rather than by GCC,
18076 although GCC avoids using @code{mflo} and @code{mfhi} if the
18077 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18078 instructions are available instead.
18081 @itemx -mno-fix-sb1
18083 Work around certain SB-1 CPU core errata.
18084 (This flag currently works around the SB-1 revision 2
18085 ``F1'' and ``F2'' floating-point errata.)
18087 @item -mr10k-cache-barrier=@var{setting}
18088 @opindex mr10k-cache-barrier
18089 Specify whether GCC should insert cache barriers to avoid the
18090 side-effects of speculation on R10K processors.
18092 In common with many processors, the R10K tries to predict the outcome
18093 of a conditional branch and speculatively executes instructions from
18094 the ``taken'' branch. It later aborts these instructions if the
18095 predicted outcome is wrong. However, on the R10K, even aborted
18096 instructions can have side effects.
18098 This problem only affects kernel stores and, depending on the system,
18099 kernel loads. As an example, a speculatively-executed store may load
18100 the target memory into cache and mark the cache line as dirty, even if
18101 the store itself is later aborted. If a DMA operation writes to the
18102 same area of memory before the ``dirty'' line is flushed, the cached
18103 data overwrites the DMA-ed data. See the R10K processor manual
18104 for a full description, including other potential problems.
18106 One workaround is to insert cache barrier instructions before every memory
18107 access that might be speculatively executed and that might have side
18108 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18109 controls GCC's implementation of this workaround. It assumes that
18110 aborted accesses to any byte in the following regions does not have
18115 the memory occupied by the current function's stack frame;
18118 the memory occupied by an incoming stack argument;
18121 the memory occupied by an object with a link-time-constant address.
18124 It is the kernel's responsibility to ensure that speculative
18125 accesses to these regions are indeed safe.
18127 If the input program contains a function declaration such as:
18133 then the implementation of @code{foo} must allow @code{j foo} and
18134 @code{jal foo} to be executed speculatively. GCC honors this
18135 restriction for functions it compiles itself. It expects non-GCC
18136 functions (such as hand-written assembly code) to do the same.
18138 The option has three forms:
18141 @item -mr10k-cache-barrier=load-store
18142 Insert a cache barrier before a load or store that might be
18143 speculatively executed and that might have side effects even
18146 @item -mr10k-cache-barrier=store
18147 Insert a cache barrier before a store that might be speculatively
18148 executed and that might have side effects even if aborted.
18150 @item -mr10k-cache-barrier=none
18151 Disable the insertion of cache barriers. This is the default setting.
18154 @item -mflush-func=@var{func}
18155 @itemx -mno-flush-func
18156 @opindex mflush-func
18157 Specifies the function to call to flush the I and D caches, or to not
18158 call any such function. If called, the function must take the same
18159 arguments as the common @code{_flush_func}, that is, the address of the
18160 memory range for which the cache is being flushed, the size of the
18161 memory range, and the number 3 (to flush both caches). The default
18162 depends on the target GCC was configured for, but commonly is either
18163 @code{_flush_func} or @code{__cpu_flush}.
18165 @item mbranch-cost=@var{num}
18166 @opindex mbranch-cost
18167 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18168 This cost is only a heuristic and is not guaranteed to produce
18169 consistent results across releases. A zero cost redundantly selects
18170 the default, which is based on the @option{-mtune} setting.
18172 @item -mbranch-likely
18173 @itemx -mno-branch-likely
18174 @opindex mbranch-likely
18175 @opindex mno-branch-likely
18176 Enable or disable use of Branch Likely instructions, regardless of the
18177 default for the selected architecture. By default, Branch Likely
18178 instructions may be generated if they are supported by the selected
18179 architecture. An exception is for the MIPS32 and MIPS64 architectures
18180 and processors that implement those architectures; for those, Branch
18181 Likely instructions are not be generated by default because the MIPS32
18182 and MIPS64 architectures specifically deprecate their use.
18184 @item -mcompact-branches=never
18185 @itemx -mcompact-branches=optimal
18186 @itemx -mcompact-branches=always
18187 @opindex mcompact-branches=never
18188 @opindex mcompact-branches=optimal
18189 @opindex mcompact-branches=always
18190 These options control which form of branches will be generated. The
18191 default is @option{-mcompact-branches=optimal}.
18193 The @option{-mcompact-branches=never} option ensures that compact branch
18194 instructions will never be generated.
18196 The @option{-mcompact-branches=always} option ensures that a compact
18197 branch instruction will be generated if available. If a compact branch
18198 instruction is not available, a delay slot form of the branch will be
18201 This option is supported from MIPS Release 6 onwards.
18203 The @option{-mcompact-branches=optimal} option will cause a delay slot
18204 branch to be used if one is available in the current ISA and the delay
18205 slot is successfully filled. If the delay slot is not filled, a compact
18206 branch will be chosen if one is available.
18208 @item -mfp-exceptions
18209 @itemx -mno-fp-exceptions
18210 @opindex mfp-exceptions
18211 Specifies whether FP exceptions are enabled. This affects how
18212 FP instructions are scheduled for some processors.
18213 The default is that FP exceptions are
18216 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18217 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18220 @item -mvr4130-align
18221 @itemx -mno-vr4130-align
18222 @opindex mvr4130-align
18223 The VR4130 pipeline is two-way superscalar, but can only issue two
18224 instructions together if the first one is 8-byte aligned. When this
18225 option is enabled, GCC aligns pairs of instructions that it
18226 thinks should execute in parallel.
18228 This option only has an effect when optimizing for the VR4130.
18229 It normally makes code faster, but at the expense of making it bigger.
18230 It is enabled by default at optimization level @option{-O3}.
18235 Enable (disable) generation of @code{synci} instructions on
18236 architectures that support it. The @code{synci} instructions (if
18237 enabled) are generated when @code{__builtin___clear_cache} is
18240 This option defaults to @option{-mno-synci}, but the default can be
18241 overridden by configuring GCC with @option{--with-synci}.
18243 When compiling code for single processor systems, it is generally safe
18244 to use @code{synci}. However, on many multi-core (SMP) systems, it
18245 does not invalidate the instruction caches on all cores and may lead
18246 to undefined behavior.
18248 @item -mrelax-pic-calls
18249 @itemx -mno-relax-pic-calls
18250 @opindex mrelax-pic-calls
18251 Try to turn PIC calls that are normally dispatched via register
18252 @code{$25} into direct calls. This is only possible if the linker can
18253 resolve the destination at link-time and if the destination is within
18254 range for a direct call.
18256 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18257 an assembler and a linker that support the @code{.reloc} assembly
18258 directive and @option{-mexplicit-relocs} is in effect. With
18259 @option{-mno-explicit-relocs}, this optimization can be performed by the
18260 assembler and the linker alone without help from the compiler.
18262 @item -mmcount-ra-address
18263 @itemx -mno-mcount-ra-address
18264 @opindex mmcount-ra-address
18265 @opindex mno-mcount-ra-address
18266 Emit (do not emit) code that allows @code{_mcount} to modify the
18267 calling function's return address. When enabled, this option extends
18268 the usual @code{_mcount} interface with a new @var{ra-address}
18269 parameter, which has type @code{intptr_t *} and is passed in register
18270 @code{$12}. @code{_mcount} can then modify the return address by
18271 doing both of the following:
18274 Returning the new address in register @code{$31}.
18276 Storing the new address in @code{*@var{ra-address}},
18277 if @var{ra-address} is nonnull.
18280 The default is @option{-mno-mcount-ra-address}.
18282 @item -mframe-header-opt
18283 @itemx -mno-frame-header-opt
18284 @opindex mframe-header-opt
18285 Enable (disable) frame header optimization in the o32 ABI. When using the
18286 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18287 function to write out register arguments. When enabled, this optimization
18288 will suppress the allocation of the frame header if it can be determined that
18291 This optimization is off by default at all optimization levels.
18296 @subsection MMIX Options
18297 @cindex MMIX Options
18299 These options are defined for the MMIX:
18303 @itemx -mno-libfuncs
18305 @opindex mno-libfuncs
18306 Specify that intrinsic library functions are being compiled, passing all
18307 values in registers, no matter the size.
18310 @itemx -mno-epsilon
18312 @opindex mno-epsilon
18313 Generate floating-point comparison instructions that compare with respect
18314 to the @code{rE} epsilon register.
18316 @item -mabi=mmixware
18318 @opindex mabi=mmixware
18320 Generate code that passes function parameters and return values that (in
18321 the called function) are seen as registers @code{$0} and up, as opposed to
18322 the GNU ABI which uses global registers @code{$231} and up.
18324 @item -mzero-extend
18325 @itemx -mno-zero-extend
18326 @opindex mzero-extend
18327 @opindex mno-zero-extend
18328 When reading data from memory in sizes shorter than 64 bits, use (do not
18329 use) zero-extending load instructions by default, rather than
18330 sign-extending ones.
18333 @itemx -mno-knuthdiv
18335 @opindex mno-knuthdiv
18336 Make the result of a division yielding a remainder have the same sign as
18337 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18338 remainder follows the sign of the dividend. Both methods are
18339 arithmetically valid, the latter being almost exclusively used.
18341 @item -mtoplevel-symbols
18342 @itemx -mno-toplevel-symbols
18343 @opindex mtoplevel-symbols
18344 @opindex mno-toplevel-symbols
18345 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18346 code can be used with the @code{PREFIX} assembly directive.
18350 Generate an executable in the ELF format, rather than the default
18351 @samp{mmo} format used by the @command{mmix} simulator.
18353 @item -mbranch-predict
18354 @itemx -mno-branch-predict
18355 @opindex mbranch-predict
18356 @opindex mno-branch-predict
18357 Use (do not use) the probable-branch instructions, when static branch
18358 prediction indicates a probable branch.
18360 @item -mbase-addresses
18361 @itemx -mno-base-addresses
18362 @opindex mbase-addresses
18363 @opindex mno-base-addresses
18364 Generate (do not generate) code that uses @emph{base addresses}. Using a
18365 base address automatically generates a request (handled by the assembler
18366 and the linker) for a constant to be set up in a global register. The
18367 register is used for one or more base address requests within the range 0
18368 to 255 from the value held in the register. The generally leads to short
18369 and fast code, but the number of different data items that can be
18370 addressed is limited. This means that a program that uses lots of static
18371 data may require @option{-mno-base-addresses}.
18373 @item -msingle-exit
18374 @itemx -mno-single-exit
18375 @opindex msingle-exit
18376 @opindex mno-single-exit
18377 Force (do not force) generated code to have a single exit point in each
18381 @node MN10300 Options
18382 @subsection MN10300 Options
18383 @cindex MN10300 options
18385 These @option{-m} options are defined for Matsushita MN10300 architectures:
18390 Generate code to avoid bugs in the multiply instructions for the MN10300
18391 processors. This is the default.
18393 @item -mno-mult-bug
18394 @opindex mno-mult-bug
18395 Do not generate code to avoid bugs in the multiply instructions for the
18396 MN10300 processors.
18400 Generate code using features specific to the AM33 processor.
18404 Do not generate code using features specific to the AM33 processor. This
18409 Generate code using features specific to the AM33/2.0 processor.
18413 Generate code using features specific to the AM34 processor.
18415 @item -mtune=@var{cpu-type}
18417 Use the timing characteristics of the indicated CPU type when
18418 scheduling instructions. This does not change the targeted processor
18419 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
18420 @samp{am33-2} or @samp{am34}.
18422 @item -mreturn-pointer-on-d0
18423 @opindex mreturn-pointer-on-d0
18424 When generating a function that returns a pointer, return the pointer
18425 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
18426 only in @code{a0}, and attempts to call such functions without a prototype
18427 result in errors. Note that this option is on by default; use
18428 @option{-mno-return-pointer-on-d0} to disable it.
18432 Do not link in the C run-time initialization object file.
18436 Indicate to the linker that it should perform a relaxation optimization pass
18437 to shorten branches, calls and absolute memory addresses. This option only
18438 has an effect when used on the command line for the final link step.
18440 This option makes symbolic debugging impossible.
18444 Allow the compiler to generate @emph{Long Instruction Word}
18445 instructions if the target is the @samp{AM33} or later. This is the
18446 default. This option defines the preprocessor macro @code{__LIW__}.
18450 Do not allow the compiler to generate @emph{Long Instruction Word}
18451 instructions. This option defines the preprocessor macro
18456 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
18457 instructions if the target is the @samp{AM33} or later. This is the
18458 default. This option defines the preprocessor macro @code{__SETLB__}.
18462 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
18463 instructions. This option defines the preprocessor macro
18464 @code{__NO_SETLB__}.
18468 @node Moxie Options
18469 @subsection Moxie Options
18470 @cindex Moxie Options
18476 Generate big-endian code. This is the default for @samp{moxie-*-*}
18481 Generate little-endian code.
18485 Generate mul.x and umul.x instructions. This is the default for
18486 @samp{moxiebox-*-*} configurations.
18490 Do not link in the C run-time initialization object file.
18494 @node MSP430 Options
18495 @subsection MSP430 Options
18496 @cindex MSP430 Options
18498 These options are defined for the MSP430:
18504 Force assembly output to always use hex constants. Normally such
18505 constants are signed decimals, but this option is available for
18506 testsuite and/or aesthetic purposes.
18510 Select the MCU to target. This is used to create a C preprocessor
18511 symbol based upon the MCU name, converted to upper case and pre- and
18512 post-fixed with @samp{__}. This in turn is used by the
18513 @file{msp430.h} header file to select an MCU-specific supplementary
18516 The option also sets the ISA to use. If the MCU name is one that is
18517 known to only support the 430 ISA then that is selected, otherwise the
18518 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
18519 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
18520 name selects the 430X ISA.
18522 In addition an MCU-specific linker script is added to the linker
18523 command line. The script's name is the name of the MCU with
18524 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
18525 command line defines the C preprocessor symbol @code{__XXX__} and
18526 cause the linker to search for a script called @file{xxx.ld}.
18528 This option is also passed on to the assembler.
18531 @itemx -mno-warn-mcu
18533 @opindex mno-warn-mcu
18534 This option enables or disables warnings about conflicts between the
18535 MCU name specified by the @option{-mmcu} option and the ISA set by the
18536 @option{-mcpu} option and/or the hardware multiply support set by the
18537 @option{-mhwmult} option. It also toggles warnings about unrecognised
18538 MCU names. This option is on by default.
18542 Specifies the ISA to use. Accepted values are @samp{msp430},
18543 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
18544 @option{-mmcu=} option should be used to select the ISA.
18548 Link to the simulator runtime libraries and linker script. Overrides
18549 any scripts that would be selected by the @option{-mmcu=} option.
18553 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
18557 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
18561 This option is passed to the assembler and linker, and allows the
18562 linker to perform certain optimizations that cannot be done until
18567 Describes the type of hardware multiply supported by the target.
18568 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
18569 for the original 16-bit-only multiply supported by early MCUs.
18570 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
18571 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
18572 A value of @samp{auto} can also be given. This tells GCC to deduce
18573 the hardware multiply support based upon the MCU name provided by the
18574 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
18575 the MCU name is not recognised then no hardware multiply support is
18576 assumed. @code{auto} is the default setting.
18578 Hardware multiplies are normally performed by calling a library
18579 routine. This saves space in the generated code. When compiling at
18580 @option{-O3} or higher however the hardware multiplier is invoked
18581 inline. This makes for bigger, but faster code.
18583 The hardware multiply routines disable interrupts whilst running and
18584 restore the previous interrupt state when they finish. This makes
18585 them safe to use inside interrupt handlers as well as in normal code.
18589 Enable the use of a minimum runtime environment - no static
18590 initializers or constructors. This is intended for memory-constrained
18591 devices. The compiler includes special symbols in some objects
18592 that tell the linker and runtime which code fragments are required.
18594 @item -mcode-region=
18595 @itemx -mdata-region=
18596 @opindex mcode-region
18597 @opindex mdata-region
18598 These options tell the compiler where to place functions and data that
18599 do not have one of the @code{lower}, @code{upper}, @code{either} or
18600 @code{section} attributes. Possible values are @code{lower},
18601 @code{upper}, @code{either} or @code{any}. The first three behave
18602 like the corresponding attribute. The fourth possible value -
18603 @code{any} - is the default. It leaves placement entirely up to the
18604 linker script and how it assigns the standard sections (.text, .data
18605 etc) to the memory regions.
18607 @item -msilicon-errata=
18608 @opindex msilicon-errata
18609 This option passes on a request to assembler to enable the fixes for
18610 the named silicon errata.
18612 @item -msilicon-errata-warn=
18613 @opindex msilicon-errata-warn
18614 This option passes on a request to the assembler to enable warning
18615 messages when a silicon errata might need to be applied.
18619 @node NDS32 Options
18620 @subsection NDS32 Options
18621 @cindex NDS32 Options
18623 These options are defined for NDS32 implementations:
18628 @opindex mbig-endian
18629 Generate code in big-endian mode.
18631 @item -mlittle-endian
18632 @opindex mlittle-endian
18633 Generate code in little-endian mode.
18635 @item -mreduced-regs
18636 @opindex mreduced-regs
18637 Use reduced-set registers for register allocation.
18640 @opindex mfull-regs
18641 Use full-set registers for register allocation.
18645 Generate conditional move instructions.
18649 Do not generate conditional move instructions.
18653 Generate performance extension instructions.
18655 @item -mno-perf-ext
18656 @opindex mno-perf-ext
18657 Do not generate performance extension instructions.
18661 Generate v3 push25/pop25 instructions.
18664 @opindex mno-v3push
18665 Do not generate v3 push25/pop25 instructions.
18669 Generate 16-bit instructions.
18672 @opindex mno-16-bit
18673 Do not generate 16-bit instructions.
18675 @item -misr-vector-size=@var{num}
18676 @opindex misr-vector-size
18677 Specify the size of each interrupt vector, which must be 4 or 16.
18679 @item -mcache-block-size=@var{num}
18680 @opindex mcache-block-size
18681 Specify the size of each cache block,
18682 which must be a power of 2 between 4 and 512.
18684 @item -march=@var{arch}
18686 Specify the name of the target architecture.
18688 @item -mcmodel=@var{code-model}
18690 Set the code model to one of
18693 All the data and read-only data segments must be within 512KB addressing space.
18694 The text segment must be within 16MB addressing space.
18695 @item @samp{medium}
18696 The data segment must be within 512KB while the read-only data segment can be
18697 within 4GB addressing space. The text segment should be still within 16MB
18700 All the text and data segments can be within 4GB addressing space.
18704 @opindex mctor-dtor
18705 Enable constructor/destructor feature.
18709 Guide linker to relax instructions.
18713 @node Nios II Options
18714 @subsection Nios II Options
18715 @cindex Nios II options
18716 @cindex Altera Nios II options
18718 These are the options defined for the Altera Nios II processor.
18724 @cindex smaller data references
18725 Put global and static objects less than or equal to @var{num} bytes
18726 into the small data or BSS sections instead of the normal data or BSS
18727 sections. The default value of @var{num} is 8.
18729 @item -mgpopt=@var{option}
18734 Generate (do not generate) GP-relative accesses. The following
18735 @var{option} names are recognized:
18740 Do not generate GP-relative accesses.
18743 Generate GP-relative accesses for small data objects that are not
18744 external, weak, or uninitialized common symbols.
18745 Also use GP-relative addressing for objects that
18746 have been explicitly placed in a small data section via a @code{section}
18750 As for @samp{local}, but also generate GP-relative accesses for
18751 small data objects that are external, weak, or common. If you use this option,
18752 you must ensure that all parts of your program (including libraries) are
18753 compiled with the same @option{-G} setting.
18756 Generate GP-relative accesses for all data objects in the program. If you
18757 use this option, the entire data and BSS segments
18758 of your program must fit in 64K of memory and you must use an appropriate
18759 linker script to allocate them within the addressible range of the
18763 Generate GP-relative addresses for function pointers as well as data
18764 pointers. If you use this option, the entire text, data, and BSS segments
18765 of your program must fit in 64K of memory and you must use an appropriate
18766 linker script to allocate them within the addressible range of the
18771 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
18772 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
18774 The default is @option{-mgpopt} except when @option{-fpic} or
18775 @option{-fPIC} is specified to generate position-independent code.
18776 Note that the Nios II ABI does not permit GP-relative accesses from
18779 You may need to specify @option{-mno-gpopt} explicitly when building
18780 programs that include large amounts of small data, including large
18781 GOT data sections. In this case, the 16-bit offset for GP-relative
18782 addressing may not be large enough to allow access to the entire
18783 small data section.
18789 Generate little-endian (default) or big-endian (experimental) code,
18792 @item -march=@var{arch}
18794 This specifies the name of the target Nios II architecture. GCC uses this
18795 name to determine what kind of instructions it can emit when generating
18796 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
18798 The preprocessor macro @code{__nios2_arch__} is available to programs,
18799 with value 1 or 2, indicating the targeted ISA level.
18801 @item -mbypass-cache
18802 @itemx -mno-bypass-cache
18803 @opindex mno-bypass-cache
18804 @opindex mbypass-cache
18805 Force all load and store instructions to always bypass cache by
18806 using I/O variants of the instructions. The default is not to
18809 @item -mno-cache-volatile
18810 @itemx -mcache-volatile
18811 @opindex mcache-volatile
18812 @opindex mno-cache-volatile
18813 Volatile memory access bypass the cache using the I/O variants of
18814 the load and store instructions. The default is not to bypass the cache.
18816 @item -mno-fast-sw-div
18817 @itemx -mfast-sw-div
18818 @opindex mno-fast-sw-div
18819 @opindex mfast-sw-div
18820 Do not use table-based fast divide for small numbers. The default
18821 is to use the fast divide at @option{-O3} and above.
18825 @itemx -mno-hw-mulx
18829 @opindex mno-hw-mul
18831 @opindex mno-hw-mulx
18833 @opindex mno-hw-div
18835 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
18836 instructions by the compiler. The default is to emit @code{mul}
18837 and not emit @code{div} and @code{mulx}.
18843 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
18844 CDX (code density) instructions. Enabling these instructions also
18845 requires @option{-march=r2}. Since these instructions are optional
18846 extensions to the R2 architecture, the default is not to emit them.
18848 @item -mcustom-@var{insn}=@var{N}
18849 @itemx -mno-custom-@var{insn}
18850 @opindex mcustom-@var{insn}
18851 @opindex mno-custom-@var{insn}
18852 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
18853 custom instruction with encoding @var{N} when generating code that uses
18854 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
18855 instruction 253 for single-precision floating-point add operations instead
18856 of the default behavior of using a library call.
18858 The following values of @var{insn} are supported. Except as otherwise
18859 noted, floating-point operations are expected to be implemented with
18860 normal IEEE 754 semantics and correspond directly to the C operators or the
18861 equivalent GCC built-in functions (@pxref{Other Builtins}).
18863 Single-precision floating point:
18866 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
18867 Binary arithmetic operations.
18873 Unary absolute value.
18875 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
18876 Comparison operations.
18878 @item @samp{fmins}, @samp{fmaxs}
18879 Floating-point minimum and maximum. These instructions are only
18880 generated if @option{-ffinite-math-only} is specified.
18882 @item @samp{fsqrts}
18883 Unary square root operation.
18885 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
18886 Floating-point trigonometric and exponential functions. These instructions
18887 are only generated if @option{-funsafe-math-optimizations} is also specified.
18891 Double-precision floating point:
18894 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
18895 Binary arithmetic operations.
18901 Unary absolute value.
18903 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
18904 Comparison operations.
18906 @item @samp{fmind}, @samp{fmaxd}
18907 Double-precision minimum and maximum. These instructions are only
18908 generated if @option{-ffinite-math-only} is specified.
18910 @item @samp{fsqrtd}
18911 Unary square root operation.
18913 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
18914 Double-precision trigonometric and exponential functions. These instructions
18915 are only generated if @option{-funsafe-math-optimizations} is also specified.
18921 @item @samp{fextsd}
18922 Conversion from single precision to double precision.
18924 @item @samp{ftruncds}
18925 Conversion from double precision to single precision.
18927 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
18928 Conversion from floating point to signed or unsigned integer types, with
18929 truncation towards zero.
18932 Conversion from single-precision floating point to signed integer,
18933 rounding to the nearest integer and ties away from zero.
18934 This corresponds to the @code{__builtin_lroundf} function when
18935 @option{-fno-math-errno} is used.
18937 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
18938 Conversion from signed or unsigned integer types to floating-point types.
18942 In addition, all of the following transfer instructions for internal
18943 registers X and Y must be provided to use any of the double-precision
18944 floating-point instructions. Custom instructions taking two
18945 double-precision source operands expect the first operand in the
18946 64-bit register X. The other operand (or only operand of a unary
18947 operation) is given to the custom arithmetic instruction with the
18948 least significant half in source register @var{src1} and the most
18949 significant half in @var{src2}. A custom instruction that returns a
18950 double-precision result returns the most significant 32 bits in the
18951 destination register and the other half in 32-bit register Y.
18952 GCC automatically generates the necessary code sequences to write
18953 register X and/or read register Y when double-precision floating-point
18954 instructions are used.
18959 Write @var{src1} into the least significant half of X and @var{src2} into
18960 the most significant half of X.
18963 Write @var{src1} into Y.
18965 @item @samp{frdxhi}, @samp{frdxlo}
18966 Read the most or least (respectively) significant half of X and store it in
18970 Read the value of Y and store it into @var{dest}.
18973 Note that you can gain more local control over generation of Nios II custom
18974 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
18975 and @code{target("no-custom-@var{insn}")} function attributes
18976 (@pxref{Function Attributes})
18977 or pragmas (@pxref{Function Specific Option Pragmas}).
18979 @item -mcustom-fpu-cfg=@var{name}
18980 @opindex mcustom-fpu-cfg
18982 This option enables a predefined, named set of custom instruction encodings
18983 (see @option{-mcustom-@var{insn}} above).
18984 Currently, the following sets are defined:
18986 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
18987 @gccoptlist{-mcustom-fmuls=252 @gol
18988 -mcustom-fadds=253 @gol
18989 -mcustom-fsubs=254 @gol
18990 -fsingle-precision-constant}
18992 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
18993 @gccoptlist{-mcustom-fmuls=252 @gol
18994 -mcustom-fadds=253 @gol
18995 -mcustom-fsubs=254 @gol
18996 -mcustom-fdivs=255 @gol
18997 -fsingle-precision-constant}
18999 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19000 @gccoptlist{-mcustom-floatus=243 @gol
19001 -mcustom-fixsi=244 @gol
19002 -mcustom-floatis=245 @gol
19003 -mcustom-fcmpgts=246 @gol
19004 -mcustom-fcmples=249 @gol
19005 -mcustom-fcmpeqs=250 @gol
19006 -mcustom-fcmpnes=251 @gol
19007 -mcustom-fmuls=252 @gol
19008 -mcustom-fadds=253 @gol
19009 -mcustom-fsubs=254 @gol
19010 -mcustom-fdivs=255 @gol
19011 -fsingle-precision-constant}
19013 Custom instruction assignments given by individual
19014 @option{-mcustom-@var{insn}=} options override those given by
19015 @option{-mcustom-fpu-cfg=}, regardless of the
19016 order of the options on the command line.
19018 Note that you can gain more local control over selection of a FPU
19019 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19020 function attribute (@pxref{Function Attributes})
19021 or pragma (@pxref{Function Specific Option Pragmas}).
19025 These additional @samp{-m} options are available for the Altera Nios II
19026 ELF (bare-metal) target:
19032 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19033 startup and termination code, and is typically used in conjunction with
19034 @option{-msys-crt0=} to specify the location of the alternate startup code
19035 provided by the HAL BSP.
19039 Link with a limited version of the C library, @option{-lsmallc}, rather than
19042 @item -msys-crt0=@var{startfile}
19044 @var{startfile} is the file name of the startfile (crt0) to use
19045 when linking. This option is only useful in conjunction with @option{-mhal}.
19047 @item -msys-lib=@var{systemlib}
19049 @var{systemlib} is the library name of the library that provides
19050 low-level system calls required by the C library,
19051 e.g. @code{read} and @code{write}.
19052 This option is typically used to link with a library provided by a HAL BSP.
19056 @node Nvidia PTX Options
19057 @subsection Nvidia PTX Options
19058 @cindex Nvidia PTX options
19059 @cindex nvptx options
19061 These options are defined for Nvidia PTX:
19069 Generate code for 32-bit or 64-bit ABI.
19072 @opindex mmainkernel
19073 Link in code for a __main kernel. This is for stand-alone instead of
19074 offloading execution.
19078 Apply partitioned execution optimizations. This is the default when any
19079 level of optimization is selected.
19083 @node PDP-11 Options
19084 @subsection PDP-11 Options
19085 @cindex PDP-11 Options
19087 These options are defined for the PDP-11:
19092 Use hardware FPP floating point. This is the default. (FIS floating
19093 point on the PDP-11/40 is not supported.)
19096 @opindex msoft-float
19097 Do not use hardware floating point.
19101 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19105 Return floating-point results in memory. This is the default.
19109 Generate code for a PDP-11/40.
19113 Generate code for a PDP-11/45. This is the default.
19117 Generate code for a PDP-11/10.
19119 @item -mbcopy-builtin
19120 @opindex mbcopy-builtin
19121 Use inline @code{movmemhi} patterns for copying memory. This is the
19126 Do not use inline @code{movmemhi} patterns for copying memory.
19132 Use 16-bit @code{int}. This is the default.
19138 Use 32-bit @code{int}.
19141 @itemx -mno-float32
19143 @opindex mno-float32
19144 Use 64-bit @code{float}. This is the default.
19147 @itemx -mno-float64
19149 @opindex mno-float64
19150 Use 32-bit @code{float}.
19154 Use @code{abshi2} pattern. This is the default.
19158 Do not use @code{abshi2} pattern.
19160 @item -mbranch-expensive
19161 @opindex mbranch-expensive
19162 Pretend that branches are expensive. This is for experimenting with
19163 code generation only.
19165 @item -mbranch-cheap
19166 @opindex mbranch-cheap
19167 Do not pretend that branches are expensive. This is the default.
19171 Use Unix assembler syntax. This is the default when configured for
19172 @samp{pdp11-*-bsd}.
19176 Use DEC assembler syntax. This is the default when configured for any
19177 PDP-11 target other than @samp{pdp11-*-bsd}.
19180 @node picoChip Options
19181 @subsection picoChip Options
19182 @cindex picoChip options
19184 These @samp{-m} options are defined for picoChip implementations:
19188 @item -mae=@var{ae_type}
19190 Set the instruction set, register set, and instruction scheduling
19191 parameters for array element type @var{ae_type}. Supported values
19192 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19194 @option{-mae=ANY} selects a completely generic AE type. Code
19195 generated with this option runs on any of the other AE types. The
19196 code is not as efficient as it would be if compiled for a specific
19197 AE type, and some types of operation (e.g., multiplication) do not
19198 work properly on all types of AE.
19200 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19201 for compiled code, and is the default.
19203 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19204 option may suffer from poor performance of byte (char) manipulation,
19205 since the DSP AE does not provide hardware support for byte load/stores.
19207 @item -msymbol-as-address
19208 Enable the compiler to directly use a symbol name as an address in a
19209 load/store instruction, without first loading it into a
19210 register. Typically, the use of this option generates larger
19211 programs, which run faster than when the option isn't used. However, the
19212 results vary from program to program, so it is left as a user option,
19213 rather than being permanently enabled.
19215 @item -mno-inefficient-warnings
19216 Disables warnings about the generation of inefficient code. These
19217 warnings can be generated, for example, when compiling code that
19218 performs byte-level memory operations on the MAC AE type. The MAC AE has
19219 no hardware support for byte-level memory operations, so all byte
19220 load/stores must be synthesized from word load/store operations. This is
19221 inefficient and a warning is generated to indicate
19222 that you should rewrite the code to avoid byte operations, or to target
19223 an AE type that has the necessary hardware support. This option disables
19228 @node PowerPC Options
19229 @subsection PowerPC Options
19230 @cindex PowerPC options
19232 These are listed under @xref{RS/6000 and PowerPC Options}.
19235 @subsection RL78 Options
19236 @cindex RL78 Options
19242 Links in additional target libraries to support operation within a
19251 Specifies the type of hardware multiplication and division support to
19252 be used. The simplest is @code{none}, which uses software for both
19253 multiplication and division. This is the default. The @code{g13}
19254 value is for the hardware multiply/divide peripheral found on the
19255 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19256 the multiplication and division instructions supported by the RL78/G14
19257 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19258 the value @code{mg10} is an alias for @code{none}.
19260 In addition a C preprocessor macro is defined, based upon the setting
19261 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19262 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19269 Specifies the RL78 core to target. The default is the G14 core, also
19270 known as an S3 core or just RL78. The G13 or S2 core does not have
19271 multiply or divide instructions, instead it uses a hardware peripheral
19272 for these operations. The G10 or S1 core does not have register
19273 banks, so it uses a different calling convention.
19275 If this option is set it also selects the type of hardware multiply
19276 support to use, unless this is overridden by an explicit
19277 @option{-mmul=none} option on the command line. Thus specifying
19278 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19279 peripheral and specifying @option{-mcpu=g10} disables the use of
19280 hardware multipications altogether.
19282 Note, although the RL78/G14 core is the default target, specifying
19283 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19284 change the behaviour of the toolchain since it also enables G14
19285 hardware multiply support. If these options are not specified on the
19286 command line then software multiplication routines will be used even
19287 though the code targets the RL78 core. This is for backwards
19288 compatibility with older toolchains which did not have hardware
19289 multiply and divide support.
19291 In addition a C preprocessor macro is defined, based upon the setting
19292 of this option. Possible values are: @code{__RL78_G10__},
19293 @code{__RL78_G13__} or @code{__RL78_G14__}.
19303 These are aliases for the corresponding @option{-mcpu=} option. They
19304 are provided for backwards compatibility.
19308 Allow the compiler to use all of the available registers. By default
19309 registers @code{r24..r31} are reserved for use in interrupt handlers.
19310 With this option enabled these registers can be used in ordinary
19313 @item -m64bit-doubles
19314 @itemx -m32bit-doubles
19315 @opindex m64bit-doubles
19316 @opindex m32bit-doubles
19317 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19318 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19319 @option{-m32bit-doubles}.
19323 @node RS/6000 and PowerPC Options
19324 @subsection IBM RS/6000 and PowerPC Options
19325 @cindex RS/6000 and PowerPC Options
19326 @cindex IBM RS/6000 and PowerPC Options
19328 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19330 @item -mpowerpc-gpopt
19331 @itemx -mno-powerpc-gpopt
19332 @itemx -mpowerpc-gfxopt
19333 @itemx -mno-powerpc-gfxopt
19336 @itemx -mno-powerpc64
19340 @itemx -mno-popcntb
19342 @itemx -mno-popcntd
19351 @itemx -mno-hard-dfp
19352 @opindex mpowerpc-gpopt
19353 @opindex mno-powerpc-gpopt
19354 @opindex mpowerpc-gfxopt
19355 @opindex mno-powerpc-gfxopt
19356 @opindex mpowerpc64
19357 @opindex mno-powerpc64
19361 @opindex mno-popcntb
19363 @opindex mno-popcntd
19369 @opindex mno-mfpgpr
19371 @opindex mno-hard-dfp
19372 You use these options to specify which instructions are available on the
19373 processor you are using. The default value of these options is
19374 determined when configuring GCC@. Specifying the
19375 @option{-mcpu=@var{cpu_type}} overrides the specification of these
19376 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
19377 rather than the options listed above.
19379 Specifying @option{-mpowerpc-gpopt} allows
19380 GCC to use the optional PowerPC architecture instructions in the
19381 General Purpose group, including floating-point square root. Specifying
19382 @option{-mpowerpc-gfxopt} allows GCC to
19383 use the optional PowerPC architecture instructions in the Graphics
19384 group, including floating-point select.
19386 The @option{-mmfcrf} option allows GCC to generate the move from
19387 condition register field instruction implemented on the POWER4
19388 processor and other processors that support the PowerPC V2.01
19390 The @option{-mpopcntb} option allows GCC to generate the popcount and
19391 double-precision FP reciprocal estimate instruction implemented on the
19392 POWER5 processor and other processors that support the PowerPC V2.02
19394 The @option{-mpopcntd} option allows GCC to generate the popcount
19395 instruction implemented on the POWER7 processor and other processors
19396 that support the PowerPC V2.06 architecture.
19397 The @option{-mfprnd} option allows GCC to generate the FP round to
19398 integer instructions implemented on the POWER5+ processor and other
19399 processors that support the PowerPC V2.03 architecture.
19400 The @option{-mcmpb} option allows GCC to generate the compare bytes
19401 instruction implemented on the POWER6 processor and other processors
19402 that support the PowerPC V2.05 architecture.
19403 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
19404 general-purpose register instructions implemented on the POWER6X
19405 processor and other processors that support the extended PowerPC V2.05
19407 The @option{-mhard-dfp} option allows GCC to generate the decimal
19408 floating-point instructions implemented on some POWER processors.
19410 The @option{-mpowerpc64} option allows GCC to generate the additional
19411 64-bit instructions that are found in the full PowerPC64 architecture
19412 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
19413 @option{-mno-powerpc64}.
19415 @item -mcpu=@var{cpu_type}
19417 Set architecture type, register usage, and
19418 instruction scheduling parameters for machine type @var{cpu_type}.
19419 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
19420 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
19421 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
19422 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
19423 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
19424 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
19425 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
19426 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
19427 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
19428 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
19429 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
19432 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
19433 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
19434 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
19435 architecture machine types, with an appropriate, generic processor
19436 model assumed for scheduling purposes.
19438 The other options specify a specific processor. Code generated under
19439 those options runs best on that processor, and may not run at all on
19442 The @option{-mcpu} options automatically enable or disable the
19445 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
19446 -mpopcntb -mpopcntd -mpowerpc64 @gol
19447 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
19448 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
19449 -mcrypto -mdirect-move -mpower8-fusion -mpower8-vector @gol
19450 -mquad-memory -mquad-memory-atomic -mmodulo -mfloat128 -mfloat128-hardware @gol
19451 -mpower9-fusion -mpower9-vector}
19453 The particular options set for any particular CPU varies between
19454 compiler versions, depending on what setting seems to produce optimal
19455 code for that CPU; it doesn't necessarily reflect the actual hardware's
19456 capabilities. If you wish to set an individual option to a particular
19457 value, you may specify it after the @option{-mcpu} option, like
19458 @option{-mcpu=970 -mno-altivec}.
19460 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
19461 not enabled or disabled by the @option{-mcpu} option at present because
19462 AIX does not have full support for these options. You may still
19463 enable or disable them individually if you're sure it'll work in your
19466 @item -mtune=@var{cpu_type}
19468 Set the instruction scheduling parameters for machine type
19469 @var{cpu_type}, but do not set the architecture type or register usage,
19470 as @option{-mcpu=@var{cpu_type}} does. The same
19471 values for @var{cpu_type} are used for @option{-mtune} as for
19472 @option{-mcpu}. If both are specified, the code generated uses the
19473 architecture and registers set by @option{-mcpu}, but the
19474 scheduling parameters set by @option{-mtune}.
19476 @item -mcmodel=small
19477 @opindex mcmodel=small
19478 Generate PowerPC64 code for the small model: The TOC is limited to
19481 @item -mcmodel=medium
19482 @opindex mcmodel=medium
19483 Generate PowerPC64 code for the medium model: The TOC and other static
19484 data may be up to a total of 4G in size.
19486 @item -mcmodel=large
19487 @opindex mcmodel=large
19488 Generate PowerPC64 code for the large model: The TOC may be up to 4G
19489 in size. Other data and code is only limited by the 64-bit address
19493 @itemx -mno-altivec
19495 @opindex mno-altivec
19496 Generate code that uses (does not use) AltiVec instructions, and also
19497 enable the use of built-in functions that allow more direct access to
19498 the AltiVec instruction set. You may also need to set
19499 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
19502 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
19503 @option{-maltivec=be}, the element order for Altivec intrinsics such
19504 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
19505 match array element order corresponding to the endianness of the
19506 target. That is, element zero identifies the leftmost element in a
19507 vector register when targeting a big-endian platform, and identifies
19508 the rightmost element in a vector register when targeting a
19509 little-endian platform.
19512 @opindex maltivec=be
19513 Generate Altivec instructions using big-endian element order,
19514 regardless of whether the target is big- or little-endian. This is
19515 the default when targeting a big-endian platform.
19517 The element order is used to interpret element numbers in Altivec
19518 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19519 @code{vec_insert}. By default, these match array element order
19520 corresponding to the endianness for the target.
19523 @opindex maltivec=le
19524 Generate Altivec instructions using little-endian element order,
19525 regardless of whether the target is big- or little-endian. This is
19526 the default when targeting a little-endian platform. This option is
19527 currently ignored when targeting a big-endian platform.
19529 The element order is used to interpret element numbers in Altivec
19530 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
19531 @code{vec_insert}. By default, these match array element order
19532 corresponding to the endianness for the target.
19537 @opindex mno-vrsave
19538 Generate VRSAVE instructions when generating AltiVec code.
19540 @item -mgen-cell-microcode
19541 @opindex mgen-cell-microcode
19542 Generate Cell microcode instructions.
19544 @item -mwarn-cell-microcode
19545 @opindex mwarn-cell-microcode
19546 Warn when a Cell microcode instruction is emitted. An example
19547 of a Cell microcode instruction is a variable shift.
19550 @opindex msecure-plt
19551 Generate code that allows @command{ld} and @command{ld.so}
19552 to build executables and shared
19553 libraries with non-executable @code{.plt} and @code{.got} sections.
19555 32-bit SYSV ABI option.
19559 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
19561 requires @code{.plt} and @code{.got}
19562 sections that are both writable and executable.
19563 This is a PowerPC 32-bit SYSV ABI option.
19569 This switch enables or disables the generation of ISEL instructions.
19571 @item -misel=@var{yes/no}
19572 This switch has been deprecated. Use @option{-misel} and
19573 @option{-mno-isel} instead.
19579 This switch enables or disables the generation of SPE simd
19585 @opindex mno-paired
19586 This switch enables or disables the generation of PAIRED simd
19589 @item -mspe=@var{yes/no}
19590 This option has been deprecated. Use @option{-mspe} and
19591 @option{-mno-spe} instead.
19597 Generate code that uses (does not use) vector/scalar (VSX)
19598 instructions, and also enable the use of built-in functions that allow
19599 more direct access to the VSX instruction set.
19604 @opindex mno-crypto
19605 Enable the use (disable) of the built-in functions that allow direct
19606 access to the cryptographic instructions that were added in version
19607 2.07 of the PowerPC ISA.
19609 @item -mdirect-move
19610 @itemx -mno-direct-move
19611 @opindex mdirect-move
19612 @opindex mno-direct-move
19613 Generate code that uses (does not use) the instructions to move data
19614 between the general purpose registers and the vector/scalar (VSX)
19615 registers that were added in version 2.07 of the PowerPC ISA.
19617 @item -mpower8-fusion
19618 @itemx -mno-power8-fusion
19619 @opindex mpower8-fusion
19620 @opindex mno-power8-fusion
19621 Generate code that keeps (does not keeps) some integer operations
19622 adjacent so that the instructions can be fused together on power8 and
19625 @item -mpower8-vector
19626 @itemx -mno-power8-vector
19627 @opindex mpower8-vector
19628 @opindex mno-power8-vector
19629 Generate code that uses (does not use) the vector and scalar
19630 instructions that were added in version 2.07 of the PowerPC ISA. Also
19631 enable the use of built-in functions that allow more direct access to
19632 the vector instructions.
19634 @item -mquad-memory
19635 @itemx -mno-quad-memory
19636 @opindex mquad-memory
19637 @opindex mno-quad-memory
19638 Generate code that uses (does not use) the non-atomic quad word memory
19639 instructions. The @option{-mquad-memory} option requires use of
19642 @item -mquad-memory-atomic
19643 @itemx -mno-quad-memory-atomic
19644 @opindex mquad-memory-atomic
19645 @opindex mno-quad-memory-atomic
19646 Generate code that uses (does not use) the atomic quad word memory
19647 instructions. The @option{-mquad-memory-atomic} option requires use of
19650 @item -mupper-regs-df
19651 @itemx -mno-upper-regs-df
19652 @opindex mupper-regs-df
19653 @opindex mno-upper-regs-df
19654 Generate code that uses (does not use) the scalar double precision
19655 instructions that target all 64 registers in the vector/scalar
19656 floating point register set that were added in version 2.06 of the
19657 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
19658 use any of the @option{-mcpu=power7}, @option{-mcpu=power8}, or
19659 @option{-mvsx} options.
19661 @item -mupper-regs-sf
19662 @itemx -mno-upper-regs-sf
19663 @opindex mupper-regs-sf
19664 @opindex mno-upper-regs-sf
19665 Generate code that uses (does not use) the scalar single precision
19666 instructions that target all 64 registers in the vector/scalar
19667 floating point register set that were added in version 2.07 of the
19668 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
19669 use either of the @option{-mcpu=power8} or @option{-mpower8-vector}
19673 @itemx -mno-upper-regs
19674 @opindex mupper-regs
19675 @opindex mno-upper-regs
19676 Generate code that uses (does not use) the scalar
19677 instructions that target all 64 registers in the vector/scalar
19678 floating point register set, depending on the model of the machine.
19680 If the @option{-mno-upper-regs} option is used, it turns off both
19681 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
19684 @itemx -mno-float128
19686 @opindex mno-float128
19687 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
19688 and use either software emulation for IEEE 128-bit floating point or
19689 hardware instructions.
19691 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
19692 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
19693 option. The @code{-mfloat128} option only works on PowerPC 64-bit
19696 @item -mfloat128-hardware
19697 @itemx -mno-float128-hardware
19698 @opindex mfloat128-hardware
19699 @opindex mno-float128-hardware
19700 Enable/disable using ISA 3.0 hardware instructions to support the
19701 @var{__float128} data type.
19706 @opindex mno-module
19707 Generate code that uses (does not use) the ISA 3.0 integer modulo
19708 instructions. The @option{-mmodulo} option is enabled by default
19709 with the @option{-mcpu=power9} option.
19711 @item -mpower9-fusion
19712 @itemx -mno-power9-fusion
19713 @opindex mpower9-fusion
19714 @opindex mno-power9-fusion
19715 Generate code that keeps (does not keeps) some operations adjacent so
19716 that the instructions can be fused together on power9 and later
19719 @item -mpower9-vector
19720 @itemx -mno-power9-vector
19721 @opindex mpower9-vector
19722 @opindex mno-power9-vector
19723 Generate code that uses (does not use) the vector and scalar
19724 instructions that were added in version 2.07 of the PowerPC ISA. Also
19725 enable the use of built-in functions that allow more direct access to
19726 the vector instructions.
19728 @item -mfloat-gprs=@var{yes/single/double/no}
19729 @itemx -mfloat-gprs
19730 @opindex mfloat-gprs
19731 This switch enables or disables the generation of floating-point
19732 operations on the general-purpose registers for architectures that
19735 The argument @samp{yes} or @samp{single} enables the use of
19736 single-precision floating-point operations.
19738 The argument @samp{double} enables the use of single and
19739 double-precision floating-point operations.
19741 The argument @samp{no} disables floating-point operations on the
19742 general-purpose registers.
19744 This option is currently only available on the MPC854x.
19750 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
19751 targets (including GNU/Linux). The 32-bit environment sets int, long
19752 and pointer to 32 bits and generates code that runs on any PowerPC
19753 variant. The 64-bit environment sets int to 32 bits and long and
19754 pointer to 64 bits, and generates code for PowerPC64, as for
19755 @option{-mpowerpc64}.
19758 @itemx -mno-fp-in-toc
19759 @itemx -mno-sum-in-toc
19760 @itemx -mminimal-toc
19762 @opindex mno-fp-in-toc
19763 @opindex mno-sum-in-toc
19764 @opindex mminimal-toc
19765 Modify generation of the TOC (Table Of Contents), which is created for
19766 every executable file. The @option{-mfull-toc} option is selected by
19767 default. In that case, GCC allocates at least one TOC entry for
19768 each unique non-automatic variable reference in your program. GCC
19769 also places floating-point constants in the TOC@. However, only
19770 16,384 entries are available in the TOC@.
19772 If you receive a linker error message that saying you have overflowed
19773 the available TOC space, you can reduce the amount of TOC space used
19774 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
19775 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
19776 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
19777 generate code to calculate the sum of an address and a constant at
19778 run time instead of putting that sum into the TOC@. You may specify one
19779 or both of these options. Each causes GCC to produce very slightly
19780 slower and larger code at the expense of conserving TOC space.
19782 If you still run out of space in the TOC even when you specify both of
19783 these options, specify @option{-mminimal-toc} instead. This option causes
19784 GCC to make only one TOC entry for every file. When you specify this
19785 option, GCC produces code that is slower and larger but which
19786 uses extremely little TOC space. You may wish to use this option
19787 only on files that contain less frequently-executed code.
19793 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
19794 @code{long} type, and the infrastructure needed to support them.
19795 Specifying @option{-maix64} implies @option{-mpowerpc64},
19796 while @option{-maix32} disables the 64-bit ABI and
19797 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
19800 @itemx -mno-xl-compat
19801 @opindex mxl-compat
19802 @opindex mno-xl-compat
19803 Produce code that conforms more closely to IBM XL compiler semantics
19804 when using AIX-compatible ABI@. Pass floating-point arguments to
19805 prototyped functions beyond the register save area (RSA) on the stack
19806 in addition to argument FPRs. Do not assume that most significant
19807 double in 128-bit long double value is properly rounded when comparing
19808 values and converting to double. Use XL symbol names for long double
19811 The AIX calling convention was extended but not initially documented to
19812 handle an obscure K&R C case of calling a function that takes the
19813 address of its arguments with fewer arguments than declared. IBM XL
19814 compilers access floating-point arguments that do not fit in the
19815 RSA from the stack when a subroutine is compiled without
19816 optimization. Because always storing floating-point arguments on the
19817 stack is inefficient and rarely needed, this option is not enabled by
19818 default and only is necessary when calling subroutines compiled by IBM
19819 XL compilers without optimization.
19823 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
19824 application written to use message passing with special startup code to
19825 enable the application to run. The system must have PE installed in the
19826 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
19827 must be overridden with the @option{-specs=} option to specify the
19828 appropriate directory location. The Parallel Environment does not
19829 support threads, so the @option{-mpe} option and the @option{-pthread}
19830 option are incompatible.
19832 @item -malign-natural
19833 @itemx -malign-power
19834 @opindex malign-natural
19835 @opindex malign-power
19836 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
19837 @option{-malign-natural} overrides the ABI-defined alignment of larger
19838 types, such as floating-point doubles, on their natural size-based boundary.
19839 The option @option{-malign-power} instructs GCC to follow the ABI-specified
19840 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
19842 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
19846 @itemx -mhard-float
19847 @opindex msoft-float
19848 @opindex mhard-float
19849 Generate code that does not use (uses) the floating-point register set.
19850 Software floating-point emulation is provided if you use the
19851 @option{-msoft-float} option, and pass the option to GCC when linking.
19853 @item -msingle-float
19854 @itemx -mdouble-float
19855 @opindex msingle-float
19856 @opindex mdouble-float
19857 Generate code for single- or double-precision floating-point operations.
19858 @option{-mdouble-float} implies @option{-msingle-float}.
19861 @opindex msimple-fpu
19862 Do not generate @code{sqrt} and @code{div} instructions for hardware
19863 floating-point unit.
19865 @item -mfpu=@var{name}
19867 Specify type of floating-point unit. Valid values for @var{name} are
19868 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
19869 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
19870 @samp{sp_full} (equivalent to @option{-msingle-float}),
19871 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
19874 @opindex mxilinx-fpu
19875 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
19878 @itemx -mno-multiple
19880 @opindex mno-multiple
19881 Generate code that uses (does not use) the load multiple word
19882 instructions and the store multiple word instructions. These
19883 instructions are generated by default on POWER systems, and not
19884 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
19885 PowerPC systems, since those instructions do not work when the
19886 processor is in little-endian mode. The exceptions are PPC740 and
19887 PPC750 which permit these instructions in little-endian mode.
19892 @opindex mno-string
19893 Generate code that uses (does not use) the load string instructions
19894 and the store string word instructions to save multiple registers and
19895 do small block moves. These instructions are generated by default on
19896 POWER systems, and not generated on PowerPC systems. Do not use
19897 @option{-mstring} on little-endian PowerPC systems, since those
19898 instructions do not work when the processor is in little-endian mode.
19899 The exceptions are PPC740 and PPC750 which permit these instructions
19900 in little-endian mode.
19905 @opindex mno-update
19906 Generate code that uses (does not use) the load or store instructions
19907 that update the base register to the address of the calculated memory
19908 location. These instructions are generated by default. If you use
19909 @option{-mno-update}, there is a small window between the time that the
19910 stack pointer is updated and the address of the previous frame is
19911 stored, which means code that walks the stack frame across interrupts or
19912 signals may get corrupted data.
19914 @item -mavoid-indexed-addresses
19915 @itemx -mno-avoid-indexed-addresses
19916 @opindex mavoid-indexed-addresses
19917 @opindex mno-avoid-indexed-addresses
19918 Generate code that tries to avoid (not avoid) the use of indexed load
19919 or store instructions. These instructions can incur a performance
19920 penalty on Power6 processors in certain situations, such as when
19921 stepping through large arrays that cross a 16M boundary. This option
19922 is enabled by default when targeting Power6 and disabled otherwise.
19925 @itemx -mno-fused-madd
19926 @opindex mfused-madd
19927 @opindex mno-fused-madd
19928 Generate code that uses (does not use) the floating-point multiply and
19929 accumulate instructions. These instructions are generated by default
19930 if hardware floating point is used. The machine-dependent
19931 @option{-mfused-madd} option is now mapped to the machine-independent
19932 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
19933 mapped to @option{-ffp-contract=off}.
19939 Generate code that uses (does not use) the half-word multiply and
19940 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
19941 These instructions are generated by default when targeting those
19948 Generate code that uses (does not use) the string-search @samp{dlmzb}
19949 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
19950 generated by default when targeting those processors.
19952 @item -mno-bit-align
19954 @opindex mno-bit-align
19955 @opindex mbit-align
19956 On System V.4 and embedded PowerPC systems do not (do) force structures
19957 and unions that contain bit-fields to be aligned to the base type of the
19960 For example, by default a structure containing nothing but 8
19961 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
19962 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
19963 the structure is aligned to a 1-byte boundary and is 1 byte in
19966 @item -mno-strict-align
19967 @itemx -mstrict-align
19968 @opindex mno-strict-align
19969 @opindex mstrict-align
19970 On System V.4 and embedded PowerPC systems do not (do) assume that
19971 unaligned memory references are handled by the system.
19973 @item -mrelocatable
19974 @itemx -mno-relocatable
19975 @opindex mrelocatable
19976 @opindex mno-relocatable
19977 Generate code that allows (does not allow) a static executable to be
19978 relocated to a different address at run time. A simple embedded
19979 PowerPC system loader should relocate the entire contents of
19980 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
19981 a table of 32-bit addresses generated by this option. For this to
19982 work, all objects linked together must be compiled with
19983 @option{-mrelocatable} or @option{-mrelocatable-lib}.
19984 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
19986 @item -mrelocatable-lib
19987 @itemx -mno-relocatable-lib
19988 @opindex mrelocatable-lib
19989 @opindex mno-relocatable-lib
19990 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
19991 @code{.fixup} section to allow static executables to be relocated at
19992 run time, but @option{-mrelocatable-lib} does not use the smaller stack
19993 alignment of @option{-mrelocatable}. Objects compiled with
19994 @option{-mrelocatable-lib} may be linked with objects compiled with
19995 any combination of the @option{-mrelocatable} options.
20001 On System V.4 and embedded PowerPC systems do not (do) assume that
20002 register 2 contains a pointer to a global area pointing to the addresses
20003 used in the program.
20006 @itemx -mlittle-endian
20008 @opindex mlittle-endian
20009 On System V.4 and embedded PowerPC systems compile code for the
20010 processor in little-endian mode. The @option{-mlittle-endian} option is
20011 the same as @option{-mlittle}.
20014 @itemx -mbig-endian
20016 @opindex mbig-endian
20017 On System V.4 and embedded PowerPC systems compile code for the
20018 processor in big-endian mode. The @option{-mbig-endian} option is
20019 the same as @option{-mbig}.
20021 @item -mdynamic-no-pic
20022 @opindex mdynamic-no-pic
20023 On Darwin and Mac OS X systems, compile code so that it is not
20024 relocatable, but that its external references are relocatable. The
20025 resulting code is suitable for applications, but not shared
20028 @item -msingle-pic-base
20029 @opindex msingle-pic-base
20030 Treat the register used for PIC addressing as read-only, rather than
20031 loading it in the prologue for each function. The runtime system is
20032 responsible for initializing this register with an appropriate value
20033 before execution begins.
20035 @item -mprioritize-restricted-insns=@var{priority}
20036 @opindex mprioritize-restricted-insns
20037 This option controls the priority that is assigned to
20038 dispatch-slot restricted instructions during the second scheduling
20039 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20040 or @samp{2} to assign no, highest, or second-highest (respectively)
20041 priority to dispatch-slot restricted
20044 @item -msched-costly-dep=@var{dependence_type}
20045 @opindex msched-costly-dep
20046 This option controls which dependences are considered costly
20047 by the target during instruction scheduling. The argument
20048 @var{dependence_type} takes one of the following values:
20052 No dependence is costly.
20055 All dependences are costly.
20057 @item @samp{true_store_to_load}
20058 A true dependence from store to load is costly.
20060 @item @samp{store_to_load}
20061 Any dependence from store to load is costly.
20064 Any dependence for which the latency is greater than or equal to
20065 @var{number} is costly.
20068 @item -minsert-sched-nops=@var{scheme}
20069 @opindex minsert-sched-nops
20070 This option controls which NOP insertion scheme is used during
20071 the second scheduling pass. The argument @var{scheme} takes one of the
20079 Pad with NOPs any dispatch group that has vacant issue slots,
20080 according to the scheduler's grouping.
20082 @item @samp{regroup_exact}
20083 Insert NOPs to force costly dependent insns into
20084 separate groups. Insert exactly as many NOPs as needed to force an insn
20085 to a new group, according to the estimated processor grouping.
20088 Insert NOPs to force costly dependent insns into
20089 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20093 @opindex mcall-sysv
20094 On System V.4 and embedded PowerPC systems compile code using calling
20095 conventions that adhere to the March 1995 draft of the System V
20096 Application Binary Interface, PowerPC processor supplement. This is the
20097 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20099 @item -mcall-sysv-eabi
20101 @opindex mcall-sysv-eabi
20102 @opindex mcall-eabi
20103 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20105 @item -mcall-sysv-noeabi
20106 @opindex mcall-sysv-noeabi
20107 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20109 @item -mcall-aixdesc
20111 On System V.4 and embedded PowerPC systems compile code for the AIX
20115 @opindex mcall-linux
20116 On System V.4 and embedded PowerPC systems compile code for the
20117 Linux-based GNU system.
20119 @item -mcall-freebsd
20120 @opindex mcall-freebsd
20121 On System V.4 and embedded PowerPC systems compile code for the
20122 FreeBSD operating system.
20124 @item -mcall-netbsd
20125 @opindex mcall-netbsd
20126 On System V.4 and embedded PowerPC systems compile code for the
20127 NetBSD operating system.
20129 @item -mcall-openbsd
20130 @opindex mcall-netbsd
20131 On System V.4 and embedded PowerPC systems compile code for the
20132 OpenBSD operating system.
20134 @item -maix-struct-return
20135 @opindex maix-struct-return
20136 Return all structures in memory (as specified by the AIX ABI)@.
20138 @item -msvr4-struct-return
20139 @opindex msvr4-struct-return
20140 Return structures smaller than 8 bytes in registers (as specified by the
20143 @item -mabi=@var{abi-type}
20145 Extend the current ABI with a particular extension, or remove such extension.
20146 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20147 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20148 @samp{elfv1}, @samp{elfv2}@.
20152 Extend the current ABI with SPE ABI extensions. This does not change
20153 the default ABI, instead it adds the SPE ABI extensions to the current
20157 @opindex mabi=no-spe
20158 Disable Book-E SPE ABI extensions for the current ABI@.
20160 @item -mabi=ibmlongdouble
20161 @opindex mabi=ibmlongdouble
20162 Change the current ABI to use IBM extended-precision long double.
20163 This is a PowerPC 32-bit SYSV ABI option.
20165 @item -mabi=ieeelongdouble
20166 @opindex mabi=ieeelongdouble
20167 Change the current ABI to use IEEE extended-precision long double.
20168 This is a PowerPC 32-bit Linux ABI option.
20171 @opindex mabi=elfv1
20172 Change the current ABI to use the ELFv1 ABI.
20173 This is the default ABI for big-endian PowerPC 64-bit Linux.
20174 Overriding the default ABI requires special system support and is
20175 likely to fail in spectacular ways.
20178 @opindex mabi=elfv2
20179 Change the current ABI to use the ELFv2 ABI.
20180 This is the default ABI for little-endian PowerPC 64-bit Linux.
20181 Overriding the default ABI requires special system support and is
20182 likely to fail in spectacular ways.
20185 @itemx -mno-prototype
20186 @opindex mprototype
20187 @opindex mno-prototype
20188 On System V.4 and embedded PowerPC systems assume that all calls to
20189 variable argument functions are properly prototyped. Otherwise, the
20190 compiler must insert an instruction before every non-prototyped call to
20191 set or clear bit 6 of the condition code register (@code{CR}) to
20192 indicate whether floating-point values are passed in the floating-point
20193 registers in case the function takes variable arguments. With
20194 @option{-mprototype}, only calls to prototyped variable argument functions
20195 set or clear the bit.
20199 On embedded PowerPC systems, assume that the startup module is called
20200 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20201 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20206 On embedded PowerPC systems, assume that the startup module is called
20207 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20212 On embedded PowerPC systems, assume that the startup module is called
20213 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20216 @item -myellowknife
20217 @opindex myellowknife
20218 On embedded PowerPC systems, assume that the startup module is called
20219 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20224 On System V.4 and embedded PowerPC systems, specify that you are
20225 compiling for a VxWorks system.
20229 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20230 header to indicate that @samp{eabi} extended relocations are used.
20236 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20237 Embedded Applications Binary Interface (EABI), which is a set of
20238 modifications to the System V.4 specifications. Selecting @option{-meabi}
20239 means that the stack is aligned to an 8-byte boundary, a function
20240 @code{__eabi} is called from @code{main} to set up the EABI
20241 environment, and the @option{-msdata} option can use both @code{r2} and
20242 @code{r13} to point to two separate small data areas. Selecting
20243 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20244 no EABI initialization function is called from @code{main}, and the
20245 @option{-msdata} option only uses @code{r13} to point to a single
20246 small data area. The @option{-meabi} option is on by default if you
20247 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20250 @opindex msdata=eabi
20251 On System V.4 and embedded PowerPC systems, put small initialized
20252 @code{const} global and static data in the @code{.sdata2} section, which
20253 is pointed to by register @code{r2}. Put small initialized
20254 non-@code{const} global and static data in the @code{.sdata} section,
20255 which is pointed to by register @code{r13}. Put small uninitialized
20256 global and static data in the @code{.sbss} section, which is adjacent to
20257 the @code{.sdata} section. The @option{-msdata=eabi} option is
20258 incompatible with the @option{-mrelocatable} option. The
20259 @option{-msdata=eabi} option also sets the @option{-memb} option.
20262 @opindex msdata=sysv
20263 On System V.4 and embedded PowerPC systems, put small global and static
20264 data in the @code{.sdata} section, which is pointed to by register
20265 @code{r13}. Put small uninitialized global and static data in the
20266 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20267 The @option{-msdata=sysv} option is incompatible with the
20268 @option{-mrelocatable} option.
20270 @item -msdata=default
20272 @opindex msdata=default
20274 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20275 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20276 same as @option{-msdata=sysv}.
20279 @opindex msdata=data
20280 On System V.4 and embedded PowerPC systems, put small global
20281 data in the @code{.sdata} section. Put small uninitialized global
20282 data in the @code{.sbss} section. Do not use register @code{r13}
20283 to address small data however. This is the default behavior unless
20284 other @option{-msdata} options are used.
20288 @opindex msdata=none
20290 On embedded PowerPC systems, put all initialized global and static data
20291 in the @code{.data} section, and all uninitialized data in the
20292 @code{.bss} section.
20294 @item -mblock-move-inline-limit=@var{num}
20295 @opindex mblock-move-inline-limit
20296 Inline all block moves (such as calls to @code{memcpy} or structure
20297 copies) less than or equal to @var{num} bytes. The minimum value for
20298 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20299 targets. The default value is target-specific.
20303 @cindex smaller data references (PowerPC)
20304 @cindex .sdata/.sdata2 references (PowerPC)
20305 On embedded PowerPC systems, put global and static items less than or
20306 equal to @var{num} bytes into the small data or BSS sections instead of
20307 the normal data or BSS section. By default, @var{num} is 8. The
20308 @option{-G @var{num}} switch is also passed to the linker.
20309 All modules should be compiled with the same @option{-G @var{num}} value.
20312 @itemx -mno-regnames
20314 @opindex mno-regnames
20315 On System V.4 and embedded PowerPC systems do (do not) emit register
20316 names in the assembly language output using symbolic forms.
20319 @itemx -mno-longcall
20321 @opindex mno-longcall
20322 By default assume that all calls are far away so that a longer and more
20323 expensive calling sequence is required. This is required for calls
20324 farther than 32 megabytes (33,554,432 bytes) from the current location.
20325 A short call is generated if the compiler knows
20326 the call cannot be that far away. This setting can be overridden by
20327 the @code{shortcall} function attribute, or by @code{#pragma
20330 Some linkers are capable of detecting out-of-range calls and generating
20331 glue code on the fly. On these systems, long calls are unnecessary and
20332 generate slower code. As of this writing, the AIX linker can do this,
20333 as can the GNU linker for PowerPC/64. It is planned to add this feature
20334 to the GNU linker for 32-bit PowerPC systems as well.
20336 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
20337 callee, L42}, plus a @dfn{branch island} (glue code). The two target
20338 addresses represent the callee and the branch island. The
20339 Darwin/PPC linker prefers the first address and generates a @code{bl
20340 callee} if the PPC @code{bl} instruction reaches the callee directly;
20341 otherwise, the linker generates @code{bl L42} to call the branch
20342 island. The branch island is appended to the body of the
20343 calling function; it computes the full 32-bit address of the callee
20346 On Mach-O (Darwin) systems, this option directs the compiler emit to
20347 the glue for every direct call, and the Darwin linker decides whether
20348 to use or discard it.
20350 In the future, GCC may ignore all longcall specifications
20351 when the linker is known to generate glue.
20353 @item -mtls-markers
20354 @itemx -mno-tls-markers
20355 @opindex mtls-markers
20356 @opindex mno-tls-markers
20357 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
20358 specifying the function argument. The relocation allows the linker to
20359 reliably associate function call with argument setup instructions for
20360 TLS optimization, which in turn allows GCC to better schedule the
20365 Adds support for multithreading with the @dfn{pthreads} library.
20366 This option sets flags for both the preprocessor and linker.
20371 This option enables use of the reciprocal estimate and
20372 reciprocal square root estimate instructions with additional
20373 Newton-Raphson steps to increase precision instead of doing a divide or
20374 square root and divide for floating-point arguments. You should use
20375 the @option{-ffast-math} option when using @option{-mrecip} (or at
20376 least @option{-funsafe-math-optimizations},
20377 @option{-finite-math-only}, @option{-freciprocal-math} and
20378 @option{-fno-trapping-math}). Note that while the throughput of the
20379 sequence is generally higher than the throughput of the non-reciprocal
20380 instruction, the precision of the sequence can be decreased by up to 2
20381 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
20384 @item -mrecip=@var{opt}
20385 @opindex mrecip=opt
20386 This option controls which reciprocal estimate instructions
20387 may be used. @var{opt} is a comma-separated list of options, which may
20388 be preceded by a @code{!} to invert the option:
20393 Enable all estimate instructions.
20396 Enable the default instructions, equivalent to @option{-mrecip}.
20399 Disable all estimate instructions, equivalent to @option{-mno-recip}.
20402 Enable the reciprocal approximation instructions for both
20403 single and double precision.
20406 Enable the single-precision reciprocal approximation instructions.
20409 Enable the double-precision reciprocal approximation instructions.
20412 Enable the reciprocal square root approximation instructions for both
20413 single and double precision.
20416 Enable the single-precision reciprocal square root approximation instructions.
20419 Enable the double-precision reciprocal square root approximation instructions.
20423 So, for example, @option{-mrecip=all,!rsqrtd} enables
20424 all of the reciprocal estimate instructions, except for the
20425 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
20426 which handle the double-precision reciprocal square root calculations.
20428 @item -mrecip-precision
20429 @itemx -mno-recip-precision
20430 @opindex mrecip-precision
20431 Assume (do not assume) that the reciprocal estimate instructions
20432 provide higher-precision estimates than is mandated by the PowerPC
20433 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
20434 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
20435 The double-precision square root estimate instructions are not generated by
20436 default on low-precision machines, since they do not provide an
20437 estimate that converges after three steps.
20439 @item -mveclibabi=@var{type}
20440 @opindex mveclibabi
20441 Specifies the ABI type to use for vectorizing intrinsics using an
20442 external library. The only type supported at present is @samp{mass},
20443 which specifies to use IBM's Mathematical Acceleration Subsystem
20444 (MASS) libraries for vectorizing intrinsics using external libraries.
20445 GCC currently emits calls to @code{acosd2}, @code{acosf4},
20446 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
20447 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
20448 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
20449 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
20450 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
20451 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
20452 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
20453 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
20454 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
20455 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
20456 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
20457 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
20458 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
20459 for power7. Both @option{-ftree-vectorize} and
20460 @option{-funsafe-math-optimizations} must also be enabled. The MASS
20461 libraries must be specified at link time.
20466 Generate (do not generate) the @code{friz} instruction when the
20467 @option{-funsafe-math-optimizations} option is used to optimize
20468 rounding of floating-point values to 64-bit integer and back to floating
20469 point. The @code{friz} instruction does not return the same value if
20470 the floating-point number is too large to fit in an integer.
20472 @item -mpointers-to-nested-functions
20473 @itemx -mno-pointers-to-nested-functions
20474 @opindex mpointers-to-nested-functions
20475 Generate (do not generate) code to load up the static chain register
20476 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
20477 systems where a function pointer points to a 3-word descriptor giving
20478 the function address, TOC value to be loaded in register @code{r2}, and
20479 static chain value to be loaded in register @code{r11}. The
20480 @option{-mpointers-to-nested-functions} is on by default. You cannot
20481 call through pointers to nested functions or pointers
20482 to functions compiled in other languages that use the static chain if
20483 you use @option{-mno-pointers-to-nested-functions}.
20485 @item -msave-toc-indirect
20486 @itemx -mno-save-toc-indirect
20487 @opindex msave-toc-indirect
20488 Generate (do not generate) code to save the TOC value in the reserved
20489 stack location in the function prologue if the function calls through
20490 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
20491 saved in the prologue, it is saved just before the call through the
20492 pointer. The @option{-mno-save-toc-indirect} option is the default.
20494 @item -mcompat-align-parm
20495 @itemx -mno-compat-align-parm
20496 @opindex mcompat-align-parm
20497 Generate (do not generate) code to pass structure parameters with a
20498 maximum alignment of 64 bits, for compatibility with older versions
20501 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
20502 structure parameter on a 128-bit boundary when that structure contained
20503 a member requiring 128-bit alignment. This is corrected in more
20504 recent versions of GCC. This option may be used to generate code
20505 that is compatible with functions compiled with older versions of
20508 The @option{-mno-compat-align-parm} option is the default.
20512 @subsection RX Options
20515 These command-line options are defined for RX targets:
20518 @item -m64bit-doubles
20519 @itemx -m32bit-doubles
20520 @opindex m64bit-doubles
20521 @opindex m32bit-doubles
20522 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
20523 or 32 bits (@option{-m32bit-doubles}) in size. The default is
20524 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
20525 works on 32-bit values, which is why the default is
20526 @option{-m32bit-doubles}.
20532 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
20533 floating-point hardware. The default is enabled for the RX600
20534 series and disabled for the RX200 series.
20536 Floating-point instructions are only generated for 32-bit floating-point
20537 values, however, so the FPU hardware is not used for doubles if the
20538 @option{-m64bit-doubles} option is used.
20540 @emph{Note} If the @option{-fpu} option is enabled then
20541 @option{-funsafe-math-optimizations} is also enabled automatically.
20542 This is because the RX FPU instructions are themselves unsafe.
20544 @item -mcpu=@var{name}
20546 Selects the type of RX CPU to be targeted. Currently three types are
20547 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
20548 the specific @samp{RX610} CPU. The default is @samp{RX600}.
20550 The only difference between @samp{RX600} and @samp{RX610} is that the
20551 @samp{RX610} does not support the @code{MVTIPL} instruction.
20553 The @samp{RX200} series does not have a hardware floating-point unit
20554 and so @option{-nofpu} is enabled by default when this type is
20557 @item -mbig-endian-data
20558 @itemx -mlittle-endian-data
20559 @opindex mbig-endian-data
20560 @opindex mlittle-endian-data
20561 Store data (but not code) in the big-endian format. The default is
20562 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
20565 @item -msmall-data-limit=@var{N}
20566 @opindex msmall-data-limit
20567 Specifies the maximum size in bytes of global and static variables
20568 which can be placed into the small data area. Using the small data
20569 area can lead to smaller and faster code, but the size of area is
20570 limited and it is up to the programmer to ensure that the area does
20571 not overflow. Also when the small data area is used one of the RX's
20572 registers (usually @code{r13}) is reserved for use pointing to this
20573 area, so it is no longer available for use by the compiler. This
20574 could result in slower and/or larger code if variables are pushed onto
20575 the stack instead of being held in this register.
20577 Note, common variables (variables that have not been initialized) and
20578 constants are not placed into the small data area as they are assigned
20579 to other sections in the output executable.
20581 The default value is zero, which disables this feature. Note, this
20582 feature is not enabled by default with higher optimization levels
20583 (@option{-O2} etc) because of the potentially detrimental effects of
20584 reserving a register. It is up to the programmer to experiment and
20585 discover whether this feature is of benefit to their program. See the
20586 description of the @option{-mpid} option for a description of how the
20587 actual register to hold the small data area pointer is chosen.
20593 Use the simulator runtime. The default is to use the libgloss
20594 board-specific runtime.
20596 @item -mas100-syntax
20597 @itemx -mno-as100-syntax
20598 @opindex mas100-syntax
20599 @opindex mno-as100-syntax
20600 When generating assembler output use a syntax that is compatible with
20601 Renesas's AS100 assembler. This syntax can also be handled by the GAS
20602 assembler, but it has some restrictions so it is not generated by default.
20604 @item -mmax-constant-size=@var{N}
20605 @opindex mmax-constant-size
20606 Specifies the maximum size, in bytes, of a constant that can be used as
20607 an operand in a RX instruction. Although the RX instruction set does
20608 allow constants of up to 4 bytes in length to be used in instructions,
20609 a longer value equates to a longer instruction. Thus in some
20610 circumstances it can be beneficial to restrict the size of constants
20611 that are used in instructions. Constants that are too big are instead
20612 placed into a constant pool and referenced via register indirection.
20614 The value @var{N} can be between 0 and 4. A value of 0 (the default)
20615 or 4 means that constants of any size are allowed.
20619 Enable linker relaxation. Linker relaxation is a process whereby the
20620 linker attempts to reduce the size of a program by finding shorter
20621 versions of various instructions. Disabled by default.
20623 @item -mint-register=@var{N}
20624 @opindex mint-register
20625 Specify the number of registers to reserve for fast interrupt handler
20626 functions. The value @var{N} can be between 0 and 4. A value of 1
20627 means that register @code{r13} is reserved for the exclusive use
20628 of fast interrupt handlers. A value of 2 reserves @code{r13} and
20629 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
20630 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
20631 A value of 0, the default, does not reserve any registers.
20633 @item -msave-acc-in-interrupts
20634 @opindex msave-acc-in-interrupts
20635 Specifies that interrupt handler functions should preserve the
20636 accumulator register. This is only necessary if normal code might use
20637 the accumulator register, for example because it performs 64-bit
20638 multiplications. The default is to ignore the accumulator as this
20639 makes the interrupt handlers faster.
20645 Enables the generation of position independent data. When enabled any
20646 access to constant data is done via an offset from a base address
20647 held in a register. This allows the location of constant data to be
20648 determined at run time without requiring the executable to be
20649 relocated, which is a benefit to embedded applications with tight
20650 memory constraints. Data that can be modified is not affected by this
20653 Note, using this feature reserves a register, usually @code{r13}, for
20654 the constant data base address. This can result in slower and/or
20655 larger code, especially in complicated functions.
20657 The actual register chosen to hold the constant data base address
20658 depends upon whether the @option{-msmall-data-limit} and/or the
20659 @option{-mint-register} command-line options are enabled. Starting
20660 with register @code{r13} and proceeding downwards, registers are
20661 allocated first to satisfy the requirements of @option{-mint-register},
20662 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
20663 is possible for the small data area register to be @code{r8} if both
20664 @option{-mint-register=4} and @option{-mpid} are specified on the
20667 By default this feature is not enabled. The default can be restored
20668 via the @option{-mno-pid} command-line option.
20670 @item -mno-warn-multiple-fast-interrupts
20671 @itemx -mwarn-multiple-fast-interrupts
20672 @opindex mno-warn-multiple-fast-interrupts
20673 @opindex mwarn-multiple-fast-interrupts
20674 Prevents GCC from issuing a warning message if it finds more than one
20675 fast interrupt handler when it is compiling a file. The default is to
20676 issue a warning for each extra fast interrupt handler found, as the RX
20677 only supports one such interrupt.
20679 @item -mallow-string-insns
20680 @itemx -mno-allow-string-insns
20681 @opindex mallow-string-insns
20682 @opindex mno-allow-string-insns
20683 Enables or disables the use of the string manipulation instructions
20684 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
20685 @code{SWHILE} and also the @code{RMPA} instruction. These
20686 instructions may prefetch data, which is not safe to do if accessing
20687 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
20688 for more information).
20690 The default is to allow these instructions, but it is not possible for
20691 GCC to reliably detect all circumstances where a string instruction
20692 might be used to access an I/O register, so their use cannot be
20693 disabled automatically. Instead it is reliant upon the programmer to
20694 use the @option{-mno-allow-string-insns} option if their program
20695 accesses I/O space.
20697 When the instructions are enabled GCC defines the C preprocessor
20698 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
20699 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
20705 Use only (or not only) @code{JSR} instructions to access functions.
20706 This option can be used when code size exceeds the range of @code{BSR}
20707 instructions. Note that @option{-mno-jsr} does not mean to not use
20708 @code{JSR} but instead means that any type of branch may be used.
20711 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
20712 has special significance to the RX port when used with the
20713 @code{interrupt} function attribute. This attribute indicates a
20714 function intended to process fast interrupts. GCC ensures
20715 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
20716 and/or @code{r13} and only provided that the normal use of the
20717 corresponding registers have been restricted via the
20718 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
20721 @node S/390 and zSeries Options
20722 @subsection S/390 and zSeries Options
20723 @cindex S/390 and zSeries Options
20725 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
20729 @itemx -msoft-float
20730 @opindex mhard-float
20731 @opindex msoft-float
20732 Use (do not use) the hardware floating-point instructions and registers
20733 for floating-point operations. When @option{-msoft-float} is specified,
20734 functions in @file{libgcc.a} are used to perform floating-point
20735 operations. When @option{-mhard-float} is specified, the compiler
20736 generates IEEE floating-point instructions. This is the default.
20739 @itemx -mno-hard-dfp
20741 @opindex mno-hard-dfp
20742 Use (do not use) the hardware decimal-floating-point instructions for
20743 decimal-floating-point operations. When @option{-mno-hard-dfp} is
20744 specified, functions in @file{libgcc.a} are used to perform
20745 decimal-floating-point operations. When @option{-mhard-dfp} is
20746 specified, the compiler generates decimal-floating-point hardware
20747 instructions. This is the default for @option{-march=z9-ec} or higher.
20749 @item -mlong-double-64
20750 @itemx -mlong-double-128
20751 @opindex mlong-double-64
20752 @opindex mlong-double-128
20753 These switches control the size of @code{long double} type. A size
20754 of 64 bits makes the @code{long double} type equivalent to the @code{double}
20755 type. This is the default.
20758 @itemx -mno-backchain
20759 @opindex mbackchain
20760 @opindex mno-backchain
20761 Store (do not store) the address of the caller's frame as backchain pointer
20762 into the callee's stack frame.
20763 A backchain may be needed to allow debugging using tools that do not understand
20764 DWARF 2 call frame information.
20765 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
20766 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
20767 the backchain is placed into the topmost word of the 96/160 byte register
20770 In general, code compiled with @option{-mbackchain} is call-compatible with
20771 code compiled with @option{-mmo-backchain}; however, use of the backchain
20772 for debugging purposes usually requires that the whole binary is built with
20773 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
20774 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20775 to build a linux kernel use @option{-msoft-float}.
20777 The default is to not maintain the backchain.
20779 @item -mpacked-stack
20780 @itemx -mno-packed-stack
20781 @opindex mpacked-stack
20782 @opindex mno-packed-stack
20783 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
20784 specified, the compiler uses the all fields of the 96/160 byte register save
20785 area only for their default purpose; unused fields still take up stack space.
20786 When @option{-mpacked-stack} is specified, register save slots are densely
20787 packed at the top of the register save area; unused space is reused for other
20788 purposes, allowing for more efficient use of the available stack space.
20789 However, when @option{-mbackchain} is also in effect, the topmost word of
20790 the save area is always used to store the backchain, and the return address
20791 register is always saved two words below the backchain.
20793 As long as the stack frame backchain is not used, code generated with
20794 @option{-mpacked-stack} is call-compatible with code generated with
20795 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
20796 S/390 or zSeries generated code that uses the stack frame backchain at run
20797 time, not just for debugging purposes. Such code is not call-compatible
20798 with code compiled with @option{-mpacked-stack}. Also, note that the
20799 combination of @option{-mbackchain},
20800 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
20801 to build a linux kernel use @option{-msoft-float}.
20803 The default is to not use the packed stack layout.
20806 @itemx -mno-small-exec
20807 @opindex msmall-exec
20808 @opindex mno-small-exec
20809 Generate (or do not generate) code using the @code{bras} instruction
20810 to do subroutine calls.
20811 This only works reliably if the total executable size does not
20812 exceed 64k. The default is to use the @code{basr} instruction instead,
20813 which does not have this limitation.
20819 When @option{-m31} is specified, generate code compliant to the
20820 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
20821 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
20822 particular to generate 64-bit instructions. For the @samp{s390}
20823 targets, the default is @option{-m31}, while the @samp{s390x}
20824 targets default to @option{-m64}.
20830 When @option{-mzarch} is specified, generate code using the
20831 instructions available on z/Architecture.
20832 When @option{-mesa} is specified, generate code using the
20833 instructions available on ESA/390. Note that @option{-mesa} is
20834 not possible with @option{-m64}.
20835 When generating code compliant to the GNU/Linux for S/390 ABI,
20836 the default is @option{-mesa}. When generating code compliant
20837 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
20843 The @option{-mhtm} option enables a set of builtins making use of
20844 instructions available with the transactional execution facility
20845 introduced with the IBM zEnterprise EC12 machine generation
20846 @ref{S/390 System z Built-in Functions}.
20847 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
20853 When @option{-mvx} is specified, generate code using the instructions
20854 available with the vector extension facility introduced with the IBM
20855 z13 machine generation.
20856 This option changes the ABI for some vector type values with regard to
20857 alignment and calling conventions. In case vector type values are
20858 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
20859 command will be added to mark the resulting binary with the ABI used.
20860 @option{-mvx} is enabled by default when using @option{-march=z13}.
20863 @itemx -mno-zvector
20865 @opindex mno-zvector
20866 The @option{-mzvector} option enables vector language extensions and
20867 builtins using instructions available with the vector extension
20868 facility introduced with the IBM z13 machine generation.
20869 This option adds support for @samp{vector} to be used as a keyword to
20870 define vector type variables and arguments. @samp{vector} is only
20871 available when GNU extensions are enabled. It will not be expanded
20872 when requesting strict standard compliance e.g. with @option{-std=c99}.
20873 In addition to the GCC low-level builtins @option{-mzvector} enables
20874 a set of builtins added for compatibility with Altivec-style
20875 implementations like Power and Cell. In order to make use of these
20876 builtins the header file @file{vecintrin.h} needs to be included.
20877 @option{-mzvector} is disabled by default.
20883 Generate (or do not generate) code using the @code{mvcle} instruction
20884 to perform block moves. When @option{-mno-mvcle} is specified,
20885 use a @code{mvc} loop instead. This is the default unless optimizing for
20892 Print (or do not print) additional debug information when compiling.
20893 The default is to not print debug information.
20895 @item -march=@var{cpu-type}
20897 Generate code that runs on @var{cpu-type}, which is the name of a
20898 system representing a certain processor type. Possible values for
20899 @var{cpu-type} are @samp{z900}, @samp{z990}, @samp{z9-109},
20900 @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12}, and @samp{z13}.
20901 The default is @option{-march=z900}. @samp{g5} and @samp{g6} are
20902 deprecated and will be removed with future releases.
20904 @item -mtune=@var{cpu-type}
20906 Tune to @var{cpu-type} everything applicable about the generated code,
20907 except for the ABI and the set of available instructions.
20908 The list of @var{cpu-type} values is the same as for @option{-march}.
20909 The default is the value used for @option{-march}.
20912 @itemx -mno-tpf-trace
20913 @opindex mtpf-trace
20914 @opindex mno-tpf-trace
20915 Generate code that adds (does not add) in TPF OS specific branches to trace
20916 routines in the operating system. This option is off by default, even
20917 when compiling for the TPF OS@.
20920 @itemx -mno-fused-madd
20921 @opindex mfused-madd
20922 @opindex mno-fused-madd
20923 Generate code that uses (does not use) the floating-point multiply and
20924 accumulate instructions. These instructions are generated by default if
20925 hardware floating point is used.
20927 @item -mwarn-framesize=@var{framesize}
20928 @opindex mwarn-framesize
20929 Emit a warning if the current function exceeds the given frame size. Because
20930 this is a compile-time check it doesn't need to be a real problem when the program
20931 runs. It is intended to identify functions that most probably cause
20932 a stack overflow. It is useful to be used in an environment with limited stack
20933 size e.g.@: the linux kernel.
20935 @item -mwarn-dynamicstack
20936 @opindex mwarn-dynamicstack
20937 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
20938 arrays. This is generally a bad idea with a limited stack size.
20940 @item -mstack-guard=@var{stack-guard}
20941 @itemx -mstack-size=@var{stack-size}
20942 @opindex mstack-guard
20943 @opindex mstack-size
20944 If these options are provided the S/390 back end emits additional instructions in
20945 the function prologue that trigger a trap if the stack size is @var{stack-guard}
20946 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
20947 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
20948 the frame size of the compiled function is chosen.
20949 These options are intended to be used to help debugging stack overflow problems.
20950 The additionally emitted code causes only little overhead and hence can also be
20951 used in production-like systems without greater performance degradation. The given
20952 values have to be exact powers of 2 and @var{stack-size} has to be greater than
20953 @var{stack-guard} without exceeding 64k.
20954 In order to be efficient the extra code makes the assumption that the stack starts
20955 at an address aligned to the value given by @var{stack-size}.
20956 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
20958 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
20960 If the hotpatch option is enabled, a ``hot-patching'' function
20961 prologue is generated for all functions in the compilation unit.
20962 The funtion label is prepended with the given number of two-byte
20963 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
20964 the label, 2 * @var{post-halfwords} bytes are appended, using the
20965 largest NOP like instructions the architecture allows (maximum
20968 If both arguments are zero, hotpatching is disabled.
20970 This option can be overridden for individual functions with the
20971 @code{hotpatch} attribute.
20974 @node Score Options
20975 @subsection Score Options
20976 @cindex Score Options
20978 These options are defined for Score implementations:
20983 Compile code for big-endian mode. This is the default.
20987 Compile code for little-endian mode.
20991 Disable generation of @code{bcnz} instructions.
20995 Enable generation of unaligned load and store instructions.
20999 Enable the use of multiply-accumulate instructions. Disabled by default.
21003 Specify the SCORE5 as the target architecture.
21007 Specify the SCORE5U of the target architecture.
21011 Specify the SCORE7 as the target architecture. This is the default.
21015 Specify the SCORE7D as the target architecture.
21019 @subsection SH Options
21021 These @samp{-m} options are defined for the SH implementations:
21026 Generate code for the SH1.
21030 Generate code for the SH2.
21033 Generate code for the SH2e.
21037 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21038 that the floating-point unit is not used.
21040 @item -m2a-single-only
21041 @opindex m2a-single-only
21042 Generate code for the SH2a-FPU, in such a way that no double-precision
21043 floating-point operations are used.
21046 @opindex m2a-single
21047 Generate code for the SH2a-FPU assuming the floating-point unit is in
21048 single-precision mode by default.
21052 Generate code for the SH2a-FPU assuming the floating-point unit is in
21053 double-precision mode by default.
21057 Generate code for the SH3.
21061 Generate code for the SH3e.
21065 Generate code for the SH4 without a floating-point unit.
21067 @item -m4-single-only
21068 @opindex m4-single-only
21069 Generate code for the SH4 with a floating-point unit that only
21070 supports single-precision arithmetic.
21074 Generate code for the SH4 assuming the floating-point unit is in
21075 single-precision mode by default.
21079 Generate code for the SH4.
21083 Generate code for SH4-100.
21085 @item -m4-100-nofpu
21086 @opindex m4-100-nofpu
21087 Generate code for SH4-100 in such a way that the
21088 floating-point unit is not used.
21090 @item -m4-100-single
21091 @opindex m4-100-single
21092 Generate code for SH4-100 assuming the floating-point unit is in
21093 single-precision mode by default.
21095 @item -m4-100-single-only
21096 @opindex m4-100-single-only
21097 Generate code for SH4-100 in such a way that no double-precision
21098 floating-point operations are used.
21102 Generate code for SH4-200.
21104 @item -m4-200-nofpu
21105 @opindex m4-200-nofpu
21106 Generate code for SH4-200 without in such a way that the
21107 floating-point unit is not used.
21109 @item -m4-200-single
21110 @opindex m4-200-single
21111 Generate code for SH4-200 assuming the floating-point unit is in
21112 single-precision mode by default.
21114 @item -m4-200-single-only
21115 @opindex m4-200-single-only
21116 Generate code for SH4-200 in such a way that no double-precision
21117 floating-point operations are used.
21121 Generate code for SH4-300.
21123 @item -m4-300-nofpu
21124 @opindex m4-300-nofpu
21125 Generate code for SH4-300 without in such a way that the
21126 floating-point unit is not used.
21128 @item -m4-300-single
21129 @opindex m4-300-single
21130 Generate code for SH4-300 in such a way that no double-precision
21131 floating-point operations are used.
21133 @item -m4-300-single-only
21134 @opindex m4-300-single-only
21135 Generate code for SH4-300 in such a way that no double-precision
21136 floating-point operations are used.
21140 Generate code for SH4-340 (no MMU, no FPU).
21144 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21149 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21150 floating-point unit is not used.
21152 @item -m4a-single-only
21153 @opindex m4a-single-only
21154 Generate code for the SH4a, in such a way that no double-precision
21155 floating-point operations are used.
21158 @opindex m4a-single
21159 Generate code for the SH4a assuming the floating-point unit is in
21160 single-precision mode by default.
21164 Generate code for the SH4a.
21168 Same as @option{-m4a-nofpu}, except that it implicitly passes
21169 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21170 instructions at the moment.
21174 Compile code for the processor in big-endian mode.
21178 Compile code for the processor in little-endian mode.
21182 Align doubles at 64-bit boundaries. Note that this changes the calling
21183 conventions, and thus some functions from the standard C library do
21184 not work unless you recompile it first with @option{-mdalign}.
21188 Shorten some address references at link time, when possible; uses the
21189 linker option @option{-relax}.
21193 Use 32-bit offsets in @code{switch} tables. The default is to use
21198 Enable the use of bit manipulation instructions on SH2A.
21202 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21203 alignment constraints.
21207 Comply with the calling conventions defined by Renesas.
21210 @opindex mno-renesas
21211 Comply with the calling conventions defined for GCC before the Renesas
21212 conventions were available. This option is the default for all
21213 targets of the SH toolchain.
21216 @opindex mnomacsave
21217 Mark the @code{MAC} register as call-clobbered, even if
21218 @option{-mrenesas} is given.
21224 Control the IEEE compliance of floating-point comparisons, which affects the
21225 handling of cases where the result of a comparison is unordered. By default
21226 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21227 enabled @option{-mno-ieee} is implicitly set, which results in faster
21228 floating-point greater-equal and less-equal comparisons. The implcit settings
21229 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21231 @item -minline-ic_invalidate
21232 @opindex minline-ic_invalidate
21233 Inline code to invalidate instruction cache entries after setting up
21234 nested function trampolines.
21235 This option has no effect if @option{-musermode} is in effect and the selected
21236 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21238 If the selected code generation option does not allow the use of the @code{icbi}
21239 instruction, and @option{-musermode} is not in effect, the inlined code
21240 manipulates the instruction cache address array directly with an associative
21241 write. This not only requires privileged mode at run time, but it also
21242 fails if the cache line had been mapped via the TLB and has become unmapped.
21246 Dump instruction size and location in the assembly code.
21249 @opindex mpadstruct
21250 This option is deprecated. It pads structures to multiple of 4 bytes,
21251 which is incompatible with the SH ABI@.
21253 @item -matomic-model=@var{model}
21254 @opindex matomic-model=@var{model}
21255 Sets the model of atomic operations and additional parameters as a comma
21256 separated list. For details on the atomic built-in functions see
21257 @ref{__atomic Builtins}. The following models and parameters are supported:
21262 Disable compiler generated atomic sequences and emit library calls for atomic
21263 operations. This is the default if the target is not @code{sh*-*-linux*}.
21266 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21267 built-in functions. The generated atomic sequences require additional support
21268 from the interrupt/exception handling code of the system and are only suitable
21269 for SH3* and SH4* single-core systems. This option is enabled by default when
21270 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21271 this option also partially utilizes the hardware atomic instructions
21272 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21273 @samp{strict} is specified.
21276 Generate software atomic sequences that use a variable in the thread control
21277 block. This is a variation of the gUSA sequences which can also be used on
21278 SH1* and SH2* targets. The generated atomic sequences require additional
21279 support from the interrupt/exception handling code of the system and are only
21280 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21281 parameter has to be specified as well.
21284 Generate software atomic sequences that temporarily disable interrupts by
21285 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21286 in privileged mode and is only suitable for single-core systems. Additional
21287 support from the interrupt/exception handling code of the system is not
21288 required. This model is enabled by default when the target is
21289 @code{sh*-*-linux*} and SH1* or SH2*.
21292 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21293 instructions only. This is only available on SH4A and is suitable for
21294 multi-core systems. Since the hardware instructions support only 32 bit atomic
21295 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21296 Code compiled with this option is also compatible with other software
21297 atomic model interrupt/exception handling systems if executed on an SH4A
21298 system. Additional support from the interrupt/exception handling code of the
21299 system is not required for this model.
21302 This parameter specifies the offset in bytes of the variable in the thread
21303 control block structure that should be used by the generated atomic sequences
21304 when the @samp{soft-tcb} model has been selected. For other models this
21305 parameter is ignored. The specified value must be an integer multiple of four
21306 and in the range 0-1020.
21309 This parameter prevents mixed usage of multiple atomic models, even if they
21310 are compatible, and makes the compiler generate atomic sequences of the
21311 specified model only.
21317 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21318 Notice that depending on the particular hardware and software configuration
21319 this can degrade overall performance due to the operand cache line flushes
21320 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21321 processors the @code{tas.b} instruction must be used with caution since it
21322 can result in data corruption for certain cache configurations.
21325 @opindex mprefergot
21326 When generating position-independent code, emit function calls using
21327 the Global Offset Table instead of the Procedure Linkage Table.
21330 @itemx -mno-usermode
21332 @opindex mno-usermode
21333 Don't allow (allow) the compiler generating privileged mode code. Specifying
21334 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21335 inlined code would not work in user mode. @option{-musermode} is the default
21336 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
21337 @option{-musermode} has no effect, since there is no user mode.
21339 @item -multcost=@var{number}
21340 @opindex multcost=@var{number}
21341 Set the cost to assume for a multiply insn.
21343 @item -mdiv=@var{strategy}
21344 @opindex mdiv=@var{strategy}
21345 Set the division strategy to be used for integer division operations.
21346 @var{strategy} can be one of:
21351 Calls a library function that uses the single-step division instruction
21352 @code{div1} to perform the operation. Division by zero calculates an
21353 unspecified result and does not trap. This is the default except for SH4,
21354 SH2A and SHcompact.
21357 Calls a library function that performs the operation in double precision
21358 floating point. Division by zero causes a floating-point exception. This is
21359 the default for SHcompact with FPU. Specifying this for targets that do not
21360 have a double precision FPU defaults to @code{call-div1}.
21363 Calls a library function that uses a lookup table for small divisors and
21364 the @code{div1} instruction with case distinction for larger divisors. Division
21365 by zero calculates an unspecified result and does not trap. This is the default
21366 for SH4. Specifying this for targets that do not have dynamic shift
21367 instructions defaults to @code{call-div1}.
21371 When a division strategy has not been specified the default strategy is
21372 selected based on the current target. For SH2A the default strategy is to
21373 use the @code{divs} and @code{divu} instructions instead of library function
21376 @item -maccumulate-outgoing-args
21377 @opindex maccumulate-outgoing-args
21378 Reserve space once for outgoing arguments in the function prologue rather
21379 than around each call. Generally beneficial for performance and size. Also
21380 needed for unwinding to avoid changing the stack frame around conditional code.
21382 @item -mdivsi3_libfunc=@var{name}
21383 @opindex mdivsi3_libfunc=@var{name}
21384 Set the name of the library function used for 32-bit signed division to
21386 This only affects the name used in the @samp{call} division strategies, and
21387 the compiler still expects the same sets of input/output/clobbered registers as
21388 if this option were not present.
21390 @item -mfixed-range=@var{register-range}
21391 @opindex mfixed-range
21392 Generate code treating the given register range as fixed registers.
21393 A fixed register is one that the register allocator can not use. This is
21394 useful when compiling kernel code. A register range is specified as
21395 two registers separated by a dash. Multiple register ranges can be
21396 specified separated by a comma.
21398 @item -mbranch-cost=@var{num}
21399 @opindex mbranch-cost=@var{num}
21400 Assume @var{num} to be the cost for a branch instruction. Higher numbers
21401 make the compiler try to generate more branch-free code if possible.
21402 If not specified the value is selected depending on the processor type that
21403 is being compiled for.
21406 @itemx -mno-zdcbranch
21407 @opindex mzdcbranch
21408 @opindex mno-zdcbranch
21409 Assume (do not assume) that zero displacement conditional branch instructions
21410 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
21411 compiler prefers zero displacement branch code sequences. This is
21412 enabled by default when generating code for SH4 and SH4A. It can be explicitly
21413 disabled by specifying @option{-mno-zdcbranch}.
21415 @item -mcbranch-force-delay-slot
21416 @opindex mcbranch-force-delay-slot
21417 Force the usage of delay slots for conditional branches, which stuffs the delay
21418 slot with a @code{nop} if a suitable instruction can't be found. By default
21419 this option is disabled. It can be enabled to work around hardware bugs as
21420 found in the original SH7055.
21423 @itemx -mno-fused-madd
21424 @opindex mfused-madd
21425 @opindex mno-fused-madd
21426 Generate code that uses (does not use) the floating-point multiply and
21427 accumulate instructions. These instructions are generated by default
21428 if hardware floating point is used. The machine-dependent
21429 @option{-mfused-madd} option is now mapped to the machine-independent
21430 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
21431 mapped to @option{-ffp-contract=off}.
21437 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
21438 and cosine approximations. The option @option{-mfsca} must be used in
21439 combination with @option{-funsafe-math-optimizations}. It is enabled by default
21440 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
21441 approximations even if @option{-funsafe-math-optimizations} is in effect.
21447 Allow or disallow the compiler to emit the @code{fsrra} instruction for
21448 reciprocal square root approximations. The option @option{-mfsrra} must be used
21449 in combination with @option{-funsafe-math-optimizations} and
21450 @option{-ffinite-math-only}. It is enabled by default when generating code for
21451 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
21452 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
21455 @item -mpretend-cmove
21456 @opindex mpretend-cmove
21457 Prefer zero-displacement conditional branches for conditional move instruction
21458 patterns. This can result in faster code on the SH4 processor.
21462 Generate code using the FDPIC ABI.
21466 @node Solaris 2 Options
21467 @subsection Solaris 2 Options
21468 @cindex Solaris 2 options
21470 These @samp{-m} options are supported on Solaris 2:
21473 @item -mclear-hwcap
21474 @opindex mclear-hwcap
21475 @option{-mclear-hwcap} tells the compiler to remove the hardware
21476 capabilities generated by the Solaris assembler. This is only necessary
21477 when object files use ISA extensions not supported by the current
21478 machine, but check at runtime whether or not to use them.
21480 @item -mimpure-text
21481 @opindex mimpure-text
21482 @option{-mimpure-text}, used in addition to @option{-shared}, tells
21483 the compiler to not pass @option{-z text} to the linker when linking a
21484 shared object. Using this option, you can link position-dependent
21485 code into a shared object.
21487 @option{-mimpure-text} suppresses the ``relocations remain against
21488 allocatable but non-writable sections'' linker error message.
21489 However, the necessary relocations trigger copy-on-write, and the
21490 shared object is not actually shared across processes. Instead of
21491 using @option{-mimpure-text}, you should compile all source code with
21492 @option{-fpic} or @option{-fPIC}.
21496 These switches are supported in addition to the above on Solaris 2:
21501 Add support for multithreading using the POSIX threads library. This
21502 option sets flags for both the preprocessor and linker. This option does
21503 not affect the thread safety of object code produced by the compiler or
21504 that of libraries supplied with it.
21508 This is a synonym for @option{-pthreads}.
21511 @node SPARC Options
21512 @subsection SPARC Options
21513 @cindex SPARC options
21515 These @samp{-m} options are supported on the SPARC:
21518 @item -mno-app-regs
21520 @opindex mno-app-regs
21522 Specify @option{-mapp-regs} to generate output using the global registers
21523 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
21524 global register 1, each global register 2 through 4 is then treated as an
21525 allocable register that is clobbered by function calls. This is the default.
21527 To be fully SVR4 ABI-compliant at the cost of some performance loss,
21528 specify @option{-mno-app-regs}. You should compile libraries and system
21529 software with this option.
21535 With @option{-mflat}, the compiler does not generate save/restore instructions
21536 and uses a ``flat'' or single register window model. This model is compatible
21537 with the regular register window model. The local registers and the input
21538 registers (0--5) are still treated as ``call-saved'' registers and are
21539 saved on the stack as needed.
21541 With @option{-mno-flat} (the default), the compiler generates save/restore
21542 instructions (except for leaf functions). This is the normal operating mode.
21545 @itemx -mhard-float
21547 @opindex mhard-float
21548 Generate output containing floating-point instructions. This is the
21552 @itemx -msoft-float
21554 @opindex msoft-float
21555 Generate output containing library calls for floating point.
21556 @strong{Warning:} the requisite libraries are not available for all SPARC
21557 targets. Normally the facilities of the machine's usual C compiler are
21558 used, but this cannot be done directly in cross-compilation. You must make
21559 your own arrangements to provide suitable library functions for
21560 cross-compilation. The embedded targets @samp{sparc-*-aout} and
21561 @samp{sparclite-*-*} do provide software floating-point support.
21563 @option{-msoft-float} changes the calling convention in the output file;
21564 therefore, it is only useful if you compile @emph{all} of a program with
21565 this option. In particular, you need to compile @file{libgcc.a}, the
21566 library that comes with GCC, with @option{-msoft-float} in order for
21569 @item -mhard-quad-float
21570 @opindex mhard-quad-float
21571 Generate output containing quad-word (long double) floating-point
21574 @item -msoft-quad-float
21575 @opindex msoft-quad-float
21576 Generate output containing library calls for quad-word (long double)
21577 floating-point instructions. The functions called are those specified
21578 in the SPARC ABI@. This is the default.
21580 As of this writing, there are no SPARC implementations that have hardware
21581 support for the quad-word floating-point instructions. They all invoke
21582 a trap handler for one of these instructions, and then the trap handler
21583 emulates the effect of the instruction. Because of the trap handler overhead,
21584 this is much slower than calling the ABI library routines. Thus the
21585 @option{-msoft-quad-float} option is the default.
21587 @item -mno-unaligned-doubles
21588 @itemx -munaligned-doubles
21589 @opindex mno-unaligned-doubles
21590 @opindex munaligned-doubles
21591 Assume that doubles have 8-byte alignment. This is the default.
21593 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
21594 alignment only if they are contained in another type, or if they have an
21595 absolute address. Otherwise, it assumes they have 4-byte alignment.
21596 Specifying this option avoids some rare compatibility problems with code
21597 generated by other compilers. It is not the default because it results
21598 in a performance loss, especially for floating-point code.
21601 @itemx -mno-user-mode
21602 @opindex muser-mode
21603 @opindex mno-user-mode
21604 Do not generate code that can only run in supervisor mode. This is relevant
21605 only for the @code{casa} instruction emitted for the LEON3 processor. This
21608 @item -mfaster-structs
21609 @itemx -mno-faster-structs
21610 @opindex mfaster-structs
21611 @opindex mno-faster-structs
21612 With @option{-mfaster-structs}, the compiler assumes that structures
21613 should have 8-byte alignment. This enables the use of pairs of
21614 @code{ldd} and @code{std} instructions for copies in structure
21615 assignment, in place of twice as many @code{ld} and @code{st} pairs.
21616 However, the use of this changed alignment directly violates the SPARC
21617 ABI@. Thus, it's intended only for use on targets where the developer
21618 acknowledges that their resulting code is not directly in line with
21619 the rules of the ABI@.
21621 @item -mstd-struct-return
21622 @itemx -mno-std-struct-return
21623 @opindex mstd-struct-return
21624 @opindex mno-std-struct-return
21625 With @option{-mstd-struct-return}, the compiler generates checking code
21626 in functions returning structures or unions to detect size mismatches
21627 between the two sides of function calls, as per the 32-bit ABI@.
21629 The default is @option{-mno-std-struct-return}. This option has no effect
21632 @item -mcpu=@var{cpu_type}
21634 Set the instruction set, register set, and instruction scheduling parameters
21635 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
21636 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
21637 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
21638 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
21639 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21640 @samp{niagara3} and @samp{niagara4}.
21642 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
21643 which selects the best architecture option for the host processor.
21644 @option{-mcpu=native} has no effect if GCC does not recognize
21647 Default instruction scheduling parameters are used for values that select
21648 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
21649 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
21651 Here is a list of each supported architecture and their supported
21659 supersparc, hypersparc, leon, leon3
21662 f930, f934, sparclite86x
21668 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4
21671 By default (unless configured otherwise), GCC generates code for the V7
21672 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
21673 additionally optimizes it for the Cypress CY7C602 chip, as used in the
21674 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
21675 SPARCStation 1, 2, IPX etc.
21677 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
21678 architecture. The only difference from V7 code is that the compiler emits
21679 the integer multiply and integer divide instructions which exist in SPARC-V8
21680 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
21681 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
21684 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
21685 the SPARC architecture. This adds the integer multiply, integer divide step
21686 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
21687 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
21688 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
21689 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
21690 MB86934 chip, which is the more recent SPARClite with FPU@.
21692 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
21693 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
21694 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
21695 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
21696 optimizes it for the TEMIC SPARClet chip.
21698 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
21699 architecture. This adds 64-bit integer and floating-point move instructions,
21700 3 additional floating-point condition code registers and conditional move
21701 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
21702 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
21703 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
21704 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
21705 @option{-mcpu=niagara}, the compiler additionally optimizes it for
21706 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
21707 additionally optimizes it for Sun UltraSPARC T2 chips. With
21708 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
21709 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
21710 additionally optimizes it for Sun UltraSPARC T4 chips.
21712 @item -mtune=@var{cpu_type}
21714 Set the instruction scheduling parameters for machine type
21715 @var{cpu_type}, but do not set the instruction set or register set that the
21716 option @option{-mcpu=@var{cpu_type}} does.
21718 The same values for @option{-mcpu=@var{cpu_type}} can be used for
21719 @option{-mtune=@var{cpu_type}}, but the only useful values are those
21720 that select a particular CPU implementation. Those are @samp{cypress},
21721 @samp{supersparc}, @samp{hypersparc}, @samp{leon}, @samp{leon3},
21722 @samp{leon3v7}, @samp{f930}, @samp{f934}, @samp{sparclite86x}, @samp{tsc701},
21723 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
21724 @samp{niagara3} and @samp{niagara4}. With native Solaris and GNU/Linux
21725 toolchains, @samp{native} can also be used.
21730 @opindex mno-v8plus
21731 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
21732 difference from the V8 ABI is that the global and out registers are
21733 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
21734 mode for all SPARC-V9 processors.
21740 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
21741 Visual Instruction Set extensions. The default is @option{-mno-vis}.
21747 With @option{-mvis2}, GCC generates code that takes advantage of
21748 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
21749 default is @option{-mvis2} when targeting a cpu that supports such
21750 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
21751 also sets @option{-mvis}.
21757 With @option{-mvis3}, GCC generates code that takes advantage of
21758 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
21759 default is @option{-mvis3} when targeting a cpu that supports such
21760 instructions, such as niagara-3 and later. Setting @option{-mvis3}
21761 also sets @option{-mvis2} and @option{-mvis}.
21766 @opindex mno-cbcond
21767 With @option{-mcbcond}, GCC generates code that takes advantage of
21768 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
21769 The default is @option{-mcbcond} when targeting a cpu that supports such
21770 instructions, such as niagara-4 and later.
21776 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
21777 population count instruction. The default is @option{-mpopc}
21778 when targeting a cpu that supports such instructions, such as Niagara-2 and
21785 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
21786 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
21787 when targeting a cpu that supports such instructions, such as Niagara-3 and
21791 @opindex mfix-at697f
21792 Enable the documented workaround for the single erratum of the Atmel AT697F
21793 processor (which corresponds to erratum #13 of the AT697E processor).
21796 @opindex mfix-ut699
21797 Enable the documented workarounds for the floating-point errata and the data
21798 cache nullify errata of the UT699 processor.
21801 These @samp{-m} options are supported in addition to the above
21802 on SPARC-V9 processors in 64-bit environments:
21809 Generate code for a 32-bit or 64-bit environment.
21810 The 32-bit environment sets int, long and pointer to 32 bits.
21811 The 64-bit environment sets int to 32 bits and long and pointer
21814 @item -mcmodel=@var{which}
21816 Set the code model to one of
21820 The Medium/Low code model: 64-bit addresses, programs
21821 must be linked in the low 32 bits of memory. Programs can be statically
21822 or dynamically linked.
21825 The Medium/Middle code model: 64-bit addresses, programs
21826 must be linked in the low 44 bits of memory, the text and data segments must
21827 be less than 2GB in size and the data segment must be located within 2GB of
21831 The Medium/Anywhere code model: 64-bit addresses, programs
21832 may be linked anywhere in memory, the text and data segments must be less
21833 than 2GB in size and the data segment must be located within 2GB of the
21837 The Medium/Anywhere code model for embedded systems:
21838 64-bit addresses, the text and data segments must be less than 2GB in
21839 size, both starting anywhere in memory (determined at link time). The
21840 global register %g4 points to the base of the data segment. Programs
21841 are statically linked and PIC is not supported.
21844 @item -mmemory-model=@var{mem-model}
21845 @opindex mmemory-model
21846 Set the memory model in force on the processor to one of
21850 The default memory model for the processor and operating system.
21853 Relaxed Memory Order
21856 Partial Store Order
21862 Sequential Consistency
21865 These memory models are formally defined in Appendix D of the Sparc V9
21866 architecture manual, as set in the processor's @code{PSTATE.MM} field.
21869 @itemx -mno-stack-bias
21870 @opindex mstack-bias
21871 @opindex mno-stack-bias
21872 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
21873 frame pointer if present, are offset by @minus{}2047 which must be added back
21874 when making stack frame references. This is the default in 64-bit mode.
21875 Otherwise, assume no such offset is present.
21879 @subsection SPU Options
21880 @cindex SPU options
21882 These @samp{-m} options are supported on the SPU:
21886 @itemx -merror-reloc
21887 @opindex mwarn-reloc
21888 @opindex merror-reloc
21890 The loader for SPU does not handle dynamic relocations. By default, GCC
21891 gives an error when it generates code that requires a dynamic
21892 relocation. @option{-mno-error-reloc} disables the error,
21893 @option{-mwarn-reloc} generates a warning instead.
21896 @itemx -munsafe-dma
21898 @opindex munsafe-dma
21900 Instructions that initiate or test completion of DMA must not be
21901 reordered with respect to loads and stores of the memory that is being
21903 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
21904 memory accesses, but that can lead to inefficient code in places where the
21905 memory is known to not change. Rather than mark the memory as volatile,
21906 you can use @option{-msafe-dma} to tell the compiler to treat
21907 the DMA instructions as potentially affecting all memory.
21909 @item -mbranch-hints
21910 @opindex mbranch-hints
21912 By default, GCC generates a branch hint instruction to avoid
21913 pipeline stalls for always-taken or probably-taken branches. A hint
21914 is not generated closer than 8 instructions away from its branch.
21915 There is little reason to disable them, except for debugging purposes,
21916 or to make an object a little bit smaller.
21920 @opindex msmall-mem
21921 @opindex mlarge-mem
21923 By default, GCC generates code assuming that addresses are never larger
21924 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
21925 a full 32-bit address.
21930 By default, GCC links against startup code that assumes the SPU-style
21931 main function interface (which has an unconventional parameter list).
21932 With @option{-mstdmain}, GCC links your program against startup
21933 code that assumes a C99-style interface to @code{main}, including a
21934 local copy of @code{argv} strings.
21936 @item -mfixed-range=@var{register-range}
21937 @opindex mfixed-range
21938 Generate code treating the given register range as fixed registers.
21939 A fixed register is one that the register allocator cannot use. This is
21940 useful when compiling kernel code. A register range is specified as
21941 two registers separated by a dash. Multiple register ranges can be
21942 specified separated by a comma.
21948 Compile code assuming that pointers to the PPU address space accessed
21949 via the @code{__ea} named address space qualifier are either 32 or 64
21950 bits wide. The default is 32 bits. As this is an ABI-changing option,
21951 all object code in an executable must be compiled with the same setting.
21953 @item -maddress-space-conversion
21954 @itemx -mno-address-space-conversion
21955 @opindex maddress-space-conversion
21956 @opindex mno-address-space-conversion
21957 Allow/disallow treating the @code{__ea} address space as superset
21958 of the generic address space. This enables explicit type casts
21959 between @code{__ea} and generic pointer as well as implicit
21960 conversions of generic pointers to @code{__ea} pointers. The
21961 default is to allow address space pointer conversions.
21963 @item -mcache-size=@var{cache-size}
21964 @opindex mcache-size
21965 This option controls the version of libgcc that the compiler links to an
21966 executable and selects a software-managed cache for accessing variables
21967 in the @code{__ea} address space with a particular cache size. Possible
21968 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
21969 and @samp{128}. The default cache size is 64KB.
21971 @item -matomic-updates
21972 @itemx -mno-atomic-updates
21973 @opindex matomic-updates
21974 @opindex mno-atomic-updates
21975 This option controls the version of libgcc that the compiler links to an
21976 executable and selects whether atomic updates to the software-managed
21977 cache of PPU-side variables are used. If you use atomic updates, changes
21978 to a PPU variable from SPU code using the @code{__ea} named address space
21979 qualifier do not interfere with changes to other PPU variables residing
21980 in the same cache line from PPU code. If you do not use atomic updates,
21981 such interference may occur; however, writing back cache lines is
21982 more efficient. The default behavior is to use atomic updates.
21985 @itemx -mdual-nops=@var{n}
21986 @opindex mdual-nops
21987 By default, GCC inserts nops to increase dual issue when it expects
21988 it to increase performance. @var{n} can be a value from 0 to 10. A
21989 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
21990 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
21992 @item -mhint-max-nops=@var{n}
21993 @opindex mhint-max-nops
21994 Maximum number of nops to insert for a branch hint. A branch hint must
21995 be at least 8 instructions away from the branch it is affecting. GCC
21996 inserts up to @var{n} nops to enforce this, otherwise it does not
21997 generate the branch hint.
21999 @item -mhint-max-distance=@var{n}
22000 @opindex mhint-max-distance
22001 The encoding of the branch hint instruction limits the hint to be within
22002 256 instructions of the branch it is affecting. By default, GCC makes
22003 sure it is within 125.
22006 @opindex msafe-hints
22007 Work around a hardware bug that causes the SPU to stall indefinitely.
22008 By default, GCC inserts the @code{hbrp} instruction to make sure
22009 this stall won't happen.
22013 @node System V Options
22014 @subsection Options for System V
22016 These additional options are available on System V Release 4 for
22017 compatibility with other compilers on those systems:
22022 Create a shared object.
22023 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22027 Identify the versions of each tool used by the compiler, in a
22028 @code{.ident} assembler directive in the output.
22032 Refrain from adding @code{.ident} directives to the output file (this is
22035 @item -YP,@var{dirs}
22037 Search the directories @var{dirs}, and no others, for libraries
22038 specified with @option{-l}.
22040 @item -Ym,@var{dir}
22042 Look in the directory @var{dir} to find the M4 preprocessor.
22043 The assembler uses this option.
22044 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22045 @c the generic assembler that comes with Solaris takes just -Ym.
22048 @node TILE-Gx Options
22049 @subsection TILE-Gx Options
22050 @cindex TILE-Gx options
22052 These @samp{-m} options are supported on the TILE-Gx:
22055 @item -mcmodel=small
22056 @opindex mcmodel=small
22057 Generate code for the small model. The distance for direct calls is
22058 limited to 500M in either direction. PC-relative addresses are 32
22059 bits. Absolute addresses support the full address range.
22061 @item -mcmodel=large
22062 @opindex mcmodel=large
22063 Generate code for the large model. There is no limitation on call
22064 distance, pc-relative addresses, or absolute addresses.
22066 @item -mcpu=@var{name}
22068 Selects the type of CPU to be targeted. Currently the only supported
22069 type is @samp{tilegx}.
22075 Generate code for a 32-bit or 64-bit environment. The 32-bit
22076 environment sets int, long, and pointer to 32 bits. The 64-bit
22077 environment sets int to 32 bits and long and pointer to 64 bits.
22080 @itemx -mlittle-endian
22081 @opindex mbig-endian
22082 @opindex mlittle-endian
22083 Generate code in big/little endian mode, respectively.
22086 @node TILEPro Options
22087 @subsection TILEPro Options
22088 @cindex TILEPro options
22090 These @samp{-m} options are supported on the TILEPro:
22093 @item -mcpu=@var{name}
22095 Selects the type of CPU to be targeted. Currently the only supported
22096 type is @samp{tilepro}.
22100 Generate code for a 32-bit environment, which sets int, long, and
22101 pointer to 32 bits. This is the only supported behavior so the flag
22102 is essentially ignored.
22106 @subsection V850 Options
22107 @cindex V850 Options
22109 These @samp{-m} options are defined for V850 implementations:
22113 @itemx -mno-long-calls
22114 @opindex mlong-calls
22115 @opindex mno-long-calls
22116 Treat all calls as being far away (near). If calls are assumed to be
22117 far away, the compiler always loads the function's address into a
22118 register, and calls indirect through the pointer.
22124 Do not optimize (do optimize) basic blocks that use the same index
22125 pointer 4 or more times to copy pointer into the @code{ep} register, and
22126 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22127 option is on by default if you optimize.
22129 @item -mno-prolog-function
22130 @itemx -mprolog-function
22131 @opindex mno-prolog-function
22132 @opindex mprolog-function
22133 Do not use (do use) external functions to save and restore registers
22134 at the prologue and epilogue of a function. The external functions
22135 are slower, but use less code space if more than one function saves
22136 the same number of registers. The @option{-mprolog-function} option
22137 is on by default if you optimize.
22141 Try to make the code as small as possible. At present, this just turns
22142 on the @option{-mep} and @option{-mprolog-function} options.
22144 @item -mtda=@var{n}
22146 Put static or global variables whose size is @var{n} bytes or less into
22147 the tiny data area that register @code{ep} points to. The tiny data
22148 area can hold up to 256 bytes in total (128 bytes for byte references).
22150 @item -msda=@var{n}
22152 Put static or global variables whose size is @var{n} bytes or less into
22153 the small data area that register @code{gp} points to. The small data
22154 area can hold up to 64 kilobytes.
22156 @item -mzda=@var{n}
22158 Put static or global variables whose size is @var{n} bytes or less into
22159 the first 32 kilobytes of memory.
22163 Specify that the target processor is the V850.
22167 Specify that the target processor is the V850E3V5. The preprocessor
22168 constant @code{__v850e3v5__} is defined if this option is used.
22172 Specify that the target processor is the V850E3V5. This is an alias for
22173 the @option{-mv850e3v5} option.
22177 Specify that the target processor is the V850E2V3. The preprocessor
22178 constant @code{__v850e2v3__} is defined if this option is used.
22182 Specify that the target processor is the V850E2. The preprocessor
22183 constant @code{__v850e2__} is defined if this option is used.
22187 Specify that the target processor is the V850E1. The preprocessor
22188 constants @code{__v850e1__} and @code{__v850e__} are defined if
22189 this option is used.
22193 Specify that the target processor is the V850ES. This is an alias for
22194 the @option{-mv850e1} option.
22198 Specify that the target processor is the V850E@. The preprocessor
22199 constant @code{__v850e__} is defined if this option is used.
22201 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22202 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22203 are defined then a default target processor is chosen and the
22204 relevant @samp{__v850*__} preprocessor constant is defined.
22206 The preprocessor constants @code{__v850} and @code{__v851__} are always
22207 defined, regardless of which processor variant is the target.
22209 @item -mdisable-callt
22210 @itemx -mno-disable-callt
22211 @opindex mdisable-callt
22212 @opindex mno-disable-callt
22213 This option suppresses generation of the @code{CALLT} instruction for the
22214 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22217 This option is enabled by default when the RH850 ABI is
22218 in use (see @option{-mrh850-abi}), and disabled by default when the
22219 GCC ABI is in use. If @code{CALLT} instructions are being generated
22220 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22226 Pass on (or do not pass on) the @option{-mrelax} command-line option
22230 @itemx -mno-long-jumps
22231 @opindex mlong-jumps
22232 @opindex mno-long-jumps
22233 Disable (or re-enable) the generation of PC-relative jump instructions.
22236 @itemx -mhard-float
22237 @opindex msoft-float
22238 @opindex mhard-float
22239 Disable (or re-enable) the generation of hardware floating point
22240 instructions. This option is only significant when the target
22241 architecture is @samp{V850E2V3} or higher. If hardware floating point
22242 instructions are being generated then the C preprocessor symbol
22243 @code{__FPU_OK__} is defined, otherwise the symbol
22244 @code{__NO_FPU__} is defined.
22248 Enables the use of the e3v5 LOOP instruction. The use of this
22249 instruction is not enabled by default when the e3v5 architecture is
22250 selected because its use is still experimental.
22254 @opindex mrh850-abi
22256 Enables support for the RH850 version of the V850 ABI. This is the
22257 default. With this version of the ABI the following rules apply:
22261 Integer sized structures and unions are returned via a memory pointer
22262 rather than a register.
22265 Large structures and unions (more than 8 bytes in size) are passed by
22269 Functions are aligned to 16-bit boundaries.
22272 The @option{-m8byte-align} command-line option is supported.
22275 The @option{-mdisable-callt} command-line option is enabled by
22276 default. The @option{-mno-disable-callt} command-line option is not
22280 When this version of the ABI is enabled the C preprocessor symbol
22281 @code{__V850_RH850_ABI__} is defined.
22285 Enables support for the old GCC version of the V850 ABI. With this
22286 version of the ABI the following rules apply:
22290 Integer sized structures and unions are returned in register @code{r10}.
22293 Large structures and unions (more than 8 bytes in size) are passed by
22297 Functions are aligned to 32-bit boundaries, unless optimizing for
22301 The @option{-m8byte-align} command-line option is not supported.
22304 The @option{-mdisable-callt} command-line option is supported but not
22305 enabled by default.
22308 When this version of the ABI is enabled the C preprocessor symbol
22309 @code{__V850_GCC_ABI__} is defined.
22311 @item -m8byte-align
22312 @itemx -mno-8byte-align
22313 @opindex m8byte-align
22314 @opindex mno-8byte-align
22315 Enables support for @code{double} and @code{long long} types to be
22316 aligned on 8-byte boundaries. The default is to restrict the
22317 alignment of all objects to at most 4-bytes. When
22318 @option{-m8byte-align} is in effect the C preprocessor symbol
22319 @code{__V850_8BYTE_ALIGN__} is defined.
22322 @opindex mbig-switch
22323 Generate code suitable for big switch tables. Use this option only if
22324 the assembler/linker complain about out of range branches within a switch
22329 This option causes r2 and r5 to be used in the code generated by
22330 the compiler. This setting is the default.
22332 @item -mno-app-regs
22333 @opindex mno-app-regs
22334 This option causes r2 and r5 to be treated as fixed registers.
22339 @subsection VAX Options
22340 @cindex VAX options
22342 These @samp{-m} options are defined for the VAX:
22347 Do not output certain jump instructions (@code{aobleq} and so on)
22348 that the Unix assembler for the VAX cannot handle across long
22353 Do output those jump instructions, on the assumption that the
22354 GNU assembler is being used.
22358 Output code for G-format floating-point numbers instead of D-format.
22361 @node Visium Options
22362 @subsection Visium Options
22363 @cindex Visium options
22369 A program which performs file I/O and is destined to run on an MCM target
22370 should be linked with this option. It causes the libraries libc.a and
22371 libdebug.a to be linked. The program should be run on the target under
22372 the control of the GDB remote debugging stub.
22376 A program which performs file I/O and is destined to run on the simulator
22377 should be linked with option. This causes libraries libc.a and libsim.a to
22381 @itemx -mhard-float
22383 @opindex mhard-float
22384 Generate code containing floating-point instructions. This is the
22388 @itemx -msoft-float
22390 @opindex msoft-float
22391 Generate code containing library calls for floating-point.
22393 @option{-msoft-float} changes the calling convention in the output file;
22394 therefore, it is only useful if you compile @emph{all} of a program with
22395 this option. In particular, you need to compile @file{libgcc.a}, the
22396 library that comes with GCC, with @option{-msoft-float} in order for
22399 @item -mcpu=@var{cpu_type}
22401 Set the instruction set, register set, and instruction scheduling parameters
22402 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22403 @samp{mcm}, @samp{gr5} and @samp{gr6}.
22405 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
22407 By default (unless configured otherwise), GCC generates code for the GR5
22408 variant of the Visium architecture.
22410 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
22411 architecture. The only difference from GR5 code is that the compiler will
22412 generate block move instructions.
22414 @item -mtune=@var{cpu_type}
22416 Set the instruction scheduling parameters for machine type @var{cpu_type},
22417 but do not set the instruction set or register set that the option
22418 @option{-mcpu=@var{cpu_type}} would.
22422 Generate code for the supervisor mode, where there are no restrictions on
22423 the access to general registers. This is the default.
22426 @opindex muser-mode
22427 Generate code for the user mode, where the access to some general registers
22428 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
22429 mode; on the GR6, only registers r29 to r31 are affected.
22433 @subsection VMS Options
22435 These @samp{-m} options are defined for the VMS implementations:
22438 @item -mvms-return-codes
22439 @opindex mvms-return-codes
22440 Return VMS condition codes from @code{main}. The default is to return POSIX-style
22441 condition (e.g.@ error) codes.
22443 @item -mdebug-main=@var{prefix}
22444 @opindex mdebug-main=@var{prefix}
22445 Flag the first routine whose name starts with @var{prefix} as the main
22446 routine for the debugger.
22450 Default to 64-bit memory allocation routines.
22452 @item -mpointer-size=@var{size}
22453 @opindex mpointer-size=@var{size}
22454 Set the default size of pointers. Possible options for @var{size} are
22455 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
22456 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
22457 The later option disables @code{pragma pointer_size}.
22460 @node VxWorks Options
22461 @subsection VxWorks Options
22462 @cindex VxWorks Options
22464 The options in this section are defined for all VxWorks targets.
22465 Options specific to the target hardware are listed with the other
22466 options for that target.
22471 GCC can generate code for both VxWorks kernels and real time processes
22472 (RTPs). This option switches from the former to the latter. It also
22473 defines the preprocessor macro @code{__RTP__}.
22476 @opindex non-static
22477 Link an RTP executable against shared libraries rather than static
22478 libraries. The options @option{-static} and @option{-shared} can
22479 also be used for RTPs (@pxref{Link Options}); @option{-static}
22486 These options are passed down to the linker. They are defined for
22487 compatibility with Diab.
22490 @opindex Xbind-lazy
22491 Enable lazy binding of function calls. This option is equivalent to
22492 @option{-Wl,-z,now} and is defined for compatibility with Diab.
22496 Disable lazy binding of function calls. This option is the default and
22497 is defined for compatibility with Diab.
22501 @subsection x86 Options
22502 @cindex x86 Options
22504 These @samp{-m} options are defined for the x86 family of computers.
22508 @item -march=@var{cpu-type}
22510 Generate instructions for the machine type @var{cpu-type}. In contrast to
22511 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
22512 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
22513 to generate code that may not run at all on processors other than the one
22514 indicated. Specifying @option{-march=@var{cpu-type}} implies
22515 @option{-mtune=@var{cpu-type}}.
22517 The choices for @var{cpu-type} are:
22521 This selects the CPU to generate code for at compilation time by determining
22522 the processor type of the compiling machine. Using @option{-march=native}
22523 enables all instruction subsets supported by the local machine (hence
22524 the result might not run on different machines). Using @option{-mtune=native}
22525 produces code optimized for the local machine under the constraints
22526 of the selected instruction set.
22529 Original Intel i386 CPU@.
22532 Intel i486 CPU@. (No scheduling is implemented for this chip.)
22536 Intel Pentium CPU with no MMX support.
22539 Intel Lakemont MCU, based on Intel Pentium CPU.
22542 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
22545 Intel Pentium Pro CPU@.
22548 When used with @option{-march}, the Pentium Pro
22549 instruction set is used, so the code runs on all i686 family chips.
22550 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
22553 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
22558 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
22562 Intel Pentium M; low-power version of Intel Pentium III CPU
22563 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
22567 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
22570 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
22574 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
22575 SSE2 and SSE3 instruction set support.
22578 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
22579 instruction set support.
22582 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22583 SSE4.1, SSE4.2 and POPCNT instruction set support.
22586 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22587 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
22590 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22591 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
22594 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
22595 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
22596 instruction set support.
22599 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22600 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22601 BMI, BMI2 and F16C instruction set support.
22604 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22605 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22606 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
22609 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22610 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22611 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
22612 XSAVES instruction set support.
22615 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
22616 instruction set support.
22619 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
22620 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
22623 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22624 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22625 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
22626 AVX512CD instruction set support.
22628 @item skylake-avx512
22629 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
22630 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
22631 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
22632 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
22635 AMD K6 CPU with MMX instruction set support.
22639 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
22642 @itemx athlon-tbird
22643 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
22649 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
22650 instruction set support.
22656 Processors based on the AMD K8 core with x86-64 instruction set support,
22657 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
22658 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
22659 instruction set extensions.)
22662 @itemx opteron-sse3
22663 @itemx athlon64-sse3
22664 Improved versions of AMD K8 cores with SSE3 instruction set support.
22668 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
22669 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
22670 instruction set extensions.)
22673 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
22674 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
22675 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
22677 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22678 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
22679 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
22682 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22683 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
22684 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
22685 64-bit instruction set extensions.
22687 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
22688 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
22689 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
22690 SSE4.2, ABM and 64-bit instruction set extensions.
22693 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
22694 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
22695 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
22696 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
22697 instruction set extensions.
22700 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
22701 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
22702 instruction set extensions.)
22705 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
22706 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
22707 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
22710 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
22714 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
22715 instruction set support.
22718 VIA C3 CPU with MMX and 3DNow!@: instruction set support. (No scheduling is
22719 implemented for this chip.)
22722 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
22724 implemented for this chip.)
22727 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
22730 @item -mtune=@var{cpu-type}
22732 Tune to @var{cpu-type} everything applicable about the generated code, except
22733 for the ABI and the set of available instructions.
22734 While picking a specific @var{cpu-type} schedules things appropriately
22735 for that particular chip, the compiler does not generate any code that
22736 cannot run on the default machine type unless you use a
22737 @option{-march=@var{cpu-type}} option.
22738 For example, if GCC is configured for i686-pc-linux-gnu
22739 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
22740 but still runs on i686 machines.
22742 The choices for @var{cpu-type} are the same as for @option{-march}.
22743 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
22747 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
22748 If you know the CPU on which your code will run, then you should use
22749 the corresponding @option{-mtune} or @option{-march} option instead of
22750 @option{-mtune=generic}. But, if you do not know exactly what CPU users
22751 of your application will have, then you should use this option.
22753 As new processors are deployed in the marketplace, the behavior of this
22754 option will change. Therefore, if you upgrade to a newer version of
22755 GCC, code generation controlled by this option will change to reflect
22757 that are most common at the time that version of GCC is released.
22759 There is no @option{-march=generic} option because @option{-march}
22760 indicates the instruction set the compiler can use, and there is no
22761 generic instruction set applicable to all processors. In contrast,
22762 @option{-mtune} indicates the processor (or, in this case, collection of
22763 processors) for which the code is optimized.
22766 Produce code optimized for the most current Intel processors, which are
22767 Haswell and Silvermont for this version of GCC. If you know the CPU
22768 on which your code will run, then you should use the corresponding
22769 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
22770 But, if you want your application performs better on both Haswell and
22771 Silvermont, then you should use this option.
22773 As new Intel processors are deployed in the marketplace, the behavior of
22774 this option will change. Therefore, if you upgrade to a newer version of
22775 GCC, code generation controlled by this option will change to reflect
22776 the most current Intel processors at the time that version of GCC is
22779 There is no @option{-march=intel} option because @option{-march} indicates
22780 the instruction set the compiler can use, and there is no common
22781 instruction set applicable to all processors. In contrast,
22782 @option{-mtune} indicates the processor (or, in this case, collection of
22783 processors) for which the code is optimized.
22786 @item -mcpu=@var{cpu-type}
22788 A deprecated synonym for @option{-mtune}.
22790 @item -mfpmath=@var{unit}
22792 Generate floating-point arithmetic for selected unit @var{unit}. The choices
22793 for @var{unit} are:
22797 Use the standard 387 floating-point coprocessor present on the majority of chips and
22798 emulated otherwise. Code compiled with this option runs almost everywhere.
22799 The temporary results are computed in 80-bit precision instead of the precision
22800 specified by the type, resulting in slightly different results compared to most
22801 of other chips. See @option{-ffloat-store} for more detailed description.
22803 This is the default choice for x86-32 targets.
22806 Use scalar floating-point instructions present in the SSE instruction set.
22807 This instruction set is supported by Pentium III and newer chips,
22808 and in the AMD line
22809 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
22810 instruction set supports only single-precision arithmetic, thus the double and
22811 extended-precision arithmetic are still done using 387. A later version, present
22812 only in Pentium 4 and AMD x86-64 chips, supports double-precision
22815 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
22816 or @option{-msse2} switches to enable SSE extensions and make this option
22817 effective. For the x86-64 compiler, these extensions are enabled by default.
22819 The resulting code should be considerably faster in the majority of cases and avoid
22820 the numerical instability problems of 387 code, but may break some existing
22821 code that expects temporaries to be 80 bits.
22823 This is the default choice for the x86-64 compiler.
22828 Attempt to utilize both instruction sets at once. This effectively doubles the
22829 amount of available registers, and on chips with separate execution units for
22830 387 and SSE the execution resources too. Use this option with care, as it is
22831 still experimental, because the GCC register allocator does not model separate
22832 functional units well, resulting in unstable performance.
22835 @item -masm=@var{dialect}
22836 @opindex masm=@var{dialect}
22837 Output assembly instructions using selected @var{dialect}. Also affects
22838 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
22839 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
22840 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
22841 not support @samp{intel}.
22844 @itemx -mno-ieee-fp
22846 @opindex mno-ieee-fp
22847 Control whether or not the compiler uses IEEE floating-point
22848 comparisons. These correctly handle the case where the result of a
22849 comparison is unordered.
22852 @opindex msoft-float
22853 Generate output containing library calls for floating point.
22855 @strong{Warning:} the requisite libraries are not part of GCC@.
22856 Normally the facilities of the machine's usual C compiler are used, but
22857 this can't be done directly in cross-compilation. You must make your
22858 own arrangements to provide suitable library functions for
22861 On machines where a function returns floating-point results in the 80387
22862 register stack, some floating-point opcodes may be emitted even if
22863 @option{-msoft-float} is used.
22865 @item -mno-fp-ret-in-387
22866 @opindex mno-fp-ret-in-387
22867 Do not use the FPU registers for return values of functions.
22869 The usual calling convention has functions return values of types
22870 @code{float} and @code{double} in an FPU register, even if there
22871 is no FPU@. The idea is that the operating system should emulate
22874 The option @option{-mno-fp-ret-in-387} causes such values to be returned
22875 in ordinary CPU registers instead.
22877 @item -mno-fancy-math-387
22878 @opindex mno-fancy-math-387
22879 Some 387 emulators do not support the @code{sin}, @code{cos} and
22880 @code{sqrt} instructions for the 387. Specify this option to avoid
22881 generating those instructions. This option is the default on
22882 OpenBSD and NetBSD@. This option is overridden when @option{-march}
22883 indicates that the target CPU always has an FPU and so the
22884 instruction does not need emulation. These
22885 instructions are not generated unless you also use the
22886 @option{-funsafe-math-optimizations} switch.
22888 @item -malign-double
22889 @itemx -mno-align-double
22890 @opindex malign-double
22891 @opindex mno-align-double
22892 Control whether GCC aligns @code{double}, @code{long double}, and
22893 @code{long long} variables on a two-word boundary or a one-word
22894 boundary. Aligning @code{double} variables on a two-word boundary
22895 produces code that runs somewhat faster on a Pentium at the
22896 expense of more memory.
22898 On x86-64, @option{-malign-double} is enabled by default.
22900 @strong{Warning:} if you use the @option{-malign-double} switch,
22901 structures containing the above types are aligned differently than
22902 the published application binary interface specifications for the x86-32
22903 and are not binary compatible with structures in code compiled
22904 without that switch.
22906 @item -m96bit-long-double
22907 @itemx -m128bit-long-double
22908 @opindex m96bit-long-double
22909 @opindex m128bit-long-double
22910 These switches control the size of @code{long double} type. The x86-32
22911 application binary interface specifies the size to be 96 bits,
22912 so @option{-m96bit-long-double} is the default in 32-bit mode.
22914 Modern architectures (Pentium and newer) prefer @code{long double}
22915 to be aligned to an 8- or 16-byte boundary. In arrays or structures
22916 conforming to the ABI, this is not possible. So specifying
22917 @option{-m128bit-long-double} aligns @code{long double}
22918 to a 16-byte boundary by padding the @code{long double} with an additional
22921 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
22922 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
22924 Notice that neither of these options enable any extra precision over the x87
22925 standard of 80 bits for a @code{long double}.
22927 @strong{Warning:} if you override the default value for your target ABI, this
22928 changes the size of
22929 structures and arrays containing @code{long double} variables,
22930 as well as modifying the function calling convention for functions taking
22931 @code{long double}. Hence they are not binary-compatible
22932 with code compiled without that switch.
22934 @item -mlong-double-64
22935 @itemx -mlong-double-80
22936 @itemx -mlong-double-128
22937 @opindex mlong-double-64
22938 @opindex mlong-double-80
22939 @opindex mlong-double-128
22940 These switches control the size of @code{long double} type. A size
22941 of 64 bits makes the @code{long double} type equivalent to the @code{double}
22942 type. This is the default for 32-bit Bionic C library. A size
22943 of 128 bits makes the @code{long double} type equivalent to the
22944 @code{__float128} type. This is the default for 64-bit Bionic C library.
22946 @strong{Warning:} if you override the default value for your target ABI, this
22947 changes the size of
22948 structures and arrays containing @code{long double} variables,
22949 as well as modifying the function calling convention for functions taking
22950 @code{long double}. Hence they are not binary-compatible
22951 with code compiled without that switch.
22953 @item -malign-data=@var{type}
22954 @opindex malign-data
22955 Control how GCC aligns variables. Supported values for @var{type} are
22956 @samp{compat} uses increased alignment value compatible uses GCC 4.8
22957 and earlier, @samp{abi} uses alignment value as specified by the
22958 psABI, and @samp{cacheline} uses increased alignment value to match
22959 the cache line size. @samp{compat} is the default.
22961 @item -mlarge-data-threshold=@var{threshold}
22962 @opindex mlarge-data-threshold
22963 When @option{-mcmodel=medium} is specified, data objects larger than
22964 @var{threshold} are placed in the large data section. This value must be the
22965 same across all objects linked into the binary, and defaults to 65535.
22969 Use a different function-calling convention, in which functions that
22970 take a fixed number of arguments return with the @code{ret @var{num}}
22971 instruction, which pops their arguments while returning. This saves one
22972 instruction in the caller since there is no need to pop the arguments
22975 You can specify that an individual function is called with this calling
22976 sequence with the function attribute @code{stdcall}. You can also
22977 override the @option{-mrtd} option by using the function attribute
22978 @code{cdecl}. @xref{Function Attributes}.
22980 @strong{Warning:} this calling convention is incompatible with the one
22981 normally used on Unix, so you cannot use it if you need to call
22982 libraries compiled with the Unix compiler.
22984 Also, you must provide function prototypes for all functions that
22985 take variable numbers of arguments (including @code{printf});
22986 otherwise incorrect code is generated for calls to those
22989 In addition, seriously incorrect code results if you call a
22990 function with too many arguments. (Normally, extra arguments are
22991 harmlessly ignored.)
22993 @item -mregparm=@var{num}
22995 Control how many registers are used to pass integer arguments. By
22996 default, no registers are used to pass arguments, and at most 3
22997 registers can be used. You can control this behavior for a specific
22998 function by using the function attribute @code{regparm}.
22999 @xref{Function Attributes}.
23001 @strong{Warning:} if you use this switch, and
23002 @var{num} is nonzero, then you must build all modules with the same
23003 value, including any libraries. This includes the system libraries and
23007 @opindex msseregparm
23008 Use SSE register passing conventions for float and double arguments
23009 and return values. You can control this behavior for a specific
23010 function by using the function attribute @code{sseregparm}.
23011 @xref{Function Attributes}.
23013 @strong{Warning:} if you use this switch then you must build all
23014 modules with the same value, including any libraries. This includes
23015 the system libraries and startup modules.
23017 @item -mvect8-ret-in-mem
23018 @opindex mvect8-ret-in-mem
23019 Return 8-byte vectors in memory instead of MMX registers. This is the
23020 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
23021 Studio compilers until version 12. Later compiler versions (starting
23022 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
23023 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
23024 you need to remain compatible with existing code produced by those
23025 previous compiler versions or older versions of GCC@.
23034 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
23035 is specified, the significands of results of floating-point operations are
23036 rounded to 24 bits (single precision); @option{-mpc64} rounds the
23037 significands of results of floating-point operations to 53 bits (double
23038 precision) and @option{-mpc80} rounds the significands of results of
23039 floating-point operations to 64 bits (extended double precision), which is
23040 the default. When this option is used, floating-point operations in higher
23041 precisions are not available to the programmer without setting the FPU
23042 control word explicitly.
23044 Setting the rounding of floating-point operations to less than the default
23045 80 bits can speed some programs by 2% or more. Note that some mathematical
23046 libraries assume that extended-precision (80-bit) floating-point operations
23047 are enabled by default; routines in such libraries could suffer significant
23048 loss of accuracy, typically through so-called ``catastrophic cancellation'',
23049 when this option is used to set the precision to less than extended precision.
23051 @item -mstackrealign
23052 @opindex mstackrealign
23053 Realign the stack at entry. On the x86, the @option{-mstackrealign}
23054 option generates an alternate prologue and epilogue that realigns the
23055 run-time stack if necessary. This supports mixing legacy codes that keep
23056 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
23057 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
23058 applicable to individual functions.
23060 @item -mpreferred-stack-boundary=@var{num}
23061 @opindex mpreferred-stack-boundary
23062 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23063 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23064 the default is 4 (16 bytes or 128 bits).
23066 @strong{Warning:} When generating code for the x86-64 architecture with
23067 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
23068 used to keep the stack boundary aligned to 8 byte boundary. Since
23069 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
23070 intended to be used in controlled environment where stack space is
23071 important limitation. This option leads to wrong code when functions
23072 compiled with 16 byte stack alignment (such as functions from a standard
23073 library) are called with misaligned stack. In this case, SSE
23074 instructions may lead to misaligned memory access traps. In addition,
23075 variable arguments are handled incorrectly for 16 byte aligned
23076 objects (including x87 long double and __int128), leading to wrong
23077 results. You must build all modules with
23078 @option{-mpreferred-stack-boundary=3}, including any libraries. This
23079 includes the system libraries and startup modules.
23081 @item -mincoming-stack-boundary=@var{num}
23082 @opindex mincoming-stack-boundary
23083 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
23084 boundary. If @option{-mincoming-stack-boundary} is not specified,
23085 the one specified by @option{-mpreferred-stack-boundary} is used.
23087 On Pentium and Pentium Pro, @code{double} and @code{long double} values
23088 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
23089 suffer significant run time performance penalties. On Pentium III, the
23090 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
23091 properly if it is not 16-byte aligned.
23093 To ensure proper alignment of this values on the stack, the stack boundary
23094 must be as aligned as that required by any value stored on the stack.
23095 Further, every function must be generated such that it keeps the stack
23096 aligned. Thus calling a function compiled with a higher preferred
23097 stack boundary from a function compiled with a lower preferred stack
23098 boundary most likely misaligns the stack. It is recommended that
23099 libraries that use callbacks always use the default setting.
23101 This extra alignment does consume extra stack space, and generally
23102 increases code size. Code that is sensitive to stack space usage, such
23103 as embedded systems and operating system kernels, may want to reduce the
23104 preferred alignment to @option{-mpreferred-stack-boundary=2}.
23161 @itemx -mavx512ifma
23162 @opindex mavx512ifma
23164 @itemx -mavx512vbmi
23165 @opindex mavx512vbmi
23177 @opindex mclfushopt
23197 @itemx -mprefetchwt1
23198 @opindex mprefetchwt1
23254 These switches enable the use of instructions in the MMX, SSE,
23255 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
23256 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
23257 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
23258 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
23259 extended instruction sets. Each has a corresponding @option{-mno-} option
23260 to disable use of these instructions.
23262 These extensions are also available as built-in functions: see
23263 @ref{x86 Built-in Functions}, for details of the functions enabled and
23264 disabled by these switches.
23266 To generate SSE/SSE2 instructions automatically from floating-point
23267 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
23269 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
23270 generates new AVX instructions or AVX equivalence for all SSEx instructions
23273 These options enable GCC to use these extended instructions in
23274 generated code, even without @option{-mfpmath=sse}. Applications that
23275 perform run-time CPU detection must compile separate files for each
23276 supported architecture, using the appropriate flags. In particular,
23277 the file containing the CPU detection code should be compiled without
23280 @item -mdump-tune-features
23281 @opindex mdump-tune-features
23282 This option instructs GCC to dump the names of the x86 performance
23283 tuning features and default settings. The names can be used in
23284 @option{-mtune-ctrl=@var{feature-list}}.
23286 @item -mtune-ctrl=@var{feature-list}
23287 @opindex mtune-ctrl=@var{feature-list}
23288 This option is used to do fine grain control of x86 code generation features.
23289 @var{feature-list} is a comma separated list of @var{feature} names. See also
23290 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
23291 on if it is not preceded with @samp{^}, otherwise, it is turned off.
23292 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
23293 developers. Using it may lead to code paths not covered by testing and can
23294 potentially result in compiler ICEs or runtime errors.
23297 @opindex mno-default
23298 This option instructs GCC to turn off all tunable features. See also
23299 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
23303 This option instructs GCC to emit a @code{cld} instruction in the prologue
23304 of functions that use string instructions. String instructions depend on
23305 the DF flag to select between autoincrement or autodecrement mode. While the
23306 ABI specifies the DF flag to be cleared on function entry, some operating
23307 systems violate this specification by not clearing the DF flag in their
23308 exception dispatchers. The exception handler can be invoked with the DF flag
23309 set, which leads to wrong direction mode when string instructions are used.
23310 This option can be enabled by default on 32-bit x86 targets by configuring
23311 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
23312 instructions can be suppressed with the @option{-mno-cld} compiler option
23316 @opindex mvzeroupper
23317 This option instructs GCC to emit a @code{vzeroupper} instruction
23318 before a transfer of control flow out of the function to minimize
23319 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
23322 @item -mprefer-avx128
23323 @opindex mprefer-avx128
23324 This option instructs GCC to use 128-bit AVX instructions instead of
23325 256-bit AVX instructions in the auto-vectorizer.
23329 This option enables GCC to generate @code{CMPXCHG16B} instructions.
23330 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
23331 (or oword) data types.
23332 This is useful for high-resolution counters that can be updated
23333 by multiple processors (or cores). This instruction is generated as part of
23334 atomic built-in functions: see @ref{__sync Builtins} or
23335 @ref{__atomic Builtins} for details.
23339 This option enables generation of @code{SAHF} instructions in 64-bit code.
23340 Early Intel Pentium 4 CPUs with Intel 64 support,
23341 prior to the introduction of Pentium 4 G1 step in December 2005,
23342 lacked the @code{LAHF} and @code{SAHF} instructions
23343 which are supported by AMD64.
23344 These are load and store instructions, respectively, for certain status flags.
23345 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
23346 @code{drem}, and @code{remainder} built-in functions;
23347 see @ref{Other Builtins} for details.
23351 This option enables use of the @code{movbe} instruction to implement
23352 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
23356 This option enables built-in functions @code{__builtin_ia32_crc32qi},
23357 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
23358 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
23362 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
23363 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
23364 with an additional Newton-Raphson step
23365 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
23366 (and their vectorized
23367 variants) for single-precision floating-point arguments. These instructions
23368 are generated only when @option{-funsafe-math-optimizations} is enabled
23369 together with @option{-finite-math-only} and @option{-fno-trapping-math}.
23370 Note that while the throughput of the sequence is higher than the throughput
23371 of the non-reciprocal instruction, the precision of the sequence can be
23372 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
23374 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
23375 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
23376 combination), and doesn't need @option{-mrecip}.
23378 Also note that GCC emits the above sequence with additional Newton-Raphson step
23379 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
23380 already with @option{-ffast-math} (or the above option combination), and
23381 doesn't need @option{-mrecip}.
23383 @item -mrecip=@var{opt}
23384 @opindex mrecip=opt
23385 This option controls which reciprocal estimate instructions
23386 may be used. @var{opt} is a comma-separated list of options, which may
23387 be preceded by a @samp{!} to invert the option:
23391 Enable all estimate instructions.
23394 Enable the default instructions, equivalent to @option{-mrecip}.
23397 Disable all estimate instructions, equivalent to @option{-mno-recip}.
23400 Enable the approximation for scalar division.
23403 Enable the approximation for vectorized division.
23406 Enable the approximation for scalar square root.
23409 Enable the approximation for vectorized square root.
23412 So, for example, @option{-mrecip=all,!sqrt} enables
23413 all of the reciprocal approximations, except for square root.
23415 @item -mveclibabi=@var{type}
23416 @opindex mveclibabi
23417 Specifies the ABI type to use for vectorizing intrinsics using an
23418 external library. Supported values for @var{type} are @samp{svml}
23419 for the Intel short
23420 vector math library and @samp{acml} for the AMD math core library.
23421 To use this option, both @option{-ftree-vectorize} and
23422 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
23423 ABI-compatible library must be specified at link time.
23425 GCC currently emits calls to @code{vmldExp2},
23426 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
23427 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
23428 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
23429 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
23430 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
23431 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
23432 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
23433 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
23434 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
23435 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
23436 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
23437 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
23438 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
23439 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
23440 when @option{-mveclibabi=acml} is used.
23442 @item -mabi=@var{name}
23444 Generate code for the specified calling convention. Permissible values
23445 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
23446 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
23447 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
23448 You can control this behavior for specific functions by
23449 using the function attributes @code{ms_abi} and @code{sysv_abi}.
23450 @xref{Function Attributes}.
23452 @item -mtls-dialect=@var{type}
23453 @opindex mtls-dialect
23454 Generate code to access thread-local storage using the @samp{gnu} or
23455 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
23456 @samp{gnu2} is more efficient, but it may add compile- and run-time
23457 requirements that cannot be satisfied on all systems.
23460 @itemx -mno-push-args
23461 @opindex mpush-args
23462 @opindex mno-push-args
23463 Use PUSH operations to store outgoing parameters. This method is shorter
23464 and usually equally fast as method using SUB/MOV operations and is enabled
23465 by default. In some cases disabling it may improve performance because of
23466 improved scheduling and reduced dependencies.
23468 @item -maccumulate-outgoing-args
23469 @opindex maccumulate-outgoing-args
23470 If enabled, the maximum amount of space required for outgoing arguments is
23471 computed in the function prologue. This is faster on most modern CPUs
23472 because of reduced dependencies, improved scheduling and reduced stack usage
23473 when the preferred stack boundary is not equal to 2. The drawback is a notable
23474 increase in code size. This switch implies @option{-mno-push-args}.
23478 Support thread-safe exception handling on MinGW. Programs that rely
23479 on thread-safe exception handling must compile and link all code with the
23480 @option{-mthreads} option. When compiling, @option{-mthreads} defines
23481 @option{-D_MT}; when linking, it links in a special thread helper library
23482 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
23484 @item -mms-bitfields
23485 @itemx -mno-ms-bitfields
23486 @opindex mms-bitfields
23487 @opindex mno-ms-bitfields
23489 Enable/disable bit-field layout compatible with the native Microsoft
23492 If @code{packed} is used on a structure, or if bit-fields are used,
23493 it may be that the Microsoft ABI lays out the structure differently
23494 than the way GCC normally does. Particularly when moving packed
23495 data between functions compiled with GCC and the native Microsoft compiler
23496 (either via function call or as data in a file), it may be necessary to access
23499 This option is enabled by default for Microsoft Windows
23500 targets. This behavior can also be controlled locally by use of variable
23501 or type attributes. For more information, see @ref{x86 Variable Attributes}
23502 and @ref{x86 Type Attributes}.
23504 The Microsoft structure layout algorithm is fairly simple with the exception
23505 of the bit-field packing.
23506 The padding and alignment of members of structures and whether a bit-field
23507 can straddle a storage-unit boundary are determine by these rules:
23510 @item Structure members are stored sequentially in the order in which they are
23511 declared: the first member has the lowest memory address and the last member
23514 @item Every data object has an alignment requirement. The alignment requirement
23515 for all data except structures, unions, and arrays is either the size of the
23516 object or the current packing size (specified with either the
23517 @code{aligned} attribute or the @code{pack} pragma),
23518 whichever is less. For structures, unions, and arrays,
23519 the alignment requirement is the largest alignment requirement of its members.
23520 Every object is allocated an offset so that:
23523 offset % alignment_requirement == 0
23526 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
23527 unit if the integral types are the same size and if the next bit-field fits
23528 into the current allocation unit without crossing the boundary imposed by the
23529 common alignment requirements of the bit-fields.
23532 MSVC interprets zero-length bit-fields in the following ways:
23535 @item If a zero-length bit-field is inserted between two bit-fields that
23536 are normally coalesced, the bit-fields are not coalesced.
23543 unsigned long bf_1 : 12;
23545 unsigned long bf_2 : 12;
23550 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
23551 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
23553 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
23554 alignment of the zero-length bit-field is greater than the member that follows it,
23555 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
23576 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
23577 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
23578 bit-field does not affect the alignment of @code{bar} or, as a result, the size
23581 Taking this into account, it is important to note the following:
23584 @item If a zero-length bit-field follows a normal bit-field, the type of the
23585 zero-length bit-field may affect the alignment of the structure as whole. For
23586 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
23587 normal bit-field, and is of type short.
23589 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
23590 still affect the alignment of the structure:
23601 Here, @code{t4} takes up 4 bytes.
23604 @item Zero-length bit-fields following non-bit-field members are ignored:
23616 Here, @code{t5} takes up 2 bytes.
23620 @item -mno-align-stringops
23621 @opindex mno-align-stringops
23622 Do not align the destination of inlined string operations. This switch reduces
23623 code size and improves performance in case the destination is already aligned,
23624 but GCC doesn't know about it.
23626 @item -minline-all-stringops
23627 @opindex minline-all-stringops
23628 By default GCC inlines string operations only when the destination is
23629 known to be aligned to least a 4-byte boundary.
23630 This enables more inlining and increases code
23631 size, but may improve performance of code that depends on fast
23632 @code{memcpy}, @code{strlen},
23633 and @code{memset} for short lengths.
23635 @item -minline-stringops-dynamically
23636 @opindex minline-stringops-dynamically
23637 For string operations of unknown size, use run-time checks with
23638 inline code for small blocks and a library call for large blocks.
23640 @item -mstringop-strategy=@var{alg}
23641 @opindex mstringop-strategy=@var{alg}
23642 Override the internal decision heuristic for the particular algorithm to use
23643 for inlining string operations. The allowed values for @var{alg} are:
23649 Expand using i386 @code{rep} prefix of the specified size.
23653 @itemx unrolled_loop
23654 Expand into an inline loop.
23657 Always use a library call.
23660 @item -mmemcpy-strategy=@var{strategy}
23661 @opindex mmemcpy-strategy=@var{strategy}
23662 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
23663 should be inlined and what inline algorithm to use when the expected size
23664 of the copy operation is known. @var{strategy}
23665 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
23666 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
23667 the max byte size with which inline algorithm @var{alg} is allowed. For the last
23668 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
23669 in the list must be specified in increasing order. The minimal byte size for
23670 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
23673 @item -mmemset-strategy=@var{strategy}
23674 @opindex mmemset-strategy=@var{strategy}
23675 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
23676 @code{__builtin_memset} expansion.
23678 @item -momit-leaf-frame-pointer
23679 @opindex momit-leaf-frame-pointer
23680 Don't keep the frame pointer in a register for leaf functions. This
23681 avoids the instructions to save, set up, and restore frame pointers and
23682 makes an extra register available in leaf functions. The option
23683 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
23684 which might make debugging harder.
23686 @item -mtls-direct-seg-refs
23687 @itemx -mno-tls-direct-seg-refs
23688 @opindex mtls-direct-seg-refs
23689 Controls whether TLS variables may be accessed with offsets from the
23690 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
23691 or whether the thread base pointer must be added. Whether or not this
23692 is valid depends on the operating system, and whether it maps the
23693 segment to cover the entire TLS area.
23695 For systems that use the GNU C Library, the default is on.
23698 @itemx -mno-sse2avx
23700 Specify that the assembler should encode SSE instructions with VEX
23701 prefix. The option @option{-mavx} turns this on by default.
23706 If profiling is active (@option{-pg}), put the profiling
23707 counter call before the prologue.
23708 Note: On x86 architectures the attribute @code{ms_hook_prologue}
23709 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
23711 @item -mrecord-mcount
23712 @itemx -mno-record-mcount
23713 @opindex mrecord-mcount
23714 If profiling is active (@option{-pg}), generate a __mcount_loc section
23715 that contains pointers to each profiling call. This is useful for
23716 automatically patching and out calls.
23719 @itemx -mno-nop-mcount
23720 @opindex mnop-mcount
23721 If profiling is active (@option{-pg}), generate the calls to
23722 the profiling functions as nops. This is useful when they
23723 should be patched in later dynamically. This is likely only
23724 useful together with @option{-mrecord-mcount}.
23726 @item -mskip-rax-setup
23727 @itemx -mno-skip-rax-setup
23728 @opindex mskip-rax-setup
23729 When generating code for the x86-64 architecture with SSE extensions
23730 disabled, @option{-skip-rax-setup} can be used to skip setting up RAX
23731 register when there are no variable arguments passed in vector registers.
23733 @strong{Warning:} Since RAX register is used to avoid unnecessarily
23734 saving vector registers on stack when passing variable arguments, the
23735 impacts of this option are callees may waste some stack space,
23736 misbehave or jump to a random location. GCC 4.4 or newer don't have
23737 those issues, regardless the RAX register value.
23740 @itemx -mno-8bit-idiv
23741 @opindex m8bit-idiv
23742 On some processors, like Intel Atom, 8-bit unsigned integer divide is
23743 much faster than 32-bit/64-bit integer divide. This option generates a
23744 run-time check. If both dividend and divisor are within range of 0
23745 to 255, 8-bit unsigned integer divide is used instead of
23746 32-bit/64-bit integer divide.
23748 @item -mavx256-split-unaligned-load
23749 @itemx -mavx256-split-unaligned-store
23750 @opindex mavx256-split-unaligned-load
23751 @opindex mavx256-split-unaligned-store
23752 Split 32-byte AVX unaligned load and store.
23754 @item -mstack-protector-guard=@var{guard}
23755 @opindex mstack-protector-guard=@var{guard}
23756 Generate stack protection code using canary at @var{guard}. Supported
23757 locations are @samp{global} for global canary or @samp{tls} for per-thread
23758 canary in the TLS block (the default). This option has effect only when
23759 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
23761 @item -mmitigate-rop
23762 @opindex mmitigate-rop
23763 Try to avoid generating code sequences that contain unintended return
23764 opcodes, to mitigate against certain forms of attack. At the moment,
23765 this option is limited in what it can do and should not be relied
23766 on to provide serious protection.
23770 These @samp{-m} switches are supported in addition to the above
23771 on x86-64 processors in 64-bit environments.
23784 Generate code for a 16-bit, 32-bit or 64-bit environment.
23785 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
23787 generates code that runs on any i386 system.
23789 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
23790 types to 64 bits, and generates code for the x86-64 architecture.
23791 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
23792 and @option{-mdynamic-no-pic} options.
23794 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
23796 generates code for the x86-64 architecture.
23798 The @option{-m16} option is the same as @option{-m32}, except for that
23799 it outputs the @code{.code16gcc} assembly directive at the beginning of
23800 the assembly output so that the binary can run in 16-bit mode.
23802 The @option{-miamcu} option generates code which conforms to Intel MCU
23803 psABI. It requires the @option{-m32} option to be turned on.
23805 @item -mno-red-zone
23806 @opindex mno-red-zone
23807 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
23808 by the x86-64 ABI; it is a 128-byte area beyond the location of the
23809 stack pointer that is not modified by signal or interrupt handlers
23810 and therefore can be used for temporary data without adjusting the stack
23811 pointer. The flag @option{-mno-red-zone} disables this red zone.
23813 @item -mcmodel=small
23814 @opindex mcmodel=small
23815 Generate code for the small code model: the program and its symbols must
23816 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
23817 Programs can be statically or dynamically linked. This is the default
23820 @item -mcmodel=kernel
23821 @opindex mcmodel=kernel
23822 Generate code for the kernel code model. The kernel runs in the
23823 negative 2 GB of the address space.
23824 This model has to be used for Linux kernel code.
23826 @item -mcmodel=medium
23827 @opindex mcmodel=medium
23828 Generate code for the medium model: the program is linked in the lower 2
23829 GB of the address space. Small symbols are also placed there. Symbols
23830 with sizes larger than @option{-mlarge-data-threshold} are put into
23831 large data or BSS sections and can be located above 2GB. Programs can
23832 be statically or dynamically linked.
23834 @item -mcmodel=large
23835 @opindex mcmodel=large
23836 Generate code for the large model. This model makes no assumptions
23837 about addresses and sizes of sections.
23839 @item -maddress-mode=long
23840 @opindex maddress-mode=long
23841 Generate code for long address mode. This is only supported for 64-bit
23842 and x32 environments. It is the default address mode for 64-bit
23845 @item -maddress-mode=short
23846 @opindex maddress-mode=short
23847 Generate code for short address mode. This is only supported for 32-bit
23848 and x32 environments. It is the default address mode for 32-bit and
23852 @node x86 Windows Options
23853 @subsection x86 Windows Options
23854 @cindex x86 Windows Options
23855 @cindex Windows Options for x86
23857 These additional options are available for Microsoft Windows targets:
23863 specifies that a console application is to be generated, by
23864 instructing the linker to set the PE header subsystem type
23865 required for console applications.
23866 This option is available for Cygwin and MinGW targets and is
23867 enabled by default on those targets.
23871 This option is available for Cygwin and MinGW targets. It
23872 specifies that a DLL---a dynamic link library---is to be
23873 generated, enabling the selection of the required runtime
23874 startup object and entry point.
23876 @item -mnop-fun-dllimport
23877 @opindex mnop-fun-dllimport
23878 This option is available for Cygwin and MinGW targets. It
23879 specifies that the @code{dllimport} attribute should be ignored.
23883 This option is available for MinGW targets. It specifies
23884 that MinGW-specific thread support is to be used.
23888 This option is available for MinGW-w64 targets. It causes
23889 the @code{UNICODE} preprocessor macro to be predefined, and
23890 chooses Unicode-capable runtime startup code.
23894 This option is available for Cygwin and MinGW targets. It
23895 specifies that the typical Microsoft Windows predefined macros are to
23896 be set in the pre-processor, but does not influence the choice
23897 of runtime library/startup code.
23901 This option is available for Cygwin and MinGW targets. It
23902 specifies that a GUI application is to be generated by
23903 instructing the linker to set the PE header subsystem type
23906 @item -fno-set-stack-executable
23907 @opindex fno-set-stack-executable
23908 This option is available for MinGW targets. It specifies that
23909 the executable flag for the stack used by nested functions isn't
23910 set. This is necessary for binaries running in kernel mode of
23911 Microsoft Windows, as there the User32 API, which is used to set executable
23912 privileges, isn't available.
23914 @item -fwritable-relocated-rdata
23915 @opindex fno-writable-relocated-rdata
23916 This option is available for MinGW and Cygwin targets. It specifies
23917 that relocated-data in read-only section is put into .data
23918 section. This is a necessary for older runtimes not supporting
23919 modification of .rdata sections for pseudo-relocation.
23921 @item -mpe-aligned-commons
23922 @opindex mpe-aligned-commons
23923 This option is available for Cygwin and MinGW targets. It
23924 specifies that the GNU extension to the PE file format that
23925 permits the correct alignment of COMMON variables should be
23926 used when generating code. It is enabled by default if
23927 GCC detects that the target assembler found during configuration
23928 supports the feature.
23931 See also under @ref{x86 Options} for standard options.
23933 @node Xstormy16 Options
23934 @subsection Xstormy16 Options
23935 @cindex Xstormy16 Options
23937 These options are defined for Xstormy16:
23942 Choose startup files and linker script suitable for the simulator.
23945 @node Xtensa Options
23946 @subsection Xtensa Options
23947 @cindex Xtensa Options
23949 These options are supported for Xtensa targets:
23953 @itemx -mno-const16
23955 @opindex mno-const16
23956 Enable or disable use of @code{CONST16} instructions for loading
23957 constant values. The @code{CONST16} instruction is currently not a
23958 standard option from Tensilica. When enabled, @code{CONST16}
23959 instructions are always used in place of the standard @code{L32R}
23960 instructions. The use of @code{CONST16} is enabled by default only if
23961 the @code{L32R} instruction is not available.
23964 @itemx -mno-fused-madd
23965 @opindex mfused-madd
23966 @opindex mno-fused-madd
23967 Enable or disable use of fused multiply/add and multiply/subtract
23968 instructions in the floating-point option. This has no effect if the
23969 floating-point option is not also enabled. Disabling fused multiply/add
23970 and multiply/subtract instructions forces the compiler to use separate
23971 instructions for the multiply and add/subtract operations. This may be
23972 desirable in some cases where strict IEEE 754-compliant results are
23973 required: the fused multiply add/subtract instructions do not round the
23974 intermediate result, thereby producing results with @emph{more} bits of
23975 precision than specified by the IEEE standard. Disabling fused multiply
23976 add/subtract instructions also ensures that the program output is not
23977 sensitive to the compiler's ability to combine multiply and add/subtract
23980 @item -mserialize-volatile
23981 @itemx -mno-serialize-volatile
23982 @opindex mserialize-volatile
23983 @opindex mno-serialize-volatile
23984 When this option is enabled, GCC inserts @code{MEMW} instructions before
23985 @code{volatile} memory references to guarantee sequential consistency.
23986 The default is @option{-mserialize-volatile}. Use
23987 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
23989 @item -mforce-no-pic
23990 @opindex mforce-no-pic
23991 For targets, like GNU/Linux, where all user-mode Xtensa code must be
23992 position-independent code (PIC), this option disables PIC for compiling
23995 @item -mtext-section-literals
23996 @itemx -mno-text-section-literals
23997 @opindex mtext-section-literals
23998 @opindex mno-text-section-literals
23999 These options control the treatment of literal pools. The default is
24000 @option{-mno-text-section-literals}, which places literals in a separate
24001 section in the output file. This allows the literal pool to be placed
24002 in a data RAM/ROM, and it also allows the linker to combine literal
24003 pools from separate object files to remove redundant literals and
24004 improve code size. With @option{-mtext-section-literals}, the literals
24005 are interspersed in the text section in order to keep them as close as
24006 possible to their references. This may be necessary for large assembly
24007 files. Literals for each function are placed right before that function.
24009 @item -mauto-litpools
24010 @itemx -mno-auto-litpools
24011 @opindex mauto-litpools
24012 @opindex mno-auto-litpools
24013 These options control the treatment of literal pools. The default is
24014 @option{-mno-auto-litpools}, which places literals in a separate
24015 section in the output file unless @option{-mtext-section-literals} is
24016 used. With @option{-mauto-litpools} the literals are interspersed in
24017 the text section by the assembler. Compiler does not produce explicit
24018 @code{.literal} directives and loads literals into registers with
24019 @code{MOVI} instructions instead of @code{L32R} to let the assembler
24020 do relaxation and place literals as necessary. This option allows
24021 assembler to create several literal pools per function and assemble
24022 very big functions, which may not be possible with
24023 @option{-mtext-section-literals}.
24025 @item -mtarget-align
24026 @itemx -mno-target-align
24027 @opindex mtarget-align
24028 @opindex mno-target-align
24029 When this option is enabled, GCC instructs the assembler to
24030 automatically align instructions to reduce branch penalties at the
24031 expense of some code density. The assembler attempts to widen density
24032 instructions to align branch targets and the instructions following call
24033 instructions. If there are not enough preceding safe density
24034 instructions to align a target, no widening is performed. The
24035 default is @option{-mtarget-align}. These options do not affect the
24036 treatment of auto-aligned instructions like @code{LOOP}, which the
24037 assembler always aligns, either by widening density instructions or
24038 by inserting NOP instructions.
24041 @itemx -mno-longcalls
24042 @opindex mlongcalls
24043 @opindex mno-longcalls
24044 When this option is enabled, GCC instructs the assembler to translate
24045 direct calls to indirect calls unless it can determine that the target
24046 of a direct call is in the range allowed by the call instruction. This
24047 translation typically occurs for calls to functions in other source
24048 files. Specifically, the assembler translates a direct @code{CALL}
24049 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
24050 The default is @option{-mno-longcalls}. This option should be used in
24051 programs where the call target can potentially be out of range. This
24052 option is implemented in the assembler, not the compiler, so the
24053 assembly code generated by GCC still shows direct call
24054 instructions---look at the disassembled object code to see the actual
24055 instructions. Note that the assembler uses an indirect call for
24056 every cross-file call, not just those that really are out of range.
24059 @node zSeries Options
24060 @subsection zSeries Options
24061 @cindex zSeries options
24063 These are listed under @xref{S/390 and zSeries Options}.
24065 @node Code Gen Options
24066 @section Options for Code Generation Conventions
24067 @cindex code generation conventions
24068 @cindex options, code generation
24069 @cindex run-time options
24071 These machine-independent options control the interface conventions
24072 used in code generation.
24074 Most of them have both positive and negative forms; the negative form
24075 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
24076 one of the forms is listed---the one that is not the default. You
24077 can figure out the other form by either removing @samp{no-} or adding
24081 @item -fbounds-check
24082 @opindex fbounds-check
24083 For front ends that support it, generate additional code to check that
24084 indices used to access arrays are within the declared range. This is
24085 currently only supported by the Java and Fortran front ends, where
24086 this option defaults to true and false respectively.
24088 @item -fstack-reuse=@var{reuse-level}
24089 @opindex fstack_reuse
24090 This option controls stack space reuse for user declared local/auto variables
24091 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
24092 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
24093 local variables and temporaries, @samp{named_vars} enables the reuse only for
24094 user defined local variables with names, and @samp{none} disables stack reuse
24095 completely. The default value is @samp{all}. The option is needed when the
24096 program extends the lifetime of a scoped local variable or a compiler generated
24097 temporary beyond the end point defined by the language. When a lifetime of
24098 a variable ends, and if the variable lives in memory, the optimizing compiler
24099 has the freedom to reuse its stack space with other temporaries or scoped
24100 local variables whose live range does not overlap with it. Legacy code extending
24101 local lifetime is likely to break with the stack reuse optimization.
24120 if (*p == 10) // out of scope use of local1
24131 A(int k) : i(k), j(k) @{ @}
24138 void foo(const A& ar)
24145 foo(A(10)); // temp object's lifetime ends when foo returns
24151 ap->i+= 10; // ap references out of scope temp whose space
24152 // is reused with a. What is the value of ap->i?
24157 The lifetime of a compiler generated temporary is well defined by the C++
24158 standard. When a lifetime of a temporary ends, and if the temporary lives
24159 in memory, the optimizing compiler has the freedom to reuse its stack
24160 space with other temporaries or scoped local variables whose live range
24161 does not overlap with it. However some of the legacy code relies on
24162 the behavior of older compilers in which temporaries' stack space is
24163 not reused, the aggressive stack reuse can lead to runtime errors. This
24164 option is used to control the temporary stack reuse optimization.
24168 This option generates traps for signed overflow on addition, subtraction,
24169 multiplication operations.
24170 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
24171 @option{-ftrapv} @option{-fwrapv} on the command-line results in
24172 @option{-fwrapv} being effective. Note that only active options override, so
24173 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
24174 results in @option{-ftrapv} being effective.
24178 This option instructs the compiler to assume that signed arithmetic
24179 overflow of addition, subtraction and multiplication wraps around
24180 using twos-complement representation. This flag enables some optimizations
24181 and disables others. This option is enabled by default for the Java
24182 front end, as required by the Java language specification.
24183 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
24184 @option{-ftrapv} @option{-fwrapv} on the command-line results in
24185 @option{-fwrapv} being effective. Note that only active options override, so
24186 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
24187 results in @option{-ftrapv} being effective.
24190 @opindex fexceptions
24191 Enable exception handling. Generates extra code needed to propagate
24192 exceptions. For some targets, this implies GCC generates frame
24193 unwind information for all functions, which can produce significant data
24194 size overhead, although it does not affect execution. If you do not
24195 specify this option, GCC enables it by default for languages like
24196 C++ that normally require exception handling, and disables it for
24197 languages like C that do not normally require it. However, you may need
24198 to enable this option when compiling C code that needs to interoperate
24199 properly with exception handlers written in C++. You may also wish to
24200 disable this option if you are compiling older C++ programs that don't
24201 use exception handling.
24203 @item -fnon-call-exceptions
24204 @opindex fnon-call-exceptions
24205 Generate code that allows trapping instructions to throw exceptions.
24206 Note that this requires platform-specific runtime support that does
24207 not exist everywhere. Moreover, it only allows @emph{trapping}
24208 instructions to throw exceptions, i.e.@: memory references or floating-point
24209 instructions. It does not allow exceptions to be thrown from
24210 arbitrary signal handlers such as @code{SIGALRM}.
24212 @item -fdelete-dead-exceptions
24213 @opindex fdelete-dead-exceptions
24214 Consider that instructions that may throw exceptions but don't otherwise
24215 contribute to the execution of the program can be optimized away.
24216 This option is enabled by default for the Ada front end, as permitted by
24217 the Ada language specification.
24218 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
24220 @item -funwind-tables
24221 @opindex funwind-tables
24222 Similar to @option{-fexceptions}, except that it just generates any needed
24223 static data, but does not affect the generated code in any other way.
24224 You normally do not need to enable this option; instead, a language processor
24225 that needs this handling enables it on your behalf.
24227 @item -fasynchronous-unwind-tables
24228 @opindex fasynchronous-unwind-tables
24229 Generate unwind table in DWARF 2 format, if supported by target machine. The
24230 table is exact at each instruction boundary, so it can be used for stack
24231 unwinding from asynchronous events (such as debugger or garbage collector).
24233 @item -fno-gnu-unique
24234 @opindex fno-gnu-unique
24235 On systems with recent GNU assembler and C library, the C++ compiler
24236 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
24237 of template static data members and static local variables in inline
24238 functions are unique even in the presence of @code{RTLD_LOCAL}; this
24239 is necessary to avoid problems with a library used by two different
24240 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
24241 therefore disagreeing with the other one about the binding of the
24242 symbol. But this causes @code{dlclose} to be ignored for affected
24243 DSOs; if your program relies on reinitialization of a DSO via
24244 @code{dlclose} and @code{dlopen}, you can use
24245 @option{-fno-gnu-unique}.
24247 @item -fpcc-struct-return
24248 @opindex fpcc-struct-return
24249 Return ``short'' @code{struct} and @code{union} values in memory like
24250 longer ones, rather than in registers. This convention is less
24251 efficient, but it has the advantage of allowing intercallability between
24252 GCC-compiled files and files compiled with other compilers, particularly
24253 the Portable C Compiler (pcc).
24255 The precise convention for returning structures in memory depends
24256 on the target configuration macros.
24258 Short structures and unions are those whose size and alignment match
24259 that of some integer type.
24261 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
24262 switch is not binary compatible with code compiled with the
24263 @option{-freg-struct-return} switch.
24264 Use it to conform to a non-default application binary interface.
24266 @item -freg-struct-return
24267 @opindex freg-struct-return
24268 Return @code{struct} and @code{union} values in registers when possible.
24269 This is more efficient for small structures than
24270 @option{-fpcc-struct-return}.
24272 If you specify neither @option{-fpcc-struct-return} nor
24273 @option{-freg-struct-return}, GCC defaults to whichever convention is
24274 standard for the target. If there is no standard convention, GCC
24275 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
24276 the principal compiler. In those cases, we can choose the standard, and
24277 we chose the more efficient register return alternative.
24279 @strong{Warning:} code compiled with the @option{-freg-struct-return}
24280 switch is not binary compatible with code compiled with the
24281 @option{-fpcc-struct-return} switch.
24282 Use it to conform to a non-default application binary interface.
24284 @item -fshort-enums
24285 @opindex fshort-enums
24286 Allocate to an @code{enum} type only as many bytes as it needs for the
24287 declared range of possible values. Specifically, the @code{enum} type
24288 is equivalent to the smallest integer type that has enough room.
24290 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
24291 code that is not binary compatible with code generated without that switch.
24292 Use it to conform to a non-default application binary interface.
24294 @item -fshort-double
24295 @opindex fshort-double
24296 Use the same size for @code{double} as for @code{float}.
24298 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
24299 code that is not binary compatible with code generated without that switch.
24300 Use it to conform to a non-default application binary interface.
24302 @item -fshort-wchar
24303 @opindex fshort-wchar
24304 Override the underlying type for @code{wchar_t} to be @code{short
24305 unsigned int} instead of the default for the target. This option is
24306 useful for building programs to run under WINE@.
24308 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
24309 code that is not binary compatible with code generated without that switch.
24310 Use it to conform to a non-default application binary interface.
24313 @opindex fno-common
24314 In C code, controls the placement of uninitialized global variables.
24315 Unix C compilers have traditionally permitted multiple definitions of
24316 such variables in different compilation units by placing the variables
24318 This is the behavior specified by @option{-fcommon}, and is the default
24319 for GCC on most targets.
24320 On the other hand, this behavior is not required by ISO C, and on some
24321 targets may carry a speed or code size penalty on variable references.
24322 The @option{-fno-common} option specifies that the compiler should place
24323 uninitialized global variables in the data section of the object file,
24324 rather than generating them as common blocks.
24325 This has the effect that if the same variable is declared
24326 (without @code{extern}) in two different compilations,
24327 you get a multiple-definition error when you link them.
24328 In this case, you must compile with @option{-fcommon} instead.
24329 Compiling with @option{-fno-common} is useful on targets for which
24330 it provides better performance, or if you wish to verify that the
24331 program will work on other systems that always treat uninitialized
24332 variable declarations this way.
24336 Ignore the @code{#ident} directive.
24338 @item -finhibit-size-directive
24339 @opindex finhibit-size-directive
24340 Don't output a @code{.size} assembler directive, or anything else that
24341 would cause trouble if the function is split in the middle, and the
24342 two halves are placed at locations far apart in memory. This option is
24343 used when compiling @file{crtstuff.c}; you should not need to use it
24346 @item -fverbose-asm
24347 @opindex fverbose-asm
24348 Put extra commentary information in the generated assembly code to
24349 make it more readable. This option is generally only of use to those
24350 who actually need to read the generated assembly code (perhaps while
24351 debugging the compiler itself).
24353 @option{-fno-verbose-asm}, the default, causes the
24354 extra information to be omitted and is useful when comparing two assembler
24357 @item -frecord-gcc-switches
24358 @opindex frecord-gcc-switches
24359 This switch causes the command line used to invoke the
24360 compiler to be recorded into the object file that is being created.
24361 This switch is only implemented on some targets and the exact format
24362 of the recording is target and binary file format dependent, but it
24363 usually takes the form of a section containing ASCII text. This
24364 switch is related to the @option{-fverbose-asm} switch, but that
24365 switch only records information in the assembler output file as
24366 comments, so it never reaches the object file.
24367 See also @option{-grecord-gcc-switches} for another
24368 way of storing compiler options into the object file.
24372 @cindex global offset table
24374 Generate position-independent code (PIC) suitable for use in a shared
24375 library, if supported for the target machine. Such code accesses all
24376 constant addresses through a global offset table (GOT)@. The dynamic
24377 loader resolves the GOT entries when the program starts (the dynamic
24378 loader is not part of GCC; it is part of the operating system). If
24379 the GOT size for the linked executable exceeds a machine-specific
24380 maximum size, you get an error message from the linker indicating that
24381 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
24382 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
24383 on the m68k and RS/6000. The x86 has no such limit.)
24385 Position-independent code requires special support, and therefore works
24386 only on certain machines. For the x86, GCC supports PIC for System V
24387 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
24388 position-independent.
24390 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
24395 If supported for the target machine, emit position-independent code,
24396 suitable for dynamic linking and avoiding any limit on the size of the
24397 global offset table. This option makes a difference on AArch64, m68k,
24398 PowerPC and SPARC@.
24400 Position-independent code requires special support, and therefore works
24401 only on certain machines.
24403 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
24410 These options are similar to @option{-fpic} and @option{-fPIC}, but
24411 generated position independent code can be only linked into executables.
24412 Usually these options are used when @option{-pie} GCC option is
24413 used during linking.
24415 @option{-fpie} and @option{-fPIE} both define the macros
24416 @code{__pie__} and @code{__PIE__}. The macros have the value 1
24417 for @option{-fpie} and 2 for @option{-fPIE}.
24421 Do not use the PLT for external function calls in position-independent code.
24422 Instead, load the callee address at call sites from the GOT and branch to it.
24423 This leads to more efficient code by eliminating PLT stubs and exposing
24424 GOT loads to optimizations. On architectures such as 32-bit x86 where
24425 PLT stubs expect the GOT pointer in a specific register, this gives more
24426 register allocation freedom to the compiler.
24427 Lazy binding requires use of the PLT;
24428 with @option{-fno-plt} all external symbols are resolved at load time.
24430 Alternatively, the function attribute @code{noplt} can be used to avoid calls
24431 through the PLT for specific external functions.
24433 In position-dependent code, a few targets also convert calls to
24434 functions that are marked to not use the PLT to use the GOT instead.
24436 @item -fno-jump-tables
24437 @opindex fno-jump-tables
24438 Do not use jump tables for switch statements even where it would be
24439 more efficient than other code generation strategies. This option is
24440 of use in conjunction with @option{-fpic} or @option{-fPIC} for
24441 building code that forms part of a dynamic linker and cannot
24442 reference the address of a jump table. On some targets, jump tables
24443 do not require a GOT and this option is not needed.
24445 @item -ffixed-@var{reg}
24447 Treat the register named @var{reg} as a fixed register; generated code
24448 should never refer to it (except perhaps as a stack pointer, frame
24449 pointer or in some other fixed role).
24451 @var{reg} must be the name of a register. The register names accepted
24452 are machine-specific and are defined in the @code{REGISTER_NAMES}
24453 macro in the machine description macro file.
24455 This flag does not have a negative form, because it specifies a
24458 @item -fcall-used-@var{reg}
24459 @opindex fcall-used
24460 Treat the register named @var{reg} as an allocable register that is
24461 clobbered by function calls. It may be allocated for temporaries or
24462 variables that do not live across a call. Functions compiled this way
24463 do not save and restore the register @var{reg}.
24465 It is an error to use this flag with the frame pointer or stack pointer.
24466 Use of this flag for other registers that have fixed pervasive roles in
24467 the machine's execution model produces disastrous results.
24469 This flag does not have a negative form, because it specifies a
24472 @item -fcall-saved-@var{reg}
24473 @opindex fcall-saved
24474 Treat the register named @var{reg} as an allocable register saved by
24475 functions. It may be allocated even for temporaries or variables that
24476 live across a call. Functions compiled this way save and restore
24477 the register @var{reg} if they use it.
24479 It is an error to use this flag with the frame pointer or stack pointer.
24480 Use of this flag for other registers that have fixed pervasive roles in
24481 the machine's execution model produces disastrous results.
24483 A different sort of disaster results from the use of this flag for
24484 a register in which function values may be returned.
24486 This flag does not have a negative form, because it specifies a
24489 @item -fpack-struct[=@var{n}]
24490 @opindex fpack-struct
24491 Without a value specified, pack all structure members together without
24492 holes. When a value is specified (which must be a small power of two), pack
24493 structure members according to this value, representing the maximum
24494 alignment (that is, objects with default alignment requirements larger than
24495 this are output potentially unaligned at the next fitting location.
24497 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
24498 code that is not binary compatible with code generated without that switch.
24499 Additionally, it makes the code suboptimal.
24500 Use it to conform to a non-default application binary interface.
24502 @item -finstrument-functions
24503 @opindex finstrument-functions
24504 Generate instrumentation calls for entry and exit to functions. Just
24505 after function entry and just before function exit, the following
24506 profiling functions are called with the address of the current
24507 function and its call site. (On some platforms,
24508 @code{__builtin_return_address} does not work beyond the current
24509 function, so the call site information may not be available to the
24510 profiling functions otherwise.)
24513 void __cyg_profile_func_enter (void *this_fn,
24515 void __cyg_profile_func_exit (void *this_fn,
24519 The first argument is the address of the start of the current function,
24520 which may be looked up exactly in the symbol table.
24522 This instrumentation is also done for functions expanded inline in other
24523 functions. The profiling calls indicate where, conceptually, the
24524 inline function is entered and exited. This means that addressable
24525 versions of such functions must be available. If all your uses of a
24526 function are expanded inline, this may mean an additional expansion of
24527 code size. If you use @code{extern inline} in your C code, an
24528 addressable version of such functions must be provided. (This is
24529 normally the case anyway, but if you get lucky and the optimizer always
24530 expands the functions inline, you might have gotten away without
24531 providing static copies.)
24533 A function may be given the attribute @code{no_instrument_function}, in
24534 which case this instrumentation is not done. This can be used, for
24535 example, for the profiling functions listed above, high-priority
24536 interrupt routines, and any functions from which the profiling functions
24537 cannot safely be called (perhaps signal handlers, if the profiling
24538 routines generate output or allocate memory).
24540 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
24541 @opindex finstrument-functions-exclude-file-list
24543 Set the list of functions that are excluded from instrumentation (see
24544 the description of @option{-finstrument-functions}). If the file that
24545 contains a function definition matches with one of @var{file}, then
24546 that function is not instrumented. The match is done on substrings:
24547 if the @var{file} parameter is a substring of the file name, it is
24548 considered to be a match.
24553 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
24557 excludes any inline function defined in files whose pathnames
24558 contain @file{/bits/stl} or @file{include/sys}.
24560 If, for some reason, you want to include letter @samp{,} in one of
24561 @var{sym}, write @samp{\,}. For example,
24562 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
24563 (note the single quote surrounding the option).
24565 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
24566 @opindex finstrument-functions-exclude-function-list
24568 This is similar to @option{-finstrument-functions-exclude-file-list},
24569 but this option sets the list of function names to be excluded from
24570 instrumentation. The function name to be matched is its user-visible
24571 name, such as @code{vector<int> blah(const vector<int> &)}, not the
24572 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
24573 match is done on substrings: if the @var{sym} parameter is a substring
24574 of the function name, it is considered to be a match. For C99 and C++
24575 extended identifiers, the function name must be given in UTF-8, not
24576 using universal character names.
24578 @item -fstack-check
24579 @opindex fstack-check
24580 Generate code to verify that you do not go beyond the boundary of the
24581 stack. You should specify this flag if you are running in an
24582 environment with multiple threads, but you only rarely need to specify it in
24583 a single-threaded environment since stack overflow is automatically
24584 detected on nearly all systems if there is only one stack.
24586 Note that this switch does not actually cause checking to be done; the
24587 operating system or the language runtime must do that. The switch causes
24588 generation of code to ensure that they see the stack being extended.
24590 You can additionally specify a string parameter: @samp{no} means no
24591 checking, @samp{generic} means force the use of old-style checking,
24592 @samp{specific} means use the best checking method and is equivalent
24593 to bare @option{-fstack-check}.
24595 Old-style checking is a generic mechanism that requires no specific
24596 target support in the compiler but comes with the following drawbacks:
24600 Modified allocation strategy for large objects: they are always
24601 allocated dynamically if their size exceeds a fixed threshold.
24604 Fixed limit on the size of the static frame of functions: when it is
24605 topped by a particular function, stack checking is not reliable and
24606 a warning is issued by the compiler.
24609 Inefficiency: because of both the modified allocation strategy and the
24610 generic implementation, code performance is hampered.
24613 Note that old-style stack checking is also the fallback method for
24614 @samp{specific} if no target support has been added in the compiler.
24616 @item -fstack-limit-register=@var{reg}
24617 @itemx -fstack-limit-symbol=@var{sym}
24618 @itemx -fno-stack-limit
24619 @opindex fstack-limit-register
24620 @opindex fstack-limit-symbol
24621 @opindex fno-stack-limit
24622 Generate code to ensure that the stack does not grow beyond a certain value,
24623 either the value of a register or the address of a symbol. If a larger
24624 stack is required, a signal is raised at run time. For most targets,
24625 the signal is raised before the stack overruns the boundary, so
24626 it is possible to catch the signal without taking special precautions.
24628 For instance, if the stack starts at absolute address @samp{0x80000000}
24629 and grows downwards, you can use the flags
24630 @option{-fstack-limit-symbol=__stack_limit} and
24631 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
24632 of 128KB@. Note that this may only work with the GNU linker.
24634 @item -fsplit-stack
24635 @opindex fsplit-stack
24636 Generate code to automatically split the stack before it overflows.
24637 The resulting program has a discontiguous stack which can only
24638 overflow if the program is unable to allocate any more memory. This
24639 is most useful when running threaded programs, as it is no longer
24640 necessary to calculate a good stack size to use for each thread. This
24641 is currently only implemented for the x86 targets running
24644 When code compiled with @option{-fsplit-stack} calls code compiled
24645 without @option{-fsplit-stack}, there may not be much stack space
24646 available for the latter code to run. If compiling all code,
24647 including library code, with @option{-fsplit-stack} is not an option,
24648 then the linker can fix up these calls so that the code compiled
24649 without @option{-fsplit-stack} always has a large stack. Support for
24650 this is implemented in the gold linker in GNU binutils release 2.21
24653 @item -fleading-underscore
24654 @opindex fleading-underscore
24655 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
24656 change the way C symbols are represented in the object file. One use
24657 is to help link with legacy assembly code.
24659 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
24660 generate code that is not binary compatible with code generated without that
24661 switch. Use it to conform to a non-default application binary interface.
24662 Not all targets provide complete support for this switch.
24664 @item -ftls-model=@var{model}
24665 @opindex ftls-model
24666 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
24667 The @var{model} argument should be one of @samp{global-dynamic},
24668 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
24669 Note that the choice is subject to optimization: the compiler may use
24670 a more efficient model for symbols not visible outside of the translation
24671 unit, or if @option{-fpic} is not given on the command line.
24673 The default without @option{-fpic} is @samp{initial-exec}; with
24674 @option{-fpic} the default is @samp{global-dynamic}.
24676 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
24677 @opindex fvisibility
24678 Set the default ELF image symbol visibility to the specified option---all
24679 symbols are marked with this unless overridden within the code.
24680 Using this feature can very substantially improve linking and
24681 load times of shared object libraries, produce more optimized
24682 code, provide near-perfect API export and prevent symbol clashes.
24683 It is @strong{strongly} recommended that you use this in any shared objects
24686 Despite the nomenclature, @samp{default} always means public; i.e.,
24687 available to be linked against from outside the shared object.
24688 @samp{protected} and @samp{internal} are pretty useless in real-world
24689 usage so the only other commonly used option is @samp{hidden}.
24690 The default if @option{-fvisibility} isn't specified is
24691 @samp{default}, i.e., make every symbol public.
24693 A good explanation of the benefits offered by ensuring ELF
24694 symbols have the correct visibility is given by ``How To Write
24695 Shared Libraries'' by Ulrich Drepper (which can be found at
24696 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
24697 solution made possible by this option to marking things hidden when
24698 the default is public is to make the default hidden and mark things
24699 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
24700 and @code{__attribute__ ((visibility("default")))} instead of
24701 @code{__declspec(dllexport)} you get almost identical semantics with
24702 identical syntax. This is a great boon to those working with
24703 cross-platform projects.
24705 For those adding visibility support to existing code, you may find
24706 @code{#pragma GCC visibility} of use. This works by you enclosing
24707 the declarations you wish to set visibility for with (for example)
24708 @code{#pragma GCC visibility push(hidden)} and
24709 @code{#pragma GCC visibility pop}.
24710 Bear in mind that symbol visibility should be viewed @strong{as
24711 part of the API interface contract} and thus all new code should
24712 always specify visibility when it is not the default; i.e., declarations
24713 only for use within the local DSO should @strong{always} be marked explicitly
24714 as hidden as so to avoid PLT indirection overheads---making this
24715 abundantly clear also aids readability and self-documentation of the code.
24716 Note that due to ISO C++ specification requirements, @code{operator new} and
24717 @code{operator delete} must always be of default visibility.
24719 Be aware that headers from outside your project, in particular system
24720 headers and headers from any other library you use, may not be
24721 expecting to be compiled with visibility other than the default. You
24722 may need to explicitly say @code{#pragma GCC visibility push(default)}
24723 before including any such headers.
24725 @code{extern} declarations are not affected by @option{-fvisibility}, so
24726 a lot of code can be recompiled with @option{-fvisibility=hidden} with
24727 no modifications. However, this means that calls to @code{extern}
24728 functions with no explicit visibility use the PLT, so it is more
24729 effective to use @code{__attribute ((visibility))} and/or
24730 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
24731 declarations should be treated as hidden.
24733 Note that @option{-fvisibility} does affect C++ vague linkage
24734 entities. This means that, for instance, an exception class that is
24735 be thrown between DSOs must be explicitly marked with default
24736 visibility so that the @samp{type_info} nodes are unified between
24739 An overview of these techniques, their benefits and how to use them
24740 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
24742 @item -fstrict-volatile-bitfields
24743 @opindex fstrict-volatile-bitfields
24744 This option should be used if accesses to volatile bit-fields (or other
24745 structure fields, although the compiler usually honors those types
24746 anyway) should use a single access of the width of the
24747 field's type, aligned to a natural alignment if possible. For
24748 example, targets with memory-mapped peripheral registers might require
24749 all such accesses to be 16 bits wide; with this flag you can
24750 declare all peripheral bit-fields as @code{unsigned short} (assuming short
24751 is 16 bits on these targets) to force GCC to use 16-bit accesses
24752 instead of, perhaps, a more efficient 32-bit access.
24754 If this option is disabled, the compiler uses the most efficient
24755 instruction. In the previous example, that might be a 32-bit load
24756 instruction, even though that accesses bytes that do not contain
24757 any portion of the bit-field, or memory-mapped registers unrelated to
24758 the one being updated.
24760 In some cases, such as when the @code{packed} attribute is applied to a
24761 structure field, it may not be possible to access the field with a single
24762 read or write that is correctly aligned for the target machine. In this
24763 case GCC falls back to generating multiple accesses rather than code that
24764 will fault or truncate the result at run time.
24766 Note: Due to restrictions of the C/C++11 memory model, write accesses are
24767 not allowed to touch non bit-field members. It is therefore recommended
24768 to define all bits of the field's type as bit-field members.
24770 The default value of this option is determined by the application binary
24771 interface for the target processor.
24773 @item -fsync-libcalls
24774 @opindex fsync-libcalls
24775 This option controls whether any out-of-line instance of the @code{__sync}
24776 family of functions may be used to implement the C++11 @code{__atomic}
24777 family of functions.
24779 The default value of this option is enabled, thus the only useful form
24780 of the option is @option{-fno-sync-libcalls}. This option is used in
24781 the implementation of the @file{libatomic} runtime library.
24787 @node Environment Variables
24788 @section Environment Variables Affecting GCC
24789 @cindex environment variables
24791 @c man begin ENVIRONMENT
24792 This section describes several environment variables that affect how GCC
24793 operates. Some of them work by specifying directories or prefixes to use
24794 when searching for various kinds of files. Some are used to specify other
24795 aspects of the compilation environment.
24797 Note that you can also specify places to search using options such as
24798 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
24799 take precedence over places specified using environment variables, which
24800 in turn take precedence over those specified by the configuration of GCC@.
24801 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
24802 GNU Compiler Collection (GCC) Internals}.
24807 @c @itemx LC_COLLATE
24809 @c @itemx LC_MONETARY
24810 @c @itemx LC_NUMERIC
24815 @c @findex LC_COLLATE
24816 @findex LC_MESSAGES
24817 @c @findex LC_MONETARY
24818 @c @findex LC_NUMERIC
24822 These environment variables control the way that GCC uses
24823 localization information which allows GCC to work with different
24824 national conventions. GCC inspects the locale categories
24825 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
24826 so. These locale categories can be set to any value supported by your
24827 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
24828 Kingdom encoded in UTF-8.
24830 The @env{LC_CTYPE} environment variable specifies character
24831 classification. GCC uses it to determine the character boundaries in
24832 a string; this is needed for some multibyte encodings that contain quote
24833 and escape characters that are otherwise interpreted as a string
24836 The @env{LC_MESSAGES} environment variable specifies the language to
24837 use in diagnostic messages.
24839 If the @env{LC_ALL} environment variable is set, it overrides the value
24840 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
24841 and @env{LC_MESSAGES} default to the value of the @env{LANG}
24842 environment variable. If none of these variables are set, GCC
24843 defaults to traditional C English behavior.
24847 If @env{TMPDIR} is set, it specifies the directory to use for temporary
24848 files. GCC uses temporary files to hold the output of one stage of
24849 compilation which is to be used as input to the next stage: for example,
24850 the output of the preprocessor, which is the input to the compiler
24853 @item GCC_COMPARE_DEBUG
24854 @findex GCC_COMPARE_DEBUG
24855 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
24856 @option{-fcompare-debug} to the compiler driver. See the documentation
24857 of this option for more details.
24859 @item GCC_EXEC_PREFIX
24860 @findex GCC_EXEC_PREFIX
24861 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
24862 names of the subprograms executed by the compiler. No slash is added
24863 when this prefix is combined with the name of a subprogram, but you can
24864 specify a prefix that ends with a slash if you wish.
24866 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
24867 an appropriate prefix to use based on the pathname it is invoked with.
24869 If GCC cannot find the subprogram using the specified prefix, it
24870 tries looking in the usual places for the subprogram.
24872 The default value of @env{GCC_EXEC_PREFIX} is
24873 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
24874 the installed compiler. In many cases @var{prefix} is the value
24875 of @code{prefix} when you ran the @file{configure} script.
24877 Other prefixes specified with @option{-B} take precedence over this prefix.
24879 This prefix is also used for finding files such as @file{crt0.o} that are
24882 In addition, the prefix is used in an unusual way in finding the
24883 directories to search for header files. For each of the standard
24884 directories whose name normally begins with @samp{/usr/local/lib/gcc}
24885 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
24886 replacing that beginning with the specified prefix to produce an
24887 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
24888 @file{foo/bar} just before it searches the standard directory
24889 @file{/usr/local/lib/bar}.
24890 If a standard directory begins with the configured
24891 @var{prefix} then the value of @var{prefix} is replaced by
24892 @env{GCC_EXEC_PREFIX} when looking for header files.
24894 @item COMPILER_PATH
24895 @findex COMPILER_PATH
24896 The value of @env{COMPILER_PATH} is a colon-separated list of
24897 directories, much like @env{PATH}. GCC tries the directories thus
24898 specified when searching for subprograms, if it can't find the
24899 subprograms using @env{GCC_EXEC_PREFIX}.
24902 @findex LIBRARY_PATH
24903 The value of @env{LIBRARY_PATH} is a colon-separated list of
24904 directories, much like @env{PATH}. When configured as a native compiler,
24905 GCC tries the directories thus specified when searching for special
24906 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
24907 using GCC also uses these directories when searching for ordinary
24908 libraries for the @option{-l} option (but directories specified with
24909 @option{-L} come first).
24913 @cindex locale definition
24914 This variable is used to pass locale information to the compiler. One way in
24915 which this information is used is to determine the character set to be used
24916 when character literals, string literals and comments are parsed in C and C++.
24917 When the compiler is configured to allow multibyte characters,
24918 the following values for @env{LANG} are recognized:
24922 Recognize JIS characters.
24924 Recognize SJIS characters.
24926 Recognize EUCJP characters.
24929 If @env{LANG} is not defined, or if it has some other value, then the
24930 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
24931 recognize and translate multibyte characters.
24935 Some additional environment variables affect the behavior of the
24938 @include cppenv.texi
24942 @node Precompiled Headers
24943 @section Using Precompiled Headers
24944 @cindex precompiled headers
24945 @cindex speed of compilation
24947 Often large projects have many header files that are included in every
24948 source file. The time the compiler takes to process these header files
24949 over and over again can account for nearly all of the time required to
24950 build the project. To make builds faster, GCC allows you to
24951 @dfn{precompile} a header file.
24953 To create a precompiled header file, simply compile it as you would any
24954 other file, if necessary using the @option{-x} option to make the driver
24955 treat it as a C or C++ header file. You may want to use a
24956 tool like @command{make} to keep the precompiled header up-to-date when
24957 the headers it contains change.
24959 A precompiled header file is searched for when @code{#include} is
24960 seen in the compilation. As it searches for the included file
24961 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
24962 compiler looks for a precompiled header in each directory just before it
24963 looks for the include file in that directory. The name searched for is
24964 the name specified in the @code{#include} with @samp{.gch} appended. If
24965 the precompiled header file can't be used, it is ignored.
24967 For instance, if you have @code{#include "all.h"}, and you have
24968 @file{all.h.gch} in the same directory as @file{all.h}, then the
24969 precompiled header file is used if possible, and the original
24970 header is used otherwise.
24972 Alternatively, you might decide to put the precompiled header file in a
24973 directory and use @option{-I} to ensure that directory is searched
24974 before (or instead of) the directory containing the original header.
24975 Then, if you want to check that the precompiled header file is always
24976 used, you can put a file of the same name as the original header in this
24977 directory containing an @code{#error} command.
24979 This also works with @option{-include}. So yet another way to use
24980 precompiled headers, good for projects not designed with precompiled
24981 header files in mind, is to simply take most of the header files used by
24982 a project, include them from another header file, precompile that header
24983 file, and @option{-include} the precompiled header. If the header files
24984 have guards against multiple inclusion, they are skipped because
24985 they've already been included (in the precompiled header).
24987 If you need to precompile the same header file for different
24988 languages, targets, or compiler options, you can instead make a
24989 @emph{directory} named like @file{all.h.gch}, and put each precompiled
24990 header in the directory, perhaps using @option{-o}. It doesn't matter
24991 what you call the files in the directory; every precompiled header in
24992 the directory is considered. The first precompiled header
24993 encountered in the directory that is valid for this compilation is
24994 used; they're searched in no particular order.
24996 There are many other possibilities, limited only by your imagination,
24997 good sense, and the constraints of your build system.
24999 A precompiled header file can be used only when these conditions apply:
25003 Only one precompiled header can be used in a particular compilation.
25006 A precompiled header can't be used once the first C token is seen. You
25007 can have preprocessor directives before a precompiled header; you cannot
25008 include a precompiled header from inside another header.
25011 The precompiled header file must be produced for the same language as
25012 the current compilation. You can't use a C precompiled header for a C++
25016 The precompiled header file must have been produced by the same compiler
25017 binary as the current compilation is using.
25020 Any macros defined before the precompiled header is included must
25021 either be defined in the same way as when the precompiled header was
25022 generated, or must not affect the precompiled header, which usually
25023 means that they don't appear in the precompiled header at all.
25025 The @option{-D} option is one way to define a macro before a
25026 precompiled header is included; using a @code{#define} can also do it.
25027 There are also some options that define macros implicitly, like
25028 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
25031 @item If debugging information is output when using the precompiled
25032 header, using @option{-g} or similar, the same kind of debugging information
25033 must have been output when building the precompiled header. However,
25034 a precompiled header built using @option{-g} can be used in a compilation
25035 when no debugging information is being output.
25037 @item The same @option{-m} options must generally be used when building
25038 and using the precompiled header. @xref{Submodel Options},
25039 for any cases where this rule is relaxed.
25041 @item Each of the following options must be the same when building and using
25042 the precompiled header:
25044 @gccoptlist{-fexceptions}
25047 Some other command-line options starting with @option{-f},
25048 @option{-p}, or @option{-O} must be defined in the same way as when
25049 the precompiled header was generated. At present, it's not clear
25050 which options are safe to change and which are not; the safest choice
25051 is to use exactly the same options when generating and using the
25052 precompiled header. The following are known to be safe:
25054 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
25055 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
25056 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
25061 For all of these except the last, the compiler automatically
25062 ignores the precompiled header if the conditions aren't met. If you
25063 find an option combination that doesn't work and doesn't cause the
25064 precompiled header to be ignored, please consider filing a bug report,
25067 If you do use differing options when generating and using the
25068 precompiled header, the actual behavior is a mixture of the
25069 behavior for the options. For instance, if you use @option{-g} to
25070 generate the precompiled header but not when using it, you may or may
25071 not get debugging information for routines in the precompiled header.