1 @c Copyright (C) 1988-2016 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
10 @c man begin COPYRIGHT
11 Copyright @copyright{} 1988-2016 Free Software Foundation, Inc.
13 Permission is granted to copy, distribute and/or modify this document
14 under the terms of the GNU Free Documentation License, Version 1.3 or
15 any later version published by the Free Software Foundation; with the
16 Invariant Sections being ``GNU General Public License'' and ``Funding
17 Free Software'', the Front-Cover texts being (a) (see below), and with
18 the Back-Cover Texts being (b) (see below). A copy of the license is
19 included in the gfdl(7) man page.
21 (a) The FSF's Front-Cover Text is:
25 (b) The FSF's Back-Cover Text is:
27 You have freedom to copy and modify this GNU Manual, like GNU
28 software. Copies published by the Free Software Foundation raise
29 funds for GNU development.
31 @c Set file name and title for the man page.
33 @settitle GNU project C and C++ compiler
35 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
36 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
37 [@option{-W}@var{warn}@dots{}] [@option{-Wpedantic}]
38 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
39 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
40 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
41 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
43 Only the most useful options are listed here; see below for the
44 remainder. @command{g++} accepts mostly the same options as @command{gcc}.
47 gpl(7), gfdl(7), fsf-funding(7),
48 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
49 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
50 @file{ld}, @file{binutils} and @file{gdb}.
53 For instructions on reporting bugs, see
57 See the Info entry for @command{gcc}, or
58 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
59 for contributors to GCC@.
64 @chapter GCC Command Options
65 @cindex GCC command options
66 @cindex command options
67 @cindex options, GCC command
69 @c man begin DESCRIPTION
70 When you invoke GCC, it normally does preprocessing, compilation,
71 assembly and linking. The ``overall options'' allow you to stop this
72 process at an intermediate stage. For example, the @option{-c} option
73 says not to run the linker. Then the output consists of object files
74 output by the assembler.
75 @xref{Overall Options,,Options Controlling the Kind of Output}.
77 Other options are passed on to one or more stages of processing. Some options
78 control the preprocessor and others the compiler itself. Yet other
79 options control the assembler and linker; most of these are not
80 documented here, since you rarely need to use any of them.
82 @cindex C compilation options
83 Most of the command-line options that you can use with GCC are useful
84 for C programs; when an option is only useful with another language
85 (usually C++), the explanation says so explicitly. If the description
86 for a particular option does not mention a source language, you can use
87 that option with all supported languages.
89 @cindex cross compiling
90 @cindex specifying machine version
91 @cindex specifying compiler version and target machine
92 @cindex compiler version, specifying
93 @cindex target machine, specifying
94 The usual way to run GCC is to run the executable called @command{gcc}, or
95 @command{@var{machine}-gcc} when cross-compiling, or
96 @command{@var{machine}-gcc-@var{version}} to run a specific version of GCC.
97 When you compile C++ programs, you should invoke GCC as @command{g++}
98 instead. @xref{Invoking G++,,Compiling C++ Programs},
99 for information about the differences in behavior between @command{gcc}
100 and @code{g++} when compiling C++ programs.
102 @cindex grouping options
103 @cindex options, grouping
104 The @command{gcc} program accepts options and file names as operands. Many
105 options have multi-letter names; therefore multiple single-letter options
106 may @emph{not} be grouped: @option{-dv} is very different from @w{@samp{-d
109 @cindex order of options
110 @cindex options, order
111 You can mix options and other arguments. For the most part, the order
112 you use doesn't matter. Order does matter when you use several
113 options of the same kind; for example, if you specify @option{-L} more
114 than once, the directories are searched in the order specified. Also,
115 the placement of the @option{-l} option is significant.
117 Many options have long names starting with @samp{-f} or with
118 @samp{-W}---for example,
119 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
120 these have both positive and negative forms; the negative form of
121 @option{-ffoo} is @option{-fno-foo}. This manual documents
122 only one of these two forms, whichever one is not the default.
126 @xref{Option Index}, for an index to GCC's options.
129 * Option Summary:: Brief list of all options, without explanations.
130 * Overall Options:: Controlling the kind of output:
131 an executable, object files, assembler files,
132 or preprocessed source.
133 * Invoking G++:: Compiling C++ programs.
134 * C Dialect Options:: Controlling the variant of C language compiled.
135 * C++ Dialect Options:: Variations on C++.
136 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
138 * Diagnostic Message Formatting Options:: Controlling how diagnostics should
140 * Warning Options:: How picky should the compiler be?
141 * Debugging Options:: Producing debuggable code.
142 * Optimize Options:: How much optimization?
143 * Instrumentation Options:: Enabling profiling and extra run-time error checking.
144 * Preprocessor Options:: Controlling header files and macro definitions.
145 Also, getting dependency information for Make.
146 * Assembler Options:: Passing options to the assembler.
147 * Link Options:: Specifying libraries and so on.
148 * Directory Options:: Where to find header files and libraries.
149 Where to find the compiler executable files.
150 * Code Gen Options:: Specifying conventions for function calls, data layout
152 * Developer Options:: Printing GCC configuration info, statistics, and
154 * Submodel Options:: Target-specific options, such as compiling for a
155 specific processor variant.
156 * Spec Files:: How to pass switches to sub-processes.
157 * Environment Variables:: Env vars that affect GCC.
158 * Precompiled Headers:: Compiling a header once, and using it many times.
164 @section Option Summary
166 Here is a summary of all the options, grouped by type. Explanations are
167 in the following sections.
170 @item Overall Options
171 @xref{Overall Options,,Options Controlling the Kind of Output}.
172 @gccoptlist{-c -S -E -o @var{file} -x @var{language} @gol
173 -v -### --help@r{[}=@var{class}@r{[},@dots{}@r{]]} --target-help --version @gol
174 -pass-exit-codes -pipe -specs=@var{file} -wrapper @gol
175 @@@var{file} -fplugin=@var{file} -fplugin-arg-@var{name}=@var{arg} @gol
176 -fdump-ada-spec@r{[}-slim@r{]} -fada-spec-parent=@var{unit} -fdump-go-spec=@var{file}}
178 @item C Language Options
179 @xref{C Dialect Options,,Options Controlling C Dialect}.
180 @gccoptlist{-ansi -std=@var{standard} -fgnu89-inline @gol
181 -aux-info @var{filename} -fallow-parameterless-variadic-functions @gol
182 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
183 -fhosted -ffreestanding -fopenacc -fopenmp -fopenmp-simd @gol
184 -fms-extensions -fplan9-extensions -fsso-struct=@var{endianness}
185 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
186 -fsigned-bitfields -fsigned-char @gol
187 -funsigned-bitfields -funsigned-char @gol
188 -trigraphs -traditional -traditional-cpp}
190 @item C++ Language Options
191 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
192 @gccoptlist{-fabi-version=@var{n} -fno-access-control @gol
193 -fargs-in-order=@var{n} -fcheck-new @gol
194 -fconstexpr-depth=@var{n} -ffriend-injection @gol
195 -fno-elide-constructors @gol
196 -fno-enforce-eh-specs @gol
197 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
198 -fno-implicit-templates @gol
199 -fno-implicit-inline-templates @gol
200 -fno-implement-inlines -fms-extensions @gol
201 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names @gol
202 -fno-optional-diags -fpermissive @gol
203 -fno-pretty-templates @gol
204 -frepo -fno-rtti -fsized-deallocation @gol
205 -ftemplate-backtrace-limit=@var{n} @gol
206 -ftemplate-depth=@var{n} @gol
207 -fno-threadsafe-statics -fuse-cxa-atexit @gol
208 -fno-weak -nostdinc++ @gol
209 -fvisibility-inlines-hidden @gol
210 -fvisibility-ms-compat @gol
211 -fext-numeric-literals @gol
212 -Wabi=@var{n} -Wabi-tag -Wconversion-null -Wctor-dtor-privacy @gol
213 -Wdelete-non-virtual-dtor -Wliteral-suffix -Wmultiple-inheritance @gol
214 -Wnamespaces -Wnarrowing @gol
215 -Wnoexcept -Wnon-virtual-dtor -Wreorder @gol
216 -Weffc++ -Wstrict-null-sentinel -Wtemplates @gol
217 -Wno-non-template-friend -Wold-style-cast @gol
218 -Woverloaded-virtual -Wno-pmf-conversions @gol
219 -Wsign-promo -Wvirtual-inheritance}
221 @item Objective-C and Objective-C++ Language Options
222 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
223 Objective-C and Objective-C++ Dialects}.
224 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
225 -fgnu-runtime -fnext-runtime @gol
226 -fno-nil-receivers @gol
227 -fobjc-abi-version=@var{n} @gol
228 -fobjc-call-cxx-cdtors @gol
229 -fobjc-direct-dispatch @gol
230 -fobjc-exceptions @gol
233 -fobjc-std=objc1 @gol
234 -fno-local-ivars @gol
235 -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]} @gol
236 -freplace-objc-classes @gol
239 -Wassign-intercept @gol
240 -Wno-protocol -Wselector @gol
241 -Wstrict-selector-match @gol
242 -Wundeclared-selector}
244 @item Diagnostic Message Formatting Options
245 @xref{Diagnostic Message Formatting Options,,Options to Control Diagnostic Messages Formatting}.
246 @gccoptlist{-fmessage-length=@var{n} @gol
247 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
248 -fdiagnostics-color=@r{[}auto@r{|}never@r{|}always@r{]} @gol
249 -fno-diagnostics-show-option -fno-diagnostics-show-caret @gol
250 -fdiagnostics-parseable-fixits}
252 @item Warning Options
253 @xref{Warning Options,,Options to Request or Suppress Warnings}.
254 @gccoptlist{-fsyntax-only -fmax-errors=@var{n} -Wpedantic @gol
255 -pedantic-errors @gol
256 -w -Wextra -Wall -Waddress -Waggregate-return @gol
257 -Wno-aggressive-loop-optimizations -Warray-bounds -Warray-bounds=@var{n} @gol
258 -Wno-attributes -Wbool-compare -Wno-builtin-macro-redefined @gol
259 -Wc90-c99-compat -Wc99-c11-compat @gol
260 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wcast-align -Wcast-qual @gol
261 -Wchar-subscripts -Wclobbered -Wcomment -Wconditionally-supported @gol
262 -Wconversion -Wcoverage-mismatch -Wno-cpp -Wdangling-else -Wdate-time @gol
263 -Wdelete-incomplete @gol
264 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init @gol
265 -Wdisabled-optimization @gol
266 -Wno-discarded-qualifiers -Wno-discarded-array-qualifiers @gol
267 -Wno-div-by-zero -Wdouble-promotion -Wduplicated-cond @gol
268 -Wempty-body -Wenum-compare -Wno-endif-labels @gol
269 -Werror -Werror=* -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
270 -Wno-format-contains-nul -Wno-format-extra-args -Wformat-nonliteral @gol
271 -Wformat-security -Wformat-signedness -Wformat-y2k -Wframe-address @gol
272 -Wframe-larger-than=@var{len} -Wno-free-nonheap-object -Wjump-misses-init @gol
273 -Wignored-qualifiers -Wignored-attributes -Wincompatible-pointer-types @gol
274 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
275 -Winit-self -Winline -Wno-int-conversion @gol
276 -Wno-int-to-pointer-cast -Winvalid-memory-model -Wno-invalid-offsetof @gol
277 -Winvalid-pch -Wlarger-than=@var{len} @gol
278 -Wlogical-op -Wlogical-not-parentheses -Wlong-long @gol
279 -Wmain -Wmaybe-uninitialized -Wmemset-elt-size -Wmemset-transposed-args @gol
280 -Wmisleading-indentation -Wmissing-braces @gol
281 -Wmissing-field-initializers -Wmissing-include-dirs @gol
282 -Wno-multichar -Wnonnull -Wnonnull-compare @gol
283 -Wnormalized=@r{[}none@r{|}id@r{|}nfc@r{|}nfkc@r{]} @gol
284 -Wnull-dereference -Wodr -Wno-overflow -Wopenmp-simd @gol
285 -Woverride-init-side-effects -Woverlength-strings @gol
286 -Wpacked -Wpacked-bitfield-compat -Wpadded @gol
287 -Wparentheses -Wno-pedantic-ms-format @gol
288 -Wplacement-new -Wplacement-new=@var{n} @gol
289 -Wpointer-arith -Wno-pointer-to-int-cast @gol
290 -Wno-pragmas -Wredundant-decls -Wno-return-local-addr @gol
291 -Wreturn-type -Wsequence-point -Wshadow -Wno-shadow-ivar @gol
292 -Wshift-overflow -Wshift-overflow=@var{n} @gol
293 -Wshift-count-negative -Wshift-count-overflow -Wshift-negative-value @gol
294 -Wsign-compare -Wsign-conversion -Wfloat-conversion @gol
295 -Wno-scalar-storage-order @gol
296 -Wsizeof-pointer-memaccess -Wsizeof-array-argument @gol
297 -Wstack-protector -Wstack-usage=@var{len} -Wstrict-aliasing @gol
298 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=@var{n} @gol
299 -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]} @gol
300 -Wsuggest-final-types @gol -Wsuggest-final-methods -Wsuggest-override @gol
301 -Wmissing-format-attribute -Wsubobject-linkage @gol
302 -Wswitch -Wswitch-bool -Wswitch-default -Wswitch-enum @gol
303 -Wswitch-unreachable -Wsync-nand @gol
304 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs @gol
305 -Wtype-limits -Wundef @gol
306 -Wuninitialized -Wunknown-pragmas -Wunsafe-loop-optimizations @gol
307 -Wunsuffixed-float-constants -Wunused -Wunused-function @gol
308 -Wunused-label -Wunused-local-typedefs -Wunused-parameter @gol
309 -Wno-unused-result -Wunused-value @gol -Wunused-variable @gol
310 -Wunused-const-variable -Wunused-const-variable=@var{n} @gol
311 -Wunused-but-set-parameter -Wunused-but-set-variable @gol
312 -Wuseless-cast -Wvariadic-macros -Wvector-operation-performance @gol
313 -Wvla -Wvolatile-register-var -Wwrite-strings @gol
314 -Wzero-as-null-pointer-constant -Whsa}
316 @item C and Objective-C-only Warning Options
317 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
318 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
319 -Wold-style-declaration -Wold-style-definition @gol
320 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
321 -Wdeclaration-after-statement -Wpointer-sign}
323 @item Debugging Options
324 @xref{Debugging Options,,Options for Debugging Your Program}.
325 @gccoptlist{-g -g@var{level} -gcoff -gdwarf -gdwarf-@var{version} @gol
326 -ggdb -grecord-gcc-switches -gno-record-gcc-switches @gol
327 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf @gol
328 -gvms -gxcoff -gxcoff+ -gz@r{[}=@var{type}@r{]} @gol
329 -fdebug-prefix-map=@var{old}=@var{new} -fdebug-types-section @gol
330 -feliminate-dwarf2-dups -fno-eliminate-unused-debug-types @gol
331 -femit-struct-debug-baseonly -femit-struct-debug-reduced @gol
332 -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]} @gol
333 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
334 -fno-merge-debug-strings -fno-dwarf2-cfi-asm @gol
335 -fvar-tracking -fvar-tracking-assignments}
337 @item Optimization Options
338 @xref{Optimize Options,,Options that Control Optimization}.
339 @gccoptlist{-faggressive-loop-optimizations -falign-functions[=@var{n}] @gol
340 -falign-jumps[=@var{n}] @gol
341 -falign-labels[=@var{n}] -falign-loops[=@var{n}] @gol
342 -fassociative-math -fauto-profile -fauto-profile[=@var{path}] @gol
343 -fauto-inc-dec -fbranch-probabilities @gol
344 -fbranch-target-load-optimize -fbranch-target-load-optimize2 @gol
345 -fbtr-bb-exclusive -fcaller-saves @gol
346 -fcombine-stack-adjustments -fconserve-stack @gol
347 -fcompare-elim -fcprop-registers -fcrossjumping @gol
348 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules @gol
349 -fcx-limited-range @gol
350 -fdata-sections -fdce -fdelayed-branch @gol
351 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively @gol
352 -fdevirtualize-at-ltrans -fdse @gol
353 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects @gol
354 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=@var{style} @gol
355 -fforward-propagate -ffp-contract=@var{style} -ffunction-sections @gol
356 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity @gol
357 -fgcse-sm -fhoist-adjacent-loads -fif-conversion @gol
358 -fif-conversion2 -findirect-inlining @gol
359 -finline-functions -finline-functions-called-once -finline-limit=@var{n} @gol
360 -finline-small-functions -fipa-cp -fipa-cp-clone -fipa-cp-alignment @gol
361 -fipa-pta -fipa-profile -fipa-pure-const -fipa-reference -fipa-icf @gol
362 -fira-algorithm=@var{algorithm} @gol
363 -fira-region=@var{region} -fira-hoist-pressure @gol
364 -fira-loop-pressure -fno-ira-share-save-slots @gol
365 -fno-ira-share-spill-slots @gol
366 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute @gol
367 -fivopts -fkeep-inline-functions -fkeep-static-functions @gol
368 -fkeep-static-consts -flive-range-shrinkage @gol
369 -floop-block -floop-interchange -floop-strip-mine @gol
370 -floop-unroll-and-jam -floop-nest-optimize @gol
371 -floop-parallelize-all -flra-remat -flto -flto-compression-level @gol
372 -flto-partition=@var{alg} -fmerge-all-constants @gol
373 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves @gol
374 -fmove-loop-invariants -fno-branch-count-reg @gol
375 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse @gol
376 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole @gol
377 -fno-peephole2 -fno-sched-interblock -fno-sched-spec -fno-signed-zeros @gol
378 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
379 -fomit-frame-pointer -foptimize-sibling-calls @gol
380 -fpartial-inlining -fpeel-loops -fpredictive-commoning @gol
381 -fprefetch-loop-arrays @gol
382 -fprofile-correction @gol
383 -fprofile-use -fprofile-use=@var{path} -fprofile-values @gol
384 -fprofile-reorder-functions @gol
385 -freciprocal-math -free -frename-registers -freorder-blocks @gol
386 -freorder-blocks-algorithm=@var{algorithm} @gol
387 -freorder-blocks-and-partition -freorder-functions @gol
388 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops @gol
389 -frounding-math -fsched2-use-superblocks -fsched-pressure @gol
390 -fsched-spec-load -fsched-spec-load-dangerous @gol
391 -fsched-stalled-insns-dep[=@var{n}] -fsched-stalled-insns[=@var{n}] @gol
392 -fsched-group-heuristic -fsched-critical-path-heuristic @gol
393 -fsched-spec-insn-heuristic -fsched-rank-heuristic @gol
394 -fsched-last-insn-heuristic -fsched-dep-count-heuristic @gol
395 -fschedule-fusion @gol
396 -fschedule-insns -fschedule-insns2 -fsection-anchors @gol
397 -fselective-scheduling -fselective-scheduling2 @gol
398 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops @gol
399 -fsemantic-interposition -fshrink-wrap -fsignaling-nans @gol
400 -fsingle-precision-constant -fsplit-ivs-in-unroller @gol
402 -fsplit-wide-types -fssa-backprop -fssa-phiopt @gol
403 -fstdarg-opt -fstrict-aliasing @gol
404 -fstrict-overflow -fthread-jumps -ftracer -ftree-bit-ccp @gol
405 -ftree-builtin-call-dce -ftree-ccp -ftree-ch @gol
406 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts @gol
407 -ftree-dse -ftree-forwprop -ftree-fre -ftree-loop-if-convert @gol
408 -ftree-loop-if-convert-stores -ftree-loop-im @gol
409 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns @gol
410 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize @gol
411 -ftree-loop-vectorize @gol
412 -ftree-parallelize-loops=@var{n} -ftree-pre -ftree-partial-pre -ftree-pta @gol
413 -ftree-reassoc -ftree-sink -ftree-slsr -ftree-sra @gol
414 -ftree-switch-conversion -ftree-tail-merge -ftree-ter @gol
415 -ftree-vectorize -ftree-vrp -funconstrained-commons @gol
416 -funit-at-a-time -funroll-all-loops -funroll-loops @gol
417 -funsafe-loop-optimizations -funsafe-math-optimizations -funswitch-loops @gol
418 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt @gol
419 -fweb -fwhole-program -fwpa -fuse-linker-plugin @gol
420 --param @var{name}=@var{value}
421 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og}
423 @item Program Instrumentation Options
424 @xref{Instrumentation Options,,Program Instrumentation Options}.
425 @gccoptlist{-p -pg -fprofile-arcs --coverage -ftest-coverage @gol
426 -fprofile-dir=@var{path} -fprofile-generate -fprofile-generate=@var{path} @gol
427 -fsanitize=@var{style} -fsanitize-recover -fsanitize-recover=@var{style} @gol
428 -fasan-shadow-offset=@var{number} -fsanitize-sections=@var{s1},@var{s2},... @gol
429 -fsanitize-undefined-trap-on-error -fbounds-check @gol
430 -fcheck-pointer-bounds -fchkp-check-incomplete-type @gol
431 -fchkp-first-field-has-own-bounds -fchkp-narrow-bounds @gol
432 -fchkp-narrow-to-innermost-array -fchkp-optimize @gol
433 -fchkp-use-fast-string-functions -fchkp-use-nochk-string-functions @gol
434 -fchkp-use-static-bounds -fchkp-use-static-const-bounds @gol
435 -fchkp-treat-zero-dynamic-size-as-infinite -fchkp-check-read @gol
436 -fchkp-check-read -fchkp-check-write -fchkp-store-bounds @gol
437 -fchkp-instrument-calls -fchkp-instrument-marked-only @gol
438 -fchkp-use-wrappers @gol
439 -fstack-protector -fstack-protector-all -fstack-protector-strong @gol
440 -fstack-protector-explicit -fstack-check @gol
441 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
442 -fno-stack-limit -fsplit-stack @gol
443 -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]} @gol
444 -fvtv-counts -fvtv-debug @gol
445 -finstrument-functions @gol
446 -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{} @gol
447 -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}}
449 @item Preprocessor Options
450 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
451 @gccoptlist{-A@var{question}=@var{answer} @gol
452 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
453 -C -dD -dI -dM -dN @gol
454 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
455 -idirafter @var{dir} @gol
456 -include @var{file} -imacros @var{file} @gol
457 -iprefix @var{file} -iwithprefix @var{dir} @gol
458 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
459 -imultilib @var{dir} -isysroot @var{dir} @gol
460 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
461 -P -fdebug-cpp -ftrack-macro-expansion -fworking-directory @gol
462 -remap -trigraphs -undef -U@var{macro} @gol
463 -Wp,@var{option} -Xpreprocessor @var{option} -no-integrated-cpp}
465 @item Assembler Option
466 @xref{Assembler Options,,Passing Options to the Assembler}.
467 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
470 @xref{Link Options,,Options for Linking}.
471 @gccoptlist{@var{object-file-name} -fuse-ld=@var{linker} -l@var{library} @gol
472 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
473 -s -static -static-libgcc -static-libstdc++ @gol
474 -static-libasan -static-libtsan -static-liblsan -static-libubsan @gol
475 -static-libmpx -static-libmpxwrappers @gol
476 -shared -shared-libgcc -symbolic @gol
477 -T @var{script} -Wl,@var{option} -Xlinker @var{option} @gol
478 -u @var{symbol} -z @var{keyword}}
480 @item Directory Options
481 @xref{Directory Options,,Options for Directory Search}.
482 @gccoptlist{-B@var{prefix} -I@var{dir} -iplugindir=@var{dir} @gol
483 -iquote@var{dir} -L@var{dir} -no-canonical-prefixes -I- @gol
484 --sysroot=@var{dir} --no-sysroot-suffix}
486 @item Code Generation Options
487 @xref{Code Gen Options,,Options for Code Generation Conventions}.
488 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
489 -ffixed-@var{reg} -fexceptions @gol
490 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables @gol
491 -fasynchronous-unwind-tables @gol
493 -finhibit-size-directive -fno-common -fno-ident @gol
494 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt @gol
495 -fno-jump-tables @gol
496 -frecord-gcc-switches @gol
497 -freg-struct-return -fshort-enums -fshort-wchar @gol
498 -fverbose-asm -fpack-struct[=@var{n}] @gol
499 -fleading-underscore -ftls-model=@var{model} @gol
500 -fstack-reuse=@var{reuse_level} @gol
502 -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]} @gol
503 -fstrict-volatile-bitfields -fsync-libcalls}
505 @item Developer Options
506 @xref{Developer Options,,GCC Developer Options}.
507 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
508 -fchecking -fchecking=@var{n} -fdbg-cnt-list @gol
509 -fdbg-cnt=@var{counter-value-list} @gol
510 -fdisable-ipa-@var{pass_name} @gol
511 -fdisable-rtl-@var{pass_name} @gol
512 -fdisable-rtl-@var{pass-name}=@var{range-list} @gol
513 -fdisable-tree-@var{pass_name} @gol
514 -fdisable-tree-@var{pass-name}=@var{range-list} @gol
515 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links @gol
516 -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
517 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
518 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline @gol
520 -fdump-rtl-@var{pass} -fdump-rtl-@var{pass}=@var{filename} @gol
521 -fdump-statistics @gol
523 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
524 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
525 -fdump-tree-cfg -fdump-tree-alias @gol
527 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
528 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
529 -fdump-tree-gimple@r{[}-raw@r{]} @gol
530 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
531 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
532 -fdump-tree-phiprop@r{[}-@var{n}@r{]} @gol
533 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
534 -fdump-tree-backprop@r{[}-@var{n}@r{]} @gol
535 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
536 -fdump-tree-nrv -fdump-tree-vect @gol
537 -fdump-tree-sink @gol
538 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
539 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
540 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
541 -fdump-tree-vtable-verify @gol
542 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
543 -fdump-tree-split-paths@r{[}-@var{n}@r{]} @gol
544 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
545 -fdump-final-insns=@var{file} @gol
546 -fcompare-debug@r{[}=@var{opts}@r{]} -fcompare-debug-second @gol
547 -fenable-@var{kind}-@var{pass} @gol
548 -fenable-@var{kind}-@var{pass}=@var{range-list} @gol
549 -fira-verbose=@var{n} @gol
550 -flto-report -flto-report-wpa -fmem-report-wpa @gol
551 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report @gol
552 -fopt-info -fopt-info-@var{options}@r{[}=@var{file}@r{]} @gol
553 -fprofile-report @gol
554 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
555 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose @gol
556 -fstats -fstack-usage -ftime-report @gol
557 -fvar-tracking-assignments-toggle -gtoggle @gol
558 -print-file-name=@var{library} -print-libgcc-file-name @gol
559 -print-multi-directory -print-multi-lib -print-multi-os-directory @gol
560 -print-prog-name=@var{program} -print-search-dirs -Q @gol
561 -print-sysroot -print-sysroot-headers-suffix @gol
562 -save-temps -save-temps=cwd -save-temps=obj -time@r{[}=@var{file}@r{]}}
564 @item Machine-Dependent Options
565 @xref{Submodel Options,,Machine-Dependent Options}.
566 @c This list is ordered alphanumerically by subsection name.
567 @c Try and put the significant identifier (CPU or system) first,
568 @c so users have a clue at guessing where the ones they want will be.
570 @emph{AArch64 Options}
571 @gccoptlist{-mabi=@var{name} -mbig-endian -mlittle-endian @gol
572 -mgeneral-regs-only @gol
573 -mcmodel=tiny -mcmodel=small -mcmodel=large @gol
575 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
576 -mtls-dialect=desc -mtls-dialect=traditional @gol
577 -mtls-size=@var{size} @gol
578 -mfix-cortex-a53-835769 -mno-fix-cortex-a53-835769 @gol
579 -mfix-cortex-a53-843419 -mno-fix-cortex-a53-843419 @gol
580 -mlow-precision-recip-sqrt -mno-low-precision-recip-sqrt@gol
581 -mlow-precision-sqrt -mno-low-precision-sqrt@gol
582 -mlow-precision-div -mno-low-precision-div @gol
583 -march=@var{name} -mcpu=@var{name} -mtune=@var{name}}
585 @emph{Adapteva Epiphany Options}
586 @gccoptlist{-mhalf-reg-file -mprefer-short-insn-regs @gol
587 -mbranch-cost=@var{num} -mcmove -mnops=@var{num} -msoft-cmpsf @gol
588 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=@var{num} @gol
589 -mround-nearest -mlong-calls -mshort-calls -msmall16 @gol
590 -mfp-mode=@var{mode} -mvect-double -max-vect-align=@var{num} @gol
591 -msplit-vecmove-early -m1reg-@var{reg}}
594 @gccoptlist{-mbarrel-shifter @gol
595 -mcpu=@var{cpu} -mA6 -mARC600 -mA7 -mARC700 @gol
596 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr @gol
597 -mea -mno-mpy -mmul32x16 -mmul64 -matomic @gol
598 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap @gol
599 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape @gol
600 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof @gol
601 -mlong-calls -mmedium-calls -msdata @gol
602 -mucb-mcount -mvolatile-cache -mtp-regno=@var{regno} @gol
603 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc @gol
604 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi @gol
605 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none @gol
606 -mlra-priority-compact mlra-priority-noncompact -mno-millicode @gol
607 -mmixed-code -mq-class -mRcq -mRcw -msize-level=@var{level} @gol
608 -mtune=@var{cpu} -mmultcost=@var{num} @gol
609 -munalign-prob-threshold=@var{probability} -mmpy-option=@var{multo} @gol
610 -mdiv-rem -mcode-density -mll64 -mfpu=@var{fpu}}
613 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
614 -mabi=@var{name} @gol
615 -mapcs-stack-check -mno-apcs-stack-check @gol
616 -mapcs-float -mno-apcs-float @gol
617 -mapcs-reentrant -mno-apcs-reentrant @gol
618 -msched-prolog -mno-sched-prolog @gol
619 -mlittle-endian -mbig-endian @gol
620 -mfloat-abi=@var{name} @gol
621 -mfp16-format=@var{name}
622 -mthumb-interwork -mno-thumb-interwork @gol
623 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
624 -mtune=@var{name} -mprint-tune-info @gol
625 -mstructure-size-boundary=@var{n} @gol
626 -mabort-on-noreturn @gol
627 -mlong-calls -mno-long-calls @gol
628 -msingle-pic-base -mno-single-pic-base @gol
629 -mpic-register=@var{reg} @gol
630 -mnop-fun-dllimport @gol
631 -mpoke-function-name @gol
633 -mtpcs-frame -mtpcs-leaf-frame @gol
634 -mcaller-super-interworking -mcallee-super-interworking @gol
635 -mtp=@var{name} -mtls-dialect=@var{dialect} @gol
636 -mword-relocations @gol
637 -mfix-cortex-m3-ldrd @gol
638 -munaligned-access @gol
639 -mneon-for-64bits @gol
640 -mslow-flash-data @gol
641 -masm-syntax-unified @gol
645 @gccoptlist{-mmcu=@var{mcu} -maccumulate-args -mbranch-cost=@var{cost} @gol
646 -mcall-prologues -mint8 -mn_flash=@var{size} -mno-interrupts @gol
647 -mrelax -mrmw -mstrict-X -mtiny-stack -nodevicelib -Waddr-space-convert @gol
650 @emph{Blackfin Options}
651 @gccoptlist{-mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]} @gol
652 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
653 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
654 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
655 -mno-id-shared-library -mshared-library-id=@var{n} @gol
656 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
657 -msep-data -mno-sep-data -mlong-calls -mno-long-calls @gol
658 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram @gol
662 @gccoptlist{-mbig-endian -mlittle-endian -march=@var{cpu} @gol
663 -msim -msdata=@var{sdata-type}}
666 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
667 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
668 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
669 -mstack-align -mdata-align -mconst-align @gol
670 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
671 -melf -maout -melinux -mlinux -sim -sim2 @gol
672 -mmul-bug-workaround -mno-mul-bug-workaround}
675 @gccoptlist{-mmac @gol
676 -mcr16cplus -mcr16c @gol
677 -msim -mint32 -mbit-ops
678 -mdata-model=@var{model}}
680 @emph{Darwin Options}
681 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
682 -arch_only -bind_at_load -bundle -bundle_loader @gol
683 -client_name -compatibility_version -current_version @gol
685 -dependency-file -dylib_file -dylinker_install_name @gol
686 -dynamic -dynamiclib -exported_symbols_list @gol
687 -filelist -flat_namespace -force_cpusubtype_ALL @gol
688 -force_flat_namespace -headerpad_max_install_names @gol
690 -image_base -init -install_name -keep_private_externs @gol
691 -multi_module -multiply_defined -multiply_defined_unused @gol
692 -noall_load -no_dead_strip_inits_and_terms @gol
693 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
694 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
695 -private_bundle -read_only_relocs -sectalign @gol
696 -sectobjectsymbols -whyload -seg1addr @gol
697 -sectcreate -sectobjectsymbols -sectorder @gol
698 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
699 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
700 -segprot -segs_read_only_addr -segs_read_write_addr @gol
701 -single_module -static -sub_library -sub_umbrella @gol
702 -twolevel_namespace -umbrella -undefined @gol
703 -unexported_symbols_list -weak_reference_mismatches @gol
704 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
705 -mkernel -mone-byte-bool}
707 @emph{DEC Alpha Options}
708 @gccoptlist{-mno-fp-regs -msoft-float @gol
709 -mieee -mieee-with-inexact -mieee-conformant @gol
710 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
711 -mtrap-precision=@var{mode} -mbuild-constants @gol
712 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
713 -mbwx -mmax -mfix -mcix @gol
714 -mfloat-vax -mfloat-ieee @gol
715 -mexplicit-relocs -msmall-data -mlarge-data @gol
716 -msmall-text -mlarge-text @gol
717 -mmemory-latency=@var{time}}
720 @gccoptlist{-msmall-model -mno-lsim}
723 @gccoptlist{-msim -mlra -mnodiv}
726 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
727 -mhard-float -msoft-float @gol
728 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
729 -mdouble -mno-double @gol
730 -mmedia -mno-media -mmuladd -mno-muladd @gol
731 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
732 -mlinked-fp -mlong-calls -malign-labels @gol
733 -mlibrary-pic -macc-4 -macc-8 @gol
734 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
735 -moptimize-membar -mno-optimize-membar @gol
736 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
737 -mvliw-branch -mno-vliw-branch @gol
738 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
739 -mno-nested-cond-exec -mtomcat-stats @gol
743 @emph{GNU/Linux Options}
744 @gccoptlist{-mglibc -muclibc -mmusl -mbionic -mandroid @gol
745 -tno-android-cc -tno-android-ld}
747 @emph{H8/300 Options}
748 @gccoptlist{-mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300}
751 @gccoptlist{-march=@var{architecture-type} @gol
752 -mdisable-fpregs -mdisable-indexing @gol
753 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
754 -mfixed-range=@var{register-range} @gol
755 -mjump-in-delay -mlinker-opt -mlong-calls @gol
756 -mlong-load-store -mno-disable-fpregs @gol
757 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
758 -mno-jump-in-delay -mno-long-load-store @gol
759 -mno-portable-runtime -mno-soft-float @gol
760 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
761 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
762 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
763 -munix=@var{unix-std} -nolibdld -static -threads}
766 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
767 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata @gol
768 -mconstant-gp -mauto-pic -mfused-madd @gol
769 -minline-float-divide-min-latency @gol
770 -minline-float-divide-max-throughput @gol
771 -mno-inline-float-divide @gol
772 -minline-int-divide-min-latency @gol
773 -minline-int-divide-max-throughput @gol
774 -mno-inline-int-divide @gol
775 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
776 -mno-inline-sqrt @gol
777 -mdwarf2-asm -mearly-stop-bits @gol
778 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
779 -mtune=@var{cpu-type} -milp32 -mlp64 @gol
780 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec @gol
781 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
782 -msched-spec-ldc -msched-spec-control-ldc @gol
783 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns @gol
784 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path @gol
785 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost @gol
786 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=@var{max-insns}}
789 @gccoptlist{-mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled @gol
790 -msign-extend-enabled -muser-enabled}
792 @emph{M32R/D Options}
793 @gccoptlist{-m32r2 -m32rx -m32r @gol
795 -malign-loops -mno-align-loops @gol
796 -missue-rate=@var{number} @gol
797 -mbranch-cost=@var{number} @gol
798 -mmodel=@var{code-size-model-type} @gol
799 -msdata=@var{sdata-type} @gol
800 -mno-flush-func -mflush-func=@var{name} @gol
801 -mno-flush-trap -mflush-trap=@var{number} @gol
805 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
807 @emph{M680x0 Options}
808 @gccoptlist{-march=@var{arch} -mcpu=@var{cpu} -mtune=@var{tune} @gol
809 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
810 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 @gol
811 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 @gol
812 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort @gol
813 -mno-short -mhard-float -m68881 -msoft-float -mpcrel @gol
814 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
815 -mshared-library-id=n -mid-shared-library -mno-id-shared-library @gol
819 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
820 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
821 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
822 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
823 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
826 @gccoptlist{-mabsdiff -mall-opts -maverage -mbased=@var{n} -mbitops @gol
827 -mc=@var{n} -mclip -mconfig=@var{name} -mcop -mcop32 -mcop64 -mivc2 @gol
828 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax @gol
829 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf @gol
832 @emph{MicroBlaze Options}
833 @gccoptlist{-msoft-float -mhard-float -msmall-divides -mcpu=@var{cpu} @gol
834 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift @gol
835 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss @gol
836 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt @gol
837 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-@var{app-model}}
840 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
841 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 @gol
842 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 @gol
843 -mips16 -mno-mips16 -mflip-mips16 @gol
844 -minterlink-compressed -mno-interlink-compressed @gol
845 -minterlink-mips16 -mno-interlink-mips16 @gol
846 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
847 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot @gol
848 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float @gol
849 -mno-float -msingle-float -mdouble-float @gol
850 -modd-spreg -mno-odd-spreg @gol
851 -mabs=@var{mode} -mnan=@var{encoding} @gol
852 -mdsp -mno-dsp -mdspr2 -mno-dspr2 @gol
855 -mvirt -mno-virt @gol
857 -mmicromips -mno-micromips @gol
859 -mfpu=@var{fpu-type} @gol
860 -msmartmips -mno-smartmips @gol
861 -mpaired-single -mno-paired-single -mdmx -mno-mdmx @gol
862 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc @gol
863 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
864 -G@var{num} -mlocal-sdata -mno-local-sdata @gol
865 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt @gol
866 -membedded-data -mno-embedded-data @gol
867 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
868 -mcode-readable=@var{setting} @gol
869 -msplit-addresses -mno-split-addresses @gol
870 -mexplicit-relocs -mno-explicit-relocs @gol
871 -mcheck-zero-division -mno-check-zero-division @gol
872 -mdivide-traps -mdivide-breaks @gol
873 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
874 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp @gol
875 -mfix-24k -mno-fix-24k @gol
876 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
877 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 @gol
878 -mfix-vr4120 -mno-fix-vr4120 @gol
879 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 @gol
880 -mflush-func=@var{func} -mno-flush-func @gol
881 -mbranch-cost=@var{num} -mbranch-likely -mno-branch-likely @gol
882 -mcompact-branches=@var{policy} @gol
883 -mfp-exceptions -mno-fp-exceptions @gol
884 -mvr4130-align -mno-vr4130-align -msynci -mno-synci @gol
885 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address @gol
886 -mframe-header-opt -mno-frame-header-opt}
889 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
890 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
891 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
892 -mno-base-addresses -msingle-exit -mno-single-exit}
894 @emph{MN10300 Options}
895 @gccoptlist{-mmult-bug -mno-mult-bug @gol
896 -mno-am33 -mam33 -mam33-2 -mam34 @gol
897 -mtune=@var{cpu-type} @gol
898 -mreturn-pointer-on-d0 @gol
899 -mno-crt0 -mrelax -mliw -msetlb}
902 @gccoptlist{-meb -mel -mmul.x -mno-crt0}
904 @emph{MSP430 Options}
905 @gccoptlist{-msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax @gol
907 -mcode-region= -mdata-region= @gol
908 -msilicon-errata= -msilicon-errata-warn= @gol
912 @gccoptlist{-mbig-endian -mlittle-endian @gol
913 -mreduced-regs -mfull-regs @gol
914 -mcmov -mno-cmov @gol
915 -mperf-ext -mno-perf-ext @gol
916 -mv3push -mno-v3push @gol
917 -m16bit -mno-16bit @gol
918 -misr-vector-size=@var{num} @gol
919 -mcache-block-size=@var{num} @gol
920 -march=@var{arch} @gol
921 -mcmodel=@var{code-model} @gol
924 @emph{Nios II Options}
925 @gccoptlist{-G @var{num} -mgpopt=@var{option} -mgpopt -mno-gpopt @gol
927 -mno-bypass-cache -mbypass-cache @gol
928 -mno-cache-volatile -mcache-volatile @gol
929 -mno-fast-sw-div -mfast-sw-div @gol
930 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div @gol
931 -mcustom-@var{insn}=@var{N} -mno-custom-@var{insn} @gol
932 -mcustom-fpu-cfg=@var{name} @gol
933 -mhal -msmallc -msys-crt0=@var{name} -msys-lib=@var{name} @gol
934 -march=@var{arch} -mbmx -mno-bmx -mcdx -mno-cdx}
936 @emph{Nvidia PTX Options}
937 @gccoptlist{-m32 -m64 -mmainkernel -moptimize}
939 @emph{PDP-11 Options}
940 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
941 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
942 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
943 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
944 -mbranch-expensive -mbranch-cheap @gol
945 -munix-asm -mdec-asm}
947 @emph{picoChip Options}
948 @gccoptlist{-mae=@var{ae_type} -mvliw-lookahead=@var{N} @gol
949 -msymbol-as-address -mno-inefficient-warnings}
951 @emph{PowerPC Options}
952 See RS/6000 and PowerPC Options.
955 @gccoptlist{-msim -mmul=none -mmul=g13 -mmul=g14 -mallregs @gol
956 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 @gol
957 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts}
959 @emph{RS/6000 and PowerPC Options}
960 @gccoptlist{-mcpu=@var{cpu-type} @gol
961 -mtune=@var{cpu-type} @gol
962 -mcmodel=@var{code-model} @gol
964 -maltivec -mno-altivec @gol
965 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
966 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
967 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd @gol
968 -mfprnd -mno-fprnd @gol
969 -mcmpb -mno-cmpb -mmfpgpr -mno-mfpgpr -mhard-dfp -mno-hard-dfp @gol
970 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
971 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
972 -malign-power -malign-natural @gol
973 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
974 -msingle-float -mdouble-float -msimple-fpu @gol
975 -mstring -mno-string -mupdate -mno-update @gol
976 -mavoid-indexed-addresses -mno-avoid-indexed-addresses @gol
977 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
978 -mstrict-align -mno-strict-align -mrelocatable @gol
979 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
980 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
981 -mdynamic-no-pic -maltivec -mswdiv -msingle-pic-base @gol
982 -mprioritize-restricted-insns=@var{priority} @gol
983 -msched-costly-dep=@var{dependence_type} @gol
984 -minsert-sched-nops=@var{scheme} @gol
985 -mcall-sysv -mcall-netbsd @gol
986 -maix-struct-return -msvr4-struct-return @gol
987 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
988 -mblock-move-inline-limit=@var{num} @gol
989 -misel -mno-isel @gol
990 -misel=yes -misel=no @gol
992 -mspe=yes -mspe=no @gol
994 -mgen-cell-microcode -mwarn-cell-microcode @gol
995 -mvrsave -mno-vrsave @gol
996 -mmulhw -mno-mulhw @gol
997 -mdlmzb -mno-dlmzb @gol
998 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
999 -mprototype -mno-prototype @gol
1000 -msim -mmvme -mads -myellowknife -memb -msdata @gol
1001 -msdata=@var{opt} -mvxworks -G @var{num} -pthread @gol
1002 -mrecip -mrecip=@var{opt} -mno-recip -mrecip-precision @gol
1003 -mno-recip-precision @gol
1004 -mveclibabi=@var{type} -mfriz -mno-friz @gol
1005 -mpointers-to-nested-functions -mno-pointers-to-nested-functions @gol
1006 -msave-toc-indirect -mno-save-toc-indirect @gol
1007 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector @gol
1008 -mcrypto -mno-crypto -mhtm -mno-htm -mdirect-move -mno-direct-move @gol
1009 -mquad-memory -mno-quad-memory @gol
1010 -mquad-memory-atomic -mno-quad-memory-atomic @gol
1011 -mcompat-align-parm -mno-compat-align-parm @gol
1012 -mupper-regs-df -mno-upper-regs-df -mupper-regs-sf -mno-upper-regs-sf @gol
1013 -mupper-regs-di -mno-upper-regs-di @gol
1014 -mupper-regs -mno-upper-regs -mmodulo -mno-modulo @gol
1015 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware @gol
1016 -mpower9-fusion -mno-mpower9-fusion -mpower9-vector -mno-power9-vector @gol
1017 -mpower9-dform -mno-power9-dform -mlra -mno-lra}
1020 @gccoptlist{-m64bit-doubles -m32bit-doubles -fpu -nofpu@gol
1022 -mbig-endian-data -mlittle-endian-data @gol
1025 -mas100-syntax -mno-as100-syntax@gol
1027 -mmax-constant-size=@gol
1030 -mallow-string-insns -mno-allow-string-insns@gol
1032 -mno-warn-multiple-fast-interrupts@gol
1033 -msave-acc-in-interrupts}
1035 @emph{S/390 and zSeries Options}
1036 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1037 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp @gol
1038 -mlong-double-64 -mlong-double-128 @gol
1039 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
1040 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
1041 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
1042 -mhtm -mvx -mzvector @gol
1043 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
1044 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard @gol
1045 -mhotpatch=@var{halfwords},@var{halfwords}}
1047 @emph{Score Options}
1048 @gccoptlist{-meb -mel @gol
1052 -mscore5 -mscore5u -mscore7 -mscore7d}
1055 @gccoptlist{-m1 -m2 -m2e @gol
1056 -m2a-nofpu -m2a-single-only -m2a-single -m2a @gol
1058 -m4-nofpu -m4-single-only -m4-single -m4 @gol
1059 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
1060 -mb -ml -mdalign -mrelax @gol
1061 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave @gol
1062 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct @gol
1063 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
1064 -mdivsi3_libfunc=@var{name} -mfixed-range=@var{register-range} @gol
1065 -maccumulate-outgoing-args @gol
1066 -matomic-model=@var{atomic-model} @gol
1067 -mbranch-cost=@var{num} -mzdcbranch -mno-zdcbranch @gol
1068 -mcbranch-force-delay-slot @gol
1069 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra @gol
1070 -mpretend-cmove -mtas}
1072 @emph{Solaris 2 Options}
1073 @gccoptlist{-mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text @gol
1076 @emph{SPARC Options}
1077 @gccoptlist{-mcpu=@var{cpu-type} @gol
1078 -mtune=@var{cpu-type} @gol
1079 -mcmodel=@var{code-model} @gol
1080 -mmemory-model=@var{mem-model} @gol
1081 -m32 -m64 -mapp-regs -mno-app-regs @gol
1082 -mfaster-structs -mno-faster-structs -mflat -mno-flat @gol
1083 -mfpu -mno-fpu -mhard-float -msoft-float @gol
1084 -mhard-quad-float -msoft-quad-float @gol
1085 -mstack-bias -mno-stack-bias @gol
1086 -mstd-struct-return -mno-std-struct-return @gol
1087 -munaligned-doubles -mno-unaligned-doubles @gol
1088 -muser-mode -mno-user-mode @gol
1089 -mv8plus -mno-v8plus -mvis -mno-vis @gol
1090 -mvis2 -mno-vis2 -mvis3 -mno-vis3 @gol
1091 -mcbcond -mno-cbcond @gol
1092 -mfmaf -mno-fmaf -mpopc -mno-popc @gol
1093 -mfix-at697f -mfix-ut699}
1096 @gccoptlist{-mwarn-reloc -merror-reloc @gol
1097 -msafe-dma -munsafe-dma @gol
1099 -msmall-mem -mlarge-mem -mstdmain @gol
1100 -mfixed-range=@var{register-range} @gol
1102 -maddress-space-conversion -mno-address-space-conversion @gol
1103 -mcache-size=@var{cache-size} @gol
1104 -matomic-updates -mno-atomic-updates}
1106 @emph{System V Options}
1107 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
1109 @emph{TILE-Gx Options}
1110 @gccoptlist{-mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian @gol
1111 -mcmodel=@var{code-model}}
1113 @emph{TILEPro Options}
1114 @gccoptlist{-mcpu=@var{cpu} -m32}
1117 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
1118 -mprolog-function -mno-prolog-function -mspace @gol
1119 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
1120 -mapp-regs -mno-app-regs @gol
1121 -mdisable-callt -mno-disable-callt @gol
1122 -mv850e2v3 -mv850e2 -mv850e1 -mv850es @gol
1123 -mv850e -mv850 -mv850e3v5 @gol
1134 @gccoptlist{-mg -mgnu -munix}
1136 @emph{Visium Options}
1137 @gccoptlist{-mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float @gol
1138 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} -msv-mode -muser-mode}
1141 @gccoptlist{-mvms-return-codes -mdebug-main=@var{prefix} -mmalloc64 @gol
1142 -mpointer-size=@var{size}}
1144 @emph{VxWorks Options}
1145 @gccoptlist{-mrtp -non-static -Bstatic -Bdynamic @gol
1146 -Xbind-lazy -Xbind-now}
1149 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
1150 -mtune-ctrl=@var{feature-list} -mdump-tune-features -mno-default @gol
1151 -mfpmath=@var{unit} @gol
1152 -masm=@var{dialect} -mno-fancy-math-387 @gol
1153 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float @gol
1154 -mno-wide-multiply -mrtd -malign-double @gol
1155 -mpreferred-stack-boundary=@var{num} @gol
1156 -mincoming-stack-boundary=@var{num} @gol
1157 -mcld -mcx16 -msahf -mmovbe -mcrc32 @gol
1158 -mrecip -mrecip=@var{opt} @gol
1159 -mvzeroupper -mprefer-avx128 @gol
1160 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx @gol
1161 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl @gol
1162 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes @gol
1163 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma @gol
1164 -mprefetchwt1 -mclflushopt -mxsavec -mxsaves @gol
1165 -msse4a -m3dnow -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop -mlzcnt @gol
1166 -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mlwp -mmpx -mmwaitx -mclzero
1167 -mpku -mthreads @gol
1168 -mms-bitfields -mno-align-stringops -minline-all-stringops @gol
1169 -minline-stringops-dynamically -mstringop-strategy=@var{alg} @gol
1170 -mmemcpy-strategy=@var{strategy} -mmemset-strategy=@var{strategy} @gol
1171 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
1172 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 @gol
1173 -mregparm=@var{num} -msseregparm @gol
1174 -mveclibabi=@var{type} -mvect8-ret-in-mem @gol
1175 -mpc32 -mpc64 -mpc80 -mstackrealign @gol
1176 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
1177 -mcmodel=@var{code-model} -mabi=@var{name} -maddress-mode=@var{mode} @gol
1178 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=@var{num} @gol
1179 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv @gol
1180 -mavx256-split-unaligned-load -mavx256-split-unaligned-store @gol
1181 -malign-data=@var{type} -mstack-protector-guard=@var{guard} @gol
1182 -mmitigate-rop -mgeneral-regs-only}
1184 @emph{x86 Windows Options}
1185 @gccoptlist{-mconsole -mcygwin -mno-cygwin -mdll @gol
1186 -mnop-fun-dllimport -mthread @gol
1187 -municode -mwin32 -mwindows -fno-set-stack-executable}
1189 @emph{Xstormy16 Options}
1192 @emph{Xtensa Options}
1193 @gccoptlist{-mconst16 -mno-const16 @gol
1194 -mfused-madd -mno-fused-madd @gol
1196 -mserialize-volatile -mno-serialize-volatile @gol
1197 -mtext-section-literals -mno-text-section-literals @gol
1198 -mauto-litpools -mno-auto-litpools @gol
1199 -mtarget-align -mno-target-align @gol
1200 -mlongcalls -mno-longcalls}
1202 @emph{zSeries Options}
1203 See S/390 and zSeries Options.
1207 @node Overall Options
1208 @section Options Controlling the Kind of Output
1210 Compilation can involve up to four stages: preprocessing, compilation
1211 proper, assembly and linking, always in that order. GCC is capable of
1212 preprocessing and compiling several files either into several
1213 assembler input files, or into one assembler input file; then each
1214 assembler input file produces an object file, and linking combines all
1215 the object files (those newly compiled, and those specified as input)
1216 into an executable file.
1218 @cindex file name suffix
1219 For any given input file, the file name suffix determines what kind of
1220 compilation is done:
1224 C source code that must be preprocessed.
1227 C source code that should not be preprocessed.
1230 C++ source code that should not be preprocessed.
1233 Objective-C source code. Note that you must link with the @file{libobjc}
1234 library to make an Objective-C program work.
1237 Objective-C source code that should not be preprocessed.
1241 Objective-C++ source code. Note that you must link with the @file{libobjc}
1242 library to make an Objective-C++ program work. Note that @samp{.M} refers
1243 to a literal capital M@.
1245 @item @var{file}.mii
1246 Objective-C++ source code that should not be preprocessed.
1249 C, C++, Objective-C or Objective-C++ header file to be turned into a
1250 precompiled header (default), or C, C++ header file to be turned into an
1251 Ada spec (via the @option{-fdump-ada-spec} switch).
1254 @itemx @var{file}.cp
1255 @itemx @var{file}.cxx
1256 @itemx @var{file}.cpp
1257 @itemx @var{file}.CPP
1258 @itemx @var{file}.c++
1260 C++ source code that must be preprocessed. Note that in @samp{.cxx},
1261 the last two letters must both be literally @samp{x}. Likewise,
1262 @samp{.C} refers to a literal capital C@.
1266 Objective-C++ source code that must be preprocessed.
1268 @item @var{file}.mii
1269 Objective-C++ source code that should not be preprocessed.
1273 @itemx @var{file}.hp
1274 @itemx @var{file}.hxx
1275 @itemx @var{file}.hpp
1276 @itemx @var{file}.HPP
1277 @itemx @var{file}.h++
1278 @itemx @var{file}.tcc
1279 C++ header file to be turned into a precompiled header or Ada spec.
1282 @itemx @var{file}.for
1283 @itemx @var{file}.ftn
1284 Fixed form Fortran source code that should not be preprocessed.
1287 @itemx @var{file}.FOR
1288 @itemx @var{file}.fpp
1289 @itemx @var{file}.FPP
1290 @itemx @var{file}.FTN
1291 Fixed form Fortran source code that must be preprocessed (with the traditional
1294 @item @var{file}.f90
1295 @itemx @var{file}.f95
1296 @itemx @var{file}.f03
1297 @itemx @var{file}.f08
1298 Free form Fortran source code that should not be preprocessed.
1300 @item @var{file}.F90
1301 @itemx @var{file}.F95
1302 @itemx @var{file}.F03
1303 @itemx @var{file}.F08
1304 Free form Fortran source code that must be preprocessed (with the
1305 traditional preprocessor).
1310 @c FIXME: Descriptions of Java file types.
1316 @item @var{file}.ads
1317 Ada source code file that contains a library unit declaration (a
1318 declaration of a package, subprogram, or generic, or a generic
1319 instantiation), or a library unit renaming declaration (a package,
1320 generic, or subprogram renaming declaration). Such files are also
1323 @item @var{file}.adb
1324 Ada source code file containing a library unit body (a subprogram or
1325 package body). Such files are also called @dfn{bodies}.
1327 @c GCC also knows about some suffixes for languages not yet included:
1338 @itemx @var{file}.sx
1339 Assembler code that must be preprocessed.
1342 An object file to be fed straight into linking.
1343 Any file name with no recognized suffix is treated this way.
1347 You can specify the input language explicitly with the @option{-x} option:
1350 @item -x @var{language}
1351 Specify explicitly the @var{language} for the following input files
1352 (rather than letting the compiler choose a default based on the file
1353 name suffix). This option applies to all following input files until
1354 the next @option{-x} option. Possible values for @var{language} are:
1356 c c-header cpp-output
1357 c++ c++-header c++-cpp-output
1358 objective-c objective-c-header objective-c-cpp-output
1359 objective-c++ objective-c++-header objective-c++-cpp-output
1360 assembler assembler-with-cpp
1362 f77 f77-cpp-input f95 f95-cpp-input
1368 Turn off any specification of a language, so that subsequent files are
1369 handled according to their file name suffixes (as they are if @option{-x}
1370 has not been used at all).
1373 If you only want some of the stages of compilation, you can use
1374 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
1375 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
1376 @command{gcc} is to stop. Note that some combinations (for example,
1377 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
1382 Compile or assemble the source files, but do not link. The linking
1383 stage simply is not done. The ultimate output is in the form of an
1384 object file for each source file.
1386 By default, the object file name for a source file is made by replacing
1387 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1389 Unrecognized input files, not requiring compilation or assembly, are
1394 Stop after the stage of compilation proper; do not assemble. The output
1395 is in the form of an assembler code file for each non-assembler input
1398 By default, the assembler file name for a source file is made by
1399 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1401 Input files that don't require compilation are ignored.
1405 Stop after the preprocessing stage; do not run the compiler proper. The
1406 output is in the form of preprocessed source code, which is sent to the
1409 Input files that don't require preprocessing are ignored.
1411 @cindex output file option
1414 Place output in file @var{file}. This applies to whatever
1415 sort of output is being produced, whether it be an executable file,
1416 an object file, an assembler file or preprocessed C code.
1418 If @option{-o} is not specified, the default is to put an executable
1419 file in @file{a.out}, the object file for
1420 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1421 assembler file in @file{@var{source}.s}, a precompiled header file in
1422 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1427 Print (on standard error output) the commands executed to run the stages
1428 of compilation. Also print the version number of the compiler driver
1429 program and of the preprocessor and the compiler proper.
1433 Like @option{-v} except the commands are not executed and arguments
1434 are quoted unless they contain only alphanumeric characters or @code{./-_}.
1435 This is useful for shell scripts to capture the driver-generated command lines.
1439 Print (on the standard output) a description of the command-line options
1440 understood by @command{gcc}. If the @option{-v} option is also specified
1441 then @option{--help} is also passed on to the various processes
1442 invoked by @command{gcc}, so that they can display the command-line options
1443 they accept. If the @option{-Wextra} option has also been specified
1444 (prior to the @option{--help} option), then command-line options that
1445 have no documentation associated with them are also displayed.
1448 @opindex target-help
1449 Print (on the standard output) a description of target-specific command-line
1450 options for each tool. For some targets extra target-specific
1451 information may also be printed.
1453 @item --help=@{@var{class}@r{|[}^@r{]}@var{qualifier}@}@r{[},@dots{}@r{]}
1454 Print (on the standard output) a description of the command-line
1455 options understood by the compiler that fit into all specified classes
1456 and qualifiers. These are the supported classes:
1459 @item @samp{optimizers}
1460 Display all of the optimization options supported by the
1463 @item @samp{warnings}
1464 Display all of the options controlling warning messages
1465 produced by the compiler.
1468 Display target-specific options. Unlike the
1469 @option{--target-help} option however, target-specific options of the
1470 linker and assembler are not displayed. This is because those
1471 tools do not currently support the extended @option{--help=} syntax.
1474 Display the values recognized by the @option{--param}
1477 @item @var{language}
1478 Display the options supported for @var{language}, where
1479 @var{language} is the name of one of the languages supported in this
1483 Display the options that are common to all languages.
1486 These are the supported qualifiers:
1489 @item @samp{undocumented}
1490 Display only those options that are undocumented.
1493 Display options taking an argument that appears after an equal
1494 sign in the same continuous piece of text, such as:
1495 @samp{--help=target}.
1497 @item @samp{separate}
1498 Display options taking an argument that appears as a separate word
1499 following the original option, such as: @samp{-o output-file}.
1502 Thus for example to display all the undocumented target-specific
1503 switches supported by the compiler, use:
1506 --help=target,undocumented
1509 The sense of a qualifier can be inverted by prefixing it with the
1510 @samp{^} character, so for example to display all binary warning
1511 options (i.e., ones that are either on or off and that do not take an
1512 argument) that have a description, use:
1515 --help=warnings,^joined,^undocumented
1518 The argument to @option{--help=} should not consist solely of inverted
1521 Combining several classes is possible, although this usually
1522 restricts the output so much that there is nothing to display. One
1523 case where it does work, however, is when one of the classes is
1524 @var{target}. For example, to display all the target-specific
1525 optimization options, use:
1528 --help=target,optimizers
1531 The @option{--help=} option can be repeated on the command line. Each
1532 successive use displays its requested class of options, skipping
1533 those that have already been displayed.
1535 If the @option{-Q} option appears on the command line before the
1536 @option{--help=} option, then the descriptive text displayed by
1537 @option{--help=} is changed. Instead of describing the displayed
1538 options, an indication is given as to whether the option is enabled,
1539 disabled or set to a specific value (assuming that the compiler
1540 knows this at the point where the @option{--help=} option is used).
1542 Here is a truncated example from the ARM port of @command{gcc}:
1545 % gcc -Q -mabi=2 --help=target -c
1546 The following options are target specific:
1548 -mabort-on-noreturn [disabled]
1552 The output is sensitive to the effects of previous command-line
1553 options, so for example it is possible to find out which optimizations
1554 are enabled at @option{-O2} by using:
1557 -Q -O2 --help=optimizers
1560 Alternatively you can discover which binary optimizations are enabled
1561 by @option{-O3} by using:
1564 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts
1565 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts
1566 diff /tmp/O2-opts /tmp/O3-opts | grep enabled
1571 Display the version number and copyrights of the invoked GCC@.
1573 @item -pass-exit-codes
1574 @opindex pass-exit-codes
1575 Normally the @command{gcc} program exits with the code of 1 if any
1576 phase of the compiler returns a non-success return code. If you specify
1577 @option{-pass-exit-codes}, the @command{gcc} program instead returns with
1578 the numerically highest error produced by any phase returning an error
1579 indication. The C, C++, and Fortran front ends return 4 if an internal
1580 compiler error is encountered.
1584 Use pipes rather than temporary files for communication between the
1585 various stages of compilation. This fails to work on some systems where
1586 the assembler is unable to read from a pipe; but the GNU assembler has
1589 @item -specs=@var{file}
1591 Process @var{file} after the compiler reads in the standard @file{specs}
1592 file, in order to override the defaults which the @command{gcc} driver
1593 program uses when determining what switches to pass to @command{cc1},
1594 @command{cc1plus}, @command{as}, @command{ld}, etc. More than one
1595 @option{-specs=@var{file}} can be specified on the command line, and they
1596 are processed in order, from left to right. @xref{Spec Files}, for
1597 information about the format of the @var{file}.
1601 Invoke all subcommands under a wrapper program. The name of the
1602 wrapper program and its parameters are passed as a comma separated
1606 gcc -c t.c -wrapper gdb,--args
1610 This invokes all subprograms of @command{gcc} under
1611 @samp{gdb --args}, thus the invocation of @command{cc1} is
1612 @samp{gdb --args cc1 @dots{}}.
1614 @item -fplugin=@var{name}.so
1616 Load the plugin code in file @var{name}.so, assumed to be a
1617 shared object to be dlopen'd by the compiler. The base name of
1618 the shared object file is used to identify the plugin for the
1619 purposes of argument parsing (See
1620 @option{-fplugin-arg-@var{name}-@var{key}=@var{value}} below).
1621 Each plugin should define the callback functions specified in the
1624 @item -fplugin-arg-@var{name}-@var{key}=@var{value}
1625 @opindex fplugin-arg
1626 Define an argument called @var{key} with a value of @var{value}
1627 for the plugin called @var{name}.
1629 @item -fdump-ada-spec@r{[}-slim@r{]}
1630 @opindex fdump-ada-spec
1631 For C and C++ source and include files, generate corresponding Ada specs.
1632 @xref{Generating Ada Bindings for C and C++ headers,,, gnat_ugn,
1633 GNAT User's Guide}, which provides detailed documentation on this feature.
1635 @item -fada-spec-parent=@var{unit}
1636 @opindex fada-spec-parent
1637 In conjunction with @option{-fdump-ada-spec@r{[}-slim@r{]}} above, generate
1638 Ada specs as child units of parent @var{unit}.
1640 @item -fdump-go-spec=@var{file}
1641 @opindex fdump-go-spec
1642 For input files in any language, generate corresponding Go
1643 declarations in @var{file}. This generates Go @code{const},
1644 @code{type}, @code{var}, and @code{func} declarations which may be a
1645 useful way to start writing a Go interface to code written in some
1648 @include @value{srcdir}/../libiberty/at-file.texi
1652 @section Compiling C++ Programs
1654 @cindex suffixes for C++ source
1655 @cindex C++ source file suffixes
1656 C++ source files conventionally use one of the suffixes @samp{.C},
1657 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1658 @samp{.cxx}; C++ header files often use @samp{.hh}, @samp{.hpp},
1659 @samp{.H}, or (for shared template code) @samp{.tcc}; and
1660 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1661 files with these names and compiles them as C++ programs even if you
1662 call the compiler the same way as for compiling C programs (usually
1663 with the name @command{gcc}).
1667 However, the use of @command{gcc} does not add the C++ library.
1668 @command{g++} is a program that calls GCC and automatically specifies linking
1669 against the C++ library. It treats @samp{.c},
1670 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1671 files unless @option{-x} is used. This program is also useful when
1672 precompiling a C header file with a @samp{.h} extension for use in C++
1673 compilations. On many systems, @command{g++} is also installed with
1674 the name @command{c++}.
1676 @cindex invoking @command{g++}
1677 When you compile C++ programs, you may specify many of the same
1678 command-line options that you use for compiling programs in any
1679 language; or command-line options meaningful for C and related
1680 languages; or options that are meaningful only for C++ programs.
1681 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1682 explanations of options for languages related to C@.
1683 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1684 explanations of options that are meaningful only for C++ programs.
1686 @node C Dialect Options
1687 @section Options Controlling C Dialect
1688 @cindex dialect options
1689 @cindex language dialect options
1690 @cindex options, dialect
1692 The following options control the dialect of C (or languages derived
1693 from C, such as C++, Objective-C and Objective-C++) that the compiler
1697 @cindex ANSI support
1701 In C mode, this is equivalent to @option{-std=c90}. In C++ mode, it is
1702 equivalent to @option{-std=c++98}.
1704 This turns off certain features of GCC that are incompatible with ISO
1705 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1706 such as the @code{asm} and @code{typeof} keywords, and
1707 predefined macros such as @code{unix} and @code{vax} that identify the
1708 type of system you are using. It also enables the undesirable and
1709 rarely used ISO trigraph feature. For the C compiler,
1710 it disables recognition of C++ style @samp{//} comments as well as
1711 the @code{inline} keyword.
1713 The alternate keywords @code{__asm__}, @code{__extension__},
1714 @code{__inline__} and @code{__typeof__} continue to work despite
1715 @option{-ansi}. You would not want to use them in an ISO C program, of
1716 course, but it is useful to put them in header files that might be included
1717 in compilations done with @option{-ansi}. Alternate predefined macros
1718 such as @code{__unix__} and @code{__vax__} are also available, with or
1719 without @option{-ansi}.
1721 The @option{-ansi} option does not cause non-ISO programs to be
1722 rejected gratuitously. For that, @option{-Wpedantic} is required in
1723 addition to @option{-ansi}. @xref{Warning Options}.
1725 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1726 option is used. Some header files may notice this macro and refrain
1727 from declaring certain functions or defining certain macros that the
1728 ISO standard doesn't call for; this is to avoid interfering with any
1729 programs that might use these names for other things.
1731 Functions that are normally built in but do not have semantics
1732 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1733 functions when @option{-ansi} is used. @xref{Other Builtins,,Other
1734 built-in functions provided by GCC}, for details of the functions
1739 Determine the language standard. @xref{Standards,,Language Standards
1740 Supported by GCC}, for details of these standard versions. This option
1741 is currently only supported when compiling C or C++.
1743 The compiler can accept several base standards, such as @samp{c90} or
1744 @samp{c++98}, and GNU dialects of those standards, such as
1745 @samp{gnu90} or @samp{gnu++98}. When a base standard is specified, the
1746 compiler accepts all programs following that standard plus those
1747 using GNU extensions that do not contradict it. For example,
1748 @option{-std=c90} turns off certain features of GCC that are
1749 incompatible with ISO C90, such as the @code{asm} and @code{typeof}
1750 keywords, but not other GNU extensions that do not have a meaning in
1751 ISO C90, such as omitting the middle term of a @code{?:}
1752 expression. On the other hand, when a GNU dialect of a standard is
1753 specified, all features supported by the compiler are enabled, even when
1754 those features change the meaning of the base standard. As a result, some
1755 strict-conforming programs may be rejected. The particular standard
1756 is used by @option{-Wpedantic} to identify which features are GNU
1757 extensions given that version of the standard. For example
1758 @option{-std=gnu90 -Wpedantic} warns about C++ style @samp{//}
1759 comments, while @option{-std=gnu99 -Wpedantic} does not.
1761 A value for this option must be provided; possible values are
1767 Support all ISO C90 programs (certain GNU extensions that conflict
1768 with ISO C90 are disabled). Same as @option{-ansi} for C code.
1770 @item iso9899:199409
1771 ISO C90 as modified in amendment 1.
1777 ISO C99. This standard is substantially completely supported, modulo
1778 bugs and floating-point issues
1779 (mainly but not entirely relating to optional C99 features from
1780 Annexes F and G). See
1781 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1782 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1787 ISO C11, the 2011 revision of the ISO C standard. This standard is
1788 substantially completely supported, modulo bugs, floating-point issues
1789 (mainly but not entirely relating to optional C11 features from
1790 Annexes F and G) and the optional Annexes K (Bounds-checking
1791 interfaces) and L (Analyzability). The name @samp{c1x} is deprecated.
1795 GNU dialect of ISO C90 (including some C99 features).
1799 GNU dialect of ISO C99. The name @samp{gnu9x} is deprecated.
1803 GNU dialect of ISO C11. This is the default for C code.
1804 The name @samp{gnu1x} is deprecated.
1808 The 1998 ISO C++ standard plus the 2003 technical corrigendum and some
1809 additional defect reports. Same as @option{-ansi} for C++ code.
1813 GNU dialect of @option{-std=c++98}.
1817 The 2011 ISO C++ standard plus amendments.
1818 The name @samp{c++0x} is deprecated.
1822 GNU dialect of @option{-std=c++11}.
1823 The name @samp{gnu++0x} is deprecated.
1827 The 2014 ISO C++ standard plus amendments.
1828 The name @samp{c++1y} is deprecated.
1832 GNU dialect of @option{-std=c++14}.
1833 This is the default for C++ code.
1834 The name @samp{gnu++1y} is deprecated.
1837 The next revision of the ISO C++ standard, tentatively planned for
1838 2017. Support is highly experimental, and will almost certainly
1839 change in incompatible ways in future releases.
1842 GNU dialect of @option{-std=c++1z}. Support is highly experimental,
1843 and will almost certainly change in incompatible ways in future
1847 @item -fgnu89-inline
1848 @opindex fgnu89-inline
1849 The option @option{-fgnu89-inline} tells GCC to use the traditional
1850 GNU semantics for @code{inline} functions when in C99 mode.
1851 @xref{Inline,,An Inline Function is As Fast As a Macro}.
1852 Using this option is roughly equivalent to adding the
1853 @code{gnu_inline} function attribute to all inline functions
1854 (@pxref{Function Attributes}).
1856 The option @option{-fno-gnu89-inline} explicitly tells GCC to use the
1857 C99 semantics for @code{inline} when in C99 or gnu99 mode (i.e., it
1858 specifies the default behavior).
1859 This option is not supported in @option{-std=c90} or
1860 @option{-std=gnu90} mode.
1862 The preprocessor macros @code{__GNUC_GNU_INLINE__} and
1863 @code{__GNUC_STDC_INLINE__} may be used to check which semantics are
1864 in effect for @code{inline} functions. @xref{Common Predefined
1865 Macros,,,cpp,The C Preprocessor}.
1867 @item -aux-info @var{filename}
1869 Output to the given filename prototyped declarations for all functions
1870 declared and/or defined in a translation unit, including those in header
1871 files. This option is silently ignored in any language other than C@.
1873 Besides declarations, the file indicates, in comments, the origin of
1874 each declaration (source file and line), whether the declaration was
1875 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1876 @samp{O} for old, respectively, in the first character after the line
1877 number and the colon), and whether it came from a declaration or a
1878 definition (@samp{C} or @samp{F}, respectively, in the following
1879 character). In the case of function definitions, a K&R-style list of
1880 arguments followed by their declarations is also provided, inside
1881 comments, after the declaration.
1883 @item -fallow-parameterless-variadic-functions
1884 @opindex fallow-parameterless-variadic-functions
1885 Accept variadic functions without named parameters.
1887 Although it is possible to define such a function, this is not very
1888 useful as it is not possible to read the arguments. This is only
1889 supported for C as this construct is allowed by C++.
1893 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1894 keyword, so that code can use these words as identifiers. You can use
1895 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1896 instead. @option{-ansi} implies @option{-fno-asm}.
1898 In C++, this switch only affects the @code{typeof} keyword, since
1899 @code{asm} and @code{inline} are standard keywords. You may want to
1900 use the @option{-fno-gnu-keywords} flag instead, which has the same
1901 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1902 switch only affects the @code{asm} and @code{typeof} keywords, since
1903 @code{inline} is a standard keyword in ISO C99.
1906 @itemx -fno-builtin-@var{function}
1907 @opindex fno-builtin
1908 @cindex built-in functions
1909 Don't recognize built-in functions that do not begin with
1910 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1911 functions provided by GCC}, for details of the functions affected,
1912 including those which are not built-in functions when @option{-ansi} or
1913 @option{-std} options for strict ISO C conformance are used because they
1914 do not have an ISO standard meaning.
1916 GCC normally generates special code to handle certain built-in functions
1917 more efficiently; for instance, calls to @code{alloca} may become single
1918 instructions which adjust the stack directly, and calls to @code{memcpy}
1919 may become inline copy loops. The resulting code is often both smaller
1920 and faster, but since the function calls no longer appear as such, you
1921 cannot set a breakpoint on those calls, nor can you change the behavior
1922 of the functions by linking with a different library. In addition,
1923 when a function is recognized as a built-in function, GCC may use
1924 information about that function to warn about problems with calls to
1925 that function, or to generate more efficient code, even if the
1926 resulting code still contains calls to that function. For example,
1927 warnings are given with @option{-Wformat} for bad calls to
1928 @code{printf} when @code{printf} is built in and @code{strlen} is
1929 known not to modify global memory.
1931 With the @option{-fno-builtin-@var{function}} option
1932 only the built-in function @var{function} is
1933 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1934 function is named that is not built-in in this version of GCC, this
1935 option is ignored. There is no corresponding
1936 @option{-fbuiltin-@var{function}} option; if you wish to enable
1937 built-in functions selectively when using @option{-fno-builtin} or
1938 @option{-ffreestanding}, you may define macros such as:
1941 #define abs(n) __builtin_abs ((n))
1942 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1947 @cindex hosted environment
1949 Assert that compilation targets a hosted environment. This implies
1950 @option{-fbuiltin}. A hosted environment is one in which the
1951 entire standard library is available, and in which @code{main} has a return
1952 type of @code{int}. Examples are nearly everything except a kernel.
1953 This is equivalent to @option{-fno-freestanding}.
1955 @item -ffreestanding
1956 @opindex ffreestanding
1957 @cindex hosted environment
1959 Assert that compilation targets a freestanding environment. This
1960 implies @option{-fno-builtin}. A freestanding environment
1961 is one in which the standard library may not exist, and program startup may
1962 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1963 This is equivalent to @option{-fno-hosted}.
1965 @xref{Standards,,Language Standards Supported by GCC}, for details of
1966 freestanding and hosted environments.
1970 @cindex OpenACC accelerator programming
1971 Enable handling of OpenACC directives @code{#pragma acc} in C/C++ and
1972 @code{!$acc} in Fortran. When @option{-fopenacc} is specified, the
1973 compiler generates accelerated code according to the OpenACC Application
1974 Programming Interface v2.0 @w{@uref{http://www.openacc.org/}}. This option
1975 implies @option{-pthread}, and thus is only supported on targets that
1976 have support for @option{-pthread}.
1978 @item -fopenacc-dim=@var{geom}
1979 @opindex fopenacc-dim
1980 @cindex OpenACC accelerator programming
1981 Specify default compute dimensions for parallel offload regions that do
1982 not explicitly specify. The @var{geom} value is a triple of
1983 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A size
1984 can be omitted, to use a target-specific default value.
1988 @cindex OpenMP parallel
1989 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1990 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1991 compiler generates parallel code according to the OpenMP Application
1992 Program Interface v4.0 @w{@uref{http://www.openmp.org/}}. This option
1993 implies @option{-pthread}, and thus is only supported on targets that
1994 have support for @option{-pthread}. @option{-fopenmp} implies
1995 @option{-fopenmp-simd}.
1998 @opindex fopenmp-simd
2001 Enable handling of OpenMP's SIMD directives with @code{#pragma omp}
2002 in C/C++ and @code{!$omp} in Fortran. Other OpenMP directives
2007 @cindex Enable Cilk Plus
2008 Enable the usage of Cilk Plus language extension features for C/C++.
2009 When the option @option{-fcilkplus} is specified, enable the usage of
2010 the Cilk Plus Language extension features for C/C++. The present
2011 implementation follows ABI version 1.2. This is an experimental
2012 feature that is only partially complete, and whose interface may
2013 change in future versions of GCC as the official specification
2014 changes. Currently, all features but @code{_Cilk_for} have been
2019 When the option @option{-fgnu-tm} is specified, the compiler
2020 generates code for the Linux variant of Intel's current Transactional
2021 Memory ABI specification document (Revision 1.1, May 6 2009). This is
2022 an experimental feature whose interface may change in future versions
2023 of GCC, as the official specification changes. Please note that not
2024 all architectures are supported for this feature.
2026 For more information on GCC's support for transactional memory,
2027 @xref{Enabling libitm,,The GNU Transactional Memory Library,libitm,GNU
2028 Transactional Memory Library}.
2030 Note that the transactional memory feature is not supported with
2031 non-call exceptions (@option{-fnon-call-exceptions}).
2033 @item -fms-extensions
2034 @opindex fms-extensions
2035 Accept some non-standard constructs used in Microsoft header files.
2037 In C++ code, this allows member names in structures to be similar
2038 to previous types declarations.
2047 Some cases of unnamed fields in structures and unions are only
2048 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
2049 fields within structs/unions}, for details.
2051 Note that this option is off for all targets but x86
2052 targets using ms-abi.
2054 @item -fplan9-extensions
2055 @opindex fplan9-extensions
2056 Accept some non-standard constructs used in Plan 9 code.
2058 This enables @option{-fms-extensions}, permits passing pointers to
2059 structures with anonymous fields to functions that expect pointers to
2060 elements of the type of the field, and permits referring to anonymous
2061 fields declared using a typedef. @xref{Unnamed Fields,,Unnamed
2062 struct/union fields within structs/unions}, for details. This is only
2063 supported for C, not C++.
2067 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
2068 options for strict ISO C conformance) implies @option{-trigraphs}.
2070 @cindex traditional C language
2071 @cindex C language, traditional
2073 @itemx -traditional-cpp
2074 @opindex traditional-cpp
2075 @opindex traditional
2076 Formerly, these options caused GCC to attempt to emulate a pre-standard
2077 C compiler. They are now only supported with the @option{-E} switch.
2078 The preprocessor continues to support a pre-standard mode. See the GNU
2079 CPP manual for details.
2081 @item -fcond-mismatch
2082 @opindex fcond-mismatch
2083 Allow conditional expressions with mismatched types in the second and
2084 third arguments. The value of such an expression is void. This option
2085 is not supported for C++.
2087 @item -flax-vector-conversions
2088 @opindex flax-vector-conversions
2089 Allow implicit conversions between vectors with differing numbers of
2090 elements and/or incompatible element types. This option should not be
2093 @item -funsigned-char
2094 @opindex funsigned-char
2095 Let the type @code{char} be unsigned, like @code{unsigned char}.
2097 Each kind of machine has a default for what @code{char} should
2098 be. It is either like @code{unsigned char} by default or like
2099 @code{signed char} by default.
2101 Ideally, a portable program should always use @code{signed char} or
2102 @code{unsigned char} when it depends on the signedness of an object.
2103 But many programs have been written to use plain @code{char} and
2104 expect it to be signed, or expect it to be unsigned, depending on the
2105 machines they were written for. This option, and its inverse, let you
2106 make such a program work with the opposite default.
2108 The type @code{char} is always a distinct type from each of
2109 @code{signed char} or @code{unsigned char}, even though its behavior
2110 is always just like one of those two.
2113 @opindex fsigned-char
2114 Let the type @code{char} be signed, like @code{signed char}.
2116 Note that this is equivalent to @option{-fno-unsigned-char}, which is
2117 the negative form of @option{-funsigned-char}. Likewise, the option
2118 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
2120 @item -fsigned-bitfields
2121 @itemx -funsigned-bitfields
2122 @itemx -fno-signed-bitfields
2123 @itemx -fno-unsigned-bitfields
2124 @opindex fsigned-bitfields
2125 @opindex funsigned-bitfields
2126 @opindex fno-signed-bitfields
2127 @opindex fno-unsigned-bitfields
2128 These options control whether a bit-field is signed or unsigned, when the
2129 declaration does not use either @code{signed} or @code{unsigned}. By
2130 default, such a bit-field is signed, because this is consistent: the
2131 basic integer types such as @code{int} are signed types.
2133 @item -fsso-struct=@var{endianness}
2134 @opindex fsso-struct
2135 Set the default scalar storage order of structures and unions to the
2136 specified endianness. The accepted values are @samp{big-endian} and
2137 @samp{little-endian}. If the option is not passed, the compiler uses
2138 the native endianness of the target. This option is not supported for C++.
2140 @strong{Warning:} the @option{-fsso-struct} switch causes GCC to generate
2141 code that is not binary compatible with code generated without it if the
2142 specified endianness is not the native endianness of the target.
2145 @node C++ Dialect Options
2146 @section Options Controlling C++ Dialect
2148 @cindex compiler options, C++
2149 @cindex C++ options, command-line
2150 @cindex options, C++
2151 This section describes the command-line options that are only meaningful
2152 for C++ programs. You can also use most of the GNU compiler options
2153 regardless of what language your program is in. For example, you
2154 might compile a file @file{firstClass.C} like this:
2157 g++ -g -fstrict-enums -O -c firstClass.C
2161 In this example, only @option{-fstrict-enums} is an option meant
2162 only for C++ programs; you can use the other options with any
2163 language supported by GCC@.
2165 Some options for compiling C programs, such as @option{-std}, are also
2166 relevant for C++ programs.
2167 @xref{C Dialect Options,,Options Controlling C Dialect}.
2169 Here is a list of options that are @emph{only} for compiling C++ programs:
2173 @item -fabi-version=@var{n}
2174 @opindex fabi-version
2175 Use version @var{n} of the C++ ABI@. The default is version 0.
2177 Version 0 refers to the version conforming most closely to
2178 the C++ ABI specification. Therefore, the ABI obtained using version 0
2179 will change in different versions of G++ as ABI bugs are fixed.
2181 Version 1 is the version of the C++ ABI that first appeared in G++ 3.2.
2183 Version 2 is the version of the C++ ABI that first appeared in G++
2184 3.4, and was the default through G++ 4.9.
2186 Version 3 corrects an error in mangling a constant address as a
2189 Version 4, which first appeared in G++ 4.5, implements a standard
2190 mangling for vector types.
2192 Version 5, which first appeared in G++ 4.6, corrects the mangling of
2193 attribute const/volatile on function pointer types, decltype of a
2194 plain decl, and use of a function parameter in the declaration of
2197 Version 6, which first appeared in G++ 4.7, corrects the promotion
2198 behavior of C++11 scoped enums and the mangling of template argument
2199 packs, const/static_cast, prefix ++ and --, and a class scope function
2200 used as a template argument.
2202 Version 7, which first appeared in G++ 4.8, that treats nullptr_t as a
2203 builtin type and corrects the mangling of lambdas in default argument
2206 Version 8, which first appeared in G++ 4.9, corrects the substitution
2207 behavior of function types with function-cv-qualifiers.
2209 Version 9, which first appeared in G++ 5.2, corrects the alignment of
2212 Version 10, which first appeared in G++ 6.1, adds mangling of
2213 attributes that affect type identity, such as ia32 calling convention
2214 attributes (e.g. @samp{stdcall}).
2216 See also @option{-Wabi}.
2218 @item -fabi-compat-version=@var{n}
2219 @opindex fabi-compat-version
2220 On targets that support strong aliases, G++
2221 works around mangling changes by creating an alias with the correct
2222 mangled name when defining a symbol with an incorrect mangled name.
2223 This switch specifies which ABI version to use for the alias.
2225 With @option{-fabi-version=0} (the default), this defaults to 8 (GCC 5
2226 compatibility). If another ABI version is explicitly selected, this
2227 defaults to 0. For compatibility with GCC versions 3.2 through 4.9,
2228 use @option{-fabi-compat-version=2}.
2230 If this option is not provided but @option{-Wabi=@var{n}} is, that
2231 version is used for compatibility aliases. If this option is provided
2232 along with @option{-Wabi} (without the version), the version from this
2233 option is used for the warning.
2235 @item -fno-access-control
2236 @opindex fno-access-control
2237 Turn off all access checking. This switch is mainly useful for working
2238 around bugs in the access control code.
2240 @item -fargs-in-order
2241 @opindex fargs-in-order
2242 Evaluate function arguments and operands of some binary expressions in
2243 left-to-right order, and evaluate the right side of an assignment
2244 before the left side, as proposed in P0145R2. Enabled by default with
2245 @option{-std=c++1z}. @option{-fargs-in-order=1} implements all of the
2246 ordering requirements except function arguments.
2250 Check that the pointer returned by @code{operator new} is non-null
2251 before attempting to modify the storage allocated. This check is
2252 normally unnecessary because the C++ standard specifies that
2253 @code{operator new} only returns @code{0} if it is declared
2254 @code{throw()}, in which case the compiler always checks the
2255 return value even without this option. In all other cases, when
2256 @code{operator new} has a non-empty exception specification, memory
2257 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
2258 @samp{new (nothrow)}.
2262 Enable support for the C++ Extensions for Concepts Technical
2263 Specification, ISO 19217 (2015), which allows code like
2266 template <class T> concept bool Addable = requires (T t) @{ t + t; @};
2267 template <Addable T> T add (T a, T b) @{ return a + b; @}
2270 @item -fconstexpr-depth=@var{n}
2271 @opindex fconstexpr-depth
2272 Set the maximum nested evaluation depth for C++11 constexpr functions
2273 to @var{n}. A limit is needed to detect endless recursion during
2274 constant expression evaluation. The minimum specified by the standard
2277 @item -fdeduce-init-list
2278 @opindex fdeduce-init-list
2279 Enable deduction of a template type parameter as
2280 @code{std::initializer_list} from a brace-enclosed initializer list, i.e.@:
2283 template <class T> auto forward(T t) -> decltype (realfn (t))
2290 forward(@{1,2@}); // call forward<std::initializer_list<int>>
2294 This deduction was implemented as a possible extension to the
2295 originally proposed semantics for the C++11 standard, but was not part
2296 of the final standard, so it is disabled by default. This option is
2297 deprecated, and may be removed in a future version of G++.
2299 @item -ffriend-injection
2300 @opindex ffriend-injection
2301 Inject friend functions into the enclosing namespace, so that they are
2302 visible outside the scope of the class in which they are declared.
2303 Friend functions were documented to work this way in the old Annotated
2304 C++ Reference Manual.
2305 However, in ISO C++ a friend function that is not declared
2306 in an enclosing scope can only be found using argument dependent
2307 lookup. GCC defaults to the standard behavior.
2309 This option is for compatibility, and may be removed in a future
2312 @item -fno-elide-constructors
2313 @opindex fno-elide-constructors
2314 The C++ standard allows an implementation to omit creating a temporary
2315 that is only used to initialize another object of the same type.
2316 Specifying this option disables that optimization, and forces G++ to
2317 call the copy constructor in all cases.
2319 @item -fno-enforce-eh-specs
2320 @opindex fno-enforce-eh-specs
2321 Don't generate code to check for violation of exception specifications
2322 at run time. This option violates the C++ standard, but may be useful
2323 for reducing code size in production builds, much like defining
2324 @code{NDEBUG}. This does not give user code permission to throw
2325 exceptions in violation of the exception specifications; the compiler
2326 still optimizes based on the specifications, so throwing an
2327 unexpected exception results in undefined behavior at run time.
2329 @item -fextern-tls-init
2330 @itemx -fno-extern-tls-init
2331 @opindex fextern-tls-init
2332 @opindex fno-extern-tls-init
2333 The C++11 and OpenMP standards allow @code{thread_local} and
2334 @code{threadprivate} variables to have dynamic (runtime)
2335 initialization. To support this, any use of such a variable goes
2336 through a wrapper function that performs any necessary initialization.
2337 When the use and definition of the variable are in the same
2338 translation unit, this overhead can be optimized away, but when the
2339 use is in a different translation unit there is significant overhead
2340 even if the variable doesn't actually need dynamic initialization. If
2341 the programmer can be sure that no use of the variable in a
2342 non-defining TU needs to trigger dynamic initialization (either
2343 because the variable is statically initialized, or a use of the
2344 variable in the defining TU will be executed before any uses in
2345 another TU), they can avoid this overhead with the
2346 @option{-fno-extern-tls-init} option.
2348 On targets that support symbol aliases, the default is
2349 @option{-fextern-tls-init}. On targets that do not support symbol
2350 aliases, the default is @option{-fno-extern-tls-init}.
2353 @itemx -fno-for-scope
2355 @opindex fno-for-scope
2356 If @option{-ffor-scope} is specified, the scope of variables declared in
2357 a @i{for-init-statement} is limited to the @code{for} loop itself,
2358 as specified by the C++ standard.
2359 If @option{-fno-for-scope} is specified, the scope of variables declared in
2360 a @i{for-init-statement} extends to the end of the enclosing scope,
2361 as was the case in old versions of G++, and other (traditional)
2362 implementations of C++.
2364 If neither flag is given, the default is to follow the standard,
2365 but to allow and give a warning for old-style code that would
2366 otherwise be invalid, or have different behavior.
2368 @item -fno-gnu-keywords
2369 @opindex fno-gnu-keywords
2370 Do not recognize @code{typeof} as a keyword, so that code can use this
2371 word as an identifier. You can use the keyword @code{__typeof__} instead.
2372 This option is implied by the strict ISO C++ dialects: @option{-ansi},
2373 @option{-std=c++98}, @option{-std=c++11}, etc.
2375 @item -fno-implicit-templates
2376 @opindex fno-implicit-templates
2377 Never emit code for non-inline templates that are instantiated
2378 implicitly (i.e.@: by use); only emit code for explicit instantiations.
2379 @xref{Template Instantiation}, for more information.
2381 @item -fno-implicit-inline-templates
2382 @opindex fno-implicit-inline-templates
2383 Don't emit code for implicit instantiations of inline templates, either.
2384 The default is to handle inlines differently so that compiles with and
2385 without optimization need the same set of explicit instantiations.
2387 @item -fno-implement-inlines
2388 @opindex fno-implement-inlines
2389 To save space, do not emit out-of-line copies of inline functions
2390 controlled by @code{#pragma implementation}. This causes linker
2391 errors if these functions are not inlined everywhere they are called.
2393 @item -fms-extensions
2394 @opindex fms-extensions
2395 Disable Wpedantic warnings about constructs used in MFC, such as implicit
2396 int and getting a pointer to member function via non-standard syntax.
2398 @item -fno-nonansi-builtins
2399 @opindex fno-nonansi-builtins
2400 Disable built-in declarations of functions that are not mandated by
2401 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
2402 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
2405 @opindex fnothrow-opt
2406 Treat a @code{throw()} exception specification as if it were a
2407 @code{noexcept} specification to reduce or eliminate the text size
2408 overhead relative to a function with no exception specification. If
2409 the function has local variables of types with non-trivial
2410 destructors, the exception specification actually makes the
2411 function smaller because the EH cleanups for those variables can be
2412 optimized away. The semantic effect is that an exception thrown out of
2413 a function with such an exception specification results in a call
2414 to @code{terminate} rather than @code{unexpected}.
2416 @item -fno-operator-names
2417 @opindex fno-operator-names
2418 Do not treat the operator name keywords @code{and}, @code{bitand},
2419 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
2420 synonyms as keywords.
2422 @item -fno-optional-diags
2423 @opindex fno-optional-diags
2424 Disable diagnostics that the standard says a compiler does not need to
2425 issue. Currently, the only such diagnostic issued by G++ is the one for
2426 a name having multiple meanings within a class.
2429 @opindex fpermissive
2430 Downgrade some diagnostics about nonconformant code from errors to
2431 warnings. Thus, using @option{-fpermissive} allows some
2432 nonconforming code to compile.
2434 @item -fno-pretty-templates
2435 @opindex fno-pretty-templates
2436 When an error message refers to a specialization of a function
2437 template, the compiler normally prints the signature of the
2438 template followed by the template arguments and any typedefs or
2439 typenames in the signature (e.g. @code{void f(T) [with T = int]}
2440 rather than @code{void f(int)}) so that it's clear which template is
2441 involved. When an error message refers to a specialization of a class
2442 template, the compiler omits any template arguments that match
2443 the default template arguments for that template. If either of these
2444 behaviors make it harder to understand the error message rather than
2445 easier, you can use @option{-fno-pretty-templates} to disable them.
2449 Enable automatic template instantiation at link time. This option also
2450 implies @option{-fno-implicit-templates}. @xref{Template
2451 Instantiation}, for more information.
2455 Disable generation of information about every class with virtual
2456 functions for use by the C++ run-time type identification features
2457 (@code{dynamic_cast} and @code{typeid}). If you don't use those parts
2458 of the language, you can save some space by using this flag. Note that
2459 exception handling uses the same information, but G++ generates it as
2460 needed. The @code{dynamic_cast} operator can still be used for casts that
2461 do not require run-time type information, i.e.@: casts to @code{void *} or to
2462 unambiguous base classes.
2464 @item -fsized-deallocation
2465 @opindex fsized-deallocation
2466 Enable the built-in global declarations
2468 void operator delete (void *, std::size_t) noexcept;
2469 void operator delete[] (void *, std::size_t) noexcept;
2471 as introduced in C++14. This is useful for user-defined replacement
2472 deallocation functions that, for example, use the size of the object
2473 to make deallocation faster. Enabled by default under
2474 @option{-std=c++14} and above. The flag @option{-Wsized-deallocation}
2475 warns about places that might want to add a definition.
2477 @item -fstrict-enums
2478 @opindex fstrict-enums
2479 Allow the compiler to optimize using the assumption that a value of
2480 enumerated type can only be one of the values of the enumeration (as
2481 defined in the C++ standard; basically, a value that can be
2482 represented in the minimum number of bits needed to represent all the
2483 enumerators). This assumption may not be valid if the program uses a
2484 cast to convert an arbitrary integer value to the enumerated type.
2486 @item -ftemplate-backtrace-limit=@var{n}
2487 @opindex ftemplate-backtrace-limit
2488 Set the maximum number of template instantiation notes for a single
2489 warning or error to @var{n}. The default value is 10.
2491 @item -ftemplate-depth=@var{n}
2492 @opindex ftemplate-depth
2493 Set the maximum instantiation depth for template classes to @var{n}.
2494 A limit on the template instantiation depth is needed to detect
2495 endless recursions during template class instantiation. ANSI/ISO C++
2496 conforming programs must not rely on a maximum depth greater than 17
2497 (changed to 1024 in C++11). The default value is 900, as the compiler
2498 can run out of stack space before hitting 1024 in some situations.
2500 @item -fno-threadsafe-statics
2501 @opindex fno-threadsafe-statics
2502 Do not emit the extra code to use the routines specified in the C++
2503 ABI for thread-safe initialization of local statics. You can use this
2504 option to reduce code size slightly in code that doesn't need to be
2507 @item -fuse-cxa-atexit
2508 @opindex fuse-cxa-atexit
2509 Register destructors for objects with static storage duration with the
2510 @code{__cxa_atexit} function rather than the @code{atexit} function.
2511 This option is required for fully standards-compliant handling of static
2512 destructors, but only works if your C library supports
2513 @code{__cxa_atexit}.
2515 @item -fno-use-cxa-get-exception-ptr
2516 @opindex fno-use-cxa-get-exception-ptr
2517 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
2518 causes @code{std::uncaught_exception} to be incorrect, but is necessary
2519 if the runtime routine is not available.
2521 @item -fvisibility-inlines-hidden
2522 @opindex fvisibility-inlines-hidden
2523 This switch declares that the user does not attempt to compare
2524 pointers to inline functions or methods where the addresses of the two functions
2525 are taken in different shared objects.
2527 The effect of this is that GCC may, effectively, mark inline methods with
2528 @code{__attribute__ ((visibility ("hidden")))} so that they do not
2529 appear in the export table of a DSO and do not require a PLT indirection
2530 when used within the DSO@. Enabling this option can have a dramatic effect
2531 on load and link times of a DSO as it massively reduces the size of the
2532 dynamic export table when the library makes heavy use of templates.
2534 The behavior of this switch is not quite the same as marking the
2535 methods as hidden directly, because it does not affect static variables
2536 local to the function or cause the compiler to deduce that
2537 the function is defined in only one shared object.
2539 You may mark a method as having a visibility explicitly to negate the
2540 effect of the switch for that method. For example, if you do want to
2541 compare pointers to a particular inline method, you might mark it as
2542 having default visibility. Marking the enclosing class with explicit
2543 visibility has no effect.
2545 Explicitly instantiated inline methods are unaffected by this option
2546 as their linkage might otherwise cross a shared library boundary.
2547 @xref{Template Instantiation}.
2549 @item -fvisibility-ms-compat
2550 @opindex fvisibility-ms-compat
2551 This flag attempts to use visibility settings to make GCC's C++
2552 linkage model compatible with that of Microsoft Visual Studio.
2554 The flag makes these changes to GCC's linkage model:
2558 It sets the default visibility to @code{hidden}, like
2559 @option{-fvisibility=hidden}.
2562 Types, but not their members, are not hidden by default.
2565 The One Definition Rule is relaxed for types without explicit
2566 visibility specifications that are defined in more than one
2567 shared object: those declarations are permitted if they are
2568 permitted when this option is not used.
2571 In new code it is better to use @option{-fvisibility=hidden} and
2572 export those classes that are intended to be externally visible.
2573 Unfortunately it is possible for code to rely, perhaps accidentally,
2574 on the Visual Studio behavior.
2576 Among the consequences of these changes are that static data members
2577 of the same type with the same name but defined in different shared
2578 objects are different, so changing one does not change the other;
2579 and that pointers to function members defined in different shared
2580 objects may not compare equal. When this flag is given, it is a
2581 violation of the ODR to define types with the same name differently.
2585 Do not use weak symbol support, even if it is provided by the linker.
2586 By default, G++ uses weak symbols if they are available. This
2587 option exists only for testing, and should not be used by end-users;
2588 it results in inferior code and has no benefits. This option may
2589 be removed in a future release of G++.
2593 Do not search for header files in the standard directories specific to
2594 C++, but do still search the other standard directories. (This option
2595 is used when building the C++ library.)
2598 In addition, these optimization, warning, and code generation options
2599 have meanings only for C++ programs:
2602 @item -Wabi @r{(C, Objective-C, C++ and Objective-C++ only)}
2605 Warn when G++ it generates code that is probably not compatible with
2606 the vendor-neutral C++ ABI@. Since G++ now defaults to updating the
2607 ABI with each major release, normally @option{-Wabi} will warn only if
2608 there is a check added later in a release series for an ABI issue
2609 discovered since the initial release. @option{-Wabi} will warn about
2610 more things if an older ABI version is selected (with
2611 @option{-fabi-version=@var{n}}).
2613 @option{-Wabi} can also be used with an explicit version number to
2614 warn about compatibility with a particular @option{-fabi-version}
2615 level, e.g. @option{-Wabi=2} to warn about changes relative to
2616 @option{-fabi-version=2}.
2618 If an explicit version number is provided and
2619 @option{-fabi-compat-version} is not specified, the version number
2620 from this option is used for compatibility aliases. If no explicit
2621 version number is provided with this option, but
2622 @option{-fabi-compat-version} is specified, that version number is
2623 used for ABI warnings.
2625 Although an effort has been made to warn about
2626 all such cases, there are probably some cases that are not warned about,
2627 even though G++ is generating incompatible code. There may also be
2628 cases where warnings are emitted even though the code that is generated
2631 You should rewrite your code to avoid these warnings if you are
2632 concerned about the fact that code generated by G++ may not be binary
2633 compatible with code generated by other compilers.
2635 Known incompatibilities in @option{-fabi-version=2} (which was the
2636 default from GCC 3.4 to 4.9) include:
2641 A template with a non-type template parameter of reference type was
2642 mangled incorrectly:
2645 template <int &> struct S @{@};
2649 This was fixed in @option{-fabi-version=3}.
2652 SIMD vector types declared using @code{__attribute ((vector_size))} were
2653 mangled in a non-standard way that does not allow for overloading of
2654 functions taking vectors of different sizes.
2656 The mangling was changed in @option{-fabi-version=4}.
2659 @code{__attribute ((const))} and @code{noreturn} were mangled as type
2660 qualifiers, and @code{decltype} of a plain declaration was folded away.
2662 These mangling issues were fixed in @option{-fabi-version=5}.
2665 Scoped enumerators passed as arguments to a variadic function are
2666 promoted like unscoped enumerators, causing @code{va_arg} to complain.
2667 On most targets this does not actually affect the parameter passing
2668 ABI, as there is no way to pass an argument smaller than @code{int}.
2670 Also, the ABI changed the mangling of template argument packs,
2671 @code{const_cast}, @code{static_cast}, prefix increment/decrement, and
2672 a class scope function used as a template argument.
2674 These issues were corrected in @option{-fabi-version=6}.
2677 Lambdas in default argument scope were mangled incorrectly, and the
2678 ABI changed the mangling of @code{nullptr_t}.
2680 These issues were corrected in @option{-fabi-version=7}.
2683 When mangling a function type with function-cv-qualifiers, the
2684 un-qualified function type was incorrectly treated as a substitution
2687 This was fixed in @option{-fabi-version=8}, the default for GCC 5.1.
2690 @code{decltype(nullptr)} incorrectly had an alignment of 1, leading to
2691 unaligned accesses. Note that this did not affect the ABI of a
2692 function with a @code{nullptr_t} parameter, as parameters have a
2695 This was fixed in @option{-fabi-version=9}, the default for GCC 5.2.
2698 Target-specific attributes that affect the identity of a type, such as
2699 ia32 calling conventions on a function type (stdcall, regparm, etc.),
2700 did not affect the mangled name, leading to name collisions when
2701 function pointers were used as template arguments.
2703 This was fixed in @option{-fabi-version=10}, the default for GCC 6.1.
2707 It also warns about psABI-related changes. The known psABI changes at this
2713 For SysV/x86-64, unions with @code{long double} members are
2714 passed in memory as specified in psABI. For example:
2724 @code{union U} is always passed in memory.
2728 @item -Wabi-tag @r{(C++ and Objective-C++ only)}
2731 Warn when a type with an ABI tag is used in a context that does not
2732 have that ABI tag. See @ref{C++ Attributes} for more information
2735 @item -Wctor-dtor-privacy @r{(C++ and Objective-C++ only)}
2736 @opindex Wctor-dtor-privacy
2737 @opindex Wno-ctor-dtor-privacy
2738 Warn when a class seems unusable because all the constructors or
2739 destructors in that class are private, and it has neither friends nor
2740 public static member functions. Also warn if there are no non-private
2741 methods, and there's at least one private member function that isn't
2742 a constructor or destructor.
2744 @item -Wdelete-non-virtual-dtor @r{(C++ and Objective-C++ only)}
2745 @opindex Wdelete-non-virtual-dtor
2746 @opindex Wno-delete-non-virtual-dtor
2747 Warn when @code{delete} is used to destroy an instance of a class that
2748 has virtual functions and non-virtual destructor. It is unsafe to delete
2749 an instance of a derived class through a pointer to a base class if the
2750 base class does not have a virtual destructor. This warning is enabled
2753 @item -Wliteral-suffix @r{(C++ and Objective-C++ only)}
2754 @opindex Wliteral-suffix
2755 @opindex Wno-literal-suffix
2756 Warn when a string or character literal is followed by a ud-suffix which does
2757 not begin with an underscore. As a conforming extension, GCC treats such
2758 suffixes as separate preprocessing tokens in order to maintain backwards
2759 compatibility with code that uses formatting macros from @code{<inttypes.h>}.
2763 #define __STDC_FORMAT_MACROS
2764 #include <inttypes.h>
2769 printf("My int64: %" PRId64"\n", i64);
2773 In this case, @code{PRId64} is treated as a separate preprocessing token.
2775 This warning is enabled by default.
2777 @item -Wlto-type-mismatch
2778 @opindex Wlto-type-mismatch
2779 @opindex Wno-lto-type-mismatch
2781 During the link-time optimization warn about type mismatches in
2782 global declarations from different compilation units.
2783 Requires @option{-flto} to be enabled. Enabled by default.
2785 @item -Wnarrowing @r{(C++ and Objective-C++ only)}
2787 @opindex Wno-narrowing
2788 With @option{-std=gnu++98} or @option{-std=c++98}, warn when a narrowing
2789 conversion prohibited by C++11 occurs within
2793 int i = @{ 2.2 @}; // error: narrowing from double to int
2796 This flag is included in @option{-Wall} and @option{-Wc++11-compat}.
2798 When a later standard is in effect, e.g. when using @option{-std=c++11},
2799 narrowing conversions are diagnosed by default, as required by the standard.
2800 A narrowing conversion from a constant produces an error,
2801 and a narrowing conversion from a non-constant produces a warning,
2802 but @option{-Wno-narrowing} suppresses the diagnostic.
2803 Note that this does not affect the meaning of well-formed code;
2804 narrowing conversions are still considered ill-formed in SFINAE contexts.
2806 @item -Wnoexcept @r{(C++ and Objective-C++ only)}
2808 @opindex Wno-noexcept
2809 Warn when a noexcept-expression evaluates to false because of a call
2810 to a function that does not have a non-throwing exception
2811 specification (i.e. @code{throw()} or @code{noexcept}) but is known by
2812 the compiler to never throw an exception.
2814 @item -Wnon-virtual-dtor @r{(C++ and Objective-C++ only)}
2815 @opindex Wnon-virtual-dtor
2816 @opindex Wno-non-virtual-dtor
2817 Warn when a class has virtual functions and an accessible non-virtual
2818 destructor itself or in an accessible polymorphic base class, in which
2819 case it is possible but unsafe to delete an instance of a derived
2820 class through a pointer to the class itself or base class. This
2821 warning is automatically enabled if @option{-Weffc++} is specified.
2823 @item -Wreorder @r{(C++ and Objective-C++ only)}
2825 @opindex Wno-reorder
2826 @cindex reordering, warning
2827 @cindex warning for reordering of member initializers
2828 Warn when the order of member initializers given in the code does not
2829 match the order in which they must be executed. For instance:
2835 A(): j (0), i (1) @{ @}
2840 The compiler rearranges the member initializers for @code{i}
2841 and @code{j} to match the declaration order of the members, emitting
2842 a warning to that effect. This warning is enabled by @option{-Wall}.
2844 @item -fext-numeric-literals @r{(C++ and Objective-C++ only)}
2845 @opindex fext-numeric-literals
2846 @opindex fno-ext-numeric-literals
2847 Accept imaginary, fixed-point, or machine-defined
2848 literal number suffixes as GNU extensions.
2849 When this option is turned off these suffixes are treated
2850 as C++11 user-defined literal numeric suffixes.
2851 This is on by default for all pre-C++11 dialects and all GNU dialects:
2852 @option{-std=c++98}, @option{-std=gnu++98}, @option{-std=gnu++11},
2853 @option{-std=gnu++14}.
2854 This option is off by default
2855 for ISO C++11 onwards (@option{-std=c++11}, ...).
2858 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
2861 @item -Weffc++ @r{(C++ and Objective-C++ only)}
2864 Warn about violations of the following style guidelines from Scott Meyers'
2865 @cite{Effective C++} series of books:
2869 Define a copy constructor and an assignment operator for classes
2870 with dynamically-allocated memory.
2873 Prefer initialization to assignment in constructors.
2876 Have @code{operator=} return a reference to @code{*this}.
2879 Don't try to return a reference when you must return an object.
2882 Distinguish between prefix and postfix forms of increment and
2883 decrement operators.
2886 Never overload @code{&&}, @code{||}, or @code{,}.
2890 This option also enables @option{-Wnon-virtual-dtor}, which is also
2891 one of the effective C++ recommendations. However, the check is
2892 extended to warn about the lack of virtual destructor in accessible
2893 non-polymorphic bases classes too.
2895 When selecting this option, be aware that the standard library
2896 headers do not obey all of these guidelines; use @samp{grep -v}
2897 to filter out those warnings.
2899 @item -Wstrict-null-sentinel @r{(C++ and Objective-C++ only)}
2900 @opindex Wstrict-null-sentinel
2901 @opindex Wno-strict-null-sentinel
2902 Warn about the use of an uncasted @code{NULL} as sentinel. When
2903 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
2904 to @code{__null}. Although it is a null pointer constant rather than a
2905 null pointer, it is guaranteed to be of the same size as a pointer.
2906 But this use is not portable across different compilers.
2908 @item -Wno-non-template-friend @r{(C++ and Objective-C++ only)}
2909 @opindex Wno-non-template-friend
2910 @opindex Wnon-template-friend
2911 Disable warnings when non-templatized friend functions are declared
2912 within a template. Since the advent of explicit template specification
2913 support in G++, if the name of the friend is an unqualified-id (i.e.,
2914 @samp{friend foo(int)}), the C++ language specification demands that the
2915 friend declare or define an ordinary, nontemplate function. (Section
2916 14.5.3). Before G++ implemented explicit specification, unqualified-ids
2917 could be interpreted as a particular specialization of a templatized
2918 function. Because this non-conforming behavior is no longer the default
2919 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
2920 check existing code for potential trouble spots and is on by default.
2921 This new compiler behavior can be turned off with
2922 @option{-Wno-non-template-friend}, which keeps the conformant compiler code
2923 but disables the helpful warning.
2925 @item -Wold-style-cast @r{(C++ and Objective-C++ only)}
2926 @opindex Wold-style-cast
2927 @opindex Wno-old-style-cast
2928 Warn if an old-style (C-style) cast to a non-void type is used within
2929 a C++ program. The new-style casts (@code{dynamic_cast},
2930 @code{static_cast}, @code{reinterpret_cast}, and @code{const_cast}) are
2931 less vulnerable to unintended effects and much easier to search for.
2933 @item -Woverloaded-virtual @r{(C++ and Objective-C++ only)}
2934 @opindex Woverloaded-virtual
2935 @opindex Wno-overloaded-virtual
2936 @cindex overloaded virtual function, warning
2937 @cindex warning for overloaded virtual function
2938 Warn when a function declaration hides virtual functions from a
2939 base class. For example, in:
2946 struct B: public A @{
2951 the @code{A} class version of @code{f} is hidden in @code{B}, and code
2962 @item -Wno-pmf-conversions @r{(C++ and Objective-C++ only)}
2963 @opindex Wno-pmf-conversions
2964 @opindex Wpmf-conversions
2965 Disable the diagnostic for converting a bound pointer to member function
2968 @item -Wsign-promo @r{(C++ and Objective-C++ only)}
2969 @opindex Wsign-promo
2970 @opindex Wno-sign-promo
2971 Warn when overload resolution chooses a promotion from unsigned or
2972 enumerated type to a signed type, over a conversion to an unsigned type of
2973 the same size. Previous versions of G++ tried to preserve
2974 unsignedness, but the standard mandates the current behavior.
2976 @item -Wtemplates @r{(C++ and Objective-C++ only)}
2978 Warn when a primary template declaration is encountered. Some coding
2979 rules disallow templates, and this may be used to enforce that rule.
2980 The warning is inactive inside a system header file, such as the STL, so
2981 one can still use the STL. One may also instantiate or specialize
2984 @item -Wmultiple-inheritance @r{(C++ and Objective-C++ only)}
2985 @opindex Wmultiple-inheritance
2986 Warn when a class is defined with multiple direct base classes. Some
2987 coding rules disallow multiple inheritance, and this may be used to
2988 enforce that rule. The warning is inactive inside a system header file,
2989 such as the STL, so one can still use the STL. One may also define
2990 classes that indirectly use multiple inheritance.
2992 @item -Wvirtual-inheritance
2993 @opindex Wvirtual-inheritance
2994 Warn when a class is defined with a virtual direct base classe. Some
2995 coding rules disallow multiple inheritance, and this may be used to
2996 enforce that rule. The warning is inactive inside a system header file,
2997 such as the STL, so one can still use the STL. One may also define
2998 classes that indirectly use virtual inheritance.
3001 @opindex Wnamespaces
3002 Warn when a namespace definition is opened. Some coding rules disallow
3003 namespaces, and this may be used to enforce that rule. The warning is
3004 inactive inside a system header file, such as the STL, so one can still
3005 use the STL. One may also use using directives and qualified names.
3007 @item -Wno-terminate @r{(C++ and Objective-C++ only)}
3009 @opindex Wno-terminate
3010 Disable the warning about a throw-expression that will immediately
3011 result in a call to @code{terminate}.
3014 @node Objective-C and Objective-C++ Dialect Options
3015 @section Options Controlling Objective-C and Objective-C++ Dialects
3017 @cindex compiler options, Objective-C and Objective-C++
3018 @cindex Objective-C and Objective-C++ options, command-line
3019 @cindex options, Objective-C and Objective-C++
3020 (NOTE: This manual does not describe the Objective-C and Objective-C++
3021 languages themselves. @xref{Standards,,Language Standards
3022 Supported by GCC}, for references.)
3024 This section describes the command-line options that are only meaningful
3025 for Objective-C and Objective-C++ programs. You can also use most of
3026 the language-independent GNU compiler options.
3027 For example, you might compile a file @file{some_class.m} like this:
3030 gcc -g -fgnu-runtime -O -c some_class.m
3034 In this example, @option{-fgnu-runtime} is an option meant only for
3035 Objective-C and Objective-C++ programs; you can use the other options with
3036 any language supported by GCC@.
3038 Note that since Objective-C is an extension of the C language, Objective-C
3039 compilations may also use options specific to the C front-end (e.g.,
3040 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
3041 C++-specific options (e.g., @option{-Wabi}).
3043 Here is a list of options that are @emph{only} for compiling Objective-C
3044 and Objective-C++ programs:
3047 @item -fconstant-string-class=@var{class-name}
3048 @opindex fconstant-string-class
3049 Use @var{class-name} as the name of the class to instantiate for each
3050 literal string specified with the syntax @code{@@"@dots{}"}. The default
3051 class name is @code{NXConstantString} if the GNU runtime is being used, and
3052 @code{NSConstantString} if the NeXT runtime is being used (see below). The
3053 @option{-fconstant-cfstrings} option, if also present, overrides the
3054 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
3055 to be laid out as constant CoreFoundation strings.
3058 @opindex fgnu-runtime
3059 Generate object code compatible with the standard GNU Objective-C
3060 runtime. This is the default for most types of systems.
3062 @item -fnext-runtime
3063 @opindex fnext-runtime
3064 Generate output compatible with the NeXT runtime. This is the default
3065 for NeXT-based systems, including Darwin and Mac OS X@. The macro
3066 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
3069 @item -fno-nil-receivers
3070 @opindex fno-nil-receivers
3071 Assume that all Objective-C message dispatches (@code{[receiver
3072 message:arg]}) in this translation unit ensure that the receiver is
3073 not @code{nil}. This allows for more efficient entry points in the
3074 runtime to be used. This option is only available in conjunction with
3075 the NeXT runtime and ABI version 0 or 1.
3077 @item -fobjc-abi-version=@var{n}
3078 @opindex fobjc-abi-version
3079 Use version @var{n} of the Objective-C ABI for the selected runtime.
3080 This option is currently supported only for the NeXT runtime. In that
3081 case, Version 0 is the traditional (32-bit) ABI without support for
3082 properties and other Objective-C 2.0 additions. Version 1 is the
3083 traditional (32-bit) ABI with support for properties and other
3084 Objective-C 2.0 additions. Version 2 is the modern (64-bit) ABI. If
3085 nothing is specified, the default is Version 0 on 32-bit target
3086 machines, and Version 2 on 64-bit target machines.
3088 @item -fobjc-call-cxx-cdtors
3089 @opindex fobjc-call-cxx-cdtors
3090 For each Objective-C class, check if any of its instance variables is a
3091 C++ object with a non-trivial default constructor. If so, synthesize a
3092 special @code{- (id) .cxx_construct} instance method which runs
3093 non-trivial default constructors on any such instance variables, in order,
3094 and then return @code{self}. Similarly, check if any instance variable
3095 is a C++ object with a non-trivial destructor, and if so, synthesize a
3096 special @code{- (void) .cxx_destruct} method which runs
3097 all such default destructors, in reverse order.
3099 The @code{- (id) .cxx_construct} and @code{- (void) .cxx_destruct}
3100 methods thusly generated only operate on instance variables
3101 declared in the current Objective-C class, and not those inherited
3102 from superclasses. It is the responsibility of the Objective-C
3103 runtime to invoke all such methods in an object's inheritance
3104 hierarchy. The @code{- (id) .cxx_construct} methods are invoked
3105 by the runtime immediately after a new object instance is allocated;
3106 the @code{- (void) .cxx_destruct} methods are invoked immediately
3107 before the runtime deallocates an object instance.
3109 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
3110 support for invoking the @code{- (id) .cxx_construct} and
3111 @code{- (void) .cxx_destruct} methods.
3113 @item -fobjc-direct-dispatch
3114 @opindex fobjc-direct-dispatch
3115 Allow fast jumps to the message dispatcher. On Darwin this is
3116 accomplished via the comm page.
3118 @item -fobjc-exceptions
3119 @opindex fobjc-exceptions
3120 Enable syntactic support for structured exception handling in
3121 Objective-C, similar to what is offered by C++ and Java. This option
3122 is required to use the Objective-C keywords @code{@@try},
3123 @code{@@throw}, @code{@@catch}, @code{@@finally} and
3124 @code{@@synchronized}. This option is available with both the GNU
3125 runtime and the NeXT runtime (but not available in conjunction with
3126 the NeXT runtime on Mac OS X 10.2 and earlier).
3130 Enable garbage collection (GC) in Objective-C and Objective-C++
3131 programs. This option is only available with the NeXT runtime; the
3132 GNU runtime has a different garbage collection implementation that
3133 does not require special compiler flags.
3135 @item -fobjc-nilcheck
3136 @opindex fobjc-nilcheck
3137 For the NeXT runtime with version 2 of the ABI, check for a nil
3138 receiver in method invocations before doing the actual method call.
3139 This is the default and can be disabled using
3140 @option{-fno-objc-nilcheck}. Class methods and super calls are never
3141 checked for nil in this way no matter what this flag is set to.
3142 Currently this flag does nothing when the GNU runtime, or an older
3143 version of the NeXT runtime ABI, is used.
3145 @item -fobjc-std=objc1
3147 Conform to the language syntax of Objective-C 1.0, the language
3148 recognized by GCC 4.0. This only affects the Objective-C additions to
3149 the C/C++ language; it does not affect conformance to C/C++ standards,
3150 which is controlled by the separate C/C++ dialect option flags. When
3151 this option is used with the Objective-C or Objective-C++ compiler,
3152 any Objective-C syntax that is not recognized by GCC 4.0 is rejected.
3153 This is useful if you need to make sure that your Objective-C code can
3154 be compiled with older versions of GCC@.
3156 @item -freplace-objc-classes
3157 @opindex freplace-objc-classes
3158 Emit a special marker instructing @command{ld(1)} not to statically link in
3159 the resulting object file, and allow @command{dyld(1)} to load it in at
3160 run time instead. This is used in conjunction with the Fix-and-Continue
3161 debugging mode, where the object file in question may be recompiled and
3162 dynamically reloaded in the course of program execution, without the need
3163 to restart the program itself. Currently, Fix-and-Continue functionality
3164 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
3169 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
3170 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
3171 compile time) with static class references that get initialized at load time,
3172 which improves run-time performance. Specifying the @option{-fzero-link} flag
3173 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
3174 to be retained. This is useful in Zero-Link debugging mode, since it allows
3175 for individual class implementations to be modified during program execution.
3176 The GNU runtime currently always retains calls to @code{objc_get_class("@dots{}")}
3177 regardless of command-line options.
3179 @item -fno-local-ivars
3180 @opindex fno-local-ivars
3181 @opindex flocal-ivars
3182 By default instance variables in Objective-C can be accessed as if
3183 they were local variables from within the methods of the class they're
3184 declared in. This can lead to shadowing between instance variables
3185 and other variables declared either locally inside a class method or
3186 globally with the same name. Specifying the @option{-fno-local-ivars}
3187 flag disables this behavior thus avoiding variable shadowing issues.
3189 @item -fivar-visibility=@r{[}public@r{|}protected@r{|}private@r{|}package@r{]}
3190 @opindex fivar-visibility
3191 Set the default instance variable visibility to the specified option
3192 so that instance variables declared outside the scope of any access
3193 modifier directives default to the specified visibility.
3197 Dump interface declarations for all classes seen in the source file to a
3198 file named @file{@var{sourcename}.decl}.
3200 @item -Wassign-intercept @r{(Objective-C and Objective-C++ only)}
3201 @opindex Wassign-intercept
3202 @opindex Wno-assign-intercept
3203 Warn whenever an Objective-C assignment is being intercepted by the
3206 @item -Wno-protocol @r{(Objective-C and Objective-C++ only)}
3207 @opindex Wno-protocol
3209 If a class is declared to implement a protocol, a warning is issued for
3210 every method in the protocol that is not implemented by the class. The
3211 default behavior is to issue a warning for every method not explicitly
3212 implemented in the class, even if a method implementation is inherited
3213 from the superclass. If you use the @option{-Wno-protocol} option, then
3214 methods inherited from the superclass are considered to be implemented,
3215 and no warning is issued for them.
3217 @item -Wselector @r{(Objective-C and Objective-C++ only)}
3219 @opindex Wno-selector
3220 Warn if multiple methods of different types for the same selector are
3221 found during compilation. The check is performed on the list of methods
3222 in the final stage of compilation. Additionally, a check is performed
3223 for each selector appearing in a @code{@@selector(@dots{})}
3224 expression, and a corresponding method for that selector has been found
3225 during compilation. Because these checks scan the method table only at
3226 the end of compilation, these warnings are not produced if the final
3227 stage of compilation is not reached, for example because an error is
3228 found during compilation, or because the @option{-fsyntax-only} option is
3231 @item -Wstrict-selector-match @r{(Objective-C and Objective-C++ only)}
3232 @opindex Wstrict-selector-match
3233 @opindex Wno-strict-selector-match
3234 Warn if multiple methods with differing argument and/or return types are
3235 found for a given selector when attempting to send a message using this
3236 selector to a receiver of type @code{id} or @code{Class}. When this flag
3237 is off (which is the default behavior), the compiler omits such warnings
3238 if any differences found are confined to types that share the same size
3241 @item -Wundeclared-selector @r{(Objective-C and Objective-C++ only)}
3242 @opindex Wundeclared-selector
3243 @opindex Wno-undeclared-selector
3244 Warn if a @code{@@selector(@dots{})} expression referring to an
3245 undeclared selector is found. A selector is considered undeclared if no
3246 method with that name has been declared before the
3247 @code{@@selector(@dots{})} expression, either explicitly in an
3248 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
3249 an @code{@@implementation} section. This option always performs its
3250 checks as soon as a @code{@@selector(@dots{})} expression is found,
3251 while @option{-Wselector} only performs its checks in the final stage of
3252 compilation. This also enforces the coding style convention
3253 that methods and selectors must be declared before being used.
3255 @item -print-objc-runtime-info
3256 @opindex print-objc-runtime-info
3257 Generate C header describing the largest structure that is passed by
3262 @node Diagnostic Message Formatting Options
3263 @section Options to Control Diagnostic Messages Formatting
3264 @cindex options to control diagnostics formatting
3265 @cindex diagnostic messages
3266 @cindex message formatting
3268 Traditionally, diagnostic messages have been formatted irrespective of
3269 the output device's aspect (e.g.@: its width, @dots{}). You can use the
3270 options described below
3271 to control the formatting algorithm for diagnostic messages,
3272 e.g.@: how many characters per line, how often source location
3273 information should be reported. Note that some language front ends may not
3274 honor these options.
3277 @item -fmessage-length=@var{n}
3278 @opindex fmessage-length
3279 Try to format error messages so that they fit on lines of about
3280 @var{n} characters. If @var{n} is zero, then no line-wrapping is
3281 done; each error message appears on a single line. This is the
3282 default for all front ends.
3284 @item -fdiagnostics-show-location=once
3285 @opindex fdiagnostics-show-location
3286 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
3287 reporter to emit source location information @emph{once}; that is, in
3288 case the message is too long to fit on a single physical line and has to
3289 be wrapped, the source location won't be emitted (as prefix) again,
3290 over and over, in subsequent continuation lines. This is the default
3293 @item -fdiagnostics-show-location=every-line
3294 Only meaningful in line-wrapping mode. Instructs the diagnostic
3295 messages reporter to emit the same source location information (as
3296 prefix) for physical lines that result from the process of breaking
3297 a message which is too long to fit on a single line.
3299 @item -fdiagnostics-color[=@var{WHEN}]
3300 @itemx -fno-diagnostics-color
3301 @opindex fdiagnostics-color
3302 @cindex highlight, color
3303 @vindex GCC_COLORS @r{environment variable}
3304 Use color in diagnostics. @var{WHEN} is @samp{never}, @samp{always},
3305 or @samp{auto}. The default depends on how the compiler has been configured,
3306 it can be any of the above @var{WHEN} options or also @samp{never}
3307 if @env{GCC_COLORS} environment variable isn't present in the environment,
3308 and @samp{auto} otherwise.
3309 @samp{auto} means to use color only when the standard error is a terminal.
3310 The forms @option{-fdiagnostics-color} and @option{-fno-diagnostics-color} are
3311 aliases for @option{-fdiagnostics-color=always} and
3312 @option{-fdiagnostics-color=never}, respectively.
3314 The colors are defined by the environment variable @env{GCC_COLORS}.
3315 Its value is a colon-separated list of capabilities and Select Graphic
3316 Rendition (SGR) substrings. SGR commands are interpreted by the
3317 terminal or terminal emulator. (See the section in the documentation
3318 of your text terminal for permitted values and their meanings as
3319 character attributes.) These substring values are integers in decimal
3320 representation and can be concatenated with semicolons.
3321 Common values to concatenate include
3323 @samp{4} for underline,
3325 @samp{7} for inverse,
3326 @samp{39} for default foreground color,
3327 @samp{30} to @samp{37} for foreground colors,
3328 @samp{90} to @samp{97} for 16-color mode foreground colors,
3329 @samp{38;5;0} to @samp{38;5;255}
3330 for 88-color and 256-color modes foreground colors,
3331 @samp{49} for default background color,
3332 @samp{40} to @samp{47} for background colors,
3333 @samp{100} to @samp{107} for 16-color mode background colors,
3334 and @samp{48;5;0} to @samp{48;5;255}
3335 for 88-color and 256-color modes background colors.
3337 The default @env{GCC_COLORS} is
3339 error=01;31:warning=01;35:note=01;36:caret=01;32:locus=01:quote=01
3342 where @samp{01;31} is bold red, @samp{01;35} is bold magenta,
3343 @samp{01;36} is bold cyan, @samp{01;32} is bold green and
3344 @samp{01} is bold. Setting @env{GCC_COLORS} to the empty
3345 string disables colors.
3346 Supported capabilities are as follows.
3350 @vindex error GCC_COLORS @r{capability}
3351 SGR substring for error: markers.
3354 @vindex warning GCC_COLORS @r{capability}
3355 SGR substring for warning: markers.
3358 @vindex note GCC_COLORS @r{capability}
3359 SGR substring for note: markers.
3362 @vindex caret GCC_COLORS @r{capability}
3363 SGR substring for caret line.
3366 @vindex locus GCC_COLORS @r{capability}
3367 SGR substring for location information, @samp{file:line} or
3368 @samp{file:line:column} etc.
3371 @vindex quote GCC_COLORS @r{capability}
3372 SGR substring for information printed within quotes.
3375 @item -fno-diagnostics-show-option
3376 @opindex fno-diagnostics-show-option
3377 @opindex fdiagnostics-show-option
3378 By default, each diagnostic emitted includes text indicating the
3379 command-line option that directly controls the diagnostic (if such an
3380 option is known to the diagnostic machinery). Specifying the
3381 @option{-fno-diagnostics-show-option} flag suppresses that behavior.
3383 @item -fno-diagnostics-show-caret
3384 @opindex fno-diagnostics-show-caret
3385 @opindex fdiagnostics-show-caret
3386 By default, each diagnostic emitted includes the original source line
3387 and a caret @samp{^} indicating the column. This option suppresses this
3388 information. The source line is truncated to @var{n} characters, if
3389 the @option{-fmessage-length=n} option is given. When the output is done
3390 to the terminal, the width is limited to the width given by the
3391 @env{COLUMNS} environment variable or, if not set, to the terminal width.
3393 @item -fdiagnostics-parseable-fixits
3394 @opindex fdiagnostics-parseable-fixits
3395 Emit fix-it hints in a machine-parseable format, suitable for consumption
3396 by IDEs. For each fix-it, a line will be printed after the relevant
3397 diagnostic, starting with the string ``fix-it:''. For example:
3400 fix-it:"test.c":@{45:3-45:21@}:"gtk_widget_show_all"
3403 The location is expressed as a half-open range, expressed as a count of
3404 bytes, starting at byte 1 for the initial column. In the above example,
3405 bytes 3 through 20 of line 45 of ``test.c'' are to be replaced with the
3409 00000000011111111112222222222
3410 12345678901234567890123456789
3411 gtk_widget_showall (dlg);
3416 The filename and replacement string escape backslash as ``\\", tab as ``\t'',
3417 newline as ``\n'', double quotes as ``\"'', non-printable characters as octal
3418 (e.g. vertical tab as ``\013'').
3420 An empty replacement string indicates that the given range is to be removed.
3421 An empty range (e.g. ``45:3-45:3'') indicates that the string is to
3422 be inserted at the given position.
3426 @node Warning Options
3427 @section Options to Request or Suppress Warnings
3428 @cindex options to control warnings
3429 @cindex warning messages
3430 @cindex messages, warning
3431 @cindex suppressing warnings
3433 Warnings are diagnostic messages that report constructions that
3434 are not inherently erroneous but that are risky or suggest there
3435 may have been an error.
3437 The following language-independent options do not enable specific
3438 warnings but control the kinds of diagnostics produced by GCC@.
3441 @cindex syntax checking
3443 @opindex fsyntax-only
3444 Check the code for syntax errors, but don't do anything beyond that.
3446 @item -fmax-errors=@var{n}
3447 @opindex fmax-errors
3448 Limits the maximum number of error messages to @var{n}, at which point
3449 GCC bails out rather than attempting to continue processing the source
3450 code. If @var{n} is 0 (the default), there is no limit on the number
3451 of error messages produced. If @option{-Wfatal-errors} is also
3452 specified, then @option{-Wfatal-errors} takes precedence over this
3457 Inhibit all warning messages.
3462 Make all warnings into errors.
3467 Make the specified warning into an error. The specifier for a warning
3468 is appended; for example @option{-Werror=switch} turns the warnings
3469 controlled by @option{-Wswitch} into errors. This switch takes a
3470 negative form, to be used to negate @option{-Werror} for specific
3471 warnings; for example @option{-Wno-error=switch} makes
3472 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3475 The warning message for each controllable warning includes the
3476 option that controls the warning. That option can then be used with
3477 @option{-Werror=} and @option{-Wno-error=} as described above.
3478 (Printing of the option in the warning message can be disabled using the
3479 @option{-fno-diagnostics-show-option} flag.)
3481 Note that specifying @option{-Werror=}@var{foo} automatically implies
3482 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3485 @item -Wfatal-errors
3486 @opindex Wfatal-errors
3487 @opindex Wno-fatal-errors
3488 This option causes the compiler to abort compilation on the first error
3489 occurred rather than trying to keep going and printing further error
3494 You can request many specific warnings with options beginning with
3495 @samp{-W}, for example @option{-Wimplicit} to request warnings on
3496 implicit declarations. Each of these specific warning options also
3497 has a negative form beginning @samp{-Wno-} to turn off warnings; for
3498 example, @option{-Wno-implicit}. This manual lists only one of the
3499 two forms, whichever is not the default. For further
3500 language-specific options also refer to @ref{C++ Dialect Options} and
3501 @ref{Objective-C and Objective-C++ Dialect Options}.
3503 Some options, such as @option{-Wall} and @option{-Wextra}, turn on other
3504 options, such as @option{-Wunused}, which may turn on further options,
3505 such as @option{-Wunused-value}. The combined effect of positive and
3506 negative forms is that more specific options have priority over less
3507 specific ones, independently of their position in the command-line. For
3508 options of the same specificity, the last one takes effect. Options
3509 enabled or disabled via pragmas (@pxref{Diagnostic Pragmas}) take effect
3510 as if they appeared at the end of the command-line.
3512 When an unrecognized warning option is requested (e.g.,
3513 @option{-Wunknown-warning}), GCC emits a diagnostic stating
3514 that the option is not recognized. However, if the @option{-Wno-} form
3515 is used, the behavior is slightly different: no diagnostic is
3516 produced for @option{-Wno-unknown-warning} unless other diagnostics
3517 are being produced. This allows the use of new @option{-Wno-} options
3518 with old compilers, but if something goes wrong, the compiler
3519 warns that an unrecognized option is present.
3526 Issue all the warnings demanded by strict ISO C and ISO C++;
3527 reject all programs that use forbidden extensions, and some other
3528 programs that do not follow ISO C and ISO C++. For ISO C, follows the
3529 version of the ISO C standard specified by any @option{-std} option used.
3531 Valid ISO C and ISO C++ programs should compile properly with or without
3532 this option (though a rare few require @option{-ansi} or a
3533 @option{-std} option specifying the required version of ISO C)@. However,
3534 without this option, certain GNU extensions and traditional C and C++
3535 features are supported as well. With this option, they are rejected.
3537 @option{-Wpedantic} does not cause warning messages for use of the
3538 alternate keywords whose names begin and end with @samp{__}. Pedantic
3539 warnings are also disabled in the expression that follows
3540 @code{__extension__}. However, only system header files should use
3541 these escape routes; application programs should avoid them.
3542 @xref{Alternate Keywords}.
3544 Some users try to use @option{-Wpedantic} to check programs for strict ISO
3545 C conformance. They soon find that it does not do quite what they want:
3546 it finds some non-ISO practices, but not all---only those for which
3547 ISO C @emph{requires} a diagnostic, and some others for which
3548 diagnostics have been added.
3550 A feature to report any failure to conform to ISO C might be useful in
3551 some instances, but would require considerable additional work and would
3552 be quite different from @option{-Wpedantic}. We don't have plans to
3553 support such a feature in the near future.
3555 Where the standard specified with @option{-std} represents a GNU
3556 extended dialect of C, such as @samp{gnu90} or @samp{gnu99}, there is a
3557 corresponding @dfn{base standard}, the version of ISO C on which the GNU
3558 extended dialect is based. Warnings from @option{-Wpedantic} are given
3559 where they are required by the base standard. (It does not make sense
3560 for such warnings to be given only for features not in the specified GNU
3561 C dialect, since by definition the GNU dialects of C include all
3562 features the compiler supports with the given option, and there would be
3563 nothing to warn about.)
3565 @item -pedantic-errors
3566 @opindex pedantic-errors
3567 Give an error whenever the @dfn{base standard} (see @option{-Wpedantic})
3568 requires a diagnostic, in some cases where there is undefined behavior
3569 at compile-time and in some other cases that do not prevent compilation
3570 of programs that are valid according to the standard. This is not
3571 equivalent to @option{-Werror=pedantic}, since there are errors enabled
3572 by this option and not enabled by the latter and vice versa.
3577 This enables all the warnings about constructions that some users
3578 consider questionable, and that are easy to avoid (or modify to
3579 prevent the warning), even in conjunction with macros. This also
3580 enables some language-specific warnings described in @ref{C++ Dialect
3581 Options} and @ref{Objective-C and Objective-C++ Dialect Options}.
3583 @option{-Wall} turns on the following warning flags:
3585 @gccoptlist{-Waddress @gol
3586 -Warray-bounds=1 @r{(only with} @option{-O2}@r{)} @gol
3588 -Wc++11-compat -Wc++14-compat@gol
3589 -Wchar-subscripts @gol
3591 -Wduplicate-decl-specifier @r{(C and Objective-C only)} @gol
3592 -Wenum-compare @r{(in C/ObjC; this is on by default in C++)} @gol
3594 -Wimplicit @r{(C and Objective-C only)} @gol
3595 -Wimplicit-int @r{(C and Objective-C only)} @gol
3596 -Wimplicit-function-declaration @r{(C and Objective-C only)} @gol
3597 -Winit-self @r{(only for C++)} @gol
3598 -Wlogical-not-parentheses
3599 -Wmain @r{(only for C/ObjC and unless} @option{-ffreestanding}@r{)} @gol
3600 -Wmaybe-uninitialized @gol
3601 -Wmemset-elt-size @gol
3602 -Wmemset-transposed-args @gol
3603 -Wmisleading-indentation @r{(only for C/C++)} @gol
3604 -Wmissing-braces @r{(only for C/ObjC)} @gol
3605 -Wnarrowing @r{(only for C++)} @gol
3607 -Wnonnull-compare @gol
3613 -Wsequence-point @gol
3614 -Wsign-compare @r{(only in C++)} @gol
3615 -Wsizeof-pointer-memaccess @gol
3616 -Wstrict-aliasing @gol
3617 -Wstrict-overflow=1 @gol
3619 -Wtautological-compare @gol
3621 -Wuninitialized @gol
3622 -Wunknown-pragmas @gol
3623 -Wunused-function @gol
3626 -Wunused-variable @gol
3627 -Wvolatile-register-var @gol
3630 Note that some warning flags are not implied by @option{-Wall}. Some of
3631 them warn about constructions that users generally do not consider
3632 questionable, but which occasionally you might wish to check for;
3633 others warn about constructions that are necessary or hard to avoid in
3634 some cases, and there is no simple way to modify the code to suppress
3635 the warning. Some of them are enabled by @option{-Wextra} but many of
3636 them must be enabled individually.
3642 This enables some extra warning flags that are not enabled by
3643 @option{-Wall}. (This option used to be called @option{-W}. The older
3644 name is still supported, but the newer name is more descriptive.)
3646 @gccoptlist{-Wclobbered @gol
3648 -Wignored-qualifiers @gol
3649 -Wmissing-field-initializers @gol
3650 -Wmissing-parameter-type @r{(C only)} @gol
3651 -Wold-style-declaration @r{(C only)} @gol
3652 -Woverride-init @gol
3653 -Wsign-compare @r{(C only)} @gol
3655 -Wuninitialized @gol
3656 -Wshift-negative-value @r{(in C++03 and in C99 and newer)} @gol
3657 -Wunused-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3658 -Wunused-but-set-parameter @r{(only with} @option{-Wunused} @r{or} @option{-Wall}@r{)} @gol
3661 The option @option{-Wextra} also prints warning messages for the
3667 A pointer is compared against integer zero with @code{<}, @code{<=},
3668 @code{>}, or @code{>=}.
3671 (C++ only) An enumerator and a non-enumerator both appear in a
3672 conditional expression.
3675 (C++ only) Ambiguous virtual bases.
3678 (C++ only) Subscripting an array that has been declared @code{register}.
3681 (C++ only) Taking the address of a variable that has been declared
3685 (C++ only) A base class is not initialized in a derived class's copy
3690 @item -Wchar-subscripts
3691 @opindex Wchar-subscripts
3692 @opindex Wno-char-subscripts
3693 Warn if an array subscript has type @code{char}. This is a common cause
3694 of error, as programmers often forget that this type is signed on some
3696 This warning is enabled by @option{-Wall}.
3700 @opindex Wno-comment
3701 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
3702 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
3703 This warning is enabled by @option{-Wall}.
3705 @item -Wno-coverage-mismatch
3706 @opindex Wno-coverage-mismatch
3707 Warn if feedback profiles do not match when using the
3708 @option{-fprofile-use} option.
3709 If a source file is changed between compiling with @option{-fprofile-gen} and
3710 with @option{-fprofile-use}, the files with the profile feedback can fail
3711 to match the source file and GCC cannot use the profile feedback
3712 information. By default, this warning is enabled and is treated as an
3713 error. @option{-Wno-coverage-mismatch} can be used to disable the
3714 warning or @option{-Wno-error=coverage-mismatch} can be used to
3715 disable the error. Disabling the error for this warning can result in
3716 poorly optimized code and is useful only in the
3717 case of very minor changes such as bug fixes to an existing code-base.
3718 Completely disabling the warning is not recommended.
3721 @r{(C, Objective-C, C++, Objective-C++ and Fortran only)}
3723 Suppress warning messages emitted by @code{#warning} directives.
3725 @item -Wdouble-promotion @r{(C, C++, Objective-C and Objective-C++ only)}
3726 @opindex Wdouble-promotion
3727 @opindex Wno-double-promotion
3728 Give a warning when a value of type @code{float} is implicitly
3729 promoted to @code{double}. CPUs with a 32-bit ``single-precision''
3730 floating-point unit implement @code{float} in hardware, but emulate
3731 @code{double} in software. On such a machine, doing computations
3732 using @code{double} values is much more expensive because of the
3733 overhead required for software emulation.
3735 It is easy to accidentally do computations with @code{double} because
3736 floating-point literals are implicitly of type @code{double}. For
3740 float area(float radius)
3742 return 3.14159 * radius * radius;
3746 the compiler performs the entire computation with @code{double}
3747 because the floating-point literal is a @code{double}.
3749 @item -Wduplicate-decl-specifier @r{(C and Objective-C only)}
3750 @opindex Wduplicate-decl-specifier
3751 @opindex Wno-duplicate-decl-specifier
3752 Warn if a declaration has duplicate @code{const}, @code{volatile},
3753 @code{restrict} or @code{_Atomic} specifier. This warning is enabled by
3757 @itemx -Wformat=@var{n}
3760 @opindex ffreestanding
3761 @opindex fno-builtin
3763 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
3764 the arguments supplied have types appropriate to the format string
3765 specified, and that the conversions specified in the format string make
3766 sense. This includes standard functions, and others specified by format
3767 attributes (@pxref{Function Attributes}), in the @code{printf},
3768 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
3769 not in the C standard) families (or other target-specific families).
3770 Which functions are checked without format attributes having been
3771 specified depends on the standard version selected, and such checks of
3772 functions without the attribute specified are disabled by
3773 @option{-ffreestanding} or @option{-fno-builtin}.
3775 The formats are checked against the format features supported by GNU
3776 libc version 2.2. These include all ISO C90 and C99 features, as well
3777 as features from the Single Unix Specification and some BSD and GNU
3778 extensions. Other library implementations may not support all these
3779 features; GCC does not support warning about features that go beyond a
3780 particular library's limitations. However, if @option{-Wpedantic} is used
3781 with @option{-Wformat}, warnings are given about format features not
3782 in the selected standard version (but not for @code{strfmon} formats,
3783 since those are not in any version of the C standard). @xref{C Dialect
3784 Options,,Options Controlling C Dialect}.
3791 Option @option{-Wformat} is equivalent to @option{-Wformat=1}, and
3792 @option{-Wno-format} is equivalent to @option{-Wformat=0}. Since
3793 @option{-Wformat} also checks for null format arguments for several
3794 functions, @option{-Wformat} also implies @option{-Wnonnull}. Some
3795 aspects of this level of format checking can be disabled by the
3796 options: @option{-Wno-format-contains-nul},
3797 @option{-Wno-format-extra-args}, and @option{-Wno-format-zero-length}.
3798 @option{-Wformat} is enabled by @option{-Wall}.
3800 @item -Wno-format-contains-nul
3801 @opindex Wno-format-contains-nul
3802 @opindex Wformat-contains-nul
3803 If @option{-Wformat} is specified, do not warn about format strings that
3806 @item -Wno-format-extra-args
3807 @opindex Wno-format-extra-args
3808 @opindex Wformat-extra-args
3809 If @option{-Wformat} is specified, do not warn about excess arguments to a
3810 @code{printf} or @code{scanf} format function. The C standard specifies
3811 that such arguments are ignored.
3813 Where the unused arguments lie between used arguments that are
3814 specified with @samp{$} operand number specifications, normally
3815 warnings are still given, since the implementation could not know what
3816 type to pass to @code{va_arg} to skip the unused arguments. However,
3817 in the case of @code{scanf} formats, this option suppresses the
3818 warning if the unused arguments are all pointers, since the Single
3819 Unix Specification says that such unused arguments are allowed.
3821 @item -Wno-format-zero-length
3822 @opindex Wno-format-zero-length
3823 @opindex Wformat-zero-length
3824 If @option{-Wformat} is specified, do not warn about zero-length formats.
3825 The C standard specifies that zero-length formats are allowed.
3830 Enable @option{-Wformat} plus additional format checks. Currently
3831 equivalent to @option{-Wformat -Wformat-nonliteral -Wformat-security
3834 @item -Wformat-nonliteral
3835 @opindex Wformat-nonliteral
3836 @opindex Wno-format-nonliteral
3837 If @option{-Wformat} is specified, also warn if the format string is not a
3838 string literal and so cannot be checked, unless the format function
3839 takes its format arguments as a @code{va_list}.
3841 @item -Wformat-security
3842 @opindex Wformat-security
3843 @opindex Wno-format-security
3844 If @option{-Wformat} is specified, also warn about uses of format
3845 functions that represent possible security problems. At present, this
3846 warns about calls to @code{printf} and @code{scanf} functions where the
3847 format string is not a string literal and there are no format arguments,
3848 as in @code{printf (foo);}. This may be a security hole if the format
3849 string came from untrusted input and contains @samp{%n}. (This is
3850 currently a subset of what @option{-Wformat-nonliteral} warns about, but
3851 in future warnings may be added to @option{-Wformat-security} that are not
3852 included in @option{-Wformat-nonliteral}.)
3854 @item -Wformat-signedness
3855 @opindex Wformat-signedness
3856 @opindex Wno-format-signedness
3857 If @option{-Wformat} is specified, also warn if the format string
3858 requires an unsigned argument and the argument is signed and vice versa.
3861 @opindex Wformat-y2k
3862 @opindex Wno-format-y2k
3863 If @option{-Wformat} is specified, also warn about @code{strftime}
3864 formats that may yield only a two-digit year.
3869 @opindex Wno-nonnull
3870 Warn about passing a null pointer for arguments marked as
3871 requiring a non-null value by the @code{nonnull} function attribute.
3873 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
3874 can be disabled with the @option{-Wno-nonnull} option.
3876 @item -Wnonnull-compare
3877 @opindex Wnonnull-compare
3878 @opindex Wno-nonnull-compare
3879 Warn when comparing an argument marked with the @code{nonnull}
3880 function attribute against null inside the function.
3882 @option{-Wnonnull-compare} is included in @option{-Wall}. It
3883 can be disabled with the @option{-Wno-nonnull-compare} option.
3885 @item -Wnull-dereference
3886 @opindex Wnull-dereference
3887 @opindex Wno-null-dereference
3888 Warn if the compiler detects paths that trigger erroneous or
3889 undefined behavior due to dereferencing a null pointer. This option
3890 is only active when @option{-fdelete-null-pointer-checks} is active,
3891 which is enabled by optimizations in most targets. The precision of
3892 the warnings depends on the optimization options used.
3894 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
3896 @opindex Wno-init-self
3897 Warn about uninitialized variables that are initialized with themselves.
3898 Note this option can only be used with the @option{-Wuninitialized} option.
3900 For example, GCC warns about @code{i} being uninitialized in the
3901 following snippet only when @option{-Winit-self} has been specified:
3912 This warning is enabled by @option{-Wall} in C++.
3914 @item -Wimplicit-int @r{(C and Objective-C only)}
3915 @opindex Wimplicit-int
3916 @opindex Wno-implicit-int
3917 Warn when a declaration does not specify a type.
3918 This warning is enabled by @option{-Wall}.
3920 @item -Wimplicit-function-declaration @r{(C and Objective-C only)}
3921 @opindex Wimplicit-function-declaration
3922 @opindex Wno-implicit-function-declaration
3923 Give a warning whenever a function is used before being declared. In
3924 C99 mode (@option{-std=c99} or @option{-std=gnu99}), this warning is
3925 enabled by default and it is made into an error by
3926 @option{-pedantic-errors}. This warning is also enabled by
3929 @item -Wimplicit @r{(C and Objective-C only)}
3931 @opindex Wno-implicit
3932 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
3933 This warning is enabled by @option{-Wall}.
3935 @item -Wignored-qualifiers @r{(C and C++ only)}
3936 @opindex Wignored-qualifiers
3937 @opindex Wno-ignored-qualifiers
3938 Warn if the return type of a function has a type qualifier
3939 such as @code{const}. For ISO C such a type qualifier has no effect,
3940 since the value returned by a function is not an lvalue.
3941 For C++, the warning is only emitted for scalar types or @code{void}.
3942 ISO C prohibits qualified @code{void} return types on function
3943 definitions, so such return types always receive a warning
3944 even without this option.
3946 This warning is also enabled by @option{-Wextra}.
3948 @item -Wignored-attributes @r{(C and C++ only)}
3949 @opindex Wignored-attributes
3950 @opindex Wno-ignored-attributes
3951 Warn when an attribute is ignored. This is different from the
3952 @option{-Wattributes} option in that it warns whenever the compiler decides
3953 to drop an attribute, not that the attribute is either unknown, used in a
3954 wrong place, etc. This warning is enabled by default.
3959 Warn if the type of @code{main} is suspicious. @code{main} should be
3960 a function with external linkage, returning int, taking either zero
3961 arguments, two, or three arguments of appropriate types. This warning
3962 is enabled by default in C++ and is enabled by either @option{-Wall}
3963 or @option{-Wpedantic}.
3965 @item -Wmisleading-indentation @r{(C and C++ only)}
3966 @opindex Wmisleading-indentation
3967 @opindex Wno-misleading-indentation
3968 Warn when the indentation of the code does not reflect the block structure.
3969 Specifically, a warning is issued for @code{if}, @code{else}, @code{while}, and
3970 @code{for} clauses with a guarded statement that does not use braces,
3971 followed by an unguarded statement with the same indentation.
3973 In the following example, the call to ``bar'' is misleadingly indented as
3974 if it were guarded by the ``if'' conditional.
3977 if (some_condition ())
3979 bar (); /* Gotcha: this is not guarded by the "if". */
3982 In the case of mixed tabs and spaces, the warning uses the
3983 @option{-ftabstop=} option to determine if the statements line up
3986 The warning is not issued for code involving multiline preprocessor logic
3987 such as the following example.
3992 #if SOME_CONDITION_THAT_DOES_NOT_HOLD
3998 The warning is not issued after a @code{#line} directive, since this
3999 typically indicates autogenerated code, and no assumptions can be made
4000 about the layout of the file that the directive references.
4002 This warning is enabled by @option{-Wall} in C and C++.
4004 @item -Wmissing-braces
4005 @opindex Wmissing-braces
4006 @opindex Wno-missing-braces
4007 Warn if an aggregate or union initializer is not fully bracketed. In
4008 the following example, the initializer for @code{a} is not fully
4009 bracketed, but that for @code{b} is fully bracketed. This warning is
4010 enabled by @option{-Wall} in C.
4013 int a[2][2] = @{ 0, 1, 2, 3 @};
4014 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
4017 This warning is enabled by @option{-Wall}.
4019 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
4020 @opindex Wmissing-include-dirs
4021 @opindex Wno-missing-include-dirs
4022 Warn if a user-supplied include directory does not exist.
4025 @opindex Wparentheses
4026 @opindex Wno-parentheses
4027 Warn if parentheses are omitted in certain contexts, such
4028 as when there is an assignment in a context where a truth value
4029 is expected, or when operators are nested whose precedence people
4030 often get confused about.
4032 Also warn if a comparison like @code{x<=y<=z} appears; this is
4033 equivalent to @code{(x<=y ? 1 : 0) <= z}, which is a different
4034 interpretation from that of ordinary mathematical notation.
4036 Also warn for dangerous uses of the GNU extension to
4037 @code{?:} with omitted middle operand. When the condition
4038 in the @code{?}: operator is a boolean expression, the omitted value is
4039 always 1. Often programmers expect it to be a value computed
4040 inside the conditional expression instead.
4042 This warning is enabled by @option{-Wall}.
4044 @item -Wsequence-point
4045 @opindex Wsequence-point
4046 @opindex Wno-sequence-point
4047 Warn about code that may have undefined semantics because of violations
4048 of sequence point rules in the C and C++ standards.
4050 The C and C++ standards define the order in which expressions in a C/C++
4051 program are evaluated in terms of @dfn{sequence points}, which represent
4052 a partial ordering between the execution of parts of the program: those
4053 executed before the sequence point, and those executed after it. These
4054 occur after the evaluation of a full expression (one which is not part
4055 of a larger expression), after the evaluation of the first operand of a
4056 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
4057 function is called (but after the evaluation of its arguments and the
4058 expression denoting the called function), and in certain other places.
4059 Other than as expressed by the sequence point rules, the order of
4060 evaluation of subexpressions of an expression is not specified. All
4061 these rules describe only a partial order rather than a total order,
4062 since, for example, if two functions are called within one expression
4063 with no sequence point between them, the order in which the functions
4064 are called is not specified. However, the standards committee have
4065 ruled that function calls do not overlap.
4067 It is not specified when between sequence points modifications to the
4068 values of objects take effect. Programs whose behavior depends on this
4069 have undefined behavior; the C and C++ standards specify that ``Between
4070 the previous and next sequence point an object shall have its stored
4071 value modified at most once by the evaluation of an expression.
4072 Furthermore, the prior value shall be read only to determine the value
4073 to be stored.''. If a program breaks these rules, the results on any
4074 particular implementation are entirely unpredictable.
4076 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
4077 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
4078 diagnosed by this option, and it may give an occasional false positive
4079 result, but in general it has been found fairly effective at detecting
4080 this sort of problem in programs.
4082 The standard is worded confusingly, therefore there is some debate
4083 over the precise meaning of the sequence point rules in subtle cases.
4084 Links to discussions of the problem, including proposed formal
4085 definitions, may be found on the GCC readings page, at
4086 @uref{http://gcc.gnu.org/@/readings.html}.
4088 This warning is enabled by @option{-Wall} for C and C++.
4090 @item -Wno-return-local-addr
4091 @opindex Wno-return-local-addr
4092 @opindex Wreturn-local-addr
4093 Do not warn about returning a pointer (or in C++, a reference) to a
4094 variable that goes out of scope after the function returns.
4097 @opindex Wreturn-type
4098 @opindex Wno-return-type
4099 Warn whenever a function is defined with a return type that defaults
4100 to @code{int}. Also warn about any @code{return} statement with no
4101 return value in a function whose return type is not @code{void}
4102 (falling off the end of the function body is considered returning
4105 For C only, warn about a @code{return} statement with an expression in a
4106 function whose return type is @code{void}, unless the expression type is
4107 also @code{void}. As a GNU extension, the latter case is accepted
4108 without a warning unless @option{-Wpedantic} is used.
4110 For C++, a function without return type always produces a diagnostic
4111 message, even when @option{-Wno-return-type} is specified. The only
4112 exceptions are @code{main} and functions defined in system headers.
4114 This warning is enabled by @option{-Wall}.
4116 @item -Wshift-count-negative
4117 @opindex Wshift-count-negative
4118 @opindex Wno-shift-count-negative
4119 Warn if shift count is negative. This warning is enabled by default.
4121 @item -Wshift-count-overflow
4122 @opindex Wshift-count-overflow
4123 @opindex Wno-shift-count-overflow
4124 Warn if shift count >= width of type. This warning is enabled by default.
4126 @item -Wshift-negative-value
4127 @opindex Wshift-negative-value
4128 @opindex Wno-shift-negative-value
4129 Warn if left shifting a negative value. This warning is enabled by
4130 @option{-Wextra} in C99 and C++11 modes (and newer).
4132 @item -Wshift-overflow
4133 @itemx -Wshift-overflow=@var{n}
4134 @opindex Wshift-overflow
4135 @opindex Wno-shift-overflow
4136 Warn about left shift overflows. This warning is enabled by
4137 default in C99 and C++11 modes (and newer).
4140 @item -Wshift-overflow=1
4141 This is the warning level of @option{-Wshift-overflow} and is enabled
4142 by default in C99 and C++11 modes (and newer). This warning level does
4143 not warn about left-shifting 1 into the sign bit. (However, in C, such
4144 an overflow is still rejected in contexts where an integer constant expression
4147 @item -Wshift-overflow=2
4148 This warning level also warns about left-shifting 1 into the sign bit,
4149 unless C++14 mode is active.
4155 Warn whenever a @code{switch} statement has an index of enumerated type
4156 and lacks a @code{case} for one or more of the named codes of that
4157 enumeration. (The presence of a @code{default} label prevents this
4158 warning.) @code{case} labels outside the enumeration range also
4159 provoke warnings when this option is used (even if there is a
4160 @code{default} label).
4161 This warning is enabled by @option{-Wall}.
4163 @item -Wswitch-default
4164 @opindex Wswitch-default
4165 @opindex Wno-switch-default
4166 Warn whenever a @code{switch} statement does not have a @code{default}
4170 @opindex Wswitch-enum
4171 @opindex Wno-switch-enum
4172 Warn whenever a @code{switch} statement has an index of enumerated type
4173 and lacks a @code{case} for one or more of the named codes of that
4174 enumeration. @code{case} labels outside the enumeration range also
4175 provoke warnings when this option is used. The only difference
4176 between @option{-Wswitch} and this option is that this option gives a
4177 warning about an omitted enumeration code even if there is a
4178 @code{default} label.
4181 @opindex Wswitch-bool
4182 @opindex Wno-switch-bool
4183 Warn whenever a @code{switch} statement has an index of boolean type
4184 and the case values are outside the range of a boolean type.
4185 It is possible to suppress this warning by casting the controlling
4186 expression to a type other than @code{bool}. For example:
4189 switch ((int) (a == 4))
4195 This warning is enabled by default for C and C++ programs.
4197 @item -Wswitch-unreachable
4198 @opindex Wswitch-unreachable
4199 @opindex Wno-switch-unreachable
4200 Warn whenever a @code{switch} statement contains statements between the
4201 controlling expression and the first case label, which will never be
4202 executed. For example:
4214 @option{-Wswitch-unreachable} does not warn if the statement between the
4215 controlling expression and the first case label is just a declaration:
4228 This warning is enabled by default for C and C++ programs.
4230 @item -Wsync-nand @r{(C and C++ only)}
4232 @opindex Wno-sync-nand
4233 Warn when @code{__sync_fetch_and_nand} and @code{__sync_nand_and_fetch}
4234 built-in functions are used. These functions changed semantics in GCC 4.4.
4238 @opindex Wno-trigraphs
4239 Warn if any trigraphs are encountered that might change the meaning of
4240 the program (trigraphs within comments are not warned about).
4241 This warning is enabled by @option{-Wall}.
4243 @item -Wunused-but-set-parameter
4244 @opindex Wunused-but-set-parameter
4245 @opindex Wno-unused-but-set-parameter
4246 Warn whenever a function parameter is assigned to, but otherwise unused
4247 (aside from its declaration).
4249 To suppress this warning use the @code{unused} attribute
4250 (@pxref{Variable Attributes}).
4252 This warning is also enabled by @option{-Wunused} together with
4255 @item -Wunused-but-set-variable
4256 @opindex Wunused-but-set-variable
4257 @opindex Wno-unused-but-set-variable
4258 Warn whenever a local variable is assigned to, but otherwise unused
4259 (aside from its declaration).
4260 This warning is enabled by @option{-Wall}.
4262 To suppress this warning use the @code{unused} attribute
4263 (@pxref{Variable Attributes}).
4265 This warning is also enabled by @option{-Wunused}, which is enabled
4268 @item -Wunused-function
4269 @opindex Wunused-function
4270 @opindex Wno-unused-function
4271 Warn whenever a static function is declared but not defined or a
4272 non-inline static function is unused.
4273 This warning is enabled by @option{-Wall}.
4275 @item -Wunused-label
4276 @opindex Wunused-label
4277 @opindex Wno-unused-label
4278 Warn whenever a label is declared but not used.
4279 This warning is enabled by @option{-Wall}.
4281 To suppress this warning use the @code{unused} attribute
4282 (@pxref{Variable Attributes}).
4284 @item -Wunused-local-typedefs @r{(C, Objective-C, C++ and Objective-C++ only)}
4285 @opindex Wunused-local-typedefs
4286 Warn when a typedef locally defined in a function is not used.
4287 This warning is enabled by @option{-Wall}.
4289 @item -Wunused-parameter
4290 @opindex Wunused-parameter
4291 @opindex Wno-unused-parameter
4292 Warn whenever a function parameter is unused aside from its declaration.
4294 To suppress this warning use the @code{unused} attribute
4295 (@pxref{Variable Attributes}).
4297 @item -Wno-unused-result
4298 @opindex Wunused-result
4299 @opindex Wno-unused-result
4300 Do not warn if a caller of a function marked with attribute
4301 @code{warn_unused_result} (@pxref{Function Attributes}) does not use
4302 its return value. The default is @option{-Wunused-result}.
4304 @item -Wunused-variable
4305 @opindex Wunused-variable
4306 @opindex Wno-unused-variable
4307 Warn whenever a local or static variable is unused aside from its
4308 declaration. This option implies @option{-Wunused-const-variable=1} for C,
4309 but not for C++. This warning is enabled by @option{-Wall}.
4311 To suppress this warning use the @code{unused} attribute
4312 (@pxref{Variable Attributes}).
4314 @item -Wunused-const-variable
4315 @itemx -Wunused-const-variable=@var{n}
4316 @opindex Wunused-const-variable
4317 @opindex Wno-unused-const-variable
4318 Warn whenever a constant static variable is unused aside from its declaration.
4319 @option{-Wunused-const-variable=1} is enabled by @option{-Wunused-variable}
4320 for C, but not for C++. In C this declares variable storage, but in C++ this
4321 is not an error since const variables take the place of @code{#define}s.
4323 To suppress this warning use the @code{unused} attribute
4324 (@pxref{Variable Attributes}).
4327 @item -Wunused-const-variable=1
4328 This is the warning level that is enabled by @option{-Wunused-variable} for
4329 C. It warns only about unused static const variables defined in the main
4330 compilation unit, but not about static const variables declared in any
4333 @item -Wunused-const-variable=2
4334 This warning level also warns for unused constant static variables in
4335 headers (excluding system headers). This is the warning level of
4336 @option{-Wunused-const-variable} and must be explicitly requested since
4337 in C++ this isn't an error and in C it might be harder to clean up all
4341 @item -Wunused-value
4342 @opindex Wunused-value
4343 @opindex Wno-unused-value
4344 Warn whenever a statement computes a result that is explicitly not
4345 used. To suppress this warning cast the unused expression to
4346 @code{void}. This includes an expression-statement or the left-hand
4347 side of a comma expression that contains no side effects. For example,
4348 an expression such as @code{x[i,j]} causes a warning, while
4349 @code{x[(void)i,j]} does not.
4351 This warning is enabled by @option{-Wall}.
4356 All the above @option{-Wunused} options combined.
4358 In order to get a warning about an unused function parameter, you must
4359 either specify @option{-Wextra -Wunused} (note that @option{-Wall} implies
4360 @option{-Wunused}), or separately specify @option{-Wunused-parameter}.
4362 @item -Wuninitialized
4363 @opindex Wuninitialized
4364 @opindex Wno-uninitialized
4365 Warn if an automatic variable is used without first being initialized
4366 or if a variable may be clobbered by a @code{setjmp} call. In C++,
4367 warn if a non-static reference or non-static @code{const} member
4368 appears in a class without constructors.
4370 If you want to warn about code that uses the uninitialized value of the
4371 variable in its own initializer, use the @option{-Winit-self} option.
4373 These warnings occur for individual uninitialized or clobbered
4374 elements of structure, union or array variables as well as for
4375 variables that are uninitialized or clobbered as a whole. They do
4376 not occur for variables or elements declared @code{volatile}. Because
4377 these warnings depend on optimization, the exact variables or elements
4378 for which there are warnings depends on the precise optimization
4379 options and version of GCC used.
4381 Note that there may be no warning about a variable that is used only
4382 to compute a value that itself is never used, because such
4383 computations may be deleted by data flow analysis before the warnings
4386 @item -Winvalid-memory-model
4387 @opindex Winvalid-memory-model
4388 @opindex Wno-invalid-memory-model
4389 Warn for invocations of @ref{__atomic Builtins}, @ref{__sync Builtins},
4390 and the C11 atomic generic functions with a memory consistency argument
4391 that is either invalid for the operation or outside the range of values
4392 of the @code{memory_order} enumeration. For example, since the
4393 @code{__atomic_store} and @code{__atomic_store_n} built-ins are only
4394 defined for the relaxed, release, and sequentially consistent memory
4395 orders the following code is diagnosed:
4400 __atomic_store_n (i, 0, memory_order_consume);
4404 @option{-Winvalid-memory-model} is enabled by default.
4406 @item -Wmaybe-uninitialized
4407 @opindex Wmaybe-uninitialized
4408 @opindex Wno-maybe-uninitialized
4409 For an automatic variable, if there exists a path from the function
4410 entry to a use of the variable that is initialized, but there exist
4411 some other paths for which the variable is not initialized, the compiler
4412 emits a warning if it cannot prove the uninitialized paths are not
4413 executed at run time. These warnings are made optional because GCC is
4414 not smart enough to see all the reasons why the code might be correct
4415 in spite of appearing to have an error. Here is one example of how
4436 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
4437 always initialized, but GCC doesn't know this. To suppress the
4438 warning, you need to provide a default case with assert(0) or
4441 @cindex @code{longjmp} warnings
4442 This option also warns when a non-volatile automatic variable might be
4443 changed by a call to @code{longjmp}. These warnings as well are possible
4444 only in optimizing compilation.
4446 The compiler sees only the calls to @code{setjmp}. It cannot know
4447 where @code{longjmp} will be called; in fact, a signal handler could
4448 call it at any point in the code. As a result, you may get a warning
4449 even when there is in fact no problem because @code{longjmp} cannot
4450 in fact be called at the place that would cause a problem.
4452 Some spurious warnings can be avoided if you declare all the functions
4453 you use that never return as @code{noreturn}. @xref{Function
4456 This warning is enabled by @option{-Wall} or @option{-Wextra}.
4458 @item -Wunknown-pragmas
4459 @opindex Wunknown-pragmas
4460 @opindex Wno-unknown-pragmas
4461 @cindex warning for unknown pragmas
4462 @cindex unknown pragmas, warning
4463 @cindex pragmas, warning of unknown
4464 Warn when a @code{#pragma} directive is encountered that is not understood by
4465 GCC@. If this command-line option is used, warnings are even issued
4466 for unknown pragmas in system header files. This is not the case if
4467 the warnings are only enabled by the @option{-Wall} command-line option.
4470 @opindex Wno-pragmas
4472 Do not warn about misuses of pragmas, such as incorrect parameters,
4473 invalid syntax, or conflicts between pragmas. See also
4474 @option{-Wunknown-pragmas}.
4476 @item -Wstrict-aliasing
4477 @opindex Wstrict-aliasing
4478 @opindex Wno-strict-aliasing
4479 This option is only active when @option{-fstrict-aliasing} is active.
4480 It warns about code that might break the strict aliasing rules that the
4481 compiler is using for optimization. The warning does not catch all
4482 cases, but does attempt to catch the more common pitfalls. It is
4483 included in @option{-Wall}.
4484 It is equivalent to @option{-Wstrict-aliasing=3}
4486 @item -Wstrict-aliasing=n
4487 @opindex Wstrict-aliasing=n
4488 This option is only active when @option{-fstrict-aliasing} is active.
4489 It warns about code that might break the strict aliasing rules that the
4490 compiler is using for optimization.
4491 Higher levels correspond to higher accuracy (fewer false positives).
4492 Higher levels also correspond to more effort, similar to the way @option{-O}
4494 @option{-Wstrict-aliasing} is equivalent to @option{-Wstrict-aliasing=3}.
4496 Level 1: Most aggressive, quick, least accurate.
4497 Possibly useful when higher levels
4498 do not warn but @option{-fstrict-aliasing} still breaks the code, as it has very few
4499 false negatives. However, it has many false positives.
4500 Warns for all pointer conversions between possibly incompatible types,
4501 even if never dereferenced. Runs in the front end only.
4503 Level 2: Aggressive, quick, not too precise.
4504 May still have many false positives (not as many as level 1 though),
4505 and few false negatives (but possibly more than level 1).
4506 Unlike level 1, it only warns when an address is taken. Warns about
4507 incomplete types. Runs in the front end only.
4509 Level 3 (default for @option{-Wstrict-aliasing}):
4510 Should have very few false positives and few false
4511 negatives. Slightly slower than levels 1 or 2 when optimization is enabled.
4512 Takes care of the common pun+dereference pattern in the front end:
4513 @code{*(int*)&some_float}.
4514 If optimization is enabled, it also runs in the back end, where it deals
4515 with multiple statement cases using flow-sensitive points-to information.
4516 Only warns when the converted pointer is dereferenced.
4517 Does not warn about incomplete types.
4519 @item -Wstrict-overflow
4520 @itemx -Wstrict-overflow=@var{n}
4521 @opindex Wstrict-overflow
4522 @opindex Wno-strict-overflow
4523 This option is only active when @option{-fstrict-overflow} is active.
4524 It warns about cases where the compiler optimizes based on the
4525 assumption that signed overflow does not occur. Note that it does not
4526 warn about all cases where the code might overflow: it only warns
4527 about cases where the compiler implements some optimization. Thus
4528 this warning depends on the optimization level.
4530 An optimization that assumes that signed overflow does not occur is
4531 perfectly safe if the values of the variables involved are such that
4532 overflow never does, in fact, occur. Therefore this warning can
4533 easily give a false positive: a warning about code that is not
4534 actually a problem. To help focus on important issues, several
4535 warning levels are defined. No warnings are issued for the use of
4536 undefined signed overflow when estimating how many iterations a loop
4537 requires, in particular when determining whether a loop will be
4541 @item -Wstrict-overflow=1
4542 Warn about cases that are both questionable and easy to avoid. For
4543 example, with @option{-fstrict-overflow}, the compiler simplifies
4544 @code{x + 1 > x} to @code{1}. This level of
4545 @option{-Wstrict-overflow} is enabled by @option{-Wall}; higher levels
4546 are not, and must be explicitly requested.
4548 @item -Wstrict-overflow=2
4549 Also warn about other cases where a comparison is simplified to a
4550 constant. For example: @code{abs (x) >= 0}. This can only be
4551 simplified when @option{-fstrict-overflow} is in effect, because
4552 @code{abs (INT_MIN)} overflows to @code{INT_MIN}, which is less than
4553 zero. @option{-Wstrict-overflow} (with no level) is the same as
4554 @option{-Wstrict-overflow=2}.
4556 @item -Wstrict-overflow=3
4557 Also warn about other cases where a comparison is simplified. For
4558 example: @code{x + 1 > 1} is simplified to @code{x > 0}.
4560 @item -Wstrict-overflow=4
4561 Also warn about other simplifications not covered by the above cases.
4562 For example: @code{(x * 10) / 5} is simplified to @code{x * 2}.
4564 @item -Wstrict-overflow=5
4565 Also warn about cases where the compiler reduces the magnitude of a
4566 constant involved in a comparison. For example: @code{x + 2 > y} is
4567 simplified to @code{x + 1 >= y}. This is reported only at the
4568 highest warning level because this simplification applies to many
4569 comparisons, so this warning level gives a very large number of
4573 @item -Wsuggest-attribute=@r{[}pure@r{|}const@r{|}noreturn@r{|}format@r{]}
4574 @opindex Wsuggest-attribute=
4575 @opindex Wno-suggest-attribute=
4576 Warn for cases where adding an attribute may be beneficial. The
4577 attributes currently supported are listed below.
4580 @item -Wsuggest-attribute=pure
4581 @itemx -Wsuggest-attribute=const
4582 @itemx -Wsuggest-attribute=noreturn
4583 @opindex Wsuggest-attribute=pure
4584 @opindex Wno-suggest-attribute=pure
4585 @opindex Wsuggest-attribute=const
4586 @opindex Wno-suggest-attribute=const
4587 @opindex Wsuggest-attribute=noreturn
4588 @opindex Wno-suggest-attribute=noreturn
4590 Warn about functions that might be candidates for attributes
4591 @code{pure}, @code{const} or @code{noreturn}. The compiler only warns for
4592 functions visible in other compilation units or (in the case of @code{pure} and
4593 @code{const}) if it cannot prove that the function returns normally. A function
4594 returns normally if it doesn't contain an infinite loop or return abnormally
4595 by throwing, calling @code{abort} or trapping. This analysis requires option
4596 @option{-fipa-pure-const}, which is enabled by default at @option{-O} and
4597 higher. Higher optimization levels improve the accuracy of the analysis.
4599 @item -Wsuggest-attribute=format
4600 @itemx -Wmissing-format-attribute
4601 @opindex Wsuggest-attribute=format
4602 @opindex Wmissing-format-attribute
4603 @opindex Wno-suggest-attribute=format
4604 @opindex Wno-missing-format-attribute
4608 Warn about function pointers that might be candidates for @code{format}
4609 attributes. Note these are only possible candidates, not absolute ones.
4610 GCC guesses that function pointers with @code{format} attributes that
4611 are used in assignment, initialization, parameter passing or return
4612 statements should have a corresponding @code{format} attribute in the
4613 resulting type. I.e.@: the left-hand side of the assignment or
4614 initialization, the type of the parameter variable, or the return type
4615 of the containing function respectively should also have a @code{format}
4616 attribute to avoid the warning.
4618 GCC also warns about function definitions that might be
4619 candidates for @code{format} attributes. Again, these are only
4620 possible candidates. GCC guesses that @code{format} attributes
4621 might be appropriate for any function that calls a function like
4622 @code{vprintf} or @code{vscanf}, but this might not always be the
4623 case, and some functions for which @code{format} attributes are
4624 appropriate may not be detected.
4627 @item -Wsuggest-final-types
4628 @opindex Wno-suggest-final-types
4629 @opindex Wsuggest-final-types
4630 Warn about types with virtual methods where code quality would be improved
4631 if the type were declared with the C++11 @code{final} specifier,
4633 declared in an anonymous namespace. This allows GCC to more aggressively
4634 devirtualize the polymorphic calls. This warning is more effective with link
4635 time optimization, where the information about the class hierarchy graph is
4638 @item -Wsuggest-final-methods
4639 @opindex Wno-suggest-final-methods
4640 @opindex Wsuggest-final-methods
4641 Warn about virtual methods where code quality would be improved if the method
4642 were declared with the C++11 @code{final} specifier,
4643 or, if possible, its type were
4644 declared in an anonymous namespace or with the @code{final} specifier.
4646 more effective with link time optimization, where the information about the
4647 class hierarchy graph is more complete. It is recommended to first consider
4648 suggestions of @option{-Wsuggest-final-types} and then rebuild with new
4651 @item -Wsuggest-override
4652 Warn about overriding virtual functions that are not marked with the override
4655 @item -Warray-bounds
4656 @itemx -Warray-bounds=@var{n}
4657 @opindex Wno-array-bounds
4658 @opindex Warray-bounds
4659 This option is only active when @option{-ftree-vrp} is active
4660 (default for @option{-O2} and above). It warns about subscripts to arrays
4661 that are always out of bounds. This warning is enabled by @option{-Wall}.
4664 @item -Warray-bounds=1
4665 This is the warning level of @option{-Warray-bounds} and is enabled
4666 by @option{-Wall}; higher levels are not, and must be explicitly requested.
4668 @item -Warray-bounds=2
4669 This warning level also warns about out of bounds access for
4670 arrays at the end of a struct and for arrays accessed through
4671 pointers. This warning level may give a larger number of
4672 false positives and is deactivated by default.
4675 @item -Wbool-compare
4676 @opindex Wno-bool-compare
4677 @opindex Wbool-compare
4678 Warn about boolean expression compared with an integer value different from
4679 @code{true}/@code{false}. For instance, the following comparison is
4684 if ((n > 1) == 2) @{ @dots{} @}
4686 This warning is enabled by @option{-Wall}.
4688 @item -Wduplicated-cond
4689 @opindex Wno-duplicated-cond
4690 @opindex Wduplicated-cond
4691 Warn about duplicated conditions in an if-else-if chain. For instance,
4692 warn for the following code:
4694 if (p->q != NULL) @{ @dots{} @}
4695 else if (p->q != NULL) @{ @dots{} @}
4698 @item -Wframe-address
4699 @opindex Wno-frame-address
4700 @opindex Wframe-address
4701 Warn when the @samp{__builtin_frame_address} or @samp{__builtin_return_address}
4702 is called with an argument greater than 0. Such calls may return indeterminate
4703 values or crash the program. The warning is included in @option{-Wall}.
4705 @item -Wno-discarded-qualifiers @r{(C and Objective-C only)}
4706 @opindex Wno-discarded-qualifiers
4707 @opindex Wdiscarded-qualifiers
4708 Do not warn if type qualifiers on pointers are being discarded.
4709 Typically, the compiler warns if a @code{const char *} variable is
4710 passed to a function that takes a @code{char *} parameter. This option
4711 can be used to suppress such a warning.
4713 @item -Wno-discarded-array-qualifiers @r{(C and Objective-C only)}
4714 @opindex Wno-discarded-array-qualifiers
4715 @opindex Wdiscarded-array-qualifiers
4716 Do not warn if type qualifiers on arrays which are pointer targets
4717 are being discarded. Typically, the compiler warns if a
4718 @code{const int (*)[]} variable is passed to a function that
4719 takes a @code{int (*)[]} parameter. This option can be used to
4720 suppress such a warning.
4722 @item -Wno-incompatible-pointer-types @r{(C and Objective-C only)}
4723 @opindex Wno-incompatible-pointer-types
4724 @opindex Wincompatible-pointer-types
4725 Do not warn when there is a conversion between pointers that have incompatible
4726 types. This warning is for cases not covered by @option{-Wno-pointer-sign},
4727 which warns for pointer argument passing or assignment with different
4730 @item -Wno-int-conversion @r{(C and Objective-C only)}
4731 @opindex Wno-int-conversion
4732 @opindex Wint-conversion
4733 Do not warn about incompatible integer to pointer and pointer to integer
4734 conversions. This warning is about implicit conversions; for explicit
4735 conversions the warnings @option{-Wno-int-to-pointer-cast} and
4736 @option{-Wno-pointer-to-int-cast} may be used.
4738 @item -Wno-div-by-zero
4739 @opindex Wno-div-by-zero
4740 @opindex Wdiv-by-zero
4741 Do not warn about compile-time integer division by zero. Floating-point
4742 division by zero is not warned about, as it can be a legitimate way of
4743 obtaining infinities and NaNs.
4745 @item -Wsystem-headers
4746 @opindex Wsystem-headers
4747 @opindex Wno-system-headers
4748 @cindex warnings from system headers
4749 @cindex system headers, warnings from
4750 Print warning messages for constructs found in system header files.
4751 Warnings from system headers are normally suppressed, on the assumption
4752 that they usually do not indicate real problems and would only make the
4753 compiler output harder to read. Using this command-line option tells
4754 GCC to emit warnings from system headers as if they occurred in user
4755 code. However, note that using @option{-Wall} in conjunction with this
4756 option does @emph{not} warn about unknown pragmas in system
4757 headers---for that, @option{-Wunknown-pragmas} must also be used.
4759 @item -Wtautological-compare
4760 @opindex Wtautological-compare
4761 @opindex Wno-tautological-compare
4762 Warn if a self-comparison always evaluates to true or false. This
4763 warning detects various mistakes such as:
4767 if (i > i) @{ @dots{} @}
4769 This warning is enabled by @option{-Wall}.
4772 @opindex Wtrampolines
4773 @opindex Wno-trampolines
4774 Warn about trampolines generated for pointers to nested functions.
4775 A trampoline is a small piece of data or code that is created at run
4776 time on the stack when the address of a nested function is taken, and is
4777 used to call the nested function indirectly. For some targets, it is
4778 made up of data only and thus requires no special treatment. But, for
4779 most targets, it is made up of code and thus requires the stack to be
4780 made executable in order for the program to work properly.
4783 @opindex Wfloat-equal
4784 @opindex Wno-float-equal
4785 Warn if floating-point values are used in equality comparisons.
4787 The idea behind this is that sometimes it is convenient (for the
4788 programmer) to consider floating-point values as approximations to
4789 infinitely precise real numbers. If you are doing this, then you need
4790 to compute (by analyzing the code, or in some other way) the maximum or
4791 likely maximum error that the computation introduces, and allow for it
4792 when performing comparisons (and when producing output, but that's a
4793 different problem). In particular, instead of testing for equality, you
4794 should check to see whether the two values have ranges that overlap; and
4795 this is done with the relational operators, so equality comparisons are
4798 @item -Wtraditional @r{(C and Objective-C only)}
4799 @opindex Wtraditional
4800 @opindex Wno-traditional
4801 Warn about certain constructs that behave differently in traditional and
4802 ISO C@. Also warn about ISO C constructs that have no traditional C
4803 equivalent, and/or problematic constructs that should be avoided.
4807 Macro parameters that appear within string literals in the macro body.
4808 In traditional C macro replacement takes place within string literals,
4809 but in ISO C it does not.
4812 In traditional C, some preprocessor directives did not exist.
4813 Traditional preprocessors only considered a line to be a directive
4814 if the @samp{#} appeared in column 1 on the line. Therefore
4815 @option{-Wtraditional} warns about directives that traditional C
4816 understands but ignores because the @samp{#} does not appear as the
4817 first character on the line. It also suggests you hide directives like
4818 @code{#pragma} not understood by traditional C by indenting them. Some
4819 traditional implementations do not recognize @code{#elif}, so this option
4820 suggests avoiding it altogether.
4823 A function-like macro that appears without arguments.
4826 The unary plus operator.
4829 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating-point
4830 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
4831 constants.) Note, these suffixes appear in macros defined in the system
4832 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
4833 Use of these macros in user code might normally lead to spurious
4834 warnings, however GCC's integrated preprocessor has enough context to
4835 avoid warning in these cases.
4838 A function declared external in one block and then used after the end of
4842 A @code{switch} statement has an operand of type @code{long}.
4845 A non-@code{static} function declaration follows a @code{static} one.
4846 This construct is not accepted by some traditional C compilers.
4849 The ISO type of an integer constant has a different width or
4850 signedness from its traditional type. This warning is only issued if
4851 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
4852 typically represent bit patterns, are not warned about.
4855 Usage of ISO string concatenation is detected.
4858 Initialization of automatic aggregates.
4861 Identifier conflicts with labels. Traditional C lacks a separate
4862 namespace for labels.
4865 Initialization of unions. If the initializer is zero, the warning is
4866 omitted. This is done under the assumption that the zero initializer in
4867 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
4868 initializer warnings and relies on default initialization to zero in the
4872 Conversions by prototypes between fixed/floating-point values and vice
4873 versa. The absence of these prototypes when compiling with traditional
4874 C causes serious problems. This is a subset of the possible
4875 conversion warnings; for the full set use @option{-Wtraditional-conversion}.
4878 Use of ISO C style function definitions. This warning intentionally is
4879 @emph{not} issued for prototype declarations or variadic functions
4880 because these ISO C features appear in your code when using
4881 libiberty's traditional C compatibility macros, @code{PARAMS} and
4882 @code{VPARAMS}. This warning is also bypassed for nested functions
4883 because that feature is already a GCC extension and thus not relevant to
4884 traditional C compatibility.
4887 @item -Wtraditional-conversion @r{(C and Objective-C only)}
4888 @opindex Wtraditional-conversion
4889 @opindex Wno-traditional-conversion
4890 Warn if a prototype causes a type conversion that is different from what
4891 would happen to the same argument in the absence of a prototype. This
4892 includes conversions of fixed point to floating and vice versa, and
4893 conversions changing the width or signedness of a fixed-point argument
4894 except when the same as the default promotion.
4896 @item -Wdeclaration-after-statement @r{(C and Objective-C only)}
4897 @opindex Wdeclaration-after-statement
4898 @opindex Wno-declaration-after-statement
4899 Warn when a declaration is found after a statement in a block. This
4900 construct, known from C++, was introduced with ISO C99 and is by default
4901 allowed in GCC@. It is not supported by ISO C90. @xref{Mixed Declarations}.
4906 Warn if an undefined identifier is evaluated in an @code{#if} directive.
4908 @item -Wno-endif-labels
4909 @opindex Wno-endif-labels
4910 @opindex Wendif-labels
4911 Do not warn whenever an @code{#else} or an @code{#endif} are followed by text.
4916 Warn whenever a local variable or type declaration shadows another
4917 variable, parameter, type, class member (in C++), or instance variable
4918 (in Objective-C) or whenever a built-in function is shadowed. Note
4919 that in C++, the compiler warns if a local variable shadows an
4920 explicit typedef, but not if it shadows a struct/class/enum.
4922 @item -Wno-shadow-ivar @r{(Objective-C only)}
4923 @opindex Wno-shadow-ivar
4924 @opindex Wshadow-ivar
4925 Do not warn whenever a local variable shadows an instance variable in an
4928 @item -Wlarger-than=@var{len}
4929 @opindex Wlarger-than=@var{len}
4930 @opindex Wlarger-than-@var{len}
4931 Warn whenever an object of larger than @var{len} bytes is defined.
4933 @item -Wframe-larger-than=@var{len}
4934 @opindex Wframe-larger-than
4935 Warn if the size of a function frame is larger than @var{len} bytes.
4936 The computation done to determine the stack frame size is approximate
4937 and not conservative.
4938 The actual requirements may be somewhat greater than @var{len}
4939 even if you do not get a warning. In addition, any space allocated
4940 via @code{alloca}, variable-length arrays, or related constructs
4941 is not included by the compiler when determining
4942 whether or not to issue a warning.
4944 @item -Wno-free-nonheap-object
4945 @opindex Wno-free-nonheap-object
4946 @opindex Wfree-nonheap-object
4947 Do not warn when attempting to free an object that was not allocated
4950 @item -Wstack-usage=@var{len}
4951 @opindex Wstack-usage
4952 Warn if the stack usage of a function might be larger than @var{len} bytes.
4953 The computation done to determine the stack usage is conservative.
4954 Any space allocated via @code{alloca}, variable-length arrays, or related
4955 constructs is included by the compiler when determining whether or not to
4958 The message is in keeping with the output of @option{-fstack-usage}.
4962 If the stack usage is fully static but exceeds the specified amount, it's:
4965 warning: stack usage is 1120 bytes
4968 If the stack usage is (partly) dynamic but bounded, it's:
4971 warning: stack usage might be 1648 bytes
4974 If the stack usage is (partly) dynamic and not bounded, it's:
4977 warning: stack usage might be unbounded
4981 @item -Wunsafe-loop-optimizations
4982 @opindex Wunsafe-loop-optimizations
4983 @opindex Wno-unsafe-loop-optimizations
4984 Warn if the loop cannot be optimized because the compiler cannot
4985 assume anything on the bounds of the loop indices. With
4986 @option{-funsafe-loop-optimizations} warn if the compiler makes
4989 @item -Wno-pedantic-ms-format @r{(MinGW targets only)}
4990 @opindex Wno-pedantic-ms-format
4991 @opindex Wpedantic-ms-format
4992 When used in combination with @option{-Wformat}
4993 and @option{-pedantic} without GNU extensions, this option
4994 disables the warnings about non-ISO @code{printf} / @code{scanf} format
4995 width specifiers @code{I32}, @code{I64}, and @code{I} used on Windows targets,
4996 which depend on the MS runtime.
4998 @item -Wplacement-new
4999 @itemx -Wplacement-new=@var{n}
5000 @opindex Wplacement-new
5001 @opindex Wno-placement-new
5002 Warn about placement new expressions with undefined behavior, such as
5003 constructing an object in a buffer that is smaller than the type of
5004 the object. For example, the placement new expression below is diagnosed
5005 because it attempts to construct an array of 64 integers in a buffer only
5011 This warning is enabled by default.
5014 @item -Wplacement-new=1
5015 This is the default warning level of @option{-Wplacement-new}. At this
5016 level the warning is not issued for some strictly undefined constructs that
5017 GCC allows as extensions for compatibility with legacy code. For example,
5018 the following @code{new} expression is not diagnosed at this level even
5019 though it has undefined behavior according to the C++ standard because
5020 it writes past the end of the one-element array.
5022 struct S @{ int n, a[1]; @};
5023 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]);
5024 new (s->a)int [32]();
5027 @item -Wplacement-new=2
5028 At this level, in addition to diagnosing all the same constructs as at level
5029 1, a diagnostic is also issued for placement new expressions that construct
5030 an object in the last member of structure whose type is an array of a single
5031 element and whose size is less than the size of the object being constructed.
5032 While the previous example would be diagnosed, the following construct makes
5033 use of the flexible member array extension to avoid the warning at level 2.
5035 struct S @{ int n, a[]; @};
5036 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]);
5037 new (s->a)int [32]();
5042 @item -Wpointer-arith
5043 @opindex Wpointer-arith
5044 @opindex Wno-pointer-arith
5045 Warn about anything that depends on the ``size of'' a function type or
5046 of @code{void}. GNU C assigns these types a size of 1, for
5047 convenience in calculations with @code{void *} pointers and pointers
5048 to functions. In C++, warn also when an arithmetic operation involves
5049 @code{NULL}. This warning is also enabled by @option{-Wpedantic}.
5052 @opindex Wtype-limits
5053 @opindex Wno-type-limits
5054 Warn if a comparison is always true or always false due to the limited
5055 range of the data type, but do not warn for constant expressions. For
5056 example, warn if an unsigned variable is compared against zero with
5057 @code{<} or @code{>=}. This warning is also enabled by
5060 @item -Wbad-function-cast @r{(C and Objective-C only)}
5061 @opindex Wbad-function-cast
5062 @opindex Wno-bad-function-cast
5063 Warn when a function call is cast to a non-matching type.
5064 For example, warn if a call to a function returning an integer type
5065 is cast to a pointer type.
5067 @item -Wc90-c99-compat @r{(C and Objective-C only)}
5068 @opindex Wc90-c99-compat
5069 @opindex Wno-c90-c99-compat
5070 Warn about features not present in ISO C90, but present in ISO C99.
5071 For instance, warn about use of variable length arrays, @code{long long}
5072 type, @code{bool} type, compound literals, designated initializers, and so
5073 on. This option is independent of the standards mode. Warnings are disabled
5074 in the expression that follows @code{__extension__}.
5076 @item -Wc99-c11-compat @r{(C and Objective-C only)}
5077 @opindex Wc99-c11-compat
5078 @opindex Wno-c99-c11-compat
5079 Warn about features not present in ISO C99, but present in ISO C11.
5080 For instance, warn about use of anonymous structures and unions,
5081 @code{_Atomic} type qualifier, @code{_Thread_local} storage-class specifier,
5082 @code{_Alignas} specifier, @code{Alignof} operator, @code{_Generic} keyword,
5083 and so on. This option is independent of the standards mode. Warnings are
5084 disabled in the expression that follows @code{__extension__}.
5086 @item -Wc++-compat @r{(C and Objective-C only)}
5087 @opindex Wc++-compat
5088 Warn about ISO C constructs that are outside of the common subset of
5089 ISO C and ISO C++, e.g.@: request for implicit conversion from
5090 @code{void *} to a pointer to non-@code{void} type.
5092 @item -Wc++11-compat @r{(C++ and Objective-C++ only)}
5093 @opindex Wc++11-compat
5094 Warn about C++ constructs whose meaning differs between ISO C++ 1998
5095 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are keywords
5096 in ISO C++ 2011. This warning turns on @option{-Wnarrowing} and is
5097 enabled by @option{-Wall}.
5099 @item -Wc++14-compat @r{(C++ and Objective-C++ only)}
5100 @opindex Wc++14-compat
5101 Warn about C++ constructs whose meaning differs between ISO C++ 2011
5102 and ISO C++ 2014. This warning is enabled by @option{-Wall}.
5106 @opindex Wno-cast-qual
5107 Warn whenever a pointer is cast so as to remove a type qualifier from
5108 the target type. For example, warn if a @code{const char *} is cast
5109 to an ordinary @code{char *}.
5111 Also warn when making a cast that introduces a type qualifier in an
5112 unsafe way. For example, casting @code{char **} to @code{const char **}
5113 is unsafe, as in this example:
5116 /* p is char ** value. */
5117 const char **q = (const char **) p;
5118 /* Assignment of readonly string to const char * is OK. */
5120 /* Now char** pointer points to read-only memory. */
5125 @opindex Wcast-align
5126 @opindex Wno-cast-align
5127 Warn whenever a pointer is cast such that the required alignment of the
5128 target is increased. For example, warn if a @code{char *} is cast to
5129 an @code{int *} on machines where integers can only be accessed at
5130 two- or four-byte boundaries.
5132 @item -Wwrite-strings
5133 @opindex Wwrite-strings
5134 @opindex Wno-write-strings
5135 When compiling C, give string constants the type @code{const
5136 char[@var{length}]} so that copying the address of one into a
5137 non-@code{const} @code{char *} pointer produces a warning. These
5138 warnings help you find at compile time code that can try to write
5139 into a string constant, but only if you have been very careful about
5140 using @code{const} in declarations and prototypes. Otherwise, it is
5141 just a nuisance. This is why we did not make @option{-Wall} request
5144 When compiling C++, warn about the deprecated conversion from string
5145 literals to @code{char *}. This warning is enabled by default for C++
5150 @opindex Wno-clobbered
5151 Warn for variables that might be changed by @code{longjmp} or
5152 @code{vfork}. This warning is also enabled by @option{-Wextra}.
5154 @item -Wconditionally-supported @r{(C++ and Objective-C++ only)}
5155 @opindex Wconditionally-supported
5156 @opindex Wno-conditionally-supported
5157 Warn for conditionally-supported (C++11 [intro.defs]) constructs.
5160 @opindex Wconversion
5161 @opindex Wno-conversion
5162 Warn for implicit conversions that may alter a value. This includes
5163 conversions between real and integer, like @code{abs (x)} when
5164 @code{x} is @code{double}; conversions between signed and unsigned,
5165 like @code{unsigned ui = -1}; and conversions to smaller types, like
5166 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
5167 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
5168 changed by the conversion like in @code{abs (2.0)}. Warnings about
5169 conversions between signed and unsigned integers can be disabled by
5170 using @option{-Wno-sign-conversion}.
5172 For C++, also warn for confusing overload resolution for user-defined
5173 conversions; and conversions that never use a type conversion
5174 operator: conversions to @code{void}, the same type, a base class or a
5175 reference to them. Warnings about conversions between signed and
5176 unsigned integers are disabled by default in C++ unless
5177 @option{-Wsign-conversion} is explicitly enabled.
5179 @item -Wno-conversion-null @r{(C++ and Objective-C++ only)}
5180 @opindex Wconversion-null
5181 @opindex Wno-conversion-null
5182 Do not warn for conversions between @code{NULL} and non-pointer
5183 types. @option{-Wconversion-null} is enabled by default.
5185 @item -Wzero-as-null-pointer-constant @r{(C++ and Objective-C++ only)}
5186 @opindex Wzero-as-null-pointer-constant
5187 @opindex Wno-zero-as-null-pointer-constant
5188 Warn when a literal @samp{0} is used as null pointer constant. This can
5189 be useful to facilitate the conversion to @code{nullptr} in C++11.
5191 @item -Wsubobject-linkage @r{(C++ and Objective-C++ only)}
5192 @opindex Wsubobject-linkage
5193 @opindex Wno-subobject-linkage
5194 Warn if a class type has a base or a field whose type uses the anonymous
5195 namespace or depends on a type with no linkage. If a type A depends on
5196 a type B with no or internal linkage, defining it in multiple
5197 translation units would be an ODR violation because the meaning of B
5198 is different in each translation unit. If A only appears in a single
5199 translation unit, the best way to silence the warning is to give it
5200 internal linkage by putting it in an anonymous namespace as well. The
5201 compiler doesn't give this warning for types defined in the main .C
5202 file, as those are unlikely to have multiple definitions.
5203 @option{-Wsubobject-linkage} is enabled by default.
5205 @item -Wdangling-else
5206 @opindex Wdangling-else
5207 @opindex Wno-dangling-else
5208 Warn about constructions where there may be confusion to which
5209 @code{if} statement an @code{else} branch belongs. Here is an example of
5224 In C/C++, every @code{else} branch belongs to the innermost possible
5225 @code{if} statement, which in this example is @code{if (b)}. This is
5226 often not what the programmer expected, as illustrated in the above
5227 example by indentation the programmer chose. When there is the
5228 potential for this confusion, GCC issues a warning when this flag
5229 is specified. To eliminate the warning, add explicit braces around
5230 the innermost @code{if} statement so there is no way the @code{else}
5231 can belong to the enclosing @code{if}. The resulting code
5248 This warning is enabled by @option{-Wparentheses}.
5252 @opindex Wno-date-time
5253 Warn when macros @code{__TIME__}, @code{__DATE__} or @code{__TIMESTAMP__}
5254 are encountered as they might prevent bit-wise-identical reproducible
5257 @item -Wdelete-incomplete @r{(C++ and Objective-C++ only)}
5258 @opindex Wdelete-incomplete
5259 @opindex Wno-delete-incomplete
5260 Warn when deleting a pointer to incomplete type, which may cause
5261 undefined behavior at runtime. This warning is enabled by default.
5263 @item -Wuseless-cast @r{(C++ and Objective-C++ only)}
5264 @opindex Wuseless-cast
5265 @opindex Wno-useless-cast
5266 Warn when an expression is casted to its own type.
5269 @opindex Wempty-body
5270 @opindex Wno-empty-body
5271 Warn if an empty body occurs in an @code{if}, @code{else} or @code{do
5272 while} statement. This warning is also enabled by @option{-Wextra}.
5274 @item -Wenum-compare
5275 @opindex Wenum-compare
5276 @opindex Wno-enum-compare
5277 Warn about a comparison between values of different enumerated types.
5278 In C++ enumeral mismatches in conditional expressions are also
5279 diagnosed and the warning is enabled by default. In C this warning is
5280 enabled by @option{-Wall}.
5282 @item -Wjump-misses-init @r{(C, Objective-C only)}
5283 @opindex Wjump-misses-init
5284 @opindex Wno-jump-misses-init
5285 Warn if a @code{goto} statement or a @code{switch} statement jumps
5286 forward across the initialization of a variable, or jumps backward to a
5287 label after the variable has been initialized. This only warns about
5288 variables that are initialized when they are declared. This warning is
5289 only supported for C and Objective-C; in C++ this sort of branch is an
5292 @option{-Wjump-misses-init} is included in @option{-Wc++-compat}. It
5293 can be disabled with the @option{-Wno-jump-misses-init} option.
5295 @item -Wsign-compare
5296 @opindex Wsign-compare
5297 @opindex Wno-sign-compare
5298 @cindex warning for comparison of signed and unsigned values
5299 @cindex comparison of signed and unsigned values, warning
5300 @cindex signed and unsigned values, comparison warning
5301 Warn when a comparison between signed and unsigned values could produce
5302 an incorrect result when the signed value is converted to unsigned.
5303 In C++, this warning is also enabled by @option{-Wall}. In C, it is
5304 also enabled by @option{-Wextra}.
5306 @item -Wsign-conversion
5307 @opindex Wsign-conversion
5308 @opindex Wno-sign-conversion
5309 Warn for implicit conversions that may change the sign of an integer
5310 value, like assigning a signed integer expression to an unsigned
5311 integer variable. An explicit cast silences the warning. In C, this
5312 option is enabled also by @option{-Wconversion}.
5314 @item -Wfloat-conversion
5315 @opindex Wfloat-conversion
5316 @opindex Wno-float-conversion
5317 Warn for implicit conversions that reduce the precision of a real value.
5318 This includes conversions from real to integer, and from higher precision
5319 real to lower precision real values. This option is also enabled by
5320 @option{-Wconversion}.
5322 @item -Wno-scalar-storage-order
5323 @opindex -Wno-scalar-storage-order
5324 @opindex -Wscalar-storage-order
5325 Do not warn on suspicious constructs involving reverse scalar storage order.
5327 @item -Wsized-deallocation @r{(C++ and Objective-C++ only)}
5328 @opindex Wsized-deallocation
5329 @opindex Wno-sized-deallocation
5330 Warn about a definition of an unsized deallocation function
5332 void operator delete (void *) noexcept;
5333 void operator delete[] (void *) noexcept;
5335 without a definition of the corresponding sized deallocation function
5337 void operator delete (void *, std::size_t) noexcept;
5338 void operator delete[] (void *, std::size_t) noexcept;
5340 or vice versa. Enabled by @option{-Wextra} along with
5341 @option{-fsized-deallocation}.
5343 @item -Wsizeof-pointer-memaccess
5344 @opindex Wsizeof-pointer-memaccess
5345 @opindex Wno-sizeof-pointer-memaccess
5346 Warn for suspicious length parameters to certain string and memory built-in
5347 functions if the argument uses @code{sizeof}. This warning warns e.g.@:
5348 about @code{memset (ptr, 0, sizeof (ptr));} if @code{ptr} is not an array,
5349 but a pointer, and suggests a possible fix, or about
5350 @code{memcpy (&foo, ptr, sizeof (&foo));}. This warning is enabled by
5353 @item -Wsizeof-array-argument
5354 @opindex Wsizeof-array-argument
5355 @opindex Wno-sizeof-array-argument
5356 Warn when the @code{sizeof} operator is applied to a parameter that is
5357 declared as an array in a function definition. This warning is enabled by
5358 default for C and C++ programs.
5360 @item -Wmemset-elt-size
5361 @opindex Wmemset-elt-size
5362 @opindex Wno-memset-elt-size
5363 Warn for suspicious calls to the @code{memset} built-in function, if the
5364 first argument references an array, and the third argument is a number
5365 equal to the number of elements, but not equal to the size of the array
5366 in memory. This indicates that the user has omitted a multiplication by
5367 the element size. This warning is enabled by @option{-Wall}.
5369 @item -Wmemset-transposed-args
5370 @opindex Wmemset-transposed-args
5371 @opindex Wno-memset-transposed-args
5372 Warn for suspicious calls to the @code{memset} built-in function, if the
5373 second argument is not zero and the third argument is zero. This warns e.g.@
5374 about @code{memset (buf, sizeof buf, 0)} where most probably
5375 @code{memset (buf, 0, sizeof buf)} was meant instead. The diagnostics
5376 is only emitted if the third argument is literal zero. If it is some
5377 expression that is folded to zero, a cast of zero to some type, etc.,
5378 it is far less likely that the user has mistakenly exchanged the arguments
5379 and no warning is emitted. This warning is enabled by @option{-Wall}.
5383 @opindex Wno-address
5384 Warn about suspicious uses of memory addresses. These include using
5385 the address of a function in a conditional expression, such as
5386 @code{void func(void); if (func)}, and comparisons against the memory
5387 address of a string literal, such as @code{if (x == "abc")}. Such
5388 uses typically indicate a programmer error: the address of a function
5389 always evaluates to true, so their use in a conditional usually
5390 indicate that the programmer forgot the parentheses in a function
5391 call; and comparisons against string literals result in unspecified
5392 behavior and are not portable in C, so they usually indicate that the
5393 programmer intended to use @code{strcmp}. This warning is enabled by
5397 @opindex Wlogical-op
5398 @opindex Wno-logical-op
5399 Warn about suspicious uses of logical operators in expressions.
5400 This includes using logical operators in contexts where a
5401 bit-wise operator is likely to be expected. Also warns when
5402 the operands of a logical operator are the same:
5405 if (a < 0 && a < 0) @{ @dots{} @}
5408 @item -Wlogical-not-parentheses
5409 @opindex Wlogical-not-parentheses
5410 @opindex Wno-logical-not-parentheses
5411 Warn about logical not used on the left hand side operand of a comparison.
5412 This option does not warn if the RHS operand is of a boolean type. Its
5413 purpose is to detect suspicious code like the following:
5417 if (!a > 1) @{ @dots{} @}
5420 It is possible to suppress the warning by wrapping the LHS into
5423 if ((!a) > 1) @{ @dots{} @}
5426 This warning is enabled by @option{-Wall}.
5428 @item -Waggregate-return
5429 @opindex Waggregate-return
5430 @opindex Wno-aggregate-return
5431 Warn if any functions that return structures or unions are defined or
5432 called. (In languages where you can return an array, this also elicits
5435 @item -Wno-aggressive-loop-optimizations
5436 @opindex Wno-aggressive-loop-optimizations
5437 @opindex Waggressive-loop-optimizations
5438 Warn if in a loop with constant number of iterations the compiler detects
5439 undefined behavior in some statement during one or more of the iterations.
5441 @item -Wno-attributes
5442 @opindex Wno-attributes
5443 @opindex Wattributes
5444 Do not warn if an unexpected @code{__attribute__} is used, such as
5445 unrecognized attributes, function attributes applied to variables,
5446 etc. This does not stop errors for incorrect use of supported
5449 @item -Wno-builtin-macro-redefined
5450 @opindex Wno-builtin-macro-redefined
5451 @opindex Wbuiltin-macro-redefined
5452 Do not warn if certain built-in macros are redefined. This suppresses
5453 warnings for redefinition of @code{__TIMESTAMP__}, @code{__TIME__},
5454 @code{__DATE__}, @code{__FILE__}, and @code{__BASE_FILE__}.
5456 @item -Wstrict-prototypes @r{(C and Objective-C only)}
5457 @opindex Wstrict-prototypes
5458 @opindex Wno-strict-prototypes
5459 Warn if a function is declared or defined without specifying the
5460 argument types. (An old-style function definition is permitted without
5461 a warning if preceded by a declaration that specifies the argument
5464 @item -Wold-style-declaration @r{(C and Objective-C only)}
5465 @opindex Wold-style-declaration
5466 @opindex Wno-old-style-declaration
5467 Warn for obsolescent usages, according to the C Standard, in a
5468 declaration. For example, warn if storage-class specifiers like
5469 @code{static} are not the first things in a declaration. This warning
5470 is also enabled by @option{-Wextra}.
5472 @item -Wold-style-definition @r{(C and Objective-C only)}
5473 @opindex Wold-style-definition
5474 @opindex Wno-old-style-definition
5475 Warn if an old-style function definition is used. A warning is given
5476 even if there is a previous prototype.
5478 @item -Wmissing-parameter-type @r{(C and Objective-C only)}
5479 @opindex Wmissing-parameter-type
5480 @opindex Wno-missing-parameter-type
5481 A function parameter is declared without a type specifier in K&R-style
5488 This warning is also enabled by @option{-Wextra}.
5490 @item -Wmissing-prototypes @r{(C and Objective-C only)}
5491 @opindex Wmissing-prototypes
5492 @opindex Wno-missing-prototypes
5493 Warn if a global function is defined without a previous prototype
5494 declaration. This warning is issued even if the definition itself
5495 provides a prototype. Use this option to detect global functions
5496 that do not have a matching prototype declaration in a header file.
5497 This option is not valid for C++ because all function declarations
5498 provide prototypes and a non-matching declaration declares an
5499 overload rather than conflict with an earlier declaration.
5500 Use @option{-Wmissing-declarations} to detect missing declarations in C++.
5502 @item -Wmissing-declarations
5503 @opindex Wmissing-declarations
5504 @opindex Wno-missing-declarations
5505 Warn if a global function is defined without a previous declaration.
5506 Do so even if the definition itself provides a prototype.
5507 Use this option to detect global functions that are not declared in
5508 header files. In C, no warnings are issued for functions with previous
5509 non-prototype declarations; use @option{-Wmissing-prototypes} to detect
5510 missing prototypes. In C++, no warnings are issued for function templates,
5511 or for inline functions, or for functions in anonymous namespaces.
5513 @item -Wmissing-field-initializers
5514 @opindex Wmissing-field-initializers
5515 @opindex Wno-missing-field-initializers
5519 Warn if a structure's initializer has some fields missing. For
5520 example, the following code causes such a warning, because
5521 @code{x.h} is implicitly zero:
5524 struct s @{ int f, g, h; @};
5525 struct s x = @{ 3, 4 @};
5528 This option does not warn about designated initializers, so the following
5529 modification does not trigger a warning:
5532 struct s @{ int f, g, h; @};
5533 struct s x = @{ .f = 3, .g = 4 @};
5536 In C++ this option does not warn either about the empty @{ @}
5537 initializer, for example:
5540 struct s @{ int f, g, h; @};
5544 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
5545 warnings without this one, use @option{-Wextra -Wno-missing-field-initializers}.
5547 @item -Wno-multichar
5548 @opindex Wno-multichar
5550 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
5551 Usually they indicate a typo in the user's code, as they have
5552 implementation-defined values, and should not be used in portable code.
5554 @item -Wnormalized@r{[}=@r{<}none@r{|}id@r{|}nfc@r{|}nfkc@r{>]}
5555 @opindex Wnormalized=
5556 @opindex Wnormalized
5557 @opindex Wno-normalized
5560 @cindex character set, input normalization
5561 In ISO C and ISO C++, two identifiers are different if they are
5562 different sequences of characters. However, sometimes when characters
5563 outside the basic ASCII character set are used, you can have two
5564 different character sequences that look the same. To avoid confusion,
5565 the ISO 10646 standard sets out some @dfn{normalization rules} which
5566 when applied ensure that two sequences that look the same are turned into
5567 the same sequence. GCC can warn you if you are using identifiers that
5568 have not been normalized; this option controls that warning.
5570 There are four levels of warning supported by GCC@. The default is
5571 @option{-Wnormalized=nfc}, which warns about any identifier that is
5572 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
5573 recommended form for most uses. It is equivalent to
5574 @option{-Wnormalized}.
5576 Unfortunately, there are some characters allowed in identifiers by
5577 ISO C and ISO C++ that, when turned into NFC, are not allowed in
5578 identifiers. That is, there's no way to use these symbols in portable
5579 ISO C or C++ and have all your identifiers in NFC@.
5580 @option{-Wnormalized=id} suppresses the warning for these characters.
5581 It is hoped that future versions of the standards involved will correct
5582 this, which is why this option is not the default.
5584 You can switch the warning off for all characters by writing
5585 @option{-Wnormalized=none} or @option{-Wno-normalized}. You should
5586 only do this if you are using some other normalization scheme (like
5587 ``D''), because otherwise you can easily create bugs that are
5588 literally impossible to see.
5590 Some characters in ISO 10646 have distinct meanings but look identical
5591 in some fonts or display methodologies, especially once formatting has
5592 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
5593 LETTER N'', displays just like a regular @code{n} that has been
5594 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
5595 normalization scheme to convert all these into a standard form as
5596 well, and GCC warns if your code is not in NFKC if you use
5597 @option{-Wnormalized=nfkc}. This warning is comparable to warning
5598 about every identifier that contains the letter O because it might be
5599 confused with the digit 0, and so is not the default, but may be
5600 useful as a local coding convention if the programming environment
5601 cannot be fixed to display these characters distinctly.
5603 @item -Wno-deprecated
5604 @opindex Wno-deprecated
5605 @opindex Wdeprecated
5606 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
5608 @item -Wno-deprecated-declarations
5609 @opindex Wno-deprecated-declarations
5610 @opindex Wdeprecated-declarations
5611 Do not warn about uses of functions (@pxref{Function Attributes}),
5612 variables (@pxref{Variable Attributes}), and types (@pxref{Type
5613 Attributes}) marked as deprecated by using the @code{deprecated}
5617 @opindex Wno-overflow
5619 Do not warn about compile-time overflow in constant expressions.
5624 Warn about One Definition Rule violations during link-time optimization.
5625 Requires @option{-flto-odr-type-merging} to be enabled. Enabled by default.
5628 @opindex Wopenm-simd
5629 Warn if the vectorizer cost model overrides the OpenMP or the Cilk Plus
5630 simd directive set by user. The @option{-fsimd-cost-model=unlimited}
5631 option can be used to relax the cost model.
5633 @item -Woverride-init @r{(C and Objective-C only)}
5634 @opindex Woverride-init
5635 @opindex Wno-override-init
5639 Warn if an initialized field without side effects is overridden when
5640 using designated initializers (@pxref{Designated Inits, , Designated
5643 This warning is included in @option{-Wextra}. To get other
5644 @option{-Wextra} warnings without this one, use @option{-Wextra
5645 -Wno-override-init}.
5647 @item -Woverride-init-side-effects @r{(C and Objective-C only)}
5648 @opindex Woverride-init-side-effects
5649 @opindex Wno-override-init-side-effects
5650 Warn if an initialized field with side effects is overridden when
5651 using designated initializers (@pxref{Designated Inits, , Designated
5652 Initializers}). This warning is enabled by default.
5657 Warn if a structure is given the packed attribute, but the packed
5658 attribute has no effect on the layout or size of the structure.
5659 Such structures may be mis-aligned for little benefit. For
5660 instance, in this code, the variable @code{f.x} in @code{struct bar}
5661 is misaligned even though @code{struct bar} does not itself
5662 have the packed attribute:
5669 @} __attribute__((packed));
5677 @item -Wpacked-bitfield-compat
5678 @opindex Wpacked-bitfield-compat
5679 @opindex Wno-packed-bitfield-compat
5680 The 4.1, 4.2 and 4.3 series of GCC ignore the @code{packed} attribute
5681 on bit-fields of type @code{char}. This has been fixed in GCC 4.4 but
5682 the change can lead to differences in the structure layout. GCC
5683 informs you when the offset of such a field has changed in GCC 4.4.
5684 For example there is no longer a 4-bit padding between field @code{a}
5685 and @code{b} in this structure:
5692 @} __attribute__ ((packed));
5695 This warning is enabled by default. Use
5696 @option{-Wno-packed-bitfield-compat} to disable this warning.
5701 Warn if padding is included in a structure, either to align an element
5702 of the structure or to align the whole structure. Sometimes when this
5703 happens it is possible to rearrange the fields of the structure to
5704 reduce the padding and so make the structure smaller.
5706 @item -Wredundant-decls
5707 @opindex Wredundant-decls
5708 @opindex Wno-redundant-decls
5709 Warn if anything is declared more than once in the same scope, even in
5710 cases where multiple declaration is valid and changes nothing.
5712 @item -Wnested-externs @r{(C and Objective-C only)}
5713 @opindex Wnested-externs
5714 @opindex Wno-nested-externs
5715 Warn if an @code{extern} declaration is encountered within a function.
5717 @item -Wno-inherited-variadic-ctor
5718 @opindex Winherited-variadic-ctor
5719 @opindex Wno-inherited-variadic-ctor
5720 Suppress warnings about use of C++11 inheriting constructors when the
5721 base class inherited from has a C variadic constructor; the warning is
5722 on by default because the ellipsis is not inherited.
5727 Warn if a function that is declared as inline cannot be inlined.
5728 Even with this option, the compiler does not warn about failures to
5729 inline functions declared in system headers.
5731 The compiler uses a variety of heuristics to determine whether or not
5732 to inline a function. For example, the compiler takes into account
5733 the size of the function being inlined and the amount of inlining
5734 that has already been done in the current function. Therefore,
5735 seemingly insignificant changes in the source program can cause the
5736 warnings produced by @option{-Winline} to appear or disappear.
5738 @item -Wno-invalid-offsetof @r{(C++ and Objective-C++ only)}
5739 @opindex Wno-invalid-offsetof
5740 @opindex Winvalid-offsetof
5741 Suppress warnings from applying the @code{offsetof} macro to a non-POD
5742 type. According to the 2014 ISO C++ standard, applying @code{offsetof}
5743 to a non-standard-layout type is undefined. In existing C++ implementations,
5744 however, @code{offsetof} typically gives meaningful results.
5745 This flag is for users who are aware that they are
5746 writing nonportable code and who have deliberately chosen to ignore the
5749 The restrictions on @code{offsetof} may be relaxed in a future version
5750 of the C++ standard.
5752 @item -Wno-int-to-pointer-cast
5753 @opindex Wno-int-to-pointer-cast
5754 @opindex Wint-to-pointer-cast
5755 Suppress warnings from casts to pointer type of an integer of a
5756 different size. In C++, casting to a pointer type of smaller size is
5757 an error. @option{Wint-to-pointer-cast} is enabled by default.
5760 @item -Wno-pointer-to-int-cast @r{(C and Objective-C only)}
5761 @opindex Wno-pointer-to-int-cast
5762 @opindex Wpointer-to-int-cast
5763 Suppress warnings from casts from a pointer to an integer type of a
5767 @opindex Winvalid-pch
5768 @opindex Wno-invalid-pch
5769 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
5770 the search path but can't be used.
5774 @opindex Wno-long-long
5775 Warn if @code{long long} type is used. This is enabled by either
5776 @option{-Wpedantic} or @option{-Wtraditional} in ISO C90 and C++98
5777 modes. To inhibit the warning messages, use @option{-Wno-long-long}.
5779 @item -Wvariadic-macros
5780 @opindex Wvariadic-macros
5781 @opindex Wno-variadic-macros
5782 Warn if variadic macros are used in ISO C90 mode, or if the GNU
5783 alternate syntax is used in ISO C99 mode. This is enabled by either
5784 @option{-Wpedantic} or @option{-Wtraditional}. To inhibit the warning
5785 messages, use @option{-Wno-variadic-macros}.
5789 @opindex Wno-varargs
5790 Warn upon questionable usage of the macros used to handle variable
5791 arguments like @code{va_start}. This is default. To inhibit the
5792 warning messages, use @option{-Wno-varargs}.
5794 @item -Wvector-operation-performance
5795 @opindex Wvector-operation-performance
5796 @opindex Wno-vector-operation-performance
5797 Warn if vector operation is not implemented via SIMD capabilities of the
5798 architecture. Mainly useful for the performance tuning.
5799 Vector operation can be implemented @code{piecewise}, which means that the
5800 scalar operation is performed on every vector element;
5801 @code{in parallel}, which means that the vector operation is implemented
5802 using scalars of wider type, which normally is more performance efficient;
5803 and @code{as a single scalar}, which means that vector fits into a
5806 @item -Wno-virtual-move-assign
5807 @opindex Wvirtual-move-assign
5808 @opindex Wno-virtual-move-assign
5809 Suppress warnings about inheriting from a virtual base with a
5810 non-trivial C++11 move assignment operator. This is dangerous because
5811 if the virtual base is reachable along more than one path, it is
5812 moved multiple times, which can mean both objects end up in the
5813 moved-from state. If the move assignment operator is written to avoid
5814 moving from a moved-from object, this warning can be disabled.
5819 Warn if variable length array is used in the code.
5820 @option{-Wno-vla} prevents the @option{-Wpedantic} warning of
5821 the variable length array.
5823 @item -Wvolatile-register-var
5824 @opindex Wvolatile-register-var
5825 @opindex Wno-volatile-register-var
5826 Warn if a register variable is declared volatile. The volatile
5827 modifier does not inhibit all optimizations that may eliminate reads
5828 and/or writes to register variables. This warning is enabled by
5831 @item -Wdisabled-optimization
5832 @opindex Wdisabled-optimization
5833 @opindex Wno-disabled-optimization
5834 Warn if a requested optimization pass is disabled. This warning does
5835 not generally indicate that there is anything wrong with your code; it
5836 merely indicates that GCC's optimizers are unable to handle the code
5837 effectively. Often, the problem is that your code is too big or too
5838 complex; GCC refuses to optimize programs when the optimization
5839 itself is likely to take inordinate amounts of time.
5841 @item -Wpointer-sign @r{(C and Objective-C only)}
5842 @opindex Wpointer-sign
5843 @opindex Wno-pointer-sign
5844 Warn for pointer argument passing or assignment with different signedness.
5845 This option is only supported for C and Objective-C@. It is implied by
5846 @option{-Wall} and by @option{-Wpedantic}, which can be disabled with
5847 @option{-Wno-pointer-sign}.
5849 @item -Wstack-protector
5850 @opindex Wstack-protector
5851 @opindex Wno-stack-protector
5852 This option is only active when @option{-fstack-protector} is active. It
5853 warns about functions that are not protected against stack smashing.
5855 @item -Woverlength-strings
5856 @opindex Woverlength-strings
5857 @opindex Wno-overlength-strings
5858 Warn about string constants that are longer than the ``minimum
5859 maximum'' length specified in the C standard. Modern compilers
5860 generally allow string constants that are much longer than the
5861 standard's minimum limit, but very portable programs should avoid
5862 using longer strings.
5864 The limit applies @emph{after} string constant concatenation, and does
5865 not count the trailing NUL@. In C90, the limit was 509 characters; in
5866 C99, it was raised to 4095. C++98 does not specify a normative
5867 minimum maximum, so we do not diagnose overlength strings in C++@.
5869 This option is implied by @option{-Wpedantic}, and can be disabled with
5870 @option{-Wno-overlength-strings}.
5872 @item -Wunsuffixed-float-constants @r{(C and Objective-C only)}
5873 @opindex Wunsuffixed-float-constants
5875 Issue a warning for any floating constant that does not have
5876 a suffix. When used together with @option{-Wsystem-headers} it
5877 warns about such constants in system header files. This can be useful
5878 when preparing code to use with the @code{FLOAT_CONST_DECIMAL64} pragma
5879 from the decimal floating-point extension to C99.
5881 @item -Wno-designated-init @r{(C and Objective-C only)}
5882 Suppress warnings when a positional initializer is used to initialize
5883 a structure that has been marked with the @code{designated_init}
5887 Issue a warning when HSAIL cannot be emitted for the compiled function or
5892 @node Debugging Options
5893 @section Options for Debugging Your Program
5894 @cindex options, debugging
5895 @cindex debugging information options
5897 To tell GCC to emit extra information for use by a debugger, in almost
5898 all cases you need only to add @option{-g} to your other options.
5900 GCC allows you to use @option{-g} with
5901 @option{-O}. The shortcuts taken by optimized code may occasionally
5902 be surprising: some variables you declared may not exist
5903 at all; flow of control may briefly move where you did not expect it;
5904 some statements may not be executed because they compute constant
5905 results or their values are already at hand; some statements may
5906 execute in different places because they have been moved out of loops.
5907 Nevertheless it is possible to debug optimized output. This makes
5908 it reasonable to use the optimizer for programs that might have bugs.
5910 If you are not using some other optimization option, consider
5911 using @option{-Og} (@pxref{Optimize Options}) with @option{-g}.
5912 With no @option{-O} option at all, some compiler passes that collect
5913 information useful for debugging do not run at all, so that
5914 @option{-Og} may result in a better debugging experience.
5919 Produce debugging information in the operating system's native format
5920 (stabs, COFF, XCOFF, or DWARF)@. GDB can work with this debugging
5923 On most systems that use stabs format, @option{-g} enables use of extra
5924 debugging information that only GDB can use; this extra information
5925 makes debugging work better in GDB but probably makes other debuggers
5927 refuse to read the program. If you want to control for certain whether
5928 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
5929 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
5933 Produce debugging information for use by GDB@. This means to use the
5934 most expressive format available (DWARF, stabs, or the native format
5935 if neither of those are supported), including GDB extensions if at all
5939 @itemx -gdwarf-@var{version}
5941 Produce debugging information in DWARF format (if that is supported).
5942 The value of @var{version} may be either 2, 3, 4 or 5; the default version
5943 for most targets is 4. DWARF Version 5 is only experimental.
5945 Note that with DWARF Version 2, some ports require and always
5946 use some non-conflicting DWARF 3 extensions in the unwind tables.
5948 Version 4 may require GDB 7.0 and @option{-fvar-tracking-assignments}
5949 for maximum benefit.
5951 GCC no longer supports DWARF Version 1, which is substantially
5952 different than Version 2 and later. For historical reasons, some
5953 other DWARF-related options (including @option{-feliminate-dwarf2-dups}
5954 and @option{-fno-dwarf2-cfi-asm}) retain a reference to DWARF Version 2
5955 in their names, but apply to all currently-supported versions of DWARF.
5959 Produce debugging information in stabs format (if that is supported),
5960 without GDB extensions. This is the format used by DBX on most BSD
5961 systems. On MIPS, Alpha and System V Release 4 systems this option
5962 produces stabs debugging output that is not understood by DBX or SDB@.
5963 On System V Release 4 systems this option requires the GNU assembler.
5967 Produce debugging information in stabs format (if that is supported),
5968 using GNU extensions understood only by the GNU debugger (GDB)@. The
5969 use of these extensions is likely to make other debuggers crash or
5970 refuse to read the program.
5974 Produce debugging information in COFF format (if that is supported).
5975 This is the format used by SDB on most System V systems prior to
5980 Produce debugging information in XCOFF format (if that is supported).
5981 This is the format used by the DBX debugger on IBM RS/6000 systems.
5985 Produce debugging information in XCOFF format (if that is supported),
5986 using GNU extensions understood only by the GNU debugger (GDB)@. The
5987 use of these extensions is likely to make other debuggers crash or
5988 refuse to read the program, and may cause assemblers other than the GNU
5989 assembler (GAS) to fail with an error.
5993 Produce debugging information in Alpha/VMS debug format (if that is
5994 supported). This is the format used by DEBUG on Alpha/VMS systems.
5997 @itemx -ggdb@var{level}
5998 @itemx -gstabs@var{level}
5999 @itemx -gcoff@var{level}
6000 @itemx -gxcoff@var{level}
6001 @itemx -gvms@var{level}
6002 Request debugging information and also use @var{level} to specify how
6003 much information. The default level is 2.
6005 Level 0 produces no debug information at all. Thus, @option{-g0} negates
6008 Level 1 produces minimal information, enough for making backtraces in
6009 parts of the program that you don't plan to debug. This includes
6010 descriptions of functions and external variables, and line number
6011 tables, but no information about local variables.
6013 Level 3 includes extra information, such as all the macro definitions
6014 present in the program. Some debuggers support macro expansion when
6015 you use @option{-g3}.
6017 @option{-gdwarf} does not accept a concatenated debug level, to avoid
6018 confusion with @option{-gdwarf-@var{level}}.
6019 Instead use an additional @option{-g@var{level}} option to change the
6020 debug level for DWARF.
6022 @item -feliminate-unused-debug-symbols
6023 @opindex feliminate-unused-debug-symbols
6024 Produce debugging information in stabs format (if that is supported),
6025 for only symbols that are actually used.
6027 @item -femit-class-debug-always
6028 @opindex femit-class-debug-always
6029 Instead of emitting debugging information for a C++ class in only one
6030 object file, emit it in all object files using the class. This option
6031 should be used only with debuggers that are unable to handle the way GCC
6032 normally emits debugging information for classes because using this
6033 option increases the size of debugging information by as much as a
6036 @item -fno-merge-debug-strings
6037 @opindex fmerge-debug-strings
6038 @opindex fno-merge-debug-strings
6039 Direct the linker to not merge together strings in the debugging
6040 information that are identical in different object files. Merging is
6041 not supported by all assemblers or linkers. Merging decreases the size
6042 of the debug information in the output file at the cost of increasing
6043 link processing time. Merging is enabled by default.
6045 @item -fdebug-prefix-map=@var{old}=@var{new}
6046 @opindex fdebug-prefix-map
6047 When compiling files in directory @file{@var{old}}, record debugging
6048 information describing them as in @file{@var{new}} instead.
6050 @item -fvar-tracking
6051 @opindex fvar-tracking
6052 Run variable tracking pass. It computes where variables are stored at each
6053 position in code. Better debugging information is then generated
6054 (if the debugging information format supports this information).
6056 It is enabled by default when compiling with optimization (@option{-Os},
6057 @option{-O}, @option{-O2}, @dots{}), debugging information (@option{-g}) and
6058 the debug info format supports it.
6060 @item -fvar-tracking-assignments
6061 @opindex fvar-tracking-assignments
6062 @opindex fno-var-tracking-assignments
6063 Annotate assignments to user variables early in the compilation and
6064 attempt to carry the annotations over throughout the compilation all the
6065 way to the end, in an attempt to improve debug information while
6066 optimizing. Use of @option{-gdwarf-4} is recommended along with it.
6068 It can be enabled even if var-tracking is disabled, in which case
6069 annotations are created and maintained, but discarded at the end.
6070 By default, this flag is enabled together with @option{-fvar-tracking},
6071 except when selective scheduling is enabled.
6074 @opindex gsplit-dwarf
6075 Separate as much DWARF debugging information as possible into a
6076 separate output file with the extension @file{.dwo}. This option allows
6077 the build system to avoid linking files with debug information. To
6078 be useful, this option requires a debugger capable of reading @file{.dwo}
6083 Generate DWARF @code{.debug_pubnames} and @code{.debug_pubtypes} sections.
6085 @item -ggnu-pubnames
6086 @opindex ggnu-pubnames
6087 Generate @code{.debug_pubnames} and @code{.debug_pubtypes} sections in a format
6088 suitable for conversion into a GDB@ index. This option is only useful
6089 with a linker that can produce GDB@ index version 7.
6091 @item -fdebug-types-section
6092 @opindex fdebug-types-section
6093 @opindex fno-debug-types-section
6094 When using DWARF Version 4 or higher, type DIEs can be put into
6095 their own @code{.debug_types} section instead of making them part of the
6096 @code{.debug_info} section. It is more efficient to put them in a separate
6097 comdat sections since the linker can then remove duplicates.
6098 But not all DWARF consumers support @code{.debug_types} sections yet
6099 and on some objects @code{.debug_types} produces larger instead of smaller
6100 debugging information.
6102 @item -grecord-gcc-switches
6103 @item -gno-record-gcc-switches
6104 @opindex grecord-gcc-switches
6105 @opindex gno-record-gcc-switches
6106 This switch causes the command-line options used to invoke the
6107 compiler that may affect code generation to be appended to the
6108 DW_AT_producer attribute in DWARF debugging information. The options
6109 are concatenated with spaces separating them from each other and from
6110 the compiler version.
6111 It is enabled by default.
6112 See also @option{-frecord-gcc-switches} for another
6113 way of storing compiler options into the object file.
6115 @item -gstrict-dwarf
6116 @opindex gstrict-dwarf
6117 Disallow using extensions of later DWARF standard version than selected
6118 with @option{-gdwarf-@var{version}}. On most targets using non-conflicting
6119 DWARF extensions from later standard versions is allowed.
6121 @item -gno-strict-dwarf
6122 @opindex gno-strict-dwarf
6123 Allow using extensions of later DWARF standard version than selected with
6124 @option{-gdwarf-@var{version}}.
6126 @item -gz@r{[}=@var{type}@r{]}
6128 Produce compressed debug sections in DWARF format, if that is supported.
6129 If @var{type} is not given, the default type depends on the capabilities
6130 of the assembler and linker used. @var{type} may be one of
6131 @samp{none} (don't compress debug sections), @samp{zlib} (use zlib
6132 compression in ELF gABI format), or @samp{zlib-gnu} (use zlib
6133 compression in traditional GNU format). If the linker doesn't support
6134 writing compressed debug sections, the option is rejected. Otherwise,
6135 if the assembler does not support them, @option{-gz} is silently ignored
6136 when producing object files.
6138 @item -feliminate-dwarf2-dups
6139 @opindex feliminate-dwarf2-dups
6140 Compress DWARF debugging information by eliminating duplicated
6141 information about each symbol. This option only makes sense when
6142 generating DWARF debugging information.
6144 @item -femit-struct-debug-baseonly
6145 @opindex femit-struct-debug-baseonly
6146 Emit debug information for struct-like types
6147 only when the base name of the compilation source file
6148 matches the base name of file in which the struct is defined.
6150 This option substantially reduces the size of debugging information,
6151 but at significant potential loss in type information to the debugger.
6152 See @option{-femit-struct-debug-reduced} for a less aggressive option.
6153 See @option{-femit-struct-debug-detailed} for more detailed control.
6155 This option works only with DWARF debug output.
6157 @item -femit-struct-debug-reduced
6158 @opindex femit-struct-debug-reduced
6159 Emit debug information for struct-like types
6160 only when the base name of the compilation source file
6161 matches the base name of file in which the type is defined,
6162 unless the struct is a template or defined in a system header.
6164 This option significantly reduces the size of debugging information,
6165 with some potential loss in type information to the debugger.
6166 See @option{-femit-struct-debug-baseonly} for a more aggressive option.
6167 See @option{-femit-struct-debug-detailed} for more detailed control.
6169 This option works only with DWARF debug output.
6171 @item -femit-struct-debug-detailed@r{[}=@var{spec-list}@r{]}
6172 @opindex femit-struct-debug-detailed
6173 Specify the struct-like types
6174 for which the compiler generates debug information.
6175 The intent is to reduce duplicate struct debug information
6176 between different object files within the same program.
6178 This option is a detailed version of
6179 @option{-femit-struct-debug-reduced} and @option{-femit-struct-debug-baseonly},
6180 which serves for most needs.
6182 A specification has the syntax@*
6183 [@samp{dir:}|@samp{ind:}][@samp{ord:}|@samp{gen:}](@samp{any}|@samp{sys}|@samp{base}|@samp{none})
6185 The optional first word limits the specification to
6186 structs that are used directly (@samp{dir:}) or used indirectly (@samp{ind:}).
6187 A struct type is used directly when it is the type of a variable, member.
6188 Indirect uses arise through pointers to structs.
6189 That is, when use of an incomplete struct is valid, the use is indirect.
6191 @samp{struct one direct; struct two * indirect;}.
6193 The optional second word limits the specification to
6194 ordinary structs (@samp{ord:}) or generic structs (@samp{gen:}).
6195 Generic structs are a bit complicated to explain.
6196 For C++, these are non-explicit specializations of template classes,
6197 or non-template classes within the above.
6198 Other programming languages have generics,
6199 but @option{-femit-struct-debug-detailed} does not yet implement them.
6201 The third word specifies the source files for those
6202 structs for which the compiler should emit debug information.
6203 The values @samp{none} and @samp{any} have the normal meaning.
6204 The value @samp{base} means that
6205 the base of name of the file in which the type declaration appears
6206 must match the base of the name of the main compilation file.
6207 In practice, this means that when compiling @file{foo.c}, debug information
6208 is generated for types declared in that file and @file{foo.h},
6209 but not other header files.
6210 The value @samp{sys} means those types satisfying @samp{base}
6211 or declared in system or compiler headers.
6213 You may need to experiment to determine the best settings for your application.
6215 The default is @option{-femit-struct-debug-detailed=all}.
6217 This option works only with DWARF debug output.
6219 @item -fno-dwarf2-cfi-asm
6220 @opindex fdwarf2-cfi-asm
6221 @opindex fno-dwarf2-cfi-asm
6222 Emit DWARF unwind info as compiler generated @code{.eh_frame} section
6223 instead of using GAS @code{.cfi_*} directives.
6225 @item -fno-eliminate-unused-debug-types
6226 @opindex feliminate-unused-debug-types
6227 @opindex fno-eliminate-unused-debug-types
6228 Normally, when producing DWARF output, GCC avoids producing debug symbol
6229 output for types that are nowhere used in the source file being compiled.
6230 Sometimes it is useful to have GCC emit debugging
6231 information for all types declared in a compilation
6232 unit, regardless of whether or not they are actually used
6233 in that compilation unit, for example
6234 if, in the debugger, you want to cast a value to a type that is
6235 not actually used in your program (but is declared). More often,
6236 however, this results in a significant amount of wasted space.
6239 @node Optimize Options
6240 @section Options That Control Optimization
6241 @cindex optimize options
6242 @cindex options, optimization
6244 These options control various sorts of optimizations.
6246 Without any optimization option, the compiler's goal is to reduce the
6247 cost of compilation and to make debugging produce the expected
6248 results. Statements are independent: if you stop the program with a
6249 breakpoint between statements, you can then assign a new value to any
6250 variable or change the program counter to any other statement in the
6251 function and get exactly the results you expect from the source
6254 Turning on optimization flags makes the compiler attempt to improve
6255 the performance and/or code size at the expense of compilation time
6256 and possibly the ability to debug the program.
6258 The compiler performs optimization based on the knowledge it has of the
6259 program. Compiling multiple files at once to a single output file mode allows
6260 the compiler to use information gained from all of the files when compiling
6263 Not all optimizations are controlled directly by a flag. Only
6264 optimizations that have a flag are listed in this section.
6266 Most optimizations are only enabled if an @option{-O} level is set on
6267 the command line. Otherwise they are disabled, even if individual
6268 optimization flags are specified.
6270 Depending on the target and how GCC was configured, a slightly different
6271 set of optimizations may be enabled at each @option{-O} level than
6272 those listed here. You can invoke GCC with @option{-Q --help=optimizers}
6273 to find out the exact set of optimizations that are enabled at each level.
6274 @xref{Overall Options}, for examples.
6281 Optimize. Optimizing compilation takes somewhat more time, and a lot
6282 more memory for a large function.
6284 With @option{-O}, the compiler tries to reduce code size and execution
6285 time, without performing any optimizations that take a great deal of
6288 @option{-O} turns on the following optimization flags:
6291 -fbranch-count-reg @gol
6292 -fcombine-stack-adjustments @gol
6294 -fcprop-registers @gol
6297 -fdelayed-branch @gol
6299 -fforward-propagate @gol
6300 -fguess-branch-probability @gol
6301 -fif-conversion2 @gol
6302 -fif-conversion @gol
6303 -finline-functions-called-once @gol
6304 -fipa-pure-const @gol
6306 -fipa-reference @gol
6307 -fmerge-constants @gol
6308 -fmove-loop-invariants @gol
6309 -freorder-blocks @gol
6311 -fsplit-wide-types @gol
6317 -ftree-coalesce-vars @gol
6318 -ftree-copy-prop @gol
6320 -ftree-dominator-opts @gol
6322 -ftree-forwprop @gol
6332 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
6333 where doing so does not interfere with debugging.
6337 Optimize even more. GCC performs nearly all supported optimizations
6338 that do not involve a space-speed tradeoff.
6339 As compared to @option{-O}, this option increases both compilation time
6340 and the performance of the generated code.
6342 @option{-O2} turns on all optimization flags specified by @option{-O}. It
6343 also turns on the following optimization flags:
6344 @gccoptlist{-fthread-jumps @gol
6345 -falign-functions -falign-jumps @gol
6346 -falign-loops -falign-labels @gol
6349 -fcse-follow-jumps -fcse-skip-blocks @gol
6350 -fdelete-null-pointer-checks @gol
6351 -fdevirtualize -fdevirtualize-speculatively @gol
6352 -fexpensive-optimizations @gol
6353 -fgcse -fgcse-lm @gol
6354 -fhoist-adjacent-loads @gol
6355 -finline-small-functions @gol
6356 -findirect-inlining @gol
6358 -fipa-cp-alignment @gol
6361 -fisolate-erroneous-paths-dereference @gol
6363 -foptimize-sibling-calls @gol
6364 -foptimize-strlen @gol
6365 -fpartial-inlining @gol
6367 -freorder-blocks-algorithm=stc @gol
6368 -freorder-blocks-and-partition -freorder-functions @gol
6369 -frerun-cse-after-loop @gol
6370 -fsched-interblock -fsched-spec @gol
6371 -fschedule-insns -fschedule-insns2 @gol
6372 -fstrict-aliasing -fstrict-overflow @gol
6373 -ftree-builtin-call-dce @gol
6374 -ftree-switch-conversion -ftree-tail-merge @gol
6379 Please note the warning under @option{-fgcse} about
6380 invoking @option{-O2} on programs that use computed gotos.
6384 Optimize yet more. @option{-O3} turns on all optimizations specified
6385 by @option{-O2} and also turns on the @option{-finline-functions},
6386 @option{-funswitch-loops}, @option{-fpredictive-commoning},
6387 @option{-fgcse-after-reload}, @option{-ftree-loop-vectorize},
6388 @option{-ftree-loop-distribute-patterns}, @option{-fsplit-paths}
6389 @option{-ftree-slp-vectorize}, @option{-fvect-cost-model},
6390 @option{-ftree-partial-pre}, @option{-fpeel-loops}
6391 and @option{-fipa-cp-clone} options.
6395 Reduce compilation time and make debugging produce the expected
6396 results. This is the default.
6400 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
6401 do not typically increase code size. It also performs further
6402 optimizations designed to reduce code size.
6404 @option{-Os} disables the following optimization flags:
6405 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
6406 -falign-labels -freorder-blocks -freorder-blocks-algorithm=stc @gol
6407 -freorder-blocks-and-partition -fprefetch-loop-arrays}
6411 Disregard strict standards compliance. @option{-Ofast} enables all
6412 @option{-O3} optimizations. It also enables optimizations that are not
6413 valid for all standard-compliant programs.
6414 It turns on @option{-ffast-math} and the Fortran-specific
6415 @option{-fno-protect-parens} and @option{-fstack-arrays}.
6419 Optimize debugging experience. @option{-Og} enables optimizations
6420 that do not interfere with debugging. It should be the optimization
6421 level of choice for the standard edit-compile-debug cycle, offering
6422 a reasonable level of optimization while maintaining fast compilation
6423 and a good debugging experience.
6426 If you use multiple @option{-O} options, with or without level numbers,
6427 the last such option is the one that is effective.
6429 Options of the form @option{-f@var{flag}} specify machine-independent
6430 flags. Most flags have both positive and negative forms; the negative
6431 form of @option{-ffoo} is @option{-fno-foo}. In the table
6432 below, only one of the forms is listed---the one you typically
6433 use. You can figure out the other form by either removing @samp{no-}
6436 The following options control specific optimizations. They are either
6437 activated by @option{-O} options or are related to ones that are. You
6438 can use the following flags in the rare cases when ``fine-tuning'' of
6439 optimizations to be performed is desired.
6442 @item -fno-defer-pop
6443 @opindex fno-defer-pop
6444 Always pop the arguments to each function call as soon as that function
6445 returns. For machines that must pop arguments after a function call,
6446 the compiler normally lets arguments accumulate on the stack for several
6447 function calls and pops them all at once.
6449 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6451 @item -fforward-propagate
6452 @opindex fforward-propagate
6453 Perform a forward propagation pass on RTL@. The pass tries to combine two
6454 instructions and checks if the result can be simplified. If loop unrolling
6455 is active, two passes are performed and the second is scheduled after
6458 This option is enabled by default at optimization levels @option{-O},
6459 @option{-O2}, @option{-O3}, @option{-Os}.
6461 @item -ffp-contract=@var{style}
6462 @opindex ffp-contract
6463 @option{-ffp-contract=off} disables floating-point expression contraction.
6464 @option{-ffp-contract=fast} enables floating-point expression contraction
6465 such as forming of fused multiply-add operations if the target has
6466 native support for them.
6467 @option{-ffp-contract=on} enables floating-point expression contraction
6468 if allowed by the language standard. This is currently not implemented
6469 and treated equal to @option{-ffp-contract=off}.
6471 The default is @option{-ffp-contract=fast}.
6473 @item -fomit-frame-pointer
6474 @opindex fomit-frame-pointer
6475 Don't keep the frame pointer in a register for functions that
6476 don't need one. This avoids the instructions to save, set up and
6477 restore frame pointers; it also makes an extra register available
6478 in many functions. @strong{It also makes debugging impossible on
6481 On some machines, such as the VAX, this flag has no effect, because
6482 the standard calling sequence automatically handles the frame pointer
6483 and nothing is saved by pretending it doesn't exist. The
6484 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
6485 whether a target machine supports this flag. @xref{Registers,,Register
6486 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
6488 The default setting (when not optimizing for
6489 size) for 32-bit GNU/Linux x86 and 32-bit Darwin x86 targets is
6490 @option{-fomit-frame-pointer}. You can configure GCC with the
6491 @option{--enable-frame-pointer} configure option to change the default.
6493 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6495 @item -foptimize-sibling-calls
6496 @opindex foptimize-sibling-calls
6497 Optimize sibling and tail recursive calls.
6499 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6501 @item -foptimize-strlen
6502 @opindex foptimize-strlen
6503 Optimize various standard C string functions (e.g. @code{strlen},
6504 @code{strchr} or @code{strcpy}) and
6505 their @code{_FORTIFY_SOURCE} counterparts into faster alternatives.
6507 Enabled at levels @option{-O2}, @option{-O3}.
6511 Do not expand any functions inline apart from those marked with
6512 the @code{always_inline} attribute. This is the default when not
6515 Single functions can be exempted from inlining by marking them
6516 with the @code{noinline} attribute.
6518 @item -finline-small-functions
6519 @opindex finline-small-functions
6520 Integrate functions into their callers when their body is smaller than expected
6521 function call code (so overall size of program gets smaller). The compiler
6522 heuristically decides which functions are simple enough to be worth integrating
6523 in this way. This inlining applies to all functions, even those not declared
6526 Enabled at level @option{-O2}.
6528 @item -findirect-inlining
6529 @opindex findirect-inlining
6530 Inline also indirect calls that are discovered to be known at compile
6531 time thanks to previous inlining. This option has any effect only
6532 when inlining itself is turned on by the @option{-finline-functions}
6533 or @option{-finline-small-functions} options.
6535 Enabled at level @option{-O2}.
6537 @item -finline-functions
6538 @opindex finline-functions
6539 Consider all functions for inlining, even if they are not declared inline.
6540 The compiler heuristically decides which functions are worth integrating
6543 If all calls to a given function are integrated, and the function is
6544 declared @code{static}, then the function is normally not output as
6545 assembler code in its own right.
6547 Enabled at level @option{-O3}.
6549 @item -finline-functions-called-once
6550 @opindex finline-functions-called-once
6551 Consider all @code{static} functions called once for inlining into their
6552 caller even if they are not marked @code{inline}. If a call to a given
6553 function is integrated, then the function is not output as assembler code
6556 Enabled at levels @option{-O1}, @option{-O2}, @option{-O3} and @option{-Os}.
6558 @item -fearly-inlining
6559 @opindex fearly-inlining
6560 Inline functions marked by @code{always_inline} and functions whose body seems
6561 smaller than the function call overhead early before doing
6562 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
6563 makes profiling significantly cheaper and usually inlining faster on programs
6564 having large chains of nested wrapper functions.
6570 Perform interprocedural scalar replacement of aggregates, removal of
6571 unused parameters and replacement of parameters passed by reference
6572 by parameters passed by value.
6574 Enabled at levels @option{-O2}, @option{-O3} and @option{-Os}.
6576 @item -finline-limit=@var{n}
6577 @opindex finline-limit
6578 By default, GCC limits the size of functions that can be inlined. This flag
6579 allows coarse control of this limit. @var{n} is the size of functions that
6580 can be inlined in number of pseudo instructions.
6582 Inlining is actually controlled by a number of parameters, which may be
6583 specified individually by using @option{--param @var{name}=@var{value}}.
6584 The @option{-finline-limit=@var{n}} option sets some of these parameters
6588 @item max-inline-insns-single
6589 is set to @var{n}/2.
6590 @item max-inline-insns-auto
6591 is set to @var{n}/2.
6594 See below for a documentation of the individual
6595 parameters controlling inlining and for the defaults of these parameters.
6597 @emph{Note:} there may be no value to @option{-finline-limit} that results
6598 in default behavior.
6600 @emph{Note:} pseudo instruction represents, in this particular context, an
6601 abstract measurement of function's size. In no way does it represent a count
6602 of assembly instructions and as such its exact meaning might change from one
6603 release to an another.
6605 @item -fno-keep-inline-dllexport
6606 @opindex fno-keep-inline-dllexport
6607 This is a more fine-grained version of @option{-fkeep-inline-functions},
6608 which applies only to functions that are declared using the @code{dllexport}
6609 attribute or declspec (@xref{Function Attributes,,Declaring Attributes of
6612 @item -fkeep-inline-functions
6613 @opindex fkeep-inline-functions
6614 In C, emit @code{static} functions that are declared @code{inline}
6615 into the object file, even if the function has been inlined into all
6616 of its callers. This switch does not affect functions using the
6617 @code{extern inline} extension in GNU C90@. In C++, emit any and all
6618 inline functions into the object file.
6620 @item -fkeep-static-functions
6621 @opindex fkeep-static-functions
6622 Emit @code{static} functions into the object file, even if the function
6625 @item -fkeep-static-consts
6626 @opindex fkeep-static-consts
6627 Emit variables declared @code{static const} when optimization isn't turned
6628 on, even if the variables aren't referenced.
6630 GCC enables this option by default. If you want to force the compiler to
6631 check if a variable is referenced, regardless of whether or not
6632 optimization is turned on, use the @option{-fno-keep-static-consts} option.
6634 @item -fmerge-constants
6635 @opindex fmerge-constants
6636 Attempt to merge identical constants (string constants and floating-point
6637 constants) across compilation units.
6639 This option is the default for optimized compilation if the assembler and
6640 linker support it. Use @option{-fno-merge-constants} to inhibit this
6643 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6645 @item -fmerge-all-constants
6646 @opindex fmerge-all-constants
6647 Attempt to merge identical constants and identical variables.
6649 This option implies @option{-fmerge-constants}. In addition to
6650 @option{-fmerge-constants} this considers e.g.@: even constant initialized
6651 arrays or initialized constant variables with integral or floating-point
6652 types. Languages like C or C++ require each variable, including multiple
6653 instances of the same variable in recursive calls, to have distinct locations,
6654 so using this option results in non-conforming
6657 @item -fmodulo-sched
6658 @opindex fmodulo-sched
6659 Perform swing modulo scheduling immediately before the first scheduling
6660 pass. This pass looks at innermost loops and reorders their
6661 instructions by overlapping different iterations.
6663 @item -fmodulo-sched-allow-regmoves
6664 @opindex fmodulo-sched-allow-regmoves
6665 Perform more aggressive SMS-based modulo scheduling with register moves
6666 allowed. By setting this flag certain anti-dependences edges are
6667 deleted, which triggers the generation of reg-moves based on the
6668 life-range analysis. This option is effective only with
6669 @option{-fmodulo-sched} enabled.
6671 @item -fno-branch-count-reg
6672 @opindex fno-branch-count-reg
6673 Avoid running a pass scanning for opportunities to use ``decrement and
6674 branch'' instructions on a count register instead of generating sequences
6675 of instructions that decrement a register, compare it against zero, and
6676 then branch based upon the result. This option is only meaningful on
6677 architectures that support such instructions, which include x86, PowerPC,
6678 IA-64 and S/390. Note that the @option{-fno-branch-count-reg} option
6679 doesn't remove the decrement and branch instructions from the generated
6680 instruction stream introduced by other optimization passes.
6682 Enabled by default at @option{-O1} and higher.
6684 The default is @option{-fbranch-count-reg}.
6686 @item -fno-function-cse
6687 @opindex fno-function-cse
6688 Do not put function addresses in registers; make each instruction that
6689 calls a constant function contain the function's address explicitly.
6691 This option results in less efficient code, but some strange hacks
6692 that alter the assembler output may be confused by the optimizations
6693 performed when this option is not used.
6695 The default is @option{-ffunction-cse}
6697 @item -fno-zero-initialized-in-bss
6698 @opindex fno-zero-initialized-in-bss
6699 If the target supports a BSS section, GCC by default puts variables that
6700 are initialized to zero into BSS@. This can save space in the resulting
6703 This option turns off this behavior because some programs explicitly
6704 rely on variables going to the data section---e.g., so that the
6705 resulting executable can find the beginning of that section and/or make
6706 assumptions based on that.
6708 The default is @option{-fzero-initialized-in-bss}.
6710 @item -fthread-jumps
6711 @opindex fthread-jumps
6712 Perform optimizations that check to see if a jump branches to a
6713 location where another comparison subsumed by the first is found. If
6714 so, the first branch is redirected to either the destination of the
6715 second branch or a point immediately following it, depending on whether
6716 the condition is known to be true or false.
6718 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6720 @item -fsplit-wide-types
6721 @opindex fsplit-wide-types
6722 When using a type that occupies multiple registers, such as @code{long
6723 long} on a 32-bit system, split the registers apart and allocate them
6724 independently. This normally generates better code for those types,
6725 but may make debugging more difficult.
6727 Enabled at levels @option{-O}, @option{-O2}, @option{-O3},
6730 @item -fcse-follow-jumps
6731 @opindex fcse-follow-jumps
6732 In common subexpression elimination (CSE), scan through jump instructions
6733 when the target of the jump is not reached by any other path. For
6734 example, when CSE encounters an @code{if} statement with an
6735 @code{else} clause, CSE follows the jump when the condition
6738 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6740 @item -fcse-skip-blocks
6741 @opindex fcse-skip-blocks
6742 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
6743 follow jumps that conditionally skip over blocks. When CSE
6744 encounters a simple @code{if} statement with no else clause,
6745 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
6746 body of the @code{if}.
6748 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6750 @item -frerun-cse-after-loop
6751 @opindex frerun-cse-after-loop
6752 Re-run common subexpression elimination after loop optimizations are
6755 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6759 Perform a global common subexpression elimination pass.
6760 This pass also performs global constant and copy propagation.
6762 @emph{Note:} When compiling a program using computed gotos, a GCC
6763 extension, you may get better run-time performance if you disable
6764 the global common subexpression elimination pass by adding
6765 @option{-fno-gcse} to the command line.
6767 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6771 When @option{-fgcse-lm} is enabled, global common subexpression elimination
6772 attempts to move loads that are only killed by stores into themselves. This
6773 allows a loop containing a load/store sequence to be changed to a load outside
6774 the loop, and a copy/store within the loop.
6776 Enabled by default when @option{-fgcse} is enabled.
6780 When @option{-fgcse-sm} is enabled, a store motion pass is run after
6781 global common subexpression elimination. This pass attempts to move
6782 stores out of loops. When used in conjunction with @option{-fgcse-lm},
6783 loops containing a load/store sequence can be changed to a load before
6784 the loop and a store after the loop.
6786 Not enabled at any optimization level.
6790 When @option{-fgcse-las} is enabled, the global common subexpression
6791 elimination pass eliminates redundant loads that come after stores to the
6792 same memory location (both partial and full redundancies).
6794 Not enabled at any optimization level.
6796 @item -fgcse-after-reload
6797 @opindex fgcse-after-reload
6798 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
6799 pass is performed after reload. The purpose of this pass is to clean up
6802 @item -faggressive-loop-optimizations
6803 @opindex faggressive-loop-optimizations
6804 This option tells the loop optimizer to use language constraints to
6805 derive bounds for the number of iterations of a loop. This assumes that
6806 loop code does not invoke undefined behavior by for example causing signed
6807 integer overflows or out-of-bound array accesses. The bounds for the
6808 number of iterations of a loop are used to guide loop unrolling and peeling
6809 and loop exit test optimizations.
6810 This option is enabled by default.
6812 @item -funsafe-loop-optimizations
6813 @opindex funsafe-loop-optimizations
6814 This option tells the loop optimizer to assume that loop indices do not
6815 overflow, and that loops with nontrivial exit condition are not
6816 infinite. This enables a wider range of loop optimizations even if
6817 the loop optimizer itself cannot prove that these assumptions are valid.
6818 If you use @option{-Wunsafe-loop-optimizations}, the compiler warns you
6819 if it finds this kind of loop.
6821 @item -funconstrained-commons
6822 @opindex funconstrained-commons
6823 This option tells the compiler that variables declared in common blocks
6824 (e.g. Fortran) may later be overridden with longer trailing arrays. This
6825 prevents certain optimizations that depend on knowing the array bounds.
6827 @item -fcrossjumping
6828 @opindex fcrossjumping
6829 Perform cross-jumping transformation.
6830 This transformation unifies equivalent code and saves code size. The
6831 resulting code may or may not perform better than without cross-jumping.
6833 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6835 @item -fauto-inc-dec
6836 @opindex fauto-inc-dec
6837 Combine increments or decrements of addresses with memory accesses.
6838 This pass is always skipped on architectures that do not have
6839 instructions to support this. Enabled by default at @option{-O} and
6840 higher on architectures that support this.
6844 Perform dead code elimination (DCE) on RTL@.
6845 Enabled by default at @option{-O} and higher.
6849 Perform dead store elimination (DSE) on RTL@.
6850 Enabled by default at @option{-O} and higher.
6852 @item -fif-conversion
6853 @opindex fif-conversion
6854 Attempt to transform conditional jumps into branch-less equivalents. This
6855 includes use of conditional moves, min, max, set flags and abs instructions, and
6856 some tricks doable by standard arithmetics. The use of conditional execution
6857 on chips where it is available is controlled by @option{-fif-conversion2}.
6859 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6861 @item -fif-conversion2
6862 @opindex fif-conversion2
6863 Use conditional execution (where available) to transform conditional jumps into
6864 branch-less equivalents.
6866 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
6868 @item -fdeclone-ctor-dtor
6869 @opindex fdeclone-ctor-dtor
6870 The C++ ABI requires multiple entry points for constructors and
6871 destructors: one for a base subobject, one for a complete object, and
6872 one for a virtual destructor that calls operator delete afterwards.
6873 For a hierarchy with virtual bases, the base and complete variants are
6874 clones, which means two copies of the function. With this option, the
6875 base and complete variants are changed to be thunks that call a common
6878 Enabled by @option{-Os}.
6880 @item -fdelete-null-pointer-checks
6881 @opindex fdelete-null-pointer-checks
6882 Assume that programs cannot safely dereference null pointers, and that
6883 no code or data element resides at address zero.
6884 This option enables simple constant
6885 folding optimizations at all optimization levels. In addition, other
6886 optimization passes in GCC use this flag to control global dataflow
6887 analyses that eliminate useless checks for null pointers; these assume
6888 that a memory access to address zero always results in a trap, so
6889 that if a pointer is checked after it has already been dereferenced,
6892 Note however that in some environments this assumption is not true.
6893 Use @option{-fno-delete-null-pointer-checks} to disable this optimization
6894 for programs that depend on that behavior.
6896 This option is enabled by default on most targets. On Nios II ELF, it
6897 defaults to off. On AVR and CR16, this option is completely disabled.
6899 Passes that use the dataflow information
6900 are enabled independently at different optimization levels.
6902 @item -fdevirtualize
6903 @opindex fdevirtualize
6904 Attempt to convert calls to virtual functions to direct calls. This
6905 is done both within a procedure and interprocedurally as part of
6906 indirect inlining (@option{-findirect-inlining}) and interprocedural constant
6907 propagation (@option{-fipa-cp}).
6908 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6910 @item -fdevirtualize-speculatively
6911 @opindex fdevirtualize-speculatively
6912 Attempt to convert calls to virtual functions to speculative direct calls.
6913 Based on the analysis of the type inheritance graph, determine for a given call
6914 the set of likely targets. If the set is small, preferably of size 1, change
6915 the call into a conditional deciding between direct and indirect calls. The
6916 speculative calls enable more optimizations, such as inlining. When they seem
6917 useless after further optimization, they are converted back into original form.
6919 @item -fdevirtualize-at-ltrans
6920 @opindex fdevirtualize-at-ltrans
6921 Stream extra information needed for aggressive devirtualization when running
6922 the link-time optimizer in local transformation mode.
6923 This option enables more devirtualization but
6924 significantly increases the size of streamed data. For this reason it is
6925 disabled by default.
6927 @item -fexpensive-optimizations
6928 @opindex fexpensive-optimizations
6929 Perform a number of minor optimizations that are relatively expensive.
6931 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
6935 Attempt to remove redundant extension instructions. This is especially
6936 helpful for the x86-64 architecture, which implicitly zero-extends in 64-bit
6937 registers after writing to their lower 32-bit half.
6939 Enabled for Alpha, AArch64 and x86 at levels @option{-O2},
6940 @option{-O3}, @option{-Os}.
6942 @item -fno-lifetime-dse
6943 @opindex fno-lifetime-dse
6944 In C++ the value of an object is only affected by changes within its
6945 lifetime: when the constructor begins, the object has an indeterminate
6946 value, and any changes during the lifetime of the object are dead when
6947 the object is destroyed. Normally dead store elimination will take
6948 advantage of this; if your code relies on the value of the object
6949 storage persisting beyond the lifetime of the object, you can use this
6950 flag to disable this optimization. To preserve stores before the
6951 constructor starts (e.g. because your operator new clears the object
6952 storage) but still treat the object as dead after the destructor you,
6953 can use @option{-flifetime-dse=1}. The default behavior can be
6954 explicitly selected with @option{-flifetime-dse=2}.
6955 @option{-flifetime-dse=0} is equivalent to @option{-fno-lifetime-dse}.
6957 @item -flive-range-shrinkage
6958 @opindex flive-range-shrinkage
6959 Attempt to decrease register pressure through register live range
6960 shrinkage. This is helpful for fast processors with small or moderate
6963 @item -fira-algorithm=@var{algorithm}
6964 @opindex fira-algorithm
6965 Use the specified coloring algorithm for the integrated register
6966 allocator. The @var{algorithm} argument can be @samp{priority}, which
6967 specifies Chow's priority coloring, or @samp{CB}, which specifies
6968 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not implemented
6969 for all architectures, but for those targets that do support it, it is
6970 the default because it generates better code.
6972 @item -fira-region=@var{region}
6973 @opindex fira-region
6974 Use specified regions for the integrated register allocator. The
6975 @var{region} argument should be one of the following:
6980 Use all loops as register allocation regions.
6981 This can give the best results for machines with a small and/or
6982 irregular register set.
6985 Use all loops except for loops with small register pressure
6986 as the regions. This value usually gives
6987 the best results in most cases and for most architectures,
6988 and is enabled by default when compiling with optimization for speed
6989 (@option{-O}, @option{-O2}, @dots{}).
6992 Use all functions as a single region.
6993 This typically results in the smallest code size, and is enabled by default for
6994 @option{-Os} or @option{-O0}.
6998 @item -fira-hoist-pressure
6999 @opindex fira-hoist-pressure
7000 Use IRA to evaluate register pressure in the code hoisting pass for
7001 decisions to hoist expressions. This option usually results in smaller
7002 code, but it can slow the compiler down.
7004 This option is enabled at level @option{-Os} for all targets.
7006 @item -fira-loop-pressure
7007 @opindex fira-loop-pressure
7008 Use IRA to evaluate register pressure in loops for decisions to move
7009 loop invariants. This option usually results in generation
7010 of faster and smaller code on machines with large register files (>= 32
7011 registers), but it can slow the compiler down.
7013 This option is enabled at level @option{-O3} for some targets.
7015 @item -fno-ira-share-save-slots
7016 @opindex fno-ira-share-save-slots
7017 Disable sharing of stack slots used for saving call-used hard
7018 registers living through a call. Each hard register gets a
7019 separate stack slot, and as a result function stack frames are
7022 @item -fno-ira-share-spill-slots
7023 @opindex fno-ira-share-spill-slots
7024 Disable sharing of stack slots allocated for pseudo-registers. Each
7025 pseudo-register that does not get a hard register gets a separate
7026 stack slot, and as a result function stack frames are larger.
7030 Enable CFG-sensitive rematerialization in LRA. Instead of loading
7031 values of spilled pseudos, LRA tries to rematerialize (recalculate)
7032 values if it is profitable.
7034 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7036 @item -fdelayed-branch
7037 @opindex fdelayed-branch
7038 If supported for the target machine, attempt to reorder instructions
7039 to exploit instruction slots available after delayed branch
7042 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7044 @item -fschedule-insns
7045 @opindex fschedule-insns
7046 If supported for the target machine, attempt to reorder instructions to
7047 eliminate execution stalls due to required data being unavailable. This
7048 helps machines that have slow floating point or memory load instructions
7049 by allowing other instructions to be issued until the result of the load
7050 or floating-point instruction is required.
7052 Enabled at levels @option{-O2}, @option{-O3}.
7054 @item -fschedule-insns2
7055 @opindex fschedule-insns2
7056 Similar to @option{-fschedule-insns}, but requests an additional pass of
7057 instruction scheduling after register allocation has been done. This is
7058 especially useful on machines with a relatively small number of
7059 registers and where memory load instructions take more than one cycle.
7061 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7063 @item -fno-sched-interblock
7064 @opindex fno-sched-interblock
7065 Don't schedule instructions across basic blocks. This is normally
7066 enabled by default when scheduling before register allocation, i.e.@:
7067 with @option{-fschedule-insns} or at @option{-O2} or higher.
7069 @item -fno-sched-spec
7070 @opindex fno-sched-spec
7071 Don't allow speculative motion of non-load instructions. This is normally
7072 enabled by default when scheduling before register allocation, i.e.@:
7073 with @option{-fschedule-insns} or at @option{-O2} or higher.
7075 @item -fsched-pressure
7076 @opindex fsched-pressure
7077 Enable register pressure sensitive insn scheduling before register
7078 allocation. This only makes sense when scheduling before register
7079 allocation is enabled, i.e.@: with @option{-fschedule-insns} or at
7080 @option{-O2} or higher. Usage of this option can improve the
7081 generated code and decrease its size by preventing register pressure
7082 increase above the number of available hard registers and subsequent
7083 spills in register allocation.
7085 @item -fsched-spec-load
7086 @opindex fsched-spec-load
7087 Allow speculative motion of some load instructions. This only makes
7088 sense when scheduling before register allocation, i.e.@: with
7089 @option{-fschedule-insns} or at @option{-O2} or higher.
7091 @item -fsched-spec-load-dangerous
7092 @opindex fsched-spec-load-dangerous
7093 Allow speculative motion of more load instructions. This only makes
7094 sense when scheduling before register allocation, i.e.@: with
7095 @option{-fschedule-insns} or at @option{-O2} or higher.
7097 @item -fsched-stalled-insns
7098 @itemx -fsched-stalled-insns=@var{n}
7099 @opindex fsched-stalled-insns
7100 Define how many insns (if any) can be moved prematurely from the queue
7101 of stalled insns into the ready list during the second scheduling pass.
7102 @option{-fno-sched-stalled-insns} means that no insns are moved
7103 prematurely, @option{-fsched-stalled-insns=0} means there is no limit
7104 on how many queued insns can be moved prematurely.
7105 @option{-fsched-stalled-insns} without a value is equivalent to
7106 @option{-fsched-stalled-insns=1}.
7108 @item -fsched-stalled-insns-dep
7109 @itemx -fsched-stalled-insns-dep=@var{n}
7110 @opindex fsched-stalled-insns-dep
7111 Define how many insn groups (cycles) are examined for a dependency
7112 on a stalled insn that is a candidate for premature removal from the queue
7113 of stalled insns. This has an effect only during the second scheduling pass,
7114 and only if @option{-fsched-stalled-insns} is used.
7115 @option{-fno-sched-stalled-insns-dep} is equivalent to
7116 @option{-fsched-stalled-insns-dep=0}.
7117 @option{-fsched-stalled-insns-dep} without a value is equivalent to
7118 @option{-fsched-stalled-insns-dep=1}.
7120 @item -fsched2-use-superblocks
7121 @opindex fsched2-use-superblocks
7122 When scheduling after register allocation, use superblock scheduling.
7123 This allows motion across basic block boundaries,
7124 resulting in faster schedules. This option is experimental, as not all machine
7125 descriptions used by GCC model the CPU closely enough to avoid unreliable
7126 results from the algorithm.
7128 This only makes sense when scheduling after register allocation, i.e.@: with
7129 @option{-fschedule-insns2} or at @option{-O2} or higher.
7131 @item -fsched-group-heuristic
7132 @opindex fsched-group-heuristic
7133 Enable the group heuristic in the scheduler. This heuristic favors
7134 the instruction that belongs to a schedule group. This is enabled
7135 by default when scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7136 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7138 @item -fsched-critical-path-heuristic
7139 @opindex fsched-critical-path-heuristic
7140 Enable the critical-path heuristic in the scheduler. This heuristic favors
7141 instructions on the critical path. This is enabled by default when
7142 scheduling is enabled, i.e.@: with @option{-fschedule-insns}
7143 or @option{-fschedule-insns2} or at @option{-O2} or higher.
7145 @item -fsched-spec-insn-heuristic
7146 @opindex fsched-spec-insn-heuristic
7147 Enable the speculative instruction heuristic in the scheduler. This
7148 heuristic favors speculative instructions with greater dependency weakness.
7149 This is enabled by default when scheduling is enabled, i.e.@:
7150 with @option{-fschedule-insns} or @option{-fschedule-insns2}
7151 or at @option{-O2} or higher.
7153 @item -fsched-rank-heuristic
7154 @opindex fsched-rank-heuristic
7155 Enable the rank heuristic in the scheduler. This heuristic favors
7156 the instruction belonging to a basic block with greater size or frequency.
7157 This is enabled by default when scheduling is enabled, i.e.@:
7158 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7159 at @option{-O2} or higher.
7161 @item -fsched-last-insn-heuristic
7162 @opindex fsched-last-insn-heuristic
7163 Enable the last-instruction heuristic in the scheduler. This heuristic
7164 favors the instruction that is less dependent on the last instruction
7165 scheduled. This is enabled by default when scheduling is enabled,
7166 i.e.@: with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7167 at @option{-O2} or higher.
7169 @item -fsched-dep-count-heuristic
7170 @opindex fsched-dep-count-heuristic
7171 Enable the dependent-count heuristic in the scheduler. This heuristic
7172 favors the instruction that has more instructions depending on it.
7173 This is enabled by default when scheduling is enabled, i.e.@:
7174 with @option{-fschedule-insns} or @option{-fschedule-insns2} or
7175 at @option{-O2} or higher.
7177 @item -freschedule-modulo-scheduled-loops
7178 @opindex freschedule-modulo-scheduled-loops
7179 Modulo scheduling is performed before traditional scheduling. If a loop
7180 is modulo scheduled, later scheduling passes may change its schedule.
7181 Use this option to control that behavior.
7183 @item -fselective-scheduling
7184 @opindex fselective-scheduling
7185 Schedule instructions using selective scheduling algorithm. Selective
7186 scheduling runs instead of the first scheduler pass.
7188 @item -fselective-scheduling2
7189 @opindex fselective-scheduling2
7190 Schedule instructions using selective scheduling algorithm. Selective
7191 scheduling runs instead of the second scheduler pass.
7193 @item -fsel-sched-pipelining
7194 @opindex fsel-sched-pipelining
7195 Enable software pipelining of innermost loops during selective scheduling.
7196 This option has no effect unless one of @option{-fselective-scheduling} or
7197 @option{-fselective-scheduling2} is turned on.
7199 @item -fsel-sched-pipelining-outer-loops
7200 @opindex fsel-sched-pipelining-outer-loops
7201 When pipelining loops during selective scheduling, also pipeline outer loops.
7202 This option has no effect unless @option{-fsel-sched-pipelining} is turned on.
7204 @item -fsemantic-interposition
7205 @opindex fsemantic-interposition
7206 Some object formats, like ELF, allow interposing of symbols by the
7208 This means that for symbols exported from the DSO, the compiler cannot perform
7209 interprocedural propagation, inlining and other optimizations in anticipation
7210 that the function or variable in question may change. While this feature is
7211 useful, for example, to rewrite memory allocation functions by a debugging
7212 implementation, it is expensive in the terms of code quality.
7213 With @option{-fno-semantic-interposition} the compiler assumes that
7214 if interposition happens for functions the overwriting function will have
7215 precisely the same semantics (and side effects).
7216 Similarly if interposition happens
7217 for variables, the constructor of the variable will be the same. The flag
7218 has no effect for functions explicitly declared inline
7219 (where it is never allowed for interposition to change semantics)
7220 and for symbols explicitly declared weak.
7223 @opindex fshrink-wrap
7224 Emit function prologues only before parts of the function that need it,
7225 rather than at the top of the function. This flag is enabled by default at
7226 @option{-O} and higher.
7228 @item -fcaller-saves
7229 @opindex fcaller-saves
7230 Enable allocation of values to registers that are clobbered by
7231 function calls, by emitting extra instructions to save and restore the
7232 registers around such calls. Such allocation is done only when it
7233 seems to result in better code.
7235 This option is always enabled by default on certain machines, usually
7236 those which have no call-preserved registers to use instead.
7238 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7240 @item -fcombine-stack-adjustments
7241 @opindex fcombine-stack-adjustments
7242 Tracks stack adjustments (pushes and pops) and stack memory references
7243 and then tries to find ways to combine them.
7245 Enabled by default at @option{-O1} and higher.
7249 Use caller save registers for allocation if those registers are not used by
7250 any called function. In that case it is not necessary to save and restore
7251 them around calls. This is only possible if called functions are part of
7252 same compilation unit as current function and they are compiled before it.
7254 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7256 @item -fconserve-stack
7257 @opindex fconserve-stack
7258 Attempt to minimize stack usage. The compiler attempts to use less
7259 stack space, even if that makes the program slower. This option
7260 implies setting the @option{large-stack-frame} parameter to 100
7261 and the @option{large-stack-frame-growth} parameter to 400.
7263 @item -ftree-reassoc
7264 @opindex ftree-reassoc
7265 Perform reassociation on trees. This flag is enabled by default
7266 at @option{-O} and higher.
7270 Perform partial redundancy elimination (PRE) on trees. This flag is
7271 enabled by default at @option{-O2} and @option{-O3}.
7273 @item -ftree-partial-pre
7274 @opindex ftree-partial-pre
7275 Make partial redundancy elimination (PRE) more aggressive. This flag is
7276 enabled by default at @option{-O3}.
7278 @item -ftree-forwprop
7279 @opindex ftree-forwprop
7280 Perform forward propagation on trees. This flag is enabled by default
7281 at @option{-O} and higher.
7285 Perform full redundancy elimination (FRE) on trees. The difference
7286 between FRE and PRE is that FRE only considers expressions
7287 that are computed on all paths leading to the redundant computation.
7288 This analysis is faster than PRE, though it exposes fewer redundancies.
7289 This flag is enabled by default at @option{-O} and higher.
7291 @item -ftree-phiprop
7292 @opindex ftree-phiprop
7293 Perform hoisting of loads from conditional pointers on trees. This
7294 pass is enabled by default at @option{-O} and higher.
7296 @item -fhoist-adjacent-loads
7297 @opindex fhoist-adjacent-loads
7298 Speculatively hoist loads from both branches of an if-then-else if the
7299 loads are from adjacent locations in the same structure and the target
7300 architecture has a conditional move instruction. This flag is enabled
7301 by default at @option{-O2} and higher.
7303 @item -ftree-copy-prop
7304 @opindex ftree-copy-prop
7305 Perform copy propagation on trees. This pass eliminates unnecessary
7306 copy operations. This flag is enabled by default at @option{-O} and
7309 @item -fipa-pure-const
7310 @opindex fipa-pure-const
7311 Discover which functions are pure or constant.
7312 Enabled by default at @option{-O} and higher.
7314 @item -fipa-reference
7315 @opindex fipa-reference
7316 Discover which static variables do not escape the
7318 Enabled by default at @option{-O} and higher.
7322 Perform interprocedural pointer analysis and interprocedural modification
7323 and reference analysis. This option can cause excessive memory and
7324 compile-time usage on large compilation units. It is not enabled by
7325 default at any optimization level.
7328 @opindex fipa-profile
7329 Perform interprocedural profile propagation. The functions called only from
7330 cold functions are marked as cold. Also functions executed once (such as
7331 @code{cold}, @code{noreturn}, static constructors or destructors) are identified. Cold
7332 functions and loop less parts of functions executed once are then optimized for
7334 Enabled by default at @option{-O} and higher.
7338 Perform interprocedural constant propagation.
7339 This optimization analyzes the program to determine when values passed
7340 to functions are constants and then optimizes accordingly.
7341 This optimization can substantially increase performance
7342 if the application has constants passed to functions.
7343 This flag is enabled by default at @option{-O2}, @option{-Os} and @option{-O3}.
7345 @item -fipa-cp-clone
7346 @opindex fipa-cp-clone
7347 Perform function cloning to make interprocedural constant propagation stronger.
7348 When enabled, interprocedural constant propagation performs function cloning
7349 when externally visible function can be called with constant arguments.
7350 Because this optimization can create multiple copies of functions,
7351 it may significantly increase code size
7352 (see @option{--param ipcp-unit-growth=@var{value}}).
7353 This flag is enabled by default at @option{-O3}.
7355 @item -fipa-cp-alignment
7356 @opindex -fipa-cp-alignment
7357 When enabled, this optimization propagates alignment of function
7358 parameters to support better vectorization and string operations.
7360 This flag is enabled by default at @option{-O2} and @option{-Os}. It
7361 requires that @option{-fipa-cp} is enabled.
7365 Perform Identical Code Folding for functions and read-only variables.
7366 The optimization reduces code size and may disturb unwind stacks by replacing
7367 a function by equivalent one with a different name. The optimization works
7368 more effectively with link time optimization enabled.
7370 Nevertheless the behavior is similar to Gold Linker ICF optimization, GCC ICF
7371 works on different levels and thus the optimizations are not same - there are
7372 equivalences that are found only by GCC and equivalences found only by Gold.
7374 This flag is enabled by default at @option{-O2} and @option{-Os}.
7376 @item -fisolate-erroneous-paths-dereference
7377 @opindex fisolate-erroneous-paths-dereference
7378 Detect paths that trigger erroneous or undefined behavior due to
7379 dereferencing a null pointer. Isolate those paths from the main control
7380 flow and turn the statement with erroneous or undefined behavior into a trap.
7381 This flag is enabled by default at @option{-O2} and higher and depends on
7382 @option{-fdelete-null-pointer-checks} also being enabled.
7384 @item -fisolate-erroneous-paths-attribute
7385 @opindex fisolate-erroneous-paths-attribute
7386 Detect paths that trigger erroneous or undefined behavior due a null value
7387 being used in a way forbidden by a @code{returns_nonnull} or @code{nonnull}
7388 attribute. Isolate those paths from the main control flow and turn the
7389 statement with erroneous or undefined behavior into a trap. This is not
7390 currently enabled, but may be enabled by @option{-O2} in the future.
7394 Perform forward store motion on trees. This flag is
7395 enabled by default at @option{-O} and higher.
7397 @item -ftree-bit-ccp
7398 @opindex ftree-bit-ccp
7399 Perform sparse conditional bit constant propagation on trees and propagate
7400 pointer alignment information.
7401 This pass only operates on local scalar variables and is enabled by default
7402 at @option{-O} and higher. It requires that @option{-ftree-ccp} is enabled.
7406 Perform sparse conditional constant propagation (CCP) on trees. This
7407 pass only operates on local scalar variables and is enabled by default
7408 at @option{-O} and higher.
7410 @item -fssa-backprop
7411 @opindex fssa-backprop
7412 Propagate information about uses of a value up the definition chain
7413 in order to simplify the definitions. For example, this pass strips
7414 sign operations if the sign of a value never matters. The flag is
7415 enabled by default at @option{-O} and higher.
7418 @opindex fssa-phiopt
7419 Perform pattern matching on SSA PHI nodes to optimize conditional
7420 code. This pass is enabled by default at @option{-O} and higher.
7422 @item -ftree-switch-conversion
7423 @opindex ftree-switch-conversion
7424 Perform conversion of simple initializations in a switch to
7425 initializations from a scalar array. This flag is enabled by default
7426 at @option{-O2} and higher.
7428 @item -ftree-tail-merge
7429 @opindex ftree-tail-merge
7430 Look for identical code sequences. When found, replace one with a jump to the
7431 other. This optimization is known as tail merging or cross jumping. This flag
7432 is enabled by default at @option{-O2} and higher. The compilation time
7434 be limited using @option{max-tail-merge-comparisons} parameter and
7435 @option{max-tail-merge-iterations} parameter.
7439 Perform dead code elimination (DCE) on trees. This flag is enabled by
7440 default at @option{-O} and higher.
7442 @item -ftree-builtin-call-dce
7443 @opindex ftree-builtin-call-dce
7444 Perform conditional dead code elimination (DCE) for calls to built-in functions
7445 that may set @code{errno} but are otherwise side-effect free. This flag is
7446 enabled by default at @option{-O2} and higher if @option{-Os} is not also
7449 @item -ftree-dominator-opts
7450 @opindex ftree-dominator-opts
7451 Perform a variety of simple scalar cleanups (constant/copy
7452 propagation, redundancy elimination, range propagation and expression
7453 simplification) based on a dominator tree traversal. This also
7454 performs jump threading (to reduce jumps to jumps). This flag is
7455 enabled by default at @option{-O} and higher.
7459 Perform dead store elimination (DSE) on trees. A dead store is a store into
7460 a memory location that is later overwritten by another store without
7461 any intervening loads. In this case the earlier store can be deleted. This
7462 flag is enabled by default at @option{-O} and higher.
7466 Perform loop header copying on trees. This is beneficial since it increases
7467 effectiveness of code motion optimizations. It also saves one jump. This flag
7468 is enabled by default at @option{-O} and higher. It is not enabled
7469 for @option{-Os}, since it usually increases code size.
7471 @item -ftree-loop-optimize
7472 @opindex ftree-loop-optimize
7473 Perform loop optimizations on trees. This flag is enabled by default
7474 at @option{-O} and higher.
7476 @item -ftree-loop-linear
7477 @itemx -floop-interchange
7478 @itemx -floop-strip-mine
7480 @itemx -floop-unroll-and-jam
7481 @opindex ftree-loop-linear
7482 @opindex floop-interchange
7483 @opindex floop-strip-mine
7484 @opindex floop-block
7485 @opindex floop-unroll-and-jam
7486 Perform loop nest optimizations. Same as
7487 @option{-floop-nest-optimize}. To use this code transformation, GCC has
7488 to be configured with @option{--with-isl} to enable the Graphite loop
7489 transformation infrastructure.
7491 @item -fgraphite-identity
7492 @opindex fgraphite-identity
7493 Enable the identity transformation for graphite. For every SCoP we generate
7494 the polyhedral representation and transform it back to gimple. Using
7495 @option{-fgraphite-identity} we can check the costs or benefits of the
7496 GIMPLE -> GRAPHITE -> GIMPLE transformation. Some minimal optimizations
7497 are also performed by the code generator isl, like index splitting and
7498 dead code elimination in loops.
7500 @item -floop-nest-optimize
7501 @opindex floop-nest-optimize
7502 Enable the isl based loop nest optimizer. This is a generic loop nest
7503 optimizer based on the Pluto optimization algorithms. It calculates a loop
7504 structure optimized for data-locality and parallelism. This option
7507 @item -floop-parallelize-all
7508 @opindex floop-parallelize-all
7509 Use the Graphite data dependence analysis to identify loops that can
7510 be parallelized. Parallelize all the loops that can be analyzed to
7511 not contain loop carried dependences without checking that it is
7512 profitable to parallelize the loops.
7514 @item -ftree-coalesce-vars
7515 @opindex ftree-coalesce-vars
7516 While transforming the program out of the SSA representation, attempt to
7517 reduce copying by coalescing versions of different user-defined
7518 variables, instead of just compiler temporaries. This may severely
7519 limit the ability to debug an optimized program compiled with
7520 @option{-fno-var-tracking-assignments}. In the negated form, this flag
7521 prevents SSA coalescing of user variables. This option is enabled by
7522 default if optimization is enabled, and it does very little otherwise.
7524 @item -ftree-loop-if-convert
7525 @opindex ftree-loop-if-convert
7526 Attempt to transform conditional jumps in the innermost loops to
7527 branch-less equivalents. The intent is to remove control-flow from
7528 the innermost loops in order to improve the ability of the
7529 vectorization pass to handle these loops. This is enabled by default
7530 if vectorization is enabled.
7532 @item -ftree-loop-if-convert-stores
7533 @opindex ftree-loop-if-convert-stores
7534 Attempt to also if-convert conditional jumps containing memory writes.
7535 This transformation can be unsafe for multi-threaded programs as it
7536 transforms conditional memory writes into unconditional memory writes.
7539 for (i = 0; i < N; i++)
7545 for (i = 0; i < N; i++)
7546 A[i] = cond ? expr : A[i];
7548 potentially producing data races.
7550 @item -ftree-loop-distribution
7551 @opindex ftree-loop-distribution
7552 Perform loop distribution. This flag can improve cache performance on
7553 big loop bodies and allow further loop optimizations, like
7554 parallelization or vectorization, to take place. For example, the loop
7571 @item -ftree-loop-distribute-patterns
7572 @opindex ftree-loop-distribute-patterns
7573 Perform loop distribution of patterns that can be code generated with
7574 calls to a library. This flag is enabled by default at @option{-O3}.
7576 This pass distributes the initialization loops and generates a call to
7577 memset zero. For example, the loop
7593 and the initialization loop is transformed into a call to memset zero.
7595 @item -ftree-loop-im
7596 @opindex ftree-loop-im
7597 Perform loop invariant motion on trees. This pass moves only invariants that
7598 are hard to handle at RTL level (function calls, operations that expand to
7599 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
7600 operands of conditions that are invariant out of the loop, so that we can use
7601 just trivial invariantness analysis in loop unswitching. The pass also includes
7604 @item -ftree-loop-ivcanon
7605 @opindex ftree-loop-ivcanon
7606 Create a canonical counter for number of iterations in loops for which
7607 determining number of iterations requires complicated analysis. Later
7608 optimizations then may determine the number easily. Useful especially
7609 in connection with unrolling.
7613 Perform induction variable optimizations (strength reduction, induction
7614 variable merging and induction variable elimination) on trees.
7616 @item -ftree-parallelize-loops=n
7617 @opindex ftree-parallelize-loops
7618 Parallelize loops, i.e., split their iteration space to run in n threads.
7619 This is only possible for loops whose iterations are independent
7620 and can be arbitrarily reordered. The optimization is only
7621 profitable on multiprocessor machines, for loops that are CPU-intensive,
7622 rather than constrained e.g.@: by memory bandwidth. This option
7623 implies @option{-pthread}, and thus is only supported on targets
7624 that have support for @option{-pthread}.
7628 Perform function-local points-to analysis on trees. This flag is
7629 enabled by default at @option{-O} and higher.
7633 Perform scalar replacement of aggregates. This pass replaces structure
7634 references with scalars to prevent committing structures to memory too
7635 early. This flag is enabled by default at @option{-O} and higher.
7639 Perform temporary expression replacement during the SSA->normal phase. Single
7640 use/single def temporaries are replaced at their use location with their
7641 defining expression. This results in non-GIMPLE code, but gives the expanders
7642 much more complex trees to work on resulting in better RTL generation. This is
7643 enabled by default at @option{-O} and higher.
7647 Perform straight-line strength reduction on trees. This recognizes related
7648 expressions involving multiplications and replaces them by less expensive
7649 calculations when possible. This is enabled by default at @option{-O} and
7652 @item -ftree-vectorize
7653 @opindex ftree-vectorize
7654 Perform vectorization on trees. This flag enables @option{-ftree-loop-vectorize}
7655 and @option{-ftree-slp-vectorize} if not explicitly specified.
7657 @item -ftree-loop-vectorize
7658 @opindex ftree-loop-vectorize
7659 Perform loop vectorization on trees. This flag is enabled by default at
7660 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7662 @item -ftree-slp-vectorize
7663 @opindex ftree-slp-vectorize
7664 Perform basic block vectorization on trees. This flag is enabled by default at
7665 @option{-O3} and when @option{-ftree-vectorize} is enabled.
7667 @item -fvect-cost-model=@var{model}
7668 @opindex fvect-cost-model
7669 Alter the cost model used for vectorization. The @var{model} argument
7670 should be one of @samp{unlimited}, @samp{dynamic} or @samp{cheap}.
7671 With the @samp{unlimited} model the vectorized code-path is assumed
7672 to be profitable while with the @samp{dynamic} model a runtime check
7673 guards the vectorized code-path to enable it only for iteration
7674 counts that will likely execute faster than when executing the original
7675 scalar loop. The @samp{cheap} model disables vectorization of
7676 loops where doing so would be cost prohibitive for example due to
7677 required runtime checks for data dependence or alignment but otherwise
7678 is equal to the @samp{dynamic} model.
7679 The default cost model depends on other optimization flags and is
7680 either @samp{dynamic} or @samp{cheap}.
7682 @item -fsimd-cost-model=@var{model}
7683 @opindex fsimd-cost-model
7684 Alter the cost model used for vectorization of loops marked with the OpenMP
7685 or Cilk Plus simd directive. The @var{model} argument should be one of
7686 @samp{unlimited}, @samp{dynamic}, @samp{cheap}. All values of @var{model}
7687 have the same meaning as described in @option{-fvect-cost-model} and by
7688 default a cost model defined with @option{-fvect-cost-model} is used.
7692 Perform Value Range Propagation on trees. This is similar to the
7693 constant propagation pass, but instead of values, ranges of values are
7694 propagated. This allows the optimizers to remove unnecessary range
7695 checks like array bound checks and null pointer checks. This is
7696 enabled by default at @option{-O2} and higher. Null pointer check
7697 elimination is only done if @option{-fdelete-null-pointer-checks} is
7701 @opindex fsplit-paths
7702 Split paths leading to loop backedges. This can improve dead code
7703 elimination and common subexpression elimination. This is enabled by
7704 default at @option{-O2} and above.
7706 @item -fsplit-ivs-in-unroller
7707 @opindex fsplit-ivs-in-unroller
7708 Enables expression of values of induction variables in later iterations
7709 of the unrolled loop using the value in the first iteration. This breaks
7710 long dependency chains, thus improving efficiency of the scheduling passes.
7712 A combination of @option{-fweb} and CSE is often sufficient to obtain the
7713 same effect. However, that is not reliable in cases where the loop body
7714 is more complicated than a single basic block. It also does not work at all
7715 on some architectures due to restrictions in the CSE pass.
7717 This optimization is enabled by default.
7719 @item -fvariable-expansion-in-unroller
7720 @opindex fvariable-expansion-in-unroller
7721 With this option, the compiler creates multiple copies of some
7722 local variables when unrolling a loop, which can result in superior code.
7724 @item -fpartial-inlining
7725 @opindex fpartial-inlining
7726 Inline parts of functions. This option has any effect only
7727 when inlining itself is turned on by the @option{-finline-functions}
7728 or @option{-finline-small-functions} options.
7730 Enabled at level @option{-O2}.
7732 @item -fpredictive-commoning
7733 @opindex fpredictive-commoning
7734 Perform predictive commoning optimization, i.e., reusing computations
7735 (especially memory loads and stores) performed in previous
7736 iterations of loops.
7738 This option is enabled at level @option{-O3}.
7740 @item -fprefetch-loop-arrays
7741 @opindex fprefetch-loop-arrays
7742 If supported by the target machine, generate instructions to prefetch
7743 memory to improve the performance of loops that access large arrays.
7745 This option may generate better or worse code; results are highly
7746 dependent on the structure of loops within the source code.
7748 Disabled at level @option{-Os}.
7751 @itemx -fno-peephole2
7752 @opindex fno-peephole
7753 @opindex fno-peephole2
7754 Disable any machine-specific peephole optimizations. The difference
7755 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
7756 are implemented in the compiler; some targets use one, some use the
7757 other, a few use both.
7759 @option{-fpeephole} is enabled by default.
7760 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7762 @item -fno-guess-branch-probability
7763 @opindex fno-guess-branch-probability
7764 Do not guess branch probabilities using heuristics.
7766 GCC uses heuristics to guess branch probabilities if they are
7767 not provided by profiling feedback (@option{-fprofile-arcs}). These
7768 heuristics are based on the control flow graph. If some branch probabilities
7769 are specified by @code{__builtin_expect}, then the heuristics are
7770 used to guess branch probabilities for the rest of the control flow graph,
7771 taking the @code{__builtin_expect} info into account. The interactions
7772 between the heuristics and @code{__builtin_expect} can be complex, and in
7773 some cases, it may be useful to disable the heuristics so that the effects
7774 of @code{__builtin_expect} are easier to understand.
7776 The default is @option{-fguess-branch-probability} at levels
7777 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7779 @item -freorder-blocks
7780 @opindex freorder-blocks
7781 Reorder basic blocks in the compiled function in order to reduce number of
7782 taken branches and improve code locality.
7784 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
7786 @item -freorder-blocks-algorithm=@var{algorithm}
7787 @opindex freorder-blocks-algorithm
7788 Use the specified algorithm for basic block reordering. The
7789 @var{algorithm} argument can be @samp{simple}, which does not increase
7790 code size (except sometimes due to secondary effects like alignment),
7791 or @samp{stc}, the ``software trace cache'' algorithm, which tries to
7792 put all often executed code together, minimizing the number of branches
7793 executed by making extra copies of code.
7795 The default is @samp{simple} at levels @option{-O}, @option{-Os}, and
7796 @samp{stc} at levels @option{-O2}, @option{-O3}.
7798 @item -freorder-blocks-and-partition
7799 @opindex freorder-blocks-and-partition
7800 In addition to reordering basic blocks in the compiled function, in order
7801 to reduce number of taken branches, partitions hot and cold basic blocks
7802 into separate sections of the assembly and @file{.o} files, to improve
7803 paging and cache locality performance.
7805 This optimization is automatically turned off in the presence of
7806 exception handling, for linkonce sections, for functions with a user-defined
7807 section attribute and on any architecture that does not support named
7810 Enabled for x86 at levels @option{-O2}, @option{-O3}.
7812 @item -freorder-functions
7813 @opindex freorder-functions
7814 Reorder functions in the object file in order to
7815 improve code locality. This is implemented by using special
7816 subsections @code{.text.hot} for most frequently executed functions and
7817 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
7818 the linker so object file format must support named sections and linker must
7819 place them in a reasonable way.
7821 Also profile feedback must be available to make this option effective. See
7822 @option{-fprofile-arcs} for details.
7824 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
7826 @item -fstrict-aliasing
7827 @opindex fstrict-aliasing
7828 Allow the compiler to assume the strictest aliasing rules applicable to
7829 the language being compiled. For C (and C++), this activates
7830 optimizations based on the type of expressions. In particular, an
7831 object of one type is assumed never to reside at the same address as an
7832 object of a different type, unless the types are almost the same. For
7833 example, an @code{unsigned int} can alias an @code{int}, but not a
7834 @code{void*} or a @code{double}. A character type may alias any other
7837 @anchor{Type-punning}Pay special attention to code like this:
7850 The practice of reading from a different union member than the one most
7851 recently written to (called ``type-punning'') is common. Even with
7852 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
7853 is accessed through the union type. So, the code above works as
7854 expected. @xref{Structures unions enumerations and bit-fields
7855 implementation}. However, this code might not:
7866 Similarly, access by taking the address, casting the resulting pointer
7867 and dereferencing the result has undefined behavior, even if the cast
7868 uses a union type, e.g.:
7872 return ((union a_union *) &d)->i;
7876 The @option{-fstrict-aliasing} option is enabled at levels
7877 @option{-O2}, @option{-O3}, @option{-Os}.
7879 @item -fstrict-overflow
7880 @opindex fstrict-overflow
7881 Allow the compiler to assume strict signed overflow rules, depending
7882 on the language being compiled. For C (and C++) this means that
7883 overflow when doing arithmetic with signed numbers is undefined, which
7884 means that the compiler may assume that it does not happen. This
7885 permits various optimizations. For example, the compiler assumes
7886 that an expression like @code{i + 10 > i} is always true for
7887 signed @code{i}. This assumption is only valid if signed overflow is
7888 undefined, as the expression is false if @code{i + 10} overflows when
7889 using twos complement arithmetic. When this option is in effect any
7890 attempt to determine whether an operation on signed numbers
7891 overflows must be written carefully to not actually involve overflow.
7893 This option also allows the compiler to assume strict pointer
7894 semantics: given a pointer to an object, if adding an offset to that
7895 pointer does not produce a pointer to the same object, the addition is
7896 undefined. This permits the compiler to conclude that @code{p + u >
7897 p} is always true for a pointer @code{p} and unsigned integer
7898 @code{u}. This assumption is only valid because pointer wraparound is
7899 undefined, as the expression is false if @code{p + u} overflows using
7900 twos complement arithmetic.
7902 See also the @option{-fwrapv} option. Using @option{-fwrapv} means
7903 that integer signed overflow is fully defined: it wraps. When
7904 @option{-fwrapv} is used, there is no difference between
7905 @option{-fstrict-overflow} and @option{-fno-strict-overflow} for
7906 integers. With @option{-fwrapv} certain types of overflow are
7907 permitted. For example, if the compiler gets an overflow when doing
7908 arithmetic on constants, the overflowed value can still be used with
7909 @option{-fwrapv}, but not otherwise.
7911 The @option{-fstrict-overflow} option is enabled at levels
7912 @option{-O2}, @option{-O3}, @option{-Os}.
7914 @item -falign-functions
7915 @itemx -falign-functions=@var{n}
7916 @opindex falign-functions
7917 Align the start of functions to the next power-of-two greater than
7918 @var{n}, skipping up to @var{n} bytes. For instance,
7919 @option{-falign-functions=32} aligns functions to the next 32-byte
7920 boundary, but @option{-falign-functions=24} aligns to the next
7921 32-byte boundary only if this can be done by skipping 23 bytes or less.
7923 @option{-fno-align-functions} and @option{-falign-functions=1} are
7924 equivalent and mean that functions are not aligned.
7926 Some assemblers only support this flag when @var{n} is a power of two;
7927 in that case, it is rounded up.
7929 If @var{n} is not specified or is zero, use a machine-dependent default.
7931 Enabled at levels @option{-O2}, @option{-O3}.
7933 @item -falign-labels
7934 @itemx -falign-labels=@var{n}
7935 @opindex falign-labels
7936 Align all branch targets to a power-of-two boundary, skipping up to
7937 @var{n} bytes like @option{-falign-functions}. This option can easily
7938 make code slower, because it must insert dummy operations for when the
7939 branch target is reached in the usual flow of the code.
7941 @option{-fno-align-labels} and @option{-falign-labels=1} are
7942 equivalent and mean that labels are not aligned.
7944 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
7945 are greater than this value, then their values are used instead.
7947 If @var{n} is not specified or is zero, use a machine-dependent default
7948 which is very likely to be @samp{1}, meaning no alignment.
7950 Enabled at levels @option{-O2}, @option{-O3}.
7953 @itemx -falign-loops=@var{n}
7954 @opindex falign-loops
7955 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
7956 like @option{-falign-functions}. If the loops are
7957 executed many times, this makes up for any execution of the dummy
7960 @option{-fno-align-loops} and @option{-falign-loops=1} are
7961 equivalent and mean that loops are not aligned.
7963 If @var{n} is not specified or is zero, use a machine-dependent default.
7965 Enabled at levels @option{-O2}, @option{-O3}.
7968 @itemx -falign-jumps=@var{n}
7969 @opindex falign-jumps
7970 Align branch targets to a power-of-two boundary, for branch targets
7971 where the targets can only be reached by jumping, skipping up to @var{n}
7972 bytes like @option{-falign-functions}. In this case, no dummy operations
7975 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
7976 equivalent and mean that loops are not aligned.
7978 If @var{n} is not specified or is zero, use a machine-dependent default.
7980 Enabled at levels @option{-O2}, @option{-O3}.
7982 @item -funit-at-a-time
7983 @opindex funit-at-a-time
7984 This option is left for compatibility reasons. @option{-funit-at-a-time}
7985 has no effect, while @option{-fno-unit-at-a-time} implies
7986 @option{-fno-toplevel-reorder} and @option{-fno-section-anchors}.
7990 @item -fno-toplevel-reorder
7991 @opindex fno-toplevel-reorder
7992 Do not reorder top-level functions, variables, and @code{asm}
7993 statements. Output them in the same order that they appear in the
7994 input file. When this option is used, unreferenced static variables
7995 are not removed. This option is intended to support existing code
7996 that relies on a particular ordering. For new code, it is better to
7997 use attributes when possible.
7999 Enabled at level @option{-O0}. When disabled explicitly, it also implies
8000 @option{-fno-section-anchors}, which is otherwise enabled at @option{-O0} on some
8005 Constructs webs as commonly used for register allocation purposes and assign
8006 each web individual pseudo register. This allows the register allocation pass
8007 to operate on pseudos directly, but also strengthens several other optimization
8008 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
8009 however, make debugging impossible, since variables no longer stay in a
8012 Enabled by default with @option{-funroll-loops}.
8014 @item -fwhole-program
8015 @opindex fwhole-program
8016 Assume that the current compilation unit represents the whole program being
8017 compiled. All public functions and variables with the exception of @code{main}
8018 and those merged by attribute @code{externally_visible} become static functions
8019 and in effect are optimized more aggressively by interprocedural optimizers.
8021 This option should not be used in combination with @option{-flto}.
8022 Instead relying on a linker plugin should provide safer and more precise
8025 @item -flto[=@var{n}]
8027 This option runs the standard link-time optimizer. When invoked
8028 with source code, it generates GIMPLE (one of GCC's internal
8029 representations) and writes it to special ELF sections in the object
8030 file. When the object files are linked together, all the function
8031 bodies are read from these ELF sections and instantiated as if they
8032 had been part of the same translation unit.
8034 To use the link-time optimizer, @option{-flto} and optimization
8035 options should be specified at compile time and during the final link.
8036 It is recommended that you compile all the files participating in the
8037 same link with the same options and also specify those options at
8042 gcc -c -O2 -flto foo.c
8043 gcc -c -O2 -flto bar.c
8044 gcc -o myprog -flto -O2 foo.o bar.o
8047 The first two invocations to GCC save a bytecode representation
8048 of GIMPLE into special ELF sections inside @file{foo.o} and
8049 @file{bar.o}. The final invocation reads the GIMPLE bytecode from
8050 @file{foo.o} and @file{bar.o}, merges the two files into a single
8051 internal image, and compiles the result as usual. Since both
8052 @file{foo.o} and @file{bar.o} are merged into a single image, this
8053 causes all the interprocedural analyses and optimizations in GCC to
8054 work across the two files as if they were a single one. This means,
8055 for example, that the inliner is able to inline functions in
8056 @file{bar.o} into functions in @file{foo.o} and vice-versa.
8058 Another (simpler) way to enable link-time optimization is:
8061 gcc -o myprog -flto -O2 foo.c bar.c
8064 The above generates bytecode for @file{foo.c} and @file{bar.c},
8065 merges them together into a single GIMPLE representation and optimizes
8066 them as usual to produce @file{myprog}.
8068 The only important thing to keep in mind is that to enable link-time
8069 optimizations you need to use the GCC driver to perform the link step.
8070 GCC then automatically performs link-time optimization if any of the
8071 objects involved were compiled with the @option{-flto} command-line option.
8073 should specify the optimization options to be used for link-time
8074 optimization though GCC tries to be clever at guessing an
8075 optimization level to use from the options used at compile time
8076 if you fail to specify one at link time. You can always override
8077 the automatic decision to do link-time optimization at link time
8078 by passing @option{-fno-lto} to the link command.
8080 To make whole program optimization effective, it is necessary to make
8081 certain whole program assumptions. The compiler needs to know
8082 what functions and variables can be accessed by libraries and runtime
8083 outside of the link-time optimized unit. When supported by the linker,
8084 the linker plugin (see @option{-fuse-linker-plugin}) passes information
8085 to the compiler about used and externally visible symbols. When
8086 the linker plugin is not available, @option{-fwhole-program} should be
8087 used to allow the compiler to make these assumptions, which leads
8088 to more aggressive optimization decisions.
8090 When @option{-fuse-linker-plugin} is not enabled, when a file is
8091 compiled with @option{-flto}, the generated object file is larger than
8092 a regular object file because it contains GIMPLE bytecodes and the usual
8093 final code (see @option{-ffat-lto-objects}. This means that
8094 object files with LTO information can be linked as normal object
8095 files; if @option{-fno-lto} is passed to the linker, no
8096 interprocedural optimizations are applied. Note that when
8097 @option{-fno-fat-lto-objects} is enabled the compile stage is faster
8098 but you cannot perform a regular, non-LTO link on them.
8100 Additionally, the optimization flags used to compile individual files
8101 are not necessarily related to those used at link time. For instance,
8104 gcc -c -O0 -ffat-lto-objects -flto foo.c
8105 gcc -c -O0 -ffat-lto-objects -flto bar.c
8106 gcc -o myprog -O3 foo.o bar.o
8109 This produces individual object files with unoptimized assembler
8110 code, but the resulting binary @file{myprog} is optimized at
8111 @option{-O3}. If, instead, the final binary is generated with
8112 @option{-fno-lto}, then @file{myprog} is not optimized.
8114 When producing the final binary, GCC only
8115 applies link-time optimizations to those files that contain bytecode.
8116 Therefore, you can mix and match object files and libraries with
8117 GIMPLE bytecodes and final object code. GCC automatically selects
8118 which files to optimize in LTO mode and which files to link without
8121 There are some code generation flags preserved by GCC when
8122 generating bytecodes, as they need to be used during the final link
8123 stage. Generally options specified at link time override those
8124 specified at compile time.
8126 If you do not specify an optimization level option @option{-O} at
8127 link time, then GCC uses the highest optimization level
8128 used when compiling the object files.
8130 Currently, the following options and their settings are taken from
8131 the first object file that explicitly specifies them:
8132 @option{-fPIC}, @option{-fpic}, @option{-fpie}, @option{-fcommon},
8133 @option{-fexceptions}, @option{-fnon-call-exceptions}, @option{-fgnu-tm}
8134 and all the @option{-m} target flags.
8136 Certain ABI-changing flags are required to match in all compilation units,
8137 and trying to override this at link time with a conflicting value
8138 is ignored. This includes options such as @option{-freg-struct-return}
8139 and @option{-fpcc-struct-return}.
8141 Other options such as @option{-ffp-contract}, @option{-fno-strict-overflow},
8142 @option{-fwrapv}, @option{-fno-trapv} or @option{-fno-strict-aliasing}
8143 are passed through to the link stage and merged conservatively for
8144 conflicting translation units. Specifically
8145 @option{-fno-strict-overflow}, @option{-fwrapv} and @option{-fno-trapv} take
8146 precedence; and for example @option{-ffp-contract=off} takes precedence
8147 over @option{-ffp-contract=fast}. You can override them at link time.
8149 If LTO encounters objects with C linkage declared with incompatible
8150 types in separate translation units to be linked together (undefined
8151 behavior according to ISO C99 6.2.7), a non-fatal diagnostic may be
8152 issued. The behavior is still undefined at run time. Similar
8153 diagnostics may be raised for other languages.
8155 Another feature of LTO is that it is possible to apply interprocedural
8156 optimizations on files written in different languages:
8161 gfortran -c -flto baz.f90
8162 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran
8165 Notice that the final link is done with @command{g++} to get the C++
8166 runtime libraries and @option{-lgfortran} is added to get the Fortran
8167 runtime libraries. In general, when mixing languages in LTO mode, you
8168 should use the same link command options as when mixing languages in a
8169 regular (non-LTO) compilation.
8171 If object files containing GIMPLE bytecode are stored in a library archive, say
8172 @file{libfoo.a}, it is possible to extract and use them in an LTO link if you
8173 are using a linker with plugin support. To create static libraries suitable
8174 for LTO, use @command{gcc-ar} and @command{gcc-ranlib} instead of @command{ar}
8175 and @command{ranlib};
8176 to show the symbols of object files with GIMPLE bytecode, use
8177 @command{gcc-nm}. Those commands require that @command{ar}, @command{ranlib}
8178 and @command{nm} have been compiled with plugin support. At link time, use the the
8179 flag @option{-fuse-linker-plugin} to ensure that the library participates in
8180 the LTO optimization process:
8183 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo
8186 With the linker plugin enabled, the linker extracts the needed
8187 GIMPLE files from @file{libfoo.a} and passes them on to the running GCC
8188 to make them part of the aggregated GIMPLE image to be optimized.
8190 If you are not using a linker with plugin support and/or do not
8191 enable the linker plugin, then the objects inside @file{libfoo.a}
8192 are extracted and linked as usual, but they do not participate
8193 in the LTO optimization process. In order to make a static library suitable
8194 for both LTO optimization and usual linkage, compile its object files with
8195 @option{-flto} @option{-ffat-lto-objects}.
8197 Link-time optimizations do not require the presence of the whole program to
8198 operate. If the program does not require any symbols to be exported, it is
8199 possible to combine @option{-flto} and @option{-fwhole-program} to allow
8200 the interprocedural optimizers to use more aggressive assumptions which may
8201 lead to improved optimization opportunities.
8202 Use of @option{-fwhole-program} is not needed when linker plugin is
8203 active (see @option{-fuse-linker-plugin}).
8205 The current implementation of LTO makes no
8206 attempt to generate bytecode that is portable between different
8207 types of hosts. The bytecode files are versioned and there is a
8208 strict version check, so bytecode files generated in one version of
8209 GCC do not work with an older or newer version of GCC.
8211 Link-time optimization does not work well with generation of debugging
8212 information. Combining @option{-flto} with
8213 @option{-g} is currently experimental and expected to produce unexpected
8216 If you specify the optional @var{n}, the optimization and code
8217 generation done at link time is executed in parallel using @var{n}
8218 parallel jobs by utilizing an installed @command{make} program. The
8219 environment variable @env{MAKE} may be used to override the program
8220 used. The default value for @var{n} is 1.
8222 You can also specify @option{-flto=jobserver} to use GNU make's
8223 job server mode to determine the number of parallel jobs. This
8224 is useful when the Makefile calling GCC is already executing in parallel.
8225 You must prepend a @samp{+} to the command recipe in the parent Makefile
8226 for this to work. This option likely only works if @env{MAKE} is
8229 @item -flto-partition=@var{alg}
8230 @opindex flto-partition
8231 Specify the partitioning algorithm used by the link-time optimizer.
8232 The value is either @samp{1to1} to specify a partitioning mirroring
8233 the original source files or @samp{balanced} to specify partitioning
8234 into equally sized chunks (whenever possible) or @samp{max} to create
8235 new partition for every symbol where possible. Specifying @samp{none}
8236 as an algorithm disables partitioning and streaming completely.
8237 The default value is @samp{balanced}. While @samp{1to1} can be used
8238 as an workaround for various code ordering issues, the @samp{max}
8239 partitioning is intended for internal testing only.
8240 The value @samp{one} specifies that exactly one partition should be
8241 used while the value @samp{none} bypasses partitioning and executes
8242 the link-time optimization step directly from the WPA phase.
8244 @item -flto-odr-type-merging
8245 @opindex flto-odr-type-merging
8246 Enable streaming of mangled types names of C++ types and their unification
8247 at link time. This increases size of LTO object files, but enables
8248 diagnostics about One Definition Rule violations.
8250 @item -flto-compression-level=@var{n}
8251 @opindex flto-compression-level
8252 This option specifies the level of compression used for intermediate
8253 language written to LTO object files, and is only meaningful in
8254 conjunction with LTO mode (@option{-flto}). Valid
8255 values are 0 (no compression) to 9 (maximum compression). Values
8256 outside this range are clamped to either 0 or 9. If the option is not
8257 given, a default balanced compression setting is used.
8259 @item -fuse-linker-plugin
8260 @opindex fuse-linker-plugin
8261 Enables the use of a linker plugin during link-time optimization. This
8262 option relies on plugin support in the linker, which is available in gold
8263 or in GNU ld 2.21 or newer.
8265 This option enables the extraction of object files with GIMPLE bytecode out
8266 of library archives. This improves the quality of optimization by exposing
8267 more code to the link-time optimizer. This information specifies what
8268 symbols can be accessed externally (by non-LTO object or during dynamic
8269 linking). Resulting code quality improvements on binaries (and shared
8270 libraries that use hidden visibility) are similar to @option{-fwhole-program}.
8271 See @option{-flto} for a description of the effect of this flag and how to
8274 This option is enabled by default when LTO support in GCC is enabled
8275 and GCC was configured for use with
8276 a linker supporting plugins (GNU ld 2.21 or newer or gold).
8278 @item -ffat-lto-objects
8279 @opindex ffat-lto-objects
8280 Fat LTO objects are object files that contain both the intermediate language
8281 and the object code. This makes them usable for both LTO linking and normal
8282 linking. This option is effective only when compiling with @option{-flto}
8283 and is ignored at link time.
8285 @option{-fno-fat-lto-objects} improves compilation time over plain LTO, but
8286 requires the complete toolchain to be aware of LTO. It requires a linker with
8287 linker plugin support for basic functionality. Additionally,
8288 @command{nm}, @command{ar} and @command{ranlib}
8289 need to support linker plugins to allow a full-featured build environment
8290 (capable of building static libraries etc). GCC provides the @command{gcc-ar},
8291 @command{gcc-nm}, @command{gcc-ranlib} wrappers to pass the right options
8292 to these tools. With non fat LTO makefiles need to be modified to use them.
8294 The default is @option{-fno-fat-lto-objects} on targets with linker plugin
8297 @item -fcompare-elim
8298 @opindex fcompare-elim
8299 After register allocation and post-register allocation instruction splitting,
8300 identify arithmetic instructions that compute processor flags similar to a
8301 comparison operation based on that arithmetic. If possible, eliminate the
8302 explicit comparison operation.
8304 This pass only applies to certain targets that cannot explicitly represent
8305 the comparison operation before register allocation is complete.
8307 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8309 @item -fcprop-registers
8310 @opindex fcprop-registers
8311 After register allocation and post-register allocation instruction splitting,
8312 perform a copy-propagation pass to try to reduce scheduling dependencies
8313 and occasionally eliminate the copy.
8315 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
8317 @item -fprofile-correction
8318 @opindex fprofile-correction
8319 Profiles collected using an instrumented binary for multi-threaded programs may
8320 be inconsistent due to missed counter updates. When this option is specified,
8321 GCC uses heuristics to correct or smooth out such inconsistencies. By
8322 default, GCC emits an error message when an inconsistent profile is detected.
8325 @itemx -fprofile-use=@var{path}
8326 @opindex fprofile-use
8327 Enable profile feedback-directed optimizations,
8328 and the following optimizations
8329 which are generally profitable only with profile feedback available:
8330 @option{-fbranch-probabilities}, @option{-fvpt},
8331 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8332 @option{-ftree-vectorize}, and @option{ftree-loop-distribute-patterns}.
8334 Before you can use this option, you must first generate profiling information.
8335 @xref{Optimize Options}, for information about the @option{-fprofile-generate}
8338 By default, GCC emits an error message if the feedback profiles do not
8339 match the source code. This error can be turned into a warning by using
8340 @option{-Wcoverage-mismatch}. Note this may result in poorly optimized
8343 If @var{path} is specified, GCC looks at the @var{path} to find
8344 the profile feedback data files. See @option{-fprofile-dir}.
8346 @item -fauto-profile
8347 @itemx -fauto-profile=@var{path}
8348 @opindex fauto-profile
8349 Enable sampling-based feedback-directed optimizations,
8350 and the following optimizations
8351 which are generally profitable only with profile feedback available:
8352 @option{-fbranch-probabilities}, @option{-fvpt},
8353 @option{-funroll-loops}, @option{-fpeel-loops}, @option{-ftracer},
8354 @option{-ftree-vectorize},
8355 @option{-finline-functions}, @option{-fipa-cp}, @option{-fipa-cp-clone},
8356 @option{-fpredictive-commoning}, @option{-funswitch-loops},
8357 @option{-fgcse-after-reload}, and @option{-ftree-loop-distribute-patterns}.
8359 @var{path} is the name of a file containing AutoFDO profile information.
8360 If omitted, it defaults to @file{fbdata.afdo} in the current directory.
8362 Producing an AutoFDO profile data file requires running your program
8363 with the @command{perf} utility on a supported GNU/Linux target system.
8364 For more information, see @uref{https://perf.wiki.kernel.org/}.
8368 perf record -e br_inst_retired:near_taken -b -o perf.data \
8372 Then use the @command{create_gcov} tool to convert the raw profile data
8373 to a format that can be used by GCC.@ You must also supply the
8374 unstripped binary for your program to this tool.
8375 See @uref{https://github.com/google/autofdo}.
8379 create_gcov --binary=your_program.unstripped --profile=perf.data \
8384 The following options control compiler behavior regarding floating-point
8385 arithmetic. These options trade off between speed and
8386 correctness. All must be specifically enabled.
8390 @opindex ffloat-store
8391 Do not store floating-point variables in registers, and inhibit other
8392 options that might change whether a floating-point value is taken from a
8395 @cindex floating-point precision
8396 This option prevents undesirable excess precision on machines such as
8397 the 68000 where the floating registers (of the 68881) keep more
8398 precision than a @code{double} is supposed to have. Similarly for the
8399 x86 architecture. For most programs, the excess precision does only
8400 good, but a few programs rely on the precise definition of IEEE floating
8401 point. Use @option{-ffloat-store} for such programs, after modifying
8402 them to store all pertinent intermediate computations into variables.
8404 @item -fexcess-precision=@var{style}
8405 @opindex fexcess-precision
8406 This option allows further control over excess precision on machines
8407 where floating-point registers have more precision than the IEEE
8408 @code{float} and @code{double} types and the processor does not
8409 support operations rounding to those types. By default,
8410 @option{-fexcess-precision=fast} is in effect; this means that
8411 operations are carried out in the precision of the registers and that
8412 it is unpredictable when rounding to the types specified in the source
8413 code takes place. When compiling C, if
8414 @option{-fexcess-precision=standard} is specified then excess
8415 precision follows the rules specified in ISO C99; in particular,
8416 both casts and assignments cause values to be rounded to their
8417 semantic types (whereas @option{-ffloat-store} only affects
8418 assignments). This option is enabled by default for C if a strict
8419 conformance option such as @option{-std=c99} is used.
8422 @option{-fexcess-precision=standard} is not implemented for languages
8423 other than C, and has no effect if
8424 @option{-funsafe-math-optimizations} or @option{-ffast-math} is
8425 specified. On the x86, it also has no effect if @option{-mfpmath=sse}
8426 or @option{-mfpmath=sse+387} is specified; in the former case, IEEE
8427 semantics apply without excess precision, and in the latter, rounding
8432 Sets the options @option{-fno-math-errno}, @option{-funsafe-math-optimizations},
8433 @option{-ffinite-math-only}, @option{-fno-rounding-math},
8434 @option{-fno-signaling-nans} and @option{-fcx-limited-range}.
8436 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
8438 This option is not turned on by any @option{-O} option besides
8439 @option{-Ofast} since it can result in incorrect output for programs
8440 that depend on an exact implementation of IEEE or ISO rules/specifications
8441 for math functions. It may, however, yield faster code for programs
8442 that do not require the guarantees of these specifications.
8444 @item -fno-math-errno
8445 @opindex fno-math-errno
8446 Do not set @code{errno} after calling math functions that are executed
8447 with a single instruction, e.g., @code{sqrt}. A program that relies on
8448 IEEE exceptions for math error handling may want to use this flag
8449 for speed while maintaining IEEE arithmetic compatibility.
8451 This option is not turned on by any @option{-O} option since
8452 it can result in incorrect output for programs that depend on
8453 an exact implementation of IEEE or ISO rules/specifications for
8454 math functions. It may, however, yield faster code for programs
8455 that do not require the guarantees of these specifications.
8457 The default is @option{-fmath-errno}.
8459 On Darwin systems, the math library never sets @code{errno}. There is
8460 therefore no reason for the compiler to consider the possibility that
8461 it might, and @option{-fno-math-errno} is the default.
8463 @item -funsafe-math-optimizations
8464 @opindex funsafe-math-optimizations
8466 Allow optimizations for floating-point arithmetic that (a) assume
8467 that arguments and results are valid and (b) may violate IEEE or
8468 ANSI standards. When used at link time, it may include libraries
8469 or startup files that change the default FPU control word or other
8470 similar optimizations.
8472 This option is not turned on by any @option{-O} option since
8473 it can result in incorrect output for programs that depend on
8474 an exact implementation of IEEE or ISO rules/specifications for
8475 math functions. It may, however, yield faster code for programs
8476 that do not require the guarantees of these specifications.
8477 Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
8478 @option{-fassociative-math} and @option{-freciprocal-math}.
8480 The default is @option{-fno-unsafe-math-optimizations}.
8482 @item -fassociative-math
8483 @opindex fassociative-math
8485 Allow re-association of operands in series of floating-point operations.
8486 This violates the ISO C and C++ language standard by possibly changing
8487 computation result. NOTE: re-ordering may change the sign of zero as
8488 well as ignore NaNs and inhibit or create underflow or overflow (and
8489 thus cannot be used on code that relies on rounding behavior like
8490 @code{(x + 2**52) - 2**52}. May also reorder floating-point comparisons
8491 and thus may not be used when ordered comparisons are required.
8492 This option requires that both @option{-fno-signed-zeros} and
8493 @option{-fno-trapping-math} be in effect. Moreover, it doesn't make
8494 much sense with @option{-frounding-math}. For Fortran the option
8495 is automatically enabled when both @option{-fno-signed-zeros} and
8496 @option{-fno-trapping-math} are in effect.
8498 The default is @option{-fno-associative-math}.
8500 @item -freciprocal-math
8501 @opindex freciprocal-math
8503 Allow the reciprocal of a value to be used instead of dividing by
8504 the value if this enables optimizations. For example @code{x / y}
8505 can be replaced with @code{x * (1/y)}, which is useful if @code{(1/y)}
8506 is subject to common subexpression elimination. Note that this loses
8507 precision and increases the number of flops operating on the value.
8509 The default is @option{-fno-reciprocal-math}.
8511 @item -ffinite-math-only
8512 @opindex ffinite-math-only
8513 Allow optimizations for floating-point arithmetic that assume
8514 that arguments and results are not NaNs or +-Infs.
8516 This option is not turned on by any @option{-O} option since
8517 it can result in incorrect output for programs that depend on
8518 an exact implementation of IEEE or ISO rules/specifications for
8519 math functions. It may, however, yield faster code for programs
8520 that do not require the guarantees of these specifications.
8522 The default is @option{-fno-finite-math-only}.
8524 @item -fno-signed-zeros
8525 @opindex fno-signed-zeros
8526 Allow optimizations for floating-point arithmetic that ignore the
8527 signedness of zero. IEEE arithmetic specifies the behavior of
8528 distinct +0.0 and @minus{}0.0 values, which then prohibits simplification
8529 of expressions such as x+0.0 or 0.0*x (even with @option{-ffinite-math-only}).
8530 This option implies that the sign of a zero result isn't significant.
8532 The default is @option{-fsigned-zeros}.
8534 @item -fno-trapping-math
8535 @opindex fno-trapping-math
8536 Compile code assuming that floating-point operations cannot generate
8537 user-visible traps. These traps include division by zero, overflow,
8538 underflow, inexact result and invalid operation. This option requires
8539 that @option{-fno-signaling-nans} be in effect. Setting this option may
8540 allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
8542 This option should never be turned on by any @option{-O} option since
8543 it can result in incorrect output for programs that depend on
8544 an exact implementation of IEEE or ISO rules/specifications for
8547 The default is @option{-ftrapping-math}.
8549 @item -frounding-math
8550 @opindex frounding-math
8551 Disable transformations and optimizations that assume default floating-point
8552 rounding behavior. This is round-to-zero for all floating point
8553 to integer conversions, and round-to-nearest for all other arithmetic
8554 truncations. This option should be specified for programs that change
8555 the FP rounding mode dynamically, or that may be executed with a
8556 non-default rounding mode. This option disables constant folding of
8557 floating-point expressions at compile time (which may be affected by
8558 rounding mode) and arithmetic transformations that are unsafe in the
8559 presence of sign-dependent rounding modes.
8561 The default is @option{-fno-rounding-math}.
8563 This option is experimental and does not currently guarantee to
8564 disable all GCC optimizations that are affected by rounding mode.
8565 Future versions of GCC may provide finer control of this setting
8566 using C99's @code{FENV_ACCESS} pragma. This command-line option
8567 will be used to specify the default state for @code{FENV_ACCESS}.
8569 @item -fsignaling-nans
8570 @opindex fsignaling-nans
8571 Compile code assuming that IEEE signaling NaNs may generate user-visible
8572 traps during floating-point operations. Setting this option disables
8573 optimizations that may change the number of exceptions visible with
8574 signaling NaNs. This option implies @option{-ftrapping-math}.
8576 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
8579 The default is @option{-fno-signaling-nans}.
8581 This option is experimental and does not currently guarantee to
8582 disable all GCC optimizations that affect signaling NaN behavior.
8584 @item -fno-fp-int-builtin-inexact
8585 @opindex fno-fp-int-builtin-inexact
8586 Do not allow the built-in functions @code{ceil}, @code{floor},
8587 @code{round} and @code{trunc}, and their @code{float} and @code{long
8588 double} variants, to generate code that raises the ``inexact''
8589 floating-point exception for noninteger arguments. ISO C99 and C11
8590 allow these functions to raise the ``inexact'' exception, but ISO/IEC
8591 TS 18661-1:2014, the C bindings to IEEE 754-2008, does not allow these
8594 The default is @option{-ffp-int-builtin-inexact}, allowing the
8595 exception to be raised. This option does nothing unless
8596 @option{-ftrapping-math} is in effect.
8598 Even if @option{-fno-fp-int-builtin-inexact} is used, if the functions
8599 generate a call to a library function then the ``inexact'' exception
8600 may be raised if the library implementation does not follow TS 18661.
8602 @item -fsingle-precision-constant
8603 @opindex fsingle-precision-constant
8604 Treat floating-point constants as single precision instead of
8605 implicitly converting them to double-precision constants.
8607 @item -fcx-limited-range
8608 @opindex fcx-limited-range
8609 When enabled, this option states that a range reduction step is not
8610 needed when performing complex division. Also, there is no checking
8611 whether the result of a complex multiplication or division is @code{NaN
8612 + I*NaN}, with an attempt to rescue the situation in that case. The
8613 default is @option{-fno-cx-limited-range}, but is enabled by
8614 @option{-ffast-math}.
8616 This option controls the default setting of the ISO C99
8617 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
8620 @item -fcx-fortran-rules
8621 @opindex fcx-fortran-rules
8622 Complex multiplication and division follow Fortran rules. Range
8623 reduction is done as part of complex division, but there is no checking
8624 whether the result of a complex multiplication or division is @code{NaN
8625 + I*NaN}, with an attempt to rescue the situation in that case.
8627 The default is @option{-fno-cx-fortran-rules}.
8631 The following options control optimizations that may improve
8632 performance, but are not enabled by any @option{-O} options. This
8633 section includes experimental options that may produce broken code.
8636 @item -fbranch-probabilities
8637 @opindex fbranch-probabilities
8638 After running a program compiled with @option{-fprofile-arcs}
8639 (@pxref{Instrumentation Options}),
8640 you can compile it a second time using
8641 @option{-fbranch-probabilities}, to improve optimizations based on
8642 the number of times each branch was taken. When a program
8643 compiled with @option{-fprofile-arcs} exits, it saves arc execution
8644 counts to a file called @file{@var{sourcename}.gcda} for each source
8645 file. The information in this data file is very dependent on the
8646 structure of the generated code, so you must use the same source code
8647 and the same optimization options for both compilations.
8649 With @option{-fbranch-probabilities}, GCC puts a
8650 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
8651 These can be used to improve optimization. Currently, they are only
8652 used in one place: in @file{reorg.c}, instead of guessing which path a
8653 branch is most likely to take, the @samp{REG_BR_PROB} values are used to
8654 exactly determine which path is taken more often.
8656 @item -fprofile-values
8657 @opindex fprofile-values
8658 If combined with @option{-fprofile-arcs}, it adds code so that some
8659 data about values of expressions in the program is gathered.
8661 With @option{-fbranch-probabilities}, it reads back the data gathered
8662 from profiling values of expressions for usage in optimizations.
8664 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
8666 @item -fprofile-reorder-functions
8667 @opindex fprofile-reorder-functions
8668 Function reordering based on profile instrumentation collects
8669 first time of execution of a function and orders these functions
8672 Enabled with @option{-fprofile-use}.
8676 If combined with @option{-fprofile-arcs}, this option instructs the compiler
8677 to add code to gather information about values of expressions.
8679 With @option{-fbranch-probabilities}, it reads back the data gathered
8680 and actually performs the optimizations based on them.
8681 Currently the optimizations include specialization of division operations
8682 using the knowledge about the value of the denominator.
8684 @item -frename-registers
8685 @opindex frename-registers
8686 Attempt to avoid false dependencies in scheduled code by making use
8687 of registers left over after register allocation. This optimization
8688 most benefits processors with lots of registers. Depending on the
8689 debug information format adopted by the target, however, it can
8690 make debugging impossible, since variables no longer stay in
8691 a ``home register''.
8693 Enabled by default with @option{-funroll-loops}.
8695 @item -fschedule-fusion
8696 @opindex fschedule-fusion
8697 Performs a target dependent pass over the instruction stream to schedule
8698 instructions of same type together because target machine can execute them
8699 more efficiently if they are adjacent to each other in the instruction flow.
8701 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
8705 Perform tail duplication to enlarge superblock size. This transformation
8706 simplifies the control flow of the function allowing other optimizations to do
8709 Enabled with @option{-fprofile-use}.
8711 @item -funroll-loops
8712 @opindex funroll-loops
8713 Unroll loops whose number of iterations can be determined at compile time or
8714 upon entry to the loop. @option{-funroll-loops} implies
8715 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
8716 It also turns on complete loop peeling (i.e.@: complete removal of loops with
8717 a small constant number of iterations). This option makes code larger, and may
8718 or may not make it run faster.
8720 Enabled with @option{-fprofile-use}.
8722 @item -funroll-all-loops
8723 @opindex funroll-all-loops
8724 Unroll all loops, even if their number of iterations is uncertain when
8725 the loop is entered. This usually makes programs run more slowly.
8726 @option{-funroll-all-loops} implies the same options as
8727 @option{-funroll-loops}.
8730 @opindex fpeel-loops
8731 Peels loops for which there is enough information that they do not
8732 roll much (from profile feedback or static analysis). It also turns on
8733 complete loop peeling (i.e.@: complete removal of loops with small constant
8734 number of iterations).
8736 Enabled with @option{-O3} and/or @option{-fprofile-use}.
8738 @item -fmove-loop-invariants
8739 @opindex fmove-loop-invariants
8740 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
8741 at level @option{-O1}
8743 @item -funswitch-loops
8744 @opindex funswitch-loops
8745 Move branches with loop invariant conditions out of the loop, with duplicates
8746 of the loop on both branches (modified according to result of the condition).
8748 @item -ffunction-sections
8749 @itemx -fdata-sections
8750 @opindex ffunction-sections
8751 @opindex fdata-sections
8752 Place each function or data item into its own section in the output
8753 file if the target supports arbitrary sections. The name of the
8754 function or the name of the data item determines the section's name
8757 Use these options on systems where the linker can perform optimizations
8758 to improve locality of reference in the instruction space. Most systems
8759 using the ELF object format and SPARC processors running Solaris 2 have
8760 linkers with such optimizations. AIX may have these optimizations in
8763 Only use these options when there are significant benefits from doing
8764 so. When you specify these options, the assembler and linker
8765 create larger object and executable files and are also slower.
8766 You cannot use @command{gprof} on all systems if you
8767 specify this option, and you may have problems with debugging if
8768 you specify both this option and @option{-g}.
8770 @item -fbranch-target-load-optimize
8771 @opindex fbranch-target-load-optimize
8772 Perform branch target register load optimization before prologue / epilogue
8774 The use of target registers can typically be exposed only during reload,
8775 thus hoisting loads out of loops and doing inter-block scheduling needs
8776 a separate optimization pass.
8778 @item -fbranch-target-load-optimize2
8779 @opindex fbranch-target-load-optimize2
8780 Perform branch target register load optimization after prologue / epilogue
8783 @item -fbtr-bb-exclusive
8784 @opindex fbtr-bb-exclusive
8785 When performing branch target register load optimization, don't reuse
8786 branch target registers within any basic block.
8789 @opindex fstdarg-opt
8790 Optimize the prologue of variadic argument functions with respect to usage of
8793 @item -fsection-anchors
8794 @opindex fsection-anchors
8795 Try to reduce the number of symbolic address calculations by using
8796 shared ``anchor'' symbols to address nearby objects. This transformation
8797 can help to reduce the number of GOT entries and GOT accesses on some
8800 For example, the implementation of the following function @code{foo}:
8804 int foo (void) @{ return a + b + c; @}
8808 usually calculates the addresses of all three variables, but if you
8809 compile it with @option{-fsection-anchors}, it accesses the variables
8810 from a common anchor point instead. The effect is similar to the
8811 following pseudocode (which isn't valid C):
8816 register int *xr = &x;
8817 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
8821 Not all targets support this option.
8823 @item --param @var{name}=@var{value}
8825 In some places, GCC uses various constants to control the amount of
8826 optimization that is done. For example, GCC does not inline functions
8827 that contain more than a certain number of instructions. You can
8828 control some of these constants on the command line using the
8829 @option{--param} option.
8831 The names of specific parameters, and the meaning of the values, are
8832 tied to the internals of the compiler, and are subject to change
8833 without notice in future releases.
8835 In each case, the @var{value} is an integer. The allowable choices for
8839 @item predictable-branch-outcome
8840 When branch is predicted to be taken with probability lower than this threshold
8841 (in percent), then it is considered well predictable. The default is 10.
8843 @item max-rtl-if-conversion-insns
8844 RTL if-conversion tries to remove conditional branches around a block and
8845 replace them with conditionally executed instructions. This parameter
8846 gives the maximum number of instructions in a block which should be
8847 considered for if-conversion. The default is 10, though the compiler will
8848 also use other heuristics to decide whether if-conversion is likely to be
8851 @item max-crossjump-edges
8852 The maximum number of incoming edges to consider for cross-jumping.
8853 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
8854 the number of edges incoming to each block. Increasing values mean
8855 more aggressive optimization, making the compilation time increase with
8856 probably small improvement in executable size.
8858 @item min-crossjump-insns
8859 The minimum number of instructions that must be matched at the end
8860 of two blocks before cross-jumping is performed on them. This
8861 value is ignored in the case where all instructions in the block being
8862 cross-jumped from are matched. The default value is 5.
8864 @item max-grow-copy-bb-insns
8865 The maximum code size expansion factor when copying basic blocks
8866 instead of jumping. The expansion is relative to a jump instruction.
8867 The default value is 8.
8869 @item max-goto-duplication-insns
8870 The maximum number of instructions to duplicate to a block that jumps
8871 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
8872 passes, GCC factors computed gotos early in the compilation process,
8873 and unfactors them as late as possible. Only computed jumps at the
8874 end of a basic blocks with no more than max-goto-duplication-insns are
8875 unfactored. The default value is 8.
8877 @item max-delay-slot-insn-search
8878 The maximum number of instructions to consider when looking for an
8879 instruction to fill a delay slot. If more than this arbitrary number of
8880 instructions are searched, the time savings from filling the delay slot
8881 are minimal, so stop searching. Increasing values mean more
8882 aggressive optimization, making the compilation time increase with probably
8883 small improvement in execution time.
8885 @item max-delay-slot-live-search
8886 When trying to fill delay slots, the maximum number of instructions to
8887 consider when searching for a block with valid live register
8888 information. Increasing this arbitrarily chosen value means more
8889 aggressive optimization, increasing the compilation time. This parameter
8890 should be removed when the delay slot code is rewritten to maintain the
8893 @item max-gcse-memory
8894 The approximate maximum amount of memory that can be allocated in
8895 order to perform the global common subexpression elimination
8896 optimization. If more memory than specified is required, the
8897 optimization is not done.
8899 @item max-gcse-insertion-ratio
8900 If the ratio of expression insertions to deletions is larger than this value
8901 for any expression, then RTL PRE inserts or removes the expression and thus
8902 leaves partially redundant computations in the instruction stream. The default value is 20.
8904 @item max-pending-list-length
8905 The maximum number of pending dependencies scheduling allows
8906 before flushing the current state and starting over. Large functions
8907 with few branches or calls can create excessively large lists which
8908 needlessly consume memory and resources.
8910 @item max-modulo-backtrack-attempts
8911 The maximum number of backtrack attempts the scheduler should make
8912 when modulo scheduling a loop. Larger values can exponentially increase
8915 @item max-inline-insns-single
8916 Several parameters control the tree inliner used in GCC@.
8917 This number sets the maximum number of instructions (counted in GCC's
8918 internal representation) in a single function that the tree inliner
8919 considers for inlining. This only affects functions declared
8920 inline and methods implemented in a class declaration (C++).
8921 The default value is 400.
8923 @item max-inline-insns-auto
8924 When you use @option{-finline-functions} (included in @option{-O3}),
8925 a lot of functions that would otherwise not be considered for inlining
8926 by the compiler are investigated. To those functions, a different
8927 (more restrictive) limit compared to functions declared inline can
8929 The default value is 40.
8931 @item inline-min-speedup
8932 When estimated performance improvement of caller + callee runtime exceeds this
8933 threshold (in precent), the function can be inlined regardless the limit on
8934 @option{--param max-inline-insns-single} and @option{--param
8935 max-inline-insns-auto}.
8937 @item large-function-insns
8938 The limit specifying really large functions. For functions larger than this
8939 limit after inlining, inlining is constrained by
8940 @option{--param large-function-growth}. This parameter is useful primarily
8941 to avoid extreme compilation time caused by non-linear algorithms used by the
8943 The default value is 2700.
8945 @item large-function-growth
8946 Specifies maximal growth of large function caused by inlining in percents.
8947 The default value is 100 which limits large function growth to 2.0 times
8950 @item large-unit-insns
8951 The limit specifying large translation unit. Growth caused by inlining of
8952 units larger than this limit is limited by @option{--param inline-unit-growth}.
8953 For small units this might be too tight.
8954 For example, consider a unit consisting of function A
8955 that is inline and B that just calls A three times. If B is small relative to
8956 A, the growth of unit is 300\% and yet such inlining is very sane. For very
8957 large units consisting of small inlineable functions, however, the overall unit
8958 growth limit is needed to avoid exponential explosion of code size. Thus for
8959 smaller units, the size is increased to @option{--param large-unit-insns}
8960 before applying @option{--param inline-unit-growth}. The default is 10000.
8962 @item inline-unit-growth
8963 Specifies maximal overall growth of the compilation unit caused by inlining.
8964 The default value is 20 which limits unit growth to 1.2 times the original
8965 size. Cold functions (either marked cold via an attribute or by profile
8966 feedback) are not accounted into the unit size.
8968 @item ipcp-unit-growth
8969 Specifies maximal overall growth of the compilation unit caused by
8970 interprocedural constant propagation. The default value is 10 which limits
8971 unit growth to 1.1 times the original size.
8973 @item large-stack-frame
8974 The limit specifying large stack frames. While inlining the algorithm is trying
8975 to not grow past this limit too much. The default value is 256 bytes.
8977 @item large-stack-frame-growth
8978 Specifies maximal growth of large stack frames caused by inlining in percents.
8979 The default value is 1000 which limits large stack frame growth to 11 times
8982 @item max-inline-insns-recursive
8983 @itemx max-inline-insns-recursive-auto
8984 Specifies the maximum number of instructions an out-of-line copy of a
8985 self-recursive inline
8986 function can grow into by performing recursive inlining.
8988 @option{--param max-inline-insns-recursive} applies to functions
8990 For functions not declared inline, recursive inlining
8991 happens only when @option{-finline-functions} (included in @option{-O3}) is
8992 enabled; @option{--param max-inline-insns-recursive-auto} applies instead. The
8993 default value is 450.
8995 @item max-inline-recursive-depth
8996 @itemx max-inline-recursive-depth-auto
8997 Specifies the maximum recursion depth used for recursive inlining.
8999 @option{--param max-inline-recursive-depth} applies to functions
9000 declared inline. For functions not declared inline, recursive inlining
9001 happens only when @option{-finline-functions} (included in @option{-O3}) is
9002 enabled; @option{--param max-inline-recursive-depth-auto} applies instead. The
9005 @item min-inline-recursive-probability
9006 Recursive inlining is profitable only for function having deep recursion
9007 in average and can hurt for function having little recursion depth by
9008 increasing the prologue size or complexity of function body to other
9011 When profile feedback is available (see @option{-fprofile-generate}) the actual
9012 recursion depth can be guessed from probability that function recurses via a
9013 given call expression. This parameter limits inlining only to call expressions
9014 whose probability exceeds the given threshold (in percents).
9015 The default value is 10.
9017 @item early-inlining-insns
9018 Specify growth that the early inliner can make. In effect it increases
9019 the amount of inlining for code having a large abstraction penalty.
9020 The default value is 14.
9022 @item max-early-inliner-iterations
9023 Limit of iterations of the early inliner. This basically bounds
9024 the number of nested indirect calls the early inliner can resolve.
9025 Deeper chains are still handled by late inlining.
9027 @item comdat-sharing-probability
9028 Probability (in percent) that C++ inline function with comdat visibility
9029 are shared across multiple compilation units. The default value is 20.
9031 @item profile-func-internal-id
9032 A parameter to control whether to use function internal id in profile
9033 database lookup. If the value is 0, the compiler uses an id that
9034 is based on function assembler name and filename, which makes old profile
9035 data more tolerant to source changes such as function reordering etc.
9036 The default value is 0.
9038 @item min-vect-loop-bound
9039 The minimum number of iterations under which loops are not vectorized
9040 when @option{-ftree-vectorize} is used. The number of iterations after
9041 vectorization needs to be greater than the value specified by this option
9042 to allow vectorization. The default value is 0.
9044 @item gcse-cost-distance-ratio
9045 Scaling factor in calculation of maximum distance an expression
9046 can be moved by GCSE optimizations. This is currently supported only in the
9047 code hoisting pass. The bigger the ratio, the more aggressive code hoisting
9048 is with simple expressions, i.e., the expressions that have cost
9049 less than @option{gcse-unrestricted-cost}. Specifying 0 disables
9050 hoisting of simple expressions. The default value is 10.
9052 @item gcse-unrestricted-cost
9053 Cost, roughly measured as the cost of a single typical machine
9054 instruction, at which GCSE optimizations do not constrain
9055 the distance an expression can travel. This is currently
9056 supported only in the code hoisting pass. The lesser the cost,
9057 the more aggressive code hoisting is. Specifying 0
9058 allows all expressions to travel unrestricted distances.
9059 The default value is 3.
9061 @item max-hoist-depth
9062 The depth of search in the dominator tree for expressions to hoist.
9063 This is used to avoid quadratic behavior in hoisting algorithm.
9064 The value of 0 does not limit on the search, but may slow down compilation
9065 of huge functions. The default value is 30.
9067 @item max-tail-merge-comparisons
9068 The maximum amount of similar bbs to compare a bb with. This is used to
9069 avoid quadratic behavior in tree tail merging. The default value is 10.
9071 @item max-tail-merge-iterations
9072 The maximum amount of iterations of the pass over the function. This is used to
9073 limit compilation time in tree tail merging. The default value is 2.
9075 @item max-unrolled-insns
9076 The maximum number of instructions that a loop may have to be unrolled.
9077 If a loop is unrolled, this parameter also determines how many times
9078 the loop code is unrolled.
9080 @item max-average-unrolled-insns
9081 The maximum number of instructions biased by probabilities of their execution
9082 that a loop may have to be unrolled. If a loop is unrolled,
9083 this parameter also determines how many times the loop code is unrolled.
9085 @item max-unroll-times
9086 The maximum number of unrollings of a single loop.
9088 @item max-peeled-insns
9089 The maximum number of instructions that a loop may have to be peeled.
9090 If a loop is peeled, this parameter also determines how many times
9091 the loop code is peeled.
9093 @item max-peel-times
9094 The maximum number of peelings of a single loop.
9096 @item max-peel-branches
9097 The maximum number of branches on the hot path through the peeled sequence.
9099 @item max-completely-peeled-insns
9100 The maximum number of insns of a completely peeled loop.
9102 @item max-completely-peel-times
9103 The maximum number of iterations of a loop to be suitable for complete peeling.
9105 @item max-completely-peel-loop-nest-depth
9106 The maximum depth of a loop nest suitable for complete peeling.
9108 @item max-unswitch-insns
9109 The maximum number of insns of an unswitched loop.
9111 @item max-unswitch-level
9112 The maximum number of branches unswitched in a single loop.
9114 @item max-loop-headers-insns
9115 The maximum number of insns in loop header duplicated by he copy loop headers
9119 The minimum cost of an expensive expression in the loop invariant motion.
9121 @item iv-consider-all-candidates-bound
9122 Bound on number of candidates for induction variables, below which
9123 all candidates are considered for each use in induction variable
9124 optimizations. If there are more candidates than this,
9125 only the most relevant ones are considered to avoid quadratic time complexity.
9127 @item iv-max-considered-uses
9128 The induction variable optimizations give up on loops that contain more
9129 induction variable uses.
9131 @item iv-always-prune-cand-set-bound
9132 If the number of candidates in the set is smaller than this value,
9133 always try to remove unnecessary ivs from the set
9134 when adding a new one.
9136 @item scev-max-expr-size
9137 Bound on size of expressions used in the scalar evolutions analyzer.
9138 Large expressions slow the analyzer.
9140 @item scev-max-expr-complexity
9141 Bound on the complexity of the expressions in the scalar evolutions analyzer.
9142 Complex expressions slow the analyzer.
9144 @item max-tree-if-conversion-phi-args
9145 Maximum number of arguments in a PHI supported by TREE if conversion
9146 unless the loop is marked with simd pragma.
9148 @item vect-max-version-for-alignment-checks
9149 The maximum number of run-time checks that can be performed when
9150 doing loop versioning for alignment in the vectorizer.
9152 @item vect-max-version-for-alias-checks
9153 The maximum number of run-time checks that can be performed when
9154 doing loop versioning for alias in the vectorizer.
9156 @item vect-max-peeling-for-alignment
9157 The maximum number of loop peels to enhance access alignment
9158 for vectorizer. Value -1 means no limit.
9160 @item max-iterations-to-track
9161 The maximum number of iterations of a loop the brute-force algorithm
9162 for analysis of the number of iterations of the loop tries to evaluate.
9164 @item hot-bb-count-ws-permille
9165 A basic block profile count is considered hot if it contributes to
9166 the given permillage (i.e. 0...1000) of the entire profiled execution.
9168 @item hot-bb-frequency-fraction
9169 Select fraction of the entry block frequency of executions of basic block in
9170 function given basic block needs to have to be considered hot.
9172 @item max-predicted-iterations
9173 The maximum number of loop iterations we predict statically. This is useful
9174 in cases where a function contains a single loop with known bound and
9175 another loop with unknown bound.
9176 The known number of iterations is predicted correctly, while
9177 the unknown number of iterations average to roughly 10. This means that the
9178 loop without bounds appears artificially cold relative to the other one.
9180 @item builtin-expect-probability
9181 Control the probability of the expression having the specified value. This
9182 parameter takes a percentage (i.e. 0 ... 100) as input.
9183 The default probability of 90 is obtained empirically.
9185 @item align-threshold
9187 Select fraction of the maximal frequency of executions of a basic block in
9188 a function to align the basic block.
9190 @item align-loop-iterations
9192 A loop expected to iterate at least the selected number of iterations is
9195 @item tracer-dynamic-coverage
9196 @itemx tracer-dynamic-coverage-feedback
9198 This value is used to limit superblock formation once the given percentage of
9199 executed instructions is covered. This limits unnecessary code size
9202 The @option{tracer-dynamic-coverage-feedback} parameter
9203 is used only when profile
9204 feedback is available. The real profiles (as opposed to statically estimated
9205 ones) are much less balanced allowing the threshold to be larger value.
9207 @item tracer-max-code-growth
9208 Stop tail duplication once code growth has reached given percentage. This is
9209 a rather artificial limit, as most of the duplicates are eliminated later in
9210 cross jumping, so it may be set to much higher values than is the desired code
9213 @item tracer-min-branch-ratio
9215 Stop reverse growth when the reverse probability of best edge is less than this
9216 threshold (in percent).
9218 @item tracer-min-branch-probability
9219 @itemx tracer-min-branch-probability-feedback
9221 Stop forward growth if the best edge has probability lower than this
9224 Similarly to @option{tracer-dynamic-coverage} two parameters are
9225 provided. @option{tracer-min-branch-probability-feedback} is used for
9226 compilation with profile feedback and @option{tracer-min-branch-probability}
9227 compilation without. The value for compilation with profile feedback
9228 needs to be more conservative (higher) in order to make tracer
9231 @item max-cse-path-length
9233 The maximum number of basic blocks on path that CSE considers.
9237 The maximum number of instructions CSE processes before flushing.
9238 The default is 1000.
9240 @item ggc-min-expand
9242 GCC uses a garbage collector to manage its own memory allocation. This
9243 parameter specifies the minimum percentage by which the garbage
9244 collector's heap should be allowed to expand between collections.
9245 Tuning this may improve compilation speed; it has no effect on code
9248 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
9249 RAM >= 1GB@. If @code{getrlimit} is available, the notion of ``RAM'' is
9250 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
9251 GCC is not able to calculate RAM on a particular platform, the lower
9252 bound of 30% is used. Setting this parameter and
9253 @option{ggc-min-heapsize} to zero causes a full collection to occur at
9254 every opportunity. This is extremely slow, but can be useful for
9257 @item ggc-min-heapsize
9259 Minimum size of the garbage collector's heap before it begins bothering
9260 to collect garbage. The first collection occurs after the heap expands
9261 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
9262 tuning this may improve compilation speed, and has no effect on code
9265 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit that
9266 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
9267 with a lower bound of 4096 (four megabytes) and an upper bound of
9268 131072 (128 megabytes). If GCC is not able to calculate RAM on a
9269 particular platform, the lower bound is used. Setting this parameter
9270 very large effectively disables garbage collection. Setting this
9271 parameter and @option{ggc-min-expand} to zero causes a full collection
9272 to occur at every opportunity.
9274 @item max-reload-search-insns
9275 The maximum number of instruction reload should look backward for equivalent
9276 register. Increasing values mean more aggressive optimization, making the
9277 compilation time increase with probably slightly better performance.
9278 The default value is 100.
9280 @item max-cselib-memory-locations
9281 The maximum number of memory locations cselib should take into account.
9282 Increasing values mean more aggressive optimization, making the compilation time
9283 increase with probably slightly better performance. The default value is 500.
9285 @item max-sched-ready-insns
9286 The maximum number of instructions ready to be issued the scheduler should
9287 consider at any given time during the first scheduling pass. Increasing
9288 values mean more thorough searches, making the compilation time increase
9289 with probably little benefit. The default value is 100.
9291 @item max-sched-region-blocks
9292 The maximum number of blocks in a region to be considered for
9293 interblock scheduling. The default value is 10.
9295 @item max-pipeline-region-blocks
9296 The maximum number of blocks in a region to be considered for
9297 pipelining in the selective scheduler. The default value is 15.
9299 @item max-sched-region-insns
9300 The maximum number of insns in a region to be considered for
9301 interblock scheduling. The default value is 100.
9303 @item max-pipeline-region-insns
9304 The maximum number of insns in a region to be considered for
9305 pipelining in the selective scheduler. The default value is 200.
9308 The minimum probability (in percents) of reaching a source block
9309 for interblock speculative scheduling. The default value is 40.
9311 @item max-sched-extend-regions-iters
9312 The maximum number of iterations through CFG to extend regions.
9313 A value of 0 (the default) disables region extensions.
9315 @item max-sched-insn-conflict-delay
9316 The maximum conflict delay for an insn to be considered for speculative motion.
9317 The default value is 3.
9319 @item sched-spec-prob-cutoff
9320 The minimal probability of speculation success (in percents), so that
9321 speculative insns are scheduled.
9322 The default value is 40.
9324 @item sched-state-edge-prob-cutoff
9325 The minimum probability an edge must have for the scheduler to save its
9327 The default value is 10.
9329 @item sched-mem-true-dep-cost
9330 Minimal distance (in CPU cycles) between store and load targeting same
9331 memory locations. The default value is 1.
9333 @item selsched-max-lookahead
9334 The maximum size of the lookahead window of selective scheduling. It is a
9335 depth of search for available instructions.
9336 The default value is 50.
9338 @item selsched-max-sched-times
9339 The maximum number of times that an instruction is scheduled during
9340 selective scheduling. This is the limit on the number of iterations
9341 through which the instruction may be pipelined. The default value is 2.
9343 @item selsched-insns-to-rename
9344 The maximum number of best instructions in the ready list that are considered
9345 for renaming in the selective scheduler. The default value is 2.
9348 The minimum value of stage count that swing modulo scheduler
9349 generates. The default value is 2.
9351 @item max-last-value-rtl
9352 The maximum size measured as number of RTLs that can be recorded in an expression
9353 in combiner for a pseudo register as last known value of that register. The default
9356 @item max-combine-insns
9357 The maximum number of instructions the RTL combiner tries to combine.
9358 The default value is 2 at @option{-Og} and 4 otherwise.
9360 @item integer-share-limit
9361 Small integer constants can use a shared data structure, reducing the
9362 compiler's memory usage and increasing its speed. This sets the maximum
9363 value of a shared integer constant. The default value is 256.
9365 @item ssp-buffer-size
9366 The minimum size of buffers (i.e.@: arrays) that receive stack smashing
9367 protection when @option{-fstack-protection} is used.
9369 @item min-size-for-stack-sharing
9370 The minimum size of variables taking part in stack slot sharing when not
9371 optimizing. The default value is 32.
9373 @item max-jump-thread-duplication-stmts
9374 Maximum number of statements allowed in a block that needs to be
9375 duplicated when threading jumps.
9377 @item max-fields-for-field-sensitive
9378 Maximum number of fields in a structure treated in
9379 a field sensitive manner during pointer analysis. The default is zero
9380 for @option{-O0} and @option{-O1},
9381 and 100 for @option{-Os}, @option{-O2}, and @option{-O3}.
9383 @item prefetch-latency
9384 Estimate on average number of instructions that are executed before
9385 prefetch finishes. The distance prefetched ahead is proportional
9386 to this constant. Increasing this number may also lead to less
9387 streams being prefetched (see @option{simultaneous-prefetches}).
9389 @item simultaneous-prefetches
9390 Maximum number of prefetches that can run at the same time.
9392 @item l1-cache-line-size
9393 The size of cache line in L1 cache, in bytes.
9396 The size of L1 cache, in kilobytes.
9399 The size of L2 cache, in kilobytes.
9401 @item min-insn-to-prefetch-ratio
9402 The minimum ratio between the number of instructions and the
9403 number of prefetches to enable prefetching in a loop.
9405 @item prefetch-min-insn-to-mem-ratio
9406 The minimum ratio between the number of instructions and the
9407 number of memory references to enable prefetching in a loop.
9409 @item use-canonical-types
9410 Whether the compiler should use the ``canonical'' type system. By
9411 default, this should always be 1, which uses a more efficient internal
9412 mechanism for comparing types in C++ and Objective-C++. However, if
9413 bugs in the canonical type system are causing compilation failures,
9414 set this value to 0 to disable canonical types.
9416 @item switch-conversion-max-branch-ratio
9417 Switch initialization conversion refuses to create arrays that are
9418 bigger than @option{switch-conversion-max-branch-ratio} times the number of
9419 branches in the switch.
9421 @item max-partial-antic-length
9422 Maximum length of the partial antic set computed during the tree
9423 partial redundancy elimination optimization (@option{-ftree-pre}) when
9424 optimizing at @option{-O3} and above. For some sorts of source code
9425 the enhanced partial redundancy elimination optimization can run away,
9426 consuming all of the memory available on the host machine. This
9427 parameter sets a limit on the length of the sets that are computed,
9428 which prevents the runaway behavior. Setting a value of 0 for
9429 this parameter allows an unlimited set length.
9431 @item sccvn-max-scc-size
9432 Maximum size of a strongly connected component (SCC) during SCCVN
9433 processing. If this limit is hit, SCCVN processing for the whole
9434 function is not done and optimizations depending on it are
9435 disabled. The default maximum SCC size is 10000.
9437 @item sccvn-max-alias-queries-per-access
9438 Maximum number of alias-oracle queries we perform when looking for
9439 redundancies for loads and stores. If this limit is hit the search
9440 is aborted and the load or store is not considered redundant. The
9441 number of queries is algorithmically limited to the number of
9442 stores on all paths from the load to the function entry.
9443 The default maximum number of queries is 1000.
9445 @item ira-max-loops-num
9446 IRA uses regional register allocation by default. If a function
9447 contains more loops than the number given by this parameter, only at most
9448 the given number of the most frequently-executed loops form regions
9449 for regional register allocation. The default value of the
9452 @item ira-max-conflict-table-size
9453 Although IRA uses a sophisticated algorithm to compress the conflict
9454 table, the table can still require excessive amounts of memory for
9455 huge functions. If the conflict table for a function could be more
9456 than the size in MB given by this parameter, the register allocator
9457 instead uses a faster, simpler, and lower-quality
9458 algorithm that does not require building a pseudo-register conflict table.
9459 The default value of the parameter is 2000.
9461 @item ira-loop-reserved-regs
9462 IRA can be used to evaluate more accurate register pressure in loops
9463 for decisions to move loop invariants (see @option{-O3}). The number
9464 of available registers reserved for some other purposes is given
9465 by this parameter. The default value of the parameter is 2, which is
9466 the minimal number of registers needed by typical instructions.
9467 This value is the best found from numerous experiments.
9469 @item lra-inheritance-ebb-probability-cutoff
9470 LRA tries to reuse values reloaded in registers in subsequent insns.
9471 This optimization is called inheritance. EBB is used as a region to
9472 do this optimization. The parameter defines a minimal fall-through
9473 edge probability in percentage used to add BB to inheritance EBB in
9474 LRA. The default value of the parameter is 40. The value was chosen
9475 from numerous runs of SPEC2000 on x86-64.
9477 @item loop-invariant-max-bbs-in-loop
9478 Loop invariant motion can be very expensive, both in compilation time and
9479 in amount of needed compile-time memory, with very large loops. Loops
9480 with more basic blocks than this parameter won't have loop invariant
9481 motion optimization performed on them. The default value of the
9482 parameter is 1000 for @option{-O1} and 10000 for @option{-O2} and above.
9484 @item loop-max-datarefs-for-datadeps
9485 Building data dependencies is expensive for very large loops. This
9486 parameter limits the number of data references in loops that are
9487 considered for data dependence analysis. These large loops are no
9488 handled by the optimizations using loop data dependencies.
9489 The default value is 1000.
9491 @item max-vartrack-size
9492 Sets a maximum number of hash table slots to use during variable
9493 tracking dataflow analysis of any function. If this limit is exceeded
9494 with variable tracking at assignments enabled, analysis for that
9495 function is retried without it, after removing all debug insns from
9496 the function. If the limit is exceeded even without debug insns, var
9497 tracking analysis is completely disabled for the function. Setting
9498 the parameter to zero makes it unlimited.
9500 @item max-vartrack-expr-depth
9501 Sets a maximum number of recursion levels when attempting to map
9502 variable names or debug temporaries to value expressions. This trades
9503 compilation time for more complete debug information. If this is set too
9504 low, value expressions that are available and could be represented in
9505 debug information may end up not being used; setting this higher may
9506 enable the compiler to find more complex debug expressions, but compile
9507 time and memory use may grow. The default is 12.
9509 @item min-nondebug-insn-uid
9510 Use uids starting at this parameter for nondebug insns. The range below
9511 the parameter is reserved exclusively for debug insns created by
9512 @option{-fvar-tracking-assignments}, but debug insns may get
9513 (non-overlapping) uids above it if the reserved range is exhausted.
9515 @item ipa-sra-ptr-growth-factor
9516 IPA-SRA replaces a pointer to an aggregate with one or more new
9517 parameters only when their cumulative size is less or equal to
9518 @option{ipa-sra-ptr-growth-factor} times the size of the original
9521 @item sra-max-scalarization-size-Ospeed
9522 @item sra-max-scalarization-size-Osize
9523 The two Scalar Reduction of Aggregates passes (SRA and IPA-SRA) aim to
9524 replace scalar parts of aggregates with uses of independent scalar
9525 variables. These parameters control the maximum size, in storage units,
9526 of aggregate which is considered for replacement when compiling for
9528 (@option{sra-max-scalarization-size-Ospeed}) or size
9529 (@option{sra-max-scalarization-size-Osize}) respectively.
9531 @item tm-max-aggregate-size
9532 When making copies of thread-local variables in a transaction, this
9533 parameter specifies the size in bytes after which variables are
9534 saved with the logging functions as opposed to save/restore code
9535 sequence pairs. This option only applies when using
9538 @item graphite-max-nb-scop-params
9539 To avoid exponential effects in the Graphite loop transforms, the
9540 number of parameters in a Static Control Part (SCoP) is bounded. The
9541 default value is 10 parameters. A variable whose value is unknown at
9542 compilation time and defined outside a SCoP is a parameter of the SCoP.
9544 @item graphite-max-bbs-per-function
9545 To avoid exponential effects in the detection of SCoPs, the size of
9546 the functions analyzed by Graphite is bounded. The default value is
9549 @item loop-block-tile-size
9550 Loop blocking or strip mining transforms, enabled with
9551 @option{-floop-block} or @option{-floop-strip-mine}, strip mine each
9552 loop in the loop nest by a given number of iterations. The strip
9553 length can be changed using the @option{loop-block-tile-size}
9554 parameter. The default value is 51 iterations.
9556 @item loop-unroll-jam-size
9557 Specify the unroll factor for the @option{-floop-unroll-and-jam} option. The
9560 @item loop-unroll-jam-depth
9561 Specify the dimension to be unrolled (counting from the most inner loop)
9562 for the @option{-floop-unroll-and-jam}. The default value is 2.
9564 @item ipa-cp-value-list-size
9565 IPA-CP attempts to track all possible values and types passed to a function's
9566 parameter in order to propagate them and perform devirtualization.
9567 @option{ipa-cp-value-list-size} is the maximum number of values and types it
9568 stores per one formal parameter of a function.
9570 @item ipa-cp-eval-threshold
9571 IPA-CP calculates its own score of cloning profitability heuristics
9572 and performs those cloning opportunities with scores that exceed
9573 @option{ipa-cp-eval-threshold}.
9575 @item ipa-cp-recursion-penalty
9576 Percentage penalty the recursive functions will receive when they
9577 are evaluated for cloning.
9579 @item ipa-cp-single-call-penalty
9580 Percentage penalty functions containg a single call to another
9581 function will receive when they are evaluated for cloning.
9584 @item ipa-max-agg-items
9585 IPA-CP is also capable to propagate a number of scalar values passed
9586 in an aggregate. @option{ipa-max-agg-items} controls the maximum
9587 number of such values per one parameter.
9589 @item ipa-cp-loop-hint-bonus
9590 When IPA-CP determines that a cloning candidate would make the number
9591 of iterations of a loop known, it adds a bonus of
9592 @option{ipa-cp-loop-hint-bonus} to the profitability score of
9595 @item ipa-cp-array-index-hint-bonus
9596 When IPA-CP determines that a cloning candidate would make the index of
9597 an array access known, it adds a bonus of
9598 @option{ipa-cp-array-index-hint-bonus} to the profitability
9599 score of the candidate.
9601 @item ipa-max-aa-steps
9602 During its analysis of function bodies, IPA-CP employs alias analysis
9603 in order to track values pointed to by function parameters. In order
9604 not spend too much time analyzing huge functions, it gives up and
9605 consider all memory clobbered after examining
9606 @option{ipa-max-aa-steps} statements modifying memory.
9608 @item lto-partitions
9609 Specify desired number of partitions produced during WHOPR compilation.
9610 The number of partitions should exceed the number of CPUs used for compilation.
9611 The default value is 32.
9613 @item lto-min-partition
9614 Size of minimal partition for WHOPR (in estimated instructions).
9615 This prevents expenses of splitting very small programs into too many
9618 @item lto-max-partition
9619 Size of max partition for WHOPR (in estimated instructions).
9620 to provide an upper bound for individual size of partition.
9621 Meant to be used only with balanced partitioning.
9623 @item cxx-max-namespaces-for-diagnostic-help
9624 The maximum number of namespaces to consult for suggestions when C++
9625 name lookup fails for an identifier. The default is 1000.
9627 @item sink-frequency-threshold
9628 The maximum relative execution frequency (in percents) of the target block
9629 relative to a statement's original block to allow statement sinking of a
9630 statement. Larger numbers result in more aggressive statement sinking.
9631 The default value is 75. A small positive adjustment is applied for
9632 statements with memory operands as those are even more profitable so sink.
9634 @item max-stores-to-sink
9635 The maximum number of conditional store pairs that can be sunk. Set to 0
9636 if either vectorization (@option{-ftree-vectorize}) or if-conversion
9637 (@option{-ftree-loop-if-convert}) is disabled. The default is 2.
9639 @item allow-store-data-races
9640 Allow optimizers to introduce new data races on stores.
9641 Set to 1 to allow, otherwise to 0. This option is enabled by default
9642 at optimization level @option{-Ofast}.
9644 @item case-values-threshold
9645 The smallest number of different values for which it is best to use a
9646 jump-table instead of a tree of conditional branches. If the value is
9647 0, use the default for the machine. The default is 0.
9649 @item tree-reassoc-width
9650 Set the maximum number of instructions executed in parallel in
9651 reassociated tree. This parameter overrides target dependent
9652 heuristics used by default if has non zero value.
9654 @item sched-pressure-algorithm
9655 Choose between the two available implementations of
9656 @option{-fsched-pressure}. Algorithm 1 is the original implementation
9657 and is the more likely to prevent instructions from being reordered.
9658 Algorithm 2 was designed to be a compromise between the relatively
9659 conservative approach taken by algorithm 1 and the rather aggressive
9660 approach taken by the default scheduler. It relies more heavily on
9661 having a regular register file and accurate register pressure classes.
9662 See @file{haifa-sched.c} in the GCC sources for more details.
9664 The default choice depends on the target.
9666 @item max-slsr-cand-scan
9667 Set the maximum number of existing candidates that are considered when
9668 seeking a basis for a new straight-line strength reduction candidate.
9671 Enable buffer overflow detection for global objects. This kind
9672 of protection is enabled by default if you are using
9673 @option{-fsanitize=address} option.
9674 To disable global objects protection use @option{--param asan-globals=0}.
9677 Enable buffer overflow detection for stack objects. This kind of
9678 protection is enabled by default when using @option{-fsanitize=address}.
9679 To disable stack protection use @option{--param asan-stack=0} option.
9681 @item asan-instrument-reads
9682 Enable buffer overflow detection for memory reads. This kind of
9683 protection is enabled by default when using @option{-fsanitize=address}.
9684 To disable memory reads protection use
9685 @option{--param asan-instrument-reads=0}.
9687 @item asan-instrument-writes
9688 Enable buffer overflow detection for memory writes. This kind of
9689 protection is enabled by default when using @option{-fsanitize=address}.
9690 To disable memory writes protection use
9691 @option{--param asan-instrument-writes=0} option.
9693 @item asan-memintrin
9694 Enable detection for built-in functions. This kind of protection
9695 is enabled by default when using @option{-fsanitize=address}.
9696 To disable built-in functions protection use
9697 @option{--param asan-memintrin=0}.
9699 @item asan-use-after-return
9700 Enable detection of use-after-return. This kind of protection
9701 is enabled by default when using @option{-fsanitize=address} option.
9702 To disable use-after-return detection use
9703 @option{--param asan-use-after-return=0}.
9705 @item asan-instrumentation-with-call-threshold
9706 If number of memory accesses in function being instrumented
9707 is greater or equal to this number, use callbacks instead of inline checks.
9708 E.g. to disable inline code use
9709 @option{--param asan-instrumentation-with-call-threshold=0}.
9711 @item chkp-max-ctor-size
9712 Static constructors generated by Pointer Bounds Checker may become very
9713 large and significantly increase compile time at optimization level
9714 @option{-O1} and higher. This parameter is a maximum nubmer of statements
9715 in a single generated constructor. Default value is 5000.
9717 @item max-fsm-thread-path-insns
9718 Maximum number of instructions to copy when duplicating blocks on a
9719 finite state automaton jump thread path. The default is 100.
9721 @item max-fsm-thread-length
9722 Maximum number of basic blocks on a finite state automaton jump thread
9723 path. The default is 10.
9725 @item max-fsm-thread-paths
9726 Maximum number of new jump thread paths to create for a finite state
9727 automaton. The default is 50.
9729 @item parloops-chunk-size
9730 Chunk size of omp schedule for loops parallelized by parloops. The default
9733 @item parloops-schedule
9734 Schedule type of omp schedule for loops parallelized by parloops (static,
9735 dynamic, guided, auto, runtime). The default is static.
9737 @item max-ssa-name-query-depth
9738 Maximum depth of recursion when querying properties of SSA names in things
9739 like fold routines. One level of recursion corresponds to following a
9742 @item hsa-gen-debug-stores
9743 Enable emission of special debug stores within HSA kernels which are
9744 then read and reported by libgomp plugin. Generation of these stores
9745 is disabled by default, use @option{--param hsa-gen-debug-stores=1} to
9748 @item max-speculative-devirt-maydefs
9749 The maximum number of may-defs we analyze when looking for a must-def
9750 specifying the dynamic type of an object that invokes a virtual call
9751 we may be able to devirtualize speculatively.
9755 @node Instrumentation Options
9756 @section Program Instrumentation Options
9757 @cindex instrumentation options
9758 @cindex program instrumentation options
9759 @cindex run-time error checking options
9760 @cindex profiling options
9761 @cindex options, program instrumentation
9762 @cindex options, run-time error checking
9763 @cindex options, profiling
9765 GCC supports a number of command-line options that control adding
9766 run-time instrumentation to the code it normally generates.
9767 For example, one purpose of instrumentation is collect profiling
9768 statistics for use in finding program hot spots, code coverage
9769 analysis, or profile-guided optimizations.
9770 Another class of program instrumentation is adding run-time checking
9771 to detect programming errors like invalid pointer
9772 dereferences or out-of-bounds array accesses, as well as deliberately
9773 hostile attacks such as stack smashing or C++ vtable hijacking.
9774 There is also a general hook which can be used to implement other
9775 forms of tracing or function-level instrumentation for debug or
9776 program analysis purposes.
9779 @cindex @command{prof}
9782 Generate extra code to write profile information suitable for the
9783 analysis program @command{prof}. You must use this option when compiling
9784 the source files you want data about, and you must also use it when
9787 @cindex @command{gprof}
9790 Generate extra code to write profile information suitable for the
9791 analysis program @command{gprof}. You must use this option when compiling
9792 the source files you want data about, and you must also use it when
9795 @item -fprofile-arcs
9796 @opindex fprofile-arcs
9797 Add code so that program flow @dfn{arcs} are instrumented. During
9798 execution the program records how many times each branch and call is
9799 executed and how many times it is taken or returns. When the compiled
9800 program exits it saves this data to a file called
9801 @file{@var{auxname}.gcda} for each source file. The data may be used for
9802 profile-directed optimizations (@option{-fbranch-probabilities}), or for
9803 test coverage analysis (@option{-ftest-coverage}). Each object file's
9804 @var{auxname} is generated from the name of the output file, if
9805 explicitly specified and it is not the final executable, otherwise it is
9806 the basename of the source file. In both cases any suffix is removed
9807 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
9808 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
9809 @xref{Cross-profiling}.
9811 @cindex @command{gcov}
9815 This option is used to compile and link code instrumented for coverage
9816 analysis. The option is a synonym for @option{-fprofile-arcs}
9817 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
9818 linking). See the documentation for those options for more details.
9823 Compile the source files with @option{-fprofile-arcs} plus optimization
9824 and code generation options. For test coverage analysis, use the
9825 additional @option{-ftest-coverage} option. You do not need to profile
9826 every source file in a program.
9829 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
9830 (the latter implies the former).
9833 Run the program on a representative workload to generate the arc profile
9834 information. This may be repeated any number of times. You can run
9835 concurrent instances of your program, and provided that the file system
9836 supports locking, the data files will be correctly updated. Also
9837 @code{fork} calls are detected and correctly handled (double counting
9841 For profile-directed optimizations, compile the source files again with
9842 the same optimization and code generation options plus
9843 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
9844 Control Optimization}).
9847 For test coverage analysis, use @command{gcov} to produce human readable
9848 information from the @file{.gcno} and @file{.gcda} files. Refer to the
9849 @command{gcov} documentation for further information.
9853 With @option{-fprofile-arcs}, for each function of your program GCC
9854 creates a program flow graph, then finds a spanning tree for the graph.
9855 Only arcs that are not on the spanning tree have to be instrumented: the
9856 compiler adds code to count the number of times that these arcs are
9857 executed. When an arc is the only exit or only entrance to a block, the
9858 instrumentation code can be added to the block; otherwise, a new basic
9859 block must be created to hold the instrumentation code.
9862 @item -ftest-coverage
9863 @opindex ftest-coverage
9864 Produce a notes file that the @command{gcov} code-coverage utility
9865 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
9866 show program coverage. Each source file's note file is called
9867 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
9868 above for a description of @var{auxname} and instructions on how to
9869 generate test coverage data. Coverage data matches the source files
9870 more closely if you do not optimize.
9872 @item -fprofile-dir=@var{path}
9873 @opindex fprofile-dir
9875 Set the directory to search for the profile data files in to @var{path}.
9876 This option affects only the profile data generated by
9877 @option{-fprofile-generate}, @option{-ftest-coverage}, @option{-fprofile-arcs}
9878 and used by @option{-fprofile-use} and @option{-fbranch-probabilities}
9879 and its related options. Both absolute and relative paths can be used.
9880 By default, GCC uses the current directory as @var{path}, thus the
9881 profile data file appears in the same directory as the object file.
9883 @item -fprofile-generate
9884 @itemx -fprofile-generate=@var{path}
9885 @opindex fprofile-generate
9887 Enable options usually used for instrumenting application to produce
9888 profile useful for later recompilation with profile feedback based
9889 optimization. You must use @option{-fprofile-generate} both when
9890 compiling and when linking your program.
9892 The following options are enabled: @option{-fprofile-arcs}, @option{-fprofile-values}, @option{-fvpt}.
9894 If @var{path} is specified, GCC looks at the @var{path} to find
9895 the profile feedback data files. See @option{-fprofile-dir}.
9897 To optimize the program based on the collected profile information, use
9898 @option{-fprofile-use}. @xref{Optimize Options}, for more information.
9900 @item -fsanitize=address
9901 @opindex fsanitize=address
9902 Enable AddressSanitizer, a fast memory error detector.
9903 Memory access instructions are instrumented to detect
9904 out-of-bounds and use-after-free bugs.
9905 See @uref{https://github.com/google/sanitizers/wiki/AddressSanitizer} for
9906 more details. The run-time behavior can be influenced using the
9907 @env{ASAN_OPTIONS} environment variable. When set to @code{help=1},
9908 the available options are shown at startup of the instrumented program. See
9909 @url{https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags}
9910 for a list of supported options.
9912 @item -fsanitize=kernel-address
9913 @opindex fsanitize=kernel-address
9914 Enable AddressSanitizer for Linux kernel.
9915 See @uref{https://github.com/google/kasan/wiki} for more details.
9917 @item -fsanitize=thread
9918 @opindex fsanitize=thread
9919 Enable ThreadSanitizer, a fast data race detector.
9920 Memory access instructions are instrumented to detect
9921 data race bugs. See @uref{https://github.com/google/sanitizers/wiki#threadsanitizer} for more
9922 details. The run-time behavior can be influenced using the @env{TSAN_OPTIONS}
9923 environment variable; see
9924 @url{https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags} for a list of
9927 @item -fsanitize=leak
9928 @opindex fsanitize=leak
9929 Enable LeakSanitizer, a memory leak detector.
9930 This option only matters for linking of executables and if neither
9931 @option{-fsanitize=address} nor @option{-fsanitize=thread} is used. In that
9932 case the executable is linked against a library that overrides @code{malloc}
9933 and other allocator functions. See
9934 @uref{https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer} for more
9935 details. The run-time behavior can be influenced using the
9936 @env{LSAN_OPTIONS} environment variable.
9938 @item -fsanitize=undefined
9939 @opindex fsanitize=undefined
9940 Enable UndefinedBehaviorSanitizer, a fast undefined behavior detector.
9941 Various computations are instrumented to detect undefined behavior
9942 at runtime. Current suboptions are:
9946 @item -fsanitize=shift
9947 @opindex fsanitize=shift
9948 This option enables checking that the result of a shift operation is
9949 not undefined. Note that what exactly is considered undefined differs
9950 slightly between C and C++, as well as between ISO C90 and C99, etc.
9952 @item -fsanitize=integer-divide-by-zero
9953 @opindex fsanitize=integer-divide-by-zero
9954 Detect integer division by zero as well as @code{INT_MIN / -1} division.
9956 @item -fsanitize=unreachable
9957 @opindex fsanitize=unreachable
9958 With this option, the compiler turns the @code{__builtin_unreachable}
9959 call into a diagnostics message call instead. When reaching the
9960 @code{__builtin_unreachable} call, the behavior is undefined.
9962 @item -fsanitize=vla-bound
9963 @opindex fsanitize=vla-bound
9964 This option instructs the compiler to check that the size of a variable
9965 length array is positive.
9967 @item -fsanitize=null
9968 @opindex fsanitize=null
9969 This option enables pointer checking. Particularly, the application
9970 built with this option turned on will issue an error message when it
9971 tries to dereference a NULL pointer, or if a reference (possibly an
9972 rvalue reference) is bound to a NULL pointer, or if a method is invoked
9973 on an object pointed by a NULL pointer.
9975 @item -fsanitize=return
9976 @opindex fsanitize=return
9977 This option enables return statement checking. Programs
9978 built with this option turned on will issue an error message
9979 when the end of a non-void function is reached without actually
9980 returning a value. This option works in C++ only.
9982 @item -fsanitize=signed-integer-overflow
9983 @opindex fsanitize=signed-integer-overflow
9984 This option enables signed integer overflow checking. We check that
9985 the result of @code{+}, @code{*}, and both unary and binary @code{-}
9986 does not overflow in the signed arithmetics. Note, integer promotion
9987 rules must be taken into account. That is, the following is not an
9990 signed char a = SCHAR_MAX;
9994 @item -fsanitize=bounds
9995 @opindex fsanitize=bounds
9996 This option enables instrumentation of array bounds. Various out of bounds
9997 accesses are detected. Flexible array members, flexible array member-like
9998 arrays, and initializers of variables with static storage are not instrumented.
10000 @item -fsanitize=bounds-strict
10001 @opindex fsanitize=bounds-strict
10002 This option enables strict instrumentation of array bounds. Most out of bounds
10003 accesses are detected, including flexible array members and flexible array
10004 member-like arrays. Initializers of variables with static storage are not
10007 @item -fsanitize=alignment
10008 @opindex fsanitize=alignment
10010 This option enables checking of alignment of pointers when they are
10011 dereferenced, or when a reference is bound to insufficiently aligned target,
10012 or when a method or constructor is invoked on insufficiently aligned object.
10014 @item -fsanitize=object-size
10015 @opindex fsanitize=object-size
10016 This option enables instrumentation of memory references using the
10017 @code{__builtin_object_size} function. Various out of bounds pointer
10018 accesses are detected.
10020 @item -fsanitize=float-divide-by-zero
10021 @opindex fsanitize=float-divide-by-zero
10022 Detect floating-point division by zero. Unlike other similar options,
10023 @option{-fsanitize=float-divide-by-zero} is not enabled by
10024 @option{-fsanitize=undefined}, since floating-point division by zero can
10025 be a legitimate way of obtaining infinities and NaNs.
10027 @item -fsanitize=float-cast-overflow
10028 @opindex fsanitize=float-cast-overflow
10029 This option enables floating-point type to integer conversion checking.
10030 We check that the result of the conversion does not overflow.
10031 Unlike other similar options, @option{-fsanitize=float-cast-overflow} is
10032 not enabled by @option{-fsanitize=undefined}.
10033 This option does not work well with @code{FE_INVALID} exceptions enabled.
10035 @item -fsanitize=nonnull-attribute
10036 @opindex fsanitize=nonnull-attribute
10038 This option enables instrumentation of calls, checking whether null values
10039 are not passed to arguments marked as requiring a non-null value by the
10040 @code{nonnull} function attribute.
10042 @item -fsanitize=returns-nonnull-attribute
10043 @opindex fsanitize=returns-nonnull-attribute
10045 This option enables instrumentation of return statements in functions
10046 marked with @code{returns_nonnull} function attribute, to detect returning
10047 of null values from such functions.
10049 @item -fsanitize=bool
10050 @opindex fsanitize=bool
10052 This option enables instrumentation of loads from bool. If a value other
10053 than 0/1 is loaded, a run-time error is issued.
10055 @item -fsanitize=enum
10056 @opindex fsanitize=enum
10058 This option enables instrumentation of loads from an enum type. If
10059 a value outside the range of values for the enum type is loaded,
10060 a run-time error is issued.
10062 @item -fsanitize=vptr
10063 @opindex fsanitize=vptr
10065 This option enables instrumentation of C++ member function calls, member
10066 accesses and some conversions between pointers to base and derived classes,
10067 to verify the referenced object has the correct dynamic type.
10071 While @option{-ftrapv} causes traps for signed overflows to be emitted,
10072 @option{-fsanitize=undefined} gives a diagnostic message.
10073 This currently works only for the C family of languages.
10075 @item -fno-sanitize=all
10076 @opindex fno-sanitize=all
10078 This option disables all previously enabled sanitizers.
10079 @option{-fsanitize=all} is not allowed, as some sanitizers cannot be used
10082 @item -fasan-shadow-offset=@var{number}
10083 @opindex fasan-shadow-offset
10084 This option forces GCC to use custom shadow offset in AddressSanitizer checks.
10085 It is useful for experimenting with different shadow memory layouts in
10086 Kernel AddressSanitizer.
10088 @item -fsanitize-sections=@var{s1},@var{s2},...
10089 @opindex fsanitize-sections
10090 Sanitize global variables in selected user-defined sections. @var{si} may
10093 @item -fsanitize-recover@r{[}=@var{opts}@r{]}
10094 @opindex fsanitize-recover
10095 @opindex fno-sanitize-recover
10096 @option{-fsanitize-recover=} controls error recovery mode for sanitizers
10097 mentioned in comma-separated list of @var{opts}. Enabling this option
10098 for a sanitizer component causes it to attempt to continue
10099 running the program as if no error happened. This means multiple
10100 runtime errors can be reported in a single program run, and the exit
10101 code of the program may indicate success even when errors
10102 have been reported. The @option{-fno-sanitize-recover=} option
10103 can be used to alter
10104 this behavior: only the first detected error is reported
10105 and program then exits with a non-zero exit code.
10107 Currently this feature only works for @option{-fsanitize=undefined} (and its suboptions
10108 except for @option{-fsanitize=unreachable} and @option{-fsanitize=return}),
10109 @option{-fsanitize=float-cast-overflow}, @option{-fsanitize=float-divide-by-zero},
10110 @option{-fsanitize=kernel-address} and @option{-fsanitize=address}.
10111 For these sanitizers error recovery is turned on by default, except @option{-fsanitize=address},
10112 for which this feature is experimental.
10113 @option{-fsanitize-recover=all} and @option{-fno-sanitize-recover=all} is also
10114 accepted, the former enables recovery for all sanitizers that support it,
10115 the latter disables recovery for all sanitizers that support it.
10117 Even if a recovery mode is turned on the compiler side, it needs to be also
10118 enabled on the runtime library side, otherwise the failures are still fatal.
10119 The runtime library defaults to @code{halt_on_error=0} for
10120 ThreadSanitizer and UndefinedBehaviorSanitizer, while default value for
10121 AddressSanitizer is @code{halt_on_error=1}. This can be overridden through
10122 setting the @code{halt_on_error} flag in the corresponding environment variable.
10124 Syntax without explicit @var{opts} parameter is deprecated. It is equivalent to
10126 -fsanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
10129 Similarly @option{-fno-sanitize-recover} is equivalent to
10131 -fno-sanitize-recover=undefined,float-cast-overflow,float-divide-by-zero
10134 @item -fsanitize-undefined-trap-on-error
10135 @opindex fsanitize-undefined-trap-on-error
10136 The @option{-fsanitize-undefined-trap-on-error} option instructs the compiler to
10137 report undefined behavior using @code{__builtin_trap} rather than
10138 a @code{libubsan} library routine. The advantage of this is that the
10139 @code{libubsan} library is not needed and is not linked in, so this
10140 is usable even in freestanding environments.
10142 @item -fsanitize-coverage=trace-pc
10143 @opindex fsanitize-coverage=trace-pc
10144 Enable coverage-guided fuzzing code instrumentation.
10145 Inserts a call to @code{__sanitizer_cov_trace_pc} into every basic block.
10147 @item -fbounds-check
10148 @opindex fbounds-check
10149 For front ends that support it, generate additional code to check that
10150 indices used to access arrays are within the declared range. This is
10151 currently only supported by the Java and Fortran front ends, where
10152 this option defaults to true and false respectively.
10154 @item -fcheck-pointer-bounds
10155 @opindex fcheck-pointer-bounds
10156 @opindex fno-check-pointer-bounds
10157 @cindex Pointer Bounds Checker options
10158 Enable Pointer Bounds Checker instrumentation. Each memory reference
10159 is instrumented with checks of the pointer used for memory access against
10160 bounds associated with that pointer.
10163 is only an implementation for Intel MPX available, thus x86 GNU/Linux target
10164 and @option{-mmpx} are required to enable this feature.
10165 MPX-based instrumentation requires
10166 a runtime library to enable MPX in hardware and handle bounds
10167 violation signals. By default when @option{-fcheck-pointer-bounds}
10168 and @option{-mmpx} options are used to link a program, the GCC driver
10169 links against the @file{libmpx} and @file{libmpxwrappers} libraries.
10170 Bounds checking on calls to dynamic libraries requires a linker
10171 with @option{-z bndplt} support; if GCC was configured with a linker
10172 without support for this option (including the Gold linker and older
10173 versions of ld), a warning is given if you link with @option{-mmpx}
10174 without also specifying @option{-static}, since the overall effectiveness
10175 of the bounds checking protection is reduced.
10176 See also @option{-static-libmpxwrappers}.
10178 MPX-based instrumentation
10179 may be used for debugging and also may be included in production code
10180 to increase program security. Depending on usage, you may
10181 have different requirements for the runtime library. The current version
10182 of the MPX runtime library is more oriented for use as a debugging
10183 tool. MPX runtime library usage implies @option{-lpthread}. See
10184 also @option{-static-libmpx}. The runtime library behavior can be
10185 influenced using various @env{CHKP_RT_*} environment variables. See
10186 @uref{https://gcc.gnu.org/wiki/Intel%20MPX%20support%20in%20the%20GCC%20compiler}
10189 Generated instrumentation may be controlled by various
10190 @option{-fchkp-*} options and by the @code{bnd_variable_size}
10191 structure field attribute (@pxref{Type Attributes}) and
10192 @code{bnd_legacy}, and @code{bnd_instrument} function attributes
10193 (@pxref{Function Attributes}). GCC also provides a number of built-in
10194 functions for controlling the Pointer Bounds Checker. @xref{Pointer
10195 Bounds Checker builtins}, for more information.
10197 @item -fchkp-check-incomplete-type
10198 @opindex fchkp-check-incomplete-type
10199 @opindex fno-chkp-check-incomplete-type
10200 Generate pointer bounds checks for variables with incomplete type.
10201 Enabled by default.
10203 @item -fchkp-narrow-bounds
10204 @opindex fchkp-narrow-bounds
10205 @opindex fno-chkp-narrow-bounds
10206 Controls bounds used by Pointer Bounds Checker for pointers to object
10207 fields. If narrowing is enabled then field bounds are used. Otherwise
10208 object bounds are used. See also @option{-fchkp-narrow-to-innermost-array}
10209 and @option{-fchkp-first-field-has-own-bounds}. Enabled by default.
10211 @item -fchkp-first-field-has-own-bounds
10212 @opindex fchkp-first-field-has-own-bounds
10213 @opindex fno-chkp-first-field-has-own-bounds
10214 Forces Pointer Bounds Checker to use narrowed bounds for the address of the
10215 first field in the structure. By default a pointer to the first field has
10216 the same bounds as a pointer to the whole structure.
10218 @item -fchkp-narrow-to-innermost-array
10219 @opindex fchkp-narrow-to-innermost-array
10220 @opindex fno-chkp-narrow-to-innermost-array
10221 Forces Pointer Bounds Checker to use bounds of the innermost arrays in
10222 case of nested static array access. By default this option is disabled and
10223 bounds of the outermost array are used.
10225 @item -fchkp-optimize
10226 @opindex fchkp-optimize
10227 @opindex fno-chkp-optimize
10228 Enables Pointer Bounds Checker optimizations. Enabled by default at
10229 optimization levels @option{-O}, @option{-O2}, @option{-O3}.
10231 @item -fchkp-use-fast-string-functions
10232 @opindex fchkp-use-fast-string-functions
10233 @opindex fno-chkp-use-fast-string-functions
10234 Enables use of @code{*_nobnd} versions of string functions (not copying bounds)
10235 by Pointer Bounds Checker. Disabled by default.
10237 @item -fchkp-use-nochk-string-functions
10238 @opindex fchkp-use-nochk-string-functions
10239 @opindex fno-chkp-use-nochk-string-functions
10240 Enables use of @code{*_nochk} versions of string functions (not checking bounds)
10241 by Pointer Bounds Checker. Disabled by default.
10243 @item -fchkp-use-static-bounds
10244 @opindex fchkp-use-static-bounds
10245 @opindex fno-chkp-use-static-bounds
10246 Allow Pointer Bounds Checker to generate static bounds holding
10247 bounds of static variables. Enabled by default.
10249 @item -fchkp-use-static-const-bounds
10250 @opindex fchkp-use-static-const-bounds
10251 @opindex fno-chkp-use-static-const-bounds
10252 Use statically-initialized bounds for constant bounds instead of
10253 generating them each time they are required. By default enabled when
10254 @option{-fchkp-use-static-bounds} is enabled.
10256 @item -fchkp-treat-zero-dynamic-size-as-infinite
10257 @opindex fchkp-treat-zero-dynamic-size-as-infinite
10258 @opindex fno-chkp-treat-zero-dynamic-size-as-infinite
10259 With this option, objects with incomplete type whose
10260 dynamically-obtained size is zero are treated as having infinite size
10261 instead by Pointer Bounds
10262 Checker. This option may be helpful if a program is linked with a library
10263 missing size information for some symbols. Disabled by default.
10265 @item -fchkp-check-read
10266 @opindex fchkp-check-read
10267 @opindex fno-chkp-check-read
10268 Instructs Pointer Bounds Checker to generate checks for all read
10269 accesses to memory. Enabled by default.
10271 @item -fchkp-check-write
10272 @opindex fchkp-check-write
10273 @opindex fno-chkp-check-write
10274 Instructs Pointer Bounds Checker to generate checks for all write
10275 accesses to memory. Enabled by default.
10277 @item -fchkp-store-bounds
10278 @opindex fchkp-store-bounds
10279 @opindex fno-chkp-store-bounds
10280 Instructs Pointer Bounds Checker to generate bounds stores for
10281 pointer writes. Enabled by default.
10283 @item -fchkp-instrument-calls
10284 @opindex fchkp-instrument-calls
10285 @opindex fno-chkp-instrument-calls
10286 Instructs Pointer Bounds Checker to pass pointer bounds to calls.
10287 Enabled by default.
10289 @item -fchkp-instrument-marked-only
10290 @opindex fchkp-instrument-marked-only
10291 @opindex fno-chkp-instrument-marked-only
10292 Instructs Pointer Bounds Checker to instrument only functions
10293 marked with the @code{bnd_instrument} attribute
10294 (@pxref{Function Attributes}). Disabled by default.
10296 @item -fchkp-use-wrappers
10297 @opindex fchkp-use-wrappers
10298 @opindex fno-chkp-use-wrappers
10299 Allows Pointer Bounds Checker to replace calls to built-in functions
10300 with calls to wrapper functions. When @option{-fchkp-use-wrappers}
10301 is used to link a program, the GCC driver automatically links
10302 against @file{libmpxwrappers}. See also @option{-static-libmpxwrappers}.
10303 Enabled by default.
10305 @item -fstack-protector
10306 @opindex fstack-protector
10307 Emit extra code to check for buffer overflows, such as stack smashing
10308 attacks. This is done by adding a guard variable to functions with
10309 vulnerable objects. This includes functions that call @code{alloca}, and
10310 functions with buffers larger than 8 bytes. The guards are initialized
10311 when a function is entered and then checked when the function exits.
10312 If a guard check fails, an error message is printed and the program exits.
10314 @item -fstack-protector-all
10315 @opindex fstack-protector-all
10316 Like @option{-fstack-protector} except that all functions are protected.
10318 @item -fstack-protector-strong
10319 @opindex fstack-protector-strong
10320 Like @option{-fstack-protector} but includes additional functions to
10321 be protected --- those that have local array definitions, or have
10322 references to local frame addresses.
10324 @item -fstack-protector-explicit
10325 @opindex fstack-protector-explicit
10326 Like @option{-fstack-protector} but only protects those functions which
10327 have the @code{stack_protect} attribute.
10329 @item -fstack-check
10330 @opindex fstack-check
10331 Generate code to verify that you do not go beyond the boundary of the
10332 stack. You should specify this flag if you are running in an
10333 environment with multiple threads, but you only rarely need to specify it in
10334 a single-threaded environment since stack overflow is automatically
10335 detected on nearly all systems if there is only one stack.
10337 Note that this switch does not actually cause checking to be done; the
10338 operating system or the language runtime must do that. The switch causes
10339 generation of code to ensure that they see the stack being extended.
10341 You can additionally specify a string parameter: @samp{no} means no
10342 checking, @samp{generic} means force the use of old-style checking,
10343 @samp{specific} means use the best checking method and is equivalent
10344 to bare @option{-fstack-check}.
10346 Old-style checking is a generic mechanism that requires no specific
10347 target support in the compiler but comes with the following drawbacks:
10351 Modified allocation strategy for large objects: they are always
10352 allocated dynamically if their size exceeds a fixed threshold.
10355 Fixed limit on the size of the static frame of functions: when it is
10356 topped by a particular function, stack checking is not reliable and
10357 a warning is issued by the compiler.
10360 Inefficiency: because of both the modified allocation strategy and the
10361 generic implementation, code performance is hampered.
10364 Note that old-style stack checking is also the fallback method for
10365 @samp{specific} if no target support has been added in the compiler.
10367 @item -fstack-limit-register=@var{reg}
10368 @itemx -fstack-limit-symbol=@var{sym}
10369 @itemx -fno-stack-limit
10370 @opindex fstack-limit-register
10371 @opindex fstack-limit-symbol
10372 @opindex fno-stack-limit
10373 Generate code to ensure that the stack does not grow beyond a certain value,
10374 either the value of a register or the address of a symbol. If a larger
10375 stack is required, a signal is raised at run time. For most targets,
10376 the signal is raised before the stack overruns the boundary, so
10377 it is possible to catch the signal without taking special precautions.
10379 For instance, if the stack starts at absolute address @samp{0x80000000}
10380 and grows downwards, you can use the flags
10381 @option{-fstack-limit-symbol=__stack_limit} and
10382 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
10383 of 128KB@. Note that this may only work with the GNU linker.
10385 You can locally override stack limit checking by using the
10386 @code{no_stack_limit} function attribute (@pxref{Function Attributes}).
10388 @item -fsplit-stack
10389 @opindex fsplit-stack
10390 Generate code to automatically split the stack before it overflows.
10391 The resulting program has a discontiguous stack which can only
10392 overflow if the program is unable to allocate any more memory. This
10393 is most useful when running threaded programs, as it is no longer
10394 necessary to calculate a good stack size to use for each thread. This
10395 is currently only implemented for the x86 targets running
10398 When code compiled with @option{-fsplit-stack} calls code compiled
10399 without @option{-fsplit-stack}, there may not be much stack space
10400 available for the latter code to run. If compiling all code,
10401 including library code, with @option{-fsplit-stack} is not an option,
10402 then the linker can fix up these calls so that the code compiled
10403 without @option{-fsplit-stack} always has a large stack. Support for
10404 this is implemented in the gold linker in GNU binutils release 2.21
10407 @item -fvtable-verify=@r{[}std@r{|}preinit@r{|}none@r{]}
10408 @opindex fvtable-verify
10409 This option is only available when compiling C++ code.
10410 It turns on (or off, if using @option{-fvtable-verify=none}) the security
10411 feature that verifies at run time, for every virtual call, that
10412 the vtable pointer through which the call is made is valid for the type of
10413 the object, and has not been corrupted or overwritten. If an invalid vtable
10414 pointer is detected at run time, an error is reported and execution of the
10415 program is immediately halted.
10417 This option causes run-time data structures to be built at program startup,
10418 which are used for verifying the vtable pointers.
10419 The options @samp{std} and @samp{preinit}
10420 control the timing of when these data structures are built. In both cases the
10421 data structures are built before execution reaches @code{main}. Using
10422 @option{-fvtable-verify=std} causes the data structures to be built after
10423 shared libraries have been loaded and initialized.
10424 @option{-fvtable-verify=preinit} causes them to be built before shared
10425 libraries have been loaded and initialized.
10427 If this option appears multiple times in the command line with different
10428 values specified, @samp{none} takes highest priority over both @samp{std} and
10429 @samp{preinit}; @samp{preinit} takes priority over @samp{std}.
10432 @opindex fvtv-debug
10433 When used in conjunction with @option{-fvtable-verify=std} or
10434 @option{-fvtable-verify=preinit}, causes debug versions of the
10435 runtime functions for the vtable verification feature to be called.
10436 This flag also causes the compiler to log information about which
10437 vtable pointers it finds for each class.
10438 This information is written to a file named @file{vtv_set_ptr_data.log}
10439 in the directory named by the environment variable @env{VTV_LOGS_DIR}
10440 if that is defined or the current working directory otherwise.
10442 Note: This feature @emph{appends} data to the log file. If you want a fresh log
10443 file, be sure to delete any existing one.
10446 @opindex fvtv-counts
10447 This is a debugging flag. When used in conjunction with
10448 @option{-fvtable-verify=std} or @option{-fvtable-verify=preinit}, this
10449 causes the compiler to keep track of the total number of virtual calls
10450 it encounters and the number of verifications it inserts. It also
10451 counts the number of calls to certain run-time library functions
10452 that it inserts and logs this information for each compilation unit.
10453 The compiler writes this information to a file named
10454 @file{vtv_count_data.log} in the directory named by the environment
10455 variable @env{VTV_LOGS_DIR} if that is defined or the current working
10456 directory otherwise. It also counts the size of the vtable pointer sets
10457 for each class, and writes this information to @file{vtv_class_set_sizes.log}
10458 in the same directory.
10460 Note: This feature @emph{appends} data to the log files. To get fresh log
10461 files, be sure to delete any existing ones.
10463 @item -finstrument-functions
10464 @opindex finstrument-functions
10465 Generate instrumentation calls for entry and exit to functions. Just
10466 after function entry and just before function exit, the following
10467 profiling functions are called with the address of the current
10468 function and its call site. (On some platforms,
10469 @code{__builtin_return_address} does not work beyond the current
10470 function, so the call site information may not be available to the
10471 profiling functions otherwise.)
10474 void __cyg_profile_func_enter (void *this_fn,
10476 void __cyg_profile_func_exit (void *this_fn,
10480 The first argument is the address of the start of the current function,
10481 which may be looked up exactly in the symbol table.
10483 This instrumentation is also done for functions expanded inline in other
10484 functions. The profiling calls indicate where, conceptually, the
10485 inline function is entered and exited. This means that addressable
10486 versions of such functions must be available. If all your uses of a
10487 function are expanded inline, this may mean an additional expansion of
10488 code size. If you use @code{extern inline} in your C code, an
10489 addressable version of such functions must be provided. (This is
10490 normally the case anyway, but if you get lucky and the optimizer always
10491 expands the functions inline, you might have gotten away without
10492 providing static copies.)
10494 A function may be given the attribute @code{no_instrument_function}, in
10495 which case this instrumentation is not done. This can be used, for
10496 example, for the profiling functions listed above, high-priority
10497 interrupt routines, and any functions from which the profiling functions
10498 cannot safely be called (perhaps signal handlers, if the profiling
10499 routines generate output or allocate memory).
10501 @item -finstrument-functions-exclude-file-list=@var{file},@var{file},@dots{}
10502 @opindex finstrument-functions-exclude-file-list
10504 Set the list of functions that are excluded from instrumentation (see
10505 the description of @option{-finstrument-functions}). If the file that
10506 contains a function definition matches with one of @var{file}, then
10507 that function is not instrumented. The match is done on substrings:
10508 if the @var{file} parameter is a substring of the file name, it is
10509 considered to be a match.
10514 -finstrument-functions-exclude-file-list=/bits/stl,include/sys
10518 excludes any inline function defined in files whose pathnames
10519 contain @file{/bits/stl} or @file{include/sys}.
10521 If, for some reason, you want to include letter @samp{,} in one of
10522 @var{sym}, write @samp{\,}. For example,
10523 @option{-finstrument-functions-exclude-file-list='\,\,tmp'}
10524 (note the single quote surrounding the option).
10526 @item -finstrument-functions-exclude-function-list=@var{sym},@var{sym},@dots{}
10527 @opindex finstrument-functions-exclude-function-list
10529 This is similar to @option{-finstrument-functions-exclude-file-list},
10530 but this option sets the list of function names to be excluded from
10531 instrumentation. The function name to be matched is its user-visible
10532 name, such as @code{vector<int> blah(const vector<int> &)}, not the
10533 internal mangled name (e.g., @code{_Z4blahRSt6vectorIiSaIiEE}). The
10534 match is done on substrings: if the @var{sym} parameter is a substring
10535 of the function name, it is considered to be a match. For C99 and C++
10536 extended identifiers, the function name must be given in UTF-8, not
10537 using universal character names.
10542 @node Preprocessor Options
10543 @section Options Controlling the Preprocessor
10544 @cindex preprocessor options
10545 @cindex options, preprocessor
10547 These options control the C preprocessor, which is run on each C source
10548 file before actual compilation.
10550 If you use the @option{-E} option, nothing is done except preprocessing.
10551 Some of these options make sense only together with @option{-E} because
10552 they cause the preprocessor output to be unsuitable for actual
10556 @item -Wp,@var{option}
10558 You can use @option{-Wp,@var{option}} to bypass the compiler driver
10559 and pass @var{option} directly through to the preprocessor. If
10560 @var{option} contains commas, it is split into multiple options at the
10561 commas. However, many options are modified, translated or interpreted
10562 by the compiler driver before being passed to the preprocessor, and
10563 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
10564 interface is undocumented and subject to change, so whenever possible
10565 you should avoid using @option{-Wp} and let the driver handle the
10568 @item -Xpreprocessor @var{option}
10569 @opindex Xpreprocessor
10570 Pass @var{option} as an option to the preprocessor. You can use this to
10571 supply system-specific preprocessor options that GCC does not
10574 If you want to pass an option that takes an argument, you must use
10575 @option{-Xpreprocessor} twice, once for the option and once for the argument.
10577 @item -no-integrated-cpp
10578 @opindex no-integrated-cpp
10579 Perform preprocessing as a separate pass before compilation.
10580 By default, GCC performs preprocessing as an integrated part of
10581 input tokenization and parsing.
10582 If this option is provided, the appropriate language front end
10583 (@command{cc1}, @command{cc1plus}, or @command{cc1obj} for C, C++,
10584 and Objective-C, respectively) is instead invoked twice,
10585 once for preprocessing only and once for actual compilation
10586 of the preprocessed input.
10587 This option may be useful in conjunction with the @option{-B} or
10588 @option{-wrapper} options to specify an alternate preprocessor or
10589 perform additional processing of the program source between
10590 normal preprocessing and compilation.
10593 @include cppopts.texi
10595 @node Assembler Options
10596 @section Passing Options to the Assembler
10598 @c prevent bad page break with this line
10599 You can pass options to the assembler.
10602 @item -Wa,@var{option}
10604 Pass @var{option} as an option to the assembler. If @var{option}
10605 contains commas, it is split into multiple options at the commas.
10607 @item -Xassembler @var{option}
10608 @opindex Xassembler
10609 Pass @var{option} as an option to the assembler. You can use this to
10610 supply system-specific assembler options that GCC does not
10613 If you want to pass an option that takes an argument, you must use
10614 @option{-Xassembler} twice, once for the option and once for the argument.
10619 @section Options for Linking
10620 @cindex link options
10621 @cindex options, linking
10623 These options come into play when the compiler links object files into
10624 an executable output file. They are meaningless if the compiler is
10625 not doing a link step.
10629 @item @var{object-file-name}
10630 A file name that does not end in a special recognized suffix is
10631 considered to name an object file or library. (Object files are
10632 distinguished from libraries by the linker according to the file
10633 contents.) If linking is done, these object files are used as input
10642 If any of these options is used, then the linker is not run, and
10643 object file names should not be used as arguments. @xref{Overall
10647 @opindex fuse-ld=bfd
10648 Use the @command{bfd} linker instead of the default linker.
10650 @item -fuse-ld=gold
10651 @opindex fuse-ld=gold
10652 Use the @command{gold} linker instead of the default linker.
10655 @item -l@var{library}
10656 @itemx -l @var{library}
10658 Search the library named @var{library} when linking. (The second
10659 alternative with the library as a separate argument is only for
10660 POSIX compliance and is not recommended.)
10662 It makes a difference where in the command you write this option; the
10663 linker searches and processes libraries and object files in the order they
10664 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
10665 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
10666 to functions in @samp{z}, those functions may not be loaded.
10668 The linker searches a standard list of directories for the library,
10669 which is actually a file named @file{lib@var{library}.a}. The linker
10670 then uses this file as if it had been specified precisely by name.
10672 The directories searched include several standard system directories
10673 plus any that you specify with @option{-L}.
10675 Normally the files found this way are library files---archive files
10676 whose members are object files. The linker handles an archive file by
10677 scanning through it for members which define symbols that have so far
10678 been referenced but not defined. But if the file that is found is an
10679 ordinary object file, it is linked in the usual fashion. The only
10680 difference between using an @option{-l} option and specifying a file name
10681 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
10682 and searches several directories.
10686 You need this special case of the @option{-l} option in order to
10687 link an Objective-C or Objective-C++ program.
10689 @item -nostartfiles
10690 @opindex nostartfiles
10691 Do not use the standard system startup files when linking.
10692 The standard system libraries are used normally, unless @option{-nostdlib}
10693 or @option{-nodefaultlibs} is used.
10695 @item -nodefaultlibs
10696 @opindex nodefaultlibs
10697 Do not use the standard system libraries when linking.
10698 Only the libraries you specify are passed to the linker, and options
10699 specifying linkage of the system libraries, such as @option{-static-libgcc}
10700 or @option{-shared-libgcc}, are ignored.
10701 The standard startup files are used normally, unless @option{-nostartfiles}
10704 The compiler may generate calls to @code{memcmp},
10705 @code{memset}, @code{memcpy} and @code{memmove}.
10706 These entries are usually resolved by entries in
10707 libc. These entry points should be supplied through some other
10708 mechanism when this option is specified.
10712 Do not use the standard system startup files or libraries when linking.
10713 No startup files and only the libraries you specify are passed to
10714 the linker, and options specifying linkage of the system libraries, such as
10715 @option{-static-libgcc} or @option{-shared-libgcc}, are ignored.
10717 The compiler may generate calls to @code{memcmp}, @code{memset},
10718 @code{memcpy} and @code{memmove}.
10719 These entries are usually resolved by entries in
10720 libc. These entry points should be supplied through some other
10721 mechanism when this option is specified.
10723 @cindex @option{-lgcc}, use with @option{-nostdlib}
10724 @cindex @option{-nostdlib} and unresolved references
10725 @cindex unresolved references and @option{-nostdlib}
10726 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
10727 @cindex @option{-nodefaultlibs} and unresolved references
10728 @cindex unresolved references and @option{-nodefaultlibs}
10729 One of the standard libraries bypassed by @option{-nostdlib} and
10730 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
10731 which GCC uses to overcome shortcomings of particular machines, or special
10732 needs for some languages.
10733 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
10734 Collection (GCC) Internals},
10735 for more discussion of @file{libgcc.a}.)
10736 In most cases, you need @file{libgcc.a} even when you want to avoid
10737 other standard libraries. In other words, when you specify @option{-nostdlib}
10738 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
10739 This ensures that you have no unresolved references to internal GCC
10740 library subroutines.
10741 (An example of such an internal subroutine is @code{__main}, used to ensure C++
10742 constructors are called; @pxref{Collect2,,@code{collect2}, gccint,
10743 GNU Compiler Collection (GCC) Internals}.)
10747 Produce a position independent executable on targets that support it.
10748 For predictable results, you must also specify the same set of options
10749 used for compilation (@option{-fpie}, @option{-fPIE},
10750 or model suboptions) when you specify this linker option.
10754 Don't produce a position independent executable.
10758 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
10759 that support it. This instructs the linker to add all symbols, not
10760 only used ones, to the dynamic symbol table. This option is needed
10761 for some uses of @code{dlopen} or to allow obtaining backtraces
10762 from within a program.
10766 Remove all symbol table and relocation information from the executable.
10770 On systems that support dynamic linking, this prevents linking with the shared
10771 libraries. On other systems, this option has no effect.
10775 Produce a shared object which can then be linked with other objects to
10776 form an executable. Not all systems support this option. For predictable
10777 results, you must also specify the same set of options used for compilation
10778 (@option{-fpic}, @option{-fPIC}, or model suboptions) when
10779 you specify this linker option.@footnote{On some systems, @samp{gcc -shared}
10780 needs to build supplementary stub code for constructors to work. On
10781 multi-libbed systems, @samp{gcc -shared} must select the correct support
10782 libraries to link against. Failing to supply the correct flags may lead
10783 to subtle defects. Supplying them in cases where they are not necessary
10786 @item -shared-libgcc
10787 @itemx -static-libgcc
10788 @opindex shared-libgcc
10789 @opindex static-libgcc
10790 On systems that provide @file{libgcc} as a shared library, these options
10791 force the use of either the shared or static version, respectively.
10792 If no shared version of @file{libgcc} was built when the compiler was
10793 configured, these options have no effect.
10795 There are several situations in which an application should use the
10796 shared @file{libgcc} instead of the static version. The most common
10797 of these is when the application wishes to throw and catch exceptions
10798 across different shared libraries. In that case, each of the libraries
10799 as well as the application itself should use the shared @file{libgcc}.
10801 Therefore, the G++ and GCJ drivers automatically add
10802 @option{-shared-libgcc} whenever you build a shared library or a main
10803 executable, because C++ and Java programs typically use exceptions, so
10804 this is the right thing to do.
10806 If, instead, you use the GCC driver to create shared libraries, you may
10807 find that they are not always linked with the shared @file{libgcc}.
10808 If GCC finds, at its configuration time, that you have a non-GNU linker
10809 or a GNU linker that does not support option @option{--eh-frame-hdr},
10810 it links the shared version of @file{libgcc} into shared libraries
10811 by default. Otherwise, it takes advantage of the linker and optimizes
10812 away the linking with the shared version of @file{libgcc}, linking with
10813 the static version of libgcc by default. This allows exceptions to
10814 propagate through such shared libraries, without incurring relocation
10815 costs at library load time.
10817 However, if a library or main executable is supposed to throw or catch
10818 exceptions, you must link it using the G++ or GCJ driver, as appropriate
10819 for the languages used in the program, or using the option
10820 @option{-shared-libgcc}, such that it is linked with the shared
10823 @item -static-libasan
10824 @opindex static-libasan
10825 When the @option{-fsanitize=address} option is used to link a program,
10826 the GCC driver automatically links against @option{libasan}. If
10827 @file{libasan} is available as a shared library, and the @option{-static}
10828 option is not used, then this links against the shared version of
10829 @file{libasan}. The @option{-static-libasan} option directs the GCC
10830 driver to link @file{libasan} statically, without necessarily linking
10831 other libraries statically.
10833 @item -static-libtsan
10834 @opindex static-libtsan
10835 When the @option{-fsanitize=thread} option is used to link a program,
10836 the GCC driver automatically links against @option{libtsan}. If
10837 @file{libtsan} is available as a shared library, and the @option{-static}
10838 option is not used, then this links against the shared version of
10839 @file{libtsan}. The @option{-static-libtsan} option directs the GCC
10840 driver to link @file{libtsan} statically, without necessarily linking
10841 other libraries statically.
10843 @item -static-liblsan
10844 @opindex static-liblsan
10845 When the @option{-fsanitize=leak} option is used to link a program,
10846 the GCC driver automatically links against @option{liblsan}. If
10847 @file{liblsan} is available as a shared library, and the @option{-static}
10848 option is not used, then this links against the shared version of
10849 @file{liblsan}. The @option{-static-liblsan} option directs the GCC
10850 driver to link @file{liblsan} statically, without necessarily linking
10851 other libraries statically.
10853 @item -static-libubsan
10854 @opindex static-libubsan
10855 When the @option{-fsanitize=undefined} option is used to link a program,
10856 the GCC driver automatically links against @option{libubsan}. If
10857 @file{libubsan} is available as a shared library, and the @option{-static}
10858 option is not used, then this links against the shared version of
10859 @file{libubsan}. The @option{-static-libubsan} option directs the GCC
10860 driver to link @file{libubsan} statically, without necessarily linking
10861 other libraries statically.
10863 @item -static-libmpx
10864 @opindex static-libmpx
10865 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are
10866 used to link a program, the GCC driver automatically links against
10867 @file{libmpx}. If @file{libmpx} is available as a shared library,
10868 and the @option{-static} option is not used, then this links against
10869 the shared version of @file{libmpx}. The @option{-static-libmpx}
10870 option directs the GCC driver to link @file{libmpx} statically,
10871 without necessarily linking other libraries statically.
10873 @item -static-libmpxwrappers
10874 @opindex static-libmpxwrappers
10875 When the @option{-fcheck-pointer bounds} and @option{-mmpx} options are used
10876 to link a program without also using @option{-fno-chkp-use-wrappers}, the
10877 GCC driver automatically links against @file{libmpxwrappers}. If
10878 @file{libmpxwrappers} is available as a shared library, and the
10879 @option{-static} option is not used, then this links against the shared
10880 version of @file{libmpxwrappers}. The @option{-static-libmpxwrappers}
10881 option directs the GCC driver to link @file{libmpxwrappers} statically,
10882 without necessarily linking other libraries statically.
10884 @item -static-libstdc++
10885 @opindex static-libstdc++
10886 When the @command{g++} program is used to link a C++ program, it
10887 normally automatically links against @option{libstdc++}. If
10888 @file{libstdc++} is available as a shared library, and the
10889 @option{-static} option is not used, then this links against the
10890 shared version of @file{libstdc++}. That is normally fine. However, it
10891 is sometimes useful to freeze the version of @file{libstdc++} used by
10892 the program without going all the way to a fully static link. The
10893 @option{-static-libstdc++} option directs the @command{g++} driver to
10894 link @file{libstdc++} statically, without necessarily linking other
10895 libraries statically.
10899 Bind references to global symbols when building a shared object. Warn
10900 about any unresolved references (unless overridden by the link editor
10901 option @option{-Xlinker -z -Xlinker defs}). Only a few systems support
10904 @item -T @var{script}
10906 @cindex linker script
10907 Use @var{script} as the linker script. This option is supported by most
10908 systems using the GNU linker. On some targets, such as bare-board
10909 targets without an operating system, the @option{-T} option may be required
10910 when linking to avoid references to undefined symbols.
10912 @item -Xlinker @var{option}
10914 Pass @var{option} as an option to the linker. You can use this to
10915 supply system-specific linker options that GCC does not recognize.
10917 If you want to pass an option that takes a separate argument, you must use
10918 @option{-Xlinker} twice, once for the option and once for the argument.
10919 For example, to pass @option{-assert definitions}, you must write
10920 @option{-Xlinker -assert -Xlinker definitions}. It does not work to write
10921 @option{-Xlinker "-assert definitions"}, because this passes the entire
10922 string as a single argument, which is not what the linker expects.
10924 When using the GNU linker, it is usually more convenient to pass
10925 arguments to linker options using the @option{@var{option}=@var{value}}
10926 syntax than as separate arguments. For example, you can specify
10927 @option{-Xlinker -Map=output.map} rather than
10928 @option{-Xlinker -Map -Xlinker output.map}. Other linkers may not support
10929 this syntax for command-line options.
10931 @item -Wl,@var{option}
10933 Pass @var{option} as an option to the linker. If @var{option} contains
10934 commas, it is split into multiple options at the commas. You can use this
10935 syntax to pass an argument to the option.
10936 For example, @option{-Wl,-Map,output.map} passes @option{-Map output.map} to the
10937 linker. When using the GNU linker, you can also get the same effect with
10938 @option{-Wl,-Map=output.map}.
10940 @item -u @var{symbol}
10942 Pretend the symbol @var{symbol} is undefined, to force linking of
10943 library modules to define it. You can use @option{-u} multiple times with
10944 different symbols to force loading of additional library modules.
10946 @item -z @var{keyword}
10948 @option{-z} is passed directly on to the linker along with the keyword
10949 @var{keyword}. See the section in the documentation of your linker for
10950 permitted values and their meanings.
10953 @node Directory Options
10954 @section Options for Directory Search
10955 @cindex directory options
10956 @cindex options, directory search
10957 @cindex search path
10959 These options specify directories to search for header files, for
10960 libraries and for parts of the compiler:
10965 Add the directory @var{dir} to the head of the list of directories to be
10966 searched for header files. This can be used to override a system header
10967 file, substituting your own version, since these directories are
10968 searched before the system header file directories. However, you should
10969 not use this option to add directories that contain vendor-supplied
10970 system header files (use @option{-isystem} for that). If you use more than
10971 one @option{-I} option, the directories are scanned in left-to-right
10972 order; the standard system directories come after.
10974 If a standard system include directory, or a directory specified with
10975 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
10976 option is ignored. The directory is still searched but as a
10977 system directory at its normal position in the system include chain.
10978 This is to ensure that GCC's procedure to fix buggy system headers and
10979 the ordering for the @code{include_next} directive are not inadvertently changed.
10980 If you really need to change the search order for system directories,
10981 use the @option{-nostdinc} and/or @option{-isystem} options.
10983 @item -iplugindir=@var{dir}
10984 @opindex iplugindir=
10985 Set the directory to search for plugins that are passed
10986 by @option{-fplugin=@var{name}} instead of
10987 @option{-fplugin=@var{path}/@var{name}.so}. This option is not meant
10988 to be used by the user, but only passed by the driver.
10990 @item -iquote@var{dir}
10992 Add the directory @var{dir} to the head of the list of directories to
10993 be searched for header files only for the case of @code{#include
10994 "@var{file}"}; they are not searched for @code{#include <@var{file}>},
10995 otherwise just like @option{-I}.
10999 Add directory @var{dir} to the list of directories to be searched
11002 @item -B@var{prefix}
11004 This option specifies where to find the executables, libraries,
11005 include files, and data files of the compiler itself.
11007 The compiler driver program runs one or more of the subprograms
11008 @command{cpp}, @command{cc1}, @command{as} and @command{ld}. It tries
11009 @var{prefix} as a prefix for each program it tries to run, both with and
11010 without @samp{@var{machine}/@var{version}/} for the corresponding target
11011 machine and compiler version.
11013 For each subprogram to be run, the compiler driver first tries the
11014 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
11015 is not specified, the driver tries two standard prefixes,
11016 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
11017 those results in a file name that is found, the unmodified program
11018 name is searched for using the directories specified in your
11019 @env{PATH} environment variable.
11021 The compiler checks to see if the path provided by @option{-B}
11022 refers to a directory, and if necessary it adds a directory
11023 separator character at the end of the path.
11025 @option{-B} prefixes that effectively specify directory names also apply
11026 to libraries in the linker, because the compiler translates these
11027 options into @option{-L} options for the linker. They also apply to
11028 include files in the preprocessor, because the compiler translates these
11029 options into @option{-isystem} options for the preprocessor. In this case,
11030 the compiler appends @samp{include} to the prefix.
11032 The runtime support file @file{libgcc.a} can also be searched for using
11033 the @option{-B} prefix, if needed. If it is not found there, the two
11034 standard prefixes above are tried, and that is all. The file is left
11035 out of the link if it is not found by those means.
11037 Another way to specify a prefix much like the @option{-B} prefix is to use
11038 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
11041 As a special kludge, if the path provided by @option{-B} is
11042 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
11043 9, then it is replaced by @file{[dir/]include}. This is to help
11044 with boot-strapping the compiler.
11046 @item -no-canonical-prefixes
11047 @opindex no-canonical-prefixes
11048 Do not expand any symbolic links, resolve references to @samp{/../}
11049 or @samp{/./}, or make the path absolute when generating a relative
11052 @item --sysroot=@var{dir}
11054 Use @var{dir} as the logical root directory for headers and libraries.
11055 For example, if the compiler normally searches for headers in
11056 @file{/usr/include} and libraries in @file{/usr/lib}, it instead
11057 searches @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
11059 If you use both this option and the @option{-isysroot} option, then
11060 the @option{--sysroot} option applies to libraries, but the
11061 @option{-isysroot} option applies to header files.
11063 The GNU linker (beginning with version 2.16) has the necessary support
11064 for this option. If your linker does not support this option, the
11065 header file aspect of @option{--sysroot} still works, but the
11066 library aspect does not.
11068 @item --no-sysroot-suffix
11069 @opindex no-sysroot-suffix
11070 For some targets, a suffix is added to the root directory specified
11071 with @option{--sysroot}, depending on the other options used, so that
11072 headers may for example be found in
11073 @file{@var{dir}/@var{suffix}/usr/include} instead of
11074 @file{@var{dir}/usr/include}. This option disables the addition of
11079 This option has been deprecated. Please use @option{-iquote} instead for
11080 @option{-I} directories before the @option{-I-} and remove the @option{-I-}
11082 Any directories you specify with @option{-I} options before the @option{-I-}
11083 option are searched only for the case of @code{#include "@var{file}"};
11084 they are not searched for @code{#include <@var{file}>}.
11086 If additional directories are specified with @option{-I} options after
11087 the @option{-I-} option, these directories are searched for all @code{#include}
11088 directives. (Ordinarily @emph{all} @option{-I} directories are used
11091 In addition, the @option{-I-} option inhibits the use of the current
11092 directory (where the current input file came from) as the first search
11093 directory for @code{#include "@var{file}"}. There is no way to
11094 override this effect of @option{-I-}. With @option{-I.} you can specify
11095 searching the directory that is current when the compiler is
11096 invoked. That is not exactly the same as what the preprocessor does
11097 by default, but it is often satisfactory.
11099 @option{-I-} does not inhibit the use of the standard system directories
11100 for header files. Thus, @option{-I-} and @option{-nostdinc} are
11104 @node Code Gen Options
11105 @section Options for Code Generation Conventions
11106 @cindex code generation conventions
11107 @cindex options, code generation
11108 @cindex run-time options
11110 These machine-independent options control the interface conventions
11111 used in code generation.
11113 Most of them have both positive and negative forms; the negative form
11114 of @option{-ffoo} is @option{-fno-foo}. In the table below, only
11115 one of the forms is listed---the one that is not the default. You
11116 can figure out the other form by either removing @samp{no-} or adding
11120 @item -fstack-reuse=@var{reuse-level}
11121 @opindex fstack_reuse
11122 This option controls stack space reuse for user declared local/auto variables
11123 and compiler generated temporaries. @var{reuse_level} can be @samp{all},
11124 @samp{named_vars}, or @samp{none}. @samp{all} enables stack reuse for all
11125 local variables and temporaries, @samp{named_vars} enables the reuse only for
11126 user defined local variables with names, and @samp{none} disables stack reuse
11127 completely. The default value is @samp{all}. The option is needed when the
11128 program extends the lifetime of a scoped local variable or a compiler generated
11129 temporary beyond the end point defined by the language. When a lifetime of
11130 a variable ends, and if the variable lives in memory, the optimizing compiler
11131 has the freedom to reuse its stack space with other temporaries or scoped
11132 local variables whose live range does not overlap with it. Legacy code extending
11133 local lifetime is likely to break with the stack reuse optimization.
11152 if (*p == 10) // out of scope use of local1
11163 A(int k) : i(k), j(k) @{ @}
11170 void foo(const A& ar)
11177 foo(A(10)); // temp object's lifetime ends when foo returns
11183 ap->i+= 10; // ap references out of scope temp whose space
11184 // is reused with a. What is the value of ap->i?
11189 The lifetime of a compiler generated temporary is well defined by the C++
11190 standard. When a lifetime of a temporary ends, and if the temporary lives
11191 in memory, the optimizing compiler has the freedom to reuse its stack
11192 space with other temporaries or scoped local variables whose live range
11193 does not overlap with it. However some of the legacy code relies on
11194 the behavior of older compilers in which temporaries' stack space is
11195 not reused, the aggressive stack reuse can lead to runtime errors. This
11196 option is used to control the temporary stack reuse optimization.
11200 This option generates traps for signed overflow on addition, subtraction,
11201 multiplication operations.
11202 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11203 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11204 @option{-fwrapv} being effective. Note that only active options override, so
11205 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11206 results in @option{-ftrapv} being effective.
11210 This option instructs the compiler to assume that signed arithmetic
11211 overflow of addition, subtraction and multiplication wraps around
11212 using twos-complement representation. This flag enables some optimizations
11213 and disables others. This option is enabled by default for the Java
11214 front end, as required by the Java language specification.
11215 The options @option{-ftrapv} and @option{-fwrapv} override each other, so using
11216 @option{-ftrapv} @option{-fwrapv} on the command-line results in
11217 @option{-fwrapv} being effective. Note that only active options override, so
11218 using @option{-ftrapv} @option{-fwrapv} @option{-fno-wrapv} on the command-line
11219 results in @option{-ftrapv} being effective.
11222 @opindex fexceptions
11223 Enable exception handling. Generates extra code needed to propagate
11224 exceptions. For some targets, this implies GCC generates frame
11225 unwind information for all functions, which can produce significant data
11226 size overhead, although it does not affect execution. If you do not
11227 specify this option, GCC enables it by default for languages like
11228 C++ that normally require exception handling, and disables it for
11229 languages like C that do not normally require it. However, you may need
11230 to enable this option when compiling C code that needs to interoperate
11231 properly with exception handlers written in C++. You may also wish to
11232 disable this option if you are compiling older C++ programs that don't
11233 use exception handling.
11235 @item -fnon-call-exceptions
11236 @opindex fnon-call-exceptions
11237 Generate code that allows trapping instructions to throw exceptions.
11238 Note that this requires platform-specific runtime support that does
11239 not exist everywhere. Moreover, it only allows @emph{trapping}
11240 instructions to throw exceptions, i.e.@: memory references or floating-point
11241 instructions. It does not allow exceptions to be thrown from
11242 arbitrary signal handlers such as @code{SIGALRM}.
11244 @item -fdelete-dead-exceptions
11245 @opindex fdelete-dead-exceptions
11246 Consider that instructions that may throw exceptions but don't otherwise
11247 contribute to the execution of the program can be optimized away.
11248 This option is enabled by default for the Ada front end, as permitted by
11249 the Ada language specification.
11250 Optimization passes that cause dead exceptions to be removed are enabled independently at different optimization levels.
11252 @item -funwind-tables
11253 @opindex funwind-tables
11254 Similar to @option{-fexceptions}, except that it just generates any needed
11255 static data, but does not affect the generated code in any other way.
11256 You normally do not need to enable this option; instead, a language processor
11257 that needs this handling enables it on your behalf.
11259 @item -fasynchronous-unwind-tables
11260 @opindex fasynchronous-unwind-tables
11261 Generate unwind table in DWARF format, if supported by target machine. The
11262 table is exact at each instruction boundary, so it can be used for stack
11263 unwinding from asynchronous events (such as debugger or garbage collector).
11265 @item -fno-gnu-unique
11266 @opindex fno-gnu-unique
11267 On systems with recent GNU assembler and C library, the C++ compiler
11268 uses the @code{STB_GNU_UNIQUE} binding to make sure that definitions
11269 of template static data members and static local variables in inline
11270 functions are unique even in the presence of @code{RTLD_LOCAL}; this
11271 is necessary to avoid problems with a library used by two different
11272 @code{RTLD_LOCAL} plugins depending on a definition in one of them and
11273 therefore disagreeing with the other one about the binding of the
11274 symbol. But this causes @code{dlclose} to be ignored for affected
11275 DSOs; if your program relies on reinitialization of a DSO via
11276 @code{dlclose} and @code{dlopen}, you can use
11277 @option{-fno-gnu-unique}.
11279 @item -fpcc-struct-return
11280 @opindex fpcc-struct-return
11281 Return ``short'' @code{struct} and @code{union} values in memory like
11282 longer ones, rather than in registers. This convention is less
11283 efficient, but it has the advantage of allowing intercallability between
11284 GCC-compiled files and files compiled with other compilers, particularly
11285 the Portable C Compiler (pcc).
11287 The precise convention for returning structures in memory depends
11288 on the target configuration macros.
11290 Short structures and unions are those whose size and alignment match
11291 that of some integer type.
11293 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
11294 switch is not binary compatible with code compiled with the
11295 @option{-freg-struct-return} switch.
11296 Use it to conform to a non-default application binary interface.
11298 @item -freg-struct-return
11299 @opindex freg-struct-return
11300 Return @code{struct} and @code{union} values in registers when possible.
11301 This is more efficient for small structures than
11302 @option{-fpcc-struct-return}.
11304 If you specify neither @option{-fpcc-struct-return} nor
11305 @option{-freg-struct-return}, GCC defaults to whichever convention is
11306 standard for the target. If there is no standard convention, GCC
11307 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
11308 the principal compiler. In those cases, we can choose the standard, and
11309 we chose the more efficient register return alternative.
11311 @strong{Warning:} code compiled with the @option{-freg-struct-return}
11312 switch is not binary compatible with code compiled with the
11313 @option{-fpcc-struct-return} switch.
11314 Use it to conform to a non-default application binary interface.
11316 @item -fshort-enums
11317 @opindex fshort-enums
11318 Allocate to an @code{enum} type only as many bytes as it needs for the
11319 declared range of possible values. Specifically, the @code{enum} type
11320 is equivalent to the smallest integer type that has enough room.
11322 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
11323 code that is not binary compatible with code generated without that switch.
11324 Use it to conform to a non-default application binary interface.
11326 @item -fshort-wchar
11327 @opindex fshort-wchar
11328 Override the underlying type for @code{wchar_t} to be @code{short
11329 unsigned int} instead of the default for the target. This option is
11330 useful for building programs to run under WINE@.
11332 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
11333 code that is not binary compatible with code generated without that switch.
11334 Use it to conform to a non-default application binary interface.
11337 @opindex fno-common
11338 In C code, controls the placement of uninitialized global variables.
11339 Unix C compilers have traditionally permitted multiple definitions of
11340 such variables in different compilation units by placing the variables
11342 This is the behavior specified by @option{-fcommon}, and is the default
11343 for GCC on most targets.
11344 On the other hand, this behavior is not required by ISO C, and on some
11345 targets may carry a speed or code size penalty on variable references.
11346 The @option{-fno-common} option specifies that the compiler should place
11347 uninitialized global variables in the data section of the object file,
11348 rather than generating them as common blocks.
11349 This has the effect that if the same variable is declared
11350 (without @code{extern}) in two different compilations,
11351 you get a multiple-definition error when you link them.
11352 In this case, you must compile with @option{-fcommon} instead.
11353 Compiling with @option{-fno-common} is useful on targets for which
11354 it provides better performance, or if you wish to verify that the
11355 program will work on other systems that always treat uninitialized
11356 variable declarations this way.
11360 Ignore the @code{#ident} directive.
11362 @item -finhibit-size-directive
11363 @opindex finhibit-size-directive
11364 Don't output a @code{.size} assembler directive, or anything else that
11365 would cause trouble if the function is split in the middle, and the
11366 two halves are placed at locations far apart in memory. This option is
11367 used when compiling @file{crtstuff.c}; you should not need to use it
11370 @item -fverbose-asm
11371 @opindex fverbose-asm
11372 Put extra commentary information in the generated assembly code to
11373 make it more readable. This option is generally only of use to those
11374 who actually need to read the generated assembly code (perhaps while
11375 debugging the compiler itself).
11377 @option{-fno-verbose-asm}, the default, causes the
11378 extra information to be omitted and is useful when comparing two assembler
11381 @item -frecord-gcc-switches
11382 @opindex frecord-gcc-switches
11383 This switch causes the command line used to invoke the
11384 compiler to be recorded into the object file that is being created.
11385 This switch is only implemented on some targets and the exact format
11386 of the recording is target and binary file format dependent, but it
11387 usually takes the form of a section containing ASCII text. This
11388 switch is related to the @option{-fverbose-asm} switch, but that
11389 switch only records information in the assembler output file as
11390 comments, so it never reaches the object file.
11391 See also @option{-grecord-gcc-switches} for another
11392 way of storing compiler options into the object file.
11396 @cindex global offset table
11398 Generate position-independent code (PIC) suitable for use in a shared
11399 library, if supported for the target machine. Such code accesses all
11400 constant addresses through a global offset table (GOT)@. The dynamic
11401 loader resolves the GOT entries when the program starts (the dynamic
11402 loader is not part of GCC; it is part of the operating system). If
11403 the GOT size for the linked executable exceeds a machine-specific
11404 maximum size, you get an error message from the linker indicating that
11405 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
11406 instead. (These maximums are 8k on the SPARC, 28k on AArch64 and 32k
11407 on the m68k and RS/6000. The x86 has no such limit.)
11409 Position-independent code requires special support, and therefore works
11410 only on certain machines. For the x86, GCC supports PIC for System V
11411 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
11412 position-independent.
11414 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11419 If supported for the target machine, emit position-independent code,
11420 suitable for dynamic linking and avoiding any limit on the size of the
11421 global offset table. This option makes a difference on AArch64, m68k,
11422 PowerPC and SPARC@.
11424 Position-independent code requires special support, and therefore works
11425 only on certain machines.
11427 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
11434 These options are similar to @option{-fpic} and @option{-fPIC}, but
11435 generated position independent code can be only linked into executables.
11436 Usually these options are used when @option{-pie} GCC option is
11437 used during linking.
11439 @option{-fpie} and @option{-fPIE} both define the macros
11440 @code{__pie__} and @code{__PIE__}. The macros have the value 1
11441 for @option{-fpie} and 2 for @option{-fPIE}.
11445 Do not use the PLT for external function calls in position-independent code.
11446 Instead, load the callee address at call sites from the GOT and branch to it.
11447 This leads to more efficient code by eliminating PLT stubs and exposing
11448 GOT loads to optimizations. On architectures such as 32-bit x86 where
11449 PLT stubs expect the GOT pointer in a specific register, this gives more
11450 register allocation freedom to the compiler.
11451 Lazy binding requires use of the PLT;
11452 with @option{-fno-plt} all external symbols are resolved at load time.
11454 Alternatively, the function attribute @code{noplt} can be used to avoid calls
11455 through the PLT for specific external functions.
11457 In position-dependent code, a few targets also convert calls to
11458 functions that are marked to not use the PLT to use the GOT instead.
11460 @item -fno-jump-tables
11461 @opindex fno-jump-tables
11462 Do not use jump tables for switch statements even where it would be
11463 more efficient than other code generation strategies. This option is
11464 of use in conjunction with @option{-fpic} or @option{-fPIC} for
11465 building code that forms part of a dynamic linker and cannot
11466 reference the address of a jump table. On some targets, jump tables
11467 do not require a GOT and this option is not needed.
11469 @item -ffixed-@var{reg}
11471 Treat the register named @var{reg} as a fixed register; generated code
11472 should never refer to it (except perhaps as a stack pointer, frame
11473 pointer or in some other fixed role).
11475 @var{reg} must be the name of a register. The register names accepted
11476 are machine-specific and are defined in the @code{REGISTER_NAMES}
11477 macro in the machine description macro file.
11479 This flag does not have a negative form, because it specifies a
11482 @item -fcall-used-@var{reg}
11483 @opindex fcall-used
11484 Treat the register named @var{reg} as an allocable register that is
11485 clobbered by function calls. It may be allocated for temporaries or
11486 variables that do not live across a call. Functions compiled this way
11487 do not save and restore the register @var{reg}.
11489 It is an error to use this flag with the frame pointer or stack pointer.
11490 Use of this flag for other registers that have fixed pervasive roles in
11491 the machine's execution model produces disastrous results.
11493 This flag does not have a negative form, because it specifies a
11496 @item -fcall-saved-@var{reg}
11497 @opindex fcall-saved
11498 Treat the register named @var{reg} as an allocable register saved by
11499 functions. It may be allocated even for temporaries or variables that
11500 live across a call. Functions compiled this way save and restore
11501 the register @var{reg} if they use it.
11503 It is an error to use this flag with the frame pointer or stack pointer.
11504 Use of this flag for other registers that have fixed pervasive roles in
11505 the machine's execution model produces disastrous results.
11507 A different sort of disaster results from the use of this flag for
11508 a register in which function values may be returned.
11510 This flag does not have a negative form, because it specifies a
11513 @item -fpack-struct[=@var{n}]
11514 @opindex fpack-struct
11515 Without a value specified, pack all structure members together without
11516 holes. When a value is specified (which must be a small power of two), pack
11517 structure members according to this value, representing the maximum
11518 alignment (that is, objects with default alignment requirements larger than
11519 this are output potentially unaligned at the next fitting location.
11521 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
11522 code that is not binary compatible with code generated without that switch.
11523 Additionally, it makes the code suboptimal.
11524 Use it to conform to a non-default application binary interface.
11526 @item -fleading-underscore
11527 @opindex fleading-underscore
11528 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
11529 change the way C symbols are represented in the object file. One use
11530 is to help link with legacy assembly code.
11532 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
11533 generate code that is not binary compatible with code generated without that
11534 switch. Use it to conform to a non-default application binary interface.
11535 Not all targets provide complete support for this switch.
11537 @item -ftls-model=@var{model}
11538 @opindex ftls-model
11539 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
11540 The @var{model} argument should be one of @samp{global-dynamic},
11541 @samp{local-dynamic}, @samp{initial-exec} or @samp{local-exec}.
11542 Note that the choice is subject to optimization: the compiler may use
11543 a more efficient model for symbols not visible outside of the translation
11544 unit, or if @option{-fpic} is not given on the command line.
11546 The default without @option{-fpic} is @samp{initial-exec}; with
11547 @option{-fpic} the default is @samp{global-dynamic}.
11549 @item -fvisibility=@r{[}default@r{|}internal@r{|}hidden@r{|}protected@r{]}
11550 @opindex fvisibility
11551 Set the default ELF image symbol visibility to the specified option---all
11552 symbols are marked with this unless overridden within the code.
11553 Using this feature can very substantially improve linking and
11554 load times of shared object libraries, produce more optimized
11555 code, provide near-perfect API export and prevent symbol clashes.
11556 It is @strong{strongly} recommended that you use this in any shared objects
11559 Despite the nomenclature, @samp{default} always means public; i.e.,
11560 available to be linked against from outside the shared object.
11561 @samp{protected} and @samp{internal} are pretty useless in real-world
11562 usage so the only other commonly used option is @samp{hidden}.
11563 The default if @option{-fvisibility} isn't specified is
11564 @samp{default}, i.e., make every symbol public.
11566 A good explanation of the benefits offered by ensuring ELF
11567 symbols have the correct visibility is given by ``How To Write
11568 Shared Libraries'' by Ulrich Drepper (which can be found at
11569 @w{@uref{http://www.akkadia.org/drepper/}})---however a superior
11570 solution made possible by this option to marking things hidden when
11571 the default is public is to make the default hidden and mark things
11572 public. This is the norm with DLLs on Windows and with @option{-fvisibility=hidden}
11573 and @code{__attribute__ ((visibility("default")))} instead of
11574 @code{__declspec(dllexport)} you get almost identical semantics with
11575 identical syntax. This is a great boon to those working with
11576 cross-platform projects.
11578 For those adding visibility support to existing code, you may find
11579 @code{#pragma GCC visibility} of use. This works by you enclosing
11580 the declarations you wish to set visibility for with (for example)
11581 @code{#pragma GCC visibility push(hidden)} and
11582 @code{#pragma GCC visibility pop}.
11583 Bear in mind that symbol visibility should be viewed @strong{as
11584 part of the API interface contract} and thus all new code should
11585 always specify visibility when it is not the default; i.e., declarations
11586 only for use within the local DSO should @strong{always} be marked explicitly
11587 as hidden as so to avoid PLT indirection overheads---making this
11588 abundantly clear also aids readability and self-documentation of the code.
11589 Note that due to ISO C++ specification requirements, @code{operator new} and
11590 @code{operator delete} must always be of default visibility.
11592 Be aware that headers from outside your project, in particular system
11593 headers and headers from any other library you use, may not be
11594 expecting to be compiled with visibility other than the default. You
11595 may need to explicitly say @code{#pragma GCC visibility push(default)}
11596 before including any such headers.
11598 @code{extern} declarations are not affected by @option{-fvisibility}, so
11599 a lot of code can be recompiled with @option{-fvisibility=hidden} with
11600 no modifications. However, this means that calls to @code{extern}
11601 functions with no explicit visibility use the PLT, so it is more
11602 effective to use @code{__attribute ((visibility))} and/or
11603 @code{#pragma GCC visibility} to tell the compiler which @code{extern}
11604 declarations should be treated as hidden.
11606 Note that @option{-fvisibility} does affect C++ vague linkage
11607 entities. This means that, for instance, an exception class that is
11608 be thrown between DSOs must be explicitly marked with default
11609 visibility so that the @samp{type_info} nodes are unified between
11612 An overview of these techniques, their benefits and how to use them
11613 is at @uref{http://gcc.gnu.org/@/wiki/@/Visibility}.
11615 @item -fstrict-volatile-bitfields
11616 @opindex fstrict-volatile-bitfields
11617 This option should be used if accesses to volatile bit-fields (or other
11618 structure fields, although the compiler usually honors those types
11619 anyway) should use a single access of the width of the
11620 field's type, aligned to a natural alignment if possible. For
11621 example, targets with memory-mapped peripheral registers might require
11622 all such accesses to be 16 bits wide; with this flag you can
11623 declare all peripheral bit-fields as @code{unsigned short} (assuming short
11624 is 16 bits on these targets) to force GCC to use 16-bit accesses
11625 instead of, perhaps, a more efficient 32-bit access.
11627 If this option is disabled, the compiler uses the most efficient
11628 instruction. In the previous example, that might be a 32-bit load
11629 instruction, even though that accesses bytes that do not contain
11630 any portion of the bit-field, or memory-mapped registers unrelated to
11631 the one being updated.
11633 In some cases, such as when the @code{packed} attribute is applied to a
11634 structure field, it may not be possible to access the field with a single
11635 read or write that is correctly aligned for the target machine. In this
11636 case GCC falls back to generating multiple accesses rather than code that
11637 will fault or truncate the result at run time.
11639 Note: Due to restrictions of the C/C++11 memory model, write accesses are
11640 not allowed to touch non bit-field members. It is therefore recommended
11641 to define all bits of the field's type as bit-field members.
11643 The default value of this option is determined by the application binary
11644 interface for the target processor.
11646 @item -fsync-libcalls
11647 @opindex fsync-libcalls
11648 This option controls whether any out-of-line instance of the @code{__sync}
11649 family of functions may be used to implement the C++11 @code{__atomic}
11650 family of functions.
11652 The default value of this option is enabled, thus the only useful form
11653 of the option is @option{-fno-sync-libcalls}. This option is used in
11654 the implementation of the @file{libatomic} runtime library.
11658 @node Developer Options
11659 @section GCC Developer Options
11660 @cindex developer options
11661 @cindex debugging GCC
11662 @cindex debug dump options
11663 @cindex dump options
11664 @cindex compilation statistics
11666 This section describes command-line options that are primarily of
11667 interest to GCC developers, including options to support compiler
11668 testing and investigation of compiler bugs and compile-time
11669 performance problems. This includes options that produce debug dumps
11670 at various points in the compilation; that print statistics such as
11671 memory use and execution time; and that print information about GCC's
11672 configuration, such as where it searches for libraries. You should
11673 rarely need to use any of these options for ordinary compilation and
11678 @item -d@var{letters}
11679 @itemx -fdump-rtl-@var{pass}
11680 @itemx -fdump-rtl-@var{pass}=@var{filename}
11682 @opindex fdump-rtl-@var{pass}
11683 Says to make debugging dumps during compilation at times specified by
11684 @var{letters}. This is used for debugging the RTL-based passes of the
11685 compiler. The file names for most of the dumps are made by appending
11686 a pass number and a word to the @var{dumpname}, and the files are
11687 created in the directory of the output file. In case of
11688 @option{=@var{filename}} option, the dump is output on the given file
11689 instead of the pass numbered dump files. Note that the pass number is
11690 assigned as passes are registered into the pass manager. Most passes
11691 are registered in the order that they will execute and for these passes
11692 the number corresponds to the pass execution order. However, passes
11693 registered by plugins, passes specific to compilation targets, or
11694 passes that are otherwise registered after all the other passes are
11695 numbered higher than a pass named "final", even if they are executed
11696 earlier. @var{dumpname} is generated from the name of the output
11697 file if explicitly specified and not an executable, otherwise it is
11698 the basename of the source file. These switches may have different
11699 effects when @option{-E} is used for preprocessing.
11701 Debug dumps can be enabled with a @option{-fdump-rtl} switch or some
11702 @option{-d} option @var{letters}. Here are the possible
11703 letters for use in @var{pass} and @var{letters}, and their meanings:
11707 @item -fdump-rtl-alignments
11708 @opindex fdump-rtl-alignments
11709 Dump after branch alignments have been computed.
11711 @item -fdump-rtl-asmcons
11712 @opindex fdump-rtl-asmcons
11713 Dump after fixing rtl statements that have unsatisfied in/out constraints.
11715 @item -fdump-rtl-auto_inc_dec
11716 @opindex fdump-rtl-auto_inc_dec
11717 Dump after auto-inc-dec discovery. This pass is only run on
11718 architectures that have auto inc or auto dec instructions.
11720 @item -fdump-rtl-barriers
11721 @opindex fdump-rtl-barriers
11722 Dump after cleaning up the barrier instructions.
11724 @item -fdump-rtl-bbpart
11725 @opindex fdump-rtl-bbpart
11726 Dump after partitioning hot and cold basic blocks.
11728 @item -fdump-rtl-bbro
11729 @opindex fdump-rtl-bbro
11730 Dump after block reordering.
11732 @item -fdump-rtl-btl1
11733 @itemx -fdump-rtl-btl2
11734 @opindex fdump-rtl-btl2
11735 @opindex fdump-rtl-btl2
11736 @option{-fdump-rtl-btl1} and @option{-fdump-rtl-btl2} enable dumping
11737 after the two branch
11738 target load optimization passes.
11740 @item -fdump-rtl-bypass
11741 @opindex fdump-rtl-bypass
11742 Dump after jump bypassing and control flow optimizations.
11744 @item -fdump-rtl-combine
11745 @opindex fdump-rtl-combine
11746 Dump after the RTL instruction combination pass.
11748 @item -fdump-rtl-compgotos
11749 @opindex fdump-rtl-compgotos
11750 Dump after duplicating the computed gotos.
11752 @item -fdump-rtl-ce1
11753 @itemx -fdump-rtl-ce2
11754 @itemx -fdump-rtl-ce3
11755 @opindex fdump-rtl-ce1
11756 @opindex fdump-rtl-ce2
11757 @opindex fdump-rtl-ce3
11758 @option{-fdump-rtl-ce1}, @option{-fdump-rtl-ce2}, and
11759 @option{-fdump-rtl-ce3} enable dumping after the three
11760 if conversion passes.
11762 @item -fdump-rtl-cprop_hardreg
11763 @opindex fdump-rtl-cprop_hardreg
11764 Dump after hard register copy propagation.
11766 @item -fdump-rtl-csa
11767 @opindex fdump-rtl-csa
11768 Dump after combining stack adjustments.
11770 @item -fdump-rtl-cse1
11771 @itemx -fdump-rtl-cse2
11772 @opindex fdump-rtl-cse1
11773 @opindex fdump-rtl-cse2
11774 @option{-fdump-rtl-cse1} and @option{-fdump-rtl-cse2} enable dumping after
11775 the two common subexpression elimination passes.
11777 @item -fdump-rtl-dce
11778 @opindex fdump-rtl-dce
11779 Dump after the standalone dead code elimination passes.
11781 @item -fdump-rtl-dbr
11782 @opindex fdump-rtl-dbr
11783 Dump after delayed branch scheduling.
11785 @item -fdump-rtl-dce1
11786 @itemx -fdump-rtl-dce2
11787 @opindex fdump-rtl-dce1
11788 @opindex fdump-rtl-dce2
11789 @option{-fdump-rtl-dce1} and @option{-fdump-rtl-dce2} enable dumping after
11790 the two dead store elimination passes.
11792 @item -fdump-rtl-eh
11793 @opindex fdump-rtl-eh
11794 Dump after finalization of EH handling code.
11796 @item -fdump-rtl-eh_ranges
11797 @opindex fdump-rtl-eh_ranges
11798 Dump after conversion of EH handling range regions.
11800 @item -fdump-rtl-expand
11801 @opindex fdump-rtl-expand
11802 Dump after RTL generation.
11804 @item -fdump-rtl-fwprop1
11805 @itemx -fdump-rtl-fwprop2
11806 @opindex fdump-rtl-fwprop1
11807 @opindex fdump-rtl-fwprop2
11808 @option{-fdump-rtl-fwprop1} and @option{-fdump-rtl-fwprop2} enable
11809 dumping after the two forward propagation passes.
11811 @item -fdump-rtl-gcse1
11812 @itemx -fdump-rtl-gcse2
11813 @opindex fdump-rtl-gcse1
11814 @opindex fdump-rtl-gcse2
11815 @option{-fdump-rtl-gcse1} and @option{-fdump-rtl-gcse2} enable dumping
11816 after global common subexpression elimination.
11818 @item -fdump-rtl-init-regs
11819 @opindex fdump-rtl-init-regs
11820 Dump after the initialization of the registers.
11822 @item -fdump-rtl-initvals
11823 @opindex fdump-rtl-initvals
11824 Dump after the computation of the initial value sets.
11826 @item -fdump-rtl-into_cfglayout
11827 @opindex fdump-rtl-into_cfglayout
11828 Dump after converting to cfglayout mode.
11830 @item -fdump-rtl-ira
11831 @opindex fdump-rtl-ira
11832 Dump after iterated register allocation.
11834 @item -fdump-rtl-jump
11835 @opindex fdump-rtl-jump
11836 Dump after the second jump optimization.
11838 @item -fdump-rtl-loop2
11839 @opindex fdump-rtl-loop2
11840 @option{-fdump-rtl-loop2} enables dumping after the rtl
11841 loop optimization passes.
11843 @item -fdump-rtl-mach
11844 @opindex fdump-rtl-mach
11845 Dump after performing the machine dependent reorganization pass, if that
11848 @item -fdump-rtl-mode_sw
11849 @opindex fdump-rtl-mode_sw
11850 Dump after removing redundant mode switches.
11852 @item -fdump-rtl-rnreg
11853 @opindex fdump-rtl-rnreg
11854 Dump after register renumbering.
11856 @item -fdump-rtl-outof_cfglayout
11857 @opindex fdump-rtl-outof_cfglayout
11858 Dump after converting from cfglayout mode.
11860 @item -fdump-rtl-peephole2
11861 @opindex fdump-rtl-peephole2
11862 Dump after the peephole pass.
11864 @item -fdump-rtl-postreload
11865 @opindex fdump-rtl-postreload
11866 Dump after post-reload optimizations.
11868 @item -fdump-rtl-pro_and_epilogue
11869 @opindex fdump-rtl-pro_and_epilogue
11870 Dump after generating the function prologues and epilogues.
11872 @item -fdump-rtl-sched1
11873 @itemx -fdump-rtl-sched2
11874 @opindex fdump-rtl-sched1
11875 @opindex fdump-rtl-sched2
11876 @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2} enable dumping
11877 after the basic block scheduling passes.
11879 @item -fdump-rtl-ree
11880 @opindex fdump-rtl-ree
11881 Dump after sign/zero extension elimination.
11883 @item -fdump-rtl-seqabstr
11884 @opindex fdump-rtl-seqabstr
11885 Dump after common sequence discovery.
11887 @item -fdump-rtl-shorten
11888 @opindex fdump-rtl-shorten
11889 Dump after shortening branches.
11891 @item -fdump-rtl-sibling
11892 @opindex fdump-rtl-sibling
11893 Dump after sibling call optimizations.
11895 @item -fdump-rtl-split1
11896 @itemx -fdump-rtl-split2
11897 @itemx -fdump-rtl-split3
11898 @itemx -fdump-rtl-split4
11899 @itemx -fdump-rtl-split5
11900 @opindex fdump-rtl-split1
11901 @opindex fdump-rtl-split2
11902 @opindex fdump-rtl-split3
11903 @opindex fdump-rtl-split4
11904 @opindex fdump-rtl-split5
11905 These options enable dumping after five rounds of
11906 instruction splitting.
11908 @item -fdump-rtl-sms
11909 @opindex fdump-rtl-sms
11910 Dump after modulo scheduling. This pass is only run on some
11913 @item -fdump-rtl-stack
11914 @opindex fdump-rtl-stack
11915 Dump after conversion from GCC's ``flat register file'' registers to the
11916 x87's stack-like registers. This pass is only run on x86 variants.
11918 @item -fdump-rtl-subreg1
11919 @itemx -fdump-rtl-subreg2
11920 @opindex fdump-rtl-subreg1
11921 @opindex fdump-rtl-subreg2
11922 @option{-fdump-rtl-subreg1} and @option{-fdump-rtl-subreg2} enable dumping after
11923 the two subreg expansion passes.
11925 @item -fdump-rtl-unshare
11926 @opindex fdump-rtl-unshare
11927 Dump after all rtl has been unshared.
11929 @item -fdump-rtl-vartrack
11930 @opindex fdump-rtl-vartrack
11931 Dump after variable tracking.
11933 @item -fdump-rtl-vregs
11934 @opindex fdump-rtl-vregs
11935 Dump after converting virtual registers to hard registers.
11937 @item -fdump-rtl-web
11938 @opindex fdump-rtl-web
11939 Dump after live range splitting.
11941 @item -fdump-rtl-regclass
11942 @itemx -fdump-rtl-subregs_of_mode_init
11943 @itemx -fdump-rtl-subregs_of_mode_finish
11944 @itemx -fdump-rtl-dfinit
11945 @itemx -fdump-rtl-dfinish
11946 @opindex fdump-rtl-regclass
11947 @opindex fdump-rtl-subregs_of_mode_init
11948 @opindex fdump-rtl-subregs_of_mode_finish
11949 @opindex fdump-rtl-dfinit
11950 @opindex fdump-rtl-dfinish
11951 These dumps are defined but always produce empty files.
11954 @itemx -fdump-rtl-all
11956 @opindex fdump-rtl-all
11957 Produce all the dumps listed above.
11961 Annotate the assembler output with miscellaneous debugging information.
11965 Dump all macro definitions, at the end of preprocessing, in addition to
11970 Produce a core dump whenever an error occurs.
11974 Annotate the assembler output with a comment indicating which
11975 pattern and alternative is used. The length of each instruction is
11980 Dump the RTL in the assembler output as a comment before each instruction.
11981 Also turns on @option{-dp} annotation.
11985 Just generate RTL for a function instead of compiling it. Usually used
11986 with @option{-fdump-rtl-expand}.
11989 @item -fdump-noaddr
11990 @opindex fdump-noaddr
11991 When doing debugging dumps, suppress address output. This makes it more
11992 feasible to use diff on debugging dumps for compiler invocations with
11993 different compiler binaries and/or different
11994 text / bss / data / heap / stack / dso start locations.
11997 @opindex freport-bug
11998 Collect and dump debug information into a temporary file if an
11999 internal compiler error (ICE) occurs.
12001 @item -fdump-unnumbered
12002 @opindex fdump-unnumbered
12003 When doing debugging dumps, suppress instruction numbers and address output.
12004 This makes it more feasible to use diff on debugging dumps for compiler
12005 invocations with different options, in particular with and without
12008 @item -fdump-unnumbered-links
12009 @opindex fdump-unnumbered-links
12010 When doing debugging dumps (see @option{-d} option above), suppress
12011 instruction numbers for the links to the previous and next instructions
12014 @item -fdump-translation-unit @r{(C++ only)}
12015 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
12016 @opindex fdump-translation-unit
12017 Dump a representation of the tree structure for the entire translation
12018 unit to a file. The file name is made by appending @file{.tu} to the
12019 source file name, and the file is created in the same directory as the
12020 output file. If the @samp{-@var{options}} form is used, @var{options}
12021 controls the details of the dump as described for the
12022 @option{-fdump-tree} options.
12024 @item -fdump-class-hierarchy @r{(C++ only)}
12025 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
12026 @opindex fdump-class-hierarchy
12027 Dump a representation of each class's hierarchy and virtual function
12028 table layout to a file. The file name is made by appending
12029 @file{.class} to the source file name, and the file is created in the
12030 same directory as the output file. If the @samp{-@var{options}} form
12031 is used, @var{options} controls the details of the dump as described
12032 for the @option{-fdump-tree} options.
12034 @item -fdump-ipa-@var{switch}
12036 Control the dumping at various stages of inter-procedural analysis
12037 language tree to a file. The file name is generated by appending a
12038 switch specific suffix to the source file name, and the file is created
12039 in the same directory as the output file. The following dumps are
12044 Enables all inter-procedural analysis dumps.
12047 Dumps information about call-graph optimization, unused function removal,
12048 and inlining decisions.
12051 Dump after function inlining.
12055 @item -fdump-passes
12056 @opindex fdump-passes
12057 Dump the list of optimization passes that are turned on and off by
12058 the current command-line options.
12060 @item -fdump-statistics-@var{option}
12061 @opindex fdump-statistics
12062 Enable and control dumping of pass statistics in a separate file. The
12063 file name is generated by appending a suffix ending in
12064 @samp{.statistics} to the source file name, and the file is created in
12065 the same directory as the output file. If the @samp{-@var{option}}
12066 form is used, @samp{-stats} causes counters to be summed over the
12067 whole compilation unit while @samp{-details} dumps every event as
12068 the passes generate them. The default with no option is to sum
12069 counters for each function compiled.
12071 @item -fdump-tree-@var{switch}
12072 @itemx -fdump-tree-@var{switch}-@var{options}
12073 @itemx -fdump-tree-@var{switch}-@var{options}=@var{filename}
12074 @opindex fdump-tree
12075 Control the dumping at various stages of processing the intermediate
12076 language tree to a file. The file name is generated by appending a
12077 switch-specific suffix to the source file name, and the file is
12078 created in the same directory as the output file. In case of
12079 @option{=@var{filename}} option, the dump is output on the given file
12080 instead of the auto named dump files. If the @samp{-@var{options}}
12081 form is used, @var{options} is a list of @samp{-} separated options
12082 which control the details of the dump. Not all options are applicable
12083 to all dumps; those that are not meaningful are ignored. The
12084 following options are available
12088 Print the address of each node. Usually this is not meaningful as it
12089 changes according to the environment and source file. Its primary use
12090 is for tying up a dump file with a debug environment.
12092 If @code{DECL_ASSEMBLER_NAME} has been set for a given decl, use that
12093 in the dump instead of @code{DECL_NAME}. Its primary use is ease of
12094 use working backward from mangled names in the assembly file.
12096 When dumping front-end intermediate representations, inhibit dumping
12097 of members of a scope or body of a function merely because that scope
12098 has been reached. Only dump such items when they are directly reachable
12099 by some other path.
12101 When dumping pretty-printed trees, this option inhibits dumping the
12102 bodies of control structures.
12104 When dumping RTL, print the RTL in slim (condensed) form instead of
12105 the default LISP-like representation.
12107 Print a raw representation of the tree. By default, trees are
12108 pretty-printed into a C-like representation.
12110 Enable more detailed dumps (not honored by every dump option). Also
12111 include information from the optimization passes.
12113 Enable dumping various statistics about the pass (not honored by every dump
12116 Enable showing basic block boundaries (disabled in raw dumps).
12118 For each of the other indicated dump files (@option{-fdump-rtl-@var{pass}}),
12119 dump a representation of the control flow graph suitable for viewing with
12120 GraphViz to @file{@var{file}.@var{passid}.@var{pass}.dot}. Each function in
12121 the file is pretty-printed as a subgraph, so that GraphViz can render them
12122 all in a single plot.
12124 This option currently only works for RTL dumps, and the RTL is always
12125 dumped in slim form.
12127 Enable showing virtual operands for every statement.
12129 Enable showing line numbers for statements.
12131 Enable showing the unique ID (@code{DECL_UID}) for each variable.
12133 Enable showing the tree dump for each statement.
12135 Enable showing the EH region number holding each statement.
12137 Enable showing scalar evolution analysis details.
12139 Enable showing optimization information (only available in certain
12142 Enable showing missed optimization information (only available in certain
12145 Enable other detailed optimization information (only available in
12147 @item =@var{filename}
12148 Instead of an auto named dump file, output into the given file
12149 name. The file names @file{stdout} and @file{stderr} are treated
12150 specially and are considered already open standard streams. For
12154 gcc -O2 -ftree-vectorize -fdump-tree-vect-blocks=foo.dump
12155 -fdump-tree-pre=stderr file.c
12158 outputs vectorizer dump into @file{foo.dump}, while the PRE dump is
12159 output on to @file{stderr}. If two conflicting dump filenames are
12160 given for the same pass, then the latter option overrides the earlier
12164 @opindex fdump-tree-split-paths
12165 Dump each function after splitting paths to loop backedges. The file
12166 name is made by appending @file{.split-paths} to the source file name.
12169 Turn on all options, except @option{raw}, @option{slim}, @option{verbose}
12170 and @option{lineno}.
12173 Turn on all optimization options, i.e., @option{optimized},
12174 @option{missed}, and @option{note}.
12177 The following tree dumps are possible:
12181 @opindex fdump-tree-original
12182 Dump before any tree based optimization, to @file{@var{file}.original}.
12185 @opindex fdump-tree-optimized
12186 Dump after all tree based optimization, to @file{@var{file}.optimized}.
12189 @opindex fdump-tree-gimple
12190 Dump each function before and after the gimplification pass to a file. The
12191 file name is made by appending @file{.gimple} to the source file name.
12194 @opindex fdump-tree-cfg
12195 Dump the control flow graph of each function to a file. The file name is
12196 made by appending @file{.cfg} to the source file name.
12199 @opindex fdump-tree-ch
12200 Dump each function after copying loop headers. The file name is made by
12201 appending @file{.ch} to the source file name.
12204 @opindex fdump-tree-ssa
12205 Dump SSA related information to a file. The file name is made by appending
12206 @file{.ssa} to the source file name.
12209 @opindex fdump-tree-alias
12210 Dump aliasing information for each function. The file name is made by
12211 appending @file{.alias} to the source file name.
12214 @opindex fdump-tree-ccp
12215 Dump each function after CCP@. The file name is made by appending
12216 @file{.ccp} to the source file name.
12219 @opindex fdump-tree-storeccp
12220 Dump each function after STORE-CCP@. The file name is made by appending
12221 @file{.storeccp} to the source file name.
12224 @opindex fdump-tree-pre
12225 Dump trees after partial redundancy elimination. The file name is made
12226 by appending @file{.pre} to the source file name.
12229 @opindex fdump-tree-fre
12230 Dump trees after full redundancy elimination. The file name is made
12231 by appending @file{.fre} to the source file name.
12234 @opindex fdump-tree-copyprop
12235 Dump trees after copy propagation. The file name is made
12236 by appending @file{.copyprop} to the source file name.
12238 @item store_copyprop
12239 @opindex fdump-tree-store_copyprop
12240 Dump trees after store copy-propagation. The file name is made
12241 by appending @file{.store_copyprop} to the source file name.
12244 @opindex fdump-tree-dce
12245 Dump each function after dead code elimination. The file name is made by
12246 appending @file{.dce} to the source file name.
12249 @opindex fdump-tree-sra
12250 Dump each function after performing scalar replacement of aggregates. The
12251 file name is made by appending @file{.sra} to the source file name.
12254 @opindex fdump-tree-sink
12255 Dump each function after performing code sinking. The file name is made
12256 by appending @file{.sink} to the source file name.
12259 @opindex fdump-tree-dom
12260 Dump each function after applying dominator tree optimizations. The file
12261 name is made by appending @file{.dom} to the source file name.
12264 @opindex fdump-tree-dse
12265 Dump each function after applying dead store elimination. The file
12266 name is made by appending @file{.dse} to the source file name.
12269 @opindex fdump-tree-phiopt
12270 Dump each function after optimizing PHI nodes into straightline code. The file
12271 name is made by appending @file{.phiopt} to the source file name.
12274 @opindex fdump-tree-backprop
12275 Dump each function after back-propagating use information up the definition
12276 chain. The file name is made by appending @file{.backprop} to the
12280 @opindex fdump-tree-forwprop
12281 Dump each function after forward propagating single use variables. The file
12282 name is made by appending @file{.forwprop} to the source file name.
12285 @opindex fdump-tree-nrv
12286 Dump each function after applying the named return value optimization on
12287 generic trees. The file name is made by appending @file{.nrv} to the source
12291 @opindex fdump-tree-vect
12292 Dump each function after applying vectorization of loops. The file name is
12293 made by appending @file{.vect} to the source file name.
12296 @opindex fdump-tree-slp
12297 Dump each function after applying vectorization of basic blocks. The file name
12298 is made by appending @file{.slp} to the source file name.
12301 @opindex fdump-tree-vrp
12302 Dump each function after Value Range Propagation (VRP). The file name
12303 is made by appending @file{.vrp} to the source file name.
12306 @opindex fdump-tree-oaccdevlow
12307 Dump each function after applying device-specific OpenACC transformations.
12308 The file name is made by appending @file{.oaccdevlow} to the source file name.
12311 @opindex fdump-tree-all
12312 Enable all the available tree dumps with the flags provided in this option.
12316 @itemx -fopt-info-@var{options}
12317 @itemx -fopt-info-@var{options}=@var{filename}
12319 Controls optimization dumps from various optimization passes. If the
12320 @samp{-@var{options}} form is used, @var{options} is a list of
12321 @samp{-} separated option keywords to select the dump details and
12324 The @var{options} can be divided into two groups: options describing the
12325 verbosity of the dump, and options describing which optimizations
12326 should be included. The options from both the groups can be freely
12327 mixed as they are non-overlapping. However, in case of any conflicts,
12328 the later options override the earlier options on the command
12331 The following options control the dump verbosity:
12335 Print information when an optimization is successfully applied. It is
12336 up to a pass to decide which information is relevant. For example, the
12337 vectorizer passes print the source location of loops which are
12338 successfully vectorized.
12340 Print information about missed optimizations. Individual passes
12341 control which information to include in the output.
12343 Print verbose information about optimizations, such as certain
12344 transformations, more detailed messages about decisions etc.
12346 Print detailed optimization information. This includes
12347 @samp{optimized}, @samp{missed}, and @samp{note}.
12350 One or more of the following option keywords can be used to describe a
12351 group of optimizations:
12355 Enable dumps from all interprocedural optimizations.
12357 Enable dumps from all loop optimizations.
12359 Enable dumps from all inlining optimizations.
12361 Enable dumps from all vectorization optimizations.
12363 Enable dumps from all optimizations. This is a superset of
12364 the optimization groups listed above.
12367 If @var{options} is
12368 omitted, it defaults to @samp{optimized-optall}, which means to dump all
12369 info about successful optimizations from all the passes.
12371 If the @var{filename} is provided, then the dumps from all the
12372 applicable optimizations are concatenated into the @var{filename}.
12373 Otherwise the dump is output onto @file{stderr}. Though multiple
12374 @option{-fopt-info} options are accepted, only one of them can include
12375 a @var{filename}. If other filenames are provided then all but the
12376 first such option are ignored.
12378 Note that the output @var{filename} is overwritten
12379 in case of multiple translation units. If a combined output from
12380 multiple translation units is desired, @file{stderr} should be used
12383 In the following example, the optimization info is output to
12392 gcc -O3 -fopt-info-missed=missed.all
12396 outputs missed optimization report from all the passes into
12397 @file{missed.all}, and this one:
12400 gcc -O2 -ftree-vectorize -fopt-info-vec-missed
12404 prints information about missed optimization opportunities from
12405 vectorization passes on @file{stderr}.
12406 Note that @option{-fopt-info-vec-missed} is equivalent to
12407 @option{-fopt-info-missed-vec}.
12409 As another example,
12411 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt
12415 outputs information about missed optimizations as well as
12416 optimized locations from all the inlining passes into
12422 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt
12426 Here the two output filenames @file{vec.miss} and @file{loop.opt} are
12427 in conflict since only one output file is allowed. In this case, only
12428 the first option takes effect and the subsequent options are
12429 ignored. Thus only @file{vec.miss} is produced which contains
12430 dumps from the vectorizer about missed opportunities.
12432 @item -fsched-verbose=@var{n}
12433 @opindex fsched-verbose
12434 On targets that use instruction scheduling, this option controls the
12435 amount of debugging output the scheduler prints to the dump files.
12437 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
12438 same information as @option{-fdump-rtl-sched1} and @option{-fdump-rtl-sched2}.
12439 For @var{n} greater than one, it also output basic block probabilities,
12440 detailed ready list information and unit/insn info. For @var{n} greater
12441 than two, it includes RTL at abort point, control-flow and regions info.
12442 And for @var{n} over four, @option{-fsched-verbose} also includes
12447 @item -fenable-@var{kind}-@var{pass}
12448 @itemx -fdisable-@var{kind}-@var{pass}=@var{range-list}
12452 This is a set of options that are used to explicitly disable/enable
12453 optimization passes. These options are intended for use for debugging GCC.
12454 Compiler users should use regular options for enabling/disabling
12459 @item -fdisable-ipa-@var{pass}
12460 Disable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12461 statically invoked in the compiler multiple times, the pass name should be
12462 appended with a sequential number starting from 1.
12464 @item -fdisable-rtl-@var{pass}
12465 @itemx -fdisable-rtl-@var{pass}=@var{range-list}
12466 Disable RTL pass @var{pass}. @var{pass} is the pass name. If the same pass is
12467 statically invoked in the compiler multiple times, the pass name should be
12468 appended with a sequential number starting from 1. @var{range-list} is a
12469 comma-separated list of function ranges or assembler names. Each range is a number
12470 pair separated by a colon. The range is inclusive in both ends. If the range
12471 is trivial, the number pair can be simplified as a single number. If the
12472 function's call graph node's @var{uid} falls within one of the specified ranges,
12473 the @var{pass} is disabled for that function. The @var{uid} is shown in the
12474 function header of a dump file, and the pass names can be dumped by using
12475 option @option{-fdump-passes}.
12477 @item -fdisable-tree-@var{pass}
12478 @itemx -fdisable-tree-@var{pass}=@var{range-list}
12479 Disable tree pass @var{pass}. See @option{-fdisable-rtl} for the description of
12482 @item -fenable-ipa-@var{pass}
12483 Enable IPA pass @var{pass}. @var{pass} is the pass name. If the same pass is
12484 statically invoked in the compiler multiple times, the pass name should be
12485 appended with a sequential number starting from 1.
12487 @item -fenable-rtl-@var{pass}
12488 @itemx -fenable-rtl-@var{pass}=@var{range-list}
12489 Enable RTL pass @var{pass}. See @option{-fdisable-rtl} for option argument
12490 description and examples.
12492 @item -fenable-tree-@var{pass}
12493 @itemx -fenable-tree-@var{pass}=@var{range-list}
12494 Enable tree pass @var{pass}. See @option{-fdisable-rtl} for the description
12495 of option arguments.
12499 Here are some examples showing uses of these options.
12503 # disable ccp1 for all functions
12504 -fdisable-tree-ccp1
12505 # disable complete unroll for function whose cgraph node uid is 1
12506 -fenable-tree-cunroll=1
12507 # disable gcse2 for functions at the following ranges [1,1],
12508 # [300,400], and [400,1000]
12509 # disable gcse2 for functions foo and foo2
12510 -fdisable-rtl-gcse2=foo,foo2
12511 # disable early inlining
12512 -fdisable-tree-einline
12513 # disable ipa inlining
12514 -fdisable-ipa-inline
12515 # enable tree full unroll
12516 -fenable-tree-unroll
12521 @itemx -fchecking=@var{n}
12523 @opindex fno-checking
12524 Enable internal consistency checking. The default depends on
12525 the compiler configuration. @option{-fchecking=2} enables further
12526 internal consistency checking that might affect code generation.
12528 @item -frandom-seed=@var{string}
12529 @opindex frandom-seed
12530 This option provides a seed that GCC uses in place of
12531 random numbers in generating certain symbol names
12532 that have to be different in every compiled file. It is also used to
12533 place unique stamps in coverage data files and the object files that
12534 produce them. You can use the @option{-frandom-seed} option to produce
12535 reproducibly identical object files.
12537 The @var{string} can either be a number (decimal, octal or hex) or an
12538 arbitrary string (in which case it's converted to a number by
12541 The @var{string} should be different for every file you compile.
12544 @itemx -save-temps=cwd
12545 @opindex save-temps
12546 Store the usual ``temporary'' intermediate files permanently; place them
12547 in the current directory and name them based on the source file. Thus,
12548 compiling @file{foo.c} with @option{-c -save-temps} produces files
12549 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
12550 preprocessed @file{foo.i} output file even though the compiler now
12551 normally uses an integrated preprocessor.
12553 When used in combination with the @option{-x} command-line option,
12554 @option{-save-temps} is sensible enough to avoid over writing an
12555 input source file with the same extension as an intermediate file.
12556 The corresponding intermediate file may be obtained by renaming the
12557 source file before using @option{-save-temps}.
12559 If you invoke GCC in parallel, compiling several different source
12560 files that share a common base name in different subdirectories or the
12561 same source file compiled for multiple output destinations, it is
12562 likely that the different parallel compilers will interfere with each
12563 other, and overwrite the temporary files. For instance:
12566 gcc -save-temps -o outdir1/foo.o indir1/foo.c&
12567 gcc -save-temps -o outdir2/foo.o indir2/foo.c&
12570 may result in @file{foo.i} and @file{foo.o} being written to
12571 simultaneously by both compilers.
12573 @item -save-temps=obj
12574 @opindex save-temps=obj
12575 Store the usual ``temporary'' intermediate files permanently. If the
12576 @option{-o} option is used, the temporary files are based on the
12577 object file. If the @option{-o} option is not used, the
12578 @option{-save-temps=obj} switch behaves like @option{-save-temps}.
12583 gcc -save-temps=obj -c foo.c
12584 gcc -save-temps=obj -c bar.c -o dir/xbar.o
12585 gcc -save-temps=obj foobar.c -o dir2/yfoobar
12589 creates @file{foo.i}, @file{foo.s}, @file{dir/xbar.i},
12590 @file{dir/xbar.s}, @file{dir2/yfoobar.i}, @file{dir2/yfoobar.s}, and
12591 @file{dir2/yfoobar.o}.
12593 @item -time@r{[}=@var{file}@r{]}
12595 Report the CPU time taken by each subprocess in the compilation
12596 sequence. For C source files, this is the compiler proper and assembler
12597 (plus the linker if linking is done).
12599 Without the specification of an output file, the output looks like this:
12606 The first number on each line is the ``user time'', that is time spent
12607 executing the program itself. The second number is ``system time'',
12608 time spent executing operating system routines on behalf of the program.
12609 Both numbers are in seconds.
12611 With the specification of an output file, the output is appended to the
12612 named file, and it looks like this:
12615 0.12 0.01 cc1 @var{options}
12616 0.00 0.01 as @var{options}
12619 The ``user time'' and the ``system time'' are moved before the program
12620 name, and the options passed to the program are displayed, so that one
12621 can later tell what file was being compiled, and with which options.
12623 @item -fdump-final-insns@r{[}=@var{file}@r{]}
12624 @opindex fdump-final-insns
12625 Dump the final internal representation (RTL) to @var{file}. If the
12626 optional argument is omitted (or if @var{file} is @code{.}), the name
12627 of the dump file is determined by appending @code{.gkd} to the
12628 compilation output file name.
12630 @item -fcompare-debug@r{[}=@var{opts}@r{]}
12631 @opindex fcompare-debug
12632 @opindex fno-compare-debug
12633 If no error occurs during compilation, run the compiler a second time,
12634 adding @var{opts} and @option{-fcompare-debug-second} to the arguments
12635 passed to the second compilation. Dump the final internal
12636 representation in both compilations, and print an error if they differ.
12638 If the equal sign is omitted, the default @option{-gtoggle} is used.
12640 The environment variable @env{GCC_COMPARE_DEBUG}, if defined, non-empty
12641 and nonzero, implicitly enables @option{-fcompare-debug}. If
12642 @env{GCC_COMPARE_DEBUG} is defined to a string starting with a dash,
12643 then it is used for @var{opts}, otherwise the default @option{-gtoggle}
12646 @option{-fcompare-debug=}, with the equal sign but without @var{opts},
12647 is equivalent to @option{-fno-compare-debug}, which disables the dumping
12648 of the final representation and the second compilation, preventing even
12649 @env{GCC_COMPARE_DEBUG} from taking effect.
12651 To verify full coverage during @option{-fcompare-debug} testing, set
12652 @env{GCC_COMPARE_DEBUG} to say @option{-fcompare-debug-not-overridden},
12653 which GCC rejects as an invalid option in any actual compilation
12654 (rather than preprocessing, assembly or linking). To get just a
12655 warning, setting @env{GCC_COMPARE_DEBUG} to @samp{-w%n-fcompare-debug
12656 not overridden} will do.
12658 @item -fcompare-debug-second
12659 @opindex fcompare-debug-second
12660 This option is implicitly passed to the compiler for the second
12661 compilation requested by @option{-fcompare-debug}, along with options to
12662 silence warnings, and omitting other options that would cause
12663 side-effect compiler outputs to files or to the standard output. Dump
12664 files and preserved temporary files are renamed so as to contain the
12665 @code{.gk} additional extension during the second compilation, to avoid
12666 overwriting those generated by the first.
12668 When this option is passed to the compiler driver, it causes the
12669 @emph{first} compilation to be skipped, which makes it useful for little
12670 other than debugging the compiler proper.
12674 Turn off generation of debug info, if leaving out this option
12675 generates it, or turn it on at level 2 otherwise. The position of this
12676 argument in the command line does not matter; it takes effect after all
12677 other options are processed, and it does so only once, no matter how
12678 many times it is given. This is mainly intended to be used with
12679 @option{-fcompare-debug}.
12681 @item -fvar-tracking-assignments-toggle
12682 @opindex fvar-tracking-assignments-toggle
12683 @opindex fno-var-tracking-assignments-toggle
12684 Toggle @option{-fvar-tracking-assignments}, in the same way that
12685 @option{-gtoggle} toggles @option{-g}.
12689 Makes the compiler print out each function name as it is compiled, and
12690 print some statistics about each pass when it finishes.
12692 @item -ftime-report
12693 @opindex ftime-report
12694 Makes the compiler print some statistics about the time consumed by each
12695 pass when it finishes.
12697 @item -fira-verbose=@var{n}
12698 @opindex fira-verbose
12699 Control the verbosity of the dump file for the integrated register allocator.
12700 The default value is 5. If the value @var{n} is greater or equal to 10,
12701 the dump output is sent to stderr using the same format as @var{n} minus 10.
12704 @opindex flto-report
12705 Prints a report with internal details on the workings of the link-time
12706 optimizer. The contents of this report vary from version to version.
12707 It is meant to be useful to GCC developers when processing object
12708 files in LTO mode (via @option{-flto}).
12710 Disabled by default.
12712 @item -flto-report-wpa
12713 @opindex flto-report-wpa
12714 Like @option{-flto-report}, but only print for the WPA phase of Link
12718 @opindex fmem-report
12719 Makes the compiler print some statistics about permanent memory
12720 allocation when it finishes.
12722 @item -fmem-report-wpa
12723 @opindex fmem-report-wpa
12724 Makes the compiler print some statistics about permanent memory
12725 allocation for the WPA phase only.
12727 @item -fpre-ipa-mem-report
12728 @opindex fpre-ipa-mem-report
12729 @item -fpost-ipa-mem-report
12730 @opindex fpost-ipa-mem-report
12731 Makes the compiler print some statistics about permanent memory
12732 allocation before or after interprocedural optimization.
12734 @item -fprofile-report
12735 @opindex fprofile-report
12736 Makes the compiler print some statistics about consistency of the
12737 (estimated) profile and effect of individual passes.
12739 @item -fstack-usage
12740 @opindex fstack-usage
12741 Makes the compiler output stack usage information for the program, on a
12742 per-function basis. The filename for the dump is made by appending
12743 @file{.su} to the @var{auxname}. @var{auxname} is generated from the name of
12744 the output file, if explicitly specified and it is not an executable,
12745 otherwise it is the basename of the source file. An entry is made up
12750 The name of the function.
12754 One or more qualifiers: @code{static}, @code{dynamic}, @code{bounded}.
12757 The qualifier @code{static} means that the function manipulates the stack
12758 statically: a fixed number of bytes are allocated for the frame on function
12759 entry and released on function exit; no stack adjustments are otherwise made
12760 in the function. The second field is this fixed number of bytes.
12762 The qualifier @code{dynamic} means that the function manipulates the stack
12763 dynamically: in addition to the static allocation described above, stack
12764 adjustments are made in the body of the function, for example to push/pop
12765 arguments around function calls. If the qualifier @code{bounded} is also
12766 present, the amount of these adjustments is bounded at compile time and
12767 the second field is an upper bound of the total amount of stack used by
12768 the function. If it is not present, the amount of these adjustments is
12769 not bounded at compile time and the second field only represents the
12774 Emit statistics about front-end processing at the end of the compilation.
12775 This option is supported only by the C++ front end, and
12776 the information is generally only useful to the G++ development team.
12778 @item -fdbg-cnt-list
12779 @opindex fdbg-cnt-list
12780 Print the name and the counter upper bound for all debug counters.
12783 @item -fdbg-cnt=@var{counter-value-list}
12785 Set the internal debug counter upper bound. @var{counter-value-list}
12786 is a comma-separated list of @var{name}:@var{value} pairs
12787 which sets the upper bound of each debug counter @var{name} to @var{value}.
12788 All debug counters have the initial upper bound of @code{UINT_MAX};
12789 thus @code{dbg_cnt} returns true always unless the upper bound
12790 is set by this option.
12791 For example, with @option{-fdbg-cnt=dce:10,tail_call:0},
12792 @code{dbg_cnt(dce)} returns true only for first 10 invocations.
12794 @item -print-file-name=@var{library}
12795 @opindex print-file-name
12796 Print the full absolute name of the library file @var{library} that
12797 would be used when linking---and don't do anything else. With this
12798 option, GCC does not compile or link anything; it just prints the
12801 @item -print-multi-directory
12802 @opindex print-multi-directory
12803 Print the directory name corresponding to the multilib selected by any
12804 other switches present in the command line. This directory is supposed
12805 to exist in @env{GCC_EXEC_PREFIX}.
12807 @item -print-multi-lib
12808 @opindex print-multi-lib
12809 Print the mapping from multilib directory names to compiler switches
12810 that enable them. The directory name is separated from the switches by
12811 @samp{;}, and each switch starts with an @samp{@@} instead of the
12812 @samp{-}, without spaces between multiple switches. This is supposed to
12813 ease shell processing.
12815 @item -print-multi-os-directory
12816 @opindex print-multi-os-directory
12817 Print the path to OS libraries for the selected
12818 multilib, relative to some @file{lib} subdirectory. If OS libraries are
12819 present in the @file{lib} subdirectory and no multilibs are used, this is
12820 usually just @file{.}, if OS libraries are present in @file{lib@var{suffix}}
12821 sibling directories this prints e.g.@: @file{../lib64}, @file{../lib} or
12822 @file{../lib32}, or if OS libraries are present in @file{lib/@var{subdir}}
12823 subdirectories it prints e.g.@: @file{amd64}, @file{sparcv9} or @file{ev6}.
12825 @item -print-multiarch
12826 @opindex print-multiarch
12827 Print the path to OS libraries for the selected multiarch,
12828 relative to some @file{lib} subdirectory.
12830 @item -print-prog-name=@var{program}
12831 @opindex print-prog-name
12832 Like @option{-print-file-name}, but searches for a program such as @command{cpp}.
12834 @item -print-libgcc-file-name
12835 @opindex print-libgcc-file-name
12836 Same as @option{-print-file-name=libgcc.a}.
12838 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
12839 but you do want to link with @file{libgcc.a}. You can do:
12842 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
12845 @item -print-search-dirs
12846 @opindex print-search-dirs
12847 Print the name of the configured installation directory and a list of
12848 program and library directories @command{gcc} searches---and don't do anything else.
12850 This is useful when @command{gcc} prints the error message
12851 @samp{installation problem, cannot exec cpp0: No such file or directory}.
12852 To resolve this you either need to put @file{cpp0} and the other compiler
12853 components where @command{gcc} expects to find them, or you can set the environment
12854 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
12855 Don't forget the trailing @samp{/}.
12856 @xref{Environment Variables}.
12858 @item -print-sysroot
12859 @opindex print-sysroot
12860 Print the target sysroot directory that is used during
12861 compilation. This is the target sysroot specified either at configure
12862 time or using the @option{--sysroot} option, possibly with an extra
12863 suffix that depends on compilation options. If no target sysroot is
12864 specified, the option prints nothing.
12866 @item -print-sysroot-headers-suffix
12867 @opindex print-sysroot-headers-suffix
12868 Print the suffix added to the target sysroot when searching for
12869 headers, or give an error if the compiler is not configured with such
12870 a suffix---and don't do anything else.
12873 @opindex dumpmachine
12874 Print the compiler's target machine (for example,
12875 @samp{i686-pc-linux-gnu})---and don't do anything else.
12878 @opindex dumpversion
12879 Print the compiler version (for example, @code{3.0})---and don't do
12884 Print the compiler's built-in specs---and don't do anything else. (This
12885 is used when GCC itself is being built.) @xref{Spec Files}.
12888 @node Submodel Options
12889 @section Machine-Dependent Options
12890 @cindex submodel options
12891 @cindex specifying hardware config
12892 @cindex hardware models and configurations, specifying
12893 @cindex target-dependent options
12894 @cindex machine-dependent options
12896 Each target machine supported by GCC can have its own options---for
12897 example, to allow you to compile for a particular processor variant or
12898 ABI, or to control optimizations specific to that machine. By
12899 convention, the names of machine-specific options start with
12902 Some configurations of the compiler also support additional target-specific
12903 options, usually for compatibility with other compilers on the same
12906 @c This list is ordered alphanumerically by subsection name.
12907 @c It should be the same order and spelling as these options are listed
12908 @c in Machine Dependent Options
12911 * AArch64 Options::
12912 * Adapteva Epiphany Options::
12916 * Blackfin Options::
12921 * DEC Alpha Options::
12925 * GNU/Linux Options::
12935 * MicroBlaze Options::
12938 * MN10300 Options::
12942 * Nios II Options::
12943 * Nvidia PTX Options::
12945 * picoChip Options::
12946 * PowerPC Options::
12948 * RS/6000 and PowerPC Options::
12950 * S/390 and zSeries Options::
12953 * Solaris 2 Options::
12956 * System V Options::
12957 * TILE-Gx Options::
12958 * TILEPro Options::
12963 * VxWorks Options::
12965 * x86 Windows Options::
12966 * Xstormy16 Options::
12968 * zSeries Options::
12971 @node AArch64 Options
12972 @subsection AArch64 Options
12973 @cindex AArch64 Options
12975 These options are defined for AArch64 implementations:
12979 @item -mabi=@var{name}
12981 Generate code for the specified data model. Permissible values
12982 are @samp{ilp32} for SysV-like data model where int, long int and pointers
12983 are 32 bits, and @samp{lp64} for SysV-like data model where int is 32 bits,
12984 but long int and pointers are 64 bits.
12986 The default depends on the specific target configuration. Note that
12987 the LP64 and ILP32 ABIs are not link-compatible; you must compile your
12988 entire program with the same ABI, and link with a compatible set of libraries.
12991 @opindex mbig-endian
12992 Generate big-endian code. This is the default when GCC is configured for an
12993 @samp{aarch64_be-*-*} target.
12995 @item -mgeneral-regs-only
12996 @opindex mgeneral-regs-only
12997 Generate code which uses only the general-purpose registers. This will prevent
12998 the compiler from using floating-point and Advanced SIMD registers but will not
12999 impose any restrictions on the assembler.
13001 @item -mlittle-endian
13002 @opindex mlittle-endian
13003 Generate little-endian code. This is the default when GCC is configured for an
13004 @samp{aarch64-*-*} but not an @samp{aarch64_be-*-*} target.
13006 @item -mcmodel=tiny
13007 @opindex mcmodel=tiny
13008 Generate code for the tiny code model. The program and its statically defined
13009 symbols must be within 1MB of each other. Programs can be statically or
13010 dynamically linked.
13012 @item -mcmodel=small
13013 @opindex mcmodel=small
13014 Generate code for the small code model. The program and its statically defined
13015 symbols must be within 4GB of each other. Programs can be statically or
13016 dynamically linked. This is the default code model.
13018 @item -mcmodel=large
13019 @opindex mcmodel=large
13020 Generate code for the large code model. This makes no assumptions about
13021 addresses and sizes of sections. Programs can be statically linked only.
13023 @item -mstrict-align
13024 @opindex mstrict-align
13025 Avoid generating memory accesses that may not be aligned on a natural object
13026 boundary as described in the architecture specification.
13028 @item -momit-leaf-frame-pointer
13029 @itemx -mno-omit-leaf-frame-pointer
13030 @opindex momit-leaf-frame-pointer
13031 @opindex mno-omit-leaf-frame-pointer
13032 Omit or keep the frame pointer in leaf functions. The former behavior is the
13035 @item -mtls-dialect=desc
13036 @opindex mtls-dialect=desc
13037 Use TLS descriptors as the thread-local storage mechanism for dynamic accesses
13038 of TLS variables. This is the default.
13040 @item -mtls-dialect=traditional
13041 @opindex mtls-dialect=traditional
13042 Use traditional TLS as the thread-local storage mechanism for dynamic accesses
13045 @item -mtls-size=@var{size}
13047 Specify bit size of immediate TLS offsets. Valid values are 12, 24, 32, 48.
13048 This option requires binutils 2.26 or newer.
13050 @item -mfix-cortex-a53-835769
13051 @itemx -mno-fix-cortex-a53-835769
13052 @opindex mfix-cortex-a53-835769
13053 @opindex mno-fix-cortex-a53-835769
13054 Enable or disable the workaround for the ARM Cortex-A53 erratum number 835769.
13055 This involves inserting a NOP instruction between memory instructions and
13056 64-bit integer multiply-accumulate instructions.
13058 @item -mfix-cortex-a53-843419
13059 @itemx -mno-fix-cortex-a53-843419
13060 @opindex mfix-cortex-a53-843419
13061 @opindex mno-fix-cortex-a53-843419
13062 Enable or disable the workaround for the ARM Cortex-A53 erratum number 843419.
13063 This erratum workaround is made at link time and this will only pass the
13064 corresponding flag to the linker.
13066 @item -mlow-precision-recip-sqrt
13067 @item -mno-low-precision-recip-sqrt
13068 @opindex mlow-precision-recip-sqrt
13069 @opindex mno-low-precision-recip-sqrt
13070 Enable or disable the reciprocal square root approximation.
13071 This option only has an effect if @option{-ffast-math} or
13072 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13073 precision of reciprocal square root results to about 16 bits for
13074 single precision and to 32 bits for double precision.
13076 @item -mlow-precision-sqrt
13077 @item -mno-low-precision-sqrt
13078 @opindex -mlow-precision-sqrt
13079 @opindex -mno-low-precision-sqrt
13080 Enable or disable the square root approximation.
13081 This option only has an effect if @option{-ffast-math} or
13082 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13083 precision of square root results to about 16 bits for
13084 single precision and to 32 bits for double precision.
13085 If enabled, it implies @option{-mlow-precision-recip-sqrt}.
13087 @item -mlow-precision-div
13088 @item -mno-low-precision-div
13089 @opindex -mlow-precision-div
13090 @opindex -mno-low-precision-div
13091 Enable or disable the division approximation.
13092 This option only has an effect if @option{-ffast-math} or
13093 @option{-funsafe-math-optimizations} is used as well. Enabling this reduces
13094 precision of division results to about 16 bits for
13095 single precision and to 32 bits for double precision.
13097 @item -march=@var{name}
13099 Specify the name of the target architecture and, optionally, one or
13100 more feature modifiers. This option has the form
13101 @option{-march=@var{arch}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}.
13103 The permissible values for @var{arch} are @samp{armv8-a},
13104 @samp{armv8.1-a} or @var{native}.
13106 The value @samp{armv8.1-a} implies @samp{armv8-a} and enables compiler
13107 support for the ARMv8.1 architecture extension. In particular, it
13108 enables the @samp{+crc} and @samp{+lse} features.
13110 The value @samp{native} is available on native AArch64 GNU/Linux and
13111 causes the compiler to pick the architecture of the host system. This
13112 option has no effect if the compiler is unable to recognize the
13113 architecture of the host system,
13115 The permissible values for @var{feature} are listed in the sub-section
13116 on @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13117 Feature Modifiers}. Where conflicting feature modifiers are
13118 specified, the right-most feature is used.
13120 GCC uses @var{name} to determine what kind of instructions it can emit
13121 when generating assembly code. If @option{-march} is specified
13122 without either of @option{-mtune} or @option{-mcpu} also being
13123 specified, the code is tuned to perform well across a range of target
13124 processors implementing the target architecture.
13126 @item -mtune=@var{name}
13128 Specify the name of the target processor for which GCC should tune the
13129 performance of the code. Permissible values for this option are:
13130 @samp{generic}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
13131 @samp{cortex-a72}, @samp{cortex-a73}, @samp{exynos-m1}, @samp{qdf24xx},
13132 @samp{thunderx}, @samp{xgene1}, @samp{vulcan}, @samp{cortex-a57.cortex-a53},
13133 @samp{cortex-a72.cortex-a53}, @samp{cortex-a73.cortex-a35},
13134 @samp{cortex-a73.cortex-a53}, @samp{native}.
13136 The values @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
13137 @samp{cortex-a73.cortex-a35}, @samp{cortex-a73.cortex-a53}
13138 specify that GCC should tune for a big.LITTLE system.
13140 Additionally on native AArch64 GNU/Linux systems the value
13141 @samp{native} tunes performance to the host system. This option has no effect
13142 if the compiler is unable to recognize the processor of the host system.
13144 Where none of @option{-mtune=}, @option{-mcpu=} or @option{-march=}
13145 are specified, the code is tuned to perform well across a range
13146 of target processors.
13148 This option cannot be suffixed by feature modifiers.
13150 @item -mcpu=@var{name}
13152 Specify the name of the target processor, optionally suffixed by one
13153 or more feature modifiers. This option has the form
13154 @option{-mcpu=@var{cpu}@r{@{}+@r{[}no@r{]}@var{feature}@r{@}*}}, where
13155 the permissible values for @var{cpu} are the same as those available
13156 for @option{-mtune}. The permissible values for @var{feature} are
13157 documented in the sub-section on
13158 @ref{aarch64-feature-modifiers,,@option{-march} and @option{-mcpu}
13159 Feature Modifiers}. Where conflicting feature modifiers are
13160 specified, the right-most feature is used.
13162 GCC uses @var{name} to determine what kind of instructions it can emit when
13163 generating assembly code (as if by @option{-march}) and to determine
13164 the target processor for which to tune for performance (as if
13165 by @option{-mtune}). Where this option is used in conjunction
13166 with @option{-march} or @option{-mtune}, those options take precedence
13167 over the appropriate part of this option.
13169 @item -moverride=@var{string}
13171 Override tuning decisions made by the back-end in response to a
13172 @option{-mtune=} switch. The syntax, semantics, and accepted values
13173 for @var{string} in this option are not guaranteed to be consistent
13176 This option is only intended to be useful when developing GCC.
13178 @item -mpc-relative-literal-loads
13179 @opindex mpc-relative-literal-loads
13180 Enable PC-relative literal loads. With this option literal pools are
13181 accessed using a single instruction and emitted after each function. This
13182 limits the maximum size of functions to 1MB. This is enabled by default for
13183 @option{-mcmodel=tiny}.
13187 @subsubsection @option{-march} and @option{-mcpu} Feature Modifiers
13188 @anchor{aarch64-feature-modifiers}
13189 @cindex @option{-march} feature modifiers
13190 @cindex @option{-mcpu} feature modifiers
13191 Feature modifiers used with @option{-march} and @option{-mcpu} can be any of
13192 the following and their inverses @option{no@var{feature}}:
13196 Enable CRC extension. This is on by default for
13197 @option{-march=armv8.1-a}.
13199 Enable Crypto extension. This also enables Advanced SIMD and floating-point
13202 Enable floating-point instructions. This is on by default for all possible
13203 values for options @option{-march} and @option{-mcpu}.
13205 Enable Advanced SIMD instructions. This also enables floating-point
13206 instructions. This is on by default for all possible values for options
13207 @option{-march} and @option{-mcpu}.
13209 Enable Large System Extension instructions. This is on by default for
13210 @option{-march=armv8.1-a}.
13214 Feature @option{crypto} implies @option{simd}, which implies @option{fp}.
13215 Conversely, @option{nofp} implies @option{nosimd}, which implies
13218 @node Adapteva Epiphany Options
13219 @subsection Adapteva Epiphany Options
13221 These @samp{-m} options are defined for Adapteva Epiphany:
13224 @item -mhalf-reg-file
13225 @opindex mhalf-reg-file
13226 Don't allocate any register in the range @code{r32}@dots{}@code{r63}.
13227 That allows code to run on hardware variants that lack these registers.
13229 @item -mprefer-short-insn-regs
13230 @opindex mprefer-short-insn-regs
13231 Preferentially allocate registers that allow short instruction generation.
13232 This can result in increased instruction count, so this may either reduce or
13233 increase overall code size.
13235 @item -mbranch-cost=@var{num}
13236 @opindex mbranch-cost
13237 Set the cost of branches to roughly @var{num} ``simple'' instructions.
13238 This cost is only a heuristic and is not guaranteed to produce
13239 consistent results across releases.
13243 Enable the generation of conditional moves.
13245 @item -mnops=@var{num}
13247 Emit @var{num} NOPs before every other generated instruction.
13249 @item -mno-soft-cmpsf
13250 @opindex mno-soft-cmpsf
13251 For single-precision floating-point comparisons, emit an @code{fsub} instruction
13252 and test the flags. This is faster than a software comparison, but can
13253 get incorrect results in the presence of NaNs, or when two different small
13254 numbers are compared such that their difference is calculated as zero.
13255 The default is @option{-msoft-cmpsf}, which uses slower, but IEEE-compliant,
13256 software comparisons.
13258 @item -mstack-offset=@var{num}
13259 @opindex mstack-offset
13260 Set the offset between the top of the stack and the stack pointer.
13261 E.g., a value of 8 means that the eight bytes in the range @code{sp+0@dots{}sp+7}
13262 can be used by leaf functions without stack allocation.
13263 Values other than @samp{8} or @samp{16} are untested and unlikely to work.
13264 Note also that this option changes the ABI; compiling a program with a
13265 different stack offset than the libraries have been compiled with
13266 generally does not work.
13267 This option can be useful if you want to evaluate if a different stack
13268 offset would give you better code, but to actually use a different stack
13269 offset to build working programs, it is recommended to configure the
13270 toolchain with the appropriate @option{--with-stack-offset=@var{num}} option.
13272 @item -mno-round-nearest
13273 @opindex mno-round-nearest
13274 Make the scheduler assume that the rounding mode has been set to
13275 truncating. The default is @option{-mround-nearest}.
13278 @opindex mlong-calls
13279 If not otherwise specified by an attribute, assume all calls might be beyond
13280 the offset range of the @code{b} / @code{bl} instructions, and therefore load the
13281 function address into a register before performing a (otherwise direct) call.
13282 This is the default.
13284 @item -mshort-calls
13285 @opindex short-calls
13286 If not otherwise specified by an attribute, assume all direct calls are
13287 in the range of the @code{b} / @code{bl} instructions, so use these instructions
13288 for direct calls. The default is @option{-mlong-calls}.
13292 Assume addresses can be loaded as 16-bit unsigned values. This does not
13293 apply to function addresses for which @option{-mlong-calls} semantics
13296 @item -mfp-mode=@var{mode}
13298 Set the prevailing mode of the floating-point unit.
13299 This determines the floating-point mode that is provided and expected
13300 at function call and return time. Making this mode match the mode you
13301 predominantly need at function start can make your programs smaller and
13302 faster by avoiding unnecessary mode switches.
13304 @var{mode} can be set to one the following values:
13308 Any mode at function entry is valid, and retained or restored when
13309 the function returns, and when it calls other functions.
13310 This mode is useful for compiling libraries or other compilation units
13311 you might want to incorporate into different programs with different
13312 prevailing FPU modes, and the convenience of being able to use a single
13313 object file outweighs the size and speed overhead for any extra
13314 mode switching that might be needed, compared with what would be needed
13315 with a more specific choice of prevailing FPU mode.
13318 This is the mode used for floating-point calculations with
13319 truncating (i.e.@: round towards zero) rounding mode. That includes
13320 conversion from floating point to integer.
13322 @item round-nearest
13323 This is the mode used for floating-point calculations with
13324 round-to-nearest-or-even rounding mode.
13327 This is the mode used to perform integer calculations in the FPU, e.g.@:
13328 integer multiply, or integer multiply-and-accumulate.
13331 The default is @option{-mfp-mode=caller}
13333 @item -mnosplit-lohi
13334 @itemx -mno-postinc
13335 @itemx -mno-postmodify
13336 @opindex mnosplit-lohi
13337 @opindex mno-postinc
13338 @opindex mno-postmodify
13339 Code generation tweaks that disable, respectively, splitting of 32-bit
13340 loads, generation of post-increment addresses, and generation of
13341 post-modify addresses. The defaults are @option{msplit-lohi},
13342 @option{-mpost-inc}, and @option{-mpost-modify}.
13344 @item -mnovect-double
13345 @opindex mno-vect-double
13346 Change the preferred SIMD mode to SImode. The default is
13347 @option{-mvect-double}, which uses DImode as preferred SIMD mode.
13349 @item -max-vect-align=@var{num}
13350 @opindex max-vect-align
13351 The maximum alignment for SIMD vector mode types.
13352 @var{num} may be 4 or 8. The default is 8.
13353 Note that this is an ABI change, even though many library function
13354 interfaces are unaffected if they don't use SIMD vector modes
13355 in places that affect size and/or alignment of relevant types.
13357 @item -msplit-vecmove-early
13358 @opindex msplit-vecmove-early
13359 Split vector moves into single word moves before reload. In theory this
13360 can give better register allocation, but so far the reverse seems to be
13361 generally the case.
13363 @item -m1reg-@var{reg}
13365 Specify a register to hold the constant @minus{}1, which makes loading small negative
13366 constants and certain bitmasks faster.
13367 Allowable values for @var{reg} are @samp{r43} and @samp{r63},
13368 which specify use of that register as a fixed register,
13369 and @samp{none}, which means that no register is used for this
13370 purpose. The default is @option{-m1reg-none}.
13375 @subsection ARC Options
13376 @cindex ARC options
13378 The following options control the architecture variant for which code
13381 @c architecture variants
13384 @item -mbarrel-shifter
13385 @opindex mbarrel-shifter
13386 Generate instructions supported by barrel shifter. This is the default
13387 unless @option{-mcpu=ARC601} or @samp{-mcpu=ARCEM} is in effect.
13389 @item -mcpu=@var{cpu}
13391 Set architecture type, register usage, and instruction scheduling
13392 parameters for @var{cpu}. There are also shortcut alias options
13393 available for backward compatibility and convenience. Supported
13394 values for @var{cpu} are
13401 Compile for ARC600. Aliases: @option{-mA6}, @option{-mARC600}.
13406 Compile for ARC601. Alias: @option{-mARC601}.
13412 Compile for ARC700. Aliases: @option{-mA7}, @option{-mARC700}.
13413 This is the default when configured with @option{--with-cpu=arc700}@.
13417 Compile for ARC EM.
13421 Compile for ARC HS.
13426 @itemx -mdpfp-compact
13427 @opindex mdpfp-compact
13428 FPX: Generate Double Precision FPX instructions, tuned for the compact
13432 @opindex mdpfp-fast
13433 FPX: Generate Double Precision FPX instructions, tuned for the fast
13436 @item -mno-dpfp-lrsr
13437 @opindex mno-dpfp-lrsr
13438 Disable LR and SR instructions from using FPX extension aux registers.
13442 Generate Extended arithmetic instructions. Currently only
13443 @code{divaw}, @code{adds}, @code{subs}, and @code{sat16} are
13444 supported. This is always enabled for @option{-mcpu=ARC700}.
13448 Do not generate mpy instructions for ARC700.
13452 Generate 32x16 bit multiply and mac instructions.
13456 Generate mul64 and mulu64 instructions. Only valid for @option{-mcpu=ARC600}.
13460 Generate norm instruction. This is the default if @option{-mcpu=ARC700}
13465 @itemx -mspfp-compact
13466 @opindex mspfp-compact
13467 FPX: Generate Single Precision FPX instructions, tuned for the compact
13471 @opindex mspfp-fast
13472 FPX: Generate Single Precision FPX instructions, tuned for the fast
13477 Enable generation of ARC SIMD instructions via target-specific
13478 builtins. Only valid for @option{-mcpu=ARC700}.
13481 @opindex msoft-float
13482 This option ignored; it is provided for compatibility purposes only.
13483 Software floating point code is emitted by default, and this default
13484 can overridden by FPX options; @samp{mspfp}, @samp{mspfp-compact}, or
13485 @samp{mspfp-fast} for single precision, and @samp{mdpfp},
13486 @samp{mdpfp-compact}, or @samp{mdpfp-fast} for double precision.
13490 Generate swap instructions.
13494 This enables Locked Load/Store Conditional extension to implement
13495 atomic memopry built-in functions. Not available for ARC 6xx or ARC
13500 Enable DIV/REM instructions for ARCv2 cores.
13502 @item -mcode-density
13503 @opindex mcode-density
13504 Enable code density instructions for ARC EM, default on for ARC HS.
13508 Enable double load/store operations for ARC HS cores.
13510 @item -mtp-regno=@var{regno}
13512 Specify thread pointer register number.
13514 @item -mmpy-option=@var{multo}
13515 @opindex mmpy-option
13516 Compile ARCv2 code with a multiplier design option. @samp{wlh1} is
13517 the default value. The recognized values for @var{multo} are:
13521 No multiplier available.
13525 The multiply option is set to w: 16x16 multiplier, fully pipelined.
13526 The following instructions are enabled: MPYW, and MPYUW.
13530 The multiply option is set to wlh1: 32x32 multiplier, fully
13531 pipelined (1 stage). The following instructions are additionally
13532 enabled: MPY, MPYU, MPYM, MPYMU, and MPY_S.
13536 The multiply option is set to wlh2: 32x32 multiplier, fully pipelined
13537 (2 stages). The following instructions are additionally enabled: MPY,
13538 MPYU, MPYM, MPYMU, and MPY_S.
13542 The multiply option is set to wlh3: Two 16x16 multiplier, blocking,
13543 sequential. The following instructions are additionally enabled: MPY,
13544 MPYU, MPYM, MPYMU, and MPY_S.
13548 The multiply option is set to wlh4: One 16x16 multiplier, blocking,
13549 sequential. The following instructions are additionally enabled: MPY,
13550 MPYU, MPYM, MPYMU, and MPY_S.
13554 The multiply option is set to wlh5: One 32x4 multiplier, blocking,
13555 sequential. The following instructions are additionally enabled: MPY,
13556 MPYU, MPYM, MPYMU, and MPY_S.
13560 This option is only available for ARCv2 cores@.
13562 @item -mfpu=@var{fpu}
13564 Enables specific floating-point hardware extension for ARCv2
13565 core. Supported values for @var{fpu} are:
13571 Enables support for single precision floating point hardware
13576 Enables support for double precision floating point hardware
13577 extensions. The single precision floating point extension is also
13578 enabled. Not available for ARC EM@.
13582 Enables support for double precision floating point hardware
13583 extensions using double precision assist instructions. The single
13584 precision floating point extension is also enabled. This option is
13585 only available for ARC EM@.
13589 Enables support for double precision floating point hardware
13590 extensions using double precision assist instructions, and simple
13591 precision square-root and divide hardware extensions. The single
13592 precision floating point extension is also enabled. This option is
13593 only available for ARC EM@.
13597 Enables support for double precision floating point hardware
13598 extensions using double precision assist instructions, and simple
13599 precision fused multiple and add hardware extension. The single
13600 precision floating point extension is also enabled. This option is
13601 only available for ARC EM@.
13605 Enables support for double precision floating point hardware
13606 extensions using double precision assist instructions, and all simple
13607 precision hardware extensions. The single precision floating point
13608 extension is also enabled. This option is only available for ARC EM@.
13612 Enables support for single precision floating point, and single
13613 precision square-root and divide hardware extensions@.
13617 Enables support for double precision floating point, and double
13618 precision square-root and divide hardware extensions. This option
13619 includes option @samp{fpus_div}. Not available for ARC EM@.
13623 Enables support for single precision floating point, and single
13624 precision fused multiple and add hardware extensions@.
13628 Enables support for double precision floating point, and double
13629 precision fused multiple and add hardware extensions. This option
13630 includes option @samp{fpus_fma}. Not available for ARC EM@.
13634 Enables support for all single precision floating point hardware
13639 Enables support for all single and double precision floating point
13640 hardware extensions. Not available for ARC EM@.
13646 The following options are passed through to the assembler, and also
13647 define preprocessor macro symbols.
13649 @c Flags used by the assembler, but for which we define preprocessor
13650 @c macro symbols as well.
13653 @opindex mdsp-packa
13654 Passed down to the assembler to enable the DSP Pack A extensions.
13655 Also sets the preprocessor symbol @code{__Xdsp_packa}.
13659 Passed down to the assembler to enable the dual viterbi butterfly
13660 extension. Also sets the preprocessor symbol @code{__Xdvbf}.
13662 @c ARC700 4.10 extension instruction
13665 Passed down to the assembler to enable the Locked Load/Store
13666 Conditional extension. Also sets the preprocessor symbol
13671 Passed down to the assembler. Also sets the preprocessor symbol
13672 @code{__Xxmac_d16}.
13676 Passed down to the assembler. Also sets the preprocessor symbol
13679 @c ARC700 4.10 extension instruction
13682 Passed down to the assembler to enable the 64-bit Time-Stamp Counter
13683 extension instruction. Also sets the preprocessor symbol
13686 @c ARC700 4.10 extension instruction
13689 Passed down to the assembler to enable the swap byte ordering
13690 extension instruction. Also sets the preprocessor symbol
13694 @opindex mtelephony
13695 Passed down to the assembler to enable dual and single operand
13696 instructions for telephony. Also sets the preprocessor symbol
13697 @code{__Xtelephony}.
13701 Passed down to the assembler to enable the XY Memory extension. Also
13702 sets the preprocessor symbol @code{__Xxy}.
13706 The following options control how the assembly code is annotated:
13708 @c Assembly annotation options
13712 Annotate assembler instructions with estimated addresses.
13714 @item -mannotate-align
13715 @opindex mannotate-align
13716 Explain what alignment considerations lead to the decision to make an
13717 instruction short or long.
13721 The following options are passed through to the linker:
13723 @c options passed through to the linker
13727 Passed through to the linker, to specify use of the @code{arclinux} emulation.
13728 This option is enabled by default in tool chains built for
13729 @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets
13730 when profiling is not requested.
13732 @item -marclinux_prof
13733 @opindex marclinux_prof
13734 Passed through to the linker, to specify use of the
13735 @code{arclinux_prof} emulation. This option is enabled by default in
13736 tool chains built for @w{@code{arc-linux-uclibc}} and
13737 @w{@code{arceb-linux-uclibc}} targets when profiling is requested.
13741 The following options control the semantics of generated code:
13743 @c semantically relevant code generation options
13746 @opindex mlong-calls
13747 Generate call insns as register indirect calls, thus providing access
13748 to the full 32-bit address range.
13750 @item -mmedium-calls
13751 @opindex mmedium-calls
13752 Don't use less than 25 bit addressing range for calls, which is the
13753 offset available for an unconditional branch-and-link
13754 instruction. Conditional execution of function calls is suppressed, to
13755 allow use of the 25-bit range, rather than the 21-bit range with
13756 conditional branch-and-link. This is the default for tool chains built
13757 for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}} targets.
13761 Do not generate sdata references. This is the default for tool chains
13762 built for @w{@code{arc-linux-uclibc}} and @w{@code{arceb-linux-uclibc}}
13766 @opindex mucb-mcount
13767 Instrument with mcount calls as used in UCB code. I.e. do the
13768 counting in the callee, not the caller. By default ARC instrumentation
13769 counts in the caller.
13771 @item -mvolatile-cache
13772 @opindex mvolatile-cache
13773 Use ordinarily cached memory accesses for volatile references. This is the
13776 @item -mno-volatile-cache
13777 @opindex mno-volatile-cache
13778 Enable cache bypass for volatile references.
13782 The following options fine tune code generation:
13783 @c code generation tuning options
13786 @opindex malign-call
13787 Do alignment optimizations for call instructions.
13789 @item -mauto-modify-reg
13790 @opindex mauto-modify-reg
13791 Enable the use of pre/post modify with register displacement.
13793 @item -mbbit-peephole
13794 @opindex mbbit-peephole
13795 Enable bbit peephole2.
13799 This option disables a target-specific pass in @file{arc_reorg} to
13800 generate @code{BRcc} instructions. It has no effect on @code{BRcc}
13801 generation driven by the combiner pass.
13803 @item -mcase-vector-pcrel
13804 @opindex mcase-vector-pcrel
13805 Use pc-relative switch case tables - this enables case table shortening.
13806 This is the default for @option{-Os}.
13808 @item -mcompact-casesi
13809 @opindex mcompact-casesi
13810 Enable compact casesi pattern.
13811 This is the default for @option{-Os}.
13813 @item -mno-cond-exec
13814 @opindex mno-cond-exec
13815 Disable ARCompact specific pass to generate conditional execution instructions.
13816 Due to delay slot scheduling and interactions between operand numbers,
13817 literal sizes, instruction lengths, and the support for conditional execution,
13818 the target-independent pass to generate conditional execution is often lacking,
13819 so the ARC port has kept a special pass around that tries to find more
13820 conditional execution generating opportunities after register allocation,
13821 branch shortening, and delay slot scheduling have been done. This pass
13822 generally, but not always, improves performance and code size, at the cost of
13823 extra compilation time, which is why there is an option to switch it off.
13824 If you have a problem with call instructions exceeding their allowable
13825 offset range because they are conditionalized, you should consider using
13826 @option{-mmedium-calls} instead.
13828 @item -mearly-cbranchsi
13829 @opindex mearly-cbranchsi
13830 Enable pre-reload use of the cbranchsi pattern.
13832 @item -mexpand-adddi
13833 @opindex mexpand-adddi
13834 Expand @code{adddi3} and @code{subdi3} at rtl generation time into
13835 @code{add.f}, @code{adc} etc.
13837 @item -mindexed-loads
13838 @opindex mindexed-loads
13839 Enable the use of indexed loads. This can be problematic because some
13840 optimizers then assume that indexed stores exist, which is not
13844 Enable Local Register Allocation. This is still experimental for ARC,
13845 so by default the compiler uses standard reload
13846 (i.e. @option{-mno-lra}).
13848 @item -mlra-priority-none
13849 @opindex mlra-priority-none
13850 Don't indicate any priority for target registers.
13852 @item -mlra-priority-compact
13853 @opindex mlra-priority-compact
13854 Indicate target register priority for r0..r3 / r12..r15.
13856 @item -mlra-priority-noncompact
13857 @opindex mlra-priority-noncompact
13858 Reduce target register priority for r0..r3 / r12..r15.
13860 @item -mno-millicode
13861 @opindex mno-millicode
13862 When optimizing for size (using @option{-Os}), prologues and epilogues
13863 that have to save or restore a large number of registers are often
13864 shortened by using call to a special function in libgcc; this is
13865 referred to as a @emph{millicode} call. As these calls can pose
13866 performance issues, and/or cause linking issues when linking in a
13867 nonstandard way, this option is provided to turn off millicode call
13871 @opindex mmixed-code
13872 Tweak register allocation to help 16-bit instruction generation.
13873 This generally has the effect of decreasing the average instruction size
13874 while increasing the instruction count.
13878 Enable 'q' instruction alternatives.
13879 This is the default for @option{-Os}.
13883 Enable Rcq constraint handling - most short code generation depends on this.
13884 This is the default.
13888 Enable Rcw constraint handling - ccfsm condexec mostly depends on this.
13889 This is the default.
13891 @item -msize-level=@var{level}
13892 @opindex msize-level
13893 Fine-tune size optimization with regards to instruction lengths and alignment.
13894 The recognized values for @var{level} are:
13897 No size optimization. This level is deprecated and treated like @samp{1}.
13900 Short instructions are used opportunistically.
13903 In addition, alignment of loops and of code after barriers are dropped.
13906 In addition, optional data alignment is dropped, and the option @option{Os} is enabled.
13910 This defaults to @samp{3} when @option{-Os} is in effect. Otherwise,
13911 the behavior when this is not set is equivalent to level @samp{1}.
13913 @item -mtune=@var{cpu}
13915 Set instruction scheduling parameters for @var{cpu}, overriding any implied
13916 by @option{-mcpu=}.
13918 Supported values for @var{cpu} are
13922 Tune for ARC600 cpu.
13925 Tune for ARC601 cpu.
13928 Tune for ARC700 cpu with standard multiplier block.
13931 Tune for ARC700 cpu with XMAC block.
13934 Tune for ARC725D cpu.
13937 Tune for ARC750D cpu.
13941 @item -mmultcost=@var{num}
13943 Cost to assume for a multiply instruction, with @samp{4} being equal to a
13944 normal instruction.
13946 @item -munalign-prob-threshold=@var{probability}
13947 @opindex munalign-prob-threshold
13948 Set probability threshold for unaligning branches.
13949 When tuning for @samp{ARC700} and optimizing for speed, branches without
13950 filled delay slot are preferably emitted unaligned and long, unless
13951 profiling indicates that the probability for the branch to be taken
13952 is below @var{probability}. @xref{Cross-profiling}.
13953 The default is (REG_BR_PROB_BASE/2), i.e.@: 5000.
13957 The following options are maintained for backward compatibility, but
13958 are now deprecated and will be removed in a future release:
13960 @c Deprecated options
13968 @opindex mbig-endian
13971 Compile code for big endian targets. Use of these options is now
13972 deprecated. Users wanting big-endian code, should use the
13973 @w{@code{arceb-elf32}} and @w{@code{arceb-linux-uclibc}} targets when
13974 building the tool chain, for which big-endian is the default.
13976 @item -mlittle-endian
13977 @opindex mlittle-endian
13980 Compile code for little endian targets. Use of these options is now
13981 deprecated. Users wanting little-endian code should use the
13982 @w{@code{arc-elf32}} and @w{@code{arc-linux-uclibc}} targets when
13983 building the tool chain, for which little-endian is the default.
13985 @item -mbarrel_shifter
13986 @opindex mbarrel_shifter
13987 Replaced by @option{-mbarrel-shifter}.
13989 @item -mdpfp_compact
13990 @opindex mdpfp_compact
13991 Replaced by @option{-mdpfp-compact}.
13994 @opindex mdpfp_fast
13995 Replaced by @option{-mdpfp-fast}.
13998 @opindex mdsp_packa
13999 Replaced by @option{-mdsp-packa}.
14003 Replaced by @option{-mea}.
14007 Replaced by @option{-mmac-24}.
14011 Replaced by @option{-mmac-d16}.
14013 @item -mspfp_compact
14014 @opindex mspfp_compact
14015 Replaced by @option{-mspfp-compact}.
14018 @opindex mspfp_fast
14019 Replaced by @option{-mspfp-fast}.
14021 @item -mtune=@var{cpu}
14023 Values @samp{arc600}, @samp{arc601}, @samp{arc700} and
14024 @samp{arc700-xmac} for @var{cpu} are replaced by @samp{ARC600},
14025 @samp{ARC601}, @samp{ARC700} and @samp{ARC700-xmac} respectively
14027 @item -multcost=@var{num}
14029 Replaced by @option{-mmultcost}.
14034 @subsection ARM Options
14035 @cindex ARM options
14037 These @samp{-m} options are defined for the ARM port:
14040 @item -mabi=@var{name}
14042 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
14043 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
14046 @opindex mapcs-frame
14047 Generate a stack frame that is compliant with the ARM Procedure Call
14048 Standard for all functions, even if this is not strictly necessary for
14049 correct execution of the code. Specifying @option{-fomit-frame-pointer}
14050 with this option causes the stack frames not to be generated for
14051 leaf functions. The default is @option{-mno-apcs-frame}.
14052 This option is deprecated.
14056 This is a synonym for @option{-mapcs-frame} and is deprecated.
14059 @c not currently implemented
14060 @item -mapcs-stack-check
14061 @opindex mapcs-stack-check
14062 Generate code to check the amount of stack space available upon entry to
14063 every function (that actually uses some stack space). If there is
14064 insufficient space available then either the function
14065 @code{__rt_stkovf_split_small} or @code{__rt_stkovf_split_big} is
14066 called, depending upon the amount of stack space required. The runtime
14067 system is required to provide these functions. The default is
14068 @option{-mno-apcs-stack-check}, since this produces smaller code.
14070 @c not currently implemented
14072 @opindex mapcs-float
14073 Pass floating-point arguments using the floating-point registers. This is
14074 one of the variants of the APCS@. This option is recommended if the
14075 target hardware has a floating-point unit or if a lot of floating-point
14076 arithmetic is going to be performed by the code. The default is
14077 @option{-mno-apcs-float}, since the size of integer-only code is
14078 slightly increased if @option{-mapcs-float} is used.
14080 @c not currently implemented
14081 @item -mapcs-reentrant
14082 @opindex mapcs-reentrant
14083 Generate reentrant, position-independent code. The default is
14084 @option{-mno-apcs-reentrant}.
14087 @item -mthumb-interwork
14088 @opindex mthumb-interwork
14089 Generate code that supports calling between the ARM and Thumb
14090 instruction sets. Without this option, on pre-v5 architectures, the
14091 two instruction sets cannot be reliably used inside one program. The
14092 default is @option{-mno-thumb-interwork}, since slightly larger code
14093 is generated when @option{-mthumb-interwork} is specified. In AAPCS
14094 configurations this option is meaningless.
14096 @item -mno-sched-prolog
14097 @opindex mno-sched-prolog
14098 Prevent the reordering of instructions in the function prologue, or the
14099 merging of those instruction with the instructions in the function's
14100 body. This means that all functions start with a recognizable set
14101 of instructions (or in fact one of a choice from a small set of
14102 different function prologues), and this information can be used to
14103 locate the start of functions inside an executable piece of code. The
14104 default is @option{-msched-prolog}.
14106 @item -mfloat-abi=@var{name}
14107 @opindex mfloat-abi
14108 Specifies which floating-point ABI to use. Permissible values
14109 are: @samp{soft}, @samp{softfp} and @samp{hard}.
14111 Specifying @samp{soft} causes GCC to generate output containing
14112 library calls for floating-point operations.
14113 @samp{softfp} allows the generation of code using hardware floating-point
14114 instructions, but still uses the soft-float calling conventions.
14115 @samp{hard} allows generation of floating-point instructions
14116 and uses FPU-specific calling conventions.
14118 The default depends on the specific target configuration. Note that
14119 the hard-float and soft-float ABIs are not link-compatible; you must
14120 compile your entire program with the same ABI, and link with a
14121 compatible set of libraries.
14123 @item -mlittle-endian
14124 @opindex mlittle-endian
14125 Generate code for a processor running in little-endian mode. This is
14126 the default for all standard configurations.
14129 @opindex mbig-endian
14130 Generate code for a processor running in big-endian mode; the default is
14131 to compile code for a little-endian processor.
14133 @item -march=@var{name}
14135 This specifies the name of the target ARM architecture. GCC uses this
14136 name to determine what kind of instructions it can emit when generating
14137 assembly code. This option can be used in conjunction with or instead
14138 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
14139 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
14140 @samp{armv5}, @samp{armv5e}, @samp{armv5t}, @samp{armv5te},
14141 @samp{armv6}, @samp{armv6-m}, @samp{armv6j}, @samp{armv6k},
14142 @samp{armv6kz}, @samp{armv6s-m},
14143 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk},
14144 @samp{armv7}, @samp{armv7-a}, @samp{armv7-m}, @samp{armv7-r}, @samp{armv7e-m},
14145 @samp{armv7ve}, @samp{armv8-a}, @samp{armv8-a+crc}, @samp{armv8.1-a},
14146 @samp{armv8.1-a+crc}, @samp{iwmmxt}, @samp{iwmmxt2}.
14148 Architecture revisions older than @samp{armv4t} are deprecated.
14150 @option{-march=armv6s-m} is the @samp{armv6-m} architecture with support for
14151 the (now mandatory) SVC instruction.
14153 @option{-march=armv6zk} is an alias for @samp{armv6kz}, existing for backwards
14156 @option{-march=armv7ve} is the @samp{armv7-a} architecture with virtualization
14159 @option{-march=armv8-a+crc} enables code generation for the ARMv8-A
14160 architecture together with the optional CRC32 extensions.
14162 @option{-march=native} causes the compiler to auto-detect the architecture
14163 of the build computer. At present, this feature is only supported on
14164 GNU/Linux, and not all architectures are recognized. If the auto-detect
14165 is unsuccessful the option has no effect.
14167 @item -mtune=@var{name}
14169 This option specifies the name of the target ARM processor for
14170 which GCC should tune the performance of the code.
14171 For some ARM implementations better performance can be obtained by using
14173 Permissible names are: @samp{arm2}, @samp{arm250},
14174 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
14175 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
14176 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
14177 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
14179 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
14180 @samp{arm710t}, @samp{arm720t}, @samp{arm740t},
14181 @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
14182 @samp{strongarm1110},
14183 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
14184 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
14185 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
14186 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
14187 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
14188 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
14189 @samp{arm1156t2-s}, @samp{arm1156t2f-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
14190 @samp{generic-armv7-a}, @samp{cortex-a5}, @samp{cortex-a7}, @samp{cortex-a8},
14191 @samp{cortex-a9}, @samp{cortex-a12}, @samp{cortex-a15}, @samp{cortex-a17},
14192 @samp{cortex-a32}, @samp{cortex-a35}, @samp{cortex-a53}, @samp{cortex-a57},
14193 @samp{cortex-a72}, @samp{cortex-a73}, @samp{cortex-r4},
14194 @samp{cortex-r4f}, @samp{cortex-r5}, @samp{cortex-r7}, @samp{cortex-r8},
14200 @samp{cortex-m0plus},
14201 @samp{cortex-m1.small-multiply},
14202 @samp{cortex-m0.small-multiply},
14203 @samp{cortex-m0plus.small-multiply},
14206 @samp{marvell-pj4},
14207 @samp{xscale}, @samp{iwmmxt}, @samp{iwmmxt2}, @samp{ep9312},
14208 @samp{fa526}, @samp{fa626},
14209 @samp{fa606te}, @samp{fa626te}, @samp{fmp626}, @samp{fa726te},
14212 Additionally, this option can specify that GCC should tune the performance
14213 of the code for a big.LITTLE system. Permissible names are:
14214 @samp{cortex-a15.cortex-a7}, @samp{cortex-a17.cortex-a7},
14215 @samp{cortex-a57.cortex-a53}, @samp{cortex-a72.cortex-a53},
14216 @samp{cortex-a72.cortex-a35}, @samp{cortex-a73.cortex-a53}.
14218 @option{-mtune=generic-@var{arch}} specifies that GCC should tune the
14219 performance for a blend of processors within architecture @var{arch}.
14220 The aim is to generate code that run well on the current most popular
14221 processors, balancing between optimizations that benefit some CPUs in the
14222 range, and avoiding performance pitfalls of other CPUs. The effects of
14223 this option may change in future GCC versions as CPU models come and go.
14225 @option{-mtune=native} causes the compiler to auto-detect the CPU
14226 of the build computer. At present, this feature is only supported on
14227 GNU/Linux, and not all architectures are recognized. If the auto-detect is
14228 unsuccessful the option has no effect.
14230 @item -mcpu=@var{name}
14232 This specifies the name of the target ARM processor. GCC uses this name
14233 to derive the name of the target ARM architecture (as if specified
14234 by @option{-march}) and the ARM processor type for which to tune for
14235 performance (as if specified by @option{-mtune}). Where this option
14236 is used in conjunction with @option{-march} or @option{-mtune},
14237 those options take precedence over the appropriate part of this option.
14239 Permissible names for this option are the same as those for
14242 @option{-mcpu=generic-@var{arch}} is also permissible, and is
14243 equivalent to @option{-march=@var{arch} -mtune=generic-@var{arch}}.
14244 See @option{-mtune} for more information.
14246 @option{-mcpu=native} causes the compiler to auto-detect the CPU
14247 of the build computer. At present, this feature is only supported on
14248 GNU/Linux, and not all architectures are recognized. If the auto-detect
14249 is unsuccessful the option has no effect.
14251 @item -mfpu=@var{name}
14253 This specifies what floating-point hardware (or hardware emulation) is
14254 available on the target. Permissible names are: @samp{vfp}, @samp{vfpv3},
14255 @samp{vfpv3-fp16}, @samp{vfpv3-d16}, @samp{vfpv3-d16-fp16}, @samp{vfpv3xd},
14256 @samp{vfpv3xd-fp16}, @samp{neon}, @samp{neon-fp16}, @samp{vfpv4},
14257 @samp{vfpv4-d16}, @samp{fpv4-sp-d16}, @samp{neon-vfpv4},
14258 @samp{fpv5-d16}, @samp{fpv5-sp-d16},
14259 @samp{fp-armv8}, @samp{neon-fp-armv8} and @samp{crypto-neon-fp-armv8}.
14261 If @option{-msoft-float} is specified this specifies the format of
14262 floating-point values.
14264 If the selected floating-point hardware includes the NEON extension
14265 (e.g. @option{-mfpu}=@samp{neon}), note that floating-point
14266 operations are not generated by GCC's auto-vectorization pass unless
14267 @option{-funsafe-math-optimizations} is also specified. This is
14268 because NEON hardware does not fully implement the IEEE 754 standard for
14269 floating-point arithmetic (in particular denormal values are treated as
14270 zero), so the use of NEON instructions may lead to a loss of precision.
14272 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}).
14274 @item -mfp16-format=@var{name}
14275 @opindex mfp16-format
14276 Specify the format of the @code{__fp16} half-precision floating-point type.
14277 Permissible names are @samp{none}, @samp{ieee}, and @samp{alternative};
14278 the default is @samp{none}, in which case the @code{__fp16} type is not
14279 defined. @xref{Half-Precision}, for more information.
14281 @item -mstructure-size-boundary=@var{n}
14282 @opindex mstructure-size-boundary
14283 The sizes of all structures and unions are rounded up to a multiple
14284 of the number of bits set by this option. Permissible values are 8, 32
14285 and 64. The default value varies for different toolchains. For the COFF
14286 targeted toolchain the default value is 8. A value of 64 is only allowed
14287 if the underlying ABI supports it.
14289 Specifying a larger number can produce faster, more efficient code, but
14290 can also increase the size of the program. Different values are potentially
14291 incompatible. Code compiled with one value cannot necessarily expect to
14292 work with code or libraries compiled with another value, if they exchange
14293 information using structures or unions.
14295 @item -mabort-on-noreturn
14296 @opindex mabort-on-noreturn
14297 Generate a call to the function @code{abort} at the end of a
14298 @code{noreturn} function. It is executed if the function tries to
14302 @itemx -mno-long-calls
14303 @opindex mlong-calls
14304 @opindex mno-long-calls
14305 Tells the compiler to perform function calls by first loading the
14306 address of the function into a register and then performing a subroutine
14307 call on this register. This switch is needed if the target function
14308 lies outside of the 64-megabyte addressing range of the offset-based
14309 version of subroutine call instruction.
14311 Even if this switch is enabled, not all function calls are turned
14312 into long calls. The heuristic is that static functions, functions
14313 that have the @code{short_call} attribute, functions that are inside
14314 the scope of a @code{#pragma no_long_calls} directive, and functions whose
14315 definitions have already been compiled within the current compilation
14316 unit are not turned into long calls. The exceptions to this rule are
14317 that weak function definitions, functions with the @code{long_call}
14318 attribute or the @code{section} attribute, and functions that are within
14319 the scope of a @code{#pragma long_calls} directive are always
14320 turned into long calls.
14322 This feature is not enabled by default. Specifying
14323 @option{-mno-long-calls} restores the default behavior, as does
14324 placing the function calls within the scope of a @code{#pragma
14325 long_calls_off} directive. Note these switches have no effect on how
14326 the compiler generates code to handle function calls via function
14329 @item -msingle-pic-base
14330 @opindex msingle-pic-base
14331 Treat the register used for PIC addressing as read-only, rather than
14332 loading it in the prologue for each function. The runtime system is
14333 responsible for initializing this register with an appropriate value
14334 before execution begins.
14336 @item -mpic-register=@var{reg}
14337 @opindex mpic-register
14338 Specify the register to be used for PIC addressing.
14339 For standard PIC base case, the default is any suitable register
14340 determined by compiler. For single PIC base case, the default is
14341 @samp{R9} if target is EABI based or stack-checking is enabled,
14342 otherwise the default is @samp{R10}.
14344 @item -mpic-data-is-text-relative
14345 @opindex mpic-data-is-text-relative
14346 Assume that each data segments are relative to text segment at load time.
14347 Therefore, it permits addressing data using PC-relative operations.
14348 This option is on by default for targets other than VxWorks RTP.
14350 @item -mpoke-function-name
14351 @opindex mpoke-function-name
14352 Write the name of each function into the text section, directly
14353 preceding the function prologue. The generated code is similar to this:
14357 .ascii "arm_poke_function_name", 0
14360 .word 0xff000000 + (t1 - t0)
14361 arm_poke_function_name
14363 stmfd sp!, @{fp, ip, lr, pc@}
14367 When performing a stack backtrace, code can inspect the value of
14368 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
14369 location @code{pc - 12} and the top 8 bits are set, then we know that
14370 there is a function name embedded immediately preceding this location
14371 and has length @code{((pc[-3]) & 0xff000000)}.
14378 Select between generating code that executes in ARM and Thumb
14379 states. The default for most configurations is to generate code
14380 that executes in ARM state, but the default can be changed by
14381 configuring GCC with the @option{--with-mode=}@var{state}
14384 You can also override the ARM and Thumb mode for each function
14385 by using the @code{target("thumb")} and @code{target("arm")} function attributes
14386 (@pxref{ARM Function Attributes}) or pragmas (@pxref{Function Specific Option Pragmas}).
14389 @opindex mtpcs-frame
14390 Generate a stack frame that is compliant with the Thumb Procedure Call
14391 Standard for all non-leaf functions. (A leaf function is one that does
14392 not call any other functions.) The default is @option{-mno-tpcs-frame}.
14394 @item -mtpcs-leaf-frame
14395 @opindex mtpcs-leaf-frame
14396 Generate a stack frame that is compliant with the Thumb Procedure Call
14397 Standard for all leaf functions. (A leaf function is one that does
14398 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
14400 @item -mcallee-super-interworking
14401 @opindex mcallee-super-interworking
14402 Gives all externally visible functions in the file being compiled an ARM
14403 instruction set header which switches to Thumb mode before executing the
14404 rest of the function. This allows these functions to be called from
14405 non-interworking code. This option is not valid in AAPCS configurations
14406 because interworking is enabled by default.
14408 @item -mcaller-super-interworking
14409 @opindex mcaller-super-interworking
14410 Allows calls via function pointers (including virtual functions) to
14411 execute correctly regardless of whether the target code has been
14412 compiled for interworking or not. There is a small overhead in the cost
14413 of executing a function pointer if this option is enabled. This option
14414 is not valid in AAPCS configurations because interworking is enabled
14417 @item -mtp=@var{name}
14419 Specify the access model for the thread local storage pointer. The valid
14420 models are @samp{soft}, which generates calls to @code{__aeabi_read_tp},
14421 @samp{cp15}, which fetches the thread pointer from @code{cp15} directly
14422 (supported in the arm6k architecture), and @samp{auto}, which uses the
14423 best available method for the selected processor. The default setting is
14426 @item -mtls-dialect=@var{dialect}
14427 @opindex mtls-dialect
14428 Specify the dialect to use for accessing thread local storage. Two
14429 @var{dialect}s are supported---@samp{gnu} and @samp{gnu2}. The
14430 @samp{gnu} dialect selects the original GNU scheme for supporting
14431 local and global dynamic TLS models. The @samp{gnu2} dialect
14432 selects the GNU descriptor scheme, which provides better performance
14433 for shared libraries. The GNU descriptor scheme is compatible with
14434 the original scheme, but does require new assembler, linker and
14435 library support. Initial and local exec TLS models are unaffected by
14436 this option and always use the original scheme.
14438 @item -mword-relocations
14439 @opindex mword-relocations
14440 Only generate absolute relocations on word-sized values (i.e. R_ARM_ABS32).
14441 This is enabled by default on targets (uClinux, SymbianOS) where the runtime
14442 loader imposes this restriction, and when @option{-fpic} or @option{-fPIC}
14445 @item -mfix-cortex-m3-ldrd
14446 @opindex mfix-cortex-m3-ldrd
14447 Some Cortex-M3 cores can cause data corruption when @code{ldrd} instructions
14448 with overlapping destination and base registers are used. This option avoids
14449 generating these instructions. This option is enabled by default when
14450 @option{-mcpu=cortex-m3} is specified.
14452 @item -munaligned-access
14453 @itemx -mno-unaligned-access
14454 @opindex munaligned-access
14455 @opindex mno-unaligned-access
14456 Enables (or disables) reading and writing of 16- and 32- bit values
14457 from addresses that are not 16- or 32- bit aligned. By default
14458 unaligned access is disabled for all pre-ARMv6 and all ARMv6-M
14459 architectures, and enabled for all other architectures. If unaligned
14460 access is not enabled then words in packed data structures are
14461 accessed a byte at a time.
14463 The ARM attribute @code{Tag_CPU_unaligned_access} is set in the
14464 generated object file to either true or false, depending upon the
14465 setting of this option. If unaligned access is enabled then the
14466 preprocessor symbol @code{__ARM_FEATURE_UNALIGNED} is also
14469 @item -mneon-for-64bits
14470 @opindex mneon-for-64bits
14471 Enables using Neon to handle scalar 64-bits operations. This is
14472 disabled by default since the cost of moving data from core registers
14475 @item -mslow-flash-data
14476 @opindex mslow-flash-data
14477 Assume loading data from flash is slower than fetching instruction.
14478 Therefore literal load is minimized for better performance.
14479 This option is only supported when compiling for ARMv7 M-profile and
14482 @item -masm-syntax-unified
14483 @opindex masm-syntax-unified
14484 Assume inline assembler is using unified asm syntax. The default is
14485 currently off which implies divided syntax. This option has no impact
14486 on Thumb2. However, this may change in future releases of GCC.
14487 Divided syntax should be considered deprecated.
14489 @item -mrestrict-it
14490 @opindex mrestrict-it
14491 Restricts generation of IT blocks to conform to the rules of ARMv8.
14492 IT blocks can only contain a single 16-bit instruction from a select
14493 set of instructions. This option is on by default for ARMv8 Thumb mode.
14495 @item -mprint-tune-info
14496 @opindex mprint-tune-info
14497 Print CPU tuning information as comment in assembler file. This is
14498 an option used only for regression testing of the compiler and not
14499 intended for ordinary use in compiling code. This option is disabled
14504 @subsection AVR Options
14505 @cindex AVR Options
14507 These options are defined for AVR implementations:
14510 @item -mmcu=@var{mcu}
14512 Specify Atmel AVR instruction set architectures (ISA) or MCU type.
14514 The default for this option is@tie{}@samp{avr2}.
14516 GCC supports the following AVR devices and ISAs:
14518 @include avr-mmcu.texi
14520 @item -maccumulate-args
14521 @opindex maccumulate-args
14522 Accumulate outgoing function arguments and acquire/release the needed
14523 stack space for outgoing function arguments once in function
14524 prologue/epilogue. Without this option, outgoing arguments are pushed
14525 before calling a function and popped afterwards.
14527 Popping the arguments after the function call can be expensive on
14528 AVR so that accumulating the stack space might lead to smaller
14529 executables because arguments need not to be removed from the
14530 stack after such a function call.
14532 This option can lead to reduced code size for functions that perform
14533 several calls to functions that get their arguments on the stack like
14534 calls to printf-like functions.
14536 @item -mbranch-cost=@var{cost}
14537 @opindex mbranch-cost
14538 Set the branch costs for conditional branch instructions to
14539 @var{cost}. Reasonable values for @var{cost} are small, non-negative
14540 integers. The default branch cost is 0.
14542 @item -mcall-prologues
14543 @opindex mcall-prologues
14544 Functions prologues/epilogues are expanded as calls to appropriate
14545 subroutines. Code size is smaller.
14549 Assume @code{int} to be 8-bit integer. This affects the sizes of all types: a
14550 @code{char} is 1 byte, an @code{int} is 1 byte, a @code{long} is 2 bytes,
14551 and @code{long long} is 4 bytes. Please note that this option does not
14552 conform to the C standards, but it results in smaller code
14555 @item -mn-flash=@var{num}
14557 Assume that the flash memory has a size of
14558 @var{num} times 64@tie{}KiB.
14560 @item -mno-interrupts
14561 @opindex mno-interrupts
14562 Generated code is not compatible with hardware interrupts.
14563 Code size is smaller.
14567 Try to replace @code{CALL} resp.@: @code{JMP} instruction by the shorter
14568 @code{RCALL} resp.@: @code{RJMP} instruction if applicable.
14569 Setting @option{-mrelax} just adds the @option{--mlink-relax} option to
14570 the assembler's command line and the @option{--relax} option to the
14571 linker's command line.
14573 Jump relaxing is performed by the linker because jump offsets are not
14574 known before code is located. Therefore, the assembler code generated by the
14575 compiler is the same, but the instructions in the executable may
14576 differ from instructions in the assembler code.
14578 Relaxing must be turned on if linker stubs are needed, see the
14579 section on @code{EIND} and linker stubs below.
14583 Assume that the device supports the Read-Modify-Write
14584 instructions @code{XCH}, @code{LAC}, @code{LAS} and @code{LAT}.
14588 Treat the stack pointer register as an 8-bit register,
14589 i.e.@: assume the high byte of the stack pointer is zero.
14590 In general, you don't need to set this option by hand.
14592 This option is used internally by the compiler to select and
14593 build multilibs for architectures @code{avr2} and @code{avr25}.
14594 These architectures mix devices with and without @code{SPH}.
14595 For any setting other than @option{-mmcu=avr2} or @option{-mmcu=avr25}
14596 the compiler driver adds or removes this option from the compiler
14597 proper's command line, because the compiler then knows if the device
14598 or architecture has an 8-bit stack pointer and thus no @code{SPH}
14603 Use address register @code{X} in a way proposed by the hardware. This means
14604 that @code{X} is only used in indirect, post-increment or
14605 pre-decrement addressing.
14607 Without this option, the @code{X} register may be used in the same way
14608 as @code{Y} or @code{Z} which then is emulated by additional
14610 For example, loading a value with @code{X+const} addressing with a
14611 small non-negative @code{const < 64} to a register @var{Rn} is
14615 adiw r26, const ; X += const
14616 ld @var{Rn}, X ; @var{Rn} = *X
14617 sbiw r26, const ; X -= const
14621 @opindex mtiny-stack
14622 Only change the lower 8@tie{}bits of the stack pointer.
14625 @opindex nodevicelib
14626 Don't link against AVR-LibC's device specific library @code{lib<mcu>.a}.
14628 @item -Waddr-space-convert
14629 @opindex Waddr-space-convert
14630 Warn about conversions between address spaces in the case where the
14631 resulting address space is not contained in the incoming address space.
14633 @item -Wmisspelled-isr
14634 @opindex Wmisspelled-isr
14635 Warn if the ISR is misspelled, i.e. without __vector prefix.
14636 Enabled by default.
14639 @subsubsection @code{EIND} and Devices with More Than 128 Ki Bytes of Flash
14640 @cindex @code{EIND}
14641 Pointers in the implementation are 16@tie{}bits wide.
14642 The address of a function or label is represented as word address so
14643 that indirect jumps and calls can target any code address in the
14644 range of 64@tie{}Ki words.
14646 In order to facilitate indirect jump on devices with more than 128@tie{}Ki
14647 bytes of program memory space, there is a special function register called
14648 @code{EIND} that serves as most significant part of the target address
14649 when @code{EICALL} or @code{EIJMP} instructions are used.
14651 Indirect jumps and calls on these devices are handled as follows by
14652 the compiler and are subject to some limitations:
14657 The compiler never sets @code{EIND}.
14660 The compiler uses @code{EIND} implicitly in @code{EICALL}/@code{EIJMP}
14661 instructions or might read @code{EIND} directly in order to emulate an
14662 indirect call/jump by means of a @code{RET} instruction.
14665 The compiler assumes that @code{EIND} never changes during the startup
14666 code or during the application. In particular, @code{EIND} is not
14667 saved/restored in function or interrupt service routine
14671 For indirect calls to functions and computed goto, the linker
14672 generates @emph{stubs}. Stubs are jump pads sometimes also called
14673 @emph{trampolines}. Thus, the indirect call/jump jumps to such a stub.
14674 The stub contains a direct jump to the desired address.
14677 Linker relaxation must be turned on so that the linker generates
14678 the stubs correctly in all situations. See the compiler option
14679 @option{-mrelax} and the linker option @option{--relax}.
14680 There are corner cases where the linker is supposed to generate stubs
14681 but aborts without relaxation and without a helpful error message.
14684 The default linker script is arranged for code with @code{EIND = 0}.
14685 If code is supposed to work for a setup with @code{EIND != 0}, a custom
14686 linker script has to be used in order to place the sections whose
14687 name start with @code{.trampolines} into the segment where @code{EIND}
14691 The startup code from libgcc never sets @code{EIND}.
14692 Notice that startup code is a blend of code from libgcc and AVR-LibC.
14693 For the impact of AVR-LibC on @code{EIND}, see the
14694 @w{@uref{http://nongnu.org/avr-libc/user-manual/,AVR-LibC user manual}}.
14697 It is legitimate for user-specific startup code to set up @code{EIND}
14698 early, for example by means of initialization code located in
14699 section @code{.init3}. Such code runs prior to general startup code
14700 that initializes RAM and calls constructors, but after the bit
14701 of startup code from AVR-LibC that sets @code{EIND} to the segment
14702 where the vector table is located.
14704 #include <avr/io.h>
14707 __attribute__((section(".init3"),naked,used,no_instrument_function))
14708 init3_set_eind (void)
14710 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t"
14711 "out %i0,r24" :: "n" (&EIND) : "r24","memory");
14716 The @code{__trampolines_start} symbol is defined in the linker script.
14719 Stubs are generated automatically by the linker if
14720 the following two conditions are met:
14723 @item The address of a label is taken by means of the @code{gs} modifier
14724 (short for @emph{generate stubs}) like so:
14726 LDI r24, lo8(gs(@var{func}))
14727 LDI r25, hi8(gs(@var{func}))
14729 @item The final location of that label is in a code segment
14730 @emph{outside} the segment where the stubs are located.
14734 The compiler emits such @code{gs} modifiers for code labels in the
14735 following situations:
14737 @item Taking address of a function or code label.
14738 @item Computed goto.
14739 @item If prologue-save function is used, see @option{-mcall-prologues}
14740 command-line option.
14741 @item Switch/case dispatch tables. If you do not want such dispatch
14742 tables you can specify the @option{-fno-jump-tables} command-line option.
14743 @item C and C++ constructors/destructors called during startup/shutdown.
14744 @item If the tools hit a @code{gs()} modifier explained above.
14748 Jumping to non-symbolic addresses like so is @emph{not} supported:
14753 /* Call function at word address 0x2 */
14754 return ((int(*)(void)) 0x2)();
14758 Instead, a stub has to be set up, i.e.@: the function has to be called
14759 through a symbol (@code{func_4} in the example):
14764 extern int func_4 (void);
14766 /* Call function at byte address 0x4 */
14771 and the application be linked with @option{-Wl,--defsym,func_4=0x4}.
14772 Alternatively, @code{func_4} can be defined in the linker script.
14775 @subsubsection Handling of the @code{RAMPD}, @code{RAMPX}, @code{RAMPY} and @code{RAMPZ} Special Function Registers
14776 @cindex @code{RAMPD}
14777 @cindex @code{RAMPX}
14778 @cindex @code{RAMPY}
14779 @cindex @code{RAMPZ}
14780 Some AVR devices support memories larger than the 64@tie{}KiB range
14781 that can be accessed with 16-bit pointers. To access memory locations
14782 outside this 64@tie{}KiB range, the contentent of a @code{RAMP}
14783 register is used as high part of the address:
14784 The @code{X}, @code{Y}, @code{Z} address register is concatenated
14785 with the @code{RAMPX}, @code{RAMPY}, @code{RAMPZ} special function
14786 register, respectively, to get a wide address. Similarly,
14787 @code{RAMPD} is used together with direct addressing.
14791 The startup code initializes the @code{RAMP} special function
14792 registers with zero.
14795 If a @ref{AVR Named Address Spaces,named address space} other than
14796 generic or @code{__flash} is used, then @code{RAMPZ} is set
14797 as needed before the operation.
14800 If the device supports RAM larger than 64@tie{}KiB and the compiler
14801 needs to change @code{RAMPZ} to accomplish an operation, @code{RAMPZ}
14802 is reset to zero after the operation.
14805 If the device comes with a specific @code{RAMP} register, the ISR
14806 prologue/epilogue saves/restores that SFR and initializes it with
14807 zero in case the ISR code might (implicitly) use it.
14810 RAM larger than 64@tie{}KiB is not supported by GCC for AVR targets.
14811 If you use inline assembler to read from locations outside the
14812 16-bit address range and change one of the @code{RAMP} registers,
14813 you must reset it to zero after the access.
14817 @subsubsection AVR Built-in Macros
14819 GCC defines several built-in macros so that the user code can test
14820 for the presence or absence of features. Almost any of the following
14821 built-in macros are deduced from device capabilities and thus
14822 triggered by the @option{-mmcu=} command-line option.
14824 For even more AVR-specific built-in macros see
14825 @ref{AVR Named Address Spaces} and @ref{AVR Built-in Functions}.
14830 Build-in macro that resolves to a decimal number that identifies the
14831 architecture and depends on the @option{-mmcu=@var{mcu}} option.
14832 Possible values are:
14834 @code{2}, @code{25}, @code{3}, @code{31}, @code{35},
14835 @code{4}, @code{5}, @code{51}, @code{6}
14837 for @var{mcu}=@code{avr2}, @code{avr25}, @code{avr3}, @code{avr31},
14838 @code{avr35}, @code{avr4}, @code{avr5}, @code{avr51}, @code{avr6},
14842 @code{100}, @code{102}, @code{104},
14843 @code{105}, @code{106}, @code{107}
14845 for @var{mcu}=@code{avrtiny}, @code{avrxmega2}, @code{avrxmega4},
14846 @code{avrxmega5}, @code{avrxmega6}, @code{avrxmega7}, respectively.
14847 If @var{mcu} specifies a device, this built-in macro is set
14848 accordingly. For example, with @option{-mmcu=atmega8} the macro is
14849 defined to @code{4}.
14851 @item __AVR_@var{Device}__
14852 Setting @option{-mmcu=@var{device}} defines this built-in macro which reflects
14853 the device's name. For example, @option{-mmcu=atmega8} defines the
14854 built-in macro @code{__AVR_ATmega8__}, @option{-mmcu=attiny261a} defines
14855 @code{__AVR_ATtiny261A__}, etc.
14857 The built-in macros' names follow
14858 the scheme @code{__AVR_@var{Device}__} where @var{Device} is
14859 the device name as from the AVR user manual. The difference between
14860 @var{Device} in the built-in macro and @var{device} in
14861 @option{-mmcu=@var{device}} is that the latter is always lowercase.
14863 If @var{device} is not a device but only a core architecture like
14864 @samp{avr51}, this macro is not defined.
14866 @item __AVR_DEVICE_NAME__
14867 Setting @option{-mmcu=@var{device}} defines this built-in macro to
14868 the device's name. For example, with @option{-mmcu=atmega8} the macro
14869 is defined to @code{atmega8}.
14871 If @var{device} is not a device but only a core architecture like
14872 @samp{avr51}, this macro is not defined.
14874 @item __AVR_XMEGA__
14875 The device / architecture belongs to the XMEGA family of devices.
14877 @item __AVR_HAVE_ELPM__
14878 The device has the @code{ELPM} instruction.
14880 @item __AVR_HAVE_ELPMX__
14881 The device has the @code{ELPM R@var{n},Z} and @code{ELPM
14882 R@var{n},Z+} instructions.
14884 @item __AVR_HAVE_MOVW__
14885 The device has the @code{MOVW} instruction to perform 16-bit
14886 register-register moves.
14888 @item __AVR_HAVE_LPMX__
14889 The device has the @code{LPM R@var{n},Z} and
14890 @code{LPM R@var{n},Z+} instructions.
14892 @item __AVR_HAVE_MUL__
14893 The device has a hardware multiplier.
14895 @item __AVR_HAVE_JMP_CALL__
14896 The device has the @code{JMP} and @code{CALL} instructions.
14897 This is the case for devices with at least 16@tie{}KiB of program
14900 @item __AVR_HAVE_EIJMP_EICALL__
14901 @itemx __AVR_3_BYTE_PC__
14902 The device has the @code{EIJMP} and @code{EICALL} instructions.
14903 This is the case for devices with more than 128@tie{}KiB of program memory.
14904 This also means that the program counter
14905 (PC) is 3@tie{}bytes wide.
14907 @item __AVR_2_BYTE_PC__
14908 The program counter (PC) is 2@tie{}bytes wide. This is the case for devices
14909 with up to 128@tie{}KiB of program memory.
14911 @item __AVR_HAVE_8BIT_SP__
14912 @itemx __AVR_HAVE_16BIT_SP__
14913 The stack pointer (SP) register is treated as 8-bit respectively
14914 16-bit register by the compiler.
14915 The definition of these macros is affected by @option{-mtiny-stack}.
14917 @item __AVR_HAVE_SPH__
14919 The device has the SPH (high part of stack pointer) special function
14920 register or has an 8-bit stack pointer, respectively.
14921 The definition of these macros is affected by @option{-mmcu=} and
14922 in the cases of @option{-mmcu=avr2} and @option{-mmcu=avr25} also
14925 @item __AVR_HAVE_RAMPD__
14926 @itemx __AVR_HAVE_RAMPX__
14927 @itemx __AVR_HAVE_RAMPY__
14928 @itemx __AVR_HAVE_RAMPZ__
14929 The device has the @code{RAMPD}, @code{RAMPX}, @code{RAMPY},
14930 @code{RAMPZ} special function register, respectively.
14932 @item __NO_INTERRUPTS__
14933 This macro reflects the @option{-mno-interrupts} command-line option.
14935 @item __AVR_ERRATA_SKIP__
14936 @itemx __AVR_ERRATA_SKIP_JMP_CALL__
14937 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit
14938 instructions because of a hardware erratum. Skip instructions are
14939 @code{SBRS}, @code{SBRC}, @code{SBIS}, @code{SBIC} and @code{CPSE}.
14940 The second macro is only defined if @code{__AVR_HAVE_JMP_CALL__} is also
14943 @item __AVR_ISA_RMW__
14944 The device has Read-Modify-Write instructions (XCH, LAC, LAS and LAT).
14946 @item __AVR_SFR_OFFSET__=@var{offset}
14947 Instructions that can address I/O special function registers directly
14948 like @code{IN}, @code{OUT}, @code{SBI}, etc.@: may use a different
14949 address as if addressed by an instruction to access RAM like @code{LD}
14950 or @code{STS}. This offset depends on the device architecture and has
14951 to be subtracted from the RAM address in order to get the
14952 respective I/O@tie{}address.
14954 @item __WITH_AVRLIBC__
14955 The compiler is configured to be used together with AVR-Libc.
14956 See the @option{--with-avrlibc} configure option.
14960 @node Blackfin Options
14961 @subsection Blackfin Options
14962 @cindex Blackfin Options
14965 @item -mcpu=@var{cpu}@r{[}-@var{sirevision}@r{]}
14967 Specifies the name of the target Blackfin processor. Currently, @var{cpu}
14968 can be one of @samp{bf512}, @samp{bf514}, @samp{bf516}, @samp{bf518},
14969 @samp{bf522}, @samp{bf523}, @samp{bf524}, @samp{bf525}, @samp{bf526},
14970 @samp{bf527}, @samp{bf531}, @samp{bf532}, @samp{bf533},
14971 @samp{bf534}, @samp{bf536}, @samp{bf537}, @samp{bf538}, @samp{bf539},
14972 @samp{bf542}, @samp{bf544}, @samp{bf547}, @samp{bf548}, @samp{bf549},
14973 @samp{bf542m}, @samp{bf544m}, @samp{bf547m}, @samp{bf548m}, @samp{bf549m},
14974 @samp{bf561}, @samp{bf592}.
14976 The optional @var{sirevision} specifies the silicon revision of the target
14977 Blackfin processor. Any workarounds available for the targeted silicon revision
14978 are enabled. If @var{sirevision} is @samp{none}, no workarounds are enabled.
14979 If @var{sirevision} is @samp{any}, all workarounds for the targeted processor
14980 are enabled. The @code{__SILICON_REVISION__} macro is defined to two
14981 hexadecimal digits representing the major and minor numbers in the silicon
14982 revision. If @var{sirevision} is @samp{none}, the @code{__SILICON_REVISION__}
14983 is not defined. If @var{sirevision} is @samp{any}, the
14984 @code{__SILICON_REVISION__} is defined to be @code{0xffff}.
14985 If this optional @var{sirevision} is not used, GCC assumes the latest known
14986 silicon revision of the targeted Blackfin processor.
14988 GCC defines a preprocessor macro for the specified @var{cpu}.
14989 For the @samp{bfin-elf} toolchain, this option causes the hardware BSP
14990 provided by libgloss to be linked in if @option{-msim} is not given.
14992 Without this option, @samp{bf532} is used as the processor by default.
14994 Note that support for @samp{bf561} is incomplete. For @samp{bf561},
14995 only the preprocessor macro is defined.
14999 Specifies that the program will be run on the simulator. This causes
15000 the simulator BSP provided by libgloss to be linked in. This option
15001 has effect only for @samp{bfin-elf} toolchain.
15002 Certain other options, such as @option{-mid-shared-library} and
15003 @option{-mfdpic}, imply @option{-msim}.
15005 @item -momit-leaf-frame-pointer
15006 @opindex momit-leaf-frame-pointer
15007 Don't keep the frame pointer in a register for leaf functions. This
15008 avoids the instructions to save, set up and restore frame pointers and
15009 makes an extra register available in leaf functions. The option
15010 @option{-fomit-frame-pointer} removes the frame pointer for all functions,
15011 which might make debugging harder.
15013 @item -mspecld-anomaly
15014 @opindex mspecld-anomaly
15015 When enabled, the compiler ensures that the generated code does not
15016 contain speculative loads after jump instructions. If this option is used,
15017 @code{__WORKAROUND_SPECULATIVE_LOADS} is defined.
15019 @item -mno-specld-anomaly
15020 @opindex mno-specld-anomaly
15021 Don't generate extra code to prevent speculative loads from occurring.
15023 @item -mcsync-anomaly
15024 @opindex mcsync-anomaly
15025 When enabled, the compiler ensures that the generated code does not
15026 contain CSYNC or SSYNC instructions too soon after conditional branches.
15027 If this option is used, @code{__WORKAROUND_SPECULATIVE_SYNCS} is defined.
15029 @item -mno-csync-anomaly
15030 @opindex mno-csync-anomaly
15031 Don't generate extra code to prevent CSYNC or SSYNC instructions from
15032 occurring too soon after a conditional branch.
15036 When enabled, the compiler is free to take advantage of the knowledge that
15037 the entire program fits into the low 64k of memory.
15040 @opindex mno-low-64k
15041 Assume that the program is arbitrarily large. This is the default.
15043 @item -mstack-check-l1
15044 @opindex mstack-check-l1
15045 Do stack checking using information placed into L1 scratchpad memory by the
15048 @item -mid-shared-library
15049 @opindex mid-shared-library
15050 Generate code that supports shared libraries via the library ID method.
15051 This allows for execute in place and shared libraries in an environment
15052 without virtual memory management. This option implies @option{-fPIC}.
15053 With a @samp{bfin-elf} target, this option implies @option{-msim}.
15055 @item -mno-id-shared-library
15056 @opindex mno-id-shared-library
15057 Generate code that doesn't assume ID-based shared libraries are being used.
15058 This is the default.
15060 @item -mleaf-id-shared-library
15061 @opindex mleaf-id-shared-library
15062 Generate code that supports shared libraries via the library ID method,
15063 but assumes that this library or executable won't link against any other
15064 ID shared libraries. That allows the compiler to use faster code for jumps
15067 @item -mno-leaf-id-shared-library
15068 @opindex mno-leaf-id-shared-library
15069 Do not assume that the code being compiled won't link against any ID shared
15070 libraries. Slower code is generated for jump and call insns.
15072 @item -mshared-library-id=n
15073 @opindex mshared-library-id
15074 Specifies the identification number of the ID-based shared library being
15075 compiled. Specifying a value of 0 generates more compact code; specifying
15076 other values forces the allocation of that number to the current
15077 library but is no more space- or time-efficient than omitting this option.
15081 Generate code that allows the data segment to be located in a different
15082 area of memory from the text segment. This allows for execute in place in
15083 an environment without virtual memory management by eliminating relocations
15084 against the text section.
15086 @item -mno-sep-data
15087 @opindex mno-sep-data
15088 Generate code that assumes that the data segment follows the text segment.
15089 This is the default.
15092 @itemx -mno-long-calls
15093 @opindex mlong-calls
15094 @opindex mno-long-calls
15095 Tells the compiler to perform function calls by first loading the
15096 address of the function into a register and then performing a subroutine
15097 call on this register. This switch is needed if the target function
15098 lies outside of the 24-bit addressing range of the offset-based
15099 version of subroutine call instruction.
15101 This feature is not enabled by default. Specifying
15102 @option{-mno-long-calls} restores the default behavior. Note these
15103 switches have no effect on how the compiler generates code to handle
15104 function calls via function pointers.
15108 Link with the fast floating-point library. This library relaxes some of
15109 the IEEE floating-point standard's rules for checking inputs against
15110 Not-a-Number (NAN), in the interest of performance.
15113 @opindex minline-plt
15114 Enable inlining of PLT entries in function calls to functions that are
15115 not known to bind locally. It has no effect without @option{-mfdpic}.
15118 @opindex mmulticore
15119 Build a standalone application for multicore Blackfin processors.
15120 This option causes proper start files and link scripts supporting
15121 multicore to be used, and defines the macro @code{__BFIN_MULTICORE}.
15122 It can only be used with @option{-mcpu=bf561@r{[}-@var{sirevision}@r{]}}.
15124 This option can be used with @option{-mcorea} or @option{-mcoreb}, which
15125 selects the one-application-per-core programming model. Without
15126 @option{-mcorea} or @option{-mcoreb}, the single-application/dual-core
15127 programming model is used. In this model, the main function of Core B
15128 should be named as @code{coreb_main}.
15130 If this option is not used, the single-core application programming
15135 Build a standalone application for Core A of BF561 when using
15136 the one-application-per-core programming model. Proper start files
15137 and link scripts are used to support Core A, and the macro
15138 @code{__BFIN_COREA} is defined.
15139 This option can only be used in conjunction with @option{-mmulticore}.
15143 Build a standalone application for Core B of BF561 when using
15144 the one-application-per-core programming model. Proper start files
15145 and link scripts are used to support Core B, and the macro
15146 @code{__BFIN_COREB} is defined. When this option is used, @code{coreb_main}
15147 should be used instead of @code{main}.
15148 This option can only be used in conjunction with @option{-mmulticore}.
15152 Build a standalone application for SDRAM. Proper start files and
15153 link scripts are used to put the application into SDRAM, and the macro
15154 @code{__BFIN_SDRAM} is defined.
15155 The loader should initialize SDRAM before loading the application.
15159 Assume that ICPLBs are enabled at run time. This has an effect on certain
15160 anomaly workarounds. For Linux targets, the default is to assume ICPLBs
15161 are enabled; for standalone applications the default is off.
15165 @subsection C6X Options
15166 @cindex C6X Options
15169 @item -march=@var{name}
15171 This specifies the name of the target architecture. GCC uses this
15172 name to determine what kind of instructions it can emit when generating
15173 assembly code. Permissible names are: @samp{c62x},
15174 @samp{c64x}, @samp{c64x+}, @samp{c67x}, @samp{c67x+}, @samp{c674x}.
15177 @opindex mbig-endian
15178 Generate code for a big-endian target.
15180 @item -mlittle-endian
15181 @opindex mlittle-endian
15182 Generate code for a little-endian target. This is the default.
15186 Choose startup files and linker script suitable for the simulator.
15188 @item -msdata=default
15189 @opindex msdata=default
15190 Put small global and static data in the @code{.neardata} section,
15191 which is pointed to by register @code{B14}. Put small uninitialized
15192 global and static data in the @code{.bss} section, which is adjacent
15193 to the @code{.neardata} section. Put small read-only data into the
15194 @code{.rodata} section. The corresponding sections used for large
15195 pieces of data are @code{.fardata}, @code{.far} and @code{.const}.
15198 @opindex msdata=all
15199 Put all data, not just small objects, into the sections reserved for
15200 small data, and use addressing relative to the @code{B14} register to
15204 @opindex msdata=none
15205 Make no use of the sections reserved for small data, and use absolute
15206 addresses to access all data. Put all initialized global and static
15207 data in the @code{.fardata} section, and all uninitialized data in the
15208 @code{.far} section. Put all constant data into the @code{.const}
15213 @subsection CRIS Options
15214 @cindex CRIS Options
15216 These options are defined specifically for the CRIS ports.
15219 @item -march=@var{architecture-type}
15220 @itemx -mcpu=@var{architecture-type}
15223 Generate code for the specified architecture. The choices for
15224 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
15225 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
15226 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
15229 @item -mtune=@var{architecture-type}
15231 Tune to @var{architecture-type} everything applicable about the generated
15232 code, except for the ABI and the set of available instructions. The
15233 choices for @var{architecture-type} are the same as for
15234 @option{-march=@var{architecture-type}}.
15236 @item -mmax-stack-frame=@var{n}
15237 @opindex mmax-stack-frame
15238 Warn when the stack frame of a function exceeds @var{n} bytes.
15244 The options @option{-metrax4} and @option{-metrax100} are synonyms for
15245 @option{-march=v3} and @option{-march=v8} respectively.
15247 @item -mmul-bug-workaround
15248 @itemx -mno-mul-bug-workaround
15249 @opindex mmul-bug-workaround
15250 @opindex mno-mul-bug-workaround
15251 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
15252 models where it applies. This option is active by default.
15256 Enable CRIS-specific verbose debug-related information in the assembly
15257 code. This option also has the effect of turning off the @samp{#NO_APP}
15258 formatted-code indicator to the assembler at the beginning of the
15263 Do not use condition-code results from previous instruction; always emit
15264 compare and test instructions before use of condition codes.
15266 @item -mno-side-effects
15267 @opindex mno-side-effects
15268 Do not emit instructions with side effects in addressing modes other than
15271 @item -mstack-align
15272 @itemx -mno-stack-align
15273 @itemx -mdata-align
15274 @itemx -mno-data-align
15275 @itemx -mconst-align
15276 @itemx -mno-const-align
15277 @opindex mstack-align
15278 @opindex mno-stack-align
15279 @opindex mdata-align
15280 @opindex mno-data-align
15281 @opindex mconst-align
15282 @opindex mno-const-align
15283 These options (@samp{no-} options) arrange (eliminate arrangements) for the
15284 stack frame, individual data and constants to be aligned for the maximum
15285 single data access size for the chosen CPU model. The default is to
15286 arrange for 32-bit alignment. ABI details such as structure layout are
15287 not affected by these options.
15295 Similar to the stack- data- and const-align options above, these options
15296 arrange for stack frame, writable data and constants to all be 32-bit,
15297 16-bit or 8-bit aligned. The default is 32-bit alignment.
15299 @item -mno-prologue-epilogue
15300 @itemx -mprologue-epilogue
15301 @opindex mno-prologue-epilogue
15302 @opindex mprologue-epilogue
15303 With @option{-mno-prologue-epilogue}, the normal function prologue and
15304 epilogue which set up the stack frame are omitted and no return
15305 instructions or return sequences are generated in the code. Use this
15306 option only together with visual inspection of the compiled code: no
15307 warnings or errors are generated when call-saved registers must be saved,
15308 or storage for local variables needs to be allocated.
15312 @opindex mno-gotplt
15314 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
15315 instruction sequences that load addresses for functions from the PLT part
15316 of the GOT rather than (traditional on other architectures) calls to the
15317 PLT@. The default is @option{-mgotplt}.
15321 Legacy no-op option only recognized with the cris-axis-elf and
15322 cris-axis-linux-gnu targets.
15326 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
15330 This option, recognized for the cris-axis-elf, arranges
15331 to link with input-output functions from a simulator library. Code,
15332 initialized data and zero-initialized data are allocated consecutively.
15336 Like @option{-sim}, but pass linker options to locate initialized data at
15337 0x40000000 and zero-initialized data at 0x80000000.
15341 @subsection CR16 Options
15342 @cindex CR16 Options
15344 These options are defined specifically for the CR16 ports.
15350 Enable the use of multiply-accumulate instructions. Disabled by default.
15354 @opindex mcr16cplus
15356 Generate code for CR16C or CR16C+ architecture. CR16C+ architecture
15361 Links the library libsim.a which is in compatible with simulator. Applicable
15362 to ELF compiler only.
15366 Choose integer type as 32-bit wide.
15370 Generates @code{sbit}/@code{cbit} instructions for bit manipulations.
15372 @item -mdata-model=@var{model}
15373 @opindex mdata-model
15374 Choose a data model. The choices for @var{model} are @samp{near},
15375 @samp{far} or @samp{medium}. @samp{medium} is default.
15376 However, @samp{far} is not valid with @option{-mcr16c}, as the
15377 CR16C architecture does not support the far data model.
15380 @node Darwin Options
15381 @subsection Darwin Options
15382 @cindex Darwin options
15384 These options are defined for all architectures running the Darwin operating
15387 FSF GCC on Darwin does not create ``fat'' object files; it creates
15388 an object file for the single architecture that GCC was built to
15389 target. Apple's GCC on Darwin does create ``fat'' files if multiple
15390 @option{-arch} options are used; it does so by running the compiler or
15391 linker multiple times and joining the results together with
15394 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
15395 @samp{i686}) is determined by the flags that specify the ISA
15396 that GCC is targeting, like @option{-mcpu} or @option{-march}. The
15397 @option{-force_cpusubtype_ALL} option can be used to override this.
15399 The Darwin tools vary in their behavior when presented with an ISA
15400 mismatch. The assembler, @file{as}, only permits instructions to
15401 be used that are valid for the subtype of the file it is generating,
15402 so you cannot put 64-bit instructions in a @samp{ppc750} object file.
15403 The linker for shared libraries, @file{/usr/bin/libtool}, fails
15404 and prints an error if asked to create a shared library with a less
15405 restrictive subtype than its input files (for instance, trying to put
15406 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
15407 for executables, @command{ld}, quietly gives the executable the most
15408 restrictive subtype of any of its input files.
15413 Add the framework directory @var{dir} to the head of the list of
15414 directories to be searched for header files. These directories are
15415 interleaved with those specified by @option{-I} options and are
15416 scanned in a left-to-right order.
15418 A framework directory is a directory with frameworks in it. A
15419 framework is a directory with a @file{Headers} and/or
15420 @file{PrivateHeaders} directory contained directly in it that ends
15421 in @file{.framework}. The name of a framework is the name of this
15422 directory excluding the @file{.framework}. Headers associated with
15423 the framework are found in one of those two directories, with
15424 @file{Headers} being searched first. A subframework is a framework
15425 directory that is in a framework's @file{Frameworks} directory.
15426 Includes of subframework headers can only appear in a header of a
15427 framework that contains the subframework, or in a sibling subframework
15428 header. Two subframeworks are siblings if they occur in the same
15429 framework. A subframework should not have the same name as a
15430 framework; a warning is issued if this is violated. Currently a
15431 subframework cannot have subframeworks; in the future, the mechanism
15432 may be extended to support this. The standard frameworks can be found
15433 in @file{/System/Library/Frameworks} and
15434 @file{/Library/Frameworks}. An example include looks like
15435 @code{#include <Framework/header.h>}, where @file{Framework} denotes
15436 the name of the framework and @file{header.h} is found in the
15437 @file{PrivateHeaders} or @file{Headers} directory.
15439 @item -iframework@var{dir}
15440 @opindex iframework
15441 Like @option{-F} except the directory is a treated as a system
15442 directory. The main difference between this @option{-iframework} and
15443 @option{-F} is that with @option{-iframework} the compiler does not
15444 warn about constructs contained within header files found via
15445 @var{dir}. This option is valid only for the C family of languages.
15449 Emit debugging information for symbols that are used. For stabs
15450 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
15451 This is by default ON@.
15455 Emit debugging information for all symbols and types.
15457 @item -mmacosx-version-min=@var{version}
15458 The earliest version of MacOS X that this executable will run on
15459 is @var{version}. Typical values of @var{version} include @code{10.1},
15460 @code{10.2}, and @code{10.3.9}.
15462 If the compiler was built to use the system's headers by default,
15463 then the default for this option is the system version on which the
15464 compiler is running, otherwise the default is to make choices that
15465 are compatible with as many systems and code bases as possible.
15469 Enable kernel development mode. The @option{-mkernel} option sets
15470 @option{-static}, @option{-fno-common}, @option{-fno-use-cxa-atexit},
15471 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
15472 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
15473 applicable. This mode also sets @option{-mno-altivec},
15474 @option{-msoft-float}, @option{-fno-builtin} and
15475 @option{-mlong-branch} for PowerPC targets.
15477 @item -mone-byte-bool
15478 @opindex mone-byte-bool
15479 Override the defaults for @code{bool} so that @code{sizeof(bool)==1}.
15480 By default @code{sizeof(bool)} is @code{4} when compiling for
15481 Darwin/PowerPC and @code{1} when compiling for Darwin/x86, so this
15482 option has no effect on x86.
15484 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
15485 to generate code that is not binary compatible with code generated
15486 without that switch. Using this switch may require recompiling all
15487 other modules in a program, including system libraries. Use this
15488 switch to conform to a non-default data model.
15490 @item -mfix-and-continue
15491 @itemx -ffix-and-continue
15492 @itemx -findirect-data
15493 @opindex mfix-and-continue
15494 @opindex ffix-and-continue
15495 @opindex findirect-data
15496 Generate code suitable for fast turnaround development, such as to
15497 allow GDB to dynamically load @file{.o} files into already-running
15498 programs. @option{-findirect-data} and @option{-ffix-and-continue}
15499 are provided for backwards compatibility.
15503 Loads all members of static archive libraries.
15504 See man ld(1) for more information.
15506 @item -arch_errors_fatal
15507 @opindex arch_errors_fatal
15508 Cause the errors having to do with files that have the wrong architecture
15511 @item -bind_at_load
15512 @opindex bind_at_load
15513 Causes the output file to be marked such that the dynamic linker will
15514 bind all undefined references when the file is loaded or launched.
15518 Produce a Mach-o bundle format file.
15519 See man ld(1) for more information.
15521 @item -bundle_loader @var{executable}
15522 @opindex bundle_loader
15523 This option specifies the @var{executable} that will load the build
15524 output file being linked. See man ld(1) for more information.
15527 @opindex dynamiclib
15528 When passed this option, GCC produces a dynamic library instead of
15529 an executable when linking, using the Darwin @file{libtool} command.
15531 @item -force_cpusubtype_ALL
15532 @opindex force_cpusubtype_ALL
15533 This causes GCC's output file to have the @samp{ALL} subtype, instead of
15534 one controlled by the @option{-mcpu} or @option{-march} option.
15536 @item -allowable_client @var{client_name}
15537 @itemx -client_name
15538 @itemx -compatibility_version
15539 @itemx -current_version
15541 @itemx -dependency-file
15543 @itemx -dylinker_install_name
15545 @itemx -exported_symbols_list
15548 @itemx -flat_namespace
15549 @itemx -force_flat_namespace
15550 @itemx -headerpad_max_install_names
15553 @itemx -install_name
15554 @itemx -keep_private_externs
15555 @itemx -multi_module
15556 @itemx -multiply_defined
15557 @itemx -multiply_defined_unused
15560 @itemx -no_dead_strip_inits_and_terms
15561 @itemx -nofixprebinding
15562 @itemx -nomultidefs
15564 @itemx -noseglinkedit
15565 @itemx -pagezero_size
15567 @itemx -prebind_all_twolevel_modules
15568 @itemx -private_bundle
15570 @itemx -read_only_relocs
15572 @itemx -sectobjectsymbols
15576 @itemx -sectobjectsymbols
15579 @itemx -segs_read_only_addr
15581 @itemx -segs_read_write_addr
15582 @itemx -seg_addr_table
15583 @itemx -seg_addr_table_filename
15584 @itemx -seglinkedit
15586 @itemx -segs_read_only_addr
15587 @itemx -segs_read_write_addr
15588 @itemx -single_module
15590 @itemx -sub_library
15592 @itemx -sub_umbrella
15593 @itemx -twolevel_namespace
15596 @itemx -unexported_symbols_list
15597 @itemx -weak_reference_mismatches
15598 @itemx -whatsloaded
15599 @opindex allowable_client
15600 @opindex client_name
15601 @opindex compatibility_version
15602 @opindex current_version
15603 @opindex dead_strip
15604 @opindex dependency-file
15605 @opindex dylib_file
15606 @opindex dylinker_install_name
15608 @opindex exported_symbols_list
15610 @opindex flat_namespace
15611 @opindex force_flat_namespace
15612 @opindex headerpad_max_install_names
15613 @opindex image_base
15615 @opindex install_name
15616 @opindex keep_private_externs
15617 @opindex multi_module
15618 @opindex multiply_defined
15619 @opindex multiply_defined_unused
15620 @opindex noall_load
15621 @opindex no_dead_strip_inits_and_terms
15622 @opindex nofixprebinding
15623 @opindex nomultidefs
15625 @opindex noseglinkedit
15626 @opindex pagezero_size
15628 @opindex prebind_all_twolevel_modules
15629 @opindex private_bundle
15630 @opindex read_only_relocs
15632 @opindex sectobjectsymbols
15635 @opindex sectcreate
15636 @opindex sectobjectsymbols
15639 @opindex segs_read_only_addr
15640 @opindex segs_read_write_addr
15641 @opindex seg_addr_table
15642 @opindex seg_addr_table_filename
15643 @opindex seglinkedit
15645 @opindex segs_read_only_addr
15646 @opindex segs_read_write_addr
15647 @opindex single_module
15649 @opindex sub_library
15650 @opindex sub_umbrella
15651 @opindex twolevel_namespace
15654 @opindex unexported_symbols_list
15655 @opindex weak_reference_mismatches
15656 @opindex whatsloaded
15657 These options are passed to the Darwin linker. The Darwin linker man page
15658 describes them in detail.
15661 @node DEC Alpha Options
15662 @subsection DEC Alpha Options
15664 These @samp{-m} options are defined for the DEC Alpha implementations:
15667 @item -mno-soft-float
15668 @itemx -msoft-float
15669 @opindex mno-soft-float
15670 @opindex msoft-float
15671 Use (do not use) the hardware floating-point instructions for
15672 floating-point operations. When @option{-msoft-float} is specified,
15673 functions in @file{libgcc.a} are used to perform floating-point
15674 operations. Unless they are replaced by routines that emulate the
15675 floating-point operations, or compiled in such a way as to call such
15676 emulations routines, these routines issue floating-point
15677 operations. If you are compiling for an Alpha without floating-point
15678 operations, you must ensure that the library is built so as not to call
15681 Note that Alpha implementations without floating-point operations are
15682 required to have floating-point registers.
15685 @itemx -mno-fp-regs
15687 @opindex mno-fp-regs
15688 Generate code that uses (does not use) the floating-point register set.
15689 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
15690 register set is not used, floating-point operands are passed in integer
15691 registers as if they were integers and floating-point results are passed
15692 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
15693 so any function with a floating-point argument or return value called by code
15694 compiled with @option{-mno-fp-regs} must also be compiled with that
15697 A typical use of this option is building a kernel that does not use,
15698 and hence need not save and restore, any floating-point registers.
15702 The Alpha architecture implements floating-point hardware optimized for
15703 maximum performance. It is mostly compliant with the IEEE floating-point
15704 standard. However, for full compliance, software assistance is
15705 required. This option generates code fully IEEE-compliant code
15706 @emph{except} that the @var{inexact-flag} is not maintained (see below).
15707 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
15708 defined during compilation. The resulting code is less efficient but is
15709 able to correctly support denormalized numbers and exceptional IEEE
15710 values such as not-a-number and plus/minus infinity. Other Alpha
15711 compilers call this option @option{-ieee_with_no_inexact}.
15713 @item -mieee-with-inexact
15714 @opindex mieee-with-inexact
15715 This is like @option{-mieee} except the generated code also maintains
15716 the IEEE @var{inexact-flag}. Turning on this option causes the
15717 generated code to implement fully-compliant IEEE math. In addition to
15718 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
15719 macro. On some Alpha implementations the resulting code may execute
15720 significantly slower than the code generated by default. Since there is
15721 very little code that depends on the @var{inexact-flag}, you should
15722 normally not specify this option. Other Alpha compilers call this
15723 option @option{-ieee_with_inexact}.
15725 @item -mfp-trap-mode=@var{trap-mode}
15726 @opindex mfp-trap-mode
15727 This option controls what floating-point related traps are enabled.
15728 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
15729 The trap mode can be set to one of four values:
15733 This is the default (normal) setting. The only traps that are enabled
15734 are the ones that cannot be disabled in software (e.g., division by zero
15738 In addition to the traps enabled by @samp{n}, underflow traps are enabled
15742 Like @samp{u}, but the instructions are marked to be safe for software
15743 completion (see Alpha architecture manual for details).
15746 Like @samp{su}, but inexact traps are enabled as well.
15749 @item -mfp-rounding-mode=@var{rounding-mode}
15750 @opindex mfp-rounding-mode
15751 Selects the IEEE rounding mode. Other Alpha compilers call this option
15752 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
15757 Normal IEEE rounding mode. Floating-point numbers are rounded towards
15758 the nearest machine number or towards the even machine number in case
15762 Round towards minus infinity.
15765 Chopped rounding mode. Floating-point numbers are rounded towards zero.
15768 Dynamic rounding mode. A field in the floating-point control register
15769 (@var{fpcr}, see Alpha architecture reference manual) controls the
15770 rounding mode in effect. The C library initializes this register for
15771 rounding towards plus infinity. Thus, unless your program modifies the
15772 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
15775 @item -mtrap-precision=@var{trap-precision}
15776 @opindex mtrap-precision
15777 In the Alpha architecture, floating-point traps are imprecise. This
15778 means without software assistance it is impossible to recover from a
15779 floating trap and program execution normally needs to be terminated.
15780 GCC can generate code that can assist operating system trap handlers
15781 in determining the exact location that caused a floating-point trap.
15782 Depending on the requirements of an application, different levels of
15783 precisions can be selected:
15787 Program precision. This option is the default and means a trap handler
15788 can only identify which program caused a floating-point exception.
15791 Function precision. The trap handler can determine the function that
15792 caused a floating-point exception.
15795 Instruction precision. The trap handler can determine the exact
15796 instruction that caused a floating-point exception.
15799 Other Alpha compilers provide the equivalent options called
15800 @option{-scope_safe} and @option{-resumption_safe}.
15802 @item -mieee-conformant
15803 @opindex mieee-conformant
15804 This option marks the generated code as IEEE conformant. You must not
15805 use this option unless you also specify @option{-mtrap-precision=i} and either
15806 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
15807 is to emit the line @samp{.eflag 48} in the function prologue of the
15808 generated assembly file.
15810 @item -mbuild-constants
15811 @opindex mbuild-constants
15812 Normally GCC examines a 32- or 64-bit integer constant to
15813 see if it can construct it from smaller constants in two or three
15814 instructions. If it cannot, it outputs the constant as a literal and
15815 generates code to load it from the data segment at run time.
15817 Use this option to require GCC to construct @emph{all} integer constants
15818 using code, even if it takes more instructions (the maximum is six).
15820 You typically use this option to build a shared library dynamic
15821 loader. Itself a shared library, it must relocate itself in memory
15822 before it can find the variables and constants in its own data segment.
15840 Indicate whether GCC should generate code to use the optional BWX,
15841 CIX, FIX and MAX instruction sets. The default is to use the instruction
15842 sets supported by the CPU type specified via @option{-mcpu=} option or that
15843 of the CPU on which GCC was built if none is specified.
15846 @itemx -mfloat-ieee
15847 @opindex mfloat-vax
15848 @opindex mfloat-ieee
15849 Generate code that uses (does not use) VAX F and G floating-point
15850 arithmetic instead of IEEE single and double precision.
15852 @item -mexplicit-relocs
15853 @itemx -mno-explicit-relocs
15854 @opindex mexplicit-relocs
15855 @opindex mno-explicit-relocs
15856 Older Alpha assemblers provided no way to generate symbol relocations
15857 except via assembler macros. Use of these macros does not allow
15858 optimal instruction scheduling. GNU binutils as of version 2.12
15859 supports a new syntax that allows the compiler to explicitly mark
15860 which relocations should apply to which instructions. This option
15861 is mostly useful for debugging, as GCC detects the capabilities of
15862 the assembler when it is built and sets the default accordingly.
15865 @itemx -mlarge-data
15866 @opindex msmall-data
15867 @opindex mlarge-data
15868 When @option{-mexplicit-relocs} is in effect, static data is
15869 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
15870 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
15871 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
15872 16-bit relocations off of the @code{$gp} register. This limits the
15873 size of the small data area to 64KB, but allows the variables to be
15874 directly accessed via a single instruction.
15876 The default is @option{-mlarge-data}. With this option the data area
15877 is limited to just below 2GB@. Programs that require more than 2GB of
15878 data must use @code{malloc} or @code{mmap} to allocate the data in the
15879 heap instead of in the program's data segment.
15881 When generating code for shared libraries, @option{-fpic} implies
15882 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
15885 @itemx -mlarge-text
15886 @opindex msmall-text
15887 @opindex mlarge-text
15888 When @option{-msmall-text} is used, the compiler assumes that the
15889 code of the entire program (or shared library) fits in 4MB, and is
15890 thus reachable with a branch instruction. When @option{-msmall-data}
15891 is used, the compiler can assume that all local symbols share the
15892 same @code{$gp} value, and thus reduce the number of instructions
15893 required for a function call from 4 to 1.
15895 The default is @option{-mlarge-text}.
15897 @item -mcpu=@var{cpu_type}
15899 Set the instruction set and instruction scheduling parameters for
15900 machine type @var{cpu_type}. You can specify either the @samp{EV}
15901 style name or the corresponding chip number. GCC supports scheduling
15902 parameters for the EV4, EV5 and EV6 family of processors and
15903 chooses the default values for the instruction set from the processor
15904 you specify. If you do not specify a processor type, GCC defaults
15905 to the processor on which the compiler was built.
15907 Supported values for @var{cpu_type} are
15913 Schedules as an EV4 and has no instruction set extensions.
15917 Schedules as an EV5 and has no instruction set extensions.
15921 Schedules as an EV5 and supports the BWX extension.
15926 Schedules as an EV5 and supports the BWX and MAX extensions.
15930 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
15934 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
15937 Native toolchains also support the value @samp{native},
15938 which selects the best architecture option for the host processor.
15939 @option{-mcpu=native} has no effect if GCC does not recognize
15942 @item -mtune=@var{cpu_type}
15944 Set only the instruction scheduling parameters for machine type
15945 @var{cpu_type}. The instruction set is not changed.
15947 Native toolchains also support the value @samp{native},
15948 which selects the best architecture option for the host processor.
15949 @option{-mtune=native} has no effect if GCC does not recognize
15952 @item -mmemory-latency=@var{time}
15953 @opindex mmemory-latency
15954 Sets the latency the scheduler should assume for typical memory
15955 references as seen by the application. This number is highly
15956 dependent on the memory access patterns used by the application
15957 and the size of the external cache on the machine.
15959 Valid options for @var{time} are
15963 A decimal number representing clock cycles.
15969 The compiler contains estimates of the number of clock cycles for
15970 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
15971 (also called Dcache, Scache, and Bcache), as well as to main memory.
15972 Note that L3 is only valid for EV5.
15978 @subsection FR30 Options
15979 @cindex FR30 Options
15981 These options are defined specifically for the FR30 port.
15985 @item -msmall-model
15986 @opindex msmall-model
15987 Use the small address space model. This can produce smaller code, but
15988 it does assume that all symbolic values and addresses fit into a
15993 Assume that runtime support has been provided and so there is no need
15994 to include the simulator library (@file{libsim.a}) on the linker
16000 @subsection FT32 Options
16001 @cindex FT32 Options
16003 These options are defined specifically for the FT32 port.
16009 Specifies that the program will be run on the simulator. This causes
16010 an alternate runtime startup and library to be linked.
16011 You must not use this option when generating programs that will run on
16012 real hardware; you must provide your own runtime library for whatever
16013 I/O functions are needed.
16017 Enable Local Register Allocation. This is still experimental for FT32,
16018 so by default the compiler uses standard reload.
16022 Do not use div and mod instructions.
16027 @subsection FRV Options
16028 @cindex FRV Options
16034 Only use the first 32 general-purpose registers.
16039 Use all 64 general-purpose registers.
16044 Use only the first 32 floating-point registers.
16049 Use all 64 floating-point registers.
16052 @opindex mhard-float
16054 Use hardware instructions for floating-point operations.
16057 @opindex msoft-float
16059 Use library routines for floating-point operations.
16064 Dynamically allocate condition code registers.
16069 Do not try to dynamically allocate condition code registers, only
16070 use @code{icc0} and @code{fcc0}.
16075 Change ABI to use double word insns.
16080 Do not use double word instructions.
16085 Use floating-point double instructions.
16088 @opindex mno-double
16090 Do not use floating-point double instructions.
16095 Use media instructions.
16100 Do not use media instructions.
16105 Use multiply and add/subtract instructions.
16108 @opindex mno-muladd
16110 Do not use multiply and add/subtract instructions.
16115 Select the FDPIC ABI, which uses function descriptors to represent
16116 pointers to functions. Without any PIC/PIE-related options, it
16117 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
16118 assumes GOT entries and small data are within a 12-bit range from the
16119 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
16120 are computed with 32 bits.
16121 With a @samp{bfin-elf} target, this option implies @option{-msim}.
16124 @opindex minline-plt
16126 Enable inlining of PLT entries in function calls to functions that are
16127 not known to bind locally. It has no effect without @option{-mfdpic}.
16128 It's enabled by default if optimizing for speed and compiling for
16129 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
16130 optimization option such as @option{-O3} or above is present in the
16136 Assume a large TLS segment when generating thread-local code.
16141 Do not assume a large TLS segment when generating thread-local code.
16146 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
16147 that is known to be in read-only sections. It's enabled by default,
16148 except for @option{-fpic} or @option{-fpie}: even though it may help
16149 make the global offset table smaller, it trades 1 instruction for 4.
16150 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
16151 one of which may be shared by multiple symbols, and it avoids the need
16152 for a GOT entry for the referenced symbol, so it's more likely to be a
16153 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
16155 @item -multilib-library-pic
16156 @opindex multilib-library-pic
16158 Link with the (library, not FD) pic libraries. It's implied by
16159 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
16160 @option{-fpic} without @option{-mfdpic}. You should never have to use
16164 @opindex mlinked-fp
16166 Follow the EABI requirement of always creating a frame pointer whenever
16167 a stack frame is allocated. This option is enabled by default and can
16168 be disabled with @option{-mno-linked-fp}.
16171 @opindex mlong-calls
16173 Use indirect addressing to call functions outside the current
16174 compilation unit. This allows the functions to be placed anywhere
16175 within the 32-bit address space.
16177 @item -malign-labels
16178 @opindex malign-labels
16180 Try to align labels to an 8-byte boundary by inserting NOPs into the
16181 previous packet. This option only has an effect when VLIW packing
16182 is enabled. It doesn't create new packets; it merely adds NOPs to
16185 @item -mlibrary-pic
16186 @opindex mlibrary-pic
16188 Generate position-independent EABI code.
16193 Use only the first four media accumulator registers.
16198 Use all eight media accumulator registers.
16203 Pack VLIW instructions.
16208 Do not pack VLIW instructions.
16211 @opindex mno-eflags
16213 Do not mark ABI switches in e_flags.
16216 @opindex mcond-move
16218 Enable the use of conditional-move instructions (default).
16220 This switch is mainly for debugging the compiler and will likely be removed
16221 in a future version.
16223 @item -mno-cond-move
16224 @opindex mno-cond-move
16226 Disable the use of conditional-move instructions.
16228 This switch is mainly for debugging the compiler and will likely be removed
16229 in a future version.
16234 Enable the use of conditional set instructions (default).
16236 This switch is mainly for debugging the compiler and will likely be removed
16237 in a future version.
16242 Disable the use of conditional set instructions.
16244 This switch is mainly for debugging the compiler and will likely be removed
16245 in a future version.
16248 @opindex mcond-exec
16250 Enable the use of conditional execution (default).
16252 This switch is mainly for debugging the compiler and will likely be removed
16253 in a future version.
16255 @item -mno-cond-exec
16256 @opindex mno-cond-exec
16258 Disable the use of conditional execution.
16260 This switch is mainly for debugging the compiler and will likely be removed
16261 in a future version.
16263 @item -mvliw-branch
16264 @opindex mvliw-branch
16266 Run a pass to pack branches into VLIW instructions (default).
16268 This switch is mainly for debugging the compiler and will likely be removed
16269 in a future version.
16271 @item -mno-vliw-branch
16272 @opindex mno-vliw-branch
16274 Do not run a pass to pack branches into VLIW instructions.
16276 This switch is mainly for debugging the compiler and will likely be removed
16277 in a future version.
16279 @item -mmulti-cond-exec
16280 @opindex mmulti-cond-exec
16282 Enable optimization of @code{&&} and @code{||} in conditional execution
16285 This switch is mainly for debugging the compiler and will likely be removed
16286 in a future version.
16288 @item -mno-multi-cond-exec
16289 @opindex mno-multi-cond-exec
16291 Disable optimization of @code{&&} and @code{||} in conditional execution.
16293 This switch is mainly for debugging the compiler and will likely be removed
16294 in a future version.
16296 @item -mnested-cond-exec
16297 @opindex mnested-cond-exec
16299 Enable nested conditional execution optimizations (default).
16301 This switch is mainly for debugging the compiler and will likely be removed
16302 in a future version.
16304 @item -mno-nested-cond-exec
16305 @opindex mno-nested-cond-exec
16307 Disable nested conditional execution optimizations.
16309 This switch is mainly for debugging the compiler and will likely be removed
16310 in a future version.
16312 @item -moptimize-membar
16313 @opindex moptimize-membar
16315 This switch removes redundant @code{membar} instructions from the
16316 compiler-generated code. It is enabled by default.
16318 @item -mno-optimize-membar
16319 @opindex mno-optimize-membar
16321 This switch disables the automatic removal of redundant @code{membar}
16322 instructions from the generated code.
16324 @item -mtomcat-stats
16325 @opindex mtomcat-stats
16327 Cause gas to print out tomcat statistics.
16329 @item -mcpu=@var{cpu}
16332 Select the processor type for which to generate code. Possible values are
16333 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
16334 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
16338 @node GNU/Linux Options
16339 @subsection GNU/Linux Options
16341 These @samp{-m} options are defined for GNU/Linux targets:
16346 Use the GNU C library. This is the default except
16347 on @samp{*-*-linux-*uclibc*}, @samp{*-*-linux-*musl*} and
16348 @samp{*-*-linux-*android*} targets.
16352 Use uClibc C library. This is the default on
16353 @samp{*-*-linux-*uclibc*} targets.
16357 Use the musl C library. This is the default on
16358 @samp{*-*-linux-*musl*} targets.
16362 Use Bionic C library. This is the default on
16363 @samp{*-*-linux-*android*} targets.
16367 Compile code compatible with Android platform. This is the default on
16368 @samp{*-*-linux-*android*} targets.
16370 When compiling, this option enables @option{-mbionic}, @option{-fPIC},
16371 @option{-fno-exceptions} and @option{-fno-rtti} by default. When linking,
16372 this option makes the GCC driver pass Android-specific options to the linker.
16373 Finally, this option causes the preprocessor macro @code{__ANDROID__}
16376 @item -tno-android-cc
16377 @opindex tno-android-cc
16378 Disable compilation effects of @option{-mandroid}, i.e., do not enable
16379 @option{-mbionic}, @option{-fPIC}, @option{-fno-exceptions} and
16380 @option{-fno-rtti} by default.
16382 @item -tno-android-ld
16383 @opindex tno-android-ld
16384 Disable linking effects of @option{-mandroid}, i.e., pass standard Linux
16385 linking options to the linker.
16389 @node H8/300 Options
16390 @subsection H8/300 Options
16392 These @samp{-m} options are defined for the H8/300 implementations:
16397 Shorten some address references at link time, when possible; uses the
16398 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
16399 ld, Using ld}, for a fuller description.
16403 Generate code for the H8/300H@.
16407 Generate code for the H8S@.
16411 Generate code for the H8S and H8/300H in the normal mode. This switch
16412 must be used either with @option{-mh} or @option{-ms}.
16416 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
16420 Extended registers are stored on stack before execution of function
16421 with monitor attribute. Default option is @option{-mexr}.
16422 This option is valid only for H8S targets.
16426 Extended registers are not stored on stack before execution of function
16427 with monitor attribute. Default option is @option{-mno-exr}.
16428 This option is valid only for H8S targets.
16432 Make @code{int} data 32 bits by default.
16435 @opindex malign-300
16436 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
16437 The default for the H8/300H and H8S is to align longs and floats on
16439 @option{-malign-300} causes them to be aligned on 2-byte boundaries.
16440 This option has no effect on the H8/300.
16444 @subsection HPPA Options
16445 @cindex HPPA Options
16447 These @samp{-m} options are defined for the HPPA family of computers:
16450 @item -march=@var{architecture-type}
16452 Generate code for the specified architecture. The choices for
16453 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
16454 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
16455 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
16456 architecture option for your machine. Code compiled for lower numbered
16457 architectures runs on higher numbered architectures, but not the
16460 @item -mpa-risc-1-0
16461 @itemx -mpa-risc-1-1
16462 @itemx -mpa-risc-2-0
16463 @opindex mpa-risc-1-0
16464 @opindex mpa-risc-1-1
16465 @opindex mpa-risc-2-0
16466 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
16468 @item -mjump-in-delay
16469 @opindex mjump-in-delay
16470 This option is ignored and provided for compatibility purposes only.
16472 @item -mdisable-fpregs
16473 @opindex mdisable-fpregs
16474 Prevent floating-point registers from being used in any manner. This is
16475 necessary for compiling kernels that perform lazy context switching of
16476 floating-point registers. If you use this option and attempt to perform
16477 floating-point operations, the compiler aborts.
16479 @item -mdisable-indexing
16480 @opindex mdisable-indexing
16481 Prevent the compiler from using indexing address modes. This avoids some
16482 rather obscure problems when compiling MIG generated code under MACH@.
16484 @item -mno-space-regs
16485 @opindex mno-space-regs
16486 Generate code that assumes the target has no space registers. This allows
16487 GCC to generate faster indirect calls and use unscaled index address modes.
16489 Such code is suitable for level 0 PA systems and kernels.
16491 @item -mfast-indirect-calls
16492 @opindex mfast-indirect-calls
16493 Generate code that assumes calls never cross space boundaries. This
16494 allows GCC to emit code that performs faster indirect calls.
16496 This option does not work in the presence of shared libraries or nested
16499 @item -mfixed-range=@var{register-range}
16500 @opindex mfixed-range
16501 Generate code treating the given register range as fixed registers.
16502 A fixed register is one that the register allocator cannot use. This is
16503 useful when compiling kernel code. A register range is specified as
16504 two registers separated by a dash. Multiple register ranges can be
16505 specified separated by a comma.
16507 @item -mlong-load-store
16508 @opindex mlong-load-store
16509 Generate 3-instruction load and store sequences as sometimes required by
16510 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
16513 @item -mportable-runtime
16514 @opindex mportable-runtime
16515 Use the portable calling conventions proposed by HP for ELF systems.
16519 Enable the use of assembler directives only GAS understands.
16521 @item -mschedule=@var{cpu-type}
16523 Schedule code according to the constraints for the machine type
16524 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
16525 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
16526 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
16527 proper scheduling option for your machine. The default scheduling is
16531 @opindex mlinker-opt
16532 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
16533 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
16534 linkers in which they give bogus error messages when linking some programs.
16537 @opindex msoft-float
16538 Generate output containing library calls for floating point.
16539 @strong{Warning:} the requisite libraries are not available for all HPPA
16540 targets. Normally the facilities of the machine's usual C compiler are
16541 used, but this cannot be done directly in cross-compilation. You must make
16542 your own arrangements to provide suitable library functions for
16545 @option{-msoft-float} changes the calling convention in the output file;
16546 therefore, it is only useful if you compile @emph{all} of a program with
16547 this option. In particular, you need to compile @file{libgcc.a}, the
16548 library that comes with GCC, with @option{-msoft-float} in order for
16553 Generate the predefine, @code{_SIO}, for server IO@. The default is
16554 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
16555 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
16556 options are available under HP-UX and HI-UX@.
16560 Use options specific to GNU @command{ld}.
16561 This passes @option{-shared} to @command{ld} when
16562 building a shared library. It is the default when GCC is configured,
16563 explicitly or implicitly, with the GNU linker. This option does not
16564 affect which @command{ld} is called; it only changes what parameters
16565 are passed to that @command{ld}.
16566 The @command{ld} that is called is determined by the
16567 @option{--with-ld} configure option, GCC's program search path, and
16568 finally by the user's @env{PATH}. The linker used by GCC can be printed
16569 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
16570 on the 64-bit HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16574 Use options specific to HP @command{ld}.
16575 This passes @option{-b} to @command{ld} when building
16576 a shared library and passes @option{+Accept TypeMismatch} to @command{ld} on all
16577 links. It is the default when GCC is configured, explicitly or
16578 implicitly, with the HP linker. This option does not affect
16579 which @command{ld} is called; it only changes what parameters are passed to that
16581 The @command{ld} that is called is determined by the @option{--with-ld}
16582 configure option, GCC's program search path, and finally by the user's
16583 @env{PATH}. The linker used by GCC can be printed using @samp{which
16584 `gcc -print-prog-name=ld`}. This option is only available on the 64-bit
16585 HP-UX GCC, i.e.@: configured with @samp{hppa*64*-*-hpux*}.
16588 @opindex mno-long-calls
16589 Generate code that uses long call sequences. This ensures that a call
16590 is always able to reach linker generated stubs. The default is to generate
16591 long calls only when the distance from the call site to the beginning
16592 of the function or translation unit, as the case may be, exceeds a
16593 predefined limit set by the branch type being used. The limits for
16594 normal calls are 7,600,000 and 240,000 bytes, respectively for the
16595 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
16598 Distances are measured from the beginning of functions when using the
16599 @option{-ffunction-sections} option, or when using the @option{-mgas}
16600 and @option{-mno-portable-runtime} options together under HP-UX with
16603 It is normally not desirable to use this option as it degrades
16604 performance. However, it may be useful in large applications,
16605 particularly when partial linking is used to build the application.
16607 The types of long calls used depends on the capabilities of the
16608 assembler and linker, and the type of code being generated. The
16609 impact on systems that support long absolute calls, and long pic
16610 symbol-difference or pc-relative calls should be relatively small.
16611 However, an indirect call is used on 32-bit ELF systems in pic code
16612 and it is quite long.
16614 @item -munix=@var{unix-std}
16616 Generate compiler predefines and select a startfile for the specified
16617 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
16618 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
16619 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
16620 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
16621 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
16624 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
16625 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
16626 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
16627 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
16628 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
16629 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
16631 It is @emph{important} to note that this option changes the interfaces
16632 for various library routines. It also affects the operational behavior
16633 of the C library. Thus, @emph{extreme} care is needed in using this
16636 Library code that is intended to operate with more than one UNIX
16637 standard must test, set and restore the variable @code{__xpg4_extended_mask}
16638 as appropriate. Most GNU software doesn't provide this capability.
16642 Suppress the generation of link options to search libdld.sl when the
16643 @option{-static} option is specified on HP-UX 10 and later.
16647 The HP-UX implementation of setlocale in libc has a dependency on
16648 libdld.sl. There isn't an archive version of libdld.sl. Thus,
16649 when the @option{-static} option is specified, special link options
16650 are needed to resolve this dependency.
16652 On HP-UX 10 and later, the GCC driver adds the necessary options to
16653 link with libdld.sl when the @option{-static} option is specified.
16654 This causes the resulting binary to be dynamic. On the 64-bit port,
16655 the linkers generate dynamic binaries by default in any case. The
16656 @option{-nolibdld} option can be used to prevent the GCC driver from
16657 adding these link options.
16661 Add support for multithreading with the @dfn{dce thread} library
16662 under HP-UX@. This option sets flags for both the preprocessor and
16666 @node IA-64 Options
16667 @subsection IA-64 Options
16668 @cindex IA-64 Options
16670 These are the @samp{-m} options defined for the Intel IA-64 architecture.
16674 @opindex mbig-endian
16675 Generate code for a big-endian target. This is the default for HP-UX@.
16677 @item -mlittle-endian
16678 @opindex mlittle-endian
16679 Generate code for a little-endian target. This is the default for AIX5
16685 @opindex mno-gnu-as
16686 Generate (or don't) code for the GNU assembler. This is the default.
16687 @c Also, this is the default if the configure option @option{--with-gnu-as}
16693 @opindex mno-gnu-ld
16694 Generate (or don't) code for the GNU linker. This is the default.
16695 @c Also, this is the default if the configure option @option{--with-gnu-ld}
16700 Generate code that does not use a global pointer register. The result
16701 is not position independent code, and violates the IA-64 ABI@.
16703 @item -mvolatile-asm-stop
16704 @itemx -mno-volatile-asm-stop
16705 @opindex mvolatile-asm-stop
16706 @opindex mno-volatile-asm-stop
16707 Generate (or don't) a stop bit immediately before and after volatile asm
16710 @item -mregister-names
16711 @itemx -mno-register-names
16712 @opindex mregister-names
16713 @opindex mno-register-names
16714 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
16715 the stacked registers. This may make assembler output more readable.
16721 Disable (or enable) optimizations that use the small data section. This may
16722 be useful for working around optimizer bugs.
16724 @item -mconstant-gp
16725 @opindex mconstant-gp
16726 Generate code that uses a single constant global pointer value. This is
16727 useful when compiling kernel code.
16731 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
16732 This is useful when compiling firmware code.
16734 @item -minline-float-divide-min-latency
16735 @opindex minline-float-divide-min-latency
16736 Generate code for inline divides of floating-point values
16737 using the minimum latency algorithm.
16739 @item -minline-float-divide-max-throughput
16740 @opindex minline-float-divide-max-throughput
16741 Generate code for inline divides of floating-point values
16742 using the maximum throughput algorithm.
16744 @item -mno-inline-float-divide
16745 @opindex mno-inline-float-divide
16746 Do not generate inline code for divides of floating-point values.
16748 @item -minline-int-divide-min-latency
16749 @opindex minline-int-divide-min-latency
16750 Generate code for inline divides of integer values
16751 using the minimum latency algorithm.
16753 @item -minline-int-divide-max-throughput
16754 @opindex minline-int-divide-max-throughput
16755 Generate code for inline divides of integer values
16756 using the maximum throughput algorithm.
16758 @item -mno-inline-int-divide
16759 @opindex mno-inline-int-divide
16760 Do not generate inline code for divides of integer values.
16762 @item -minline-sqrt-min-latency
16763 @opindex minline-sqrt-min-latency
16764 Generate code for inline square roots
16765 using the minimum latency algorithm.
16767 @item -minline-sqrt-max-throughput
16768 @opindex minline-sqrt-max-throughput
16769 Generate code for inline square roots
16770 using the maximum throughput algorithm.
16772 @item -mno-inline-sqrt
16773 @opindex mno-inline-sqrt
16774 Do not generate inline code for @code{sqrt}.
16777 @itemx -mno-fused-madd
16778 @opindex mfused-madd
16779 @opindex mno-fused-madd
16780 Do (don't) generate code that uses the fused multiply/add or multiply/subtract
16781 instructions. The default is to use these instructions.
16783 @item -mno-dwarf2-asm
16784 @itemx -mdwarf2-asm
16785 @opindex mno-dwarf2-asm
16786 @opindex mdwarf2-asm
16787 Don't (or do) generate assembler code for the DWARF line number debugging
16788 info. This may be useful when not using the GNU assembler.
16790 @item -mearly-stop-bits
16791 @itemx -mno-early-stop-bits
16792 @opindex mearly-stop-bits
16793 @opindex mno-early-stop-bits
16794 Allow stop bits to be placed earlier than immediately preceding the
16795 instruction that triggered the stop bit. This can improve instruction
16796 scheduling, but does not always do so.
16798 @item -mfixed-range=@var{register-range}
16799 @opindex mfixed-range
16800 Generate code treating the given register range as fixed registers.
16801 A fixed register is one that the register allocator cannot use. This is
16802 useful when compiling kernel code. A register range is specified as
16803 two registers separated by a dash. Multiple register ranges can be
16804 specified separated by a comma.
16806 @item -mtls-size=@var{tls-size}
16808 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
16811 @item -mtune=@var{cpu-type}
16813 Tune the instruction scheduling for a particular CPU, Valid values are
16814 @samp{itanium}, @samp{itanium1}, @samp{merced}, @samp{itanium2},
16815 and @samp{mckinley}.
16821 Generate code for a 32-bit or 64-bit environment.
16822 The 32-bit environment sets int, long and pointer to 32 bits.
16823 The 64-bit environment sets int to 32 bits and long and pointer
16824 to 64 bits. These are HP-UX specific flags.
16826 @item -mno-sched-br-data-spec
16827 @itemx -msched-br-data-spec
16828 @opindex mno-sched-br-data-spec
16829 @opindex msched-br-data-spec
16830 (Dis/En)able data speculative scheduling before reload.
16831 This results in generation of @code{ld.a} instructions and
16832 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16833 The default setting is disabled.
16835 @item -msched-ar-data-spec
16836 @itemx -mno-sched-ar-data-spec
16837 @opindex msched-ar-data-spec
16838 @opindex mno-sched-ar-data-spec
16839 (En/Dis)able data speculative scheduling after reload.
16840 This results in generation of @code{ld.a} instructions and
16841 the corresponding check instructions (@code{ld.c} / @code{chk.a}).
16842 The default setting is enabled.
16844 @item -mno-sched-control-spec
16845 @itemx -msched-control-spec
16846 @opindex mno-sched-control-spec
16847 @opindex msched-control-spec
16848 (Dis/En)able control speculative scheduling. This feature is
16849 available only during region scheduling (i.e.@: before reload).
16850 This results in generation of the @code{ld.s} instructions and
16851 the corresponding check instructions @code{chk.s}.
16852 The default setting is disabled.
16854 @item -msched-br-in-data-spec
16855 @itemx -mno-sched-br-in-data-spec
16856 @opindex msched-br-in-data-spec
16857 @opindex mno-sched-br-in-data-spec
16858 (En/Dis)able speculative scheduling of the instructions that
16859 are dependent on the data speculative loads before reload.
16860 This is effective only with @option{-msched-br-data-spec} enabled.
16861 The default setting is enabled.
16863 @item -msched-ar-in-data-spec
16864 @itemx -mno-sched-ar-in-data-spec
16865 @opindex msched-ar-in-data-spec
16866 @opindex mno-sched-ar-in-data-spec
16867 (En/Dis)able speculative scheduling of the instructions that
16868 are dependent on the data speculative loads after reload.
16869 This is effective only with @option{-msched-ar-data-spec} enabled.
16870 The default setting is enabled.
16872 @item -msched-in-control-spec
16873 @itemx -mno-sched-in-control-spec
16874 @opindex msched-in-control-spec
16875 @opindex mno-sched-in-control-spec
16876 (En/Dis)able speculative scheduling of the instructions that
16877 are dependent on the control speculative loads.
16878 This is effective only with @option{-msched-control-spec} enabled.
16879 The default setting is enabled.
16881 @item -mno-sched-prefer-non-data-spec-insns
16882 @itemx -msched-prefer-non-data-spec-insns
16883 @opindex mno-sched-prefer-non-data-spec-insns
16884 @opindex msched-prefer-non-data-spec-insns
16885 If enabled, data-speculative instructions are chosen for schedule
16886 only if there are no other choices at the moment. This makes
16887 the use of the data speculation much more conservative.
16888 The default setting is disabled.
16890 @item -mno-sched-prefer-non-control-spec-insns
16891 @itemx -msched-prefer-non-control-spec-insns
16892 @opindex mno-sched-prefer-non-control-spec-insns
16893 @opindex msched-prefer-non-control-spec-insns
16894 If enabled, control-speculative instructions are chosen for schedule
16895 only if there are no other choices at the moment. This makes
16896 the use of the control speculation much more conservative.
16897 The default setting is disabled.
16899 @item -mno-sched-count-spec-in-critical-path
16900 @itemx -msched-count-spec-in-critical-path
16901 @opindex mno-sched-count-spec-in-critical-path
16902 @opindex msched-count-spec-in-critical-path
16903 If enabled, speculative dependencies are considered during
16904 computation of the instructions priorities. This makes the use of the
16905 speculation a bit more conservative.
16906 The default setting is disabled.
16908 @item -msched-spec-ldc
16909 @opindex msched-spec-ldc
16910 Use a simple data speculation check. This option is on by default.
16912 @item -msched-control-spec-ldc
16913 @opindex msched-spec-ldc
16914 Use a simple check for control speculation. This option is on by default.
16916 @item -msched-stop-bits-after-every-cycle
16917 @opindex msched-stop-bits-after-every-cycle
16918 Place a stop bit after every cycle when scheduling. This option is on
16921 @item -msched-fp-mem-deps-zero-cost
16922 @opindex msched-fp-mem-deps-zero-cost
16923 Assume that floating-point stores and loads are not likely to cause a conflict
16924 when placed into the same instruction group. This option is disabled by
16927 @item -msel-sched-dont-check-control-spec
16928 @opindex msel-sched-dont-check-control-spec
16929 Generate checks for control speculation in selective scheduling.
16930 This flag is disabled by default.
16932 @item -msched-max-memory-insns=@var{max-insns}
16933 @opindex msched-max-memory-insns
16934 Limit on the number of memory insns per instruction group, giving lower
16935 priority to subsequent memory insns attempting to schedule in the same
16936 instruction group. Frequently useful to prevent cache bank conflicts.
16937 The default value is 1.
16939 @item -msched-max-memory-insns-hard-limit
16940 @opindex msched-max-memory-insns-hard-limit
16941 Makes the limit specified by @option{msched-max-memory-insns} a hard limit,
16942 disallowing more than that number in an instruction group.
16943 Otherwise, the limit is ``soft'', meaning that non-memory operations
16944 are preferred when the limit is reached, but memory operations may still
16950 @subsection LM32 Options
16951 @cindex LM32 options
16953 These @option{-m} options are defined for the LatticeMico32 architecture:
16956 @item -mbarrel-shift-enabled
16957 @opindex mbarrel-shift-enabled
16958 Enable barrel-shift instructions.
16960 @item -mdivide-enabled
16961 @opindex mdivide-enabled
16962 Enable divide and modulus instructions.
16964 @item -mmultiply-enabled
16965 @opindex multiply-enabled
16966 Enable multiply instructions.
16968 @item -msign-extend-enabled
16969 @opindex msign-extend-enabled
16970 Enable sign extend instructions.
16972 @item -muser-enabled
16973 @opindex muser-enabled
16974 Enable user-defined instructions.
16979 @subsection M32C Options
16980 @cindex M32C options
16983 @item -mcpu=@var{name}
16985 Select the CPU for which code is generated. @var{name} may be one of
16986 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
16987 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
16988 the M32C/80 series.
16992 Specifies that the program will be run on the simulator. This causes
16993 an alternate runtime library to be linked in which supports, for
16994 example, file I/O@. You must not use this option when generating
16995 programs that will run on real hardware; you must provide your own
16996 runtime library for whatever I/O functions are needed.
16998 @item -memregs=@var{number}
17000 Specifies the number of memory-based pseudo-registers GCC uses
17001 during code generation. These pseudo-registers are used like real
17002 registers, so there is a tradeoff between GCC's ability to fit the
17003 code into available registers, and the performance penalty of using
17004 memory instead of registers. Note that all modules in a program must
17005 be compiled with the same value for this option. Because of that, you
17006 must not use this option with GCC's default runtime libraries.
17010 @node M32R/D Options
17011 @subsection M32R/D Options
17012 @cindex M32R/D options
17014 These @option{-m} options are defined for Renesas M32R/D architectures:
17019 Generate code for the M32R/2@.
17023 Generate code for the M32R/X@.
17027 Generate code for the M32R@. This is the default.
17029 @item -mmodel=small
17030 @opindex mmodel=small
17031 Assume all objects live in the lower 16MB of memory (so that their addresses
17032 can be loaded with the @code{ld24} instruction), and assume all subroutines
17033 are reachable with the @code{bl} instruction.
17034 This is the default.
17036 The addressability of a particular object can be set with the
17037 @code{model} attribute.
17039 @item -mmodel=medium
17040 @opindex mmodel=medium
17041 Assume objects may be anywhere in the 32-bit address space (the compiler
17042 generates @code{seth/add3} instructions to load their addresses), and
17043 assume all subroutines are reachable with the @code{bl} instruction.
17045 @item -mmodel=large
17046 @opindex mmodel=large
17047 Assume objects may be anywhere in the 32-bit address space (the compiler
17048 generates @code{seth/add3} instructions to load their addresses), and
17049 assume subroutines may not be reachable with the @code{bl} instruction
17050 (the compiler generates the much slower @code{seth/add3/jl}
17051 instruction sequence).
17054 @opindex msdata=none
17055 Disable use of the small data area. Variables are put into
17056 one of @code{.data}, @code{.bss}, or @code{.rodata} (unless the
17057 @code{section} attribute has been specified).
17058 This is the default.
17060 The small data area consists of sections @code{.sdata} and @code{.sbss}.
17061 Objects may be explicitly put in the small data area with the
17062 @code{section} attribute using one of these sections.
17064 @item -msdata=sdata
17065 @opindex msdata=sdata
17066 Put small global and static data in the small data area, but do not
17067 generate special code to reference them.
17070 @opindex msdata=use
17071 Put small global and static data in the small data area, and generate
17072 special instructions to reference them.
17076 @cindex smaller data references
17077 Put global and static objects less than or equal to @var{num} bytes
17078 into the small data or BSS sections instead of the normal data or BSS
17079 sections. The default value of @var{num} is 8.
17080 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
17081 for this option to have any effect.
17083 All modules should be compiled with the same @option{-G @var{num}} value.
17084 Compiling with different values of @var{num} may or may not work; if it
17085 doesn't the linker gives an error message---incorrect code is not
17090 Makes the M32R-specific code in the compiler display some statistics
17091 that might help in debugging programs.
17093 @item -malign-loops
17094 @opindex malign-loops
17095 Align all loops to a 32-byte boundary.
17097 @item -mno-align-loops
17098 @opindex mno-align-loops
17099 Do not enforce a 32-byte alignment for loops. This is the default.
17101 @item -missue-rate=@var{number}
17102 @opindex missue-rate=@var{number}
17103 Issue @var{number} instructions per cycle. @var{number} can only be 1
17106 @item -mbranch-cost=@var{number}
17107 @opindex mbranch-cost=@var{number}
17108 @var{number} can only be 1 or 2. If it is 1 then branches are
17109 preferred over conditional code, if it is 2, then the opposite applies.
17111 @item -mflush-trap=@var{number}
17112 @opindex mflush-trap=@var{number}
17113 Specifies the trap number to use to flush the cache. The default is
17114 12. Valid numbers are between 0 and 15 inclusive.
17116 @item -mno-flush-trap
17117 @opindex mno-flush-trap
17118 Specifies that the cache cannot be flushed by using a trap.
17120 @item -mflush-func=@var{name}
17121 @opindex mflush-func=@var{name}
17122 Specifies the name of the operating system function to call to flush
17123 the cache. The default is @samp{_flush_cache}, but a function call
17124 is only used if a trap is not available.
17126 @item -mno-flush-func
17127 @opindex mno-flush-func
17128 Indicates that there is no OS function for flushing the cache.
17132 @node M680x0 Options
17133 @subsection M680x0 Options
17134 @cindex M680x0 options
17136 These are the @samp{-m} options defined for M680x0 and ColdFire processors.
17137 The default settings depend on which architecture was selected when
17138 the compiler was configured; the defaults for the most common choices
17142 @item -march=@var{arch}
17144 Generate code for a specific M680x0 or ColdFire instruction set
17145 architecture. Permissible values of @var{arch} for M680x0
17146 architectures are: @samp{68000}, @samp{68010}, @samp{68020},
17147 @samp{68030}, @samp{68040}, @samp{68060} and @samp{cpu32}. ColdFire
17148 architectures are selected according to Freescale's ISA classification
17149 and the permissible values are: @samp{isaa}, @samp{isaaplus},
17150 @samp{isab} and @samp{isac}.
17152 GCC defines a macro @code{__mcf@var{arch}__} whenever it is generating
17153 code for a ColdFire target. The @var{arch} in this macro is one of the
17154 @option{-march} arguments given above.
17156 When used together, @option{-march} and @option{-mtune} select code
17157 that runs on a family of similar processors but that is optimized
17158 for a particular microarchitecture.
17160 @item -mcpu=@var{cpu}
17162 Generate code for a specific M680x0 or ColdFire processor.
17163 The M680x0 @var{cpu}s are: @samp{68000}, @samp{68010}, @samp{68020},
17164 @samp{68030}, @samp{68040}, @samp{68060}, @samp{68302}, @samp{68332}
17165 and @samp{cpu32}. The ColdFire @var{cpu}s are given by the table
17166 below, which also classifies the CPUs into families:
17168 @multitable @columnfractions 0.20 0.80
17169 @item @strong{Family} @tab @strong{@samp{-mcpu} arguments}
17170 @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}
17171 @item @samp{5206} @tab @samp{5202} @samp{5204} @samp{5206}
17172 @item @samp{5206e} @tab @samp{5206e}
17173 @item @samp{5208} @tab @samp{5207} @samp{5208}
17174 @item @samp{5211a} @tab @samp{5210a} @samp{5211a}
17175 @item @samp{5213} @tab @samp{5211} @samp{5212} @samp{5213}
17176 @item @samp{5216} @tab @samp{5214} @samp{5216}
17177 @item @samp{52235} @tab @samp{52230} @samp{52231} @samp{52232} @samp{52233} @samp{52234} @samp{52235}
17178 @item @samp{5225} @tab @samp{5224} @samp{5225}
17179 @item @samp{52259} @tab @samp{52252} @samp{52254} @samp{52255} @samp{52256} @samp{52258} @samp{52259}
17180 @item @samp{5235} @tab @samp{5232} @samp{5233} @samp{5234} @samp{5235} @samp{523x}
17181 @item @samp{5249} @tab @samp{5249}
17182 @item @samp{5250} @tab @samp{5250}
17183 @item @samp{5271} @tab @samp{5270} @samp{5271}
17184 @item @samp{5272} @tab @samp{5272}
17185 @item @samp{5275} @tab @samp{5274} @samp{5275}
17186 @item @samp{5282} @tab @samp{5280} @samp{5281} @samp{5282} @samp{528x}
17187 @item @samp{53017} @tab @samp{53011} @samp{53012} @samp{53013} @samp{53014} @samp{53015} @samp{53016} @samp{53017}
17188 @item @samp{5307} @tab @samp{5307}
17189 @item @samp{5329} @tab @samp{5327} @samp{5328} @samp{5329} @samp{532x}
17190 @item @samp{5373} @tab @samp{5372} @samp{5373} @samp{537x}
17191 @item @samp{5407} @tab @samp{5407}
17192 @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}
17195 @option{-mcpu=@var{cpu}} overrides @option{-march=@var{arch}} if
17196 @var{arch} is compatible with @var{cpu}. Other combinations of
17197 @option{-mcpu} and @option{-march} are rejected.
17199 GCC defines the macro @code{__mcf_cpu_@var{cpu}} when ColdFire target
17200 @var{cpu} is selected. It also defines @code{__mcf_family_@var{family}},
17201 where the value of @var{family} is given by the table above.
17203 @item -mtune=@var{tune}
17205 Tune the code for a particular microarchitecture within the
17206 constraints set by @option{-march} and @option{-mcpu}.
17207 The M680x0 microarchitectures are: @samp{68000}, @samp{68010},
17208 @samp{68020}, @samp{68030}, @samp{68040}, @samp{68060}
17209 and @samp{cpu32}. The ColdFire microarchitectures
17210 are: @samp{cfv1}, @samp{cfv2}, @samp{cfv3}, @samp{cfv4} and @samp{cfv4e}.
17212 You can also use @option{-mtune=68020-40} for code that needs
17213 to run relatively well on 68020, 68030 and 68040 targets.
17214 @option{-mtune=68020-60} is similar but includes 68060 targets
17215 as well. These two options select the same tuning decisions as
17216 @option{-m68020-40} and @option{-m68020-60} respectively.
17218 GCC defines the macros @code{__mc@var{arch}} and @code{__mc@var{arch}__}
17219 when tuning for 680x0 architecture @var{arch}. It also defines
17220 @code{mc@var{arch}} unless either @option{-ansi} or a non-GNU @option{-std}
17221 option is used. If GCC is tuning for a range of architectures,
17222 as selected by @option{-mtune=68020-40} or @option{-mtune=68020-60},
17223 it defines the macros for every architecture in the range.
17225 GCC also defines the macro @code{__m@var{uarch}__} when tuning for
17226 ColdFire microarchitecture @var{uarch}, where @var{uarch} is one
17227 of the arguments given above.
17233 Generate output for a 68000. This is the default
17234 when the compiler is configured for 68000-based systems.
17235 It is equivalent to @option{-march=68000}.
17237 Use this option for microcontrollers with a 68000 or EC000 core,
17238 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
17242 Generate output for a 68010. This is the default
17243 when the compiler is configured for 68010-based systems.
17244 It is equivalent to @option{-march=68010}.
17250 Generate output for a 68020. This is the default
17251 when the compiler is configured for 68020-based systems.
17252 It is equivalent to @option{-march=68020}.
17256 Generate output for a 68030. This is the default when the compiler is
17257 configured for 68030-based systems. It is equivalent to
17258 @option{-march=68030}.
17262 Generate output for a 68040. This is the default when the compiler is
17263 configured for 68040-based systems. It is equivalent to
17264 @option{-march=68040}.
17266 This option inhibits the use of 68881/68882 instructions that have to be
17267 emulated by software on the 68040. Use this option if your 68040 does not
17268 have code to emulate those instructions.
17272 Generate output for a 68060. This is the default when the compiler is
17273 configured for 68060-based systems. It is equivalent to
17274 @option{-march=68060}.
17276 This option inhibits the use of 68020 and 68881/68882 instructions that
17277 have to be emulated by software on the 68060. Use this option if your 68060
17278 does not have code to emulate those instructions.
17282 Generate output for a CPU32. This is the default
17283 when the compiler is configured for CPU32-based systems.
17284 It is equivalent to @option{-march=cpu32}.
17286 Use this option for microcontrollers with a
17287 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
17288 68336, 68340, 68341, 68349 and 68360.
17292 Generate output for a 520X ColdFire CPU@. This is the default
17293 when the compiler is configured for 520X-based systems.
17294 It is equivalent to @option{-mcpu=5206}, and is now deprecated
17295 in favor of that option.
17297 Use this option for microcontroller with a 5200 core, including
17298 the MCF5202, MCF5203, MCF5204 and MCF5206.
17302 Generate output for a 5206e ColdFire CPU@. The option is now
17303 deprecated in favor of the equivalent @option{-mcpu=5206e}.
17307 Generate output for a member of the ColdFire 528X family.
17308 The option is now deprecated in favor of the equivalent
17309 @option{-mcpu=528x}.
17313 Generate output for a ColdFire 5307 CPU@. The option is now deprecated
17314 in favor of the equivalent @option{-mcpu=5307}.
17318 Generate output for a ColdFire 5407 CPU@. The option is now deprecated
17319 in favor of the equivalent @option{-mcpu=5407}.
17323 Generate output for a ColdFire V4e family CPU (e.g.@: 547x/548x).
17324 This includes use of hardware floating-point instructions.
17325 The option is equivalent to @option{-mcpu=547x}, and is now
17326 deprecated in favor of that option.
17330 Generate output for a 68040, without using any of the new instructions.
17331 This results in code that can run relatively efficiently on either a
17332 68020/68881 or a 68030 or a 68040. The generated code does use the
17333 68881 instructions that are emulated on the 68040.
17335 The option is equivalent to @option{-march=68020} @option{-mtune=68020-40}.
17339 Generate output for a 68060, without using any of the new instructions.
17340 This results in code that can run relatively efficiently on either a
17341 68020/68881 or a 68030 or a 68040. The generated code does use the
17342 68881 instructions that are emulated on the 68060.
17344 The option is equivalent to @option{-march=68020} @option{-mtune=68020-60}.
17348 @opindex mhard-float
17350 Generate floating-point instructions. This is the default for 68020
17351 and above, and for ColdFire devices that have an FPU@. It defines the
17352 macro @code{__HAVE_68881__} on M680x0 targets and @code{__mcffpu__}
17353 on ColdFire targets.
17356 @opindex msoft-float
17357 Do not generate floating-point instructions; use library calls instead.
17358 This is the default for 68000, 68010, and 68832 targets. It is also
17359 the default for ColdFire devices that have no FPU.
17365 Generate (do not generate) ColdFire hardware divide and remainder
17366 instructions. If @option{-march} is used without @option{-mcpu},
17367 the default is ``on'' for ColdFire architectures and ``off'' for M680x0
17368 architectures. Otherwise, the default is taken from the target CPU
17369 (either the default CPU, or the one specified by @option{-mcpu}). For
17370 example, the default is ``off'' for @option{-mcpu=5206} and ``on'' for
17371 @option{-mcpu=5206e}.
17373 GCC defines the macro @code{__mcfhwdiv__} when this option is enabled.
17377 Consider type @code{int} to be 16 bits wide, like @code{short int}.
17378 Additionally, parameters passed on the stack are also aligned to a
17379 16-bit boundary even on targets whose API mandates promotion to 32-bit.
17383 Do not consider type @code{int} to be 16 bits wide. This is the default.
17386 @itemx -mno-bitfield
17387 @opindex mnobitfield
17388 @opindex mno-bitfield
17389 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
17390 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
17394 Do use the bit-field instructions. The @option{-m68020} option implies
17395 @option{-mbitfield}. This is the default if you use a configuration
17396 designed for a 68020.
17400 Use a different function-calling convention, in which functions
17401 that take a fixed number of arguments return with the @code{rtd}
17402 instruction, which pops their arguments while returning. This
17403 saves one instruction in the caller since there is no need to pop
17404 the arguments there.
17406 This calling convention is incompatible with the one normally
17407 used on Unix, so you cannot use it if you need to call libraries
17408 compiled with the Unix compiler.
17410 Also, you must provide function prototypes for all functions that
17411 take variable numbers of arguments (including @code{printf});
17412 otherwise incorrect code is generated for calls to those
17415 In addition, seriously incorrect code results if you call a
17416 function with too many arguments. (Normally, extra arguments are
17417 harmlessly ignored.)
17419 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
17420 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
17424 Do not use the calling conventions selected by @option{-mrtd}.
17425 This is the default.
17428 @itemx -mno-align-int
17429 @opindex malign-int
17430 @opindex mno-align-int
17431 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
17432 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
17433 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
17434 Aligning variables on 32-bit boundaries produces code that runs somewhat
17435 faster on processors with 32-bit busses at the expense of more memory.
17437 @strong{Warning:} if you use the @option{-malign-int} switch, GCC
17438 aligns structures containing the above types differently than
17439 most published application binary interface specifications for the m68k.
17443 Use the pc-relative addressing mode of the 68000 directly, instead of
17444 using a global offset table. At present, this option implies @option{-fpic},
17445 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
17446 not presently supported with @option{-mpcrel}, though this could be supported for
17447 68020 and higher processors.
17449 @item -mno-strict-align
17450 @itemx -mstrict-align
17451 @opindex mno-strict-align
17452 @opindex mstrict-align
17453 Do not (do) assume that unaligned memory references are handled by
17457 Generate code that allows the data segment to be located in a different
17458 area of memory from the text segment. This allows for execute-in-place in
17459 an environment without virtual memory management. This option implies
17462 @item -mno-sep-data
17463 Generate code that assumes that the data segment follows the text segment.
17464 This is the default.
17466 @item -mid-shared-library
17467 Generate code that supports shared libraries via the library ID method.
17468 This allows for execute-in-place and shared libraries in an environment
17469 without virtual memory management. This option implies @option{-fPIC}.
17471 @item -mno-id-shared-library
17472 Generate code that doesn't assume ID-based shared libraries are being used.
17473 This is the default.
17475 @item -mshared-library-id=n
17476 Specifies the identification number of the ID-based shared library being
17477 compiled. Specifying a value of 0 generates more compact code; specifying
17478 other values forces the allocation of that number to the current
17479 library, but is no more space- or time-efficient than omitting this option.
17485 When generating position-independent code for ColdFire, generate code
17486 that works if the GOT has more than 8192 entries. This code is
17487 larger and slower than code generated without this option. On M680x0
17488 processors, this option is not needed; @option{-fPIC} suffices.
17490 GCC normally uses a single instruction to load values from the GOT@.
17491 While this is relatively efficient, it only works if the GOT
17492 is smaller than about 64k. Anything larger causes the linker
17493 to report an error such as:
17495 @cindex relocation truncated to fit (ColdFire)
17497 relocation truncated to fit: R_68K_GOT16O foobar
17500 If this happens, you should recompile your code with @option{-mxgot}.
17501 It should then work with very large GOTs. However, code generated with
17502 @option{-mxgot} is less efficient, since it takes 4 instructions to fetch
17503 the value of a global symbol.
17505 Note that some linkers, including newer versions of the GNU linker,
17506 can create multiple GOTs and sort GOT entries. If you have such a linker,
17507 you should only need to use @option{-mxgot} when compiling a single
17508 object file that accesses more than 8192 GOT entries. Very few do.
17510 These options have no effect unless GCC is generating
17511 position-independent code.
17515 @node MCore Options
17516 @subsection MCore Options
17517 @cindex MCore options
17519 These are the @samp{-m} options defined for the Motorola M*Core
17525 @itemx -mno-hardlit
17527 @opindex mno-hardlit
17528 Inline constants into the code stream if it can be done in two
17529 instructions or less.
17535 Use the divide instruction. (Enabled by default).
17537 @item -mrelax-immediate
17538 @itemx -mno-relax-immediate
17539 @opindex mrelax-immediate
17540 @opindex mno-relax-immediate
17541 Allow arbitrary-sized immediates in bit operations.
17543 @item -mwide-bitfields
17544 @itemx -mno-wide-bitfields
17545 @opindex mwide-bitfields
17546 @opindex mno-wide-bitfields
17547 Always treat bit-fields as @code{int}-sized.
17549 @item -m4byte-functions
17550 @itemx -mno-4byte-functions
17551 @opindex m4byte-functions
17552 @opindex mno-4byte-functions
17553 Force all functions to be aligned to a 4-byte boundary.
17555 @item -mcallgraph-data
17556 @itemx -mno-callgraph-data
17557 @opindex mcallgraph-data
17558 @opindex mno-callgraph-data
17559 Emit callgraph information.
17562 @itemx -mno-slow-bytes
17563 @opindex mslow-bytes
17564 @opindex mno-slow-bytes
17565 Prefer word access when reading byte quantities.
17567 @item -mlittle-endian
17568 @itemx -mbig-endian
17569 @opindex mlittle-endian
17570 @opindex mbig-endian
17571 Generate code for a little-endian target.
17577 Generate code for the 210 processor.
17581 Assume that runtime support has been provided and so omit the
17582 simulator library (@file{libsim.a)} from the linker command line.
17584 @item -mstack-increment=@var{size}
17585 @opindex mstack-increment
17586 Set the maximum amount for a single stack increment operation. Large
17587 values can increase the speed of programs that contain functions
17588 that need a large amount of stack space, but they can also trigger a
17589 segmentation fault if the stack is extended too much. The default
17595 @subsection MeP Options
17596 @cindex MeP options
17602 Enables the @code{abs} instruction, which is the absolute difference
17603 between two registers.
17607 Enables all the optional instructions---average, multiply, divide, bit
17608 operations, leading zero, absolute difference, min/max, clip, and
17614 Enables the @code{ave} instruction, which computes the average of two
17617 @item -mbased=@var{n}
17619 Variables of size @var{n} bytes or smaller are placed in the
17620 @code{.based} section by default. Based variables use the @code{$tp}
17621 register as a base register, and there is a 128-byte limit to the
17622 @code{.based} section.
17626 Enables the bit operation instructions---bit test (@code{btstm}), set
17627 (@code{bsetm}), clear (@code{bclrm}), invert (@code{bnotm}), and
17628 test-and-set (@code{tas}).
17630 @item -mc=@var{name}
17632 Selects which section constant data is placed in. @var{name} may
17633 be @samp{tiny}, @samp{near}, or @samp{far}.
17637 Enables the @code{clip} instruction. Note that @option{-mclip} is not
17638 useful unless you also provide @option{-mminmax}.
17640 @item -mconfig=@var{name}
17642 Selects one of the built-in core configurations. Each MeP chip has
17643 one or more modules in it; each module has a core CPU and a variety of
17644 coprocessors, optional instructions, and peripherals. The
17645 @code{MeP-Integrator} tool, not part of GCC, provides these
17646 configurations through this option; using this option is the same as
17647 using all the corresponding command-line options. The default
17648 configuration is @samp{default}.
17652 Enables the coprocessor instructions. By default, this is a 32-bit
17653 coprocessor. Note that the coprocessor is normally enabled via the
17654 @option{-mconfig=} option.
17658 Enables the 32-bit coprocessor's instructions.
17662 Enables the 64-bit coprocessor's instructions.
17666 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor.
17670 Causes constant variables to be placed in the @code{.near} section.
17674 Enables the @code{div} and @code{divu} instructions.
17678 Generate big-endian code.
17682 Generate little-endian code.
17684 @item -mio-volatile
17685 @opindex mio-volatile
17686 Tells the compiler that any variable marked with the @code{io}
17687 attribute is to be considered volatile.
17691 Causes variables to be assigned to the @code{.far} section by default.
17695 Enables the @code{leadz} (leading zero) instruction.
17699 Causes variables to be assigned to the @code{.near} section by default.
17703 Enables the @code{min} and @code{max} instructions.
17707 Enables the multiplication and multiply-accumulate instructions.
17711 Disables all the optional instructions enabled by @option{-mall-opts}.
17715 Enables the @code{repeat} and @code{erepeat} instructions, used for
17716 low-overhead looping.
17720 Causes all variables to default to the @code{.tiny} section. Note
17721 that there is a 65536-byte limit to this section. Accesses to these
17722 variables use the @code{%gp} base register.
17726 Enables the saturation instructions. Note that the compiler does not
17727 currently generate these itself, but this option is included for
17728 compatibility with other tools, like @code{as}.
17732 Link the SDRAM-based runtime instead of the default ROM-based runtime.
17736 Link the simulator run-time libraries.
17740 Link the simulator runtime libraries, excluding built-in support
17741 for reset and exception vectors and tables.
17745 Causes all functions to default to the @code{.far} section. Without
17746 this option, functions default to the @code{.near} section.
17748 @item -mtiny=@var{n}
17750 Variables that are @var{n} bytes or smaller are allocated to the
17751 @code{.tiny} section. These variables use the @code{$gp} base
17752 register. The default for this option is 4, but note that there's a
17753 65536-byte limit to the @code{.tiny} section.
17757 @node MicroBlaze Options
17758 @subsection MicroBlaze Options
17759 @cindex MicroBlaze Options
17764 @opindex msoft-float
17765 Use software emulation for floating point (default).
17768 @opindex mhard-float
17769 Use hardware floating-point instructions.
17773 Do not optimize block moves, use @code{memcpy}.
17775 @item -mno-clearbss
17776 @opindex mno-clearbss
17777 This option is deprecated. Use @option{-fno-zero-initialized-in-bss} instead.
17779 @item -mcpu=@var{cpu-type}
17781 Use features of, and schedule code for, the given CPU.
17782 Supported values are in the format @samp{v@var{X}.@var{YY}.@var{Z}},
17783 where @var{X} is a major version, @var{YY} is the minor version, and
17784 @var{Z} is compatibility code. Example values are @samp{v3.00.a},
17785 @samp{v4.00.b}, @samp{v5.00.a}, @samp{v5.00.b}, @samp{v5.00.b}, @samp{v6.00.a}.
17787 @item -mxl-soft-mul
17788 @opindex mxl-soft-mul
17789 Use software multiply emulation (default).
17791 @item -mxl-soft-div
17792 @opindex mxl-soft-div
17793 Use software emulation for divides (default).
17795 @item -mxl-barrel-shift
17796 @opindex mxl-barrel-shift
17797 Use the hardware barrel shifter.
17799 @item -mxl-pattern-compare
17800 @opindex mxl-pattern-compare
17801 Use pattern compare instructions.
17803 @item -msmall-divides
17804 @opindex msmall-divides
17805 Use table lookup optimization for small signed integer divisions.
17807 @item -mxl-stack-check
17808 @opindex mxl-stack-check
17809 This option is deprecated. Use @option{-fstack-check} instead.
17812 @opindex mxl-gp-opt
17813 Use GP-relative @code{.sdata}/@code{.sbss} sections.
17815 @item -mxl-multiply-high
17816 @opindex mxl-multiply-high
17817 Use multiply high instructions for high part of 32x32 multiply.
17819 @item -mxl-float-convert
17820 @opindex mxl-float-convert
17821 Use hardware floating-point conversion instructions.
17823 @item -mxl-float-sqrt
17824 @opindex mxl-float-sqrt
17825 Use hardware floating-point square root instruction.
17828 @opindex mbig-endian
17829 Generate code for a big-endian target.
17831 @item -mlittle-endian
17832 @opindex mlittle-endian
17833 Generate code for a little-endian target.
17836 @opindex mxl-reorder
17837 Use reorder instructions (swap and byte reversed load/store).
17839 @item -mxl-mode-@var{app-model}
17840 Select application model @var{app-model}. Valid models are
17843 normal executable (default), uses startup code @file{crt0.o}.
17846 for use with Xilinx Microprocessor Debugger (XMD) based
17847 software intrusive debug agent called xmdstub. This uses startup file
17848 @file{crt1.o} and sets the start address of the program to 0x800.
17851 for applications that are loaded using a bootloader.
17852 This model uses startup file @file{crt2.o} which does not contain a processor
17853 reset vector handler. This is suitable for transferring control on a
17854 processor reset to the bootloader rather than the application.
17857 for applications that do not require any of the
17858 MicroBlaze vectors. This option may be useful for applications running
17859 within a monitoring application. This model uses @file{crt3.o} as a startup file.
17862 Option @option{-xl-mode-@var{app-model}} is a deprecated alias for
17863 @option{-mxl-mode-@var{app-model}}.
17868 @subsection MIPS Options
17869 @cindex MIPS options
17875 Generate big-endian code.
17879 Generate little-endian code. This is the default for @samp{mips*el-*-*}
17882 @item -march=@var{arch}
17884 Generate code that runs on @var{arch}, which can be the name of a
17885 generic MIPS ISA, or the name of a particular processor.
17887 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
17888 @samp{mips32}, @samp{mips32r2}, @samp{mips32r3}, @samp{mips32r5},
17889 @samp{mips32r6}, @samp{mips64}, @samp{mips64r2}, @samp{mips64r3},
17890 @samp{mips64r5} and @samp{mips64r6}.
17891 The processor names are:
17892 @samp{4kc}, @samp{4km}, @samp{4kp}, @samp{4ksc},
17893 @samp{4kec}, @samp{4kem}, @samp{4kep}, @samp{4ksd},
17894 @samp{5kc}, @samp{5kf},
17896 @samp{24kc}, @samp{24kf2_1}, @samp{24kf1_1},
17897 @samp{24kec}, @samp{24kef2_1}, @samp{24kef1_1},
17898 @samp{34kc}, @samp{34kf2_1}, @samp{34kf1_1}, @samp{34kn},
17899 @samp{74kc}, @samp{74kf2_1}, @samp{74kf1_1}, @samp{74kf3_2},
17900 @samp{1004kc}, @samp{1004kf2_1}, @samp{1004kf1_1},
17903 @samp{loongson2e}, @samp{loongson2f}, @samp{loongson3a},
17905 @samp{m14k}, @samp{m14kc}, @samp{m14ke}, @samp{m14kec},
17906 @samp{m5100}, @samp{m5101},
17907 @samp{octeon}, @samp{octeon+}, @samp{octeon2}, @samp{octeon3},
17910 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
17911 @samp{r4600}, @samp{r4650}, @samp{r4700}, @samp{r6000}, @samp{r8000},
17912 @samp{rm7000}, @samp{rm9000},
17913 @samp{r10000}, @samp{r12000}, @samp{r14000}, @samp{r16000},
17916 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
17917 @samp{vr5000}, @samp{vr5400}, @samp{vr5500},
17918 @samp{xlr} and @samp{xlp}.
17919 The special value @samp{from-abi} selects the
17920 most compatible architecture for the selected ABI (that is,
17921 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
17923 The native Linux/GNU toolchain also supports the value @samp{native},
17924 which selects the best architecture option for the host processor.
17925 @option{-march=native} has no effect if GCC does not recognize
17928 In processor names, a final @samp{000} can be abbreviated as @samp{k}
17929 (for example, @option{-march=r2k}). Prefixes are optional, and
17930 @samp{vr} may be written @samp{r}.
17932 Names of the form @samp{@var{n}f2_1} refer to processors with
17933 FPUs clocked at half the rate of the core, names of the form
17934 @samp{@var{n}f1_1} refer to processors with FPUs clocked at the same
17935 rate as the core, and names of the form @samp{@var{n}f3_2} refer to
17936 processors with FPUs clocked a ratio of 3:2 with respect to the core.
17937 For compatibility reasons, @samp{@var{n}f} is accepted as a synonym
17938 for @samp{@var{n}f2_1} while @samp{@var{n}x} and @samp{@var{b}fx} are
17939 accepted as synonyms for @samp{@var{n}f1_1}.
17941 GCC defines two macros based on the value of this option. The first
17942 is @code{_MIPS_ARCH}, which gives the name of target architecture, as
17943 a string. The second has the form @code{_MIPS_ARCH_@var{foo}},
17944 where @var{foo} is the capitalized value of @code{_MIPS_ARCH}@.
17945 For example, @option{-march=r2000} sets @code{_MIPS_ARCH}
17946 to @code{"r2000"} and defines the macro @code{_MIPS_ARCH_R2000}.
17948 Note that the @code{_MIPS_ARCH} macro uses the processor names given
17949 above. In other words, it has the full prefix and does not
17950 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
17951 the macro names the resolved architecture (either @code{"mips1"} or
17952 @code{"mips3"}). It names the default architecture when no
17953 @option{-march} option is given.
17955 @item -mtune=@var{arch}
17957 Optimize for @var{arch}. Among other things, this option controls
17958 the way instructions are scheduled, and the perceived cost of arithmetic
17959 operations. The list of @var{arch} values is the same as for
17962 When this option is not used, GCC optimizes for the processor
17963 specified by @option{-march}. By using @option{-march} and
17964 @option{-mtune} together, it is possible to generate code that
17965 runs on a family of processors, but optimize the code for one
17966 particular member of that family.
17968 @option{-mtune} defines the macros @code{_MIPS_TUNE} and
17969 @code{_MIPS_TUNE_@var{foo}}, which work in the same way as the
17970 @option{-march} ones described above.
17974 Equivalent to @option{-march=mips1}.
17978 Equivalent to @option{-march=mips2}.
17982 Equivalent to @option{-march=mips3}.
17986 Equivalent to @option{-march=mips4}.
17990 Equivalent to @option{-march=mips32}.
17994 Equivalent to @option{-march=mips32r3}.
17998 Equivalent to @option{-march=mips32r5}.
18002 Equivalent to @option{-march=mips32r6}.
18006 Equivalent to @option{-march=mips64}.
18010 Equivalent to @option{-march=mips64r2}.
18014 Equivalent to @option{-march=mips64r3}.
18018 Equivalent to @option{-march=mips64r5}.
18022 Equivalent to @option{-march=mips64r6}.
18027 @opindex mno-mips16
18028 Generate (do not generate) MIPS16 code. If GCC is targeting a
18029 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE@.
18031 MIPS16 code generation can also be controlled on a per-function basis
18032 by means of @code{mips16} and @code{nomips16} attributes.
18033 @xref{Function Attributes}, for more information.
18035 @item -mflip-mips16
18036 @opindex mflip-mips16
18037 Generate MIPS16 code on alternating functions. This option is provided
18038 for regression testing of mixed MIPS16/non-MIPS16 code generation, and is
18039 not intended for ordinary use in compiling user code.
18041 @item -minterlink-compressed
18042 @item -mno-interlink-compressed
18043 @opindex minterlink-compressed
18044 @opindex mno-interlink-compressed
18045 Require (do not require) that code using the standard (uncompressed) MIPS ISA
18046 be link-compatible with MIPS16 and microMIPS code, and vice versa.
18048 For example, code using the standard ISA encoding cannot jump directly
18049 to MIPS16 or microMIPS code; it must either use a call or an indirect jump.
18050 @option{-minterlink-compressed} therefore disables direct jumps unless GCC
18051 knows that the target of the jump is not compressed.
18053 @item -minterlink-mips16
18054 @itemx -mno-interlink-mips16
18055 @opindex minterlink-mips16
18056 @opindex mno-interlink-mips16
18057 Aliases of @option{-minterlink-compressed} and
18058 @option{-mno-interlink-compressed}. These options predate the microMIPS ASE
18059 and are retained for backwards compatibility.
18071 Generate code for the given ABI@.
18073 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
18074 generates 64-bit code when you select a 64-bit architecture, but you
18075 can use @option{-mgp32} to get 32-bit code instead.
18077 For information about the O64 ABI, see
18078 @uref{http://gcc.gnu.org/@/projects/@/mipso64-abi.html}.
18080 GCC supports a variant of the o32 ABI in which floating-point registers
18081 are 64 rather than 32 bits wide. You can select this combination with
18082 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @code{mthc1}
18083 and @code{mfhc1} instructions and is therefore only supported for
18084 MIPS32R2, MIPS32R3 and MIPS32R5 processors.
18086 The register assignments for arguments and return values remain the
18087 same, but each scalar value is passed in a single 64-bit register
18088 rather than a pair of 32-bit registers. For example, scalar
18089 floating-point values are returned in @samp{$f0} only, not a
18090 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
18091 remains the same in that the even-numbered double-precision registers
18094 Two additional variants of the o32 ABI are supported to enable
18095 a transition from 32-bit to 64-bit registers. These are FPXX
18096 (@option{-mfpxx}) and FP64A (@option{-mfp64} @option{-mno-odd-spreg}).
18097 The FPXX extension mandates that all code must execute correctly
18098 when run using 32-bit or 64-bit registers. The code can be interlinked
18099 with either FP32 or FP64, but not both.
18100 The FP64A extension is similar to the FP64 extension but forbids the
18101 use of odd-numbered single-precision registers. This can be used
18102 in conjunction with the @code{FRE} mode of FPUs in MIPS32R5
18103 processors and allows both FP32 and FP64A code to interlink and
18104 run in the same process without changing FPU modes.
18107 @itemx -mno-abicalls
18109 @opindex mno-abicalls
18110 Generate (do not generate) code that is suitable for SVR4-style
18111 dynamic objects. @option{-mabicalls} is the default for SVR4-based
18116 Generate (do not generate) code that is fully position-independent,
18117 and that can therefore be linked into shared libraries. This option
18118 only affects @option{-mabicalls}.
18120 All @option{-mabicalls} code has traditionally been position-independent,
18121 regardless of options like @option{-fPIC} and @option{-fpic}. However,
18122 as an extension, the GNU toolchain allows executables to use absolute
18123 accesses for locally-binding symbols. It can also use shorter GP
18124 initialization sequences and generate direct calls to locally-defined
18125 functions. This mode is selected by @option{-mno-shared}.
18127 @option{-mno-shared} depends on binutils 2.16 or higher and generates
18128 objects that can only be linked by the GNU linker. However, the option
18129 does not affect the ABI of the final executable; it only affects the ABI
18130 of relocatable objects. Using @option{-mno-shared} generally makes
18131 executables both smaller and quicker.
18133 @option{-mshared} is the default.
18139 Assume (do not assume) that the static and dynamic linkers
18140 support PLTs and copy relocations. This option only affects
18141 @option{-mno-shared -mabicalls}. For the n64 ABI, this option
18142 has no effect without @option{-msym32}.
18144 You can make @option{-mplt} the default by configuring
18145 GCC with @option{--with-mips-plt}. The default is
18146 @option{-mno-plt} otherwise.
18152 Lift (do not lift) the usual restrictions on the size of the global
18155 GCC normally uses a single instruction to load values from the GOT@.
18156 While this is relatively efficient, it only works if the GOT
18157 is smaller than about 64k. Anything larger causes the linker
18158 to report an error such as:
18160 @cindex relocation truncated to fit (MIPS)
18162 relocation truncated to fit: R_MIPS_GOT16 foobar
18165 If this happens, you should recompile your code with @option{-mxgot}.
18166 This works with very large GOTs, although the code is also
18167 less efficient, since it takes three instructions to fetch the
18168 value of a global symbol.
18170 Note that some linkers can create multiple GOTs. If you have such a
18171 linker, you should only need to use @option{-mxgot} when a single object
18172 file accesses more than 64k's worth of GOT entries. Very few do.
18174 These options have no effect unless GCC is generating position
18179 Assume that general-purpose registers are 32 bits wide.
18183 Assume that general-purpose registers are 64 bits wide.
18187 Assume that floating-point registers are 32 bits wide.
18191 Assume that floating-point registers are 64 bits wide.
18195 Do not assume the width of floating-point registers.
18198 @opindex mhard-float
18199 Use floating-point coprocessor instructions.
18202 @opindex msoft-float
18203 Do not use floating-point coprocessor instructions. Implement
18204 floating-point calculations using library calls instead.
18208 Equivalent to @option{-msoft-float}, but additionally asserts that the
18209 program being compiled does not perform any floating-point operations.
18210 This option is presently supported only by some bare-metal MIPS
18211 configurations, where it may select a special set of libraries
18212 that lack all floating-point support (including, for example, the
18213 floating-point @code{printf} formats).
18214 If code compiled with @option{-mno-float} accidentally contains
18215 floating-point operations, it is likely to suffer a link-time
18216 or run-time failure.
18218 @item -msingle-float
18219 @opindex msingle-float
18220 Assume that the floating-point coprocessor only supports single-precision
18223 @item -mdouble-float
18224 @opindex mdouble-float
18225 Assume that the floating-point coprocessor supports double-precision
18226 operations. This is the default.
18229 @itemx -mno-odd-spreg
18230 @opindex modd-spreg
18231 @opindex mno-odd-spreg
18232 Enable the use of odd-numbered single-precision floating-point registers
18233 for the o32 ABI. This is the default for processors that are known to
18234 support these registers. When using the o32 FPXX ABI, @option{-mno-odd-spreg}
18238 @itemx -mabs=legacy
18240 @opindex mabs=legacy
18241 These options control the treatment of the special not-a-number (NaN)
18242 IEEE 754 floating-point data with the @code{abs.@i{fmt}} and
18243 @code{neg.@i{fmt}} machine instructions.
18245 By default or when @option{-mabs=legacy} is used the legacy
18246 treatment is selected. In this case these instructions are considered
18247 arithmetic and avoided where correct operation is required and the
18248 input operand might be a NaN. A longer sequence of instructions that
18249 manipulate the sign bit of floating-point datum manually is used
18250 instead unless the @option{-ffinite-math-only} option has also been
18253 The @option{-mabs=2008} option selects the IEEE 754-2008 treatment. In
18254 this case these instructions are considered non-arithmetic and therefore
18255 operating correctly in all cases, including in particular where the
18256 input operand is a NaN. These instructions are therefore always used
18257 for the respective operations.
18260 @itemx -mnan=legacy
18262 @opindex mnan=legacy
18263 These options control the encoding of the special not-a-number (NaN)
18264 IEEE 754 floating-point data.
18266 The @option{-mnan=legacy} option selects the legacy encoding. In this
18267 case quiet NaNs (qNaNs) are denoted by the first bit of their trailing
18268 significand field being 0, whereas signaling NaNs (sNaNs) are denoted
18269 by the first bit of their trailing significand field being 1.
18271 The @option{-mnan=2008} option selects the IEEE 754-2008 encoding. In
18272 this case qNaNs are denoted by the first bit of their trailing
18273 significand field being 1, whereas sNaNs are denoted by the first bit of
18274 their trailing significand field being 0.
18276 The default is @option{-mnan=legacy} unless GCC has been configured with
18277 @option{--with-nan=2008}.
18283 Use (do not use) @samp{ll}, @samp{sc}, and @samp{sync} instructions to
18284 implement atomic memory built-in functions. When neither option is
18285 specified, GCC uses the instructions if the target architecture
18288 @option{-mllsc} is useful if the runtime environment can emulate the
18289 instructions and @option{-mno-llsc} can be useful when compiling for
18290 nonstandard ISAs. You can make either option the default by
18291 configuring GCC with @option{--with-llsc} and @option{--without-llsc}
18292 respectively. @option{--with-llsc} is the default for some
18293 configurations; see the installation documentation for details.
18299 Use (do not use) revision 1 of the MIPS DSP ASE@.
18300 @xref{MIPS DSP Built-in Functions}. This option defines the
18301 preprocessor macro @code{__mips_dsp}. It also defines
18302 @code{__mips_dsp_rev} to 1.
18308 Use (do not use) revision 2 of the MIPS DSP ASE@.
18309 @xref{MIPS DSP Built-in Functions}. This option defines the
18310 preprocessor macros @code{__mips_dsp} and @code{__mips_dspr2}.
18311 It also defines @code{__mips_dsp_rev} to 2.
18314 @itemx -mno-smartmips
18315 @opindex msmartmips
18316 @opindex mno-smartmips
18317 Use (do not use) the MIPS SmartMIPS ASE.
18319 @item -mpaired-single
18320 @itemx -mno-paired-single
18321 @opindex mpaired-single
18322 @opindex mno-paired-single
18323 Use (do not use) paired-single floating-point instructions.
18324 @xref{MIPS Paired-Single Support}. This option requires
18325 hardware floating-point support to be enabled.
18331 Use (do not use) MIPS Digital Media Extension instructions.
18332 This option can only be used when generating 64-bit code and requires
18333 hardware floating-point support to be enabled.
18338 @opindex mno-mips3d
18339 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
18340 The option @option{-mips3d} implies @option{-mpaired-single}.
18343 @itemx -mno-micromips
18344 @opindex mmicromips
18345 @opindex mno-mmicromips
18346 Generate (do not generate) microMIPS code.
18348 MicroMIPS code generation can also be controlled on a per-function basis
18349 by means of @code{micromips} and @code{nomicromips} attributes.
18350 @xref{Function Attributes}, for more information.
18356 Use (do not use) MT Multithreading instructions.
18362 Use (do not use) the MIPS MCU ASE instructions.
18368 Use (do not use) the MIPS Enhanced Virtual Addressing instructions.
18374 Use (do not use) the MIPS Virtualization Application Specific instructions.
18380 Use (do not use) the MIPS eXtended Physical Address (XPA) instructions.
18384 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
18385 an explanation of the default and the way that the pointer size is
18390 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
18392 The default size of @code{int}s, @code{long}s and pointers depends on
18393 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
18394 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
18395 32-bit @code{long}s. Pointers are the same size as @code{long}s,
18396 or the same size as integer registers, whichever is smaller.
18402 Assume (do not assume) that all symbols have 32-bit values, regardless
18403 of the selected ABI@. This option is useful in combination with
18404 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
18405 to generate shorter and faster references to symbolic addresses.
18409 Put definitions of externally-visible data in a small data section
18410 if that data is no bigger than @var{num} bytes. GCC can then generate
18411 more efficient accesses to the data; see @option{-mgpopt} for details.
18413 The default @option{-G} option depends on the configuration.
18415 @item -mlocal-sdata
18416 @itemx -mno-local-sdata
18417 @opindex mlocal-sdata
18418 @opindex mno-local-sdata
18419 Extend (do not extend) the @option{-G} behavior to local data too,
18420 such as to static variables in C@. @option{-mlocal-sdata} is the
18421 default for all configurations.
18423 If the linker complains that an application is using too much small data,
18424 you might want to try rebuilding the less performance-critical parts with
18425 @option{-mno-local-sdata}. You might also want to build large
18426 libraries with @option{-mno-local-sdata}, so that the libraries leave
18427 more room for the main program.
18429 @item -mextern-sdata
18430 @itemx -mno-extern-sdata
18431 @opindex mextern-sdata
18432 @opindex mno-extern-sdata
18433 Assume (do not assume) that externally-defined data is in
18434 a small data section if the size of that data is within the @option{-G} limit.
18435 @option{-mextern-sdata} is the default for all configurations.
18437 If you compile a module @var{Mod} with @option{-mextern-sdata} @option{-G
18438 @var{num}} @option{-mgpopt}, and @var{Mod} references a variable @var{Var}
18439 that is no bigger than @var{num} bytes, you must make sure that @var{Var}
18440 is placed in a small data section. If @var{Var} is defined by another
18441 module, you must either compile that module with a high-enough
18442 @option{-G} setting or attach a @code{section} attribute to @var{Var}'s
18443 definition. If @var{Var} is common, you must link the application
18444 with a high-enough @option{-G} setting.
18446 The easiest way of satisfying these restrictions is to compile
18447 and link every module with the same @option{-G} option. However,
18448 you may wish to build a library that supports several different
18449 small data limits. You can do this by compiling the library with
18450 the highest supported @option{-G} setting and additionally using
18451 @option{-mno-extern-sdata} to stop the library from making assumptions
18452 about externally-defined data.
18458 Use (do not use) GP-relative accesses for symbols that are known to be
18459 in a small data section; see @option{-G}, @option{-mlocal-sdata} and
18460 @option{-mextern-sdata}. @option{-mgpopt} is the default for all
18463 @option{-mno-gpopt} is useful for cases where the @code{$gp} register
18464 might not hold the value of @code{_gp}. For example, if the code is
18465 part of a library that might be used in a boot monitor, programs that
18466 call boot monitor routines pass an unknown value in @code{$gp}.
18467 (In such situations, the boot monitor itself is usually compiled
18468 with @option{-G0}.)
18470 @option{-mno-gpopt} implies @option{-mno-local-sdata} and
18471 @option{-mno-extern-sdata}.
18473 @item -membedded-data
18474 @itemx -mno-embedded-data
18475 @opindex membedded-data
18476 @opindex mno-embedded-data
18477 Allocate variables to the read-only data section first if possible, then
18478 next in the small data section if possible, otherwise in data. This gives
18479 slightly slower code than the default, but reduces the amount of RAM required
18480 when executing, and thus may be preferred for some embedded systems.
18482 @item -muninit-const-in-rodata
18483 @itemx -mno-uninit-const-in-rodata
18484 @opindex muninit-const-in-rodata
18485 @opindex mno-uninit-const-in-rodata
18486 Put uninitialized @code{const} variables in the read-only data section.
18487 This option is only meaningful in conjunction with @option{-membedded-data}.
18489 @item -mcode-readable=@var{setting}
18490 @opindex mcode-readable
18491 Specify whether GCC may generate code that reads from executable sections.
18492 There are three possible settings:
18495 @item -mcode-readable=yes
18496 Instructions may freely access executable sections. This is the
18499 @item -mcode-readable=pcrel
18500 MIPS16 PC-relative load instructions can access executable sections,
18501 but other instructions must not do so. This option is useful on 4KSc
18502 and 4KSd processors when the code TLBs have the Read Inhibit bit set.
18503 It is also useful on processors that can be configured to have a dual
18504 instruction/data SRAM interface and that, like the M4K, automatically
18505 redirect PC-relative loads to the instruction RAM.
18507 @item -mcode-readable=no
18508 Instructions must not access executable sections. This option can be
18509 useful on targets that are configured to have a dual instruction/data
18510 SRAM interface but that (unlike the M4K) do not automatically redirect
18511 PC-relative loads to the instruction RAM.
18514 @item -msplit-addresses
18515 @itemx -mno-split-addresses
18516 @opindex msplit-addresses
18517 @opindex mno-split-addresses
18518 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
18519 relocation operators. This option has been superseded by
18520 @option{-mexplicit-relocs} but is retained for backwards compatibility.
18522 @item -mexplicit-relocs
18523 @itemx -mno-explicit-relocs
18524 @opindex mexplicit-relocs
18525 @opindex mno-explicit-relocs
18526 Use (do not use) assembler relocation operators when dealing with symbolic
18527 addresses. The alternative, selected by @option{-mno-explicit-relocs},
18528 is to use assembler macros instead.
18530 @option{-mexplicit-relocs} is the default if GCC was configured
18531 to use an assembler that supports relocation operators.
18533 @item -mcheck-zero-division
18534 @itemx -mno-check-zero-division
18535 @opindex mcheck-zero-division
18536 @opindex mno-check-zero-division
18537 Trap (do not trap) on integer division by zero.
18539 The default is @option{-mcheck-zero-division}.
18541 @item -mdivide-traps
18542 @itemx -mdivide-breaks
18543 @opindex mdivide-traps
18544 @opindex mdivide-breaks
18545 MIPS systems check for division by zero by generating either a
18546 conditional trap or a break instruction. Using traps results in
18547 smaller code, but is only supported on MIPS II and later. Also, some
18548 versions of the Linux kernel have a bug that prevents trap from
18549 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
18550 allow conditional traps on architectures that support them and
18551 @option{-mdivide-breaks} to force the use of breaks.
18553 The default is usually @option{-mdivide-traps}, but this can be
18554 overridden at configure time using @option{--with-divide=breaks}.
18555 Divide-by-zero checks can be completely disabled using
18556 @option{-mno-check-zero-division}.
18561 @opindex mno-memcpy
18562 Force (do not force) the use of @code{memcpy} for non-trivial block
18563 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
18564 most constant-sized copies.
18567 @itemx -mno-long-calls
18568 @opindex mlong-calls
18569 @opindex mno-long-calls
18570 Disable (do not disable) use of the @code{jal} instruction. Calling
18571 functions using @code{jal} is more efficient but requires the caller
18572 and callee to be in the same 256 megabyte segment.
18574 This option has no effect on abicalls code. The default is
18575 @option{-mno-long-calls}.
18581 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
18582 instructions, as provided by the R4650 ISA@.
18588 Enable (disable) use of the @code{madd} and @code{msub} integer
18589 instructions. The default is @option{-mimadd} on architectures
18590 that support @code{madd} and @code{msub} except for the 74k
18591 architecture where it was found to generate slower code.
18594 @itemx -mno-fused-madd
18595 @opindex mfused-madd
18596 @opindex mno-fused-madd
18597 Enable (disable) use of the floating-point multiply-accumulate
18598 instructions, when they are available. The default is
18599 @option{-mfused-madd}.
18601 On the R8000 CPU when multiply-accumulate instructions are used,
18602 the intermediate product is calculated to infinite precision
18603 and is not subject to the FCSR Flush to Zero bit. This may be
18604 undesirable in some circumstances. On other processors the result
18605 is numerically identical to the equivalent computation using
18606 separate multiply, add, subtract and negate instructions.
18610 Tell the MIPS assembler to not run its preprocessor over user
18611 assembler files (with a @samp{.s} suffix) when assembling them.
18616 @opindex mno-fix-24k
18617 Work around the 24K E48 (lost data on stores during refill) errata.
18618 The workarounds are implemented by the assembler rather than by GCC@.
18621 @itemx -mno-fix-r4000
18622 @opindex mfix-r4000
18623 @opindex mno-fix-r4000
18624 Work around certain R4000 CPU errata:
18627 A double-word or a variable shift may give an incorrect result if executed
18628 immediately after starting an integer division.
18630 A double-word or a variable shift may give an incorrect result if executed
18631 while an integer multiplication is in progress.
18633 An integer division may give an incorrect result if started in a delay slot
18634 of a taken branch or a jump.
18638 @itemx -mno-fix-r4400
18639 @opindex mfix-r4400
18640 @opindex mno-fix-r4400
18641 Work around certain R4400 CPU errata:
18644 A double-word or a variable shift may give an incorrect result if executed
18645 immediately after starting an integer division.
18649 @itemx -mno-fix-r10000
18650 @opindex mfix-r10000
18651 @opindex mno-fix-r10000
18652 Work around certain R10000 errata:
18655 @code{ll}/@code{sc} sequences may not behave atomically on revisions
18656 prior to 3.0. They may deadlock on revisions 2.6 and earlier.
18659 This option can only be used if the target architecture supports
18660 branch-likely instructions. @option{-mfix-r10000} is the default when
18661 @option{-march=r10000} is used; @option{-mno-fix-r10000} is the default
18665 @itemx -mno-fix-rm7000
18666 @opindex mfix-rm7000
18667 Work around the RM7000 @code{dmult}/@code{dmultu} errata. The
18668 workarounds are implemented by the assembler rather than by GCC@.
18671 @itemx -mno-fix-vr4120
18672 @opindex mfix-vr4120
18673 Work around certain VR4120 errata:
18676 @code{dmultu} does not always produce the correct result.
18678 @code{div} and @code{ddiv} do not always produce the correct result if one
18679 of the operands is negative.
18681 The workarounds for the division errata rely on special functions in
18682 @file{libgcc.a}. At present, these functions are only provided by
18683 the @code{mips64vr*-elf} configurations.
18685 Other VR4120 errata require a NOP to be inserted between certain pairs of
18686 instructions. These errata are handled by the assembler, not by GCC itself.
18689 @opindex mfix-vr4130
18690 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
18691 workarounds are implemented by the assembler rather than by GCC,
18692 although GCC avoids using @code{mflo} and @code{mfhi} if the
18693 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
18694 instructions are available instead.
18697 @itemx -mno-fix-sb1
18699 Work around certain SB-1 CPU core errata.
18700 (This flag currently works around the SB-1 revision 2
18701 ``F1'' and ``F2'' floating-point errata.)
18703 @item -mr10k-cache-barrier=@var{setting}
18704 @opindex mr10k-cache-barrier
18705 Specify whether GCC should insert cache barriers to avoid the
18706 side-effects of speculation on R10K processors.
18708 In common with many processors, the R10K tries to predict the outcome
18709 of a conditional branch and speculatively executes instructions from
18710 the ``taken'' branch. It later aborts these instructions if the
18711 predicted outcome is wrong. However, on the R10K, even aborted
18712 instructions can have side effects.
18714 This problem only affects kernel stores and, depending on the system,
18715 kernel loads. As an example, a speculatively-executed store may load
18716 the target memory into cache and mark the cache line as dirty, even if
18717 the store itself is later aborted. If a DMA operation writes to the
18718 same area of memory before the ``dirty'' line is flushed, the cached
18719 data overwrites the DMA-ed data. See the R10K processor manual
18720 for a full description, including other potential problems.
18722 One workaround is to insert cache barrier instructions before every memory
18723 access that might be speculatively executed and that might have side
18724 effects even if aborted. @option{-mr10k-cache-barrier=@var{setting}}
18725 controls GCC's implementation of this workaround. It assumes that
18726 aborted accesses to any byte in the following regions does not have
18731 the memory occupied by the current function's stack frame;
18734 the memory occupied by an incoming stack argument;
18737 the memory occupied by an object with a link-time-constant address.
18740 It is the kernel's responsibility to ensure that speculative
18741 accesses to these regions are indeed safe.
18743 If the input program contains a function declaration such as:
18749 then the implementation of @code{foo} must allow @code{j foo} and
18750 @code{jal foo} to be executed speculatively. GCC honors this
18751 restriction for functions it compiles itself. It expects non-GCC
18752 functions (such as hand-written assembly code) to do the same.
18754 The option has three forms:
18757 @item -mr10k-cache-barrier=load-store
18758 Insert a cache barrier before a load or store that might be
18759 speculatively executed and that might have side effects even
18762 @item -mr10k-cache-barrier=store
18763 Insert a cache barrier before a store that might be speculatively
18764 executed and that might have side effects even if aborted.
18766 @item -mr10k-cache-barrier=none
18767 Disable the insertion of cache barriers. This is the default setting.
18770 @item -mflush-func=@var{func}
18771 @itemx -mno-flush-func
18772 @opindex mflush-func
18773 Specifies the function to call to flush the I and D caches, or to not
18774 call any such function. If called, the function must take the same
18775 arguments as the common @code{_flush_func}, that is, the address of the
18776 memory range for which the cache is being flushed, the size of the
18777 memory range, and the number 3 (to flush both caches). The default
18778 depends on the target GCC was configured for, but commonly is either
18779 @code{_flush_func} or @code{__cpu_flush}.
18781 @item mbranch-cost=@var{num}
18782 @opindex mbranch-cost
18783 Set the cost of branches to roughly @var{num} ``simple'' instructions.
18784 This cost is only a heuristic and is not guaranteed to produce
18785 consistent results across releases. A zero cost redundantly selects
18786 the default, which is based on the @option{-mtune} setting.
18788 @item -mbranch-likely
18789 @itemx -mno-branch-likely
18790 @opindex mbranch-likely
18791 @opindex mno-branch-likely
18792 Enable or disable use of Branch Likely instructions, regardless of the
18793 default for the selected architecture. By default, Branch Likely
18794 instructions may be generated if they are supported by the selected
18795 architecture. An exception is for the MIPS32 and MIPS64 architectures
18796 and processors that implement those architectures; for those, Branch
18797 Likely instructions are not be generated by default because the MIPS32
18798 and MIPS64 architectures specifically deprecate their use.
18800 @item -mcompact-branches=never
18801 @itemx -mcompact-branches=optimal
18802 @itemx -mcompact-branches=always
18803 @opindex mcompact-branches=never
18804 @opindex mcompact-branches=optimal
18805 @opindex mcompact-branches=always
18806 These options control which form of branches will be generated. The
18807 default is @option{-mcompact-branches=optimal}.
18809 The @option{-mcompact-branches=never} option ensures that compact branch
18810 instructions will never be generated.
18812 The @option{-mcompact-branches=always} option ensures that a compact
18813 branch instruction will be generated if available. If a compact branch
18814 instruction is not available, a delay slot form of the branch will be
18817 This option is supported from MIPS Release 6 onwards.
18819 The @option{-mcompact-branches=optimal} option will cause a delay slot
18820 branch to be used if one is available in the current ISA and the delay
18821 slot is successfully filled. If the delay slot is not filled, a compact
18822 branch will be chosen if one is available.
18824 @item -mfp-exceptions
18825 @itemx -mno-fp-exceptions
18826 @opindex mfp-exceptions
18827 Specifies whether FP exceptions are enabled. This affects how
18828 FP instructions are scheduled for some processors.
18829 The default is that FP exceptions are
18832 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
18833 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
18836 @item -mvr4130-align
18837 @itemx -mno-vr4130-align
18838 @opindex mvr4130-align
18839 The VR4130 pipeline is two-way superscalar, but can only issue two
18840 instructions together if the first one is 8-byte aligned. When this
18841 option is enabled, GCC aligns pairs of instructions that it
18842 thinks should execute in parallel.
18844 This option only has an effect when optimizing for the VR4130.
18845 It normally makes code faster, but at the expense of making it bigger.
18846 It is enabled by default at optimization level @option{-O3}.
18851 Enable (disable) generation of @code{synci} instructions on
18852 architectures that support it. The @code{synci} instructions (if
18853 enabled) are generated when @code{__builtin___clear_cache} is
18856 This option defaults to @option{-mno-synci}, but the default can be
18857 overridden by configuring GCC with @option{--with-synci}.
18859 When compiling code for single processor systems, it is generally safe
18860 to use @code{synci}. However, on many multi-core (SMP) systems, it
18861 does not invalidate the instruction caches on all cores and may lead
18862 to undefined behavior.
18864 @item -mrelax-pic-calls
18865 @itemx -mno-relax-pic-calls
18866 @opindex mrelax-pic-calls
18867 Try to turn PIC calls that are normally dispatched via register
18868 @code{$25} into direct calls. This is only possible if the linker can
18869 resolve the destination at link time and if the destination is within
18870 range for a direct call.
18872 @option{-mrelax-pic-calls} is the default if GCC was configured to use
18873 an assembler and a linker that support the @code{.reloc} assembly
18874 directive and @option{-mexplicit-relocs} is in effect. With
18875 @option{-mno-explicit-relocs}, this optimization can be performed by the
18876 assembler and the linker alone without help from the compiler.
18878 @item -mmcount-ra-address
18879 @itemx -mno-mcount-ra-address
18880 @opindex mmcount-ra-address
18881 @opindex mno-mcount-ra-address
18882 Emit (do not emit) code that allows @code{_mcount} to modify the
18883 calling function's return address. When enabled, this option extends
18884 the usual @code{_mcount} interface with a new @var{ra-address}
18885 parameter, which has type @code{intptr_t *} and is passed in register
18886 @code{$12}. @code{_mcount} can then modify the return address by
18887 doing both of the following:
18890 Returning the new address in register @code{$31}.
18892 Storing the new address in @code{*@var{ra-address}},
18893 if @var{ra-address} is nonnull.
18896 The default is @option{-mno-mcount-ra-address}.
18898 @item -mframe-header-opt
18899 @itemx -mno-frame-header-opt
18900 @opindex mframe-header-opt
18901 Enable (disable) frame header optimization in the o32 ABI. When using the
18902 o32 ABI, calling functions will allocate 16 bytes on the stack for the called
18903 function to write out register arguments. When enabled, this optimization
18904 will suppress the allocation of the frame header if it can be determined that
18907 This optimization is off by default at all optimization levels.
18912 @subsection MMIX Options
18913 @cindex MMIX Options
18915 These options are defined for the MMIX:
18919 @itemx -mno-libfuncs
18921 @opindex mno-libfuncs
18922 Specify that intrinsic library functions are being compiled, passing all
18923 values in registers, no matter the size.
18926 @itemx -mno-epsilon
18928 @opindex mno-epsilon
18929 Generate floating-point comparison instructions that compare with respect
18930 to the @code{rE} epsilon register.
18932 @item -mabi=mmixware
18934 @opindex mabi=mmixware
18936 Generate code that passes function parameters and return values that (in
18937 the called function) are seen as registers @code{$0} and up, as opposed to
18938 the GNU ABI which uses global registers @code{$231} and up.
18940 @item -mzero-extend
18941 @itemx -mno-zero-extend
18942 @opindex mzero-extend
18943 @opindex mno-zero-extend
18944 When reading data from memory in sizes shorter than 64 bits, use (do not
18945 use) zero-extending load instructions by default, rather than
18946 sign-extending ones.
18949 @itemx -mno-knuthdiv
18951 @opindex mno-knuthdiv
18952 Make the result of a division yielding a remainder have the same sign as
18953 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
18954 remainder follows the sign of the dividend. Both methods are
18955 arithmetically valid, the latter being almost exclusively used.
18957 @item -mtoplevel-symbols
18958 @itemx -mno-toplevel-symbols
18959 @opindex mtoplevel-symbols
18960 @opindex mno-toplevel-symbols
18961 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
18962 code can be used with the @code{PREFIX} assembly directive.
18966 Generate an executable in the ELF format, rather than the default
18967 @samp{mmo} format used by the @command{mmix} simulator.
18969 @item -mbranch-predict
18970 @itemx -mno-branch-predict
18971 @opindex mbranch-predict
18972 @opindex mno-branch-predict
18973 Use (do not use) the probable-branch instructions, when static branch
18974 prediction indicates a probable branch.
18976 @item -mbase-addresses
18977 @itemx -mno-base-addresses
18978 @opindex mbase-addresses
18979 @opindex mno-base-addresses
18980 Generate (do not generate) code that uses @emph{base addresses}. Using a
18981 base address automatically generates a request (handled by the assembler
18982 and the linker) for a constant to be set up in a global register. The
18983 register is used for one or more base address requests within the range 0
18984 to 255 from the value held in the register. The generally leads to short
18985 and fast code, but the number of different data items that can be
18986 addressed is limited. This means that a program that uses lots of static
18987 data may require @option{-mno-base-addresses}.
18989 @item -msingle-exit
18990 @itemx -mno-single-exit
18991 @opindex msingle-exit
18992 @opindex mno-single-exit
18993 Force (do not force) generated code to have a single exit point in each
18997 @node MN10300 Options
18998 @subsection MN10300 Options
18999 @cindex MN10300 options
19001 These @option{-m} options are defined for Matsushita MN10300 architectures:
19006 Generate code to avoid bugs in the multiply instructions for the MN10300
19007 processors. This is the default.
19009 @item -mno-mult-bug
19010 @opindex mno-mult-bug
19011 Do not generate code to avoid bugs in the multiply instructions for the
19012 MN10300 processors.
19016 Generate code using features specific to the AM33 processor.
19020 Do not generate code using features specific to the AM33 processor. This
19025 Generate code using features specific to the AM33/2.0 processor.
19029 Generate code using features specific to the AM34 processor.
19031 @item -mtune=@var{cpu-type}
19033 Use the timing characteristics of the indicated CPU type when
19034 scheduling instructions. This does not change the targeted processor
19035 type. The CPU type must be one of @samp{mn10300}, @samp{am33},
19036 @samp{am33-2} or @samp{am34}.
19038 @item -mreturn-pointer-on-d0
19039 @opindex mreturn-pointer-on-d0
19040 When generating a function that returns a pointer, return the pointer
19041 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
19042 only in @code{a0}, and attempts to call such functions without a prototype
19043 result in errors. Note that this option is on by default; use
19044 @option{-mno-return-pointer-on-d0} to disable it.
19048 Do not link in the C run-time initialization object file.
19052 Indicate to the linker that it should perform a relaxation optimization pass
19053 to shorten branches, calls and absolute memory addresses. This option only
19054 has an effect when used on the command line for the final link step.
19056 This option makes symbolic debugging impossible.
19060 Allow the compiler to generate @emph{Long Instruction Word}
19061 instructions if the target is the @samp{AM33} or later. This is the
19062 default. This option defines the preprocessor macro @code{__LIW__}.
19066 Do not allow the compiler to generate @emph{Long Instruction Word}
19067 instructions. This option defines the preprocessor macro
19072 Allow the compiler to generate the @emph{SETLB} and @emph{Lcc}
19073 instructions if the target is the @samp{AM33} or later. This is the
19074 default. This option defines the preprocessor macro @code{__SETLB__}.
19078 Do not allow the compiler to generate @emph{SETLB} or @emph{Lcc}
19079 instructions. This option defines the preprocessor macro
19080 @code{__NO_SETLB__}.
19084 @node Moxie Options
19085 @subsection Moxie Options
19086 @cindex Moxie Options
19092 Generate big-endian code. This is the default for @samp{moxie-*-*}
19097 Generate little-endian code.
19101 Generate mul.x and umul.x instructions. This is the default for
19102 @samp{moxiebox-*-*} configurations.
19106 Do not link in the C run-time initialization object file.
19110 @node MSP430 Options
19111 @subsection MSP430 Options
19112 @cindex MSP430 Options
19114 These options are defined for the MSP430:
19120 Force assembly output to always use hex constants. Normally such
19121 constants are signed decimals, but this option is available for
19122 testsuite and/or aesthetic purposes.
19126 Select the MCU to target. This is used to create a C preprocessor
19127 symbol based upon the MCU name, converted to upper case and pre- and
19128 post-fixed with @samp{__}. This in turn is used by the
19129 @file{msp430.h} header file to select an MCU-specific supplementary
19132 The option also sets the ISA to use. If the MCU name is one that is
19133 known to only support the 430 ISA then that is selected, otherwise the
19134 430X ISA is selected. A generic MCU name of @samp{msp430} can also be
19135 used to select the 430 ISA. Similarly the generic @samp{msp430x} MCU
19136 name selects the 430X ISA.
19138 In addition an MCU-specific linker script is added to the linker
19139 command line. The script's name is the name of the MCU with
19140 @file{.ld} appended. Thus specifying @option{-mmcu=xxx} on the @command{gcc}
19141 command line defines the C preprocessor symbol @code{__XXX__} and
19142 cause the linker to search for a script called @file{xxx.ld}.
19144 This option is also passed on to the assembler.
19147 @itemx -mno-warn-mcu
19149 @opindex mno-warn-mcu
19150 This option enables or disables warnings about conflicts between the
19151 MCU name specified by the @option{-mmcu} option and the ISA set by the
19152 @option{-mcpu} option and/or the hardware multiply support set by the
19153 @option{-mhwmult} option. It also toggles warnings about unrecognized
19154 MCU names. This option is on by default.
19158 Specifies the ISA to use. Accepted values are @samp{msp430},
19159 @samp{msp430x} and @samp{msp430xv2}. This option is deprecated. The
19160 @option{-mmcu=} option should be used to select the ISA.
19164 Link to the simulator runtime libraries and linker script. Overrides
19165 any scripts that would be selected by the @option{-mmcu=} option.
19169 Use large-model addressing (20-bit pointers, 32-bit @code{size_t}).
19173 Use small-model addressing (16-bit pointers, 16-bit @code{size_t}).
19177 This option is passed to the assembler and linker, and allows the
19178 linker to perform certain optimizations that cannot be done until
19183 Describes the type of hardware multiply supported by the target.
19184 Accepted values are @samp{none} for no hardware multiply, @samp{16bit}
19185 for the original 16-bit-only multiply supported by early MCUs.
19186 @samp{32bit} for the 16/32-bit multiply supported by later MCUs and
19187 @samp{f5series} for the 16/32-bit multiply supported by F5-series MCUs.
19188 A value of @samp{auto} can also be given. This tells GCC to deduce
19189 the hardware multiply support based upon the MCU name provided by the
19190 @option{-mmcu} option. If no @option{-mmcu} option is specified or if
19191 the MCU name is not recognized then no hardware multiply support is
19192 assumed. @code{auto} is the default setting.
19194 Hardware multiplies are normally performed by calling a library
19195 routine. This saves space in the generated code. When compiling at
19196 @option{-O3} or higher however the hardware multiplier is invoked
19197 inline. This makes for bigger, but faster code.
19199 The hardware multiply routines disable interrupts whilst running and
19200 restore the previous interrupt state when they finish. This makes
19201 them safe to use inside interrupt handlers as well as in normal code.
19205 Enable the use of a minimum runtime environment - no static
19206 initializers or constructors. This is intended for memory-constrained
19207 devices. The compiler includes special symbols in some objects
19208 that tell the linker and runtime which code fragments are required.
19210 @item -mcode-region=
19211 @itemx -mdata-region=
19212 @opindex mcode-region
19213 @opindex mdata-region
19214 These options tell the compiler where to place functions and data that
19215 do not have one of the @code{lower}, @code{upper}, @code{either} or
19216 @code{section} attributes. Possible values are @code{lower},
19217 @code{upper}, @code{either} or @code{any}. The first three behave
19218 like the corresponding attribute. The fourth possible value -
19219 @code{any} - is the default. It leaves placement entirely up to the
19220 linker script and how it assigns the standard sections
19221 (@code{.text}, @code{.data}, etc) to the memory regions.
19223 @item -msilicon-errata=
19224 @opindex msilicon-errata
19225 This option passes on a request to assembler to enable the fixes for
19226 the named silicon errata.
19228 @item -msilicon-errata-warn=
19229 @opindex msilicon-errata-warn
19230 This option passes on a request to the assembler to enable warning
19231 messages when a silicon errata might need to be applied.
19235 @node NDS32 Options
19236 @subsection NDS32 Options
19237 @cindex NDS32 Options
19239 These options are defined for NDS32 implementations:
19244 @opindex mbig-endian
19245 Generate code in big-endian mode.
19247 @item -mlittle-endian
19248 @opindex mlittle-endian
19249 Generate code in little-endian mode.
19251 @item -mreduced-regs
19252 @opindex mreduced-regs
19253 Use reduced-set registers for register allocation.
19256 @opindex mfull-regs
19257 Use full-set registers for register allocation.
19261 Generate conditional move instructions.
19265 Do not generate conditional move instructions.
19269 Generate performance extension instructions.
19271 @item -mno-perf-ext
19272 @opindex mno-perf-ext
19273 Do not generate performance extension instructions.
19277 Generate v3 push25/pop25 instructions.
19280 @opindex mno-v3push
19281 Do not generate v3 push25/pop25 instructions.
19285 Generate 16-bit instructions.
19288 @opindex mno-16-bit
19289 Do not generate 16-bit instructions.
19291 @item -misr-vector-size=@var{num}
19292 @opindex misr-vector-size
19293 Specify the size of each interrupt vector, which must be 4 or 16.
19295 @item -mcache-block-size=@var{num}
19296 @opindex mcache-block-size
19297 Specify the size of each cache block,
19298 which must be a power of 2 between 4 and 512.
19300 @item -march=@var{arch}
19302 Specify the name of the target architecture.
19304 @item -mcmodel=@var{code-model}
19306 Set the code model to one of
19309 All the data and read-only data segments must be within 512KB addressing space.
19310 The text segment must be within 16MB addressing space.
19311 @item @samp{medium}
19312 The data segment must be within 512KB while the read-only data segment can be
19313 within 4GB addressing space. The text segment should be still within 16MB
19316 All the text and data segments can be within 4GB addressing space.
19320 @opindex mctor-dtor
19321 Enable constructor/destructor feature.
19325 Guide linker to relax instructions.
19329 @node Nios II Options
19330 @subsection Nios II Options
19331 @cindex Nios II options
19332 @cindex Altera Nios II options
19334 These are the options defined for the Altera Nios II processor.
19340 @cindex smaller data references
19341 Put global and static objects less than or equal to @var{num} bytes
19342 into the small data or BSS sections instead of the normal data or BSS
19343 sections. The default value of @var{num} is 8.
19345 @item -mgpopt=@var{option}
19350 Generate (do not generate) GP-relative accesses. The following
19351 @var{option} names are recognized:
19356 Do not generate GP-relative accesses.
19359 Generate GP-relative accesses for small data objects that are not
19360 external, weak, or uninitialized common symbols.
19361 Also use GP-relative addressing for objects that
19362 have been explicitly placed in a small data section via a @code{section}
19366 As for @samp{local}, but also generate GP-relative accesses for
19367 small data objects that are external, weak, or common. If you use this option,
19368 you must ensure that all parts of your program (including libraries) are
19369 compiled with the same @option{-G} setting.
19372 Generate GP-relative accesses for all data objects in the program. If you
19373 use this option, the entire data and BSS segments
19374 of your program must fit in 64K of memory and you must use an appropriate
19375 linker script to allocate them within the addressable range of the
19379 Generate GP-relative addresses for function pointers as well as data
19380 pointers. If you use this option, the entire text, data, and BSS segments
19381 of your program must fit in 64K of memory and you must use an appropriate
19382 linker script to allocate them within the addressable range of the
19387 @option{-mgpopt} is equivalent to @option{-mgpopt=local}, and
19388 @option{-mno-gpopt} is equivalent to @option{-mgpopt=none}.
19390 The default is @option{-mgpopt} except when @option{-fpic} or
19391 @option{-fPIC} is specified to generate position-independent code.
19392 Note that the Nios II ABI does not permit GP-relative accesses from
19395 You may need to specify @option{-mno-gpopt} explicitly when building
19396 programs that include large amounts of small data, including large
19397 GOT data sections. In this case, the 16-bit offset for GP-relative
19398 addressing may not be large enough to allow access to the entire
19399 small data section.
19405 Generate little-endian (default) or big-endian (experimental) code,
19408 @item -march=@var{arch}
19410 This specifies the name of the target Nios II architecture. GCC uses this
19411 name to determine what kind of instructions it can emit when generating
19412 assembly code. Permissible names are: @samp{r1}, @samp{r2}.
19414 The preprocessor macro @code{__nios2_arch__} is available to programs,
19415 with value 1 or 2, indicating the targeted ISA level.
19417 @item -mbypass-cache
19418 @itemx -mno-bypass-cache
19419 @opindex mno-bypass-cache
19420 @opindex mbypass-cache
19421 Force all load and store instructions to always bypass cache by
19422 using I/O variants of the instructions. The default is not to
19425 @item -mno-cache-volatile
19426 @itemx -mcache-volatile
19427 @opindex mcache-volatile
19428 @opindex mno-cache-volatile
19429 Volatile memory access bypass the cache using the I/O variants of
19430 the load and store instructions. The default is not to bypass the cache.
19432 @item -mno-fast-sw-div
19433 @itemx -mfast-sw-div
19434 @opindex mno-fast-sw-div
19435 @opindex mfast-sw-div
19436 Do not use table-based fast divide for small numbers. The default
19437 is to use the fast divide at @option{-O3} and above.
19441 @itemx -mno-hw-mulx
19445 @opindex mno-hw-mul
19447 @opindex mno-hw-mulx
19449 @opindex mno-hw-div
19451 Enable or disable emitting @code{mul}, @code{mulx} and @code{div} family of
19452 instructions by the compiler. The default is to emit @code{mul}
19453 and not emit @code{div} and @code{mulx}.
19459 Enable or disable generation of Nios II R2 BMX (bit manipulation) and
19460 CDX (code density) instructions. Enabling these instructions also
19461 requires @option{-march=r2}. Since these instructions are optional
19462 extensions to the R2 architecture, the default is not to emit them.
19464 @item -mcustom-@var{insn}=@var{N}
19465 @itemx -mno-custom-@var{insn}
19466 @opindex mcustom-@var{insn}
19467 @opindex mno-custom-@var{insn}
19468 Each @option{-mcustom-@var{insn}=@var{N}} option enables use of a
19469 custom instruction with encoding @var{N} when generating code that uses
19470 @var{insn}. For example, @option{-mcustom-fadds=253} generates custom
19471 instruction 253 for single-precision floating-point add operations instead
19472 of the default behavior of using a library call.
19474 The following values of @var{insn} are supported. Except as otherwise
19475 noted, floating-point operations are expected to be implemented with
19476 normal IEEE 754 semantics and correspond directly to the C operators or the
19477 equivalent GCC built-in functions (@pxref{Other Builtins}).
19479 Single-precision floating point:
19482 @item @samp{fadds}, @samp{fsubs}, @samp{fdivs}, @samp{fmuls}
19483 Binary arithmetic operations.
19489 Unary absolute value.
19491 @item @samp{fcmpeqs}, @samp{fcmpges}, @samp{fcmpgts}, @samp{fcmples}, @samp{fcmplts}, @samp{fcmpnes}
19492 Comparison operations.
19494 @item @samp{fmins}, @samp{fmaxs}
19495 Floating-point minimum and maximum. These instructions are only
19496 generated if @option{-ffinite-math-only} is specified.
19498 @item @samp{fsqrts}
19499 Unary square root operation.
19501 @item @samp{fcoss}, @samp{fsins}, @samp{ftans}, @samp{fatans}, @samp{fexps}, @samp{flogs}
19502 Floating-point trigonometric and exponential functions. These instructions
19503 are only generated if @option{-funsafe-math-optimizations} is also specified.
19507 Double-precision floating point:
19510 @item @samp{faddd}, @samp{fsubd}, @samp{fdivd}, @samp{fmuld}
19511 Binary arithmetic operations.
19517 Unary absolute value.
19519 @item @samp{fcmpeqd}, @samp{fcmpged}, @samp{fcmpgtd}, @samp{fcmpled}, @samp{fcmpltd}, @samp{fcmpned}
19520 Comparison operations.
19522 @item @samp{fmind}, @samp{fmaxd}
19523 Double-precision minimum and maximum. These instructions are only
19524 generated if @option{-ffinite-math-only} is specified.
19526 @item @samp{fsqrtd}
19527 Unary square root operation.
19529 @item @samp{fcosd}, @samp{fsind}, @samp{ftand}, @samp{fatand}, @samp{fexpd}, @samp{flogd}
19530 Double-precision trigonometric and exponential functions. These instructions
19531 are only generated if @option{-funsafe-math-optimizations} is also specified.
19537 @item @samp{fextsd}
19538 Conversion from single precision to double precision.
19540 @item @samp{ftruncds}
19541 Conversion from double precision to single precision.
19543 @item @samp{fixsi}, @samp{fixsu}, @samp{fixdi}, @samp{fixdu}
19544 Conversion from floating point to signed or unsigned integer types, with
19545 truncation towards zero.
19548 Conversion from single-precision floating point to signed integer,
19549 rounding to the nearest integer and ties away from zero.
19550 This corresponds to the @code{__builtin_lroundf} function when
19551 @option{-fno-math-errno} is used.
19553 @item @samp{floatis}, @samp{floatus}, @samp{floatid}, @samp{floatud}
19554 Conversion from signed or unsigned integer types to floating-point types.
19558 In addition, all of the following transfer instructions for internal
19559 registers X and Y must be provided to use any of the double-precision
19560 floating-point instructions. Custom instructions taking two
19561 double-precision source operands expect the first operand in the
19562 64-bit register X. The other operand (or only operand of a unary
19563 operation) is given to the custom arithmetic instruction with the
19564 least significant half in source register @var{src1} and the most
19565 significant half in @var{src2}. A custom instruction that returns a
19566 double-precision result returns the most significant 32 bits in the
19567 destination register and the other half in 32-bit register Y.
19568 GCC automatically generates the necessary code sequences to write
19569 register X and/or read register Y when double-precision floating-point
19570 instructions are used.
19575 Write @var{src1} into the least significant half of X and @var{src2} into
19576 the most significant half of X.
19579 Write @var{src1} into Y.
19581 @item @samp{frdxhi}, @samp{frdxlo}
19582 Read the most or least (respectively) significant half of X and store it in
19586 Read the value of Y and store it into @var{dest}.
19589 Note that you can gain more local control over generation of Nios II custom
19590 instructions by using the @code{target("custom-@var{insn}=@var{N}")}
19591 and @code{target("no-custom-@var{insn}")} function attributes
19592 (@pxref{Function Attributes})
19593 or pragmas (@pxref{Function Specific Option Pragmas}).
19595 @item -mcustom-fpu-cfg=@var{name}
19596 @opindex mcustom-fpu-cfg
19598 This option enables a predefined, named set of custom instruction encodings
19599 (see @option{-mcustom-@var{insn}} above).
19600 Currently, the following sets are defined:
19602 @option{-mcustom-fpu-cfg=60-1} is equivalent to:
19603 @gccoptlist{-mcustom-fmuls=252 @gol
19604 -mcustom-fadds=253 @gol
19605 -mcustom-fsubs=254 @gol
19606 -fsingle-precision-constant}
19608 @option{-mcustom-fpu-cfg=60-2} is equivalent to:
19609 @gccoptlist{-mcustom-fmuls=252 @gol
19610 -mcustom-fadds=253 @gol
19611 -mcustom-fsubs=254 @gol
19612 -mcustom-fdivs=255 @gol
19613 -fsingle-precision-constant}
19615 @option{-mcustom-fpu-cfg=72-3} is equivalent to:
19616 @gccoptlist{-mcustom-floatus=243 @gol
19617 -mcustom-fixsi=244 @gol
19618 -mcustom-floatis=245 @gol
19619 -mcustom-fcmpgts=246 @gol
19620 -mcustom-fcmples=249 @gol
19621 -mcustom-fcmpeqs=250 @gol
19622 -mcustom-fcmpnes=251 @gol
19623 -mcustom-fmuls=252 @gol
19624 -mcustom-fadds=253 @gol
19625 -mcustom-fsubs=254 @gol
19626 -mcustom-fdivs=255 @gol
19627 -fsingle-precision-constant}
19629 Custom instruction assignments given by individual
19630 @option{-mcustom-@var{insn}=} options override those given by
19631 @option{-mcustom-fpu-cfg=}, regardless of the
19632 order of the options on the command line.
19634 Note that you can gain more local control over selection of a FPU
19635 configuration by using the @code{target("custom-fpu-cfg=@var{name}")}
19636 function attribute (@pxref{Function Attributes})
19637 or pragma (@pxref{Function Specific Option Pragmas}).
19641 These additional @samp{-m} options are available for the Altera Nios II
19642 ELF (bare-metal) target:
19648 Link with HAL BSP. This suppresses linking with the GCC-provided C runtime
19649 startup and termination code, and is typically used in conjunction with
19650 @option{-msys-crt0=} to specify the location of the alternate startup code
19651 provided by the HAL BSP.
19655 Link with a limited version of the C library, @option{-lsmallc}, rather than
19658 @item -msys-crt0=@var{startfile}
19660 @var{startfile} is the file name of the startfile (crt0) to use
19661 when linking. This option is only useful in conjunction with @option{-mhal}.
19663 @item -msys-lib=@var{systemlib}
19665 @var{systemlib} is the library name of the library that provides
19666 low-level system calls required by the C library,
19667 e.g. @code{read} and @code{write}.
19668 This option is typically used to link with a library provided by a HAL BSP.
19672 @node Nvidia PTX Options
19673 @subsection Nvidia PTX Options
19674 @cindex Nvidia PTX options
19675 @cindex nvptx options
19677 These options are defined for Nvidia PTX:
19685 Generate code for 32-bit or 64-bit ABI.
19688 @opindex mmainkernel
19689 Link in code for a __main kernel. This is for stand-alone instead of
19690 offloading execution.
19694 Apply partitioned execution optimizations. This is the default when any
19695 level of optimization is selected.
19699 @node PDP-11 Options
19700 @subsection PDP-11 Options
19701 @cindex PDP-11 Options
19703 These options are defined for the PDP-11:
19708 Use hardware FPP floating point. This is the default. (FIS floating
19709 point on the PDP-11/40 is not supported.)
19712 @opindex msoft-float
19713 Do not use hardware floating point.
19717 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
19721 Return floating-point results in memory. This is the default.
19725 Generate code for a PDP-11/40.
19729 Generate code for a PDP-11/45. This is the default.
19733 Generate code for a PDP-11/10.
19735 @item -mbcopy-builtin
19736 @opindex mbcopy-builtin
19737 Use inline @code{movmemhi} patterns for copying memory. This is the
19742 Do not use inline @code{movmemhi} patterns for copying memory.
19748 Use 16-bit @code{int}. This is the default.
19754 Use 32-bit @code{int}.
19757 @itemx -mno-float32
19759 @opindex mno-float32
19760 Use 64-bit @code{float}. This is the default.
19763 @itemx -mno-float64
19765 @opindex mno-float64
19766 Use 32-bit @code{float}.
19770 Use @code{abshi2} pattern. This is the default.
19774 Do not use @code{abshi2} pattern.
19776 @item -mbranch-expensive
19777 @opindex mbranch-expensive
19778 Pretend that branches are expensive. This is for experimenting with
19779 code generation only.
19781 @item -mbranch-cheap
19782 @opindex mbranch-cheap
19783 Do not pretend that branches are expensive. This is the default.
19787 Use Unix assembler syntax. This is the default when configured for
19788 @samp{pdp11-*-bsd}.
19792 Use DEC assembler syntax. This is the default when configured for any
19793 PDP-11 target other than @samp{pdp11-*-bsd}.
19796 @node picoChip Options
19797 @subsection picoChip Options
19798 @cindex picoChip options
19800 These @samp{-m} options are defined for picoChip implementations:
19804 @item -mae=@var{ae_type}
19806 Set the instruction set, register set, and instruction scheduling
19807 parameters for array element type @var{ae_type}. Supported values
19808 for @var{ae_type} are @samp{ANY}, @samp{MUL}, and @samp{MAC}.
19810 @option{-mae=ANY} selects a completely generic AE type. Code
19811 generated with this option runs on any of the other AE types. The
19812 code is not as efficient as it would be if compiled for a specific
19813 AE type, and some types of operation (e.g., multiplication) do not
19814 work properly on all types of AE.
19816 @option{-mae=MUL} selects a MUL AE type. This is the most useful AE type
19817 for compiled code, and is the default.
19819 @option{-mae=MAC} selects a DSP-style MAC AE. Code compiled with this
19820 option may suffer from poor performance of byte (char) manipulation,
19821 since the DSP AE does not provide hardware support for byte load/stores.
19823 @item -msymbol-as-address
19824 Enable the compiler to directly use a symbol name as an address in a
19825 load/store instruction, without first loading it into a
19826 register. Typically, the use of this option generates larger
19827 programs, which run faster than when the option isn't used. However, the
19828 results vary from program to program, so it is left as a user option,
19829 rather than being permanently enabled.
19831 @item -mno-inefficient-warnings
19832 Disables warnings about the generation of inefficient code. These
19833 warnings can be generated, for example, when compiling code that
19834 performs byte-level memory operations on the MAC AE type. The MAC AE has
19835 no hardware support for byte-level memory operations, so all byte
19836 load/stores must be synthesized from word load/store operations. This is
19837 inefficient and a warning is generated to indicate
19838 that you should rewrite the code to avoid byte operations, or to target
19839 an AE type that has the necessary hardware support. This option disables
19844 @node PowerPC Options
19845 @subsection PowerPC Options
19846 @cindex PowerPC options
19848 These are listed under @xref{RS/6000 and PowerPC Options}.
19851 @subsection RL78 Options
19852 @cindex RL78 Options
19858 Links in additional target libraries to support operation within a
19867 Specifies the type of hardware multiplication and division support to
19868 be used. The simplest is @code{none}, which uses software for both
19869 multiplication and division. This is the default. The @code{g13}
19870 value is for the hardware multiply/divide peripheral found on the
19871 RL78/G13 (S2 core) targets. The @code{g14} value selects the use of
19872 the multiplication and division instructions supported by the RL78/G14
19873 (S3 core) parts. The value @code{rl78} is an alias for @code{g14} and
19874 the value @code{mg10} is an alias for @code{none}.
19876 In addition a C preprocessor macro is defined, based upon the setting
19877 of this option. Possible values are: @code{__RL78_MUL_NONE__},
19878 @code{__RL78_MUL_G13__} or @code{__RL78_MUL_G14__}.
19885 Specifies the RL78 core to target. The default is the G14 core, also
19886 known as an S3 core or just RL78. The G13 or S2 core does not have
19887 multiply or divide instructions, instead it uses a hardware peripheral
19888 for these operations. The G10 or S1 core does not have register
19889 banks, so it uses a different calling convention.
19891 If this option is set it also selects the type of hardware multiply
19892 support to use, unless this is overridden by an explicit
19893 @option{-mmul=none} option on the command line. Thus specifying
19894 @option{-mcpu=g13} enables the use of the G13 hardware multiply
19895 peripheral and specifying @option{-mcpu=g10} disables the use of
19896 hardware multiplications altogether.
19898 Note, although the RL78/G14 core is the default target, specifying
19899 @option{-mcpu=g14} or @option{-mcpu=rl78} on the command line does
19900 change the behavior of the toolchain since it also enables G14
19901 hardware multiply support. If these options are not specified on the
19902 command line then software multiplication routines will be used even
19903 though the code targets the RL78 core. This is for backwards
19904 compatibility with older toolchains which did not have hardware
19905 multiply and divide support.
19907 In addition a C preprocessor macro is defined, based upon the setting
19908 of this option. Possible values are: @code{__RL78_G10__},
19909 @code{__RL78_G13__} or @code{__RL78_G14__}.
19919 These are aliases for the corresponding @option{-mcpu=} option. They
19920 are provided for backwards compatibility.
19924 Allow the compiler to use all of the available registers. By default
19925 registers @code{r24..r31} are reserved for use in interrupt handlers.
19926 With this option enabled these registers can be used in ordinary
19929 @item -m64bit-doubles
19930 @itemx -m32bit-doubles
19931 @opindex m64bit-doubles
19932 @opindex m32bit-doubles
19933 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
19934 or 32 bits (@option{-m32bit-doubles}) in size. The default is
19935 @option{-m32bit-doubles}.
19937 @item -msave-mduc-in-interrupts
19938 @item -mno-save-mduc-in-interrupts
19939 @opindex msave-mduc-in-interrupts
19940 @opindex mno-save-mduc-in-interrupts
19941 Specifies that interrupt handler functions should preserve the
19942 MDUC registers. This is only necessary if normal code might use
19943 the MDUC registers, for example because it performs multiplication
19944 and division operations. The default is to ignore the MDUC registers
19945 as this makes the interrupt handlers faster. The target option -mg13
19946 needs to be passed for this to work as this feature is only available
19947 on the G13 target (S2 core). The MDUC registers will only be saved
19948 if the interrupt handler performs a multiplication or division
19949 operation or it calls another function.
19953 @node RS/6000 and PowerPC Options
19954 @subsection IBM RS/6000 and PowerPC Options
19955 @cindex RS/6000 and PowerPC Options
19956 @cindex IBM RS/6000 and PowerPC Options
19958 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
19960 @item -mpowerpc-gpopt
19961 @itemx -mno-powerpc-gpopt
19962 @itemx -mpowerpc-gfxopt
19963 @itemx -mno-powerpc-gfxopt
19966 @itemx -mno-powerpc64
19970 @itemx -mno-popcntb
19972 @itemx -mno-popcntd
19981 @itemx -mno-hard-dfp
19982 @opindex mpowerpc-gpopt
19983 @opindex mno-powerpc-gpopt
19984 @opindex mpowerpc-gfxopt
19985 @opindex mno-powerpc-gfxopt
19986 @opindex mpowerpc64
19987 @opindex mno-powerpc64
19991 @opindex mno-popcntb
19993 @opindex mno-popcntd
19999 @opindex mno-mfpgpr
20001 @opindex mno-hard-dfp
20002 You use these options to specify which instructions are available on the
20003 processor you are using. The default value of these options is
20004 determined when configuring GCC@. Specifying the
20005 @option{-mcpu=@var{cpu_type}} overrides the specification of these
20006 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
20007 rather than the options listed above.
20009 Specifying @option{-mpowerpc-gpopt} allows
20010 GCC to use the optional PowerPC architecture instructions in the
20011 General Purpose group, including floating-point square root. Specifying
20012 @option{-mpowerpc-gfxopt} allows GCC to
20013 use the optional PowerPC architecture instructions in the Graphics
20014 group, including floating-point select.
20016 The @option{-mmfcrf} option allows GCC to generate the move from
20017 condition register field instruction implemented on the POWER4
20018 processor and other processors that support the PowerPC V2.01
20020 The @option{-mpopcntb} option allows GCC to generate the popcount and
20021 double-precision FP reciprocal estimate instruction implemented on the
20022 POWER5 processor and other processors that support the PowerPC V2.02
20024 The @option{-mpopcntd} option allows GCC to generate the popcount
20025 instruction implemented on the POWER7 processor and other processors
20026 that support the PowerPC V2.06 architecture.
20027 The @option{-mfprnd} option allows GCC to generate the FP round to
20028 integer instructions implemented on the POWER5+ processor and other
20029 processors that support the PowerPC V2.03 architecture.
20030 The @option{-mcmpb} option allows GCC to generate the compare bytes
20031 instruction implemented on the POWER6 processor and other processors
20032 that support the PowerPC V2.05 architecture.
20033 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
20034 general-purpose register instructions implemented on the POWER6X
20035 processor and other processors that support the extended PowerPC V2.05
20037 The @option{-mhard-dfp} option allows GCC to generate the decimal
20038 floating-point instructions implemented on some POWER processors.
20040 The @option{-mpowerpc64} option allows GCC to generate the additional
20041 64-bit instructions that are found in the full PowerPC64 architecture
20042 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
20043 @option{-mno-powerpc64}.
20045 @item -mcpu=@var{cpu_type}
20047 Set architecture type, register usage, and
20048 instruction scheduling parameters for machine type @var{cpu_type}.
20049 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
20050 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{464}, @samp{464fp},
20051 @samp{476}, @samp{476fp}, @samp{505}, @samp{601}, @samp{602}, @samp{603},
20052 @samp{603e}, @samp{604}, @samp{604e}, @samp{620}, @samp{630}, @samp{740},
20053 @samp{7400}, @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
20054 @samp{860}, @samp{970}, @samp{8540}, @samp{a2}, @samp{e300c2},
20055 @samp{e300c3}, @samp{e500mc}, @samp{e500mc64}, @samp{e5500},
20056 @samp{e6500}, @samp{ec603e}, @samp{G3}, @samp{G4}, @samp{G5},
20057 @samp{titan}, @samp{power3}, @samp{power4}, @samp{power5}, @samp{power5+},
20058 @samp{power6}, @samp{power6x}, @samp{power7}, @samp{power8},
20059 @samp{power9}, @samp{powerpc}, @samp{powerpc64}, @samp{powerpc64le},
20062 @option{-mcpu=powerpc}, @option{-mcpu=powerpc64}, and
20063 @option{-mcpu=powerpc64le} specify pure 32-bit PowerPC (either
20064 endian), 64-bit big endian PowerPC and 64-bit little endian PowerPC
20065 architecture machine types, with an appropriate, generic processor
20066 model assumed for scheduling purposes.
20068 The other options specify a specific processor. Code generated under
20069 those options runs best on that processor, and may not run at all on
20072 The @option{-mcpu} options automatically enable or disable the
20075 @gccoptlist{-maltivec -mfprnd -mhard-float -mmfcrf -mmultiple @gol
20076 -mpopcntb -mpopcntd -mpowerpc64 @gol
20077 -mpowerpc-gpopt -mpowerpc-gfxopt -msingle-float -mdouble-float @gol
20078 -msimple-fpu -mstring -mmulhw -mdlmzb -mmfpgpr -mvsx @gol
20079 -mcrypto -mdirect-move -mhtm -mpower8-fusion -mpower8-vector @gol
20080 -mquad-memory -mquad-memory-atomic -mmodulo -mfloat128 -mfloat128-hardware @gol
20081 -mpower9-fusion -mpower9-vector -mpower9-dform}
20083 The particular options set for any particular CPU varies between
20084 compiler versions, depending on what setting seems to produce optimal
20085 code for that CPU; it doesn't necessarily reflect the actual hardware's
20086 capabilities. If you wish to set an individual option to a particular
20087 value, you may specify it after the @option{-mcpu} option, like
20088 @option{-mcpu=970 -mno-altivec}.
20090 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
20091 not enabled or disabled by the @option{-mcpu} option at present because
20092 AIX does not have full support for these options. You may still
20093 enable or disable them individually if you're sure it'll work in your
20096 @item -mtune=@var{cpu_type}
20098 Set the instruction scheduling parameters for machine type
20099 @var{cpu_type}, but do not set the architecture type or register usage,
20100 as @option{-mcpu=@var{cpu_type}} does. The same
20101 values for @var{cpu_type} are used for @option{-mtune} as for
20102 @option{-mcpu}. If both are specified, the code generated uses the
20103 architecture and registers set by @option{-mcpu}, but the
20104 scheduling parameters set by @option{-mtune}.
20106 @item -mcmodel=small
20107 @opindex mcmodel=small
20108 Generate PowerPC64 code for the small model: The TOC is limited to
20111 @item -mcmodel=medium
20112 @opindex mcmodel=medium
20113 Generate PowerPC64 code for the medium model: The TOC and other static
20114 data may be up to a total of 4G in size.
20116 @item -mcmodel=large
20117 @opindex mcmodel=large
20118 Generate PowerPC64 code for the large model: The TOC may be up to 4G
20119 in size. Other data and code is only limited by the 64-bit address
20123 @itemx -mno-altivec
20125 @opindex mno-altivec
20126 Generate code that uses (does not use) AltiVec instructions, and also
20127 enable the use of built-in functions that allow more direct access to
20128 the AltiVec instruction set. You may also need to set
20129 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
20132 When @option{-maltivec} is used, rather than @option{-maltivec=le} or
20133 @option{-maltivec=be}, the element order for AltiVec intrinsics such
20134 as @code{vec_splat}, @code{vec_extract}, and @code{vec_insert}
20135 match array element order corresponding to the endianness of the
20136 target. That is, element zero identifies the leftmost element in a
20137 vector register when targeting a big-endian platform, and identifies
20138 the rightmost element in a vector register when targeting a
20139 little-endian platform.
20142 @opindex maltivec=be
20143 Generate AltiVec instructions using big-endian element order,
20144 regardless of whether the target is big- or little-endian. This is
20145 the default when targeting a big-endian platform.
20147 The element order is used to interpret element numbers in AltiVec
20148 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20149 @code{vec_insert}. By default, these match array element order
20150 corresponding to the endianness for the target.
20153 @opindex maltivec=le
20154 Generate AltiVec instructions using little-endian element order,
20155 regardless of whether the target is big- or little-endian. This is
20156 the default when targeting a little-endian platform. This option is
20157 currently ignored when targeting a big-endian platform.
20159 The element order is used to interpret element numbers in AltiVec
20160 intrinsics such as @code{vec_splat}, @code{vec_extract}, and
20161 @code{vec_insert}. By default, these match array element order
20162 corresponding to the endianness for the target.
20167 @opindex mno-vrsave
20168 Generate VRSAVE instructions when generating AltiVec code.
20170 @item -mgen-cell-microcode
20171 @opindex mgen-cell-microcode
20172 Generate Cell microcode instructions.
20174 @item -mwarn-cell-microcode
20175 @opindex mwarn-cell-microcode
20176 Warn when a Cell microcode instruction is emitted. An example
20177 of a Cell microcode instruction is a variable shift.
20180 @opindex msecure-plt
20181 Generate code that allows @command{ld} and @command{ld.so}
20182 to build executables and shared
20183 libraries with non-executable @code{.plt} and @code{.got} sections.
20185 32-bit SYSV ABI option.
20189 Generate code that uses a BSS @code{.plt} section that @command{ld.so}
20191 requires @code{.plt} and @code{.got}
20192 sections that are both writable and executable.
20193 This is a PowerPC 32-bit SYSV ABI option.
20199 This switch enables or disables the generation of ISEL instructions.
20201 @item -misel=@var{yes/no}
20202 This switch has been deprecated. Use @option{-misel} and
20203 @option{-mno-isel} instead.
20207 Enable Local Register Allocation. This is still experimental for PowerPC,
20208 so by default the compiler uses standard reload
20209 (i.e. @option{-mno-lra}).
20215 This switch enables or disables the generation of SPE simd
20221 @opindex mno-paired
20222 This switch enables or disables the generation of PAIRED simd
20225 @item -mspe=@var{yes/no}
20226 This option has been deprecated. Use @option{-mspe} and
20227 @option{-mno-spe} instead.
20233 Generate code that uses (does not use) vector/scalar (VSX)
20234 instructions, and also enable the use of built-in functions that allow
20235 more direct access to the VSX instruction set.
20240 @opindex mno-crypto
20241 Enable the use (disable) of the built-in functions that allow direct
20242 access to the cryptographic instructions that were added in version
20243 2.07 of the PowerPC ISA.
20245 @item -mdirect-move
20246 @itemx -mno-direct-move
20247 @opindex mdirect-move
20248 @opindex mno-direct-move
20249 Generate code that uses (does not use) the instructions to move data
20250 between the general purpose registers and the vector/scalar (VSX)
20251 registers that were added in version 2.07 of the PowerPC ISA.
20257 Enable (disable) the use of the built-in functions that allow direct
20258 access to the Hardware Transactional Memory (HTM) instructions that
20259 were added in version 2.07 of the PowerPC ISA.
20261 @item -mpower8-fusion
20262 @itemx -mno-power8-fusion
20263 @opindex mpower8-fusion
20264 @opindex mno-power8-fusion
20265 Generate code that keeps (does not keeps) some integer operations
20266 adjacent so that the instructions can be fused together on power8 and
20269 @item -mpower8-vector
20270 @itemx -mno-power8-vector
20271 @opindex mpower8-vector
20272 @opindex mno-power8-vector
20273 Generate code that uses (does not use) the vector and scalar
20274 instructions that were added in version 2.07 of the PowerPC ISA. Also
20275 enable the use of built-in functions that allow more direct access to
20276 the vector instructions.
20278 @item -mquad-memory
20279 @itemx -mno-quad-memory
20280 @opindex mquad-memory
20281 @opindex mno-quad-memory
20282 Generate code that uses (does not use) the non-atomic quad word memory
20283 instructions. The @option{-mquad-memory} option requires use of
20286 @item -mquad-memory-atomic
20287 @itemx -mno-quad-memory-atomic
20288 @opindex mquad-memory-atomic
20289 @opindex mno-quad-memory-atomic
20290 Generate code that uses (does not use) the atomic quad word memory
20291 instructions. The @option{-mquad-memory-atomic} option requires use of
20294 @item -mupper-regs-di
20295 @itemx -mno-upper-regs-di
20296 @opindex mupper-regs-di
20297 @opindex mno-upper-regs-di
20298 Generate code that uses (does not use) the scalar instructions that
20299 target all 64 registers in the vector/scalar floating point register
20300 set that were added in version 2.06 of the PowerPC ISA when processing
20301 integers. @option{-mupper-regs-di} is turned on by default if you use
20302 any of the @option{-mcpu=power7}, @option{-mcpu=power8},
20303 @option{-mcpu=power9}, or @option{-mvsx} options.
20305 @item -mupper-regs-df
20306 @itemx -mno-upper-regs-df
20307 @opindex mupper-regs-df
20308 @opindex mno-upper-regs-df
20309 Generate code that uses (does not use) the scalar double precision
20310 instructions that target all 64 registers in the vector/scalar
20311 floating point register set that were added in version 2.06 of the
20312 PowerPC ISA. @option{-mupper-regs-df} is turned on by default if you
20313 use any of the @option{-mcpu=power7}, @option{-mcpu=power8},
20314 @option{-mcpu=power9}, or @option{-mvsx} options.
20316 @item -mupper-regs-sf
20317 @itemx -mno-upper-regs-sf
20318 @opindex mupper-regs-sf
20319 @opindex mno-upper-regs-sf
20320 Generate code that uses (does not use) the scalar single precision
20321 instructions that target all 64 registers in the vector/scalar
20322 floating point register set that were added in version 2.07 of the
20323 PowerPC ISA. @option{-mupper-regs-sf} is turned on by default if you
20324 use either of the @option{-mcpu=power8}, @option{-mpower8-vector}, or
20325 @option{-mpower9} options.
20328 @itemx -mno-upper-regs
20329 @opindex mupper-regs
20330 @opindex mno-upper-regs
20331 Generate code that uses (does not use) the scalar
20332 instructions that target all 64 registers in the vector/scalar
20333 floating point register set, depending on the model of the machine.
20335 If the @option{-mno-upper-regs} option is used, it turns off both
20336 @option{-mupper-regs-sf} and @option{-mupper-regs-df} options.
20339 @itemx -mno-float128
20341 @opindex mno-float128
20342 Enable/disable the @var{__float128} keyword for IEEE 128-bit floating point
20343 and use either software emulation for IEEE 128-bit floating point or
20344 hardware instructions.
20346 The VSX instruction set (@option{-mvsx}, @option{-mcpu=power7}, or
20347 @option{-mcpu=power8}) must be enabled to use the @option{-mfloat128}
20348 option. The @code{-mfloat128} option only works on PowerPC 64-bit
20351 @item -mfloat128-hardware
20352 @itemx -mno-float128-hardware
20353 @opindex mfloat128-hardware
20354 @opindex mno-float128-hardware
20355 Enable/disable using ISA 3.0 hardware instructions to support the
20356 @var{__float128} data type.
20361 @opindex mno-module
20362 Generate code that uses (does not use) the ISA 3.0 integer modulo
20363 instructions. The @option{-mmodulo} option is enabled by default
20364 with the @option{-mcpu=power9} option.
20366 @item -mpower9-fusion
20367 @itemx -mno-power9-fusion
20368 @opindex mpower9-fusion
20369 @opindex mno-power9-fusion
20370 Generate code that keeps (does not keeps) some operations adjacent so
20371 that the instructions can be fused together on power9 and later
20374 @item -mpower9-vector
20375 @itemx -mno-power9-vector
20376 @opindex mpower9-vector
20377 @opindex mno-power9-vector
20378 Generate code that uses (does not use) the vector and scalar
20379 instructions that were added in version 3.0 of the PowerPC ISA. Also
20380 enable the use of built-in functions that allow more direct access to
20381 the vector instructions.
20383 @item -mpower9-dform
20384 @itemx -mno-power9-dform
20385 @opindex mpower9-dform
20386 @opindex mno-power9-dform
20387 Enable (disable) scalar d-form (register + offset) memory instructions
20388 to load/store traditional Altivec registers. If the @var{LRA} register
20389 allocator is enabled, also enable (disable) vector d-form memory
20392 @item -mfloat-gprs=@var{yes/single/double/no}
20393 @itemx -mfloat-gprs
20394 @opindex mfloat-gprs
20395 This switch enables or disables the generation of floating-point
20396 operations on the general-purpose registers for architectures that
20399 The argument @samp{yes} or @samp{single} enables the use of
20400 single-precision floating-point operations.
20402 The argument @samp{double} enables the use of single and
20403 double-precision floating-point operations.
20405 The argument @samp{no} disables floating-point operations on the
20406 general-purpose registers.
20408 This option is currently only available on the MPC854x.
20414 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
20415 targets (including GNU/Linux). The 32-bit environment sets int, long
20416 and pointer to 32 bits and generates code that runs on any PowerPC
20417 variant. The 64-bit environment sets int to 32 bits and long and
20418 pointer to 64 bits, and generates code for PowerPC64, as for
20419 @option{-mpowerpc64}.
20422 @itemx -mno-fp-in-toc
20423 @itemx -mno-sum-in-toc
20424 @itemx -mminimal-toc
20426 @opindex mno-fp-in-toc
20427 @opindex mno-sum-in-toc
20428 @opindex mminimal-toc
20429 Modify generation of the TOC (Table Of Contents), which is created for
20430 every executable file. The @option{-mfull-toc} option is selected by
20431 default. In that case, GCC allocates at least one TOC entry for
20432 each unique non-automatic variable reference in your program. GCC
20433 also places floating-point constants in the TOC@. However, only
20434 16,384 entries are available in the TOC@.
20436 If you receive a linker error message that saying you have overflowed
20437 the available TOC space, you can reduce the amount of TOC space used
20438 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
20439 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
20440 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
20441 generate code to calculate the sum of an address and a constant at
20442 run time instead of putting that sum into the TOC@. You may specify one
20443 or both of these options. Each causes GCC to produce very slightly
20444 slower and larger code at the expense of conserving TOC space.
20446 If you still run out of space in the TOC even when you specify both of
20447 these options, specify @option{-mminimal-toc} instead. This option causes
20448 GCC to make only one TOC entry for every file. When you specify this
20449 option, GCC produces code that is slower and larger but which
20450 uses extremely little TOC space. You may wish to use this option
20451 only on files that contain less frequently-executed code.
20457 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
20458 @code{long} type, and the infrastructure needed to support them.
20459 Specifying @option{-maix64} implies @option{-mpowerpc64},
20460 while @option{-maix32} disables the 64-bit ABI and
20461 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
20464 @itemx -mno-xl-compat
20465 @opindex mxl-compat
20466 @opindex mno-xl-compat
20467 Produce code that conforms more closely to IBM XL compiler semantics
20468 when using AIX-compatible ABI@. Pass floating-point arguments to
20469 prototyped functions beyond the register save area (RSA) on the stack
20470 in addition to argument FPRs. Do not assume that most significant
20471 double in 128-bit long double value is properly rounded when comparing
20472 values and converting to double. Use XL symbol names for long double
20475 The AIX calling convention was extended but not initially documented to
20476 handle an obscure K&R C case of calling a function that takes the
20477 address of its arguments with fewer arguments than declared. IBM XL
20478 compilers access floating-point arguments that do not fit in the
20479 RSA from the stack when a subroutine is compiled without
20480 optimization. Because always storing floating-point arguments on the
20481 stack is inefficient and rarely needed, this option is not enabled by
20482 default and only is necessary when calling subroutines compiled by IBM
20483 XL compilers without optimization.
20487 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
20488 application written to use message passing with special startup code to
20489 enable the application to run. The system must have PE installed in the
20490 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
20491 must be overridden with the @option{-specs=} option to specify the
20492 appropriate directory location. The Parallel Environment does not
20493 support threads, so the @option{-mpe} option and the @option{-pthread}
20494 option are incompatible.
20496 @item -malign-natural
20497 @itemx -malign-power
20498 @opindex malign-natural
20499 @opindex malign-power
20500 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
20501 @option{-malign-natural} overrides the ABI-defined alignment of larger
20502 types, such as floating-point doubles, on their natural size-based boundary.
20503 The option @option{-malign-power} instructs GCC to follow the ABI-specified
20504 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
20506 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
20510 @itemx -mhard-float
20511 @opindex msoft-float
20512 @opindex mhard-float
20513 Generate code that does not use (uses) the floating-point register set.
20514 Software floating-point emulation is provided if you use the
20515 @option{-msoft-float} option, and pass the option to GCC when linking.
20517 @item -msingle-float
20518 @itemx -mdouble-float
20519 @opindex msingle-float
20520 @opindex mdouble-float
20521 Generate code for single- or double-precision floating-point operations.
20522 @option{-mdouble-float} implies @option{-msingle-float}.
20525 @opindex msimple-fpu
20526 Do not generate @code{sqrt} and @code{div} instructions for hardware
20527 floating-point unit.
20529 @item -mfpu=@var{name}
20531 Specify type of floating-point unit. Valid values for @var{name} are
20532 @samp{sp_lite} (equivalent to @option{-msingle-float -msimple-fpu}),
20533 @samp{dp_lite} (equivalent to @option{-mdouble-float -msimple-fpu}),
20534 @samp{sp_full} (equivalent to @option{-msingle-float}),
20535 and @samp{dp_full} (equivalent to @option{-mdouble-float}).
20538 @opindex mxilinx-fpu
20539 Perform optimizations for the floating-point unit on Xilinx PPC 405/440.
20542 @itemx -mno-multiple
20544 @opindex mno-multiple
20545 Generate code that uses (does not use) the load multiple word
20546 instructions and the store multiple word instructions. These
20547 instructions are generated by default on POWER systems, and not
20548 generated on PowerPC systems. Do not use @option{-mmultiple} on little-endian
20549 PowerPC systems, since those instructions do not work when the
20550 processor is in little-endian mode. The exceptions are PPC740 and
20551 PPC750 which permit these instructions in little-endian mode.
20556 @opindex mno-string
20557 Generate code that uses (does not use) the load string instructions
20558 and the store string word instructions to save multiple registers and
20559 do small block moves. These instructions are generated by default on
20560 POWER systems, and not generated on PowerPC systems. Do not use
20561 @option{-mstring} on little-endian PowerPC systems, since those
20562 instructions do not work when the processor is in little-endian mode.
20563 The exceptions are PPC740 and PPC750 which permit these instructions
20564 in little-endian mode.
20569 @opindex mno-update
20570 Generate code that uses (does not use) the load or store instructions
20571 that update the base register to the address of the calculated memory
20572 location. These instructions are generated by default. If you use
20573 @option{-mno-update}, there is a small window between the time that the
20574 stack pointer is updated and the address of the previous frame is
20575 stored, which means code that walks the stack frame across interrupts or
20576 signals may get corrupted data.
20578 @item -mavoid-indexed-addresses
20579 @itemx -mno-avoid-indexed-addresses
20580 @opindex mavoid-indexed-addresses
20581 @opindex mno-avoid-indexed-addresses
20582 Generate code that tries to avoid (not avoid) the use of indexed load
20583 or store instructions. These instructions can incur a performance
20584 penalty on Power6 processors in certain situations, such as when
20585 stepping through large arrays that cross a 16M boundary. This option
20586 is enabled by default when targeting Power6 and disabled otherwise.
20589 @itemx -mno-fused-madd
20590 @opindex mfused-madd
20591 @opindex mno-fused-madd
20592 Generate code that uses (does not use) the floating-point multiply and
20593 accumulate instructions. These instructions are generated by default
20594 if hardware floating point is used. The machine-dependent
20595 @option{-mfused-madd} option is now mapped to the machine-independent
20596 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
20597 mapped to @option{-ffp-contract=off}.
20603 Generate code that uses (does not use) the half-word multiply and
20604 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 processors.
20605 These instructions are generated by default when targeting those
20612 Generate code that uses (does not use) the string-search @samp{dlmzb}
20613 instruction on the IBM 405, 440, 464 and 476 processors. This instruction is
20614 generated by default when targeting those processors.
20616 @item -mno-bit-align
20618 @opindex mno-bit-align
20619 @opindex mbit-align
20620 On System V.4 and embedded PowerPC systems do not (do) force structures
20621 and unions that contain bit-fields to be aligned to the base type of the
20624 For example, by default a structure containing nothing but 8
20625 @code{unsigned} bit-fields of length 1 is aligned to a 4-byte
20626 boundary and has a size of 4 bytes. By using @option{-mno-bit-align},
20627 the structure is aligned to a 1-byte boundary and is 1 byte in
20630 @item -mno-strict-align
20631 @itemx -mstrict-align
20632 @opindex mno-strict-align
20633 @opindex mstrict-align
20634 On System V.4 and embedded PowerPC systems do not (do) assume that
20635 unaligned memory references are handled by the system.
20637 @item -mrelocatable
20638 @itemx -mno-relocatable
20639 @opindex mrelocatable
20640 @opindex mno-relocatable
20641 Generate code that allows (does not allow) a static executable to be
20642 relocated to a different address at run time. A simple embedded
20643 PowerPC system loader should relocate the entire contents of
20644 @code{.got2} and 4-byte locations listed in the @code{.fixup} section,
20645 a table of 32-bit addresses generated by this option. For this to
20646 work, all objects linked together must be compiled with
20647 @option{-mrelocatable} or @option{-mrelocatable-lib}.
20648 @option{-mrelocatable} code aligns the stack to an 8-byte boundary.
20650 @item -mrelocatable-lib
20651 @itemx -mno-relocatable-lib
20652 @opindex mrelocatable-lib
20653 @opindex mno-relocatable-lib
20654 Like @option{-mrelocatable}, @option{-mrelocatable-lib} generates a
20655 @code{.fixup} section to allow static executables to be relocated at
20656 run time, but @option{-mrelocatable-lib} does not use the smaller stack
20657 alignment of @option{-mrelocatable}. Objects compiled with
20658 @option{-mrelocatable-lib} may be linked with objects compiled with
20659 any combination of the @option{-mrelocatable} options.
20665 On System V.4 and embedded PowerPC systems do not (do) assume that
20666 register 2 contains a pointer to a global area pointing to the addresses
20667 used in the program.
20670 @itemx -mlittle-endian
20672 @opindex mlittle-endian
20673 On System V.4 and embedded PowerPC systems compile code for the
20674 processor in little-endian mode. The @option{-mlittle-endian} option is
20675 the same as @option{-mlittle}.
20678 @itemx -mbig-endian
20680 @opindex mbig-endian
20681 On System V.4 and embedded PowerPC systems compile code for the
20682 processor in big-endian mode. The @option{-mbig-endian} option is
20683 the same as @option{-mbig}.
20685 @item -mdynamic-no-pic
20686 @opindex mdynamic-no-pic
20687 On Darwin and Mac OS X systems, compile code so that it is not
20688 relocatable, but that its external references are relocatable. The
20689 resulting code is suitable for applications, but not shared
20692 @item -msingle-pic-base
20693 @opindex msingle-pic-base
20694 Treat the register used for PIC addressing as read-only, rather than
20695 loading it in the prologue for each function. The runtime system is
20696 responsible for initializing this register with an appropriate value
20697 before execution begins.
20699 @item -mprioritize-restricted-insns=@var{priority}
20700 @opindex mprioritize-restricted-insns
20701 This option controls the priority that is assigned to
20702 dispatch-slot restricted instructions during the second scheduling
20703 pass. The argument @var{priority} takes the value @samp{0}, @samp{1},
20704 or @samp{2} to assign no, highest, or second-highest (respectively)
20705 priority to dispatch-slot restricted
20708 @item -msched-costly-dep=@var{dependence_type}
20709 @opindex msched-costly-dep
20710 This option controls which dependences are considered costly
20711 by the target during instruction scheduling. The argument
20712 @var{dependence_type} takes one of the following values:
20716 No dependence is costly.
20719 All dependences are costly.
20721 @item @samp{true_store_to_load}
20722 A true dependence from store to load is costly.
20724 @item @samp{store_to_load}
20725 Any dependence from store to load is costly.
20728 Any dependence for which the latency is greater than or equal to
20729 @var{number} is costly.
20732 @item -minsert-sched-nops=@var{scheme}
20733 @opindex minsert-sched-nops
20734 This option controls which NOP insertion scheme is used during
20735 the second scheduling pass. The argument @var{scheme} takes one of the
20743 Pad with NOPs any dispatch group that has vacant issue slots,
20744 according to the scheduler's grouping.
20746 @item @samp{regroup_exact}
20747 Insert NOPs to force costly dependent insns into
20748 separate groups. Insert exactly as many NOPs as needed to force an insn
20749 to a new group, according to the estimated processor grouping.
20752 Insert NOPs to force costly dependent insns into
20753 separate groups. Insert @var{number} NOPs to force an insn to a new group.
20757 @opindex mcall-sysv
20758 On System V.4 and embedded PowerPC systems compile code using calling
20759 conventions that adhere to the March 1995 draft of the System V
20760 Application Binary Interface, PowerPC processor supplement. This is the
20761 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
20763 @item -mcall-sysv-eabi
20765 @opindex mcall-sysv-eabi
20766 @opindex mcall-eabi
20767 Specify both @option{-mcall-sysv} and @option{-meabi} options.
20769 @item -mcall-sysv-noeabi
20770 @opindex mcall-sysv-noeabi
20771 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
20773 @item -mcall-aixdesc
20775 On System V.4 and embedded PowerPC systems compile code for the AIX
20779 @opindex mcall-linux
20780 On System V.4 and embedded PowerPC systems compile code for the
20781 Linux-based GNU system.
20783 @item -mcall-freebsd
20784 @opindex mcall-freebsd
20785 On System V.4 and embedded PowerPC systems compile code for the
20786 FreeBSD operating system.
20788 @item -mcall-netbsd
20789 @opindex mcall-netbsd
20790 On System V.4 and embedded PowerPC systems compile code for the
20791 NetBSD operating system.
20793 @item -mcall-openbsd
20794 @opindex mcall-netbsd
20795 On System V.4 and embedded PowerPC systems compile code for the
20796 OpenBSD operating system.
20798 @item -maix-struct-return
20799 @opindex maix-struct-return
20800 Return all structures in memory (as specified by the AIX ABI)@.
20802 @item -msvr4-struct-return
20803 @opindex msvr4-struct-return
20804 Return structures smaller than 8 bytes in registers (as specified by the
20807 @item -mabi=@var{abi-type}
20809 Extend the current ABI with a particular extension, or remove such extension.
20810 Valid values are @samp{altivec}, @samp{no-altivec}, @samp{spe},
20811 @samp{no-spe}, @samp{ibmlongdouble}, @samp{ieeelongdouble},
20812 @samp{elfv1}, @samp{elfv2}@.
20816 Extend the current ABI with SPE ABI extensions. This does not change
20817 the default ABI, instead it adds the SPE ABI extensions to the current
20821 @opindex mabi=no-spe
20822 Disable Book-E SPE ABI extensions for the current ABI@.
20824 @item -mabi=ibmlongdouble
20825 @opindex mabi=ibmlongdouble
20826 Change the current ABI to use IBM extended-precision long double.
20827 This is a PowerPC 32-bit SYSV ABI option.
20829 @item -mabi=ieeelongdouble
20830 @opindex mabi=ieeelongdouble
20831 Change the current ABI to use IEEE extended-precision long double.
20832 This is a PowerPC 32-bit Linux ABI option.
20835 @opindex mabi=elfv1
20836 Change the current ABI to use the ELFv1 ABI.
20837 This is the default ABI for big-endian PowerPC 64-bit Linux.
20838 Overriding the default ABI requires special system support and is
20839 likely to fail in spectacular ways.
20842 @opindex mabi=elfv2
20843 Change the current ABI to use the ELFv2 ABI.
20844 This is the default ABI for little-endian PowerPC 64-bit Linux.
20845 Overriding the default ABI requires special system support and is
20846 likely to fail in spectacular ways.
20849 @itemx -mno-prototype
20850 @opindex mprototype
20851 @opindex mno-prototype
20852 On System V.4 and embedded PowerPC systems assume that all calls to
20853 variable argument functions are properly prototyped. Otherwise, the
20854 compiler must insert an instruction before every non-prototyped call to
20855 set or clear bit 6 of the condition code register (@code{CR}) to
20856 indicate whether floating-point values are passed in the floating-point
20857 registers in case the function takes variable arguments. With
20858 @option{-mprototype}, only calls to prototyped variable argument functions
20859 set or clear the bit.
20863 On embedded PowerPC systems, assume that the startup module is called
20864 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
20865 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}
20870 On embedded PowerPC systems, assume that the startup module is called
20871 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
20876 On embedded PowerPC systems, assume that the startup module is called
20877 @file{crt0.o} and the standard C libraries are @file{libads.a} and
20880 @item -myellowknife
20881 @opindex myellowknife
20882 On embedded PowerPC systems, assume that the startup module is called
20883 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
20888 On System V.4 and embedded PowerPC systems, specify that you are
20889 compiling for a VxWorks system.
20893 On embedded PowerPC systems, set the @code{PPC_EMB} bit in the ELF flags
20894 header to indicate that @samp{eabi} extended relocations are used.
20900 On System V.4 and embedded PowerPC systems do (do not) adhere to the
20901 Embedded Applications Binary Interface (EABI), which is a set of
20902 modifications to the System V.4 specifications. Selecting @option{-meabi}
20903 means that the stack is aligned to an 8-byte boundary, a function
20904 @code{__eabi} is called from @code{main} to set up the EABI
20905 environment, and the @option{-msdata} option can use both @code{r2} and
20906 @code{r13} to point to two separate small data areas. Selecting
20907 @option{-mno-eabi} means that the stack is aligned to a 16-byte boundary,
20908 no EABI initialization function is called from @code{main}, and the
20909 @option{-msdata} option only uses @code{r13} to point to a single
20910 small data area. The @option{-meabi} option is on by default if you
20911 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
20914 @opindex msdata=eabi
20915 On System V.4 and embedded PowerPC systems, put small initialized
20916 @code{const} global and static data in the @code{.sdata2} section, which
20917 is pointed to by register @code{r2}. Put small initialized
20918 non-@code{const} global and static data in the @code{.sdata} section,
20919 which is pointed to by register @code{r13}. Put small uninitialized
20920 global and static data in the @code{.sbss} section, which is adjacent to
20921 the @code{.sdata} section. The @option{-msdata=eabi} option is
20922 incompatible with the @option{-mrelocatable} option. The
20923 @option{-msdata=eabi} option also sets the @option{-memb} option.
20926 @opindex msdata=sysv
20927 On System V.4 and embedded PowerPC systems, put small global and static
20928 data in the @code{.sdata} section, which is pointed to by register
20929 @code{r13}. Put small uninitialized global and static data in the
20930 @code{.sbss} section, which is adjacent to the @code{.sdata} section.
20931 The @option{-msdata=sysv} option is incompatible with the
20932 @option{-mrelocatable} option.
20934 @item -msdata=default
20936 @opindex msdata=default
20938 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
20939 compile code the same as @option{-msdata=eabi}, otherwise compile code the
20940 same as @option{-msdata=sysv}.
20943 @opindex msdata=data
20944 On System V.4 and embedded PowerPC systems, put small global
20945 data in the @code{.sdata} section. Put small uninitialized global
20946 data in the @code{.sbss} section. Do not use register @code{r13}
20947 to address small data however. This is the default behavior unless
20948 other @option{-msdata} options are used.
20952 @opindex msdata=none
20954 On embedded PowerPC systems, put all initialized global and static data
20955 in the @code{.data} section, and all uninitialized data in the
20956 @code{.bss} section.
20958 @item -mblock-move-inline-limit=@var{num}
20959 @opindex mblock-move-inline-limit
20960 Inline all block moves (such as calls to @code{memcpy} or structure
20961 copies) less than or equal to @var{num} bytes. The minimum value for
20962 @var{num} is 32 bytes on 32-bit targets and 64 bytes on 64-bit
20963 targets. The default value is target-specific.
20967 @cindex smaller data references (PowerPC)
20968 @cindex .sdata/.sdata2 references (PowerPC)
20969 On embedded PowerPC systems, put global and static items less than or
20970 equal to @var{num} bytes into the small data or BSS sections instead of
20971 the normal data or BSS section. By default, @var{num} is 8. The
20972 @option{-G @var{num}} switch is also passed to the linker.
20973 All modules should be compiled with the same @option{-G @var{num}} value.
20976 @itemx -mno-regnames
20978 @opindex mno-regnames
20979 On System V.4 and embedded PowerPC systems do (do not) emit register
20980 names in the assembly language output using symbolic forms.
20983 @itemx -mno-longcall
20985 @opindex mno-longcall
20986 By default assume that all calls are far away so that a longer and more
20987 expensive calling sequence is required. This is required for calls
20988 farther than 32 megabytes (33,554,432 bytes) from the current location.
20989 A short call is generated if the compiler knows
20990 the call cannot be that far away. This setting can be overridden by
20991 the @code{shortcall} function attribute, or by @code{#pragma
20994 Some linkers are capable of detecting out-of-range calls and generating
20995 glue code on the fly. On these systems, long calls are unnecessary and
20996 generate slower code. As of this writing, the AIX linker can do this,
20997 as can the GNU linker for PowerPC/64. It is planned to add this feature
20998 to the GNU linker for 32-bit PowerPC systems as well.
21000 On Darwin/PPC systems, @code{#pragma longcall} generates @code{jbsr
21001 callee, L42}, plus a @dfn{branch island} (glue code). The two target
21002 addresses represent the callee and the branch island. The
21003 Darwin/PPC linker prefers the first address and generates a @code{bl
21004 callee} if the PPC @code{bl} instruction reaches the callee directly;
21005 otherwise, the linker generates @code{bl L42} to call the branch
21006 island. The branch island is appended to the body of the
21007 calling function; it computes the full 32-bit address of the callee
21010 On Mach-O (Darwin) systems, this option directs the compiler emit to
21011 the glue for every direct call, and the Darwin linker decides whether
21012 to use or discard it.
21014 In the future, GCC may ignore all longcall specifications
21015 when the linker is known to generate glue.
21017 @item -mtls-markers
21018 @itemx -mno-tls-markers
21019 @opindex mtls-markers
21020 @opindex mno-tls-markers
21021 Mark (do not mark) calls to @code{__tls_get_addr} with a relocation
21022 specifying the function argument. The relocation allows the linker to
21023 reliably associate function call with argument setup instructions for
21024 TLS optimization, which in turn allows GCC to better schedule the
21029 Adds support for multithreading with the @dfn{pthreads} library.
21030 This option sets flags for both the preprocessor and linker.
21035 This option enables use of the reciprocal estimate and
21036 reciprocal square root estimate instructions with additional
21037 Newton-Raphson steps to increase precision instead of doing a divide or
21038 square root and divide for floating-point arguments. You should use
21039 the @option{-ffast-math} option when using @option{-mrecip} (or at
21040 least @option{-funsafe-math-optimizations},
21041 @option{-ffinite-math-only}, @option{-freciprocal-math} and
21042 @option{-fno-trapping-math}). Note that while the throughput of the
21043 sequence is generally higher than the throughput of the non-reciprocal
21044 instruction, the precision of the sequence can be decreased by up to 2
21045 ulp (i.e.@: the inverse of 1.0 equals 0.99999994) for reciprocal square
21048 @item -mrecip=@var{opt}
21049 @opindex mrecip=opt
21050 This option controls which reciprocal estimate instructions
21051 may be used. @var{opt} is a comma-separated list of options, which may
21052 be preceded by a @code{!} to invert the option:
21057 Enable all estimate instructions.
21060 Enable the default instructions, equivalent to @option{-mrecip}.
21063 Disable all estimate instructions, equivalent to @option{-mno-recip}.
21066 Enable the reciprocal approximation instructions for both
21067 single and double precision.
21070 Enable the single-precision reciprocal approximation instructions.
21073 Enable the double-precision reciprocal approximation instructions.
21076 Enable the reciprocal square root approximation instructions for both
21077 single and double precision.
21080 Enable the single-precision reciprocal square root approximation instructions.
21083 Enable the double-precision reciprocal square root approximation instructions.
21087 So, for example, @option{-mrecip=all,!rsqrtd} enables
21088 all of the reciprocal estimate instructions, except for the
21089 @code{FRSQRTE}, @code{XSRSQRTEDP}, and @code{XVRSQRTEDP} instructions
21090 which handle the double-precision reciprocal square root calculations.
21092 @item -mrecip-precision
21093 @itemx -mno-recip-precision
21094 @opindex mrecip-precision
21095 Assume (do not assume) that the reciprocal estimate instructions
21096 provide higher-precision estimates than is mandated by the PowerPC
21097 ABI. Selecting @option{-mcpu=power6}, @option{-mcpu=power7} or
21098 @option{-mcpu=power8} automatically selects @option{-mrecip-precision}.
21099 The double-precision square root estimate instructions are not generated by
21100 default on low-precision machines, since they do not provide an
21101 estimate that converges after three steps.
21103 @item -mveclibabi=@var{type}
21104 @opindex mveclibabi
21105 Specifies the ABI type to use for vectorizing intrinsics using an
21106 external library. The only type supported at present is @samp{mass},
21107 which specifies to use IBM's Mathematical Acceleration Subsystem
21108 (MASS) libraries for vectorizing intrinsics using external libraries.
21109 GCC currently emits calls to @code{acosd2}, @code{acosf4},
21110 @code{acoshd2}, @code{acoshf4}, @code{asind2}, @code{asinf4},
21111 @code{asinhd2}, @code{asinhf4}, @code{atan2d2}, @code{atan2f4},
21112 @code{atand2}, @code{atanf4}, @code{atanhd2}, @code{atanhf4},
21113 @code{cbrtd2}, @code{cbrtf4}, @code{cosd2}, @code{cosf4},
21114 @code{coshd2}, @code{coshf4}, @code{erfcd2}, @code{erfcf4},
21115 @code{erfd2}, @code{erff4}, @code{exp2d2}, @code{exp2f4},
21116 @code{expd2}, @code{expf4}, @code{expm1d2}, @code{expm1f4},
21117 @code{hypotd2}, @code{hypotf4}, @code{lgammad2}, @code{lgammaf4},
21118 @code{log10d2}, @code{log10f4}, @code{log1pd2}, @code{log1pf4},
21119 @code{log2d2}, @code{log2f4}, @code{logd2}, @code{logf4},
21120 @code{powd2}, @code{powf4}, @code{sind2}, @code{sinf4}, @code{sinhd2},
21121 @code{sinhf4}, @code{sqrtd2}, @code{sqrtf4}, @code{tand2},
21122 @code{tanf4}, @code{tanhd2}, and @code{tanhf4} when generating code
21123 for power7. Both @option{-ftree-vectorize} and
21124 @option{-funsafe-math-optimizations} must also be enabled. The MASS
21125 libraries must be specified at link time.
21130 Generate (do not generate) the @code{friz} instruction when the
21131 @option{-funsafe-math-optimizations} option is used to optimize
21132 rounding of floating-point values to 64-bit integer and back to floating
21133 point. The @code{friz} instruction does not return the same value if
21134 the floating-point number is too large to fit in an integer.
21136 @item -mpointers-to-nested-functions
21137 @itemx -mno-pointers-to-nested-functions
21138 @opindex mpointers-to-nested-functions
21139 Generate (do not generate) code to load up the static chain register
21140 (@code{r11}) when calling through a pointer on AIX and 64-bit Linux
21141 systems where a function pointer points to a 3-word descriptor giving
21142 the function address, TOC value to be loaded in register @code{r2}, and
21143 static chain value to be loaded in register @code{r11}. The
21144 @option{-mpointers-to-nested-functions} is on by default. You cannot
21145 call through pointers to nested functions or pointers
21146 to functions compiled in other languages that use the static chain if
21147 you use @option{-mno-pointers-to-nested-functions}.
21149 @item -msave-toc-indirect
21150 @itemx -mno-save-toc-indirect
21151 @opindex msave-toc-indirect
21152 Generate (do not generate) code to save the TOC value in the reserved
21153 stack location in the function prologue if the function calls through
21154 a pointer on AIX and 64-bit Linux systems. If the TOC value is not
21155 saved in the prologue, it is saved just before the call through the
21156 pointer. The @option{-mno-save-toc-indirect} option is the default.
21158 @item -mcompat-align-parm
21159 @itemx -mno-compat-align-parm
21160 @opindex mcompat-align-parm
21161 Generate (do not generate) code to pass structure parameters with a
21162 maximum alignment of 64 bits, for compatibility with older versions
21165 Older versions of GCC (prior to 4.9.0) incorrectly did not align a
21166 structure parameter on a 128-bit boundary when that structure contained
21167 a member requiring 128-bit alignment. This is corrected in more
21168 recent versions of GCC. This option may be used to generate code
21169 that is compatible with functions compiled with older versions of
21172 The @option{-mno-compat-align-parm} option is the default.
21176 @subsection RX Options
21179 These command-line options are defined for RX targets:
21182 @item -m64bit-doubles
21183 @itemx -m32bit-doubles
21184 @opindex m64bit-doubles
21185 @opindex m32bit-doubles
21186 Make the @code{double} data type be 64 bits (@option{-m64bit-doubles})
21187 or 32 bits (@option{-m32bit-doubles}) in size. The default is
21188 @option{-m32bit-doubles}. @emph{Note} RX floating-point hardware only
21189 works on 32-bit values, which is why the default is
21190 @option{-m32bit-doubles}.
21196 Enables (@option{-fpu}) or disables (@option{-nofpu}) the use of RX
21197 floating-point hardware. The default is enabled for the RX600
21198 series and disabled for the RX200 series.
21200 Floating-point instructions are only generated for 32-bit floating-point
21201 values, however, so the FPU hardware is not used for doubles if the
21202 @option{-m64bit-doubles} option is used.
21204 @emph{Note} If the @option{-fpu} option is enabled then
21205 @option{-funsafe-math-optimizations} is also enabled automatically.
21206 This is because the RX FPU instructions are themselves unsafe.
21208 @item -mcpu=@var{name}
21210 Selects the type of RX CPU to be targeted. Currently three types are
21211 supported, the generic @samp{RX600} and @samp{RX200} series hardware and
21212 the specific @samp{RX610} CPU. The default is @samp{RX600}.
21214 The only difference between @samp{RX600} and @samp{RX610} is that the
21215 @samp{RX610} does not support the @code{MVTIPL} instruction.
21217 The @samp{RX200} series does not have a hardware floating-point unit
21218 and so @option{-nofpu} is enabled by default when this type is
21221 @item -mbig-endian-data
21222 @itemx -mlittle-endian-data
21223 @opindex mbig-endian-data
21224 @opindex mlittle-endian-data
21225 Store data (but not code) in the big-endian format. The default is
21226 @option{-mlittle-endian-data}, i.e.@: to store data in the little-endian
21229 @item -msmall-data-limit=@var{N}
21230 @opindex msmall-data-limit
21231 Specifies the maximum size in bytes of global and static variables
21232 which can be placed into the small data area. Using the small data
21233 area can lead to smaller and faster code, but the size of area is
21234 limited and it is up to the programmer to ensure that the area does
21235 not overflow. Also when the small data area is used one of the RX's
21236 registers (usually @code{r13}) is reserved for use pointing to this
21237 area, so it is no longer available for use by the compiler. This
21238 could result in slower and/or larger code if variables are pushed onto
21239 the stack instead of being held in this register.
21241 Note, common variables (variables that have not been initialized) and
21242 constants are not placed into the small data area as they are assigned
21243 to other sections in the output executable.
21245 The default value is zero, which disables this feature. Note, this
21246 feature is not enabled by default with higher optimization levels
21247 (@option{-O2} etc) because of the potentially detrimental effects of
21248 reserving a register. It is up to the programmer to experiment and
21249 discover whether this feature is of benefit to their program. See the
21250 description of the @option{-mpid} option for a description of how the
21251 actual register to hold the small data area pointer is chosen.
21257 Use the simulator runtime. The default is to use the libgloss
21258 board-specific runtime.
21260 @item -mas100-syntax
21261 @itemx -mno-as100-syntax
21262 @opindex mas100-syntax
21263 @opindex mno-as100-syntax
21264 When generating assembler output use a syntax that is compatible with
21265 Renesas's AS100 assembler. This syntax can also be handled by the GAS
21266 assembler, but it has some restrictions so it is not generated by default.
21268 @item -mmax-constant-size=@var{N}
21269 @opindex mmax-constant-size
21270 Specifies the maximum size, in bytes, of a constant that can be used as
21271 an operand in a RX instruction. Although the RX instruction set does
21272 allow constants of up to 4 bytes in length to be used in instructions,
21273 a longer value equates to a longer instruction. Thus in some
21274 circumstances it can be beneficial to restrict the size of constants
21275 that are used in instructions. Constants that are too big are instead
21276 placed into a constant pool and referenced via register indirection.
21278 The value @var{N} can be between 0 and 4. A value of 0 (the default)
21279 or 4 means that constants of any size are allowed.
21283 Enable linker relaxation. Linker relaxation is a process whereby the
21284 linker attempts to reduce the size of a program by finding shorter
21285 versions of various instructions. Disabled by default.
21287 @item -mint-register=@var{N}
21288 @opindex mint-register
21289 Specify the number of registers to reserve for fast interrupt handler
21290 functions. The value @var{N} can be between 0 and 4. A value of 1
21291 means that register @code{r13} is reserved for the exclusive use
21292 of fast interrupt handlers. A value of 2 reserves @code{r13} and
21293 @code{r12}. A value of 3 reserves @code{r13}, @code{r12} and
21294 @code{r11}, and a value of 4 reserves @code{r13} through @code{r10}.
21295 A value of 0, the default, does not reserve any registers.
21297 @item -msave-acc-in-interrupts
21298 @opindex msave-acc-in-interrupts
21299 Specifies that interrupt handler functions should preserve the
21300 accumulator register. This is only necessary if normal code might use
21301 the accumulator register, for example because it performs 64-bit
21302 multiplications. The default is to ignore the accumulator as this
21303 makes the interrupt handlers faster.
21309 Enables the generation of position independent data. When enabled any
21310 access to constant data is done via an offset from a base address
21311 held in a register. This allows the location of constant data to be
21312 determined at run time without requiring the executable to be
21313 relocated, which is a benefit to embedded applications with tight
21314 memory constraints. Data that can be modified is not affected by this
21317 Note, using this feature reserves a register, usually @code{r13}, for
21318 the constant data base address. This can result in slower and/or
21319 larger code, especially in complicated functions.
21321 The actual register chosen to hold the constant data base address
21322 depends upon whether the @option{-msmall-data-limit} and/or the
21323 @option{-mint-register} command-line options are enabled. Starting
21324 with register @code{r13} and proceeding downwards, registers are
21325 allocated first to satisfy the requirements of @option{-mint-register},
21326 then @option{-mpid} and finally @option{-msmall-data-limit}. Thus it
21327 is possible for the small data area register to be @code{r8} if both
21328 @option{-mint-register=4} and @option{-mpid} are specified on the
21331 By default this feature is not enabled. The default can be restored
21332 via the @option{-mno-pid} command-line option.
21334 @item -mno-warn-multiple-fast-interrupts
21335 @itemx -mwarn-multiple-fast-interrupts
21336 @opindex mno-warn-multiple-fast-interrupts
21337 @opindex mwarn-multiple-fast-interrupts
21338 Prevents GCC from issuing a warning message if it finds more than one
21339 fast interrupt handler when it is compiling a file. The default is to
21340 issue a warning for each extra fast interrupt handler found, as the RX
21341 only supports one such interrupt.
21343 @item -mallow-string-insns
21344 @itemx -mno-allow-string-insns
21345 @opindex mallow-string-insns
21346 @opindex mno-allow-string-insns
21347 Enables or disables the use of the string manipulation instructions
21348 @code{SMOVF}, @code{SCMPU}, @code{SMOVB}, @code{SMOVU}, @code{SUNTIL}
21349 @code{SWHILE} and also the @code{RMPA} instruction. These
21350 instructions may prefetch data, which is not safe to do if accessing
21351 an I/O register. (See section 12.2.7 of the RX62N Group User's Manual
21352 for more information).
21354 The default is to allow these instructions, but it is not possible for
21355 GCC to reliably detect all circumstances where a string instruction
21356 might be used to access an I/O register, so their use cannot be
21357 disabled automatically. Instead it is reliant upon the programmer to
21358 use the @option{-mno-allow-string-insns} option if their program
21359 accesses I/O space.
21361 When the instructions are enabled GCC defines the C preprocessor
21362 symbol @code{__RX_ALLOW_STRING_INSNS__}, otherwise it defines the
21363 symbol @code{__RX_DISALLOW_STRING_INSNS__}.
21369 Use only (or not only) @code{JSR} instructions to access functions.
21370 This option can be used when code size exceeds the range of @code{BSR}
21371 instructions. Note that @option{-mno-jsr} does not mean to not use
21372 @code{JSR} but instead means that any type of branch may be used.
21375 @emph{Note:} The generic GCC command-line option @option{-ffixed-@var{reg}}
21376 has special significance to the RX port when used with the
21377 @code{interrupt} function attribute. This attribute indicates a
21378 function intended to process fast interrupts. GCC ensures
21379 that it only uses the registers @code{r10}, @code{r11}, @code{r12}
21380 and/or @code{r13} and only provided that the normal use of the
21381 corresponding registers have been restricted via the
21382 @option{-ffixed-@var{reg}} or @option{-mint-register} command-line
21385 @node S/390 and zSeries Options
21386 @subsection S/390 and zSeries Options
21387 @cindex S/390 and zSeries Options
21389 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
21393 @itemx -msoft-float
21394 @opindex mhard-float
21395 @opindex msoft-float
21396 Use (do not use) the hardware floating-point instructions and registers
21397 for floating-point operations. When @option{-msoft-float} is specified,
21398 functions in @file{libgcc.a} are used to perform floating-point
21399 operations. When @option{-mhard-float} is specified, the compiler
21400 generates IEEE floating-point instructions. This is the default.
21403 @itemx -mno-hard-dfp
21405 @opindex mno-hard-dfp
21406 Use (do not use) the hardware decimal-floating-point instructions for
21407 decimal-floating-point operations. When @option{-mno-hard-dfp} is
21408 specified, functions in @file{libgcc.a} are used to perform
21409 decimal-floating-point operations. When @option{-mhard-dfp} is
21410 specified, the compiler generates decimal-floating-point hardware
21411 instructions. This is the default for @option{-march=z9-ec} or higher.
21413 @item -mlong-double-64
21414 @itemx -mlong-double-128
21415 @opindex mlong-double-64
21416 @opindex mlong-double-128
21417 These switches control the size of @code{long double} type. A size
21418 of 64 bits makes the @code{long double} type equivalent to the @code{double}
21419 type. This is the default.
21422 @itemx -mno-backchain
21423 @opindex mbackchain
21424 @opindex mno-backchain
21425 Store (do not store) the address of the caller's frame as backchain pointer
21426 into the callee's stack frame.
21427 A backchain may be needed to allow debugging using tools that do not understand
21428 DWARF call frame information.
21429 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
21430 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
21431 the backchain is placed into the topmost word of the 96/160 byte register
21434 In general, code compiled with @option{-mbackchain} is call-compatible with
21435 code compiled with @option{-mmo-backchain}; however, use of the backchain
21436 for debugging purposes usually requires that the whole binary is built with
21437 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
21438 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21439 to build a linux kernel use @option{-msoft-float}.
21441 The default is to not maintain the backchain.
21443 @item -mpacked-stack
21444 @itemx -mno-packed-stack
21445 @opindex mpacked-stack
21446 @opindex mno-packed-stack
21447 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
21448 specified, the compiler uses the all fields of the 96/160 byte register save
21449 area only for their default purpose; unused fields still take up stack space.
21450 When @option{-mpacked-stack} is specified, register save slots are densely
21451 packed at the top of the register save area; unused space is reused for other
21452 purposes, allowing for more efficient use of the available stack space.
21453 However, when @option{-mbackchain} is also in effect, the topmost word of
21454 the save area is always used to store the backchain, and the return address
21455 register is always saved two words below the backchain.
21457 As long as the stack frame backchain is not used, code generated with
21458 @option{-mpacked-stack} is call-compatible with code generated with
21459 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
21460 S/390 or zSeries generated code that uses the stack frame backchain at run
21461 time, not just for debugging purposes. Such code is not call-compatible
21462 with code compiled with @option{-mpacked-stack}. Also, note that the
21463 combination of @option{-mbackchain},
21464 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
21465 to build a linux kernel use @option{-msoft-float}.
21467 The default is to not use the packed stack layout.
21470 @itemx -mno-small-exec
21471 @opindex msmall-exec
21472 @opindex mno-small-exec
21473 Generate (or do not generate) code using the @code{bras} instruction
21474 to do subroutine calls.
21475 This only works reliably if the total executable size does not
21476 exceed 64k. The default is to use the @code{basr} instruction instead,
21477 which does not have this limitation.
21483 When @option{-m31} is specified, generate code compliant to the
21484 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
21485 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
21486 particular to generate 64-bit instructions. For the @samp{s390}
21487 targets, the default is @option{-m31}, while the @samp{s390x}
21488 targets default to @option{-m64}.
21494 When @option{-mzarch} is specified, generate code using the
21495 instructions available on z/Architecture.
21496 When @option{-mesa} is specified, generate code using the
21497 instructions available on ESA/390. Note that @option{-mesa} is
21498 not possible with @option{-m64}.
21499 When generating code compliant to the GNU/Linux for S/390 ABI,
21500 the default is @option{-mesa}. When generating code compliant
21501 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
21507 The @option{-mhtm} option enables a set of builtins making use of
21508 instructions available with the transactional execution facility
21509 introduced with the IBM zEnterprise EC12 machine generation
21510 @ref{S/390 System z Built-in Functions}.
21511 @option{-mhtm} is enabled by default when using @option{-march=zEC12}.
21517 When @option{-mvx} is specified, generate code using the instructions
21518 available with the vector extension facility introduced with the IBM
21519 z13 machine generation.
21520 This option changes the ABI for some vector type values with regard to
21521 alignment and calling conventions. In case vector type values are
21522 being used in an ABI-relevant context a GAS @samp{.gnu_attribute}
21523 command will be added to mark the resulting binary with the ABI used.
21524 @option{-mvx} is enabled by default when using @option{-march=z13}.
21527 @itemx -mno-zvector
21529 @opindex mno-zvector
21530 The @option{-mzvector} option enables vector language extensions and
21531 builtins using instructions available with the vector extension
21532 facility introduced with the IBM z13 machine generation.
21533 This option adds support for @samp{vector} to be used as a keyword to
21534 define vector type variables and arguments. @samp{vector} is only
21535 available when GNU extensions are enabled. It will not be expanded
21536 when requesting strict standard compliance e.g. with @option{-std=c99}.
21537 In addition to the GCC low-level builtins @option{-mzvector} enables
21538 a set of builtins added for compatibility with AltiVec-style
21539 implementations like Power and Cell. In order to make use of these
21540 builtins the header file @file{vecintrin.h} needs to be included.
21541 @option{-mzvector} is disabled by default.
21547 Generate (or do not generate) code using the @code{mvcle} instruction
21548 to perform block moves. When @option{-mno-mvcle} is specified,
21549 use a @code{mvc} loop instead. This is the default unless optimizing for
21556 Print (or do not print) additional debug information when compiling.
21557 The default is to not print debug information.
21559 @item -march=@var{cpu-type}
21561 Generate code that runs on @var{cpu-type}, which is the name of a
21562 system representing a certain processor type. Possible values for
21563 @var{cpu-type} are @samp{z900}, @samp{z990}, @samp{z9-109},
21564 @samp{z9-ec}, @samp{z10}, @samp{z196}, @samp{zEC12}, and @samp{z13}.
21565 The default is @option{-march=z900}. @samp{g5} and @samp{g6} are
21566 deprecated and will be removed with future releases.
21568 @item -mtune=@var{cpu-type}
21570 Tune to @var{cpu-type} everything applicable about the generated code,
21571 except for the ABI and the set of available instructions.
21572 The list of @var{cpu-type} values is the same as for @option{-march}.
21573 The default is the value used for @option{-march}.
21576 @itemx -mno-tpf-trace
21577 @opindex mtpf-trace
21578 @opindex mno-tpf-trace
21579 Generate code that adds (does not add) in TPF OS specific branches to trace
21580 routines in the operating system. This option is off by default, even
21581 when compiling for the TPF OS@.
21584 @itemx -mno-fused-madd
21585 @opindex mfused-madd
21586 @opindex mno-fused-madd
21587 Generate code that uses (does not use) the floating-point multiply and
21588 accumulate instructions. These instructions are generated by default if
21589 hardware floating point is used.
21591 @item -mwarn-framesize=@var{framesize}
21592 @opindex mwarn-framesize
21593 Emit a warning if the current function exceeds the given frame size. Because
21594 this is a compile-time check it doesn't need to be a real problem when the program
21595 runs. It is intended to identify functions that most probably cause
21596 a stack overflow. It is useful to be used in an environment with limited stack
21597 size e.g.@: the linux kernel.
21599 @item -mwarn-dynamicstack
21600 @opindex mwarn-dynamicstack
21601 Emit a warning if the function calls @code{alloca} or uses dynamically-sized
21602 arrays. This is generally a bad idea with a limited stack size.
21604 @item -mstack-guard=@var{stack-guard}
21605 @itemx -mstack-size=@var{stack-size}
21606 @opindex mstack-guard
21607 @opindex mstack-size
21608 If these options are provided the S/390 back end emits additional instructions in
21609 the function prologue that trigger a trap if the stack size is @var{stack-guard}
21610 bytes above the @var{stack-size} (remember that the stack on S/390 grows downward).
21611 If the @var{stack-guard} option is omitted the smallest power of 2 larger than
21612 the frame size of the compiled function is chosen.
21613 These options are intended to be used to help debugging stack overflow problems.
21614 The additionally emitted code causes only little overhead and hence can also be
21615 used in production-like systems without greater performance degradation. The given
21616 values have to be exact powers of 2 and @var{stack-size} has to be greater than
21617 @var{stack-guard} without exceeding 64k.
21618 In order to be efficient the extra code makes the assumption that the stack starts
21619 at an address aligned to the value given by @var{stack-size}.
21620 The @var{stack-guard} option can only be used in conjunction with @var{stack-size}.
21622 @item -mhotpatch=@var{pre-halfwords},@var{post-halfwords}
21624 If the hotpatch option is enabled, a ``hot-patching'' function
21625 prologue is generated for all functions in the compilation unit.
21626 The funtion label is prepended with the given number of two-byte
21627 NOP instructions (@var{pre-halfwords}, maximum 1000000). After
21628 the label, 2 * @var{post-halfwords} bytes are appended, using the
21629 largest NOP like instructions the architecture allows (maximum
21632 If both arguments are zero, hotpatching is disabled.
21634 This option can be overridden for individual functions with the
21635 @code{hotpatch} attribute.
21638 @node Score Options
21639 @subsection Score Options
21640 @cindex Score Options
21642 These options are defined for Score implementations:
21647 Compile code for big-endian mode. This is the default.
21651 Compile code for little-endian mode.
21655 Disable generation of @code{bcnz} instructions.
21659 Enable generation of unaligned load and store instructions.
21663 Enable the use of multiply-accumulate instructions. Disabled by default.
21667 Specify the SCORE5 as the target architecture.
21671 Specify the SCORE5U of the target architecture.
21675 Specify the SCORE7 as the target architecture. This is the default.
21679 Specify the SCORE7D as the target architecture.
21683 @subsection SH Options
21685 These @samp{-m} options are defined for the SH implementations:
21690 Generate code for the SH1.
21694 Generate code for the SH2.
21697 Generate code for the SH2e.
21701 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a way
21702 that the floating-point unit is not used.
21704 @item -m2a-single-only
21705 @opindex m2a-single-only
21706 Generate code for the SH2a-FPU, in such a way that no double-precision
21707 floating-point operations are used.
21710 @opindex m2a-single
21711 Generate code for the SH2a-FPU assuming the floating-point unit is in
21712 single-precision mode by default.
21716 Generate code for the SH2a-FPU assuming the floating-point unit is in
21717 double-precision mode by default.
21721 Generate code for the SH3.
21725 Generate code for the SH3e.
21729 Generate code for the SH4 without a floating-point unit.
21731 @item -m4-single-only
21732 @opindex m4-single-only
21733 Generate code for the SH4 with a floating-point unit that only
21734 supports single-precision arithmetic.
21738 Generate code for the SH4 assuming the floating-point unit is in
21739 single-precision mode by default.
21743 Generate code for the SH4.
21747 Generate code for SH4-100.
21749 @item -m4-100-nofpu
21750 @opindex m4-100-nofpu
21751 Generate code for SH4-100 in such a way that the
21752 floating-point unit is not used.
21754 @item -m4-100-single
21755 @opindex m4-100-single
21756 Generate code for SH4-100 assuming the floating-point unit is in
21757 single-precision mode by default.
21759 @item -m4-100-single-only
21760 @opindex m4-100-single-only
21761 Generate code for SH4-100 in such a way that no double-precision
21762 floating-point operations are used.
21766 Generate code for SH4-200.
21768 @item -m4-200-nofpu
21769 @opindex m4-200-nofpu
21770 Generate code for SH4-200 without in such a way that the
21771 floating-point unit is not used.
21773 @item -m4-200-single
21774 @opindex m4-200-single
21775 Generate code for SH4-200 assuming the floating-point unit is in
21776 single-precision mode by default.
21778 @item -m4-200-single-only
21779 @opindex m4-200-single-only
21780 Generate code for SH4-200 in such a way that no double-precision
21781 floating-point operations are used.
21785 Generate code for SH4-300.
21787 @item -m4-300-nofpu
21788 @opindex m4-300-nofpu
21789 Generate code for SH4-300 without in such a way that the
21790 floating-point unit is not used.
21792 @item -m4-300-single
21793 @opindex m4-300-single
21794 Generate code for SH4-300 in such a way that no double-precision
21795 floating-point operations are used.
21797 @item -m4-300-single-only
21798 @opindex m4-300-single-only
21799 Generate code for SH4-300 in such a way that no double-precision
21800 floating-point operations are used.
21804 Generate code for SH4-340 (no MMU, no FPU).
21808 Generate code for SH4-500 (no FPU). Passes @option{-isa=sh4-nofpu} to the
21813 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
21814 floating-point unit is not used.
21816 @item -m4a-single-only
21817 @opindex m4a-single-only
21818 Generate code for the SH4a, in such a way that no double-precision
21819 floating-point operations are used.
21822 @opindex m4a-single
21823 Generate code for the SH4a assuming the floating-point unit is in
21824 single-precision mode by default.
21828 Generate code for the SH4a.
21832 Same as @option{-m4a-nofpu}, except that it implicitly passes
21833 @option{-dsp} to the assembler. GCC doesn't generate any DSP
21834 instructions at the moment.
21838 Compile code for the processor in big-endian mode.
21842 Compile code for the processor in little-endian mode.
21846 Align doubles at 64-bit boundaries. Note that this changes the calling
21847 conventions, and thus some functions from the standard C library do
21848 not work unless you recompile it first with @option{-mdalign}.
21852 Shorten some address references at link time, when possible; uses the
21853 linker option @option{-relax}.
21857 Use 32-bit offsets in @code{switch} tables. The default is to use
21862 Enable the use of bit manipulation instructions on SH2A.
21866 Enable the use of the instruction @code{fmovd}. Check @option{-mdalign} for
21867 alignment constraints.
21871 Comply with the calling conventions defined by Renesas.
21874 @opindex mno-renesas
21875 Comply with the calling conventions defined for GCC before the Renesas
21876 conventions were available. This option is the default for all
21877 targets of the SH toolchain.
21880 @opindex mnomacsave
21881 Mark the @code{MAC} register as call-clobbered, even if
21882 @option{-mrenesas} is given.
21888 Control the IEEE compliance of floating-point comparisons, which affects the
21889 handling of cases where the result of a comparison is unordered. By default
21890 @option{-mieee} is implicitly enabled. If @option{-ffinite-math-only} is
21891 enabled @option{-mno-ieee} is implicitly set, which results in faster
21892 floating-point greater-equal and less-equal comparisons. The implicit settings
21893 can be overridden by specifying either @option{-mieee} or @option{-mno-ieee}.
21895 @item -minline-ic_invalidate
21896 @opindex minline-ic_invalidate
21897 Inline code to invalidate instruction cache entries after setting up
21898 nested function trampolines.
21899 This option has no effect if @option{-musermode} is in effect and the selected
21900 code generation option (e.g. @option{-m4}) does not allow the use of the @code{icbi}
21902 If the selected code generation option does not allow the use of the @code{icbi}
21903 instruction, and @option{-musermode} is not in effect, the inlined code
21904 manipulates the instruction cache address array directly with an associative
21905 write. This not only requires privileged mode at run time, but it also
21906 fails if the cache line had been mapped via the TLB and has become unmapped.
21910 Dump instruction size and location in the assembly code.
21913 @opindex mpadstruct
21914 This option is deprecated. It pads structures to multiple of 4 bytes,
21915 which is incompatible with the SH ABI@.
21917 @item -matomic-model=@var{model}
21918 @opindex matomic-model=@var{model}
21919 Sets the model of atomic operations and additional parameters as a comma
21920 separated list. For details on the atomic built-in functions see
21921 @ref{__atomic Builtins}. The following models and parameters are supported:
21926 Disable compiler generated atomic sequences and emit library calls for atomic
21927 operations. This is the default if the target is not @code{sh*-*-linux*}.
21930 Generate GNU/Linux compatible gUSA software atomic sequences for the atomic
21931 built-in functions. The generated atomic sequences require additional support
21932 from the interrupt/exception handling code of the system and are only suitable
21933 for SH3* and SH4* single-core systems. This option is enabled by default when
21934 the target is @code{sh*-*-linux*} and SH3* or SH4*. When the target is SH4A,
21935 this option also partially utilizes the hardware atomic instructions
21936 @code{movli.l} and @code{movco.l} to create more efficient code, unless
21937 @samp{strict} is specified.
21940 Generate software atomic sequences that use a variable in the thread control
21941 block. This is a variation of the gUSA sequences which can also be used on
21942 SH1* and SH2* targets. The generated atomic sequences require additional
21943 support from the interrupt/exception handling code of the system and are only
21944 suitable for single-core systems. When using this model, the @samp{gbr-offset=}
21945 parameter has to be specified as well.
21948 Generate software atomic sequences that temporarily disable interrupts by
21949 setting @code{SR.IMASK = 1111}. This model works only when the program runs
21950 in privileged mode and is only suitable for single-core systems. Additional
21951 support from the interrupt/exception handling code of the system is not
21952 required. This model is enabled by default when the target is
21953 @code{sh*-*-linux*} and SH1* or SH2*.
21956 Generate hardware atomic sequences using the @code{movli.l} and @code{movco.l}
21957 instructions only. This is only available on SH4A and is suitable for
21958 multi-core systems. Since the hardware instructions support only 32 bit atomic
21959 variables access to 8 or 16 bit variables is emulated with 32 bit accesses.
21960 Code compiled with this option is also compatible with other software
21961 atomic model interrupt/exception handling systems if executed on an SH4A
21962 system. Additional support from the interrupt/exception handling code of the
21963 system is not required for this model.
21966 This parameter specifies the offset in bytes of the variable in the thread
21967 control block structure that should be used by the generated atomic sequences
21968 when the @samp{soft-tcb} model has been selected. For other models this
21969 parameter is ignored. The specified value must be an integer multiple of four
21970 and in the range 0-1020.
21973 This parameter prevents mixed usage of multiple atomic models, even if they
21974 are compatible, and makes the compiler generate atomic sequences of the
21975 specified model only.
21981 Generate the @code{tas.b} opcode for @code{__atomic_test_and_set}.
21982 Notice that depending on the particular hardware and software configuration
21983 this can degrade overall performance due to the operand cache line flushes
21984 that are implied by the @code{tas.b} instruction. On multi-core SH4A
21985 processors the @code{tas.b} instruction must be used with caution since it
21986 can result in data corruption for certain cache configurations.
21989 @opindex mprefergot
21990 When generating position-independent code, emit function calls using
21991 the Global Offset Table instead of the Procedure Linkage Table.
21994 @itemx -mno-usermode
21996 @opindex mno-usermode
21997 Don't allow (allow) the compiler generating privileged mode code. Specifying
21998 @option{-musermode} also implies @option{-mno-inline-ic_invalidate} if the
21999 inlined code would not work in user mode. @option{-musermode} is the default
22000 when the target is @code{sh*-*-linux*}. If the target is SH1* or SH2*
22001 @option{-musermode} has no effect, since there is no user mode.
22003 @item -multcost=@var{number}
22004 @opindex multcost=@var{number}
22005 Set the cost to assume for a multiply insn.
22007 @item -mdiv=@var{strategy}
22008 @opindex mdiv=@var{strategy}
22009 Set the division strategy to be used for integer division operations.
22010 @var{strategy} can be one of:
22015 Calls a library function that uses the single-step division instruction
22016 @code{div1} to perform the operation. Division by zero calculates an
22017 unspecified result and does not trap. This is the default except for SH4,
22018 SH2A and SHcompact.
22021 Calls a library function that performs the operation in double precision
22022 floating point. Division by zero causes a floating-point exception. This is
22023 the default for SHcompact with FPU. Specifying this for targets that do not
22024 have a double precision FPU defaults to @code{call-div1}.
22027 Calls a library function that uses a lookup table for small divisors and
22028 the @code{div1} instruction with case distinction for larger divisors. Division
22029 by zero calculates an unspecified result and does not trap. This is the default
22030 for SH4. Specifying this for targets that do not have dynamic shift
22031 instructions defaults to @code{call-div1}.
22035 When a division strategy has not been specified the default strategy is
22036 selected based on the current target. For SH2A the default strategy is to
22037 use the @code{divs} and @code{divu} instructions instead of library function
22040 @item -maccumulate-outgoing-args
22041 @opindex maccumulate-outgoing-args
22042 Reserve space once for outgoing arguments in the function prologue rather
22043 than around each call. Generally beneficial for performance and size. Also
22044 needed for unwinding to avoid changing the stack frame around conditional code.
22046 @item -mdivsi3_libfunc=@var{name}
22047 @opindex mdivsi3_libfunc=@var{name}
22048 Set the name of the library function used for 32-bit signed division to
22050 This only affects the name used in the @samp{call} division strategies, and
22051 the compiler still expects the same sets of input/output/clobbered registers as
22052 if this option were not present.
22054 @item -mfixed-range=@var{register-range}
22055 @opindex mfixed-range
22056 Generate code treating the given register range as fixed registers.
22057 A fixed register is one that the register allocator can not use. This is
22058 useful when compiling kernel code. A register range is specified as
22059 two registers separated by a dash. Multiple register ranges can be
22060 specified separated by a comma.
22062 @item -mbranch-cost=@var{num}
22063 @opindex mbranch-cost=@var{num}
22064 Assume @var{num} to be the cost for a branch instruction. Higher numbers
22065 make the compiler try to generate more branch-free code if possible.
22066 If not specified the value is selected depending on the processor type that
22067 is being compiled for.
22070 @itemx -mno-zdcbranch
22071 @opindex mzdcbranch
22072 @opindex mno-zdcbranch
22073 Assume (do not assume) that zero displacement conditional branch instructions
22074 @code{bt} and @code{bf} are fast. If @option{-mzdcbranch} is specified, the
22075 compiler prefers zero displacement branch code sequences. This is
22076 enabled by default when generating code for SH4 and SH4A. It can be explicitly
22077 disabled by specifying @option{-mno-zdcbranch}.
22079 @item -mcbranch-force-delay-slot
22080 @opindex mcbranch-force-delay-slot
22081 Force the usage of delay slots for conditional branches, which stuffs the delay
22082 slot with a @code{nop} if a suitable instruction can't be found. By default
22083 this option is disabled. It can be enabled to work around hardware bugs as
22084 found in the original SH7055.
22087 @itemx -mno-fused-madd
22088 @opindex mfused-madd
22089 @opindex mno-fused-madd
22090 Generate code that uses (does not use) the floating-point multiply and
22091 accumulate instructions. These instructions are generated by default
22092 if hardware floating point is used. The machine-dependent
22093 @option{-mfused-madd} option is now mapped to the machine-independent
22094 @option{-ffp-contract=fast} option, and @option{-mno-fused-madd} is
22095 mapped to @option{-ffp-contract=off}.
22101 Allow or disallow the compiler to emit the @code{fsca} instruction for sine
22102 and cosine approximations. The option @option{-mfsca} must be used in
22103 combination with @option{-funsafe-math-optimizations}. It is enabled by default
22104 when generating code for SH4A. Using @option{-mno-fsca} disables sine and cosine
22105 approximations even if @option{-funsafe-math-optimizations} is in effect.
22111 Allow or disallow the compiler to emit the @code{fsrra} instruction for
22112 reciprocal square root approximations. The option @option{-mfsrra} must be used
22113 in combination with @option{-funsafe-math-optimizations} and
22114 @option{-ffinite-math-only}. It is enabled by default when generating code for
22115 SH4A. Using @option{-mno-fsrra} disables reciprocal square root approximations
22116 even if @option{-funsafe-math-optimizations} and @option{-ffinite-math-only} are
22119 @item -mpretend-cmove
22120 @opindex mpretend-cmove
22121 Prefer zero-displacement conditional branches for conditional move instruction
22122 patterns. This can result in faster code on the SH4 processor.
22126 Generate code using the FDPIC ABI.
22130 @node Solaris 2 Options
22131 @subsection Solaris 2 Options
22132 @cindex Solaris 2 options
22134 These @samp{-m} options are supported on Solaris 2:
22137 @item -mclear-hwcap
22138 @opindex mclear-hwcap
22139 @option{-mclear-hwcap} tells the compiler to remove the hardware
22140 capabilities generated by the Solaris assembler. This is only necessary
22141 when object files use ISA extensions not supported by the current
22142 machine, but check at runtime whether or not to use them.
22144 @item -mimpure-text
22145 @opindex mimpure-text
22146 @option{-mimpure-text}, used in addition to @option{-shared}, tells
22147 the compiler to not pass @option{-z text} to the linker when linking a
22148 shared object. Using this option, you can link position-dependent
22149 code into a shared object.
22151 @option{-mimpure-text} suppresses the ``relocations remain against
22152 allocatable but non-writable sections'' linker error message.
22153 However, the necessary relocations trigger copy-on-write, and the
22154 shared object is not actually shared across processes. Instead of
22155 using @option{-mimpure-text}, you should compile all source code with
22156 @option{-fpic} or @option{-fPIC}.
22160 These switches are supported in addition to the above on Solaris 2:
22165 Add support for multithreading using the POSIX threads library. This
22166 option sets flags for both the preprocessor and linker. This option does
22167 not affect the thread safety of object code produced by the compiler or
22168 that of libraries supplied with it.
22172 This is a synonym for @option{-pthreads}.
22175 @node SPARC Options
22176 @subsection SPARC Options
22177 @cindex SPARC options
22179 These @samp{-m} options are supported on the SPARC:
22182 @item -mno-app-regs
22184 @opindex mno-app-regs
22186 Specify @option{-mapp-regs} to generate output using the global registers
22187 2 through 4, which the SPARC SVR4 ABI reserves for applications. Like the
22188 global register 1, each global register 2 through 4 is then treated as an
22189 allocable register that is clobbered by function calls. This is the default.
22191 To be fully SVR4 ABI-compliant at the cost of some performance loss,
22192 specify @option{-mno-app-regs}. You should compile libraries and system
22193 software with this option.
22199 With @option{-mflat}, the compiler does not generate save/restore instructions
22200 and uses a ``flat'' or single register window model. This model is compatible
22201 with the regular register window model. The local registers and the input
22202 registers (0--5) are still treated as ``call-saved'' registers and are
22203 saved on the stack as needed.
22205 With @option{-mno-flat} (the default), the compiler generates save/restore
22206 instructions (except for leaf functions). This is the normal operating mode.
22209 @itemx -mhard-float
22211 @opindex mhard-float
22212 Generate output containing floating-point instructions. This is the
22216 @itemx -msoft-float
22218 @opindex msoft-float
22219 Generate output containing library calls for floating point.
22220 @strong{Warning:} the requisite libraries are not available for all SPARC
22221 targets. Normally the facilities of the machine's usual C compiler are
22222 used, but this cannot be done directly in cross-compilation. You must make
22223 your own arrangements to provide suitable library functions for
22224 cross-compilation. The embedded targets @samp{sparc-*-aout} and
22225 @samp{sparclite-*-*} do provide software floating-point support.
22227 @option{-msoft-float} changes the calling convention in the output file;
22228 therefore, it is only useful if you compile @emph{all} of a program with
22229 this option. In particular, you need to compile @file{libgcc.a}, the
22230 library that comes with GCC, with @option{-msoft-float} in order for
22233 @item -mhard-quad-float
22234 @opindex mhard-quad-float
22235 Generate output containing quad-word (long double) floating-point
22238 @item -msoft-quad-float
22239 @opindex msoft-quad-float
22240 Generate output containing library calls for quad-word (long double)
22241 floating-point instructions. The functions called are those specified
22242 in the SPARC ABI@. This is the default.
22244 As of this writing, there are no SPARC implementations that have hardware
22245 support for the quad-word floating-point instructions. They all invoke
22246 a trap handler for one of these instructions, and then the trap handler
22247 emulates the effect of the instruction. Because of the trap handler overhead,
22248 this is much slower than calling the ABI library routines. Thus the
22249 @option{-msoft-quad-float} option is the default.
22251 @item -mno-unaligned-doubles
22252 @itemx -munaligned-doubles
22253 @opindex mno-unaligned-doubles
22254 @opindex munaligned-doubles
22255 Assume that doubles have 8-byte alignment. This is the default.
22257 With @option{-munaligned-doubles}, GCC assumes that doubles have 8-byte
22258 alignment only if they are contained in another type, or if they have an
22259 absolute address. Otherwise, it assumes they have 4-byte alignment.
22260 Specifying this option avoids some rare compatibility problems with code
22261 generated by other compilers. It is not the default because it results
22262 in a performance loss, especially for floating-point code.
22265 @itemx -mno-user-mode
22266 @opindex muser-mode
22267 @opindex mno-user-mode
22268 Do not generate code that can only run in supervisor mode. This is relevant
22269 only for the @code{casa} instruction emitted for the LEON3 processor. This
22272 @item -mfaster-structs
22273 @itemx -mno-faster-structs
22274 @opindex mfaster-structs
22275 @opindex mno-faster-structs
22276 With @option{-mfaster-structs}, the compiler assumes that structures
22277 should have 8-byte alignment. This enables the use of pairs of
22278 @code{ldd} and @code{std} instructions for copies in structure
22279 assignment, in place of twice as many @code{ld} and @code{st} pairs.
22280 However, the use of this changed alignment directly violates the SPARC
22281 ABI@. Thus, it's intended only for use on targets where the developer
22282 acknowledges that their resulting code is not directly in line with
22283 the rules of the ABI@.
22285 @item -mstd-struct-return
22286 @itemx -mno-std-struct-return
22287 @opindex mstd-struct-return
22288 @opindex mno-std-struct-return
22289 With @option{-mstd-struct-return}, the compiler generates checking code
22290 in functions returning structures or unions to detect size mismatches
22291 between the two sides of function calls, as per the 32-bit ABI@.
22293 The default is @option{-mno-std-struct-return}. This option has no effect
22296 @item -mcpu=@var{cpu_type}
22298 Set the instruction set, register set, and instruction scheduling parameters
22299 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
22300 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{hypersparc},
22301 @samp{leon}, @samp{leon3}, @samp{leon3v7}, @samp{sparclite}, @samp{f930},
22302 @samp{f934}, @samp{sparclite86x}, @samp{sparclet}, @samp{tsc701}, @samp{v9},
22303 @samp{ultrasparc}, @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2},
22304 @samp{niagara3}, @samp{niagara4} and @samp{niagara7}.
22306 Native Solaris and GNU/Linux toolchains also support the value @samp{native},
22307 which selects the best architecture option for the host processor.
22308 @option{-mcpu=native} has no effect if GCC does not recognize
22311 Default instruction scheduling parameters are used for values that select
22312 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
22313 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
22315 Here is a list of each supported architecture and their supported
22323 supersparc, hypersparc, leon, leon3
22326 f930, f934, sparclite86x
22332 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, niagara4, niagara7
22335 By default (unless configured otherwise), GCC generates code for the V7
22336 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
22337 additionally optimizes it for the Cypress CY7C602 chip, as used in the
22338 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
22339 SPARCStation 1, 2, IPX etc.
22341 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
22342 architecture. The only difference from V7 code is that the compiler emits
22343 the integer multiply and integer divide instructions which exist in SPARC-V8
22344 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
22345 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
22348 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
22349 the SPARC architecture. This adds the integer multiply, integer divide step
22350 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
22351 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
22352 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
22353 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
22354 MB86934 chip, which is the more recent SPARClite with FPU@.
22356 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
22357 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
22358 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
22359 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
22360 optimizes it for the TEMIC SPARClet chip.
22362 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
22363 architecture. This adds 64-bit integer and floating-point move instructions,
22364 3 additional floating-point condition code registers and conditional move
22365 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
22366 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
22367 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
22368 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
22369 @option{-mcpu=niagara}, the compiler additionally optimizes it for
22370 Sun UltraSPARC T1 chips. With @option{-mcpu=niagara2}, the compiler
22371 additionally optimizes it for Sun UltraSPARC T2 chips. With
22372 @option{-mcpu=niagara3}, the compiler additionally optimizes it for Sun
22373 UltraSPARC T3 chips. With @option{-mcpu=niagara4}, the compiler
22374 additionally optimizes it for Sun UltraSPARC T4 chips. With
22375 @option{-mcpu=niagara7}, the compiler additionally optimizes it for
22376 Oracle SPARC M7 chips.
22378 @item -mtune=@var{cpu_type}
22380 Set the instruction scheduling parameters for machine type
22381 @var{cpu_type}, but do not set the instruction set or register set that the
22382 option @option{-mcpu=@var{cpu_type}} does.
22384 The same values for @option{-mcpu=@var{cpu_type}} can be used for
22385 @option{-mtune=@var{cpu_type}}, but the only useful values are those
22386 that select a particular CPU implementation. Those are
22387 @samp{cypress}, @samp{supersparc}, @samp{hypersparc}, @samp{leon},
22388 @samp{leon3}, @samp{leon3v7}, @samp{f930}, @samp{f934},
22389 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
22390 @samp{ultrasparc3}, @samp{niagara}, @samp{niagara2}, @samp{niagara3},
22391 @samp{niagara4} and @samp{niagara7}. With native Solaris and
22392 GNU/Linux toolchains, @samp{native} can also be used.
22397 @opindex mno-v8plus
22398 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
22399 difference from the V8 ABI is that the global and out registers are
22400 considered 64 bits wide. This is enabled by default on Solaris in 32-bit
22401 mode for all SPARC-V9 processors.
22407 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
22408 Visual Instruction Set extensions. The default is @option{-mno-vis}.
22414 With @option{-mvis2}, GCC generates code that takes advantage of
22415 version 2.0 of the UltraSPARC Visual Instruction Set extensions. The
22416 default is @option{-mvis2} when targeting a cpu that supports such
22417 instructions, such as UltraSPARC-III and later. Setting @option{-mvis2}
22418 also sets @option{-mvis}.
22424 With @option{-mvis3}, GCC generates code that takes advantage of
22425 version 3.0 of the UltraSPARC Visual Instruction Set extensions. The
22426 default is @option{-mvis3} when targeting a cpu that supports such
22427 instructions, such as niagara-3 and later. Setting @option{-mvis3}
22428 also sets @option{-mvis2} and @option{-mvis}.
22434 With @option{-mvis4}, GCC generates code that takes advantage of
22435 version 4.0 of the UltraSPARC Visual Instruction Set extensions. The
22436 default is @option{-mvis4} when targeting a cpu that supports such
22437 instructions, such as niagara-7 and later. Setting @option{-mvis4}
22438 also sets @option{-mvis3}, @option{-mvis2} and @option{-mvis}.
22443 @opindex mno-cbcond
22444 With @option{-mcbcond}, GCC generates code that takes advantage of
22445 compare-and-branch instructions, as defined in the Sparc Architecture 2011.
22446 The default is @option{-mcbcond} when targeting a cpu that supports such
22447 instructions, such as niagara-4 and later.
22453 With @option{-mpopc}, GCC generates code that takes advantage of the UltraSPARC
22454 population count instruction. The default is @option{-mpopc}
22455 when targeting a cpu that supports such instructions, such as Niagara-2 and
22462 With @option{-mfmaf}, GCC generates code that takes advantage of the UltraSPARC
22463 Fused Multiply-Add Floating-point extensions. The default is @option{-mfmaf}
22464 when targeting a cpu that supports such instructions, such as Niagara-3 and
22468 @opindex mfix-at697f
22469 Enable the documented workaround for the single erratum of the Atmel AT697F
22470 processor (which corresponds to erratum #13 of the AT697E processor).
22473 @opindex mfix-ut699
22474 Enable the documented workarounds for the floating-point errata and the data
22475 cache nullify errata of the UT699 processor.
22478 These @samp{-m} options are supported in addition to the above
22479 on SPARC-V9 processors in 64-bit environments:
22486 Generate code for a 32-bit or 64-bit environment.
22487 The 32-bit environment sets int, long and pointer to 32 bits.
22488 The 64-bit environment sets int to 32 bits and long and pointer
22491 @item -mcmodel=@var{which}
22493 Set the code model to one of
22497 The Medium/Low code model: 64-bit addresses, programs
22498 must be linked in the low 32 bits of memory. Programs can be statically
22499 or dynamically linked.
22502 The Medium/Middle code model: 64-bit addresses, programs
22503 must be linked in the low 44 bits of memory, the text and data segments must
22504 be less than 2GB in size and the data segment must be located within 2GB of
22508 The Medium/Anywhere code model: 64-bit addresses, programs
22509 may be linked anywhere in memory, the text and data segments must be less
22510 than 2GB in size and the data segment must be located within 2GB of the
22514 The Medium/Anywhere code model for embedded systems:
22515 64-bit addresses, the text and data segments must be less than 2GB in
22516 size, both starting anywhere in memory (determined at link time). The
22517 global register %g4 points to the base of the data segment. Programs
22518 are statically linked and PIC is not supported.
22521 @item -mmemory-model=@var{mem-model}
22522 @opindex mmemory-model
22523 Set the memory model in force on the processor to one of
22527 The default memory model for the processor and operating system.
22530 Relaxed Memory Order
22533 Partial Store Order
22539 Sequential Consistency
22542 These memory models are formally defined in Appendix D of the Sparc V9
22543 architecture manual, as set in the processor's @code{PSTATE.MM} field.
22546 @itemx -mno-stack-bias
22547 @opindex mstack-bias
22548 @opindex mno-stack-bias
22549 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
22550 frame pointer if present, are offset by @minus{}2047 which must be added back
22551 when making stack frame references. This is the default in 64-bit mode.
22552 Otherwise, assume no such offset is present.
22556 @subsection SPU Options
22557 @cindex SPU options
22559 These @samp{-m} options are supported on the SPU:
22563 @itemx -merror-reloc
22564 @opindex mwarn-reloc
22565 @opindex merror-reloc
22567 The loader for SPU does not handle dynamic relocations. By default, GCC
22568 gives an error when it generates code that requires a dynamic
22569 relocation. @option{-mno-error-reloc} disables the error,
22570 @option{-mwarn-reloc} generates a warning instead.
22573 @itemx -munsafe-dma
22575 @opindex munsafe-dma
22577 Instructions that initiate or test completion of DMA must not be
22578 reordered with respect to loads and stores of the memory that is being
22580 With @option{-munsafe-dma} you must use the @code{volatile} keyword to protect
22581 memory accesses, but that can lead to inefficient code in places where the
22582 memory is known to not change. Rather than mark the memory as volatile,
22583 you can use @option{-msafe-dma} to tell the compiler to treat
22584 the DMA instructions as potentially affecting all memory.
22586 @item -mbranch-hints
22587 @opindex mbranch-hints
22589 By default, GCC generates a branch hint instruction to avoid
22590 pipeline stalls for always-taken or probably-taken branches. A hint
22591 is not generated closer than 8 instructions away from its branch.
22592 There is little reason to disable them, except for debugging purposes,
22593 or to make an object a little bit smaller.
22597 @opindex msmall-mem
22598 @opindex mlarge-mem
22600 By default, GCC generates code assuming that addresses are never larger
22601 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
22602 a full 32-bit address.
22607 By default, GCC links against startup code that assumes the SPU-style
22608 main function interface (which has an unconventional parameter list).
22609 With @option{-mstdmain}, GCC links your program against startup
22610 code that assumes a C99-style interface to @code{main}, including a
22611 local copy of @code{argv} strings.
22613 @item -mfixed-range=@var{register-range}
22614 @opindex mfixed-range
22615 Generate code treating the given register range as fixed registers.
22616 A fixed register is one that the register allocator cannot use. This is
22617 useful when compiling kernel code. A register range is specified as
22618 two registers separated by a dash. Multiple register ranges can be
22619 specified separated by a comma.
22625 Compile code assuming that pointers to the PPU address space accessed
22626 via the @code{__ea} named address space qualifier are either 32 or 64
22627 bits wide. The default is 32 bits. As this is an ABI-changing option,
22628 all object code in an executable must be compiled with the same setting.
22630 @item -maddress-space-conversion
22631 @itemx -mno-address-space-conversion
22632 @opindex maddress-space-conversion
22633 @opindex mno-address-space-conversion
22634 Allow/disallow treating the @code{__ea} address space as superset
22635 of the generic address space. This enables explicit type casts
22636 between @code{__ea} and generic pointer as well as implicit
22637 conversions of generic pointers to @code{__ea} pointers. The
22638 default is to allow address space pointer conversions.
22640 @item -mcache-size=@var{cache-size}
22641 @opindex mcache-size
22642 This option controls the version of libgcc that the compiler links to an
22643 executable and selects a software-managed cache for accessing variables
22644 in the @code{__ea} address space with a particular cache size. Possible
22645 options for @var{cache-size} are @samp{8}, @samp{16}, @samp{32}, @samp{64}
22646 and @samp{128}. The default cache size is 64KB.
22648 @item -matomic-updates
22649 @itemx -mno-atomic-updates
22650 @opindex matomic-updates
22651 @opindex mno-atomic-updates
22652 This option controls the version of libgcc that the compiler links to an
22653 executable and selects whether atomic updates to the software-managed
22654 cache of PPU-side variables are used. If you use atomic updates, changes
22655 to a PPU variable from SPU code using the @code{__ea} named address space
22656 qualifier do not interfere with changes to other PPU variables residing
22657 in the same cache line from PPU code. If you do not use atomic updates,
22658 such interference may occur; however, writing back cache lines is
22659 more efficient. The default behavior is to use atomic updates.
22662 @itemx -mdual-nops=@var{n}
22663 @opindex mdual-nops
22664 By default, GCC inserts nops to increase dual issue when it expects
22665 it to increase performance. @var{n} can be a value from 0 to 10. A
22666 smaller @var{n} inserts fewer nops. 10 is the default, 0 is the
22667 same as @option{-mno-dual-nops}. Disabled with @option{-Os}.
22669 @item -mhint-max-nops=@var{n}
22670 @opindex mhint-max-nops
22671 Maximum number of nops to insert for a branch hint. A branch hint must
22672 be at least 8 instructions away from the branch it is affecting. GCC
22673 inserts up to @var{n} nops to enforce this, otherwise it does not
22674 generate the branch hint.
22676 @item -mhint-max-distance=@var{n}
22677 @opindex mhint-max-distance
22678 The encoding of the branch hint instruction limits the hint to be within
22679 256 instructions of the branch it is affecting. By default, GCC makes
22680 sure it is within 125.
22683 @opindex msafe-hints
22684 Work around a hardware bug that causes the SPU to stall indefinitely.
22685 By default, GCC inserts the @code{hbrp} instruction to make sure
22686 this stall won't happen.
22690 @node System V Options
22691 @subsection Options for System V
22693 These additional options are available on System V Release 4 for
22694 compatibility with other compilers on those systems:
22699 Create a shared object.
22700 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
22704 Identify the versions of each tool used by the compiler, in a
22705 @code{.ident} assembler directive in the output.
22709 Refrain from adding @code{.ident} directives to the output file (this is
22712 @item -YP,@var{dirs}
22714 Search the directories @var{dirs}, and no others, for libraries
22715 specified with @option{-l}.
22717 @item -Ym,@var{dir}
22719 Look in the directory @var{dir} to find the M4 preprocessor.
22720 The assembler uses this option.
22721 @c This is supposed to go with a -Yd for predefined M4 macro files, but
22722 @c the generic assembler that comes with Solaris takes just -Ym.
22725 @node TILE-Gx Options
22726 @subsection TILE-Gx Options
22727 @cindex TILE-Gx options
22729 These @samp{-m} options are supported on the TILE-Gx:
22732 @item -mcmodel=small
22733 @opindex mcmodel=small
22734 Generate code for the small model. The distance for direct calls is
22735 limited to 500M in either direction. PC-relative addresses are 32
22736 bits. Absolute addresses support the full address range.
22738 @item -mcmodel=large
22739 @opindex mcmodel=large
22740 Generate code for the large model. There is no limitation on call
22741 distance, pc-relative addresses, or absolute addresses.
22743 @item -mcpu=@var{name}
22745 Selects the type of CPU to be targeted. Currently the only supported
22746 type is @samp{tilegx}.
22752 Generate code for a 32-bit or 64-bit environment. The 32-bit
22753 environment sets int, long, and pointer to 32 bits. The 64-bit
22754 environment sets int to 32 bits and long and pointer to 64 bits.
22757 @itemx -mlittle-endian
22758 @opindex mbig-endian
22759 @opindex mlittle-endian
22760 Generate code in big/little endian mode, respectively.
22763 @node TILEPro Options
22764 @subsection TILEPro Options
22765 @cindex TILEPro options
22767 These @samp{-m} options are supported on the TILEPro:
22770 @item -mcpu=@var{name}
22772 Selects the type of CPU to be targeted. Currently the only supported
22773 type is @samp{tilepro}.
22777 Generate code for a 32-bit environment, which sets int, long, and
22778 pointer to 32 bits. This is the only supported behavior so the flag
22779 is essentially ignored.
22783 @subsection V850 Options
22784 @cindex V850 Options
22786 These @samp{-m} options are defined for V850 implementations:
22790 @itemx -mno-long-calls
22791 @opindex mlong-calls
22792 @opindex mno-long-calls
22793 Treat all calls as being far away (near). If calls are assumed to be
22794 far away, the compiler always loads the function's address into a
22795 register, and calls indirect through the pointer.
22801 Do not optimize (do optimize) basic blocks that use the same index
22802 pointer 4 or more times to copy pointer into the @code{ep} register, and
22803 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
22804 option is on by default if you optimize.
22806 @item -mno-prolog-function
22807 @itemx -mprolog-function
22808 @opindex mno-prolog-function
22809 @opindex mprolog-function
22810 Do not use (do use) external functions to save and restore registers
22811 at the prologue and epilogue of a function. The external functions
22812 are slower, but use less code space if more than one function saves
22813 the same number of registers. The @option{-mprolog-function} option
22814 is on by default if you optimize.
22818 Try to make the code as small as possible. At present, this just turns
22819 on the @option{-mep} and @option{-mprolog-function} options.
22821 @item -mtda=@var{n}
22823 Put static or global variables whose size is @var{n} bytes or less into
22824 the tiny data area that register @code{ep} points to. The tiny data
22825 area can hold up to 256 bytes in total (128 bytes for byte references).
22827 @item -msda=@var{n}
22829 Put static or global variables whose size is @var{n} bytes or less into
22830 the small data area that register @code{gp} points to. The small data
22831 area can hold up to 64 kilobytes.
22833 @item -mzda=@var{n}
22835 Put static or global variables whose size is @var{n} bytes or less into
22836 the first 32 kilobytes of memory.
22840 Specify that the target processor is the V850.
22844 Specify that the target processor is the V850E3V5. The preprocessor
22845 constant @code{__v850e3v5__} is defined if this option is used.
22849 Specify that the target processor is the V850E3V5. This is an alias for
22850 the @option{-mv850e3v5} option.
22854 Specify that the target processor is the V850E2V3. The preprocessor
22855 constant @code{__v850e2v3__} is defined if this option is used.
22859 Specify that the target processor is the V850E2. The preprocessor
22860 constant @code{__v850e2__} is defined if this option is used.
22864 Specify that the target processor is the V850E1. The preprocessor
22865 constants @code{__v850e1__} and @code{__v850e__} are defined if
22866 this option is used.
22870 Specify that the target processor is the V850ES. This is an alias for
22871 the @option{-mv850e1} option.
22875 Specify that the target processor is the V850E@. The preprocessor
22876 constant @code{__v850e__} is defined if this option is used.
22878 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
22879 nor @option{-mv850e2} nor @option{-mv850e2v3} nor @option{-mv850e3v5}
22880 are defined then a default target processor is chosen and the
22881 relevant @samp{__v850*__} preprocessor constant is defined.
22883 The preprocessor constants @code{__v850} and @code{__v851__} are always
22884 defined, regardless of which processor variant is the target.
22886 @item -mdisable-callt
22887 @itemx -mno-disable-callt
22888 @opindex mdisable-callt
22889 @opindex mno-disable-callt
22890 This option suppresses generation of the @code{CALLT} instruction for the
22891 v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the v850
22894 This option is enabled by default when the RH850 ABI is
22895 in use (see @option{-mrh850-abi}), and disabled by default when the
22896 GCC ABI is in use. If @code{CALLT} instructions are being generated
22897 then the C preprocessor symbol @code{__V850_CALLT__} is defined.
22903 Pass on (or do not pass on) the @option{-mrelax} command-line option
22907 @itemx -mno-long-jumps
22908 @opindex mlong-jumps
22909 @opindex mno-long-jumps
22910 Disable (or re-enable) the generation of PC-relative jump instructions.
22913 @itemx -mhard-float
22914 @opindex msoft-float
22915 @opindex mhard-float
22916 Disable (or re-enable) the generation of hardware floating point
22917 instructions. This option is only significant when the target
22918 architecture is @samp{V850E2V3} or higher. If hardware floating point
22919 instructions are being generated then the C preprocessor symbol
22920 @code{__FPU_OK__} is defined, otherwise the symbol
22921 @code{__NO_FPU__} is defined.
22925 Enables the use of the e3v5 LOOP instruction. The use of this
22926 instruction is not enabled by default when the e3v5 architecture is
22927 selected because its use is still experimental.
22931 @opindex mrh850-abi
22933 Enables support for the RH850 version of the V850 ABI. This is the
22934 default. With this version of the ABI the following rules apply:
22938 Integer sized structures and unions are returned via a memory pointer
22939 rather than a register.
22942 Large structures and unions (more than 8 bytes in size) are passed by
22946 Functions are aligned to 16-bit boundaries.
22949 The @option{-m8byte-align} command-line option is supported.
22952 The @option{-mdisable-callt} command-line option is enabled by
22953 default. The @option{-mno-disable-callt} command-line option is not
22957 When this version of the ABI is enabled the C preprocessor symbol
22958 @code{__V850_RH850_ABI__} is defined.
22962 Enables support for the old GCC version of the V850 ABI. With this
22963 version of the ABI the following rules apply:
22967 Integer sized structures and unions are returned in register @code{r10}.
22970 Large structures and unions (more than 8 bytes in size) are passed by
22974 Functions are aligned to 32-bit boundaries, unless optimizing for
22978 The @option{-m8byte-align} command-line option is not supported.
22981 The @option{-mdisable-callt} command-line option is supported but not
22982 enabled by default.
22985 When this version of the ABI is enabled the C preprocessor symbol
22986 @code{__V850_GCC_ABI__} is defined.
22988 @item -m8byte-align
22989 @itemx -mno-8byte-align
22990 @opindex m8byte-align
22991 @opindex mno-8byte-align
22992 Enables support for @code{double} and @code{long long} types to be
22993 aligned on 8-byte boundaries. The default is to restrict the
22994 alignment of all objects to at most 4-bytes. When
22995 @option{-m8byte-align} is in effect the C preprocessor symbol
22996 @code{__V850_8BYTE_ALIGN__} is defined.
22999 @opindex mbig-switch
23000 Generate code suitable for big switch tables. Use this option only if
23001 the assembler/linker complain about out of range branches within a switch
23006 This option causes r2 and r5 to be used in the code generated by
23007 the compiler. This setting is the default.
23009 @item -mno-app-regs
23010 @opindex mno-app-regs
23011 This option causes r2 and r5 to be treated as fixed registers.
23016 @subsection VAX Options
23017 @cindex VAX options
23019 These @samp{-m} options are defined for the VAX:
23024 Do not output certain jump instructions (@code{aobleq} and so on)
23025 that the Unix assembler for the VAX cannot handle across long
23030 Do output those jump instructions, on the assumption that the
23031 GNU assembler is being used.
23035 Output code for G-format floating-point numbers instead of D-format.
23038 @node Visium Options
23039 @subsection Visium Options
23040 @cindex Visium options
23046 A program which performs file I/O and is destined to run on an MCM target
23047 should be linked with this option. It causes the libraries libc.a and
23048 libdebug.a to be linked. The program should be run on the target under
23049 the control of the GDB remote debugging stub.
23053 A program which performs file I/O and is destined to run on the simulator
23054 should be linked with option. This causes libraries libc.a and libsim.a to
23058 @itemx -mhard-float
23060 @opindex mhard-float
23061 Generate code containing floating-point instructions. This is the
23065 @itemx -msoft-float
23067 @opindex msoft-float
23068 Generate code containing library calls for floating-point.
23070 @option{-msoft-float} changes the calling convention in the output file;
23071 therefore, it is only useful if you compile @emph{all} of a program with
23072 this option. In particular, you need to compile @file{libgcc.a}, the
23073 library that comes with GCC, with @option{-msoft-float} in order for
23076 @item -mcpu=@var{cpu_type}
23078 Set the instruction set, register set, and instruction scheduling parameters
23079 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
23080 @samp{mcm}, @samp{gr5} and @samp{gr6}.
23082 @samp{mcm} is a synonym of @samp{gr5} present for backward compatibility.
23084 By default (unless configured otherwise), GCC generates code for the GR5
23085 variant of the Visium architecture.
23087 With @option{-mcpu=gr6}, GCC generates code for the GR6 variant of the Visium
23088 architecture. The only difference from GR5 code is that the compiler will
23089 generate block move instructions.
23091 @item -mtune=@var{cpu_type}
23093 Set the instruction scheduling parameters for machine type @var{cpu_type},
23094 but do not set the instruction set or register set that the option
23095 @option{-mcpu=@var{cpu_type}} would.
23099 Generate code for the supervisor mode, where there are no restrictions on
23100 the access to general registers. This is the default.
23103 @opindex muser-mode
23104 Generate code for the user mode, where the access to some general registers
23105 is forbidden: on the GR5, registers r24 to r31 cannot be accessed in this
23106 mode; on the GR6, only registers r29 to r31 are affected.
23110 @subsection VMS Options
23112 These @samp{-m} options are defined for the VMS implementations:
23115 @item -mvms-return-codes
23116 @opindex mvms-return-codes
23117 Return VMS condition codes from @code{main}. The default is to return POSIX-style
23118 condition (e.g.@ error) codes.
23120 @item -mdebug-main=@var{prefix}
23121 @opindex mdebug-main=@var{prefix}
23122 Flag the first routine whose name starts with @var{prefix} as the main
23123 routine for the debugger.
23127 Default to 64-bit memory allocation routines.
23129 @item -mpointer-size=@var{size}
23130 @opindex mpointer-size=@var{size}
23131 Set the default size of pointers. Possible options for @var{size} are
23132 @samp{32} or @samp{short} for 32 bit pointers, @samp{64} or @samp{long}
23133 for 64 bit pointers, and @samp{no} for supporting only 32 bit pointers.
23134 The later option disables @code{pragma pointer_size}.
23137 @node VxWorks Options
23138 @subsection VxWorks Options
23139 @cindex VxWorks Options
23141 The options in this section are defined for all VxWorks targets.
23142 Options specific to the target hardware are listed with the other
23143 options for that target.
23148 GCC can generate code for both VxWorks kernels and real time processes
23149 (RTPs). This option switches from the former to the latter. It also
23150 defines the preprocessor macro @code{__RTP__}.
23153 @opindex non-static
23154 Link an RTP executable against shared libraries rather than static
23155 libraries. The options @option{-static} and @option{-shared} can
23156 also be used for RTPs (@pxref{Link Options}); @option{-static}
23163 These options are passed down to the linker. They are defined for
23164 compatibility with Diab.
23167 @opindex Xbind-lazy
23168 Enable lazy binding of function calls. This option is equivalent to
23169 @option{-Wl,-z,now} and is defined for compatibility with Diab.
23173 Disable lazy binding of function calls. This option is the default and
23174 is defined for compatibility with Diab.
23178 @subsection x86 Options
23179 @cindex x86 Options
23181 These @samp{-m} options are defined for the x86 family of computers.
23185 @item -march=@var{cpu-type}
23187 Generate instructions for the machine type @var{cpu-type}. In contrast to
23188 @option{-mtune=@var{cpu-type}}, which merely tunes the generated code
23189 for the specified @var{cpu-type}, @option{-march=@var{cpu-type}} allows GCC
23190 to generate code that may not run at all on processors other than the one
23191 indicated. Specifying @option{-march=@var{cpu-type}} implies
23192 @option{-mtune=@var{cpu-type}}.
23194 The choices for @var{cpu-type} are:
23198 This selects the CPU to generate code for at compilation time by determining
23199 the processor type of the compiling machine. Using @option{-march=native}
23200 enables all instruction subsets supported by the local machine (hence
23201 the result might not run on different machines). Using @option{-mtune=native}
23202 produces code optimized for the local machine under the constraints
23203 of the selected instruction set.
23206 Original Intel i386 CPU@.
23209 Intel i486 CPU@. (No scheduling is implemented for this chip.)
23213 Intel Pentium CPU with no MMX support.
23216 Intel Lakemont MCU, based on Intel Pentium CPU.
23219 Intel Pentium MMX CPU, based on Pentium core with MMX instruction set support.
23222 Intel Pentium Pro CPU@.
23225 When used with @option{-march}, the Pentium Pro
23226 instruction set is used, so the code runs on all i686 family chips.
23227 When used with @option{-mtune}, it has the same meaning as @samp{generic}.
23230 Intel Pentium II CPU, based on Pentium Pro core with MMX instruction set
23235 Intel Pentium III CPU, based on Pentium Pro core with MMX and SSE instruction
23239 Intel Pentium M; low-power version of Intel Pentium III CPU
23240 with MMX, SSE and SSE2 instruction set support. Used by Centrino notebooks.
23244 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set support.
23247 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 and SSE3 instruction
23251 Improved version of Intel Pentium 4 CPU with 64-bit extensions, MMX, SSE,
23252 SSE2 and SSE3 instruction set support.
23255 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
23256 instruction set support.
23259 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23260 SSE4.1, SSE4.2 and POPCNT instruction set support.
23263 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23264 SSE4.1, SSE4.2, POPCNT, AES and PCLMUL instruction set support.
23267 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23268 SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL instruction set support.
23271 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3, SSSE3,
23272 SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, FSGSBASE, RDRND and F16C
23273 instruction set support.
23276 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23277 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23278 BMI, BMI2 and F16C instruction set support.
23281 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23282 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23283 BMI, BMI2, F16C, RDSEED, ADCX and PREFETCHW instruction set support.
23286 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23287 SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23288 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC and
23289 XSAVES instruction set support.
23292 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3 and SSSE3
23293 instruction set support.
23296 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3, SSSE3,
23297 SSE4.1, SSE4.2, POPCNT, AES, PCLMUL and RDRND instruction set support.
23300 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23301 SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23302 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, AVX512F, AVX512PF, AVX512ER and
23303 AVX512CD instruction set support.
23305 @item skylake-avx512
23306 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, SSE, SSE2, SSE3,
23307 SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA,
23308 BMI, BMI2, F16C, RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F,
23309 AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction set support.
23312 AMD K6 CPU with MMX instruction set support.
23316 Improved versions of AMD K6 CPU with MMX and 3DNow!@: instruction set support.
23319 @itemx athlon-tbird
23320 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow!@: and SSE prefetch instructions
23326 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow!@: and full SSE
23327 instruction set support.
23333 Processors based on the AMD K8 core with x86-64 instruction set support,
23334 including the AMD Opteron, Athlon 64, and Athlon 64 FX processors.
23335 (This supersets MMX, SSE, SSE2, 3DNow!, enhanced 3DNow!@: and 64-bit
23336 instruction set extensions.)
23339 @itemx opteron-sse3
23340 @itemx athlon64-sse3
23341 Improved versions of AMD K8 cores with SSE3 instruction set support.
23345 CPUs based on AMD Family 10h cores with x86-64 instruction set support. (This
23346 supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, enhanced 3DNow!, ABM and 64-bit
23347 instruction set extensions.)
23350 CPUs based on AMD Family 15h cores with x86-64 instruction set support. (This
23351 supersets FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A,
23352 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.)
23354 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23355 supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, LWP, AES, PCL_MUL, CX16, MMX,
23356 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set
23359 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23360 supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, XOP, LWP, AES,
23361 PCL_MUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and
23362 64-bit instruction set extensions.
23364 AMD Family 15h core based CPUs with x86-64 instruction set support. (This
23365 supersets BMI, BMI2, TBM, F16C, FMA, FMA4, FSGSBASE, AVX, AVX2, XOP, LWP,
23366 AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1,
23367 SSE4.2, ABM and 64-bit instruction set extensions.
23370 AMD Family 17h core based CPUs with x86-64 instruction set support. (This
23371 supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX,
23372 SHA, CLZERO, AES, PCL_MUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3,
23373 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit
23374 instruction set extensions.
23377 CPUs based on AMD Family 14h cores with x86-64 instruction set support. (This
23378 supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, CX16, ABM and 64-bit
23379 instruction set extensions.)
23382 CPUs based on AMD Family 16h cores with x86-64 instruction set support. This
23383 includes MOVBE, F16C, BMI, AVX, PCL_MUL, AES, SSE4.2, SSE4.1, CX16, ABM,
23384 SSE4A, SSSE3, SSE3, SSE2, SSE, MMX and 64-bit instruction set extensions.
23387 IDT WinChip C6 CPU, dealt in same way as i486 with additional MMX instruction
23391 IDT WinChip 2 CPU, dealt in same way as i486 with additional MMX and 3DNow!@:
23392 instruction set support.
23395 VIA C3 CPU with MMX and 3DNow!@: instruction set support.
23396 (No scheduling is implemented for this chip.)
23399 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set support.
23400 (No scheduling is implemented for this chip.)
23403 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
23404 (No scheduling is implemented for this chip.)
23407 VIA Eden Samuel 2 CPU with MMX and 3DNow!@: instruction set support.
23408 (No scheduling is implemented for this chip.)
23411 VIA Eden Nehemiah CPU with MMX and SSE instruction set support.
23412 (No scheduling is implemented for this chip.)
23415 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction set support.
23416 (No scheduling is implemented for this chip.)
23419 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 instruction set support.
23420 (No scheduling is implemented for this chip.)
23423 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2,
23424 AVX and AVX2 instruction set support.
23425 (No scheduling is implemented for this chip.)
23428 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
23429 instruction set support.
23430 (No scheduling is implemented for this chip.)
23433 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
23434 instruction set support.
23435 (No scheduling is implemented for this chip.)
23438 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3
23439 instruction set support.
23440 (No scheduling is implemented for this chip.)
23443 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
23444 instruction set support.
23445 (No scheduling is implemented for this chip.)
23448 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
23449 instruction set support.
23450 (No scheduling is implemented for this chip.)
23453 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and SSE4.1
23454 instruction set support.
23455 (No scheduling is implemented for this chip.)
23458 AMD Geode embedded processor with MMX and 3DNow!@: instruction set support.
23461 @item -mtune=@var{cpu-type}
23463 Tune to @var{cpu-type} everything applicable about the generated code, except
23464 for the ABI and the set of available instructions.
23465 While picking a specific @var{cpu-type} schedules things appropriately
23466 for that particular chip, the compiler does not generate any code that
23467 cannot run on the default machine type unless you use a
23468 @option{-march=@var{cpu-type}} option.
23469 For example, if GCC is configured for i686-pc-linux-gnu
23470 then @option{-mtune=pentium4} generates code that is tuned for Pentium 4
23471 but still runs on i686 machines.
23473 The choices for @var{cpu-type} are the same as for @option{-march}.
23474 In addition, @option{-mtune} supports 2 extra choices for @var{cpu-type}:
23478 Produce code optimized for the most common IA32/@/AMD64/@/EM64T processors.
23479 If you know the CPU on which your code will run, then you should use
23480 the corresponding @option{-mtune} or @option{-march} option instead of
23481 @option{-mtune=generic}. But, if you do not know exactly what CPU users
23482 of your application will have, then you should use this option.
23484 As new processors are deployed in the marketplace, the behavior of this
23485 option will change. Therefore, if you upgrade to a newer version of
23486 GCC, code generation controlled by this option will change to reflect
23488 that are most common at the time that version of GCC is released.
23490 There is no @option{-march=generic} option because @option{-march}
23491 indicates the instruction set the compiler can use, and there is no
23492 generic instruction set applicable to all processors. In contrast,
23493 @option{-mtune} indicates the processor (or, in this case, collection of
23494 processors) for which the code is optimized.
23497 Produce code optimized for the most current Intel processors, which are
23498 Haswell and Silvermont for this version of GCC. If you know the CPU
23499 on which your code will run, then you should use the corresponding
23500 @option{-mtune} or @option{-march} option instead of @option{-mtune=intel}.
23501 But, if you want your application performs better on both Haswell and
23502 Silvermont, then you should use this option.
23504 As new Intel processors are deployed in the marketplace, the behavior of
23505 this option will change. Therefore, if you upgrade to a newer version of
23506 GCC, code generation controlled by this option will change to reflect
23507 the most current Intel processors at the time that version of GCC is
23510 There is no @option{-march=intel} option because @option{-march} indicates
23511 the instruction set the compiler can use, and there is no common
23512 instruction set applicable to all processors. In contrast,
23513 @option{-mtune} indicates the processor (or, in this case, collection of
23514 processors) for which the code is optimized.
23517 @item -mcpu=@var{cpu-type}
23519 A deprecated synonym for @option{-mtune}.
23521 @item -mfpmath=@var{unit}
23523 Generate floating-point arithmetic for selected unit @var{unit}. The choices
23524 for @var{unit} are:
23528 Use the standard 387 floating-point coprocessor present on the majority of chips and
23529 emulated otherwise. Code compiled with this option runs almost everywhere.
23530 The temporary results are computed in 80-bit precision instead of the precision
23531 specified by the type, resulting in slightly different results compared to most
23532 of other chips. See @option{-ffloat-store} for more detailed description.
23534 This is the default choice for x86-32 targets.
23537 Use scalar floating-point instructions present in the SSE instruction set.
23538 This instruction set is supported by Pentium III and newer chips,
23539 and in the AMD line
23540 by Athlon-4, Athlon XP and Athlon MP chips. The earlier version of the SSE
23541 instruction set supports only single-precision arithmetic, thus the double and
23542 extended-precision arithmetic are still done using 387. A later version, present
23543 only in Pentium 4 and AMD x86-64 chips, supports double-precision
23546 For the x86-32 compiler, you must use @option{-march=@var{cpu-type}}, @option{-msse}
23547 or @option{-msse2} switches to enable SSE extensions and make this option
23548 effective. For the x86-64 compiler, these extensions are enabled by default.
23550 The resulting code should be considerably faster in the majority of cases and avoid
23551 the numerical instability problems of 387 code, but may break some existing
23552 code that expects temporaries to be 80 bits.
23554 This is the default choice for the x86-64 compiler.
23559 Attempt to utilize both instruction sets at once. This effectively doubles the
23560 amount of available registers, and on chips with separate execution units for
23561 387 and SSE the execution resources too. Use this option with care, as it is
23562 still experimental, because the GCC register allocator does not model separate
23563 functional units well, resulting in unstable performance.
23566 @item -masm=@var{dialect}
23567 @opindex masm=@var{dialect}
23568 Output assembly instructions using selected @var{dialect}. Also affects
23569 which dialect is used for basic @code{asm} (@pxref{Basic Asm}) and
23570 extended @code{asm} (@pxref{Extended Asm}). Supported choices (in dialect
23571 order) are @samp{att} or @samp{intel}. The default is @samp{att}. Darwin does
23572 not support @samp{intel}.
23575 @itemx -mno-ieee-fp
23577 @opindex mno-ieee-fp
23578 Control whether or not the compiler uses IEEE floating-point
23579 comparisons. These correctly handle the case where the result of a
23580 comparison is unordered.
23585 @opindex mhard-float
23586 Generate output containing 80387 instructions for floating point.
23591 @opindex msoft-float
23592 Generate output containing library calls for floating point.
23594 @strong{Warning:} the requisite libraries are not part of GCC@.
23595 Normally the facilities of the machine's usual C compiler are used, but
23596 this can't be done directly in cross-compilation. You must make your
23597 own arrangements to provide suitable library functions for
23600 On machines where a function returns floating-point results in the 80387
23601 register stack, some floating-point opcodes may be emitted even if
23602 @option{-msoft-float} is used.
23604 @item -mno-fp-ret-in-387
23605 @opindex mno-fp-ret-in-387
23606 Do not use the FPU registers for return values of functions.
23608 The usual calling convention has functions return values of types
23609 @code{float} and @code{double} in an FPU register, even if there
23610 is no FPU@. The idea is that the operating system should emulate
23613 The option @option{-mno-fp-ret-in-387} causes such values to be returned
23614 in ordinary CPU registers instead.
23616 @item -mno-fancy-math-387
23617 @opindex mno-fancy-math-387
23618 Some 387 emulators do not support the @code{sin}, @code{cos} and
23619 @code{sqrt} instructions for the 387. Specify this option to avoid
23620 generating those instructions. This option is the default on
23621 OpenBSD and NetBSD@. This option is overridden when @option{-march}
23622 indicates that the target CPU always has an FPU and so the
23623 instruction does not need emulation. These
23624 instructions are not generated unless you also use the
23625 @option{-funsafe-math-optimizations} switch.
23627 @item -malign-double
23628 @itemx -mno-align-double
23629 @opindex malign-double
23630 @opindex mno-align-double
23631 Control whether GCC aligns @code{double}, @code{long double}, and
23632 @code{long long} variables on a two-word boundary or a one-word
23633 boundary. Aligning @code{double} variables on a two-word boundary
23634 produces code that runs somewhat faster on a Pentium at the
23635 expense of more memory.
23637 On x86-64, @option{-malign-double} is enabled by default.
23639 @strong{Warning:} if you use the @option{-malign-double} switch,
23640 structures containing the above types are aligned differently than
23641 the published application binary interface specifications for the x86-32
23642 and are not binary compatible with structures in code compiled
23643 without that switch.
23645 @item -m96bit-long-double
23646 @itemx -m128bit-long-double
23647 @opindex m96bit-long-double
23648 @opindex m128bit-long-double
23649 These switches control the size of @code{long double} type. The x86-32
23650 application binary interface specifies the size to be 96 bits,
23651 so @option{-m96bit-long-double} is the default in 32-bit mode.
23653 Modern architectures (Pentium and newer) prefer @code{long double}
23654 to be aligned to an 8- or 16-byte boundary. In arrays or structures
23655 conforming to the ABI, this is not possible. So specifying
23656 @option{-m128bit-long-double} aligns @code{long double}
23657 to a 16-byte boundary by padding the @code{long double} with an additional
23660 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
23661 its ABI specifies that @code{long double} is aligned on 16-byte boundary.
23663 Notice that neither of these options enable any extra precision over the x87
23664 standard of 80 bits for a @code{long double}.
23666 @strong{Warning:} if you override the default value for your target ABI, this
23667 changes the size of
23668 structures and arrays containing @code{long double} variables,
23669 as well as modifying the function calling convention for functions taking
23670 @code{long double}. Hence they are not binary-compatible
23671 with code compiled without that switch.
23673 @item -mlong-double-64
23674 @itemx -mlong-double-80
23675 @itemx -mlong-double-128
23676 @opindex mlong-double-64
23677 @opindex mlong-double-80
23678 @opindex mlong-double-128
23679 These switches control the size of @code{long double} type. A size
23680 of 64 bits makes the @code{long double} type equivalent to the @code{double}
23681 type. This is the default for 32-bit Bionic C library. A size
23682 of 128 bits makes the @code{long double} type equivalent to the
23683 @code{__float128} type. This is the default for 64-bit Bionic C library.
23685 @strong{Warning:} if you override the default value for your target ABI, this
23686 changes the size of
23687 structures and arrays containing @code{long double} variables,
23688 as well as modifying the function calling convention for functions taking
23689 @code{long double}. Hence they are not binary-compatible
23690 with code compiled without that switch.
23692 @item -malign-data=@var{type}
23693 @opindex malign-data
23694 Control how GCC aligns variables. Supported values for @var{type} are
23695 @samp{compat} uses increased alignment value compatible uses GCC 4.8
23696 and earlier, @samp{abi} uses alignment value as specified by the
23697 psABI, and @samp{cacheline} uses increased alignment value to match
23698 the cache line size. @samp{compat} is the default.
23700 @item -mlarge-data-threshold=@var{threshold}
23701 @opindex mlarge-data-threshold
23702 When @option{-mcmodel=medium} is specified, data objects larger than
23703 @var{threshold} are placed in the large data section. This value must be the
23704 same across all objects linked into the binary, and defaults to 65535.
23708 Use a different function-calling convention, in which functions that
23709 take a fixed number of arguments return with the @code{ret @var{num}}
23710 instruction, which pops their arguments while returning. This saves one
23711 instruction in the caller since there is no need to pop the arguments
23714 You can specify that an individual function is called with this calling
23715 sequence with the function attribute @code{stdcall}. You can also
23716 override the @option{-mrtd} option by using the function attribute
23717 @code{cdecl}. @xref{Function Attributes}.
23719 @strong{Warning:} this calling convention is incompatible with the one
23720 normally used on Unix, so you cannot use it if you need to call
23721 libraries compiled with the Unix compiler.
23723 Also, you must provide function prototypes for all functions that
23724 take variable numbers of arguments (including @code{printf});
23725 otherwise incorrect code is generated for calls to those
23728 In addition, seriously incorrect code results if you call a
23729 function with too many arguments. (Normally, extra arguments are
23730 harmlessly ignored.)
23732 @item -mregparm=@var{num}
23734 Control how many registers are used to pass integer arguments. By
23735 default, no registers are used to pass arguments, and at most 3
23736 registers can be used. You can control this behavior for a specific
23737 function by using the function attribute @code{regparm}.
23738 @xref{Function Attributes}.
23740 @strong{Warning:} if you use this switch, and
23741 @var{num} is nonzero, then you must build all modules with the same
23742 value, including any libraries. This includes the system libraries and
23746 @opindex msseregparm
23747 Use SSE register passing conventions for float and double arguments
23748 and return values. You can control this behavior for a specific
23749 function by using the function attribute @code{sseregparm}.
23750 @xref{Function Attributes}.
23752 @strong{Warning:} if you use this switch then you must build all
23753 modules with the same value, including any libraries. This includes
23754 the system libraries and startup modules.
23756 @item -mvect8-ret-in-mem
23757 @opindex mvect8-ret-in-mem
23758 Return 8-byte vectors in memory instead of MMX registers. This is the
23759 default on Solaris@tie{}8 and 9 and VxWorks to match the ABI of the Sun
23760 Studio compilers until version 12. Later compiler versions (starting
23761 with Studio 12 Update@tie{}1) follow the ABI used by other x86 targets, which
23762 is the default on Solaris@tie{}10 and later. @emph{Only} use this option if
23763 you need to remain compatible with existing code produced by those
23764 previous compiler versions or older versions of GCC@.
23773 Set 80387 floating-point precision to 32, 64 or 80 bits. When @option{-mpc32}
23774 is specified, the significands of results of floating-point operations are
23775 rounded to 24 bits (single precision); @option{-mpc64} rounds the
23776 significands of results of floating-point operations to 53 bits (double
23777 precision) and @option{-mpc80} rounds the significands of results of
23778 floating-point operations to 64 bits (extended double precision), which is
23779 the default. When this option is used, floating-point operations in higher
23780 precisions are not available to the programmer without setting the FPU
23781 control word explicitly.
23783 Setting the rounding of floating-point operations to less than the default
23784 80 bits can speed some programs by 2% or more. Note that some mathematical
23785 libraries assume that extended-precision (80-bit) floating-point operations
23786 are enabled by default; routines in such libraries could suffer significant
23787 loss of accuracy, typically through so-called ``catastrophic cancellation'',
23788 when this option is used to set the precision to less than extended precision.
23790 @item -mstackrealign
23791 @opindex mstackrealign
23792 Realign the stack at entry. On the x86, the @option{-mstackrealign}
23793 option generates an alternate prologue and epilogue that realigns the
23794 run-time stack if necessary. This supports mixing legacy codes that keep
23795 4-byte stack alignment with modern codes that keep 16-byte stack alignment for
23796 SSE compatibility. See also the attribute @code{force_align_arg_pointer},
23797 applicable to individual functions.
23799 @item -mpreferred-stack-boundary=@var{num}
23800 @opindex mpreferred-stack-boundary
23801 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
23802 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
23803 the default is 4 (16 bytes or 128 bits).
23805 @strong{Warning:} When generating code for the x86-64 architecture with
23806 SSE extensions disabled, @option{-mpreferred-stack-boundary=3} can be
23807 used to keep the stack boundary aligned to 8 byte boundary. Since
23808 x86-64 ABI require 16 byte stack alignment, this is ABI incompatible and
23809 intended to be used in controlled environment where stack space is
23810 important limitation. This option leads to wrong code when functions
23811 compiled with 16 byte stack alignment (such as functions from a standard
23812 library) are called with misaligned stack. In this case, SSE
23813 instructions may lead to misaligned memory access traps. In addition,
23814 variable arguments are handled incorrectly for 16 byte aligned
23815 objects (including x87 long double and __int128), leading to wrong
23816 results. You must build all modules with
23817 @option{-mpreferred-stack-boundary=3}, including any libraries. This
23818 includes the system libraries and startup modules.
23820 @item -mincoming-stack-boundary=@var{num}
23821 @opindex mincoming-stack-boundary
23822 Assume the incoming stack is aligned to a 2 raised to @var{num} byte
23823 boundary. If @option{-mincoming-stack-boundary} is not specified,
23824 the one specified by @option{-mpreferred-stack-boundary} is used.
23826 On Pentium and Pentium Pro, @code{double} and @code{long double} values
23827 should be aligned to an 8-byte boundary (see @option{-malign-double}) or
23828 suffer significant run time performance penalties. On Pentium III, the
23829 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
23830 properly if it is not 16-byte aligned.
23832 To ensure proper alignment of this values on the stack, the stack boundary
23833 must be as aligned as that required by any value stored on the stack.
23834 Further, every function must be generated such that it keeps the stack
23835 aligned. Thus calling a function compiled with a higher preferred
23836 stack boundary from a function compiled with a lower preferred stack
23837 boundary most likely misaligns the stack. It is recommended that
23838 libraries that use callbacks always use the default setting.
23840 This extra alignment does consume extra stack space, and generally
23841 increases code size. Code that is sensitive to stack space usage, such
23842 as embedded systems and operating system kernels, may want to reduce the
23843 preferred alignment to @option{-mpreferred-stack-boundary=2}.
23900 @itemx -mavx512ifma
23901 @opindex mavx512ifma
23903 @itemx -mavx512vbmi
23904 @opindex mavx512vbmi
23916 @opindex mclfushopt
23933 @itemx -mprefetchwt1
23934 @opindex mprefetchwt1
23990 These switches enable the use of instructions in the MMX, SSE,
23991 SSE2, SSE3, SSSE3, SSE4.1, AVX, AVX2, AVX512F, AVX512PF, AVX512ER, AVX512CD,
23992 SHA, AES, PCLMUL, FSGSBASE, RDRND, F16C, FMA, SSE4A, FMA4, XOP, LWP, ABM,
23993 AVX512VL, AVX512BW, AVX512DQ, AVX512IFMA AVX512VBMI, BMI, BMI2, FXSR,
23994 XSAVE, XSAVEOPT, LZCNT, RTM, MPX, MWAITX, PKU or 3DNow!@:
23995 extended instruction sets. Each has a corresponding @option{-mno-} option
23996 to disable use of these instructions.
23998 These extensions are also available as built-in functions: see
23999 @ref{x86 Built-in Functions}, for details of the functions enabled and
24000 disabled by these switches.
24002 To generate SSE/SSE2 instructions automatically from floating-point
24003 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
24005 GCC depresses SSEx instructions when @option{-mavx} is used. Instead, it
24006 generates new AVX instructions or AVX equivalence for all SSEx instructions
24009 These options enable GCC to use these extended instructions in
24010 generated code, even without @option{-mfpmath=sse}. Applications that
24011 perform run-time CPU detection must compile separate files for each
24012 supported architecture, using the appropriate flags. In particular,
24013 the file containing the CPU detection code should be compiled without
24016 @item -mdump-tune-features
24017 @opindex mdump-tune-features
24018 This option instructs GCC to dump the names of the x86 performance
24019 tuning features and default settings. The names can be used in
24020 @option{-mtune-ctrl=@var{feature-list}}.
24022 @item -mtune-ctrl=@var{feature-list}
24023 @opindex mtune-ctrl=@var{feature-list}
24024 This option is used to do fine grain control of x86 code generation features.
24025 @var{feature-list} is a comma separated list of @var{feature} names. See also
24026 @option{-mdump-tune-features}. When specified, the @var{feature} is turned
24027 on if it is not preceded with @samp{^}, otherwise, it is turned off.
24028 @option{-mtune-ctrl=@var{feature-list}} is intended to be used by GCC
24029 developers. Using it may lead to code paths not covered by testing and can
24030 potentially result in compiler ICEs or runtime errors.
24033 @opindex mno-default
24034 This option instructs GCC to turn off all tunable features. See also
24035 @option{-mtune-ctrl=@var{feature-list}} and @option{-mdump-tune-features}.
24039 This option instructs GCC to emit a @code{cld} instruction in the prologue
24040 of functions that use string instructions. String instructions depend on
24041 the DF flag to select between autoincrement or autodecrement mode. While the
24042 ABI specifies the DF flag to be cleared on function entry, some operating
24043 systems violate this specification by not clearing the DF flag in their
24044 exception dispatchers. The exception handler can be invoked with the DF flag
24045 set, which leads to wrong direction mode when string instructions are used.
24046 This option can be enabled by default on 32-bit x86 targets by configuring
24047 GCC with the @option{--enable-cld} configure option. Generation of @code{cld}
24048 instructions can be suppressed with the @option{-mno-cld} compiler option
24052 @opindex mvzeroupper
24053 This option instructs GCC to emit a @code{vzeroupper} instruction
24054 before a transfer of control flow out of the function to minimize
24055 the AVX to SSE transition penalty as well as remove unnecessary @code{zeroupper}
24058 @item -mprefer-avx128
24059 @opindex mprefer-avx128
24060 This option instructs GCC to use 128-bit AVX instructions instead of
24061 256-bit AVX instructions in the auto-vectorizer.
24065 This option enables GCC to generate @code{CMPXCHG16B} instructions.
24066 @code{CMPXCHG16B} allows for atomic operations on 128-bit double quadword
24067 (or oword) data types.
24068 This is useful for high-resolution counters that can be updated
24069 by multiple processors (or cores). This instruction is generated as part of
24070 atomic built-in functions: see @ref{__sync Builtins} or
24071 @ref{__atomic Builtins} for details.
24075 This option enables generation of @code{SAHF} instructions in 64-bit code.
24076 Early Intel Pentium 4 CPUs with Intel 64 support,
24077 prior to the introduction of Pentium 4 G1 step in December 2005,
24078 lacked the @code{LAHF} and @code{SAHF} instructions
24079 which are supported by AMD64.
24080 These are load and store instructions, respectively, for certain status flags.
24081 In 64-bit mode, the @code{SAHF} instruction is used to optimize @code{fmod},
24082 @code{drem}, and @code{remainder} built-in functions;
24083 see @ref{Other Builtins} for details.
24087 This option enables use of the @code{movbe} instruction to implement
24088 @code{__builtin_bswap32} and @code{__builtin_bswap64}.
24092 This option enables built-in functions @code{__builtin_ia32_crc32qi},
24093 @code{__builtin_ia32_crc32hi}, @code{__builtin_ia32_crc32si} and
24094 @code{__builtin_ia32_crc32di} to generate the @code{crc32} machine instruction.
24098 This option enables use of @code{RCPSS} and @code{RSQRTSS} instructions
24099 (and their vectorized variants @code{RCPPS} and @code{RSQRTPS})
24100 with an additional Newton-Raphson step
24101 to increase precision instead of @code{DIVSS} and @code{SQRTSS}
24102 (and their vectorized
24103 variants) for single-precision floating-point arguments. These instructions
24104 are generated only when @option{-funsafe-math-optimizations} is enabled
24105 together with @option{-ffinite-math-only} and @option{-fno-trapping-math}.
24106 Note that while the throughput of the sequence is higher than the throughput
24107 of the non-reciprocal instruction, the precision of the sequence can be
24108 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 0.99999994).
24110 Note that GCC implements @code{1.0f/sqrtf(@var{x})} in terms of @code{RSQRTSS}
24111 (or @code{RSQRTPS}) already with @option{-ffast-math} (or the above option
24112 combination), and doesn't need @option{-mrecip}.
24114 Also note that GCC emits the above sequence with additional Newton-Raphson step
24115 for vectorized single-float division and vectorized @code{sqrtf(@var{x})}
24116 already with @option{-ffast-math} (or the above option combination), and
24117 doesn't need @option{-mrecip}.
24119 @item -mrecip=@var{opt}
24120 @opindex mrecip=opt
24121 This option controls which reciprocal estimate instructions
24122 may be used. @var{opt} is a comma-separated list of options, which may
24123 be preceded by a @samp{!} to invert the option:
24127 Enable all estimate instructions.
24130 Enable the default instructions, equivalent to @option{-mrecip}.
24133 Disable all estimate instructions, equivalent to @option{-mno-recip}.
24136 Enable the approximation for scalar division.
24139 Enable the approximation for vectorized division.
24142 Enable the approximation for scalar square root.
24145 Enable the approximation for vectorized square root.
24148 So, for example, @option{-mrecip=all,!sqrt} enables
24149 all of the reciprocal approximations, except for square root.
24151 @item -mveclibabi=@var{type}
24152 @opindex mveclibabi
24153 Specifies the ABI type to use for vectorizing intrinsics using an
24154 external library. Supported values for @var{type} are @samp{svml}
24155 for the Intel short
24156 vector math library and @samp{acml} for the AMD math core library.
24157 To use this option, both @option{-ftree-vectorize} and
24158 @option{-funsafe-math-optimizations} have to be enabled, and an SVML or ACML
24159 ABI-compatible library must be specified at link time.
24161 GCC currently emits calls to @code{vmldExp2},
24162 @code{vmldLn2}, @code{vmldLog102}, @code{vmldLog102}, @code{vmldPow2},
24163 @code{vmldTanh2}, @code{vmldTan2}, @code{vmldAtan2}, @code{vmldAtanh2},
24164 @code{vmldCbrt2}, @code{vmldSinh2}, @code{vmldSin2}, @code{vmldAsinh2},
24165 @code{vmldAsin2}, @code{vmldCosh2}, @code{vmldCos2}, @code{vmldAcosh2},
24166 @code{vmldAcos2}, @code{vmlsExp4}, @code{vmlsLn4}, @code{vmlsLog104},
24167 @code{vmlsLog104}, @code{vmlsPow4}, @code{vmlsTanh4}, @code{vmlsTan4},
24168 @code{vmlsAtan4}, @code{vmlsAtanh4}, @code{vmlsCbrt4}, @code{vmlsSinh4},
24169 @code{vmlsSin4}, @code{vmlsAsinh4}, @code{vmlsAsin4}, @code{vmlsCosh4},
24170 @code{vmlsCos4}, @code{vmlsAcosh4} and @code{vmlsAcos4} for corresponding
24171 function type when @option{-mveclibabi=svml} is used, and @code{__vrd2_sin},
24172 @code{__vrd2_cos}, @code{__vrd2_exp}, @code{__vrd2_log}, @code{__vrd2_log2},
24173 @code{__vrd2_log10}, @code{__vrs4_sinf}, @code{__vrs4_cosf},
24174 @code{__vrs4_expf}, @code{__vrs4_logf}, @code{__vrs4_log2f},
24175 @code{__vrs4_log10f} and @code{__vrs4_powf} for the corresponding function type
24176 when @option{-mveclibabi=acml} is used.
24178 @item -mabi=@var{name}
24180 Generate code for the specified calling convention. Permissible values
24181 are @samp{sysv} for the ABI used on GNU/Linux and other systems, and
24182 @samp{ms} for the Microsoft ABI. The default is to use the Microsoft
24183 ABI when targeting Microsoft Windows and the SysV ABI on all other systems.
24184 You can control this behavior for specific functions by
24185 using the function attributes @code{ms_abi} and @code{sysv_abi}.
24186 @xref{Function Attributes}.
24188 @item -mtls-dialect=@var{type}
24189 @opindex mtls-dialect
24190 Generate code to access thread-local storage using the @samp{gnu} or
24191 @samp{gnu2} conventions. @samp{gnu} is the conservative default;
24192 @samp{gnu2} is more efficient, but it may add compile- and run-time
24193 requirements that cannot be satisfied on all systems.
24196 @itemx -mno-push-args
24197 @opindex mpush-args
24198 @opindex mno-push-args
24199 Use PUSH operations to store outgoing parameters. This method is shorter
24200 and usually equally fast as method using SUB/MOV operations and is enabled
24201 by default. In some cases disabling it may improve performance because of
24202 improved scheduling and reduced dependencies.
24204 @item -maccumulate-outgoing-args
24205 @opindex maccumulate-outgoing-args
24206 If enabled, the maximum amount of space required for outgoing arguments is
24207 computed in the function prologue. This is faster on most modern CPUs
24208 because of reduced dependencies, improved scheduling and reduced stack usage
24209 when the preferred stack boundary is not equal to 2. The drawback is a notable
24210 increase in code size. This switch implies @option{-mno-push-args}.
24214 Support thread-safe exception handling on MinGW. Programs that rely
24215 on thread-safe exception handling must compile and link all code with the
24216 @option{-mthreads} option. When compiling, @option{-mthreads} defines
24217 @option{-D_MT}; when linking, it links in a special thread helper library
24218 @option{-lmingwthrd} which cleans up per-thread exception-handling data.
24220 @item -mms-bitfields
24221 @itemx -mno-ms-bitfields
24222 @opindex mms-bitfields
24223 @opindex mno-ms-bitfields
24225 Enable/disable bit-field layout compatible with the native Microsoft
24228 If @code{packed} is used on a structure, or if bit-fields are used,
24229 it may be that the Microsoft ABI lays out the structure differently
24230 than the way GCC normally does. Particularly when moving packed
24231 data between functions compiled with GCC and the native Microsoft compiler
24232 (either via function call or as data in a file), it may be necessary to access
24235 This option is enabled by default for Microsoft Windows
24236 targets. This behavior can also be controlled locally by use of variable
24237 or type attributes. For more information, see @ref{x86 Variable Attributes}
24238 and @ref{x86 Type Attributes}.
24240 The Microsoft structure layout algorithm is fairly simple with the exception
24241 of the bit-field packing.
24242 The padding and alignment of members of structures and whether a bit-field
24243 can straddle a storage-unit boundary are determine by these rules:
24246 @item Structure members are stored sequentially in the order in which they are
24247 declared: the first member has the lowest memory address and the last member
24250 @item Every data object has an alignment requirement. The alignment requirement
24251 for all data except structures, unions, and arrays is either the size of the
24252 object or the current packing size (specified with either the
24253 @code{aligned} attribute or the @code{pack} pragma),
24254 whichever is less. For structures, unions, and arrays,
24255 the alignment requirement is the largest alignment requirement of its members.
24256 Every object is allocated an offset so that:
24259 offset % alignment_requirement == 0
24262 @item Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte allocation
24263 unit if the integral types are the same size and if the next bit-field fits
24264 into the current allocation unit without crossing the boundary imposed by the
24265 common alignment requirements of the bit-fields.
24268 MSVC interprets zero-length bit-fields in the following ways:
24271 @item If a zero-length bit-field is inserted between two bit-fields that
24272 are normally coalesced, the bit-fields are not coalesced.
24279 unsigned long bf_1 : 12;
24281 unsigned long bf_2 : 12;
24286 The size of @code{t1} is 8 bytes with the zero-length bit-field. If the
24287 zero-length bit-field were removed, @code{t1}'s size would be 4 bytes.
24289 @item If a zero-length bit-field is inserted after a bit-field, @code{foo}, and the
24290 alignment of the zero-length bit-field is greater than the member that follows it,
24291 @code{bar}, @code{bar} is aligned as the type of the zero-length bit-field.
24312 For @code{t2}, @code{bar} is placed at offset 2, rather than offset 1.
24313 Accordingly, the size of @code{t2} is 4. For @code{t3}, the zero-length
24314 bit-field does not affect the alignment of @code{bar} or, as a result, the size
24317 Taking this into account, it is important to note the following:
24320 @item If a zero-length bit-field follows a normal bit-field, the type of the
24321 zero-length bit-field may affect the alignment of the structure as whole. For
24322 example, @code{t2} has a size of 4 bytes, since the zero-length bit-field follows a
24323 normal bit-field, and is of type short.
24325 @item Even if a zero-length bit-field is not followed by a normal bit-field, it may
24326 still affect the alignment of the structure:
24337 Here, @code{t4} takes up 4 bytes.
24340 @item Zero-length bit-fields following non-bit-field members are ignored:
24352 Here, @code{t5} takes up 2 bytes.
24356 @item -mno-align-stringops
24357 @opindex mno-align-stringops
24358 Do not align the destination of inlined string operations. This switch reduces
24359 code size and improves performance in case the destination is already aligned,
24360 but GCC doesn't know about it.
24362 @item -minline-all-stringops
24363 @opindex minline-all-stringops
24364 By default GCC inlines string operations only when the destination is
24365 known to be aligned to least a 4-byte boundary.
24366 This enables more inlining and increases code
24367 size, but may improve performance of code that depends on fast
24368 @code{memcpy}, @code{strlen},
24369 and @code{memset} for short lengths.
24371 @item -minline-stringops-dynamically
24372 @opindex minline-stringops-dynamically
24373 For string operations of unknown size, use run-time checks with
24374 inline code for small blocks and a library call for large blocks.
24376 @item -mstringop-strategy=@var{alg}
24377 @opindex mstringop-strategy=@var{alg}
24378 Override the internal decision heuristic for the particular algorithm to use
24379 for inlining string operations. The allowed values for @var{alg} are:
24385 Expand using i386 @code{rep} prefix of the specified size.
24389 @itemx unrolled_loop
24390 Expand into an inline loop.
24393 Always use a library call.
24396 @item -mmemcpy-strategy=@var{strategy}
24397 @opindex mmemcpy-strategy=@var{strategy}
24398 Override the internal decision heuristic to decide if @code{__builtin_memcpy}
24399 should be inlined and what inline algorithm to use when the expected size
24400 of the copy operation is known. @var{strategy}
24401 is a comma-separated list of @var{alg}:@var{max_size}:@var{dest_align} triplets.
24402 @var{alg} is specified in @option{-mstringop-strategy}, @var{max_size} specifies
24403 the max byte size with which inline algorithm @var{alg} is allowed. For the last
24404 triplet, the @var{max_size} must be @code{-1}. The @var{max_size} of the triplets
24405 in the list must be specified in increasing order. The minimal byte size for
24406 @var{alg} is @code{0} for the first triplet and @code{@var{max_size} + 1} of the
24409 @item -mmemset-strategy=@var{strategy}
24410 @opindex mmemset-strategy=@var{strategy}
24411 The option is similar to @option{-mmemcpy-strategy=} except that it is to control
24412 @code{__builtin_memset} expansion.
24414 @item -momit-leaf-frame-pointer
24415 @opindex momit-leaf-frame-pointer
24416 Don't keep the frame pointer in a register for leaf functions. This
24417 avoids the instructions to save, set up, and restore frame pointers and
24418 makes an extra register available in leaf functions. The option
24419 @option{-fomit-leaf-frame-pointer} removes the frame pointer for leaf functions,
24420 which might make debugging harder.
24422 @item -mtls-direct-seg-refs
24423 @itemx -mno-tls-direct-seg-refs
24424 @opindex mtls-direct-seg-refs
24425 Controls whether TLS variables may be accessed with offsets from the
24426 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
24427 or whether the thread base pointer must be added. Whether or not this
24428 is valid depends on the operating system, and whether it maps the
24429 segment to cover the entire TLS area.
24431 For systems that use the GNU C Library, the default is on.
24434 @itemx -mno-sse2avx
24436 Specify that the assembler should encode SSE instructions with VEX
24437 prefix. The option @option{-mavx} turns this on by default.
24442 If profiling is active (@option{-pg}), put the profiling
24443 counter call before the prologue.
24444 Note: On x86 architectures the attribute @code{ms_hook_prologue}
24445 isn't possible at the moment for @option{-mfentry} and @option{-pg}.
24447 @item -mrecord-mcount
24448 @itemx -mno-record-mcount
24449 @opindex mrecord-mcount
24450 If profiling is active (@option{-pg}), generate a __mcount_loc section
24451 that contains pointers to each profiling call. This is useful for
24452 automatically patching and out calls.
24455 @itemx -mno-nop-mcount
24456 @opindex mnop-mcount
24457 If profiling is active (@option{-pg}), generate the calls to
24458 the profiling functions as nops. This is useful when they
24459 should be patched in later dynamically. This is likely only
24460 useful together with @option{-mrecord-mcount}.
24462 @item -mskip-rax-setup
24463 @itemx -mno-skip-rax-setup
24464 @opindex mskip-rax-setup
24465 When generating code for the x86-64 architecture with SSE extensions
24466 disabled, @option{-mskip-rax-setup} can be used to skip setting up RAX
24467 register when there are no variable arguments passed in vector registers.
24469 @strong{Warning:} Since RAX register is used to avoid unnecessarily
24470 saving vector registers on stack when passing variable arguments, the
24471 impacts of this option are callees may waste some stack space,
24472 misbehave or jump to a random location. GCC 4.4 or newer don't have
24473 those issues, regardless the RAX register value.
24476 @itemx -mno-8bit-idiv
24477 @opindex m8bit-idiv
24478 On some processors, like Intel Atom, 8-bit unsigned integer divide is
24479 much faster than 32-bit/64-bit integer divide. This option generates a
24480 run-time check. If both dividend and divisor are within range of 0
24481 to 255, 8-bit unsigned integer divide is used instead of
24482 32-bit/64-bit integer divide.
24484 @item -mavx256-split-unaligned-load
24485 @itemx -mavx256-split-unaligned-store
24486 @opindex mavx256-split-unaligned-load
24487 @opindex mavx256-split-unaligned-store
24488 Split 32-byte AVX unaligned load and store.
24490 @item -mstack-protector-guard=@var{guard}
24491 @opindex mstack-protector-guard=@var{guard}
24492 Generate stack protection code using canary at @var{guard}. Supported
24493 locations are @samp{global} for global canary or @samp{tls} for per-thread
24494 canary in the TLS block (the default). This option has effect only when
24495 @option{-fstack-protector} or @option{-fstack-protector-all} is specified.
24497 @item -mmitigate-rop
24498 @opindex mmitigate-rop
24499 Try to avoid generating code sequences that contain unintended return
24500 opcodes, to mitigate against certain forms of attack. At the moment,
24501 this option is limited in what it can do and should not be relied
24502 on to provide serious protection.
24504 @item -mgeneral-regs-only
24505 @opindex mgeneral-regs-only
24506 Generate code that uses only the general-purpose registers. This
24507 prevents the compiler from using floating-point, vector, mask and bound
24512 These @samp{-m} switches are supported in addition to the above
24513 on x86-64 processors in 64-bit environments.
24526 Generate code for a 16-bit, 32-bit or 64-bit environment.
24527 The @option{-m32} option sets @code{int}, @code{long}, and pointer types
24529 generates code that runs on any i386 system.
24531 The @option{-m64} option sets @code{int} to 32 bits and @code{long} and pointer
24532 types to 64 bits, and generates code for the x86-64 architecture.
24533 For Darwin only the @option{-m64} option also turns off the @option{-fno-pic}
24534 and @option{-mdynamic-no-pic} options.
24536 The @option{-mx32} option sets @code{int}, @code{long}, and pointer types
24538 generates code for the x86-64 architecture.
24540 The @option{-m16} option is the same as @option{-m32}, except for that
24541 it outputs the @code{.code16gcc} assembly directive at the beginning of
24542 the assembly output so that the binary can run in 16-bit mode.
24544 The @option{-miamcu} option generates code which conforms to Intel MCU
24545 psABI. It requires the @option{-m32} option to be turned on.
24547 @item -mno-red-zone
24548 @opindex mno-red-zone
24549 Do not use a so-called ``red zone'' for x86-64 code. The red zone is mandated
24550 by the x86-64 ABI; it is a 128-byte area beyond the location of the
24551 stack pointer that is not modified by signal or interrupt handlers
24552 and therefore can be used for temporary data without adjusting the stack
24553 pointer. The flag @option{-mno-red-zone} disables this red zone.
24555 @item -mcmodel=small
24556 @opindex mcmodel=small
24557 Generate code for the small code model: the program and its symbols must
24558 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
24559 Programs can be statically or dynamically linked. This is the default
24562 @item -mcmodel=kernel
24563 @opindex mcmodel=kernel
24564 Generate code for the kernel code model. The kernel runs in the
24565 negative 2 GB of the address space.
24566 This model has to be used for Linux kernel code.
24568 @item -mcmodel=medium
24569 @opindex mcmodel=medium
24570 Generate code for the medium model: the program is linked in the lower 2
24571 GB of the address space. Small symbols are also placed there. Symbols
24572 with sizes larger than @option{-mlarge-data-threshold} are put into
24573 large data or BSS sections and can be located above 2GB. Programs can
24574 be statically or dynamically linked.
24576 @item -mcmodel=large
24577 @opindex mcmodel=large
24578 Generate code for the large model. This model makes no assumptions
24579 about addresses and sizes of sections.
24581 @item -maddress-mode=long
24582 @opindex maddress-mode=long
24583 Generate code for long address mode. This is only supported for 64-bit
24584 and x32 environments. It is the default address mode for 64-bit
24587 @item -maddress-mode=short
24588 @opindex maddress-mode=short
24589 Generate code for short address mode. This is only supported for 32-bit
24590 and x32 environments. It is the default address mode for 32-bit and
24594 @node x86 Windows Options
24595 @subsection x86 Windows Options
24596 @cindex x86 Windows Options
24597 @cindex Windows Options for x86
24599 These additional options are available for Microsoft Windows targets:
24605 specifies that a console application is to be generated, by
24606 instructing the linker to set the PE header subsystem type
24607 required for console applications.
24608 This option is available for Cygwin and MinGW targets and is
24609 enabled by default on those targets.
24613 This option is available for Cygwin and MinGW targets. It
24614 specifies that a DLL---a dynamic link library---is to be
24615 generated, enabling the selection of the required runtime
24616 startup object and entry point.
24618 @item -mnop-fun-dllimport
24619 @opindex mnop-fun-dllimport
24620 This option is available for Cygwin and MinGW targets. It
24621 specifies that the @code{dllimport} attribute should be ignored.
24625 This option is available for MinGW targets. It specifies
24626 that MinGW-specific thread support is to be used.
24630 This option is available for MinGW-w64 targets. It causes
24631 the @code{UNICODE} preprocessor macro to be predefined, and
24632 chooses Unicode-capable runtime startup code.
24636 This option is available for Cygwin and MinGW targets. It
24637 specifies that the typical Microsoft Windows predefined macros are to
24638 be set in the pre-processor, but does not influence the choice
24639 of runtime library/startup code.
24643 This option is available for Cygwin and MinGW targets. It
24644 specifies that a GUI application is to be generated by
24645 instructing the linker to set the PE header subsystem type
24648 @item -fno-set-stack-executable
24649 @opindex fno-set-stack-executable
24650 This option is available for MinGW targets. It specifies that
24651 the executable flag for the stack used by nested functions isn't
24652 set. This is necessary for binaries running in kernel mode of
24653 Microsoft Windows, as there the User32 API, which is used to set executable
24654 privileges, isn't available.
24656 @item -fwritable-relocated-rdata
24657 @opindex fno-writable-relocated-rdata
24658 This option is available for MinGW and Cygwin targets. It specifies
24659 that relocated-data in read-only section is put into the @code{.data}
24660 section. This is a necessary for older runtimes not supporting
24661 modification of @code{.rdata} sections for pseudo-relocation.
24663 @item -mpe-aligned-commons
24664 @opindex mpe-aligned-commons
24665 This option is available for Cygwin and MinGW targets. It
24666 specifies that the GNU extension to the PE file format that
24667 permits the correct alignment of COMMON variables should be
24668 used when generating code. It is enabled by default if
24669 GCC detects that the target assembler found during configuration
24670 supports the feature.
24673 See also under @ref{x86 Options} for standard options.
24675 @node Xstormy16 Options
24676 @subsection Xstormy16 Options
24677 @cindex Xstormy16 Options
24679 These options are defined for Xstormy16:
24684 Choose startup files and linker script suitable for the simulator.
24687 @node Xtensa Options
24688 @subsection Xtensa Options
24689 @cindex Xtensa Options
24691 These options are supported for Xtensa targets:
24695 @itemx -mno-const16
24697 @opindex mno-const16
24698 Enable or disable use of @code{CONST16} instructions for loading
24699 constant values. The @code{CONST16} instruction is currently not a
24700 standard option from Tensilica. When enabled, @code{CONST16}
24701 instructions are always used in place of the standard @code{L32R}
24702 instructions. The use of @code{CONST16} is enabled by default only if
24703 the @code{L32R} instruction is not available.
24706 @itemx -mno-fused-madd
24707 @opindex mfused-madd
24708 @opindex mno-fused-madd
24709 Enable or disable use of fused multiply/add and multiply/subtract
24710 instructions in the floating-point option. This has no effect if the
24711 floating-point option is not also enabled. Disabling fused multiply/add
24712 and multiply/subtract instructions forces the compiler to use separate
24713 instructions for the multiply and add/subtract operations. This may be
24714 desirable in some cases where strict IEEE 754-compliant results are
24715 required: the fused multiply add/subtract instructions do not round the
24716 intermediate result, thereby producing results with @emph{more} bits of
24717 precision than specified by the IEEE standard. Disabling fused multiply
24718 add/subtract instructions also ensures that the program output is not
24719 sensitive to the compiler's ability to combine multiply and add/subtract
24722 @item -mserialize-volatile
24723 @itemx -mno-serialize-volatile
24724 @opindex mserialize-volatile
24725 @opindex mno-serialize-volatile
24726 When this option is enabled, GCC inserts @code{MEMW} instructions before
24727 @code{volatile} memory references to guarantee sequential consistency.
24728 The default is @option{-mserialize-volatile}. Use
24729 @option{-mno-serialize-volatile} to omit the @code{MEMW} instructions.
24731 @item -mforce-no-pic
24732 @opindex mforce-no-pic
24733 For targets, like GNU/Linux, where all user-mode Xtensa code must be
24734 position-independent code (PIC), this option disables PIC for compiling
24737 @item -mtext-section-literals
24738 @itemx -mno-text-section-literals
24739 @opindex mtext-section-literals
24740 @opindex mno-text-section-literals
24741 These options control the treatment of literal pools. The default is
24742 @option{-mno-text-section-literals}, which places literals in a separate
24743 section in the output file. This allows the literal pool to be placed
24744 in a data RAM/ROM, and it also allows the linker to combine literal
24745 pools from separate object files to remove redundant literals and
24746 improve code size. With @option{-mtext-section-literals}, the literals
24747 are interspersed in the text section in order to keep them as close as
24748 possible to their references. This may be necessary for large assembly
24749 files. Literals for each function are placed right before that function.
24751 @item -mauto-litpools
24752 @itemx -mno-auto-litpools
24753 @opindex mauto-litpools
24754 @opindex mno-auto-litpools
24755 These options control the treatment of literal pools. The default is
24756 @option{-mno-auto-litpools}, which places literals in a separate
24757 section in the output file unless @option{-mtext-section-literals} is
24758 used. With @option{-mauto-litpools} the literals are interspersed in
24759 the text section by the assembler. Compiler does not produce explicit
24760 @code{.literal} directives and loads literals into registers with
24761 @code{MOVI} instructions instead of @code{L32R} to let the assembler
24762 do relaxation and place literals as necessary. This option allows
24763 assembler to create several literal pools per function and assemble
24764 very big functions, which may not be possible with
24765 @option{-mtext-section-literals}.
24767 @item -mtarget-align
24768 @itemx -mno-target-align
24769 @opindex mtarget-align
24770 @opindex mno-target-align
24771 When this option is enabled, GCC instructs the assembler to
24772 automatically align instructions to reduce branch penalties at the
24773 expense of some code density. The assembler attempts to widen density
24774 instructions to align branch targets and the instructions following call
24775 instructions. If there are not enough preceding safe density
24776 instructions to align a target, no widening is performed. The
24777 default is @option{-mtarget-align}. These options do not affect the
24778 treatment of auto-aligned instructions like @code{LOOP}, which the
24779 assembler always aligns, either by widening density instructions or
24780 by inserting NOP instructions.
24783 @itemx -mno-longcalls
24784 @opindex mlongcalls
24785 @opindex mno-longcalls
24786 When this option is enabled, GCC instructs the assembler to translate
24787 direct calls to indirect calls unless it can determine that the target
24788 of a direct call is in the range allowed by the call instruction. This
24789 translation typically occurs for calls to functions in other source
24790 files. Specifically, the assembler translates a direct @code{CALL}
24791 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
24792 The default is @option{-mno-longcalls}. This option should be used in
24793 programs where the call target can potentially be out of range. This
24794 option is implemented in the assembler, not the compiler, so the
24795 assembly code generated by GCC still shows direct call
24796 instructions---look at the disassembled object code to see the actual
24797 instructions. Note that the assembler uses an indirect call for
24798 every cross-file call, not just those that really are out of range.
24801 @node zSeries Options
24802 @subsection zSeries Options
24803 @cindex zSeries options
24805 These are listed under @xref{S/390 and zSeries Options}.
24811 @section Specifying Subprocesses and the Switches to Pass to Them
24814 @command{gcc} is a driver program. It performs its job by invoking a
24815 sequence of other programs to do the work of compiling, assembling and
24816 linking. GCC interprets its command-line parameters and uses these to
24817 deduce which programs it should invoke, and which command-line options
24818 it ought to place on their command lines. This behavior is controlled
24819 by @dfn{spec strings}. In most cases there is one spec string for each
24820 program that GCC can invoke, but a few programs have multiple spec
24821 strings to control their behavior. The spec strings built into GCC can
24822 be overridden by using the @option{-specs=} command-line switch to specify
24825 @dfn{Spec files} are plain-text files that are used to construct spec
24826 strings. They consist of a sequence of directives separated by blank
24827 lines. The type of directive is determined by the first non-whitespace
24828 character on the line, which can be one of the following:
24831 @item %@var{command}
24832 Issues a @var{command} to the spec file processor. The commands that can
24836 @item %include <@var{file}>
24837 @cindex @code{%include}
24838 Search for @var{file} and insert its text at the current point in the
24841 @item %include_noerr <@var{file}>
24842 @cindex @code{%include_noerr}
24843 Just like @samp{%include}, but do not generate an error message if the include
24844 file cannot be found.
24846 @item %rename @var{old_name} @var{new_name}
24847 @cindex @code{%rename}
24848 Rename the spec string @var{old_name} to @var{new_name}.
24852 @item *[@var{spec_name}]:
24853 This tells the compiler to create, override or delete the named spec
24854 string. All lines after this directive up to the next directive or
24855 blank line are considered to be the text for the spec string. If this
24856 results in an empty string then the spec is deleted. (Or, if the
24857 spec did not exist, then nothing happens.) Otherwise, if the spec
24858 does not currently exist a new spec is created. If the spec does
24859 exist then its contents are overridden by the text of this
24860 directive, unless the first character of that text is the @samp{+}
24861 character, in which case the text is appended to the spec.
24863 @item [@var{suffix}]:
24864 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
24865 and up to the next directive or blank line are considered to make up the
24866 spec string for the indicated suffix. When the compiler encounters an
24867 input file with the named suffix, it processes the spec string in
24868 order to work out how to compile that file. For example:
24872 z-compile -input %i
24875 This says that any input file whose name ends in @samp{.ZZ} should be
24876 passed to the program @samp{z-compile}, which should be invoked with the
24877 command-line switch @option{-input} and with the result of performing the
24878 @samp{%i} substitution. (See below.)
24880 As an alternative to providing a spec string, the text following a
24881 suffix directive can be one of the following:
24884 @item @@@var{language}
24885 This says that the suffix is an alias for a known @var{language}. This is
24886 similar to using the @option{-x} command-line switch to GCC to specify a
24887 language explicitly. For example:
24894 Says that .ZZ files are, in fact, C++ source files.
24897 This causes an error messages saying:
24900 @var{name} compiler not installed on this system.
24904 GCC already has an extensive list of suffixes built into it.
24905 This directive adds an entry to the end of the list of suffixes, but
24906 since the list is searched from the end backwards, it is effectively
24907 possible to override earlier entries using this technique.
24911 GCC has the following spec strings built into it. Spec files can
24912 override these strings or create their own. Note that individual
24913 targets can also add their own spec strings to this list.
24916 asm Options to pass to the assembler
24917 asm_final Options to pass to the assembler post-processor
24918 cpp Options to pass to the C preprocessor
24919 cc1 Options to pass to the C compiler
24920 cc1plus Options to pass to the C++ compiler
24921 endfile Object files to include at the end of the link
24922 link Options to pass to the linker
24923 lib Libraries to include on the command line to the linker
24924 libgcc Decides which GCC support library to pass to the linker
24925 linker Sets the name of the linker
24926 predefines Defines to be passed to the C preprocessor
24927 signed_char Defines to pass to CPP to say whether @code{char} is signed
24929 startfile Object files to include at the start of the link
24932 Here is a small example of a spec file:
24935 %rename lib old_lib
24938 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
24941 This example renames the spec called @samp{lib} to @samp{old_lib} and
24942 then overrides the previous definition of @samp{lib} with a new one.
24943 The new definition adds in some extra command-line options before
24944 including the text of the old definition.
24946 @dfn{Spec strings} are a list of command-line options to be passed to their
24947 corresponding program. In addition, the spec strings can contain
24948 @samp{%}-prefixed sequences to substitute variable text or to
24949 conditionally insert text into the command line. Using these constructs
24950 it is possible to generate quite complex command lines.
24952 Here is a table of all defined @samp{%}-sequences for spec
24953 strings. Note that spaces are not generated automatically around the
24954 results of expanding these sequences. Therefore you can concatenate them
24955 together or combine them with constant text in a single argument.
24959 Substitute one @samp{%} into the program name or argument.
24962 Substitute the name of the input file being processed.
24965 Substitute the basename of the input file being processed.
24966 This is the substring up to (and not including) the last period
24967 and not including the directory.
24970 This is the same as @samp{%b}, but include the file suffix (text after
24974 Marks the argument containing or following the @samp{%d} as a
24975 temporary file name, so that that file is deleted if GCC exits
24976 successfully. Unlike @samp{%g}, this contributes no text to the
24979 @item %g@var{suffix}
24980 Substitute a file name that has suffix @var{suffix} and is chosen
24981 once per compilation, and mark the argument in the same way as
24982 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
24983 name is now chosen in a way that is hard to predict even when previously
24984 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
24985 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
24986 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
24987 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
24988 was simply substituted with a file name chosen once per compilation,
24989 without regard to any appended suffix (which was therefore treated
24990 just like ordinary text), making such attacks more likely to succeed.
24992 @item %u@var{suffix}
24993 Like @samp{%g}, but generates a new temporary file name
24994 each time it appears instead of once per compilation.
24996 @item %U@var{suffix}
24997 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
24998 new one if there is no such last file name. In the absence of any
24999 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
25000 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
25001 involves the generation of two distinct file names, one
25002 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
25003 simply substituted with a file name chosen for the previous @samp{%u},
25004 without regard to any appended suffix.
25006 @item %j@var{suffix}
25007 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
25008 writable, and if @option{-save-temps} is not used;
25009 otherwise, substitute the name
25010 of a temporary file, just like @samp{%u}. This temporary file is not
25011 meant for communication between processes, but rather as a junk
25012 disposal mechanism.
25014 @item %|@var{suffix}
25015 @itemx %m@var{suffix}
25016 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
25017 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
25018 all. These are the two most common ways to instruct a program that it
25019 should read from standard input or write to standard output. If you
25020 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
25021 construct: see for example @file{f/lang-specs.h}.
25023 @item %.@var{SUFFIX}
25024 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
25025 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
25026 terminated by the next space or %.
25029 Marks the argument containing or following the @samp{%w} as the
25030 designated output file of this compilation. This puts the argument
25031 into the sequence of arguments that @samp{%o} substitutes.
25034 Substitutes the names of all the output files, with spaces
25035 automatically placed around them. You should write spaces
25036 around the @samp{%o} as well or the results are undefined.
25037 @samp{%o} is for use in the specs for running the linker.
25038 Input files whose names have no recognized suffix are not compiled
25039 at all, but they are included among the output files, so they are
25043 Substitutes the suffix for object files. Note that this is
25044 handled specially when it immediately follows @samp{%g, %u, or %U},
25045 because of the need for those to form complete file names. The
25046 handling is such that @samp{%O} is treated exactly as if it had already
25047 been substituted, except that @samp{%g, %u, and %U} do not currently
25048 support additional @var{suffix} characters following @samp{%O} as they do
25049 following, for example, @samp{.o}.
25052 Substitutes the standard macro predefinitions for the
25053 current target machine. Use this when running @command{cpp}.
25056 Like @samp{%p}, but puts @samp{__} before and after the name of each
25057 predefined macro, except for macros that start with @samp{__} or with
25058 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
25062 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
25063 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
25064 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
25065 and @option{-imultilib} as necessary.
25068 Current argument is the name of a library or startup file of some sort.
25069 Search for that file in a standard list of directories and substitute
25070 the full name found. The current working directory is included in the
25071 list of directories scanned.
25074 Current argument is the name of a linker script. Search for that file
25075 in the current list of directories to scan for libraries. If the file
25076 is located insert a @option{--script} option into the command line
25077 followed by the full path name found. If the file is not found then
25078 generate an error message. Note: the current working directory is not
25082 Print @var{str} as an error message. @var{str} is terminated by a newline.
25083 Use this when inconsistent options are detected.
25085 @item %(@var{name})
25086 Substitute the contents of spec string @var{name} at this point.
25088 @item %x@{@var{option}@}
25089 Accumulate an option for @samp{%X}.
25092 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
25096 Output the accumulated assembler options specified by @option{-Wa}.
25099 Output the accumulated preprocessor options specified by @option{-Wp}.
25102 Process the @code{asm} spec. This is used to compute the
25103 switches to be passed to the assembler.
25106 Process the @code{asm_final} spec. This is a spec string for
25107 passing switches to an assembler post-processor, if such a program is
25111 Process the @code{link} spec. This is the spec for computing the
25112 command line passed to the linker. Typically it makes use of the
25113 @samp{%L %G %S %D and %E} sequences.
25116 Dump out a @option{-L} option for each directory that GCC believes might
25117 contain startup files. If the target supports multilibs then the
25118 current multilib directory is prepended to each of these paths.
25121 Process the @code{lib} spec. This is a spec string for deciding which
25122 libraries are included on the command line to the linker.
25125 Process the @code{libgcc} spec. This is a spec string for deciding
25126 which GCC support library is included on the command line to the linker.
25129 Process the @code{startfile} spec. This is a spec for deciding which
25130 object files are the first ones passed to the linker. Typically
25131 this might be a file named @file{crt0.o}.
25134 Process the @code{endfile} spec. This is a spec string that specifies
25135 the last object files that are passed to the linker.
25138 Process the @code{cpp} spec. This is used to construct the arguments
25139 to be passed to the C preprocessor.
25142 Process the @code{cc1} spec. This is used to construct the options to be
25143 passed to the actual C compiler (@command{cc1}).
25146 Process the @code{cc1plus} spec. This is used to construct the options to be
25147 passed to the actual C++ compiler (@command{cc1plus}).
25150 Substitute the variable part of a matched option. See below.
25151 Note that each comma in the substituted string is replaced by
25155 Remove all occurrences of @code{-S} from the command line. Note---this
25156 command is position dependent. @samp{%} commands in the spec string
25157 before this one see @code{-S}, @samp{%} commands in the spec string
25158 after this one do not.
25160 @item %:@var{function}(@var{args})
25161 Call the named function @var{function}, passing it @var{args}.
25162 @var{args} is first processed as a nested spec string, then split
25163 into an argument vector in the usual fashion. The function returns
25164 a string which is processed as if it had appeared literally as part
25165 of the current spec.
25167 The following built-in spec functions are provided:
25170 @item @code{getenv}
25171 The @code{getenv} spec function takes two arguments: an environment
25172 variable name and a string. If the environment variable is not
25173 defined, a fatal error is issued. Otherwise, the return value is the
25174 value of the environment variable concatenated with the string. For
25175 example, if @env{TOPDIR} is defined as @file{/path/to/top}, then:
25178 %:getenv(TOPDIR /include)
25181 expands to @file{/path/to/top/include}.
25183 @item @code{if-exists}
25184 The @code{if-exists} spec function takes one argument, an absolute
25185 pathname to a file. If the file exists, @code{if-exists} returns the
25186 pathname. Here is a small example of its usage:
25190 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
25193 @item @code{if-exists-else}
25194 The @code{if-exists-else} spec function is similar to the @code{if-exists}
25195 spec function, except that it takes two arguments. The first argument is
25196 an absolute pathname to a file. If the file exists, @code{if-exists-else}
25197 returns the pathname. If it does not exist, it returns the second argument.
25198 This way, @code{if-exists-else} can be used to select one file or another,
25199 based on the existence of the first. Here is a small example of its usage:
25203 crt0%O%s %:if-exists(crti%O%s) \
25204 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
25207 @item @code{replace-outfile}
25208 The @code{replace-outfile} spec function takes two arguments. It looks for the
25209 first argument in the outfiles array and replaces it with the second argument. Here
25210 is a small example of its usage:
25213 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
25216 @item @code{remove-outfile}
25217 The @code{remove-outfile} spec function takes one argument. It looks for the
25218 first argument in the outfiles array and removes it. Here is a small example
25222 %:remove-outfile(-lm)
25225 @item @code{pass-through-libs}
25226 The @code{pass-through-libs} spec function takes any number of arguments. It
25227 finds any @option{-l} options and any non-options ending in @file{.a} (which it
25228 assumes are the names of linker input library archive files) and returns a
25229 result containing all the found arguments each prepended by
25230 @option{-plugin-opt=-pass-through=} and joined by spaces. This list is
25231 intended to be passed to the LTO linker plugin.
25234 %:pass-through-libs(%G %L %G)
25237 @item @code{print-asm-header}
25238 The @code{print-asm-header} function takes no arguments and simply
25239 prints a banner like:
25245 Use "-Wa,OPTION" to pass "OPTION" to the assembler.
25248 It is used to separate compiler options from assembler options
25249 in the @option{--target-help} output.
25252 @item %@{@code{S}@}
25253 Substitutes the @code{-S} switch, if that switch is given to GCC@.
25254 If that switch is not specified, this substitutes nothing. Note that
25255 the leading dash is omitted when specifying this option, and it is
25256 automatically inserted if the substitution is performed. Thus the spec
25257 string @samp{%@{foo@}} matches the command-line option @option{-foo}
25258 and outputs the command-line option @option{-foo}.
25260 @item %W@{@code{S}@}
25261 Like %@{@code{S}@} but mark last argument supplied within as a file to be
25262 deleted on failure.
25264 @item %@{@code{S}*@}
25265 Substitutes all the switches specified to GCC whose names start
25266 with @code{-S}, but which also take an argument. This is used for
25267 switches like @option{-o}, @option{-D}, @option{-I}, etc.
25268 GCC considers @option{-o foo} as being
25269 one switch whose name starts with @samp{o}. %@{o*@} substitutes this
25270 text, including the space. Thus two arguments are generated.
25272 @item %@{@code{S}*&@code{T}*@}
25273 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
25274 (the order of @code{S} and @code{T} in the spec is not significant).
25275 There can be any number of ampersand-separated variables; for each the
25276 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
25278 @item %@{@code{S}:@code{X}@}
25279 Substitutes @code{X}, if the @option{-S} switch is given to GCC@.
25281 @item %@{!@code{S}:@code{X}@}
25282 Substitutes @code{X}, if the @option{-S} switch is @emph{not} given to GCC@.
25284 @item %@{@code{S}*:@code{X}@}
25285 Substitutes @code{X} if one or more switches whose names start with
25286 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
25287 once, no matter how many such switches appeared. However, if @code{%*}
25288 appears somewhere in @code{X}, then @code{X} is substituted once
25289 for each matching switch, with the @code{%*} replaced by the part of
25290 that switch matching the @code{*}.
25292 If @code{%*} appears as the last part of a spec sequence then a space
25293 is added after the end of the last substitution. If there is more
25294 text in the sequence, however, then a space is not generated. This
25295 allows the @code{%*} substitution to be used as part of a larger
25296 string. For example, a spec string like this:
25299 %@{mcu=*:--script=%*/memory.ld@}
25303 when matching an option like @option{-mcu=newchip} produces:
25306 --script=newchip/memory.ld
25309 @item %@{.@code{S}:@code{X}@}
25310 Substitutes @code{X}, if processing a file with suffix @code{S}.
25312 @item %@{!.@code{S}:@code{X}@}
25313 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
25315 @item %@{,@code{S}:@code{X}@}
25316 Substitutes @code{X}, if processing a file for language @code{S}.
25318 @item %@{!,@code{S}:@code{X}@}
25319 Substitutes @code{X}, if not processing a file for language @code{S}.
25321 @item %@{@code{S}|@code{P}:@code{X}@}
25322 Substitutes @code{X} if either @code{-S} or @code{-P} is given to
25323 GCC@. This may be combined with @samp{!}, @samp{.}, @samp{,}, and
25324 @code{*} sequences as well, although they have a stronger binding than
25325 the @samp{|}. If @code{%*} appears in @code{X}, all of the
25326 alternatives must be starred, and only the first matching alternative
25329 For example, a spec string like this:
25332 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
25336 outputs the following command-line options from the following input
25337 command-line options:
25342 -d fred.c -foo -baz -boggle
25343 -d jim.d -bar -baz -boggle
25346 @item %@{S:X; T:Y; :D@}
25348 If @code{S} is given to GCC, substitutes @code{X}; else if @code{T} is
25349 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
25350 be as many clauses as you need. This may be combined with @code{.},
25351 @code{,}, @code{!}, @code{|}, and @code{*} as needed.
25356 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
25357 construct may contain other nested @samp{%} constructs or spaces, or
25358 even newlines. They are processed as usual, as described above.
25359 Trailing white space in @code{X} is ignored. White space may also
25360 appear anywhere on the left side of the colon in these constructs,
25361 except between @code{.} or @code{*} and the corresponding word.
25363 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
25364 handled specifically in these constructs. If another value of
25365 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
25366 @option{-W} switch is found later in the command line, the earlier
25367 switch value is ignored, except with @{@code{S}*@} where @code{S} is
25368 just one letter, which passes all matching options.
25370 The character @samp{|} at the beginning of the predicate text is used to
25371 indicate that a command should be piped to the following command, but
25372 only if @option{-pipe} is specified.
25374 It is built into GCC which switches take arguments and which do not.
25375 (You might think it would be useful to generalize this to allow each
25376 compiler's spec to say which switches take arguments. But this cannot
25377 be done in a consistent fashion. GCC cannot even decide which input
25378 files have been specified without knowing which switches take arguments,
25379 and it must know which input files to compile in order to tell which
25382 GCC also knows implicitly that arguments starting in @option{-l} are to be
25383 treated as compiler output files, and passed to the linker in their
25384 proper position among the other output files.
25386 @node Environment Variables
25387 @section Environment Variables Affecting GCC
25388 @cindex environment variables
25390 @c man begin ENVIRONMENT
25391 This section describes several environment variables that affect how GCC
25392 operates. Some of them work by specifying directories or prefixes to use
25393 when searching for various kinds of files. Some are used to specify other
25394 aspects of the compilation environment.
25396 Note that you can also specify places to search using options such as
25397 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
25398 take precedence over places specified using environment variables, which
25399 in turn take precedence over those specified by the configuration of GCC@.
25400 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
25401 GNU Compiler Collection (GCC) Internals}.
25406 @c @itemx LC_COLLATE
25408 @c @itemx LC_MONETARY
25409 @c @itemx LC_NUMERIC
25414 @c @findex LC_COLLATE
25415 @findex LC_MESSAGES
25416 @c @findex LC_MONETARY
25417 @c @findex LC_NUMERIC
25421 These environment variables control the way that GCC uses
25422 localization information which allows GCC to work with different
25423 national conventions. GCC inspects the locale categories
25424 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
25425 so. These locale categories can be set to any value supported by your
25426 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
25427 Kingdom encoded in UTF-8.
25429 The @env{LC_CTYPE} environment variable specifies character
25430 classification. GCC uses it to determine the character boundaries in
25431 a string; this is needed for some multibyte encodings that contain quote
25432 and escape characters that are otherwise interpreted as a string
25435 The @env{LC_MESSAGES} environment variable specifies the language to
25436 use in diagnostic messages.
25438 If the @env{LC_ALL} environment variable is set, it overrides the value
25439 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
25440 and @env{LC_MESSAGES} default to the value of the @env{LANG}
25441 environment variable. If none of these variables are set, GCC
25442 defaults to traditional C English behavior.
25446 If @env{TMPDIR} is set, it specifies the directory to use for temporary
25447 files. GCC uses temporary files to hold the output of one stage of
25448 compilation which is to be used as input to the next stage: for example,
25449 the output of the preprocessor, which is the input to the compiler
25452 @item GCC_COMPARE_DEBUG
25453 @findex GCC_COMPARE_DEBUG
25454 Setting @env{GCC_COMPARE_DEBUG} is nearly equivalent to passing
25455 @option{-fcompare-debug} to the compiler driver. See the documentation
25456 of this option for more details.
25458 @item GCC_EXEC_PREFIX
25459 @findex GCC_EXEC_PREFIX
25460 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
25461 names of the subprograms executed by the compiler. No slash is added
25462 when this prefix is combined with the name of a subprogram, but you can
25463 specify a prefix that ends with a slash if you wish.
25465 If @env{GCC_EXEC_PREFIX} is not set, GCC attempts to figure out
25466 an appropriate prefix to use based on the pathname it is invoked with.
25468 If GCC cannot find the subprogram using the specified prefix, it
25469 tries looking in the usual places for the subprogram.
25471 The default value of @env{GCC_EXEC_PREFIX} is
25472 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the prefix to
25473 the installed compiler. In many cases @var{prefix} is the value
25474 of @code{prefix} when you ran the @file{configure} script.
25476 Other prefixes specified with @option{-B} take precedence over this prefix.
25478 This prefix is also used for finding files such as @file{crt0.o} that are
25481 In addition, the prefix is used in an unusual way in finding the
25482 directories to search for header files. For each of the standard
25483 directories whose name normally begins with @samp{/usr/local/lib/gcc}
25484 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
25485 replacing that beginning with the specified prefix to produce an
25486 alternate directory name. Thus, with @option{-Bfoo/}, GCC searches
25487 @file{foo/bar} just before it searches the standard directory
25488 @file{/usr/local/lib/bar}.
25489 If a standard directory begins with the configured
25490 @var{prefix} then the value of @var{prefix} is replaced by
25491 @env{GCC_EXEC_PREFIX} when looking for header files.
25493 @item COMPILER_PATH
25494 @findex COMPILER_PATH
25495 The value of @env{COMPILER_PATH} is a colon-separated list of
25496 directories, much like @env{PATH}. GCC tries the directories thus
25497 specified when searching for subprograms, if it can't find the
25498 subprograms using @env{GCC_EXEC_PREFIX}.
25501 @findex LIBRARY_PATH
25502 The value of @env{LIBRARY_PATH} is a colon-separated list of
25503 directories, much like @env{PATH}. When configured as a native compiler,
25504 GCC tries the directories thus specified when searching for special
25505 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
25506 using GCC also uses these directories when searching for ordinary
25507 libraries for the @option{-l} option (but directories specified with
25508 @option{-L} come first).
25512 @cindex locale definition
25513 This variable is used to pass locale information to the compiler. One way in
25514 which this information is used is to determine the character set to be used
25515 when character literals, string literals and comments are parsed in C and C++.
25516 When the compiler is configured to allow multibyte characters,
25517 the following values for @env{LANG} are recognized:
25521 Recognize JIS characters.
25523 Recognize SJIS characters.
25525 Recognize EUCJP characters.
25528 If @env{LANG} is not defined, or if it has some other value, then the
25529 compiler uses @code{mblen} and @code{mbtowc} as defined by the default locale to
25530 recognize and translate multibyte characters.
25534 Some additional environment variables affect the behavior of the
25537 @include cppenv.texi
25541 @node Precompiled Headers
25542 @section Using Precompiled Headers
25543 @cindex precompiled headers
25544 @cindex speed of compilation
25546 Often large projects have many header files that are included in every
25547 source file. The time the compiler takes to process these header files
25548 over and over again can account for nearly all of the time required to
25549 build the project. To make builds faster, GCC allows you to
25550 @dfn{precompile} a header file.
25552 To create a precompiled header file, simply compile it as you would any
25553 other file, if necessary using the @option{-x} option to make the driver
25554 treat it as a C or C++ header file. You may want to use a
25555 tool like @command{make} to keep the precompiled header up-to-date when
25556 the headers it contains change.
25558 A precompiled header file is searched for when @code{#include} is
25559 seen in the compilation. As it searches for the included file
25560 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
25561 compiler looks for a precompiled header in each directory just before it
25562 looks for the include file in that directory. The name searched for is
25563 the name specified in the @code{#include} with @samp{.gch} appended. If
25564 the precompiled header file can't be used, it is ignored.
25566 For instance, if you have @code{#include "all.h"}, and you have
25567 @file{all.h.gch} in the same directory as @file{all.h}, then the
25568 precompiled header file is used if possible, and the original
25569 header is used otherwise.
25571 Alternatively, you might decide to put the precompiled header file in a
25572 directory and use @option{-I} to ensure that directory is searched
25573 before (or instead of) the directory containing the original header.
25574 Then, if you want to check that the precompiled header file is always
25575 used, you can put a file of the same name as the original header in this
25576 directory containing an @code{#error} command.
25578 This also works with @option{-include}. So yet another way to use
25579 precompiled headers, good for projects not designed with precompiled
25580 header files in mind, is to simply take most of the header files used by
25581 a project, include them from another header file, precompile that header
25582 file, and @option{-include} the precompiled header. If the header files
25583 have guards against multiple inclusion, they are skipped because
25584 they've already been included (in the precompiled header).
25586 If you need to precompile the same header file for different
25587 languages, targets, or compiler options, you can instead make a
25588 @emph{directory} named like @file{all.h.gch}, and put each precompiled
25589 header in the directory, perhaps using @option{-o}. It doesn't matter
25590 what you call the files in the directory; every precompiled header in
25591 the directory is considered. The first precompiled header
25592 encountered in the directory that is valid for this compilation is
25593 used; they're searched in no particular order.
25595 There are many other possibilities, limited only by your imagination,
25596 good sense, and the constraints of your build system.
25598 A precompiled header file can be used only when these conditions apply:
25602 Only one precompiled header can be used in a particular compilation.
25605 A precompiled header can't be used once the first C token is seen. You
25606 can have preprocessor directives before a precompiled header; you cannot
25607 include a precompiled header from inside another header.
25610 The precompiled header file must be produced for the same language as
25611 the current compilation. You can't use a C precompiled header for a C++
25615 The precompiled header file must have been produced by the same compiler
25616 binary as the current compilation is using.
25619 Any macros defined before the precompiled header is included must
25620 either be defined in the same way as when the precompiled header was
25621 generated, or must not affect the precompiled header, which usually
25622 means that they don't appear in the precompiled header at all.
25624 The @option{-D} option is one way to define a macro before a
25625 precompiled header is included; using a @code{#define} can also do it.
25626 There are also some options that define macros implicitly, like
25627 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
25630 @item If debugging information is output when using the precompiled
25631 header, using @option{-g} or similar, the same kind of debugging information
25632 must have been output when building the precompiled header. However,
25633 a precompiled header built using @option{-g} can be used in a compilation
25634 when no debugging information is being output.
25636 @item The same @option{-m} options must generally be used when building
25637 and using the precompiled header. @xref{Submodel Options},
25638 for any cases where this rule is relaxed.
25640 @item Each of the following options must be the same when building and using
25641 the precompiled header:
25643 @gccoptlist{-fexceptions}
25646 Some other command-line options starting with @option{-f},
25647 @option{-p}, or @option{-O} must be defined in the same way as when
25648 the precompiled header was generated. At present, it's not clear
25649 which options are safe to change and which are not; the safest choice
25650 is to use exactly the same options when generating and using the
25651 precompiled header. The following are known to be safe:
25653 @gccoptlist{-fmessage-length= -fpreprocessed -fsched-interblock @gol
25654 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous @gol
25655 -fsched-verbose=@var{number} -fschedule-insns -fvisibility= @gol
25660 For all of these except the last, the compiler automatically
25661 ignores the precompiled header if the conditions aren't met. If you
25662 find an option combination that doesn't work and doesn't cause the
25663 precompiled header to be ignored, please consider filing a bug report,
25666 If you do use differing options when generating and using the
25667 precompiled header, the actual behavior is a mixture of the
25668 behavior for the options. For instance, if you use @option{-g} to
25669 generate the precompiled header but not when using it, you may or may
25670 not get debugging information for routines in the precompiled header.