1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
12 @c man begin COPYRIGHT
13 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
14 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
16 Permission is granted to copy, distribute and/or modify this document
17 under the terms of the GNU Free Documentation License, Version 1.2 or
18 any later version published by the Free Software Foundation; with the
19 Invariant Sections being ``GNU General Public License'' and ``Funding
20 Free Software'', the Front-Cover texts being (a) (see below), and with
21 the Back-Cover Texts being (b) (see below). A copy of the license is
22 included in the gfdl(7) man page.
24 (a) The FSF's Front-Cover Text is:
28 (b) The FSF's Back-Cover Text is:
30 You have freedom to copy and modify this GNU Manual, like GNU
31 software. Copies published by the Free Software Foundation raise
32 funds for GNU development.
34 @c Set file name and title for the man page.
36 @settitle GNU project C and C++ compiler
38 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
39 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
40 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
41 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
42 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
43 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
44 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
46 Only the most useful options are listed here; see below for the
47 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
50 gpl(7), gfdl(7), fsf-funding(7),
51 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
52 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
53 @file{ld}, @file{binutils} and @file{gdb}.
56 For instructions on reporting bugs, see
57 @w{@uref{http://gcc.gnu.org/bugs.html}}.
60 See the Info entry for @command{gcc}, or
61 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
62 for contributors to GCC@.
67 @chapter GCC Command Options
68 @cindex GCC command options
69 @cindex command options
70 @cindex options, GCC command
72 @c man begin DESCRIPTION
73 When you invoke GCC, it normally does preprocessing, compilation,
74 assembly and linking. The ``overall options'' allow you to stop this
75 process at an intermediate stage. For example, the @option{-c} option
76 says not to run the linker. Then the output consists of object files
77 output by the assembler.
79 Other options are passed on to one stage of processing. Some options
80 control the preprocessor and others the compiler itself. Yet other
81 options control the assembler and linker; most of these are not
82 documented here, since you rarely need to use any of them.
84 @cindex C compilation options
85 Most of the command line options that you can use with GCC are useful
86 for C programs; when an option is only useful with another language
87 (usually C++), the explanation says so explicitly. If the description
88 for a particular option does not mention a source language, you can use
89 that option with all supported languages.
91 @cindex C++ compilation options
92 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
93 options for compiling C++ programs.
95 @cindex grouping options
96 @cindex options, grouping
97 The @command{gcc} program accepts options and file names as operands. Many
98 options have multi-letter names; therefore multiple single-letter options
99 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
102 @cindex order of options
103 @cindex options, order
104 You can mix options and other arguments. For the most part, the order
105 you use doesn't matter. Order does matter when you use several options
106 of the same kind; for example, if you specify @option{-L} more than once,
107 the directories are searched in the order specified.
109 Many options have long names starting with @samp{-f} or with
110 @samp{-W}---for example,
111 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
112 these have both positive and negative forms; the negative form of
113 @option{-ffoo} would be @option{-fno-foo}. This manual documents
114 only one of these two forms, whichever one is not the default.
118 @xref{Option Index}, for an index to GCC's options.
121 * Option Summary:: Brief list of all options, without explanations.
122 * Overall Options:: Controlling the kind of output:
123 an executable, object files, assembler files,
124 or preprocessed source.
125 * Invoking G++:: Compiling C++ programs.
126 * C Dialect Options:: Controlling the variant of C language compiled.
127 * C++ Dialect Options:: Variations on C++.
128 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
130 * Language Independent Options:: Controlling how diagnostics should be
132 * Warning Options:: How picky should the compiler be?
133 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
134 * Optimize Options:: How much optimization?
135 * Preprocessor Options:: Controlling header files and macro definitions.
136 Also, getting dependency information for Make.
137 * Assembler Options:: Passing options to the assembler.
138 * Link Options:: Specifying libraries and so on.
139 * Directory Options:: Where to find header files and libraries.
140 Where to find the compiler executable files.
141 * Spec Files:: How to pass switches to sub-processes.
142 * Target Options:: Running a cross-compiler, or an old version of GCC.
143 * Submodel Options:: Specifying minor hardware or convention variations,
144 such as 68010 vs 68020.
145 * Code Gen Options:: Specifying conventions for function calls, data layout
147 * Environment Variables:: Env vars that affect GCC.
148 * Precompiled Headers:: Compiling a header once, and using it many times.
149 * Running Protoize:: Automatically adding or removing function prototypes.
155 @section Option Summary
157 Here is a summary of all the options, grouped by type. Explanations are
158 in the following sections.
161 @item Overall Options
162 @xref{Overall Options,,Options Controlling the Kind of Output}.
163 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
164 -x @var{language} -v -### --help --target-help --version @@@var{file}}
166 @item C Language Options
167 @xref{C Dialect Options,,Options Controlling C Dialect}.
168 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
169 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
170 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
171 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
172 -fallow-single-precision -fcond-mismatch -flax-vector-conversions @gol
173 -fsigned-bitfields -fsigned-char @gol
174 -funsigned-bitfields -funsigned-char}
176 @item C++ Language Options
177 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
178 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
179 -fconserve-space -ffriend-injection @gol
180 -fno-elide-constructors @gol
181 -fno-enforce-eh-specs @gol
182 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
183 -fno-implicit-templates @gol
184 -fno-implicit-inline-templates @gol
185 -fno-implement-inlines -fms-extensions @gol
186 -fno-nonansi-builtins -fno-operator-names @gol
187 -fno-optional-diags -fpermissive @gol
188 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
189 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
190 -fno-default-inline -fvisibility-inlines-hidden @gol
191 -Wabi -Wctor-dtor-privacy @gol
192 -Wnon-virtual-dtor -Wreorder @gol
193 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
194 -Wno-non-template-friend -Wold-style-cast @gol
195 -Woverloaded-virtual -Wno-pmf-conversions @gol
198 @item Objective-C and Objective-C++ Language Options
199 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
200 Objective-C and Objective-C++ Dialects}.
201 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
202 -fgnu-runtime -fnext-runtime @gol
203 -fno-nil-receivers @gol
204 -fobjc-call-cxx-cdtors @gol
205 -fobjc-direct-dispatch @gol
206 -fobjc-exceptions @gol
208 -freplace-objc-classes @gol
211 -Wassign-intercept @gol
212 -Wno-protocol -Wselector @gol
213 -Wstrict-selector-match @gol
214 -Wundeclared-selector}
216 @item Language Independent Options
217 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
218 @gccoptlist{-fmessage-length=@var{n} @gol
219 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
220 -fdiagnostics-show-option}
222 @item Warning Options
223 @xref{Warning Options,,Options to Request or Suppress Warnings}.
224 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
225 -w -Wextra -Wall -Waggregate-return -Walways-true -Wno-attributes @gol
226 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts @gol
227 -Wclobbered -Wcomment @gol
228 -Wconversion -Wno-deprecated-declarations @gol
229 -Wdisabled-optimization -Wno-div-by-zero @gol
230 -Wempty-body -Wno-endif-labels @gol
231 -Werror -Werror-* -Werror-implicit-function-declaration @gol
232 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
233 -Wno-format-extra-args -Wformat-nonliteral @gol
234 -Wformat-security -Wformat-y2k @gol
235 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
236 -Wimport -Wno-import -Winit-self -Winline @gol
237 -Wno-int-to-pointer-cast @gol
238 -Wno-invalid-offsetof -Winvalid-pch @gol
239 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
240 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
241 -Wmissing-format-attribute -Wmissing-include-dirs @gol
242 -Wmissing-noreturn @gol
243 -Wno-multichar -Wnonnull -Wno-overflow @gol
244 -Woverlength-strings -Wpacked -Wpadded @gol
245 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
246 -Wredundant-decls @gol
247 -Wreturn-type -Wsequence-point -Wshadow @gol
248 -Wsign-compare -Wstack-protector @gol
249 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
250 -Wstring-literal-comparison @gol
251 -Wswitch -Wswitch-default -Wswitch-enum @gol
252 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
253 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
254 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
255 -Wunused-value -Wunused-variable -Wvariadic-macros @gol
256 -Wvolatile-register-var -Wwrite-strings}
258 @item C-only Warning Options
259 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
260 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs @gol
261 -Wold-style-declaration -Wold-style-definition @gol
262 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion @gol
263 -Wdeclaration-after-statement -Wpointer-sign}
265 @item Debugging Options
266 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
267 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
268 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
269 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
270 -fdump-ipa-all -fdump-ipa-cgraph @gol
272 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
277 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
282 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
283 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
285 -fdump-tree-nrv -fdump-tree-vect @gol
286 -fdump-tree-sink @gol
287 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
288 -fdump-tree-salias @gol
289 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
290 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
291 -ftree-vectorizer-verbose=@var{n} @gol
292 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
293 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
294 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
295 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report -fprofile-arcs @gol
296 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
297 -ftest-coverage -ftime-report -fvar-tracking @gol
298 -g -g@var{level} -gcoff -gdwarf-2 @gol
299 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
300 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
301 -print-multi-directory -print-multi-lib @gol
302 -print-prog-name=@var{program} -print-search-dirs -Q @gol
305 @item Optimization Options
306 @xref{Optimize Options,,Options that Control Optimization}.
307 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
308 -falign-labels=@var{n} -falign-loops=@var{n} @gol
309 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
310 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
311 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
312 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
313 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
314 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
315 -fexpensive-optimizations -ffast-math -ffloat-store @gol
316 -fforce-addr -fforward-propagate -ffunction-sections @gol
317 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
318 -fcrossjumping -fif-conversion -fif-conversion2 @gol
319 -finline-functions -finline-functions-called-once @gol
320 -finline-limit=@var{n} -fkeep-inline-functions @gol
321 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
322 -fmodulo-sched -fno-branch-count-reg @gol
323 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
324 -fno-function-cse -fno-guess-branch-probability @gol
325 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
326 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
327 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
328 -fomit-frame-pointer -foptimize-register-move @gol
329 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
330 -fprofile-generate -fprofile-use @gol
331 -fregmove -frename-registers @gol
332 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
333 -frerun-cse-after-loop @gol
334 -frounding-math -frtl-abstract-sequences @gol
335 -fschedule-insns -fschedule-insns2 @gol
336 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
337 -fsched-spec-load-dangerous @gol
338 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
339 -fsched2-use-superblocks @gol
340 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
341 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
342 -fstack-protector -fstack-protector-all @gol
343 -fstrict-aliasing -ftracer -fthread-jumps @gol
344 -funroll-all-loops -funroll-loops -fpeel-loops @gol
345 -fsplit-ivs-in-unroller -funswitch-loops @gol
346 -fvariable-expansion-in-unroller @gol
347 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
348 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
349 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
350 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
351 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
352 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
353 --param @var{name}=@var{value}
354 -O -O0 -O1 -O2 -O3 -Os}
356 @item Preprocessor Options
357 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
358 @gccoptlist{-A@var{question}=@var{answer} @gol
359 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
360 -C -dD -dI -dM -dN @gol
361 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
362 -idirafter @var{dir} @gol
363 -include @var{file} -imacros @var{file} @gol
364 -iprefix @var{file} -iwithprefix @var{dir} @gol
365 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
366 -imultilib @var{dir} -isysroot @var{dir} @gol
367 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
368 -P -fworking-directory -remap @gol
369 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
370 -Xpreprocessor @var{option}}
372 @item Assembler Option
373 @xref{Assembler Options,,Passing Options to the Assembler}.
374 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
377 @xref{Link Options,,Options for Linking}.
378 @gccoptlist{@var{object-file-name} -l@var{library} @gol
379 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
380 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
381 -Wl,@var{option} -Xlinker @var{option} @gol
384 @item Directory Options
385 @xref{Directory Options,,Options for Directory Search}.
386 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
387 -specs=@var{file} -I- --sysroot=@var{dir}}
390 @c I wrote this xref this way to avoid overfull hbox. -- rms
391 @xref{Target Options}.
392 @gccoptlist{-V @var{version} -b @var{machine}}
394 @item Machine Dependent Options
395 @xref{Submodel Options,,Hardware Models and Configurations}.
396 @c This list is ordered alphanumerically by subsection name.
397 @c Try and put the significant identifier (CPU or system) first,
398 @c so users have a clue at guessing where the ones they want will be.
401 @gccoptlist{-EB -EL @gol
402 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
403 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
406 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
407 -mabi=@var{name} @gol
408 -mapcs-stack-check -mno-apcs-stack-check @gol
409 -mapcs-float -mno-apcs-float @gol
410 -mapcs-reentrant -mno-apcs-reentrant @gol
411 -msched-prolog -mno-sched-prolog @gol
412 -mlittle-endian -mbig-endian -mwords-little-endian @gol
413 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
414 -mthumb-interwork -mno-thumb-interwork @gol
415 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
416 -mstructure-size-boundary=@var{n} @gol
417 -mabort-on-noreturn @gol
418 -mlong-calls -mno-long-calls @gol
419 -msingle-pic-base -mno-single-pic-base @gol
420 -mpic-register=@var{reg} @gol
421 -mnop-fun-dllimport @gol
422 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
423 -mpoke-function-name @gol
425 -mtpcs-frame -mtpcs-leaf-frame @gol
426 -mcaller-super-interworking -mcallee-super-interworking @gol
430 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
431 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
433 @emph{Blackfin Options}
434 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
435 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
436 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
437 -mno-id-shared-library -mshared-library-id=@var{n} @gol
438 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
439 -msep-data -mno-sep-data -mlong-calls -mno-long-calls}
442 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
443 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
444 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
445 -mstack-align -mdata-align -mconst-align @gol
446 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
447 -melf -maout -melinux -mlinux -sim -sim2 @gol
448 -mmul-bug-workaround -mno-mul-bug-workaround}
451 @gccoptlist{-mmac -mpush-args}
453 @emph{Darwin Options}
454 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
455 -arch_only -bind_at_load -bundle -bundle_loader @gol
456 -client_name -compatibility_version -current_version @gol
458 -dependency-file -dylib_file -dylinker_install_name @gol
459 -dynamic -dynamiclib -exported_symbols_list @gol
460 -filelist -flat_namespace -force_cpusubtype_ALL @gol
461 -force_flat_namespace -headerpad_max_install_names @gol
462 -image_base -init -install_name -keep_private_externs @gol
463 -multi_module -multiply_defined -multiply_defined_unused @gol
464 -noall_load -no_dead_strip_inits_and_terms @gol
465 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
466 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
467 -private_bundle -read_only_relocs -sectalign @gol
468 -sectobjectsymbols -whyload -seg1addr @gol
469 -sectcreate -sectobjectsymbols -sectorder @gol
470 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
471 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
472 -segprot -segs_read_only_addr -segs_read_write_addr @gol
473 -single_module -static -sub_library -sub_umbrella @gol
474 -twolevel_namespace -umbrella -undefined @gol
475 -unexported_symbols_list -weak_reference_mismatches @gol
476 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
477 -mkernel -mone-byte-bool}
479 @emph{DEC Alpha Options}
480 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
481 -mieee -mieee-with-inexact -mieee-conformant @gol
482 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
483 -mtrap-precision=@var{mode} -mbuild-constants @gol
484 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
485 -mbwx -mmax -mfix -mcix @gol
486 -mfloat-vax -mfloat-ieee @gol
487 -mexplicit-relocs -msmall-data -mlarge-data @gol
488 -msmall-text -mlarge-text @gol
489 -mmemory-latency=@var{time}}
491 @emph{DEC Alpha/VMS Options}
492 @gccoptlist{-mvms-return-codes}
495 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
496 -mhard-float -msoft-float @gol
497 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
498 -mdouble -mno-double @gol
499 -mmedia -mno-media -mmuladd -mno-muladd @gol
500 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
501 -mlinked-fp -mlong-calls -malign-labels @gol
502 -mlibrary-pic -macc-4 -macc-8 @gol
503 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
504 -moptimize-membar -mno-optimize-membar @gol
505 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
506 -mvliw-branch -mno-vliw-branch @gol
507 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
508 -mno-nested-cond-exec -mtomcat-stats @gol
512 @emph{GNU/Linux Options}
513 @gccoptlist{-muclibc}
515 @emph{H8/300 Options}
516 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
519 @gccoptlist{-march=@var{architecture-type} @gol
520 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
521 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
522 -mfixed-range=@var{register-range} @gol
523 -mjump-in-delay -mlinker-opt -mlong-calls @gol
524 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
525 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
526 -mno-jump-in-delay -mno-long-load-store @gol
527 -mno-portable-runtime -mno-soft-float @gol
528 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
529 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
530 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
531 -munix=@var{unix-std} -nolibdld -static -threads}
533 @emph{i386 and x86-64 Options}
534 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
535 -mfpmath=@var{unit} @gol
536 -masm=@var{dialect} -mno-fancy-math-387 @gol
537 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
538 -mno-wide-multiply -mrtd -malign-double @gol
539 -mpreferred-stack-boundary=@var{num} @gol
540 -mmmx -msse -msse2 -msse3 -mssse3 -m3dnow @gol
541 -mthreads -mno-align-stringops -minline-all-stringops @gol
542 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
543 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
545 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
546 -mcmodel=@var{code-model} @gol
547 -m32 -m64 -mlarge-data-threshold=@var{num}}
550 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
551 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
552 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
553 -minline-float-divide-max-throughput @gol
554 -minline-int-divide-min-latency @gol
555 -minline-int-divide-max-throughput @gol
556 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
557 -mno-dwarf2-asm -mearly-stop-bits @gol
558 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
559 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
560 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
561 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
562 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
563 -mno-sched-prefer-non-data-spec-insns @gol
564 -mno-sched-prefer-non-control-spec-insns @gol
565 -mno-sched-count-spec-in-critical-path}
567 @emph{M32R/D Options}
568 @gccoptlist{-m32r2 -m32rx -m32r @gol
570 -malign-loops -mno-align-loops @gol
571 -missue-rate=@var{number} @gol
572 -mbranch-cost=@var{number} @gol
573 -mmodel=@var{code-size-model-type} @gol
574 -msdata=@var{sdata-type} @gol
575 -mno-flush-func -mflush-func=@var{name} @gol
576 -mno-flush-trap -mflush-trap=@var{number} @gol
580 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
582 @emph{M680x0 Options}
583 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
584 -m68060 -mcpu32 -m5200 -mcfv4e -m68881 -mbitfield @gol
585 -mc68000 -mc68020 @gol
586 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
587 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
588 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
590 @emph{M68hc1x Options}
591 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
592 -mauto-incdec -minmax -mlong-calls -mshort @gol
593 -msoft-reg-count=@var{count}}
596 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
597 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
598 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
599 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
600 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
603 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
604 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
605 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
606 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
607 -mfp32 -mfp64 -mhard-float -msoft-float @gol
608 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
609 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
610 -G@var{num} -membedded-data -mno-embedded-data @gol
611 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
612 -msplit-addresses -mno-split-addresses @gol
613 -mexplicit-relocs -mno-explicit-relocs @gol
614 -mcheck-zero-division -mno-check-zero-division @gol
615 -mdivide-traps -mdivide-breaks @gol
616 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
617 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
618 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
619 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
620 -mfix-sb1 -mno-fix-sb1 @gol
621 -mflush-func=@var{func} -mno-flush-func @gol
622 -mbranch-likely -mno-branch-likely @gol
623 -mfp-exceptions -mno-fp-exceptions @gol
624 -mvr4130-align -mno-vr4130-align}
627 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
628 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
629 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
630 -mno-base-addresses -msingle-exit -mno-single-exit}
632 @emph{MN10300 Options}
633 @gccoptlist{-mmult-bug -mno-mult-bug @gol
634 -mam33 -mno-am33 @gol
635 -mam33-2 -mno-am33-2 @gol
636 -mreturn-pointer-on-d0 @gol
640 @gccoptlist{-mno-crt0 -mbacc -msim @gol
641 -march=@var{cpu-type} }
643 @emph{PDP-11 Options}
644 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
645 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
646 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
647 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
648 -mbranch-expensive -mbranch-cheap @gol
649 -msplit -mno-split -munix-asm -mdec-asm}
651 @emph{PowerPC Options}
652 See RS/6000 and PowerPC Options.
654 @emph{RS/6000 and PowerPC Options}
655 @gccoptlist{-mcpu=@var{cpu-type} @gol
656 -mtune=@var{cpu-type} @gol
657 -mpower -mno-power -mpower2 -mno-power2 @gol
658 -mpowerpc -mpowerpc64 -mno-powerpc @gol
659 -maltivec -mno-altivec @gol
660 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
661 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
662 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
663 -mmfpgpr -mno-mfpgpr @gol
664 -mnew-mnemonics -mold-mnemonics @gol
665 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
666 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
667 -malign-power -malign-natural @gol
668 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
669 -mstring -mno-string -mupdate -mno-update @gol
670 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
671 -mstrict-align -mno-strict-align -mrelocatable @gol
672 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
673 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
674 -mdynamic-no-pic -maltivec -mswdiv @gol
675 -mprioritize-restricted-insns=@var{priority} @gol
676 -msched-costly-dep=@var{dependence_type} @gol
677 -minsert-sched-nops=@var{scheme} @gol
678 -mcall-sysv -mcall-netbsd @gol
679 -maix-struct-return -msvr4-struct-return @gol
680 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
681 -misel -mno-isel @gol
682 -misel=yes -misel=no @gol
684 -mspe=yes -mspe=no @gol
685 -mvrsave -mno-vrsave @gol
686 -mmulhw -mno-mulhw @gol
687 -mdlmzb -mno-dlmzb @gol
688 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
689 -mprototype -mno-prototype @gol
690 -msim -mmvme -mads -myellowknife -memb -msdata @gol
691 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
693 @emph{S/390 and zSeries Options}
694 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
695 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
696 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
697 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
698 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
699 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
700 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
703 @gccoptlist{-mel -mel @gol
708 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
709 -m4-nofpu -m4-single-only -m4-single -m4 @gol
710 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
711 -m5-64media -m5-64media-nofpu @gol
712 -m5-32media -m5-32media-nofpu @gol
713 -m5-compact -m5-compact-nofpu @gol
714 -mb -ml -mdalign -mrelax @gol
715 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
716 -mieee -misize -minline-ic_invalidate -mpadstruct -mspace @gol
717 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
718 -mdivsi3_libfunc=@var{name} @gol
719 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
723 @gccoptlist{-mcpu=@var{cpu-type} @gol
724 -mtune=@var{cpu-type} @gol
725 -mcmodel=@var{code-model} @gol
726 -m32 -m64 -mapp-regs -mno-app-regs @gol
727 -mfaster-structs -mno-faster-structs @gol
728 -mfpu -mno-fpu -mhard-float -msoft-float @gol
729 -mhard-quad-float -msoft-quad-float @gol
730 -mimpure-text -mno-impure-text -mlittle-endian @gol
731 -mstack-bias -mno-stack-bias @gol
732 -munaligned-doubles -mno-unaligned-doubles @gol
733 -mv8plus -mno-v8plus -mvis -mno-vis
734 -threads -pthreads -pthread}
737 @gccoptlist{-mwarn-reloc -merror-reloc @gol
738 -msafe-dma -munsafe-dma @gol
740 -msmall-mem -mlarge-mem -mstdmain @gol
741 -mfixed-range=@var{register-range}}
743 @emph{System V Options}
744 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
746 @emph{TMS320C3x/C4x Options}
747 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
748 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
749 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
750 -mparallel-insns -mparallel-mpy -mpreserve-float}
753 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
754 -mprolog-function -mno-prolog-function -mspace @gol
755 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
756 -mapp-regs -mno-app-regs @gol
757 -mdisable-callt -mno-disable-callt @gol
763 @gccoptlist{-mg -mgnu -munix}
765 @emph{x86-64 Options}
766 See i386 and x86-64 Options.
768 @emph{Xstormy16 Options}
771 @emph{Xtensa Options}
772 @gccoptlist{-mconst16 -mno-const16 @gol
773 -mfused-madd -mno-fused-madd @gol
774 -mtext-section-literals -mno-text-section-literals @gol
775 -mtarget-align -mno-target-align @gol
776 -mlongcalls -mno-longcalls}
778 @emph{zSeries Options}
779 See S/390 and zSeries Options.
781 @item Code Generation Options
782 @xref{Code Gen Options,,Options for Code Generation Conventions}.
783 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
784 -ffixed-@var{reg} -fexceptions @gol
785 -fnon-call-exceptions -funwind-tables @gol
786 -fasynchronous-unwind-tables @gol
787 -finhibit-size-directive -finstrument-functions @gol
788 -fno-common -fno-ident @gol
789 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
790 -fno-jump-tables @gol
791 -frecord-gcc-switches @gol
792 -freg-struct-return -fshort-enums @gol
793 -fshort-double -fshort-wchar @gol
794 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
795 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
796 -fargument-alias -fargument-noalias @gol
797 -fargument-noalias-global -fargument-noalias-anything
798 -fleading-underscore -ftls-model=@var{model} @gol
799 -ftrapv -fwrapv -fbounds-check @gol
804 * Overall Options:: Controlling the kind of output:
805 an executable, object files, assembler files,
806 or preprocessed source.
807 * C Dialect Options:: Controlling the variant of C language compiled.
808 * C++ Dialect Options:: Variations on C++.
809 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
811 * Language Independent Options:: Controlling how diagnostics should be
813 * Warning Options:: How picky should the compiler be?
814 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
815 * Optimize Options:: How much optimization?
816 * Preprocessor Options:: Controlling header files and macro definitions.
817 Also, getting dependency information for Make.
818 * Assembler Options:: Passing options to the assembler.
819 * Link Options:: Specifying libraries and so on.
820 * Directory Options:: Where to find header files and libraries.
821 Where to find the compiler executable files.
822 * Spec Files:: How to pass switches to sub-processes.
823 * Target Options:: Running a cross-compiler, or an old version of GCC.
826 @node Overall Options
827 @section Options Controlling the Kind of Output
829 Compilation can involve up to four stages: preprocessing, compilation
830 proper, assembly and linking, always in that order. GCC is capable of
831 preprocessing and compiling several files either into several
832 assembler input files, or into one assembler input file; then each
833 assembler input file produces an object file, and linking combines all
834 the object files (those newly compiled, and those specified as input)
835 into an executable file.
837 @cindex file name suffix
838 For any given input file, the file name suffix determines what kind of
843 C source code which must be preprocessed.
846 C source code which should not be preprocessed.
849 C++ source code which should not be preprocessed.
852 Objective-C source code. Note that you must link with the @file{libobjc}
853 library to make an Objective-C program work.
856 Objective-C source code which should not be preprocessed.
860 Objective-C++ source code. Note that you must link with the @file{libobjc}
861 library to make an Objective-C++ program work. Note that @samp{.M} refers
862 to a literal capital M@.
865 Objective-C++ source code which should not be preprocessed.
868 C, C++, Objective-C or Objective-C++ header file to be turned into a
873 @itemx @var{file}.cxx
874 @itemx @var{file}.cpp
875 @itemx @var{file}.CPP
876 @itemx @var{file}.c++
878 C++ source code which must be preprocessed. Note that in @samp{.cxx},
879 the last two letters must both be literally @samp{x}. Likewise,
880 @samp{.C} refers to a literal capital C@.
884 Objective-C++ source code which must be preprocessed.
887 Objective-C++ source code which should not be preprocessed.
891 C++ header file to be turned into a precompiled header.
894 @itemx @var{file}.for
895 @itemx @var{file}.FOR
896 Fixed form Fortran source code which should not be preprocessed.
899 @itemx @var{file}.fpp
900 @itemx @var{file}.FPP
901 Fixed form Fortran source code which must be preprocessed (with the traditional
905 @itemx @var{file}.f95
906 Free form Fortran source code which should not be preprocessed.
909 @itemx @var{file}.F95
910 Free form Fortran source code which must be preprocessed (with the
911 traditional preprocessor).
913 @c FIXME: Descriptions of Java file types.
920 Ada source code file which contains a library unit declaration (a
921 declaration of a package, subprogram, or generic, or a generic
922 instantiation), or a library unit renaming declaration (a package,
923 generic, or subprogram renaming declaration). Such files are also
926 @itemx @var{file}.adb
927 Ada source code file containing a library unit body (a subprogram or
928 package body). Such files are also called @dfn{bodies}.
930 @c GCC also knows about some suffixes for languages not yet included:
941 Assembler code which must be preprocessed.
944 An object file to be fed straight into linking.
945 Any file name with no recognized suffix is treated this way.
949 You can specify the input language explicitly with the @option{-x} option:
952 @item -x @var{language}
953 Specify explicitly the @var{language} for the following input files
954 (rather than letting the compiler choose a default based on the file
955 name suffix). This option applies to all following input files until
956 the next @option{-x} option. Possible values for @var{language} are:
958 c c-header c-cpp-output
959 c++ c++-header c++-cpp-output
960 objective-c objective-c-header objective-c-cpp-output
961 objective-c++ objective-c++-header objective-c++-cpp-output
962 assembler assembler-with-cpp
970 Turn off any specification of a language, so that subsequent files are
971 handled according to their file name suffixes (as they are if @option{-x}
972 has not been used at all).
974 @item -pass-exit-codes
975 @opindex pass-exit-codes
976 Normally the @command{gcc} program will exit with the code of 1 if any
977 phase of the compiler returns a non-success return code. If you specify
978 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
979 numerically highest error produced by any phase that returned an error
980 indication. The C, C++, and Fortran frontends return 4, if an internal
981 compiler error is encountered.
984 If you only want some of the stages of compilation, you can use
985 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
986 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
987 @command{gcc} is to stop. Note that some combinations (for example,
988 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
993 Compile or assemble the source files, but do not link. The linking
994 stage simply is not done. The ultimate output is in the form of an
995 object file for each source file.
997 By default, the object file name for a source file is made by replacing
998 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
1000 Unrecognized input files, not requiring compilation or assembly, are
1005 Stop after the stage of compilation proper; do not assemble. The output
1006 is in the form of an assembler code file for each non-assembler input
1009 By default, the assembler file name for a source file is made by
1010 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1012 Input files that don't require compilation are ignored.
1016 Stop after the preprocessing stage; do not run the compiler proper. The
1017 output is in the form of preprocessed source code, which is sent to the
1020 Input files which don't require preprocessing are ignored.
1022 @cindex output file option
1025 Place output in file @var{file}. This applies regardless to whatever
1026 sort of output is being produced, whether it be an executable file,
1027 an object file, an assembler file or preprocessed C code.
1029 If @option{-o} is not specified, the default is to put an executable
1030 file in @file{a.out}, the object file for
1031 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1032 assembler file in @file{@var{source}.s}, a precompiled header file in
1033 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1038 Print (on standard error output) the commands executed to run the stages
1039 of compilation. Also print the version number of the compiler driver
1040 program and of the preprocessor and the compiler proper.
1044 Like @option{-v} except the commands are not executed and all command
1045 arguments are quoted. This is useful for shell scripts to capture the
1046 driver-generated command lines.
1050 Use pipes rather than temporary files for communication between the
1051 various stages of compilation. This fails to work on some systems where
1052 the assembler is unable to read from a pipe; but the GNU assembler has
1057 If you are compiling multiple source files, this option tells the driver
1058 to pass all the source files to the compiler at once (for those
1059 languages for which the compiler can handle this). This will allow
1060 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1061 language for which this is supported is C@. If you pass source files for
1062 multiple languages to the driver, using this option, the driver will invoke
1063 the compiler(s) that support IMA once each, passing each compiler all the
1064 source files appropriate for it. For those languages that do not support
1065 IMA this option will be ignored, and the compiler will be invoked once for
1066 each source file in that language. If you use this option in conjunction
1067 with @option{-save-temps}, the compiler will generate multiple
1069 (one for each source file), but only one (combined) @file{.o} or
1074 Print (on the standard output) a description of the command line options
1075 understood by @command{gcc}. If the @option{-v} option is also specified
1076 then @option{--help} will also be passed on to the various processes
1077 invoked by @command{gcc}, so that they can display the command line options
1078 they accept. If the @option{-Wextra} option is also specified then command
1079 line options which have no documentation associated with them will also
1083 @opindex target-help
1084 Print (on the standard output) a description of target specific command
1085 line options for each tool.
1089 Display the version number and copyrights of the invoked GCC@.
1091 @include @value{srcdir}/../libiberty/at-file.texi
1095 @section Compiling C++ Programs
1097 @cindex suffixes for C++ source
1098 @cindex C++ source file suffixes
1099 C++ source files conventionally use one of the suffixes @samp{.C},
1100 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1101 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1102 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1103 files with these names and compiles them as C++ programs even if you
1104 call the compiler the same way as for compiling C programs (usually
1105 with the name @command{gcc}).
1109 However, the use of @command{gcc} does not add the C++ library.
1110 @command{g++} is a program that calls GCC and treats @samp{.c},
1111 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1112 files unless @option{-x} is used, and automatically specifies linking
1113 against the C++ library. This program is also useful when
1114 precompiling a C header file with a @samp{.h} extension for use in C++
1115 compilations. On many systems, @command{g++} is also installed with
1116 the name @command{c++}.
1118 @cindex invoking @command{g++}
1119 When you compile C++ programs, you may specify many of the same
1120 command-line options that you use for compiling programs in any
1121 language; or command-line options meaningful for C and related
1122 languages; or options that are meaningful only for C++ programs.
1123 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1124 explanations of options for languages related to C@.
1125 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1126 explanations of options that are meaningful only for C++ programs.
1128 @node C Dialect Options
1129 @section Options Controlling C Dialect
1130 @cindex dialect options
1131 @cindex language dialect options
1132 @cindex options, dialect
1134 The following options control the dialect of C (or languages derived
1135 from C, such as C++, Objective-C and Objective-C++) that the compiler
1139 @cindex ANSI support
1143 In C mode, support all ISO C90 programs. In C++ mode,
1144 remove GNU extensions that conflict with ISO C++.
1146 This turns off certain features of GCC that are incompatible with ISO
1147 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1148 such as the @code{asm} and @code{typeof} keywords, and
1149 predefined macros such as @code{unix} and @code{vax} that identify the
1150 type of system you are using. It also enables the undesirable and
1151 rarely used ISO trigraph feature. For the C compiler,
1152 it disables recognition of C++ style @samp{//} comments as well as
1153 the @code{inline} keyword.
1155 The alternate keywords @code{__asm__}, @code{__extension__},
1156 @code{__inline__} and @code{__typeof__} continue to work despite
1157 @option{-ansi}. You would not want to use them in an ISO C program, of
1158 course, but it is useful to put them in header files that might be included
1159 in compilations done with @option{-ansi}. Alternate predefined macros
1160 such as @code{__unix__} and @code{__vax__} are also available, with or
1161 without @option{-ansi}.
1163 The @option{-ansi} option does not cause non-ISO programs to be
1164 rejected gratuitously. For that, @option{-pedantic} is required in
1165 addition to @option{-ansi}. @xref{Warning Options}.
1167 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1168 option is used. Some header files may notice this macro and refrain
1169 from declaring certain functions or defining certain macros that the
1170 ISO standard doesn't call for; this is to avoid interfering with any
1171 programs that might use these names for other things.
1173 Functions which would normally be built in but do not have semantics
1174 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1175 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1176 built-in functions provided by GCC}, for details of the functions
1181 Determine the language standard. This option is currently only
1182 supported when compiling C or C++. A value for this option must be
1183 provided; possible values are
1188 ISO C90 (same as @option{-ansi}).
1190 @item iso9899:199409
1191 ISO C90 as modified in amendment 1.
1197 ISO C99. Note that this standard is not yet fully supported; see
1198 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1199 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1202 Default, ISO C90 plus GNU extensions (including some C99 features).
1206 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1207 this will become the default. The name @samp{gnu9x} is deprecated.
1210 The 1998 ISO C++ standard plus amendments.
1213 The same as @option{-std=c++98} plus GNU extensions. This is the
1214 default for C++ code.
1217 The working draft of the upcoming ISO C++0x standard. This option
1218 enables experimental features that are likely to be included in
1219 C++0x. The working draft is constantly changing, and any feature that is
1220 enabled by this flag may be removed from future versions of GCC if it is
1221 not part of the C++0x standard.
1224 The same as @option{-std=c++0x} plus GNU extensions. As with
1225 @option{-std=c++0x}, this option enables experimental features that may
1226 be removed in future versions of GCC.
1229 Even when this option is not specified, you can still use some of the
1230 features of newer standards in so far as they do not conflict with
1231 previous C standards. For example, you may use @code{__restrict__} even
1232 when @option{-std=c99} is not specified.
1234 The @option{-std} options specifying some version of ISO C have the same
1235 effects as @option{-ansi}, except that features that were not in ISO C90
1236 but are in the specified version (for example, @samp{//} comments and
1237 the @code{inline} keyword in ISO C99) are not disabled.
1239 @xref{Standards,,Language Standards Supported by GCC}, for details of
1240 these standard versions.
1242 @item -aux-info @var{filename}
1244 Output to the given filename prototyped declarations for all functions
1245 declared and/or defined in a translation unit, including those in header
1246 files. This option is silently ignored in any language other than C@.
1248 Besides declarations, the file indicates, in comments, the origin of
1249 each declaration (source file and line), whether the declaration was
1250 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1251 @samp{O} for old, respectively, in the first character after the line
1252 number and the colon), and whether it came from a declaration or a
1253 definition (@samp{C} or @samp{F}, respectively, in the following
1254 character). In the case of function definitions, a K&R-style list of
1255 arguments followed by their declarations is also provided, inside
1256 comments, after the declaration.
1260 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1261 keyword, so that code can use these words as identifiers. You can use
1262 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1263 instead. @option{-ansi} implies @option{-fno-asm}.
1265 In C++, this switch only affects the @code{typeof} keyword, since
1266 @code{asm} and @code{inline} are standard keywords. You may want to
1267 use the @option{-fno-gnu-keywords} flag instead, which has the same
1268 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1269 switch only affects the @code{asm} and @code{typeof} keywords, since
1270 @code{inline} is a standard keyword in ISO C99.
1273 @itemx -fno-builtin-@var{function}
1274 @opindex fno-builtin
1275 @cindex built-in functions
1276 Don't recognize built-in functions that do not begin with
1277 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1278 functions provided by GCC}, for details of the functions affected,
1279 including those which are not built-in functions when @option{-ansi} or
1280 @option{-std} options for strict ISO C conformance are used because they
1281 do not have an ISO standard meaning.
1283 GCC normally generates special code to handle certain built-in functions
1284 more efficiently; for instance, calls to @code{alloca} may become single
1285 instructions that adjust the stack directly, and calls to @code{memcpy}
1286 may become inline copy loops. The resulting code is often both smaller
1287 and faster, but since the function calls no longer appear as such, you
1288 cannot set a breakpoint on those calls, nor can you change the behavior
1289 of the functions by linking with a different library. In addition,
1290 when a function is recognized as a built-in function, GCC may use
1291 information about that function to warn about problems with calls to
1292 that function, or to generate more efficient code, even if the
1293 resulting code still contains calls to that function. For example,
1294 warnings are given with @option{-Wformat} for bad calls to
1295 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1296 known not to modify global memory.
1298 With the @option{-fno-builtin-@var{function}} option
1299 only the built-in function @var{function} is
1300 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1301 function is named this is not built-in in this version of GCC, this
1302 option is ignored. There is no corresponding
1303 @option{-fbuiltin-@var{function}} option; if you wish to enable
1304 built-in functions selectively when using @option{-fno-builtin} or
1305 @option{-ffreestanding}, you may define macros such as:
1308 #define abs(n) __builtin_abs ((n))
1309 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1314 @cindex hosted environment
1316 Assert that compilation takes place in a hosted environment. This implies
1317 @option{-fbuiltin}. A hosted environment is one in which the
1318 entire standard library is available, and in which @code{main} has a return
1319 type of @code{int}. Examples are nearly everything except a kernel.
1320 This is equivalent to @option{-fno-freestanding}.
1322 @item -ffreestanding
1323 @opindex ffreestanding
1324 @cindex hosted environment
1326 Assert that compilation takes place in a freestanding environment. This
1327 implies @option{-fno-builtin}. A freestanding environment
1328 is one in which the standard library may not exist, and program startup may
1329 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1330 This is equivalent to @option{-fno-hosted}.
1332 @xref{Standards,,Language Standards Supported by GCC}, for details of
1333 freestanding and hosted environments.
1337 @cindex openmp parallel
1338 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1339 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1340 compiler generates parallel code according to the OpenMP Application
1341 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1343 @item -fms-extensions
1344 @opindex fms-extensions
1345 Accept some non-standard constructs used in Microsoft header files.
1347 Some cases of unnamed fields in structures and unions are only
1348 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1349 fields within structs/unions}, for details.
1353 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1354 options for strict ISO C conformance) implies @option{-trigraphs}.
1356 @item -no-integrated-cpp
1357 @opindex no-integrated-cpp
1358 Performs a compilation in two passes: preprocessing and compiling. This
1359 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1360 @option{-B} option. The user supplied compilation step can then add in
1361 an additional preprocessing step after normal preprocessing but before
1362 compiling. The default is to use the integrated cpp (internal cpp)
1364 The semantics of this option will change if "cc1", "cc1plus", and
1365 "cc1obj" are merged.
1367 @cindex traditional C language
1368 @cindex C language, traditional
1370 @itemx -traditional-cpp
1371 @opindex traditional-cpp
1372 @opindex traditional
1373 Formerly, these options caused GCC to attempt to emulate a pre-standard
1374 C compiler. They are now only supported with the @option{-E} switch.
1375 The preprocessor continues to support a pre-standard mode. See the GNU
1376 CPP manual for details.
1378 @item -fcond-mismatch
1379 @opindex fcond-mismatch
1380 Allow conditional expressions with mismatched types in the second and
1381 third arguments. The value of such an expression is void. This option
1382 is not supported for C++.
1384 @item -flax-vector-conversions
1385 @opindex flax-vector-conversions
1386 Allow implicit conversions between vectors with differing numbers of
1387 elements and/or incompatible element types. This option should not be
1390 @item -funsigned-char
1391 @opindex funsigned-char
1392 Let the type @code{char} be unsigned, like @code{unsigned char}.
1394 Each kind of machine has a default for what @code{char} should
1395 be. It is either like @code{unsigned char} by default or like
1396 @code{signed char} by default.
1398 Ideally, a portable program should always use @code{signed char} or
1399 @code{unsigned char} when it depends on the signedness of an object.
1400 But many programs have been written to use plain @code{char} and
1401 expect it to be signed, or expect it to be unsigned, depending on the
1402 machines they were written for. This option, and its inverse, let you
1403 make such a program work with the opposite default.
1405 The type @code{char} is always a distinct type from each of
1406 @code{signed char} or @code{unsigned char}, even though its behavior
1407 is always just like one of those two.
1410 @opindex fsigned-char
1411 Let the type @code{char} be signed, like @code{signed char}.
1413 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1414 the negative form of @option{-funsigned-char}. Likewise, the option
1415 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1417 @item -fsigned-bitfields
1418 @itemx -funsigned-bitfields
1419 @itemx -fno-signed-bitfields
1420 @itemx -fno-unsigned-bitfields
1421 @opindex fsigned-bitfields
1422 @opindex funsigned-bitfields
1423 @opindex fno-signed-bitfields
1424 @opindex fno-unsigned-bitfields
1425 These options control whether a bit-field is signed or unsigned, when the
1426 declaration does not use either @code{signed} or @code{unsigned}. By
1427 default, such a bit-field is signed, because this is consistent: the
1428 basic integer types such as @code{int} are signed types.
1431 @node C++ Dialect Options
1432 @section Options Controlling C++ Dialect
1434 @cindex compiler options, C++
1435 @cindex C++ options, command line
1436 @cindex options, C++
1437 This section describes the command-line options that are only meaningful
1438 for C++ programs; but you can also use most of the GNU compiler options
1439 regardless of what language your program is in. For example, you
1440 might compile a file @code{firstClass.C} like this:
1443 g++ -g -frepo -O -c firstClass.C
1447 In this example, only @option{-frepo} is an option meant
1448 only for C++ programs; you can use the other options with any
1449 language supported by GCC@.
1451 Here is a list of options that are @emph{only} for compiling C++ programs:
1455 @item -fabi-version=@var{n}
1456 @opindex fabi-version
1457 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1458 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1459 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1460 the version that conforms most closely to the C++ ABI specification.
1461 Therefore, the ABI obtained using version 0 will change as ABI bugs
1464 The default is version 2.
1466 @item -fno-access-control
1467 @opindex fno-access-control
1468 Turn off all access checking. This switch is mainly useful for working
1469 around bugs in the access control code.
1473 Check that the pointer returned by @code{operator new} is non-null
1474 before attempting to modify the storage allocated. This check is
1475 normally unnecessary because the C++ standard specifies that
1476 @code{operator new} will only return @code{0} if it is declared
1477 @samp{throw()}, in which case the compiler will always check the
1478 return value even without this option. In all other cases, when
1479 @code{operator new} has a non-empty exception specification, memory
1480 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1481 @samp{new (nothrow)}.
1483 @item -fconserve-space
1484 @opindex fconserve-space
1485 Put uninitialized or runtime-initialized global variables into the
1486 common segment, as C does. This saves space in the executable at the
1487 cost of not diagnosing duplicate definitions. If you compile with this
1488 flag and your program mysteriously crashes after @code{main()} has
1489 completed, you may have an object that is being destroyed twice because
1490 two definitions were merged.
1492 This option is no longer useful on most targets, now that support has
1493 been added for putting variables into BSS without making them common.
1495 @item -ffriend-injection
1496 @opindex ffriend-injection
1497 Inject friend functions into the enclosing namespace, so that they are
1498 visible outside the scope of the class in which they are declared.
1499 Friend functions were documented to work this way in the old Annotated
1500 C++ Reference Manual, and versions of G++ before 4.1 always worked
1501 that way. However, in ISO C++ a friend function which is not declared
1502 in an enclosing scope can only be found using argument dependent
1503 lookup. This option causes friends to be injected as they were in
1506 This option is for compatibility, and may be removed in a future
1509 @item -fno-elide-constructors
1510 @opindex fno-elide-constructors
1511 The C++ standard allows an implementation to omit creating a temporary
1512 which is only used to initialize another object of the same type.
1513 Specifying this option disables that optimization, and forces G++ to
1514 call the copy constructor in all cases.
1516 @item -fno-enforce-eh-specs
1517 @opindex fno-enforce-eh-specs
1518 Don't generate code to check for violation of exception specifications
1519 at runtime. This option violates the C++ standard, but may be useful
1520 for reducing code size in production builds, much like defining
1521 @samp{NDEBUG}. This does not give user code permission to throw
1522 exceptions in violation of the exception specifications; the compiler
1523 will still optimize based on the specifications, so throwing an
1524 unexpected exception will result in undefined behavior.
1527 @itemx -fno-for-scope
1529 @opindex fno-for-scope
1530 If @option{-ffor-scope} is specified, the scope of variables declared in
1531 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1532 as specified by the C++ standard.
1533 If @option{-fno-for-scope} is specified, the scope of variables declared in
1534 a @i{for-init-statement} extends to the end of the enclosing scope,
1535 as was the case in old versions of G++, and other (traditional)
1536 implementations of C++.
1538 The default if neither flag is given to follow the standard,
1539 but to allow and give a warning for old-style code that would
1540 otherwise be invalid, or have different behavior.
1542 @item -fno-gnu-keywords
1543 @opindex fno-gnu-keywords
1544 Do not recognize @code{typeof} as a keyword, so that code can use this
1545 word as an identifier. You can use the keyword @code{__typeof__} instead.
1546 @option{-ansi} implies @option{-fno-gnu-keywords}.
1548 @item -fno-implicit-templates
1549 @opindex fno-implicit-templates
1550 Never emit code for non-inline templates which are instantiated
1551 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1552 @xref{Template Instantiation}, for more information.
1554 @item -fno-implicit-inline-templates
1555 @opindex fno-implicit-inline-templates
1556 Don't emit code for implicit instantiations of inline templates, either.
1557 The default is to handle inlines differently so that compiles with and
1558 without optimization will need the same set of explicit instantiations.
1560 @item -fno-implement-inlines
1561 @opindex fno-implement-inlines
1562 To save space, do not emit out-of-line copies of inline functions
1563 controlled by @samp{#pragma implementation}. This will cause linker
1564 errors if these functions are not inlined everywhere they are called.
1566 @item -fms-extensions
1567 @opindex fms-extensions
1568 Disable pedantic warnings about constructs used in MFC, such as implicit
1569 int and getting a pointer to member function via non-standard syntax.
1571 @item -fno-nonansi-builtins
1572 @opindex fno-nonansi-builtins
1573 Disable built-in declarations of functions that are not mandated by
1574 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1575 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1577 @item -fno-operator-names
1578 @opindex fno-operator-names
1579 Do not treat the operator name keywords @code{and}, @code{bitand},
1580 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1581 synonyms as keywords.
1583 @item -fno-optional-diags
1584 @opindex fno-optional-diags
1585 Disable diagnostics that the standard says a compiler does not need to
1586 issue. Currently, the only such diagnostic issued by G++ is the one for
1587 a name having multiple meanings within a class.
1590 @opindex fpermissive
1591 Downgrade some diagnostics about nonconformant code from errors to
1592 warnings. Thus, using @option{-fpermissive} will allow some
1593 nonconforming code to compile.
1597 Enable automatic template instantiation at link time. This option also
1598 implies @option{-fno-implicit-templates}. @xref{Template
1599 Instantiation}, for more information.
1603 Disable generation of information about every class with virtual
1604 functions for use by the C++ runtime type identification features
1605 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1606 of the language, you can save some space by using this flag. Note that
1607 exception handling uses the same information, but it will generate it as
1608 needed. The @samp{dynamic_cast} operator can still be used for casts that
1609 do not require runtime type information, i.e. casts to @code{void *} or to
1610 unambiguous base classes.
1614 Emit statistics about front-end processing at the end of the compilation.
1615 This information is generally only useful to the G++ development team.
1617 @item -ftemplate-depth-@var{n}
1618 @opindex ftemplate-depth
1619 Set the maximum instantiation depth for template classes to @var{n}.
1620 A limit on the template instantiation depth is needed to detect
1621 endless recursions during template class instantiation. ANSI/ISO C++
1622 conforming programs must not rely on a maximum depth greater than 17.
1624 @item -fno-threadsafe-statics
1625 @opindex fno-threadsafe-statics
1626 Do not emit the extra code to use the routines specified in the C++
1627 ABI for thread-safe initialization of local statics. You can use this
1628 option to reduce code size slightly in code that doesn't need to be
1631 @item -fuse-cxa-atexit
1632 @opindex fuse-cxa-atexit
1633 Register destructors for objects with static storage duration with the
1634 @code{__cxa_atexit} function rather than the @code{atexit} function.
1635 This option is required for fully standards-compliant handling of static
1636 destructors, but will only work if your C library supports
1637 @code{__cxa_atexit}.
1639 @item -fno-use-cxa-get-exception-ptr
1640 @opindex fno-use-cxa-get-exception-ptr
1641 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1642 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1643 if the runtime routine is not available.
1645 @item -fvisibility-inlines-hidden
1646 @opindex fvisibility-inlines-hidden
1647 This switch declares that the user does not attempt to compare
1648 pointers to inline methods where the addresses of the two functions
1649 were taken in different shared objects.
1651 The effect of this is that GCC may, effectively, mark inline methods with
1652 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1653 appear in the export table of a DSO and do not require a PLT indirection
1654 when used within the DSO@. Enabling this option can have a dramatic effect
1655 on load and link times of a DSO as it massively reduces the size of the
1656 dynamic export table when the library makes heavy use of templates.
1658 The behaviour of this switch is not quite the same as marking the
1659 methods as hidden directly, because it does not affect static variables
1660 local to the function or cause the compiler to deduce that
1661 the function is defined in only one shared object.
1663 You may mark a method as having a visibility explicitly to negate the
1664 effect of the switch for that method. For example, if you do want to
1665 compare pointers to a particular inline method, you might mark it as
1666 having default visibility. Marking the enclosing class with explicit
1667 visibility will have no effect.
1669 Explicitly instantiated inline methods are unaffected by this option
1670 as their linkage might otherwise cross a shared library boundary.
1671 @xref{Template Instantiation}.
1675 Do not use weak symbol support, even if it is provided by the linker.
1676 By default, G++ will use weak symbols if they are available. This
1677 option exists only for testing, and should not be used by end-users;
1678 it will result in inferior code and has no benefits. This option may
1679 be removed in a future release of G++.
1683 Do not search for header files in the standard directories specific to
1684 C++, but do still search the other standard directories. (This option
1685 is used when building the C++ library.)
1688 In addition, these optimization, warning, and code generation options
1689 have meanings only for C++ programs:
1692 @item -fno-default-inline
1693 @opindex fno-default-inline
1694 Do not assume @samp{inline} for functions defined inside a class scope.
1695 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1696 functions will have linkage like inline functions; they just won't be
1699 @item -Wabi @r{(C++ only)}
1701 Warn when G++ generates code that is probably not compatible with the
1702 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1703 all such cases, there are probably some cases that are not warned about,
1704 even though G++ is generating incompatible code. There may also be
1705 cases where warnings are emitted even though the code that is generated
1708 You should rewrite your code to avoid these warnings if you are
1709 concerned about the fact that code generated by G++ may not be binary
1710 compatible with code generated by other compilers.
1712 The known incompatibilities at this point include:
1717 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1718 pack data into the same byte as a base class. For example:
1721 struct A @{ virtual void f(); int f1 : 1; @};
1722 struct B : public A @{ int f2 : 1; @};
1726 In this case, G++ will place @code{B::f2} into the same byte
1727 as@code{A::f1}; other compilers will not. You can avoid this problem
1728 by explicitly padding @code{A} so that its size is a multiple of the
1729 byte size on your platform; that will cause G++ and other compilers to
1730 layout @code{B} identically.
1733 Incorrect handling of tail-padding for virtual bases. G++ does not use
1734 tail padding when laying out virtual bases. For example:
1737 struct A @{ virtual void f(); char c1; @};
1738 struct B @{ B(); char c2; @};
1739 struct C : public A, public virtual B @{@};
1743 In this case, G++ will not place @code{B} into the tail-padding for
1744 @code{A}; other compilers will. You can avoid this problem by
1745 explicitly padding @code{A} so that its size is a multiple of its
1746 alignment (ignoring virtual base classes); that will cause G++ and other
1747 compilers to layout @code{C} identically.
1750 Incorrect handling of bit-fields with declared widths greater than that
1751 of their underlying types, when the bit-fields appear in a union. For
1755 union U @{ int i : 4096; @};
1759 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1760 union too small by the number of bits in an @code{int}.
1763 Empty classes can be placed at incorrect offsets. For example:
1773 struct C : public B, public A @{@};
1777 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1778 it should be placed at offset zero. G++ mistakenly believes that the
1779 @code{A} data member of @code{B} is already at offset zero.
1782 Names of template functions whose types involve @code{typename} or
1783 template template parameters can be mangled incorrectly.
1786 template <typename Q>
1787 void f(typename Q::X) @{@}
1789 template <template <typename> class Q>
1790 void f(typename Q<int>::X) @{@}
1794 Instantiations of these templates may be mangled incorrectly.
1798 @item -Wctor-dtor-privacy @r{(C++ only)}
1799 @opindex Wctor-dtor-privacy
1800 Warn when a class seems unusable because all the constructors or
1801 destructors in that class are private, and it has neither friends nor
1802 public static member functions.
1804 @item -Wnon-virtual-dtor @r{(C++ only)}
1805 @opindex Wnon-virtual-dtor
1806 Warn when a class appears to be polymorphic, thereby requiring a virtual
1807 destructor, yet it declares a non-virtual one. This warning is also
1808 enabled if -Weffc++ is specified.
1810 @item -Wreorder @r{(C++ only)}
1812 @cindex reordering, warning
1813 @cindex warning for reordering of member initializers
1814 Warn when the order of member initializers given in the code does not
1815 match the order in which they must be executed. For instance:
1821 A(): j (0), i (1) @{ @}
1825 The compiler will rearrange the member initializers for @samp{i}
1826 and @samp{j} to match the declaration order of the members, emitting
1827 a warning to that effect. This warning is enabled by @option{-Wall}.
1830 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1833 @item -Weffc++ @r{(C++ only)}
1835 Warn about violations of the following style guidelines from Scott Meyers'
1836 @cite{Effective C++} book:
1840 Item 11: Define a copy constructor and an assignment operator for classes
1841 with dynamically allocated memory.
1844 Item 12: Prefer initialization to assignment in constructors.
1847 Item 14: Make destructors virtual in base classes.
1850 Item 15: Have @code{operator=} return a reference to @code{*this}.
1853 Item 23: Don't try to return a reference when you must return an object.
1857 Also warn about violations of the following style guidelines from
1858 Scott Meyers' @cite{More Effective C++} book:
1862 Item 6: Distinguish between prefix and postfix forms of increment and
1863 decrement operators.
1866 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1870 When selecting this option, be aware that the standard library
1871 headers do not obey all of these guidelines; use @samp{grep -v}
1872 to filter out those warnings.
1874 @item -Wno-deprecated @r{(C++ only)}
1875 @opindex Wno-deprecated
1876 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1878 @item -Wstrict-null-sentinel @r{(C++ only)}
1879 @opindex Wstrict-null-sentinel
1880 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1881 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1882 to @code{__null}. Although it is a null pointer constant not a null pointer,
1883 it is guaranteed to of the same size as a pointer. But this use is
1884 not portable across different compilers.
1886 @item -Wno-non-template-friend @r{(C++ only)}
1887 @opindex Wno-non-template-friend
1888 Disable warnings when non-templatized friend functions are declared
1889 within a template. Since the advent of explicit template specification
1890 support in G++, if the name of the friend is an unqualified-id (i.e.,
1891 @samp{friend foo(int)}), the C++ language specification demands that the
1892 friend declare or define an ordinary, nontemplate function. (Section
1893 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1894 could be interpreted as a particular specialization of a templatized
1895 function. Because this non-conforming behavior is no longer the default
1896 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1897 check existing code for potential trouble spots and is on by default.
1898 This new compiler behavior can be turned off with
1899 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1900 but disables the helpful warning.
1902 @item -Wold-style-cast @r{(C++ only)}
1903 @opindex Wold-style-cast
1904 Warn if an old-style (C-style) cast to a non-void type is used within
1905 a C++ program. The new-style casts (@samp{dynamic_cast},
1906 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1907 less vulnerable to unintended effects and much easier to search for.
1909 @item -Woverloaded-virtual @r{(C++ only)}
1910 @opindex Woverloaded-virtual
1911 @cindex overloaded virtual fn, warning
1912 @cindex warning for overloaded virtual fn
1913 Warn when a function declaration hides virtual functions from a
1914 base class. For example, in:
1921 struct B: public A @{
1926 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1934 will fail to compile.
1936 @item -Wno-pmf-conversions @r{(C++ only)}
1937 @opindex Wno-pmf-conversions
1938 Disable the diagnostic for converting a bound pointer to member function
1941 @item -Wsign-promo @r{(C++ only)}
1942 @opindex Wsign-promo
1943 Warn when overload resolution chooses a promotion from unsigned or
1944 enumerated type to a signed type, over a conversion to an unsigned type of
1945 the same size. Previous versions of G++ would try to preserve
1946 unsignedness, but the standard mandates the current behavior.
1951 A& operator = (int);
1961 In this example, G++ will synthesize a default @samp{A& operator =
1962 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1965 @node Objective-C and Objective-C++ Dialect Options
1966 @section Options Controlling Objective-C and Objective-C++ Dialects
1968 @cindex compiler options, Objective-C and Objective-C++
1969 @cindex Objective-C and Objective-C++ options, command line
1970 @cindex options, Objective-C and Objective-C++
1971 (NOTE: This manual does not describe the Objective-C and Objective-C++
1972 languages themselves. See @xref{Standards,,Language Standards
1973 Supported by GCC}, for references.)
1975 This section describes the command-line options that are only meaningful
1976 for Objective-C and Objective-C++ programs, but you can also use most of
1977 the language-independent GNU compiler options.
1978 For example, you might compile a file @code{some_class.m} like this:
1981 gcc -g -fgnu-runtime -O -c some_class.m
1985 In this example, @option{-fgnu-runtime} is an option meant only for
1986 Objective-C and Objective-C++ programs; you can use the other options with
1987 any language supported by GCC@.
1989 Note that since Objective-C is an extension of the C language, Objective-C
1990 compilations may also use options specific to the C front-end (e.g.,
1991 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1992 C++-specific options (e.g., @option{-Wabi}).
1994 Here is a list of options that are @emph{only} for compiling Objective-C
1995 and Objective-C++ programs:
1998 @item -fconstant-string-class=@var{class-name}
1999 @opindex fconstant-string-class
2000 Use @var{class-name} as the name of the class to instantiate for each
2001 literal string specified with the syntax @code{@@"@dots{}"}. The default
2002 class name is @code{NXConstantString} if the GNU runtime is being used, and
2003 @code{NSConstantString} if the NeXT runtime is being used (see below). The
2004 @option{-fconstant-cfstrings} option, if also present, will override the
2005 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
2006 to be laid out as constant CoreFoundation strings.
2009 @opindex fgnu-runtime
2010 Generate object code compatible with the standard GNU Objective-C
2011 runtime. This is the default for most types of systems.
2013 @item -fnext-runtime
2014 @opindex fnext-runtime
2015 Generate output compatible with the NeXT runtime. This is the default
2016 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2017 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2020 @item -fno-nil-receivers
2021 @opindex fno-nil-receivers
2022 Assume that all Objective-C message dispatches (e.g.,
2023 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2024 is not @code{nil}. This allows for more efficient entry points in the runtime
2025 to be used. Currently, this option is only available in conjunction with
2026 the NeXT runtime on Mac OS X 10.3 and later.
2028 @item -fobjc-call-cxx-cdtors
2029 @opindex fobjc-call-cxx-cdtors
2030 For each Objective-C class, check if any of its instance variables is a
2031 C++ object with a non-trivial default constructor. If so, synthesize a
2032 special @code{- (id) .cxx_construct} instance method that will run
2033 non-trivial default constructors on any such instance variables, in order,
2034 and then return @code{self}. Similarly, check if any instance variable
2035 is a C++ object with a non-trivial destructor, and if so, synthesize a
2036 special @code{- (void) .cxx_destruct} method that will run
2037 all such default destructors, in reverse order.
2039 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2040 thusly generated will only operate on instance variables declared in the
2041 current Objective-C class, and not those inherited from superclasses. It
2042 is the responsibility of the Objective-C runtime to invoke all such methods
2043 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2044 will be invoked by the runtime immediately after a new object
2045 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2046 be invoked immediately before the runtime deallocates an object instance.
2048 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2049 support for invoking the @code{- (id) .cxx_construct} and
2050 @code{- (void) .cxx_destruct} methods.
2052 @item -fobjc-direct-dispatch
2053 @opindex fobjc-direct-dispatch
2054 Allow fast jumps to the message dispatcher. On Darwin this is
2055 accomplished via the comm page.
2057 @item -fobjc-exceptions
2058 @opindex fobjc-exceptions
2059 Enable syntactic support for structured exception handling in Objective-C,
2060 similar to what is offered by C++ and Java. This option is
2061 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2070 @@catch (AnObjCClass *exc) @{
2077 @@catch (AnotherClass *exc) @{
2080 @@catch (id allOthers) @{
2090 The @code{@@throw} statement may appear anywhere in an Objective-C or
2091 Objective-C++ program; when used inside of a @code{@@catch} block, the
2092 @code{@@throw} may appear without an argument (as shown above), in which case
2093 the object caught by the @code{@@catch} will be rethrown.
2095 Note that only (pointers to) Objective-C objects may be thrown and
2096 caught using this scheme. When an object is thrown, it will be caught
2097 by the nearest @code{@@catch} clause capable of handling objects of that type,
2098 analogously to how @code{catch} blocks work in C++ and Java. A
2099 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2100 any and all Objective-C exceptions not caught by previous @code{@@catch}
2103 The @code{@@finally} clause, if present, will be executed upon exit from the
2104 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2105 regardless of whether any exceptions are thrown, caught or rethrown
2106 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2107 of the @code{finally} clause in Java.
2109 There are several caveats to using the new exception mechanism:
2113 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2114 idioms provided by the @code{NSException} class, the new
2115 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2116 systems, due to additional functionality needed in the (NeXT) Objective-C
2120 As mentioned above, the new exceptions do not support handling
2121 types other than Objective-C objects. Furthermore, when used from
2122 Objective-C++, the Objective-C exception model does not interoperate with C++
2123 exceptions at this time. This means you cannot @code{@@throw} an exception
2124 from Objective-C and @code{catch} it in C++, or vice versa
2125 (i.e., @code{throw @dots{} @@catch}).
2128 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2129 blocks for thread-safe execution:
2132 @@synchronized (ObjCClass *guard) @{
2137 Upon entering the @code{@@synchronized} block, a thread of execution shall
2138 first check whether a lock has been placed on the corresponding @code{guard}
2139 object by another thread. If it has, the current thread shall wait until
2140 the other thread relinquishes its lock. Once @code{guard} becomes available,
2141 the current thread will place its own lock on it, execute the code contained in
2142 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2143 making @code{guard} available to other threads).
2145 Unlike Java, Objective-C does not allow for entire methods to be marked
2146 @code{@@synchronized}. Note that throwing exceptions out of
2147 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2148 to be unlocked properly.
2152 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2154 @item -freplace-objc-classes
2155 @opindex freplace-objc-classes
2156 Emit a special marker instructing @command{ld(1)} not to statically link in
2157 the resulting object file, and allow @command{dyld(1)} to load it in at
2158 run time instead. This is used in conjunction with the Fix-and-Continue
2159 debugging mode, where the object file in question may be recompiled and
2160 dynamically reloaded in the course of program execution, without the need
2161 to restart the program itself. Currently, Fix-and-Continue functionality
2162 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2167 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2168 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2169 compile time) with static class references that get initialized at load time,
2170 which improves run-time performance. Specifying the @option{-fzero-link} flag
2171 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2172 to be retained. This is useful in Zero-Link debugging mode, since it allows
2173 for individual class implementations to be modified during program execution.
2177 Dump interface declarations for all classes seen in the source file to a
2178 file named @file{@var{sourcename}.decl}.
2180 @item -Wassign-intercept
2181 @opindex Wassign-intercept
2182 Warn whenever an Objective-C assignment is being intercepted by the
2186 @opindex Wno-protocol
2187 If a class is declared to implement a protocol, a warning is issued for
2188 every method in the protocol that is not implemented by the class. The
2189 default behavior is to issue a warning for every method not explicitly
2190 implemented in the class, even if a method implementation is inherited
2191 from the superclass. If you use the @option{-Wno-protocol} option, then
2192 methods inherited from the superclass are considered to be implemented,
2193 and no warning is issued for them.
2197 Warn if multiple methods of different types for the same selector are
2198 found during compilation. The check is performed on the list of methods
2199 in the final stage of compilation. Additionally, a check is performed
2200 for each selector appearing in a @code{@@selector(@dots{})}
2201 expression, and a corresponding method for that selector has been found
2202 during compilation. Because these checks scan the method table only at
2203 the end of compilation, these warnings are not produced if the final
2204 stage of compilation is not reached, for example because an error is
2205 found during compilation, or because the @option{-fsyntax-only} option is
2208 @item -Wstrict-selector-match
2209 @opindex Wstrict-selector-match
2210 Warn if multiple methods with differing argument and/or return types are
2211 found for a given selector when attempting to send a message using this
2212 selector to a receiver of type @code{id} or @code{Class}. When this flag
2213 is off (which is the default behavior), the compiler will omit such warnings
2214 if any differences found are confined to types which share the same size
2217 @item -Wundeclared-selector
2218 @opindex Wundeclared-selector
2219 Warn if a @code{@@selector(@dots{})} expression referring to an
2220 undeclared selector is found. A selector is considered undeclared if no
2221 method with that name has been declared before the
2222 @code{@@selector(@dots{})} expression, either explicitly in an
2223 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2224 an @code{@@implementation} section. This option always performs its
2225 checks as soon as a @code{@@selector(@dots{})} expression is found,
2226 while @option{-Wselector} only performs its checks in the final stage of
2227 compilation. This also enforces the coding style convention
2228 that methods and selectors must be declared before being used.
2230 @item -print-objc-runtime-info
2231 @opindex print-objc-runtime-info
2232 Generate C header describing the largest structure that is passed by
2237 @node Language Independent Options
2238 @section Options to Control Diagnostic Messages Formatting
2239 @cindex options to control diagnostics formatting
2240 @cindex diagnostic messages
2241 @cindex message formatting
2243 Traditionally, diagnostic messages have been formatted irrespective of
2244 the output device's aspect (e.g.@: its width, @dots{}). The options described
2245 below can be used to control the diagnostic messages formatting
2246 algorithm, e.g.@: how many characters per line, how often source location
2247 information should be reported. Right now, only the C++ front end can
2248 honor these options. However it is expected, in the near future, that
2249 the remaining front ends would be able to digest them correctly.
2252 @item -fmessage-length=@var{n}
2253 @opindex fmessage-length
2254 Try to format error messages so that they fit on lines of about @var{n}
2255 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2256 the front ends supported by GCC@. If @var{n} is zero, then no
2257 line-wrapping will be done; each error message will appear on a single
2260 @opindex fdiagnostics-show-location
2261 @item -fdiagnostics-show-location=once
2262 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2263 reporter to emit @emph{once} source location information; that is, in
2264 case the message is too long to fit on a single physical line and has to
2265 be wrapped, the source location won't be emitted (as prefix) again,
2266 over and over, in subsequent continuation lines. This is the default
2269 @item -fdiagnostics-show-location=every-line
2270 Only meaningful in line-wrapping mode. Instructs the diagnostic
2271 messages reporter to emit the same source location information (as
2272 prefix) for physical lines that result from the process of breaking
2273 a message which is too long to fit on a single line.
2275 @item -fdiagnostics-show-option
2276 @opindex fdiagnostics-show-option
2277 This option instructs the diagnostic machinery to add text to each
2278 diagnostic emitted, which indicates which command line option directly
2279 controls that diagnostic, when such an option is known to the
2280 diagnostic machinery.
2284 @node Warning Options
2285 @section Options to Request or Suppress Warnings
2286 @cindex options to control warnings
2287 @cindex warning messages
2288 @cindex messages, warning
2289 @cindex suppressing warnings
2291 Warnings are diagnostic messages that report constructions which
2292 are not inherently erroneous but which are risky or suggest there
2293 may have been an error.
2295 You can request many specific warnings with options beginning @samp{-W},
2296 for example @option{-Wimplicit} to request warnings on implicit
2297 declarations. Each of these specific warning options also has a
2298 negative form beginning @samp{-Wno-} to turn off warnings;
2299 for example, @option{-Wno-implicit}. This manual lists only one of the
2300 two forms, whichever is not the default.
2302 The following options control the amount and kinds of warnings produced
2303 by GCC; for further, language-specific options also refer to
2304 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2308 @cindex syntax checking
2310 @opindex fsyntax-only
2311 Check the code for syntax errors, but don't do anything beyond that.
2315 Issue all the warnings demanded by strict ISO C and ISO C++;
2316 reject all programs that use forbidden extensions, and some other
2317 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2318 version of the ISO C standard specified by any @option{-std} option used.
2320 Valid ISO C and ISO C++ programs should compile properly with or without
2321 this option (though a rare few will require @option{-ansi} or a
2322 @option{-std} option specifying the required version of ISO C)@. However,
2323 without this option, certain GNU extensions and traditional C and C++
2324 features are supported as well. With this option, they are rejected.
2326 @option{-pedantic} does not cause warning messages for use of the
2327 alternate keywords whose names begin and end with @samp{__}. Pedantic
2328 warnings are also disabled in the expression that follows
2329 @code{__extension__}. However, only system header files should use
2330 these escape routes; application programs should avoid them.
2331 @xref{Alternate Keywords}.
2333 Some users try to use @option{-pedantic} to check programs for strict ISO
2334 C conformance. They soon find that it does not do quite what they want:
2335 it finds some non-ISO practices, but not all---only those for which
2336 ISO C @emph{requires} a diagnostic, and some others for which
2337 diagnostics have been added.
2339 A feature to report any failure to conform to ISO C might be useful in
2340 some instances, but would require considerable additional work and would
2341 be quite different from @option{-pedantic}. We don't have plans to
2342 support such a feature in the near future.
2344 Where the standard specified with @option{-std} represents a GNU
2345 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2346 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2347 extended dialect is based. Warnings from @option{-pedantic} are given
2348 where they are required by the base standard. (It would not make sense
2349 for such warnings to be given only for features not in the specified GNU
2350 C dialect, since by definition the GNU dialects of C include all
2351 features the compiler supports with the given option, and there would be
2352 nothing to warn about.)
2354 @item -pedantic-errors
2355 @opindex pedantic-errors
2356 Like @option{-pedantic}, except that errors are produced rather than
2361 Inhibit all warning messages.
2365 Inhibit warning messages about the use of @samp{#import}.
2367 @item -Wchar-subscripts
2368 @opindex Wchar-subscripts
2369 Warn if an array subscript has type @code{char}. This is a common cause
2370 of error, as programmers often forget that this type is signed on some
2372 This warning is enabled by @option{-Wall}.
2376 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2377 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2378 This warning is enabled by @option{-Wall}.
2380 @item -Wfatal-errors
2381 @opindex Wfatal-errors
2382 This option causes the compiler to abort compilation on the first error
2383 occurred rather than trying to keep going and printing further error
2388 @opindex ffreestanding
2389 @opindex fno-builtin
2390 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2391 the arguments supplied have types appropriate to the format string
2392 specified, and that the conversions specified in the format string make
2393 sense. This includes standard functions, and others specified by format
2394 attributes (@pxref{Function Attributes}), in the @code{printf},
2395 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2396 not in the C standard) families (or other target-specific families).
2397 Which functions are checked without format attributes having been
2398 specified depends on the standard version selected, and such checks of
2399 functions without the attribute specified are disabled by
2400 @option{-ffreestanding} or @option{-fno-builtin}.
2402 The formats are checked against the format features supported by GNU
2403 libc version 2.2. These include all ISO C90 and C99 features, as well
2404 as features from the Single Unix Specification and some BSD and GNU
2405 extensions. Other library implementations may not support all these
2406 features; GCC does not support warning about features that go beyond a
2407 particular library's limitations. However, if @option{-pedantic} is used
2408 with @option{-Wformat}, warnings will be given about format features not
2409 in the selected standard version (but not for @code{strfmon} formats,
2410 since those are not in any version of the C standard). @xref{C Dialect
2411 Options,,Options Controlling C Dialect}.
2413 Since @option{-Wformat} also checks for null format arguments for
2414 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2416 @option{-Wformat} is included in @option{-Wall}. For more control over some
2417 aspects of format checking, the options @option{-Wformat-y2k},
2418 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2419 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2420 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2423 @opindex Wformat-y2k
2424 If @option{-Wformat} is specified, also warn about @code{strftime}
2425 formats which may yield only a two-digit year.
2427 @item -Wno-format-extra-args
2428 @opindex Wno-format-extra-args
2429 If @option{-Wformat} is specified, do not warn about excess arguments to a
2430 @code{printf} or @code{scanf} format function. The C standard specifies
2431 that such arguments are ignored.
2433 Where the unused arguments lie between used arguments that are
2434 specified with @samp{$} operand number specifications, normally
2435 warnings are still given, since the implementation could not know what
2436 type to pass to @code{va_arg} to skip the unused arguments. However,
2437 in the case of @code{scanf} formats, this option will suppress the
2438 warning if the unused arguments are all pointers, since the Single
2439 Unix Specification says that such unused arguments are allowed.
2441 @item -Wno-format-zero-length
2442 @opindex Wno-format-zero-length
2443 If @option{-Wformat} is specified, do not warn about zero-length formats.
2444 The C standard specifies that zero-length formats are allowed.
2446 @item -Wformat-nonliteral
2447 @opindex Wformat-nonliteral
2448 If @option{-Wformat} is specified, also warn if the format string is not a
2449 string literal and so cannot be checked, unless the format function
2450 takes its format arguments as a @code{va_list}.
2452 @item -Wformat-security
2453 @opindex Wformat-security
2454 If @option{-Wformat} is specified, also warn about uses of format
2455 functions that represent possible security problems. At present, this
2456 warns about calls to @code{printf} and @code{scanf} functions where the
2457 format string is not a string literal and there are no format arguments,
2458 as in @code{printf (foo);}. This may be a security hole if the format
2459 string came from untrusted input and contains @samp{%n}. (This is
2460 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2461 in future warnings may be added to @option{-Wformat-security} that are not
2462 included in @option{-Wformat-nonliteral}.)
2466 Enable @option{-Wformat} plus format checks not included in
2467 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2468 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2472 Warn about passing a null pointer for arguments marked as
2473 requiring a non-null value by the @code{nonnull} function attribute.
2475 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2476 can be disabled with the @option{-Wno-nonnull} option.
2478 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2480 Warn about uninitialized variables which are initialized with themselves.
2481 Note this option can only be used with the @option{-Wuninitialized} option,
2482 which in turn only works with @option{-O1} and above.
2484 For example, GCC will warn about @code{i} being uninitialized in the
2485 following snippet only when @option{-Winit-self} has been specified:
2496 @item -Wimplicit-int
2497 @opindex Wimplicit-int
2498 Warn when a declaration does not specify a type.
2499 This warning is enabled by @option{-Wall}.
2501 @item -Wimplicit-function-declaration
2502 @itemx -Werror-implicit-function-declaration
2503 @opindex Wimplicit-function-declaration
2504 @opindex Werror-implicit-function-declaration
2505 Give a warning (or error) whenever a function is used before being
2506 declared. The form @option{-Wno-error-implicit-function-declaration}
2508 This warning is enabled by @option{-Wall} (as a warning, not an error).
2512 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2513 This warning is enabled by @option{-Wall}.
2517 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2518 function with external linkage, returning int, taking either zero
2519 arguments, two, or three arguments of appropriate types.
2520 This warning is enabled by @option{-Wall}.
2522 @item -Wmissing-braces
2523 @opindex Wmissing-braces
2524 Warn if an aggregate or union initializer is not fully bracketed. In
2525 the following example, the initializer for @samp{a} is not fully
2526 bracketed, but that for @samp{b} is fully bracketed.
2529 int a[2][2] = @{ 0, 1, 2, 3 @};
2530 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2533 This warning is enabled by @option{-Wall}.
2535 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2536 @opindex Wmissing-include-dirs
2537 Warn if a user-supplied include directory does not exist.
2540 @opindex Wparentheses
2541 Warn if parentheses are omitted in certain contexts, such
2542 as when there is an assignment in a context where a truth value
2543 is expected, or when operators are nested whose precedence people
2544 often get confused about.
2546 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2547 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2548 interpretation from that of ordinary mathematical notation.
2550 Also warn about constructions where there may be confusion to which
2551 @code{if} statement an @code{else} branch belongs. Here is an example of
2566 In C/C++, every @code{else} branch belongs to the innermost possible
2567 @code{if} statement, which in this example is @code{if (b)}. This is
2568 often not what the programmer expected, as illustrated in the above
2569 example by indentation the programmer chose. When there is the
2570 potential for this confusion, GCC will issue a warning when this flag
2571 is specified. To eliminate the warning, add explicit braces around
2572 the innermost @code{if} statement so there is no way the @code{else}
2573 could belong to the enclosing @code{if}. The resulting code would
2590 This warning is enabled by @option{-Wall}.
2592 @item -Wsequence-point
2593 @opindex Wsequence-point
2594 Warn about code that may have undefined semantics because of violations
2595 of sequence point rules in the C and C++ standards.
2597 The C and C++ standards defines the order in which expressions in a C/C++
2598 program are evaluated in terms of @dfn{sequence points}, which represent
2599 a partial ordering between the execution of parts of the program: those
2600 executed before the sequence point, and those executed after it. These
2601 occur after the evaluation of a full expression (one which is not part
2602 of a larger expression), after the evaluation of the first operand of a
2603 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2604 function is called (but after the evaluation of its arguments and the
2605 expression denoting the called function), and in certain other places.
2606 Other than as expressed by the sequence point rules, the order of
2607 evaluation of subexpressions of an expression is not specified. All
2608 these rules describe only a partial order rather than a total order,
2609 since, for example, if two functions are called within one expression
2610 with no sequence point between them, the order in which the functions
2611 are called is not specified. However, the standards committee have
2612 ruled that function calls do not overlap.
2614 It is not specified when between sequence points modifications to the
2615 values of objects take effect. Programs whose behavior depends on this
2616 have undefined behavior; the C and C++ standards specify that ``Between
2617 the previous and next sequence point an object shall have its stored
2618 value modified at most once by the evaluation of an expression.
2619 Furthermore, the prior value shall be read only to determine the value
2620 to be stored.''. If a program breaks these rules, the results on any
2621 particular implementation are entirely unpredictable.
2623 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2624 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2625 diagnosed by this option, and it may give an occasional false positive
2626 result, but in general it has been found fairly effective at detecting
2627 this sort of problem in programs.
2629 The standard is worded confusingly, therefore there is some debate
2630 over the precise meaning of the sequence point rules in subtle cases.
2631 Links to discussions of the problem, including proposed formal
2632 definitions, may be found on the GCC readings page, at
2633 @w{@uref{http://gcc.gnu.org/readings.html}}.
2635 This warning is enabled by @option{-Wall} for C and C++.
2638 @opindex Wreturn-type
2639 Warn whenever a function is defined with a return-type that defaults to
2640 @code{int}. Also warn about any @code{return} statement with no
2641 return-value in a function whose return-type is not @code{void}.
2643 For C, also warn if the return type of a function has a type qualifier
2644 such as @code{const}. Such a type qualifier has no effect, since the
2645 value returned by a function is not an lvalue. ISO C prohibits
2646 qualified @code{void} return types on function definitions, so such
2647 return types always receive a warning even without this option.
2649 For C++, a function without return type always produces a diagnostic
2650 message, even when @option{-Wno-return-type} is specified. The only
2651 exceptions are @samp{main} and functions defined in system headers.
2653 This warning is enabled by @option{-Wall}.
2657 Warn whenever a @code{switch} statement has an index of enumerated type
2658 and lacks a @code{case} for one or more of the named codes of that
2659 enumeration. (The presence of a @code{default} label prevents this
2660 warning.) @code{case} labels outside the enumeration range also
2661 provoke warnings when this option is used.
2662 This warning is enabled by @option{-Wall}.
2664 @item -Wswitch-default
2665 @opindex Wswitch-switch
2666 Warn whenever a @code{switch} statement does not have a @code{default}
2670 @opindex Wswitch-enum
2671 Warn whenever a @code{switch} statement has an index of enumerated type
2672 and lacks a @code{case} for one or more of the named codes of that
2673 enumeration. @code{case} labels outside the enumeration range also
2674 provoke warnings when this option is used.
2678 Warn if any trigraphs are encountered that might change the meaning of
2679 the program (trigraphs within comments are not warned about).
2680 This warning is enabled by @option{-Wall}.
2682 @item -Wunused-function
2683 @opindex Wunused-function
2684 Warn whenever a static function is declared but not defined or a
2685 non-inline static function is unused.
2686 This warning is enabled by @option{-Wall}.
2688 @item -Wunused-label
2689 @opindex Wunused-label
2690 Warn whenever a label is declared but not used.
2691 This warning is enabled by @option{-Wall}.
2693 To suppress this warning use the @samp{unused} attribute
2694 (@pxref{Variable Attributes}).
2696 @item -Wunused-parameter
2697 @opindex Wunused-parameter
2698 Warn whenever a function parameter is unused aside from its declaration.
2700 To suppress this warning use the @samp{unused} attribute
2701 (@pxref{Variable Attributes}).
2703 @item -Wunused-variable
2704 @opindex Wunused-variable
2705 Warn whenever a local variable or non-constant static variable is unused
2706 aside from its declaration.
2707 This warning is enabled by @option{-Wall}.
2709 To suppress this warning use the @samp{unused} attribute
2710 (@pxref{Variable Attributes}).
2712 @item -Wunused-value
2713 @opindex Wunused-value
2714 Warn whenever a statement computes a result that is explicitly not used.
2715 This warning is enabled by @option{-Wall}.
2717 To suppress this warning cast the expression to @samp{void}.
2721 All the above @option{-Wunused} options combined.
2723 In order to get a warning about an unused function parameter, you must
2724 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2725 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2727 @item -Wuninitialized
2728 @opindex Wuninitialized
2729 Warn if an automatic variable is used without first being initialized or
2730 if a variable may be clobbered by a @code{setjmp} call.
2732 These warnings are possible only in optimizing compilation,
2733 because they require data flow information that is computed only
2734 when optimizing. If you do not specify @option{-O}, you will not get
2735 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2736 requiring @option{-O}.
2738 If you want to warn about code which uses the uninitialized value of the
2739 variable in its own initializer, use the @option{-Winit-self} option.
2741 These warnings occur for individual uninitialized or clobbered
2742 elements of structure, union or array variables as well as for
2743 variables which are uninitialized or clobbered as a whole. They do
2744 not occur for variables or elements declared @code{volatile}. Because
2745 these warnings depend on optimization, the exact variables or elements
2746 for which there are warnings will depend on the precise optimization
2747 options and version of GCC used.
2749 Note that there may be no warning about a variable that is used only
2750 to compute a value that itself is never used, because such
2751 computations may be deleted by data flow analysis before the warnings
2754 These warnings are made optional because GCC is not smart
2755 enough to see all the reasons why the code might be correct
2756 despite appearing to have an error. Here is one example of how
2777 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2778 always initialized, but GCC doesn't know this. Here is
2779 another common case:
2784 if (change_y) save_y = y, y = new_y;
2786 if (change_y) y = save_y;
2791 This has no bug because @code{save_y} is used only if it is set.
2793 @cindex @code{longjmp} warnings
2794 This option also warns when a non-volatile automatic variable might be
2795 changed by a call to @code{longjmp}. These warnings as well are possible
2796 only in optimizing compilation.
2798 The compiler sees only the calls to @code{setjmp}. It cannot know
2799 where @code{longjmp} will be called; in fact, a signal handler could
2800 call it at any point in the code. As a result, you may get a warning
2801 even when there is in fact no problem because @code{longjmp} cannot
2802 in fact be called at the place which would cause a problem.
2804 Some spurious warnings can be avoided if you declare all the functions
2805 you use that never return as @code{noreturn}. @xref{Function
2808 This warning is enabled by @option{-Wall}.
2810 @item -Wunknown-pragmas
2811 @opindex Wunknown-pragmas
2812 @cindex warning for unknown pragmas
2813 @cindex unknown pragmas, warning
2814 @cindex pragmas, warning of unknown
2815 Warn when a #pragma directive is encountered which is not understood by
2816 GCC@. If this command line option is used, warnings will even be issued
2817 for unknown pragmas in system header files. This is not the case if
2818 the warnings were only enabled by the @option{-Wall} command line option.
2821 @opindex Wno-pragmas
2823 Do not warn about misuses of pragmas, such as incorrect parameters,
2824 invalid syntax, or conflicts between pragmas. See also
2825 @samp{-Wunknown-pragmas}.
2827 @item -Wstrict-aliasing
2828 @opindex Wstrict-aliasing
2829 This option is only active when @option{-fstrict-aliasing} is active.
2830 It warns about code which might break the strict aliasing rules that the
2831 compiler is using for optimization. The warning does not catch all
2832 cases, but does attempt to catch the more common pitfalls. It is
2833 included in @option{-Wall}.
2835 @item -Wstrict-aliasing=2
2836 @opindex Wstrict-aliasing=2
2837 This option is only active when @option{-fstrict-aliasing} is active.
2838 It warns about code which might break the strict aliasing rules that the
2839 compiler is using for optimization. This warning catches more cases than
2840 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2841 cases that are safe.
2845 All of the above @samp{-W} options combined. This enables all the
2846 warnings about constructions that some users consider questionable, and
2847 that are easy to avoid (or modify to prevent the warning), even in
2848 conjunction with macros. This also enables some language-specific
2849 warnings described in @ref{C++ Dialect Options} and
2850 @ref{Objective-C and Objective-C++ Dialect Options}.
2853 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2854 Some of them warn about constructions that users generally do not
2855 consider questionable, but which occasionally you might wish to check
2856 for; others warn about constructions that are necessary or hard to avoid
2857 in some cases, and there is no simple way to modify the code to suppress
2864 (This option used to be called @option{-W}. The older name is still
2865 supported, but the newer name is more descriptive.) Print extra warning
2866 messages for these events:
2870 A function can return either with or without a value. (Falling
2871 off the end of the function body is considered returning without
2872 a value.) For example, this function would evoke such a
2886 An expression-statement or the left-hand side of a comma expression
2887 contains no side effects.
2888 To suppress the warning, cast the unused expression to void.
2889 For example, an expression such as @samp{x[i,j]} will cause a warning,
2890 but @samp{x[(void)i,j]} will not.
2893 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2896 Storage-class specifiers like @code{static} are not the first things
2897 in a declaration. According to the C Standard, this usage is
2898 obsolescent. This warning can be independently controlled by
2899 @option{-Wold-style-declaration}.
2902 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2906 A comparison between signed and unsigned values could produce an
2907 incorrect result when the signed value is converted to unsigned.
2908 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2911 An aggregate has an initializer which does not initialize all members.
2912 This warning can be independently controlled by
2913 @option{-Wmissing-field-initializers}.
2916 An initialized field without side effects is overridden when using
2917 designated initializers (@pxref{Designated Inits, , Designated
2918 Initializers}). This warning can be independently controlled by
2919 @option{-Woverride-init}.
2922 A function parameter is declared without a type specifier in K&R-style
2923 functions. This warning can be independently controlled by
2924 @option{-Wmissing-parameter-type}.
2927 An empty body occurs in an @samp{if} or @samp{else} statement. This
2928 warning can be independently controlled by @option{-Wempty-body}.
2931 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2932 @samp{>}, or @samp{>=}.
2935 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2936 This warning can be independently controlled by @option{-Wclobbered}.
2939 Any of several floating-point events that often indicate errors, such as
2940 overflow, underflow, loss of precision, etc.
2942 @item @r{(C++ only)}
2943 An enumerator and a non-enumerator both appear in a conditional expression.
2945 @item @r{(C++ only)}
2946 A non-static reference or non-static @samp{const} member appears in a
2947 class without constructors.
2949 @item @r{(C++ only)}
2950 Ambiguous virtual bases.
2952 @item @r{(C++ only)}
2953 Subscripting an array which has been declared @samp{register}.
2955 @item @r{(C++ only)}
2956 Taking the address of a variable which has been declared @samp{register}.
2958 @item @r{(C++ only)}
2959 A base class is not initialized in a derived class' copy constructor.
2962 @item -Wno-div-by-zero
2963 @opindex Wno-div-by-zero
2964 @opindex Wdiv-by-zero
2965 Do not warn about compile-time integer division by zero. Floating point
2966 division by zero is not warned about, as it can be a legitimate way of
2967 obtaining infinities and NaNs.
2969 @item -Wsystem-headers
2970 @opindex Wsystem-headers
2971 @cindex warnings from system headers
2972 @cindex system headers, warnings from
2973 Print warning messages for constructs found in system header files.
2974 Warnings from system headers are normally suppressed, on the assumption
2975 that they usually do not indicate real problems and would only make the
2976 compiler output harder to read. Using this command line option tells
2977 GCC to emit warnings from system headers as if they occurred in user
2978 code. However, note that using @option{-Wall} in conjunction with this
2979 option will @emph{not} warn about unknown pragmas in system
2980 headers---for that, @option{-Wunknown-pragmas} must also be used.
2983 @opindex Wfloat-equal
2984 Warn if floating point values are used in equality comparisons.
2986 The idea behind this is that sometimes it is convenient (for the
2987 programmer) to consider floating-point values as approximations to
2988 infinitely precise real numbers. If you are doing this, then you need
2989 to compute (by analyzing the code, or in some other way) the maximum or
2990 likely maximum error that the computation introduces, and allow for it
2991 when performing comparisons (and when producing output, but that's a
2992 different problem). In particular, instead of testing for equality, you
2993 would check to see whether the two values have ranges that overlap; and
2994 this is done with the relational operators, so equality comparisons are
2997 @item -Wtraditional @r{(C only)}
2998 @opindex Wtraditional
2999 Warn about certain constructs that behave differently in traditional and
3000 ISO C@. Also warn about ISO C constructs that have no traditional C
3001 equivalent, and/or problematic constructs which should be avoided.
3005 Macro parameters that appear within string literals in the macro body.
3006 In traditional C macro replacement takes place within string literals,
3007 but does not in ISO C@.
3010 In traditional C, some preprocessor directives did not exist.
3011 Traditional preprocessors would only consider a line to be a directive
3012 if the @samp{#} appeared in column 1 on the line. Therefore
3013 @option{-Wtraditional} warns about directives that traditional C
3014 understands but would ignore because the @samp{#} does not appear as the
3015 first character on the line. It also suggests you hide directives like
3016 @samp{#pragma} not understood by traditional C by indenting them. Some
3017 traditional implementations would not recognize @samp{#elif}, so it
3018 suggests avoiding it altogether.
3021 A function-like macro that appears without arguments.
3024 The unary plus operator.
3027 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3028 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3029 constants.) Note, these suffixes appear in macros defined in the system
3030 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3031 Use of these macros in user code might normally lead to spurious
3032 warnings, however GCC's integrated preprocessor has enough context to
3033 avoid warning in these cases.
3036 A function declared external in one block and then used after the end of
3040 A @code{switch} statement has an operand of type @code{long}.
3043 A non-@code{static} function declaration follows a @code{static} one.
3044 This construct is not accepted by some traditional C compilers.
3047 The ISO type of an integer constant has a different width or
3048 signedness from its traditional type. This warning is only issued if
3049 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3050 typically represent bit patterns, are not warned about.
3053 Usage of ISO string concatenation is detected.
3056 Initialization of automatic aggregates.
3059 Identifier conflicts with labels. Traditional C lacks a separate
3060 namespace for labels.
3063 Initialization of unions. If the initializer is zero, the warning is
3064 omitted. This is done under the assumption that the zero initializer in
3065 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3066 initializer warnings and relies on default initialization to zero in the
3070 Conversions by prototypes between fixed/floating point values and vice
3071 versa. The absence of these prototypes when compiling with traditional
3072 C would cause serious problems. This is a subset of the possible
3073 conversion warnings, for the full set use @option{-Wtraditional-conversion}.
3076 Use of ISO C style function definitions. This warning intentionally is
3077 @emph{not} issued for prototype declarations or variadic functions
3078 because these ISO C features will appear in your code when using
3079 libiberty's traditional C compatibility macros, @code{PARAMS} and
3080 @code{VPARAMS}. This warning is also bypassed for nested functions
3081 because that feature is already a GCC extension and thus not relevant to
3082 traditional C compatibility.
3085 @item -Wtraditional-conversion @r{(C only)}
3086 @opindex Wtraditional-conversion
3087 Warn if a prototype causes a type conversion that is different from what
3088 would happen to the same argument in the absence of a prototype. This
3089 includes conversions of fixed point to floating and vice versa, and
3090 conversions changing the width or signedness of a fixed point argument
3091 except when the same as the default promotion.
3093 @item -Wdeclaration-after-statement @r{(C only)}
3094 @opindex Wdeclaration-after-statement
3095 Warn when a declaration is found after a statement in a block. This
3096 construct, known from C++, was introduced with ISO C99 and is by default
3097 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3098 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3102 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3104 @item -Wno-endif-labels
3105 @opindex Wno-endif-labels
3106 @opindex Wendif-labels
3107 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3111 Warn whenever a local variable shadows another local variable, parameter or
3112 global variable or whenever a built-in function is shadowed.
3114 @item -Wlarger-than-@var{len}
3115 @opindex Wlarger-than
3116 Warn whenever an object of larger than @var{len} bytes is defined.
3118 @item -Wunsafe-loop-optimizations
3119 @opindex Wunsafe-loop-optimizations
3120 Warn if the loop cannot be optimized because the compiler could not
3121 assume anything on the bounds of the loop indices. With
3122 @option{-funsafe-loop-optimizations} warn if the compiler made
3125 @item -Wpointer-arith
3126 @opindex Wpointer-arith
3127 Warn about anything that depends on the ``size of'' a function type or
3128 of @code{void}. GNU C assigns these types a size of 1, for
3129 convenience in calculations with @code{void *} pointers and pointers
3132 @item -Wbad-function-cast @r{(C only)}
3133 @opindex Wbad-function-cast
3134 Warn whenever a function call is cast to a non-matching type.
3135 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3138 Warn about ISO C constructs that are outside of the common subset of
3139 ISO C and ISO C++, e.g.@: request for implicit conversion from
3140 @code{void *} to a pointer to non-@code{void} type.
3144 Warn whenever a pointer is cast so as to remove a type qualifier from
3145 the target type. For example, warn if a @code{const char *} is cast
3146 to an ordinary @code{char *}.
3149 @opindex Wcast-align
3150 Warn whenever a pointer is cast such that the required alignment of the
3151 target is increased. For example, warn if a @code{char *} is cast to
3152 an @code{int *} on machines where integers can only be accessed at
3153 two- or four-byte boundaries.
3155 @item -Wwrite-strings
3156 @opindex Wwrite-strings
3157 When compiling C, give string constants the type @code{const
3158 char[@var{length}]} so that
3159 copying the address of one into a non-@code{const} @code{char *}
3160 pointer will get a warning; when compiling C++, warn about the
3161 deprecated conversion from string literals to @code{char *}. This
3162 warning, by default, is enabled for C++ programs.
3163 These warnings will help you find at
3164 compile time code that can try to write into a string constant, but
3165 only if you have been very careful about using @code{const} in
3166 declarations and prototypes. Otherwise, it will just be a nuisance;
3167 this is why we did not make @option{-Wall} request these warnings.
3171 Warn for variables that might be changed by @samp{longjmp} or
3172 @samp{vfork}. This warning is also enabled by @option{-Wextra}.
3175 @opindex Wconversion
3176 Warn for implicit conversions that may alter a value. This includes
3177 conversions between real and integer, like @code{abs (x)} when
3178 @code{x} is @code{double}; conversions between signed and unsigned,
3179 like @code{unsigned ui = -1}; and conversions to smaller types, like
3180 @code{sqrtf (M_PI)}. Do not warn for explicit casts like @code{abs
3181 ((int) x)} and @code{ui = (unsigned) -1}, or if the value is not
3182 changed by the conversion like in @code{abs (2.0)}.
3185 @opindex Wempty-body
3186 An empty body occurs in an @samp{if} or @samp{else} statement.
3187 This warning is also enabled by @option{-Wextra}.
3189 @item -Wsign-compare
3190 @opindex Wsign-compare
3191 @cindex warning for comparison of signed and unsigned values
3192 @cindex comparison of signed and unsigned values, warning
3193 @cindex signed and unsigned values, comparison warning
3194 Warn when a comparison between signed and unsigned values could produce
3195 an incorrect result when the signed value is converted to unsigned.
3196 This warning is also enabled by @option{-Wextra}; to get the other warnings
3197 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3199 @item -Waggregate-return
3200 @opindex Waggregate-return
3201 Warn if any functions that return structures or unions are defined or
3202 called. (In languages where you can return an array, this also elicits
3206 @opindex Walways-true
3207 Warn about comparisons which are always true such as testing if
3208 unsigned values are greater than or equal to zero. This warning is
3209 enabled by @option{-Wall}.
3211 @item -Wno-attributes
3212 @opindex Wno-attributes
3213 @opindex Wattributes
3214 Do not warn if an unexpected @code{__attribute__} is used, such as
3215 unrecognized attributes, function attributes applied to variables,
3216 etc. This will not stop errors for incorrect use of supported
3219 @item -Wstrict-prototypes @r{(C only)}
3220 @opindex Wstrict-prototypes
3221 Warn if a function is declared or defined without specifying the
3222 argument types. (An old-style function definition is permitted without
3223 a warning if preceded by a declaration which specifies the argument
3226 @item -Wold-style-declaration @r{(C only)}
3227 @opindex Wold-style-declaration
3228 Warn for obsolescent usages, according to the C Standard, in a
3229 declaration. For example, warn if storage-class specifiers like
3230 @code{static} are not the first things in a declaration. This warning
3231 is also enabled by @option{-Wextra}.
3233 @item -Wold-style-definition @r{(C only)}
3234 @opindex Wold-style-definition
3235 Warn if an old-style function definition is used. A warning is given
3236 even if there is a previous prototype.
3238 @item -Wmissing-parameter-type @r{(C only)}
3239 @opindex Wmissing-parameter-type
3240 A function parameter is declared without a type specifier in K&R-style
3247 This warning is also enabled by @option{-Wextra}.
3249 @item -Wmissing-prototypes @r{(C only)}
3250 @opindex Wmissing-prototypes
3251 Warn if a global function is defined without a previous prototype
3252 declaration. This warning is issued even if the definition itself
3253 provides a prototype. The aim is to detect global functions that fail
3254 to be declared in header files.
3256 @item -Wmissing-declarations @r{(C and C++ only)}
3257 @opindex Wmissing-declarations
3258 Warn if a global function is defined without a previous declaration.
3259 Do so even if the definition itself provides a prototype.
3260 Use this option to detect global functions that are not declared in
3261 header files. In C++, no warnings are issued for function templates,
3262 or for inline functions, or for functions in anonymous namespaces.
3264 @item -Wmissing-field-initializers
3265 @opindex Wmissing-field-initializers
3268 Warn if a structure's initializer has some fields missing. For
3269 example, the following code would cause such a warning, because
3270 @code{x.h} is implicitly zero:
3273 struct s @{ int f, g, h; @};
3274 struct s x = @{ 3, 4 @};
3277 This option does not warn about designated initializers, so the following
3278 modification would not trigger a warning:
3281 struct s @{ int f, g, h; @};
3282 struct s x = @{ .f = 3, .g = 4 @};
3285 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3286 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3288 @item -Wmissing-noreturn
3289 @opindex Wmissing-noreturn
3290 Warn about functions which might be candidates for attribute @code{noreturn}.
3291 Note these are only possible candidates, not absolute ones. Care should
3292 be taken to manually verify functions actually do not ever return before
3293 adding the @code{noreturn} attribute, otherwise subtle code generation
3294 bugs could be introduced. You will not get a warning for @code{main} in
3295 hosted C environments.
3297 @item -Wmissing-format-attribute
3298 @opindex Wmissing-format-attribute
3300 Warn about function pointers which might be candidates for @code{format}
3301 attributes. Note these are only possible candidates, not absolute ones.
3302 GCC will guess that function pointers with @code{format} attributes that
3303 are used in assignment, initialization, parameter passing or return
3304 statements should have a corresponding @code{format} attribute in the
3305 resulting type. I.e.@: the left-hand side of the assignment or
3306 initialization, the type of the parameter variable, or the return type
3307 of the containing function respectively should also have a @code{format}
3308 attribute to avoid the warning.
3310 GCC will also warn about function definitions which might be
3311 candidates for @code{format} attributes. Again, these are only
3312 possible candidates. GCC will guess that @code{format} attributes
3313 might be appropriate for any function that calls a function like
3314 @code{vprintf} or @code{vscanf}, but this might not always be the
3315 case, and some functions for which @code{format} attributes are
3316 appropriate may not be detected.
3318 @item -Wno-multichar
3319 @opindex Wno-multichar
3321 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3322 Usually they indicate a typo in the user's code, as they have
3323 implementation-defined values, and should not be used in portable code.
3325 @item -Wnormalized=<none|id|nfc|nfkc>
3326 @opindex Wnormalized
3329 @cindex character set, input normalization
3330 In ISO C and ISO C++, two identifiers are different if they are
3331 different sequences of characters. However, sometimes when characters
3332 outside the basic ASCII character set are used, you can have two
3333 different character sequences that look the same. To avoid confusion,
3334 the ISO 10646 standard sets out some @dfn{normalization rules} which
3335 when applied ensure that two sequences that look the same are turned into
3336 the same sequence. GCC can warn you if you are using identifiers which
3337 have not been normalized; this option controls that warning.
3339 There are four levels of warning that GCC supports. The default is
3340 @option{-Wnormalized=nfc}, which warns about any identifier which is
3341 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3342 recommended form for most uses.
3344 Unfortunately, there are some characters which ISO C and ISO C++ allow
3345 in identifiers that when turned into NFC aren't allowable as
3346 identifiers. That is, there's no way to use these symbols in portable
3347 ISO C or C++ and have all your identifiers in NFC.
3348 @option{-Wnormalized=id} suppresses the warning for these characters.
3349 It is hoped that future versions of the standards involved will correct
3350 this, which is why this option is not the default.
3352 You can switch the warning off for all characters by writing
3353 @option{-Wnormalized=none}. You would only want to do this if you
3354 were using some other normalization scheme (like ``D''), because
3355 otherwise you can easily create bugs that are literally impossible to see.
3357 Some characters in ISO 10646 have distinct meanings but look identical
3358 in some fonts or display methodologies, especially once formatting has
3359 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3360 LETTER N'', will display just like a regular @code{n} which has been
3361 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3362 normalization scheme to convert all these into a standard form as
3363 well, and GCC will warn if your code is not in NFKC if you use
3364 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3365 about every identifier that contains the letter O because it might be
3366 confused with the digit 0, and so is not the default, but may be
3367 useful as a local coding convention if the programming environment is
3368 unable to be fixed to display these characters distinctly.
3370 @item -Wno-deprecated-declarations
3371 @opindex Wno-deprecated-declarations
3372 Do not warn about uses of functions (@pxref{Function Attributes}),
3373 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3374 Attributes}) marked as deprecated by using the @code{deprecated}
3378 @opindex Wno-overflow
3379 Do not warn about compile-time overflow in constant expressions.
3381 @item -Woverride-init
3382 @opindex Woverride-init
3385 Warn if an initialized field without side effects is overridden when
3386 using designated initializers (@pxref{Designated Inits, , Designated
3389 This warning is included in @option{-Wextra}. To get other
3390 @option{-Wextra} warnings without this one, use @samp{-Wextra
3391 -Wno-override-init}.
3395 Warn if a structure is given the packed attribute, but the packed
3396 attribute has no effect on the layout or size of the structure.
3397 Such structures may be mis-aligned for little benefit. For
3398 instance, in this code, the variable @code{f.x} in @code{struct bar}
3399 will be misaligned even though @code{struct bar} does not itself
3400 have the packed attribute:
3407 @} __attribute__((packed));
3417 Warn if padding is included in a structure, either to align an element
3418 of the structure or to align the whole structure. Sometimes when this
3419 happens it is possible to rearrange the fields of the structure to
3420 reduce the padding and so make the structure smaller.
3422 @item -Wredundant-decls
3423 @opindex Wredundant-decls
3424 Warn if anything is declared more than once in the same scope, even in
3425 cases where multiple declaration is valid and changes nothing.
3427 @item -Wnested-externs @r{(C only)}
3428 @opindex Wnested-externs
3429 Warn if an @code{extern} declaration is encountered within a function.
3431 @item -Wunreachable-code
3432 @opindex Wunreachable-code
3433 Warn if the compiler detects that code will never be executed.
3435 This option is intended to warn when the compiler detects that at
3436 least a whole line of source code will never be executed, because
3437 some condition is never satisfied or because it is after a
3438 procedure that never returns.
3440 It is possible for this option to produce a warning even though there
3441 are circumstances under which part of the affected line can be executed,
3442 so care should be taken when removing apparently-unreachable code.
3444 For instance, when a function is inlined, a warning may mean that the
3445 line is unreachable in only one inlined copy of the function.
3447 This option is not made part of @option{-Wall} because in a debugging
3448 version of a program there is often substantial code which checks
3449 correct functioning of the program and is, hopefully, unreachable
3450 because the program does work. Another common use of unreachable
3451 code is to provide behavior which is selectable at compile-time.
3455 Warn if a function can not be inlined and it was declared as inline.
3456 Even with this option, the compiler will not warn about failures to
3457 inline functions declared in system headers.
3459 The compiler uses a variety of heuristics to determine whether or not
3460 to inline a function. For example, the compiler takes into account
3461 the size of the function being inlined and the amount of inlining
3462 that has already been done in the current function. Therefore,
3463 seemingly insignificant changes in the source program can cause the
3464 warnings produced by @option{-Winline} to appear or disappear.
3466 @item -Wno-invalid-offsetof @r{(C++ only)}
3467 @opindex Wno-invalid-offsetof
3468 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3469 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3470 to a non-POD type is undefined. In existing C++ implementations,
3471 however, @samp{offsetof} typically gives meaningful results even when
3472 applied to certain kinds of non-POD types. (Such as a simple
3473 @samp{struct} that fails to be a POD type only by virtue of having a
3474 constructor.) This flag is for users who are aware that they are
3475 writing nonportable code and who have deliberately chosen to ignore the
3478 The restrictions on @samp{offsetof} may be relaxed in a future version
3479 of the C++ standard.
3481 @item -Wno-int-to-pointer-cast @r{(C only)}
3482 @opindex Wno-int-to-pointer-cast
3483 Suppress warnings from casts to pointer type of an integer of a
3486 @item -Wno-pointer-to-int-cast @r{(C only)}
3487 @opindex Wno-pointer-to-int-cast
3488 Suppress warnings from casts from a pointer to an integer type of a
3492 @opindex Winvalid-pch
3493 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3494 the search path but can't be used.
3498 @opindex Wno-long-long
3499 Warn if @samp{long long} type is used. This is default. To inhibit
3500 the warning messages, use @option{-Wno-long-long}. Flags
3501 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3502 only when @option{-pedantic} flag is used.
3504 @item -Wvariadic-macros
3505 @opindex Wvariadic-macros
3506 @opindex Wno-variadic-macros
3507 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3508 alternate syntax when in pedantic ISO C99 mode. This is default.
3509 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3511 @item -Wvolatile-register-var
3512 @opindex Wvolatile-register-var
3513 @opindex Wno-volatile-register-var
3514 Warn if a register variable is declared volatile. The volatile
3515 modifier does not inhibit all optimizations that may eliminate reads
3516 and/or writes to register variables.
3518 @item -Wdisabled-optimization
3519 @opindex Wdisabled-optimization
3520 Warn if a requested optimization pass is disabled. This warning does
3521 not generally indicate that there is anything wrong with your code; it
3522 merely indicates that GCC's optimizers were unable to handle the code
3523 effectively. Often, the problem is that your code is too big or too
3524 complex; GCC will refuse to optimize programs when the optimization
3525 itself is likely to take inordinate amounts of time.
3527 @item -Wpointer-sign
3528 @opindex Wpointer-sign
3529 @opindex Wno-pointer-sign
3530 Warn for pointer argument passing or assignment with different signedness.
3531 This option is only supported for C and Objective-C@. It is implied by
3532 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3533 @option{-Wno-pointer-sign}.
3537 Make all warnings into errors.
3541 Make the specified warning into an errors. The specifier for a
3542 warning is appended, for example @option{-Werror=switch} turns the
3543 warnings controlled by @option{-Wswitch} into errors. This switch
3544 takes a negative form, to be used to negate @option{-Werror} for
3545 specific warnings, for example @option{-Wno-error=switch} makes
3546 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3547 is in effect. You can use the @option{-fdiagnostics-show-option}
3548 option to have each controllable warning amended with the option which
3549 controls it, to determine what to use with this option.
3551 Note that specifying @option{-Werror=}@var{foo} automatically implies
3552 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3555 @item -Wstack-protector
3556 @opindex Wstack-protector
3557 This option is only active when @option{-fstack-protector} is active. It
3558 warns about functions that will not be protected against stack smashing.
3560 @item -Wstring-literal-comparison
3561 @opindex Wstring-literal-comparison
3562 Warn about suspicious comparisons to string literal constants. In C,
3563 direct comparisons against the memory address of a string literal, such
3564 as @code{if (x == "abc")}, typically indicate a programmer error, and
3565 even when intentional, result in unspecified behavior and are not portable.
3566 Usually these warnings alert that the programmer intended to use
3567 @code{strcmp}. This warning is enabled by @option{-Wall}.
3569 @item -Woverlength-strings
3570 @opindex Woverlength-strings
3571 Warn about string constants which are longer than the ``minimum
3572 maximum'' length specified in the C standard. Modern compilers
3573 generally allow string constants which are much longer than the
3574 standard's minimum limit, but very portable programs should avoid
3575 using longer strings.
3577 The limit applies @emph{after} string constant concatenation, and does
3578 not count the trailing NUL@. In C89, the limit was 509 characters; in
3579 C99, it was raised to 4095. C++98 does not specify a normative
3580 minimum maximum, so we do not diagnose overlength strings in C++@.
3582 This option is implied by @option{-pedantic}, and can be disabled with
3583 @option{-Wno-overlength-strings}.
3586 @node Debugging Options
3587 @section Options for Debugging Your Program or GCC
3588 @cindex options, debugging
3589 @cindex debugging information options
3591 GCC has various special options that are used for debugging
3592 either your program or GCC:
3597 Produce debugging information in the operating system's native format
3598 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3601 On most systems that use stabs format, @option{-g} enables use of extra
3602 debugging information that only GDB can use; this extra information
3603 makes debugging work better in GDB but will probably make other debuggers
3605 refuse to read the program. If you want to control for certain whether
3606 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3607 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3609 GCC allows you to use @option{-g} with
3610 @option{-O}. The shortcuts taken by optimized code may occasionally
3611 produce surprising results: some variables you declared may not exist
3612 at all; flow of control may briefly move where you did not expect it;
3613 some statements may not be executed because they compute constant
3614 results or their values were already at hand; some statements may
3615 execute in different places because they were moved out of loops.
3617 Nevertheless it proves possible to debug optimized output. This makes
3618 it reasonable to use the optimizer for programs that might have bugs.
3620 The following options are useful when GCC is generated with the
3621 capability for more than one debugging format.
3625 Produce debugging information for use by GDB@. This means to use the
3626 most expressive format available (DWARF 2, stabs, or the native format
3627 if neither of those are supported), including GDB extensions if at all
3632 Produce debugging information in stabs format (if that is supported),
3633 without GDB extensions. This is the format used by DBX on most BSD
3634 systems. On MIPS, Alpha and System V Release 4 systems this option
3635 produces stabs debugging output which is not understood by DBX or SDB@.
3636 On System V Release 4 systems this option requires the GNU assembler.
3638 @item -feliminate-unused-debug-symbols
3639 @opindex feliminate-unused-debug-symbols
3640 Produce debugging information in stabs format (if that is supported),
3641 for only symbols that are actually used.
3643 @item -femit-class-debug-always
3644 Instead of emitting debugging information for a C++ class in only one
3645 object file, emit it in all object files using the class. This option
3646 should be used only with debuggers that are unable to handle the way GCC
3647 normally emits debugging information for classes because using this
3648 option will increase the size of debugging information by as much as a
3653 Produce debugging information in stabs format (if that is supported),
3654 using GNU extensions understood only by the GNU debugger (GDB)@. The
3655 use of these extensions is likely to make other debuggers crash or
3656 refuse to read the program.
3660 Produce debugging information in COFF format (if that is supported).
3661 This is the format used by SDB on most System V systems prior to
3666 Produce debugging information in XCOFF format (if that is supported).
3667 This is the format used by the DBX debugger on IBM RS/6000 systems.
3671 Produce debugging information in XCOFF format (if that is supported),
3672 using GNU extensions understood only by the GNU debugger (GDB)@. The
3673 use of these extensions is likely to make other debuggers crash or
3674 refuse to read the program, and may cause assemblers other than the GNU
3675 assembler (GAS) to fail with an error.
3679 Produce debugging information in DWARF version 2 format (if that is
3680 supported). This is the format used by DBX on IRIX 6. With this
3681 option, GCC uses features of DWARF version 3 when they are useful;
3682 version 3 is upward compatible with version 2, but may still cause
3683 problems for older debuggers.
3687 Produce debugging information in VMS debug format (if that is
3688 supported). This is the format used by DEBUG on VMS systems.
3691 @itemx -ggdb@var{level}
3692 @itemx -gstabs@var{level}
3693 @itemx -gcoff@var{level}
3694 @itemx -gxcoff@var{level}
3695 @itemx -gvms@var{level}
3696 Request debugging information and also use @var{level} to specify how
3697 much information. The default level is 2.
3699 Level 0 produces no debug information at all. Thus, @option{-g0} negates
3702 Level 1 produces minimal information, enough for making backtraces in
3703 parts of the program that you don't plan to debug. This includes
3704 descriptions of functions and external variables, but no information
3705 about local variables and no line numbers.
3707 Level 3 includes extra information, such as all the macro definitions
3708 present in the program. Some debuggers support macro expansion when
3709 you use @option{-g3}.
3711 @option{-gdwarf-2} does not accept a concatenated debug level, because
3712 GCC used to support an option @option{-gdwarf} that meant to generate
3713 debug information in version 1 of the DWARF format (which is very
3714 different from version 2), and it would have been too confusing. That
3715 debug format is long obsolete, but the option cannot be changed now.
3716 Instead use an additional @option{-g@var{level}} option to change the
3717 debug level for DWARF2.
3719 @item -feliminate-dwarf2-dups
3720 @opindex feliminate-dwarf2-dups
3721 Compress DWARF2 debugging information by eliminating duplicated
3722 information about each symbol. This option only makes sense when
3723 generating DWARF2 debugging information with @option{-gdwarf-2}.
3725 @cindex @command{prof}
3728 Generate extra code to write profile information suitable for the
3729 analysis program @command{prof}. You must use this option when compiling
3730 the source files you want data about, and you must also use it when
3733 @cindex @command{gprof}
3736 Generate extra code to write profile information suitable for the
3737 analysis program @command{gprof}. You must use this option when compiling
3738 the source files you want data about, and you must also use it when
3743 Makes the compiler print out each function name as it is compiled, and
3744 print some statistics about each pass when it finishes.
3747 @opindex ftime-report
3748 Makes the compiler print some statistics about the time consumed by each
3749 pass when it finishes.
3752 @opindex fmem-report
3753 Makes the compiler print some statistics about permanent memory
3754 allocation when it finishes.
3756 @item -fpre-ipa-mem-report
3757 @opindex fpre-ipa-mem-report
3758 @item -fpost-ipa-mem-report
3759 @opindex fpost-ipa-mem-report
3760 Makes the compiler print some statistics about permanent memory
3761 allocation before or after interprocedural optimization.
3763 @item -fprofile-arcs
3764 @opindex fprofile-arcs
3765 Add code so that program flow @dfn{arcs} are instrumented. During
3766 execution the program records how many times each branch and call is
3767 executed and how many times it is taken or returns. When the compiled
3768 program exits it saves this data to a file called
3769 @file{@var{auxname}.gcda} for each source file. The data may be used for
3770 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3771 test coverage analysis (@option{-ftest-coverage}). Each object file's
3772 @var{auxname} is generated from the name of the output file, if
3773 explicitly specified and it is not the final executable, otherwise it is
3774 the basename of the source file. In both cases any suffix is removed
3775 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3776 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3777 @xref{Cross-profiling}.
3779 @cindex @command{gcov}
3783 This option is used to compile and link code instrumented for coverage
3784 analysis. The option is a synonym for @option{-fprofile-arcs}
3785 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3786 linking). See the documentation for those options for more details.
3791 Compile the source files with @option{-fprofile-arcs} plus optimization
3792 and code generation options. For test coverage analysis, use the
3793 additional @option{-ftest-coverage} option. You do not need to profile
3794 every source file in a program.
3797 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3798 (the latter implies the former).
3801 Run the program on a representative workload to generate the arc profile
3802 information. This may be repeated any number of times. You can run
3803 concurrent instances of your program, and provided that the file system
3804 supports locking, the data files will be correctly updated. Also
3805 @code{fork} calls are detected and correctly handled (double counting
3809 For profile-directed optimizations, compile the source files again with
3810 the same optimization and code generation options plus
3811 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3812 Control Optimization}).
3815 For test coverage analysis, use @command{gcov} to produce human readable
3816 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3817 @command{gcov} documentation for further information.
3821 With @option{-fprofile-arcs}, for each function of your program GCC
3822 creates a program flow graph, then finds a spanning tree for the graph.
3823 Only arcs that are not on the spanning tree have to be instrumented: the
3824 compiler adds code to count the number of times that these arcs are
3825 executed. When an arc is the only exit or only entrance to a block, the
3826 instrumentation code can be added to the block; otherwise, a new basic
3827 block must be created to hold the instrumentation code.
3830 @item -ftest-coverage
3831 @opindex ftest-coverage
3832 Produce a notes file that the @command{gcov} code-coverage utility
3833 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3834 show program coverage. Each source file's note file is called
3835 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3836 above for a description of @var{auxname} and instructions on how to
3837 generate test coverage data. Coverage data will match the source files
3838 more closely, if you do not optimize.
3840 @item -d@var{letters}
3841 @item -fdump-rtl-@var{pass}
3843 Says to make debugging dumps during compilation at times specified by
3844 @var{letters}. This is used for debugging the RTL-based passes of the
3845 compiler. The file names for most of the dumps are made by appending a
3846 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3847 from the name of the output file, if explicitly specified and it is not
3848 an executable, otherwise it is the basename of the source file.
3850 Most debug dumps can be enabled either passing a letter to the @option{-d}
3851 option, or with a long @option{-fdump-rtl} switch; here are the possible
3852 letters for use in @var{letters} and @var{pass}, and their meanings:
3857 Annotate the assembler output with miscellaneous debugging information.
3860 @itemx -fdump-rtl-bbro
3862 @opindex fdump-rtl-bbro
3863 Dump after block reordering, to @file{@var{file}.148r.bbro}.
3866 @itemx -fdump-rtl-combine
3868 @opindex fdump-rtl-combine
3869 Dump after instruction combination, to the file @file{@var{file}.129r.combine}.
3872 @itemx -fdump-rtl-ce1
3873 @itemx -fdump-rtl-ce2
3875 @opindex fdump-rtl-ce1
3876 @opindex fdump-rtl-ce2
3877 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3878 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
3879 and @option{-fdump-rtl-ce2} enable dumping after the second if
3880 conversion, to the file @file{@var{file}.130r.ce2}.
3883 @itemx -fdump-rtl-btl
3884 @itemx -fdump-rtl-dbr
3886 @opindex fdump-rtl-btl
3887 @opindex fdump-rtl-dbr
3888 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3889 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3890 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3891 scheduling, to @file{@var{file}.36.dbr}.
3895 Dump all macro definitions, at the end of preprocessing, in addition to
3899 @itemx -fdump-rtl-ce3
3901 @opindex fdump-rtl-ce3
3902 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
3905 @itemx -fdump-rtl-cfg
3906 @itemx -fdump-rtl-life
3908 @opindex fdump-rtl-cfg
3909 @opindex fdump-rtl-life
3910 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3911 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
3912 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3913 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
3916 @itemx -fdump-rtl-greg
3918 @opindex fdump-rtl-greg
3919 Dump after global register allocation, to @file{@var{file}.139r.greg}.
3922 @itemx -fdump-rtl-gcse
3923 @itemx -fdump-rtl-bypass
3925 @opindex fdump-rtl-gcse
3926 @opindex fdump-rtl-bypass
3927 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3928 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3929 enable dumping after jump bypassing and control flow optimizations, to
3930 @file{@var{file}.115r.bypass}.
3933 @itemx -fdump-rtl-eh
3935 @opindex fdump-rtl-eh
3936 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3939 @itemx -fdump-rtl-sibling
3941 @opindex fdump-rtl-sibling
3942 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
3945 @itemx -fdump-rtl-jump
3947 @opindex fdump-rtl-jump
3948 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
3951 @itemx -fdump-rtl-stack
3953 @opindex fdump-rtl-stack
3954 Dump after conversion from registers to stack, to @file{@var{file}.152r.stack}.
3957 @itemx -fdump-rtl-lreg
3959 @opindex fdump-rtl-lreg
3960 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
3963 @itemx -fdump-rtl-loop2
3965 @opindex fdump-rtl-loop2
3966 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
3967 loop optimization pass, to @file{@var{file}.119r.loop2},
3968 @file{@var{file}.120r.loop2_init},
3969 @file{@var{file}.121r.loop2_invariant}, and
3970 @file{@var{file}.125r.loop2_done}.
3973 @itemx -fdump-rtl-sms
3975 @opindex fdump-rtl-sms
3976 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
3979 @itemx -fdump-rtl-mach
3981 @opindex fdump-rtl-mach
3982 Dump after performing the machine dependent reorganization pass, to
3983 @file{@var{file}.155r.mach}.
3986 @itemx -fdump-rtl-rnreg
3988 @opindex fdump-rtl-rnreg
3989 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
3992 @itemx -fdump-rtl-regmove
3994 @opindex fdump-rtl-regmove
3995 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
3998 @itemx -fdump-rtl-postreload
4000 @opindex fdump-rtl-postreload
4001 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
4004 @itemx -fdump-rtl-expand
4006 @opindex fdump-rtl-expand
4007 Dump after RTL generation, to @file{@var{file}.104r.expand}.
4010 @itemx -fdump-rtl-sched2
4012 @opindex fdump-rtl-sched2
4013 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
4016 @itemx -fdump-rtl-cse
4018 @opindex fdump-rtl-cse
4019 Dump after CSE (including the jump optimization that sometimes follows
4020 CSE), to @file{@var{file}.113r.cse}.
4023 @itemx -fdump-rtl-sched1
4025 @opindex fdump-rtl-sched1
4026 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
4029 @itemx -fdump-rtl-cse2
4031 @opindex fdump-rtl-cse2
4032 Dump after the second CSE pass (including the jump optimization that
4033 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
4036 @itemx -fdump-rtl-tracer
4038 @opindex fdump-rtl-tracer
4039 Dump after running tracer, to @file{@var{file}.118r.tracer}.
4042 @itemx -fdump-rtl-vpt
4043 @itemx -fdump-rtl-vartrack
4045 @opindex fdump-rtl-vpt
4046 @opindex fdump-rtl-vartrack
4047 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
4048 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
4049 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
4050 to @file{@var{file}.154r.vartrack}.
4053 @itemx -fdump-rtl-flow2
4055 @opindex fdump-rtl-flow2
4056 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4059 @itemx -fdump-rtl-peephole2
4061 @opindex fdump-rtl-peephole2
4062 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4065 @itemx -fdump-rtl-web
4067 @opindex fdump-rtl-web
4068 Dump after live range splitting, to @file{@var{file}.126r.web}.
4071 @itemx -fdump-rtl-all
4073 @opindex fdump-rtl-all
4074 Produce all the dumps listed above.
4078 Produce a core dump whenever an error occurs.
4082 Print statistics on memory usage, at the end of the run, to
4087 Annotate the assembler output with a comment indicating which
4088 pattern and alternative was used. The length of each instruction is
4093 Dump the RTL in the assembler output as a comment before each instruction.
4094 Also turns on @option{-dp} annotation.
4098 For each of the other indicated dump files (either with @option{-d} or
4099 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4100 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4104 Just generate RTL for a function instead of compiling it. Usually used
4105 with @samp{r} (@option{-fdump-rtl-expand}).
4109 Dump debugging information during parsing, to standard error.
4113 @opindex fdump-noaddr
4114 When doing debugging dumps (see @option{-d} option above), suppress
4115 address output. This makes it more feasible to use diff on debugging
4116 dumps for compiler invocations with different compiler binaries and/or
4117 different text / bss / data / heap / stack / dso start locations.
4119 @item -fdump-unnumbered
4120 @opindex fdump-unnumbered
4121 When doing debugging dumps (see @option{-d} option above), suppress instruction
4122 numbers, line number note and address output. This makes it more feasible to
4123 use diff on debugging dumps for compiler invocations with different
4124 options, in particular with and without @option{-g}.
4126 @item -fdump-translation-unit @r{(C++ only)}
4127 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4128 @opindex fdump-translation-unit
4129 Dump a representation of the tree structure for the entire translation
4130 unit to a file. The file name is made by appending @file{.tu} to the
4131 source file name. If the @samp{-@var{options}} form is used, @var{options}
4132 controls the details of the dump as described for the
4133 @option{-fdump-tree} options.
4135 @item -fdump-class-hierarchy @r{(C++ only)}
4136 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4137 @opindex fdump-class-hierarchy
4138 Dump a representation of each class's hierarchy and virtual function
4139 table layout to a file. The file name is made by appending @file{.class}
4140 to the source file name. If the @samp{-@var{options}} form is used,
4141 @var{options} controls the details of the dump as described for the
4142 @option{-fdump-tree} options.
4144 @item -fdump-ipa-@var{switch}
4146 Control the dumping at various stages of inter-procedural analysis
4147 language tree to a file. The file name is generated by appending a switch
4148 specific suffix to the source file name. The following dumps are possible:
4152 Enables all inter-procedural analysis dumps; currently the only produced
4153 dump is the @samp{cgraph} dump.
4156 Dumps information about call-graph optimization, unused function removal,
4157 and inlining decisions.
4160 @item -fdump-tree-@var{switch}
4161 @itemx -fdump-tree-@var{switch}-@var{options}
4163 Control the dumping at various stages of processing the intermediate
4164 language tree to a file. The file name is generated by appending a switch
4165 specific suffix to the source file name. If the @samp{-@var{options}}
4166 form is used, @var{options} is a list of @samp{-} separated options that
4167 control the details of the dump. Not all options are applicable to all
4168 dumps, those which are not meaningful will be ignored. The following
4169 options are available
4173 Print the address of each node. Usually this is not meaningful as it
4174 changes according to the environment and source file. Its primary use
4175 is for tying up a dump file with a debug environment.
4177 Inhibit dumping of members of a scope or body of a function merely
4178 because that scope has been reached. Only dump such items when they
4179 are directly reachable by some other path. When dumping pretty-printed
4180 trees, this option inhibits dumping the bodies of control structures.
4182 Print a raw representation of the tree. By default, trees are
4183 pretty-printed into a C-like representation.
4185 Enable more detailed dumps (not honored by every dump option).
4187 Enable dumping various statistics about the pass (not honored by every dump
4190 Enable showing basic block boundaries (disabled in raw dumps).
4192 Enable showing virtual operands for every statement.
4194 Enable showing line numbers for statements.
4196 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4198 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4201 The following tree dumps are possible:
4205 Dump before any tree based optimization, to @file{@var{file}.original}.
4208 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4211 Dump after function inlining, to @file{@var{file}.inlined}.
4214 @opindex fdump-tree-gimple
4215 Dump each function before and after the gimplification pass to a file. The
4216 file name is made by appending @file{.gimple} to the source file name.
4219 @opindex fdump-tree-cfg
4220 Dump the control flow graph of each function to a file. The file name is
4221 made by appending @file{.cfg} to the source file name.
4224 @opindex fdump-tree-vcg
4225 Dump the control flow graph of each function to a file in VCG format. The
4226 file name is made by appending @file{.vcg} to the source file name. Note
4227 that if the file contains more than one function, the generated file cannot
4228 be used directly by VCG@. You will need to cut and paste each function's
4229 graph into its own separate file first.
4232 @opindex fdump-tree-ch
4233 Dump each function after copying loop headers. The file name is made by
4234 appending @file{.ch} to the source file name.
4237 @opindex fdump-tree-ssa
4238 Dump SSA related information to a file. The file name is made by appending
4239 @file{.ssa} to the source file name.
4242 @opindex fdump-tree-salias
4243 Dump structure aliasing variable information to a file. This file name
4244 is made by appending @file{.salias} to the source file name.
4247 @opindex fdump-tree-alias
4248 Dump aliasing information for each function. The file name is made by
4249 appending @file{.alias} to the source file name.
4252 @opindex fdump-tree-ccp
4253 Dump each function after CCP@. The file name is made by appending
4254 @file{.ccp} to the source file name.
4257 @opindex fdump-tree-storeccp
4258 Dump each function after STORE-CCP. The file name is made by appending
4259 @file{.storeccp} to the source file name.
4262 @opindex fdump-tree-pre
4263 Dump trees after partial redundancy elimination. The file name is made
4264 by appending @file{.pre} to the source file name.
4267 @opindex fdump-tree-fre
4268 Dump trees after full redundancy elimination. The file name is made
4269 by appending @file{.fre} to the source file name.
4272 @opindex fdump-tree-copyprop
4273 Dump trees after copy propagation. The file name is made
4274 by appending @file{.copyprop} to the source file name.
4276 @item store_copyprop
4277 @opindex fdump-tree-store_copyprop
4278 Dump trees after store copy-propagation. The file name is made
4279 by appending @file{.store_copyprop} to the source file name.
4282 @opindex fdump-tree-dce
4283 Dump each function after dead code elimination. The file name is made by
4284 appending @file{.dce} to the source file name.
4287 @opindex fdump-tree-mudflap
4288 Dump each function after adding mudflap instrumentation. The file name is
4289 made by appending @file{.mudflap} to the source file name.
4292 @opindex fdump-tree-sra
4293 Dump each function after performing scalar replacement of aggregates. The
4294 file name is made by appending @file{.sra} to the source file name.
4297 @opindex fdump-tree-sink
4298 Dump each function after performing code sinking. The file name is made
4299 by appending @file{.sink} to the source file name.
4302 @opindex fdump-tree-dom
4303 Dump each function after applying dominator tree optimizations. The file
4304 name is made by appending @file{.dom} to the source file name.
4307 @opindex fdump-tree-dse
4308 Dump each function after applying dead store elimination. The file
4309 name is made by appending @file{.dse} to the source file name.
4312 @opindex fdump-tree-phiopt
4313 Dump each function after optimizing PHI nodes into straightline code. The file
4314 name is made by appending @file{.phiopt} to the source file name.
4317 @opindex fdump-tree-forwprop
4318 Dump each function after forward propagating single use variables. The file
4319 name is made by appending @file{.forwprop} to the source file name.
4322 @opindex fdump-tree-copyrename
4323 Dump each function after applying the copy rename optimization. The file
4324 name is made by appending @file{.copyrename} to the source file name.
4327 @opindex fdump-tree-nrv
4328 Dump each function after applying the named return value optimization on
4329 generic trees. The file name is made by appending @file{.nrv} to the source
4333 @opindex fdump-tree-vect
4334 Dump each function after applying vectorization of loops. The file name is
4335 made by appending @file{.vect} to the source file name.
4338 @opindex fdump-tree-vrp
4339 Dump each function after Value Range Propagation (VRP). The file name
4340 is made by appending @file{.vrp} to the source file name.
4343 @opindex fdump-tree-all
4344 Enable all the available tree dumps with the flags provided in this option.
4347 @item -ftree-vectorizer-verbose=@var{n}
4348 @opindex ftree-vectorizer-verbose
4349 This option controls the amount of debugging output the vectorizer prints.
4350 This information is written to standard error, unless
4351 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4352 in which case it is output to the usual dump listing file, @file{.vect}.
4353 For @var{n}=0 no diagnostic information is reported.
4354 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4355 and the total number of loops that got vectorized.
4356 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4357 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4358 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4359 level that @option{-fdump-tree-vect-stats} uses.
4360 Higher verbosity levels mean either more information dumped for each
4361 reported loop, or same amount of information reported for more loops:
4362 If @var{n}=3, alignment related information is added to the reports.
4363 If @var{n}=4, data-references related information (e.g. memory dependences,
4364 memory access-patterns) is added to the reports.
4365 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4366 that did not pass the first analysis phase (i.e. may not be countable, or
4367 may have complicated control-flow).
4368 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4369 For @var{n}=7, all the information the vectorizer generates during its
4370 analysis and transformation is reported. This is the same verbosity level
4371 that @option{-fdump-tree-vect-details} uses.
4373 @item -frandom-seed=@var{string}
4374 @opindex frandom-string
4375 This option provides a seed that GCC uses when it would otherwise use
4376 random numbers. It is used to generate certain symbol names
4377 that have to be different in every compiled file. It is also used to
4378 place unique stamps in coverage data files and the object files that
4379 produce them. You can use the @option{-frandom-seed} option to produce
4380 reproducibly identical object files.
4382 The @var{string} should be different for every file you compile.
4384 @item -fsched-verbose=@var{n}
4385 @opindex fsched-verbose
4386 On targets that use instruction scheduling, this option controls the
4387 amount of debugging output the scheduler prints. This information is
4388 written to standard error, unless @option{-dS} or @option{-dR} is
4389 specified, in which case it is output to the usual dump
4390 listing file, @file{.sched} or @file{.sched2} respectively. However
4391 for @var{n} greater than nine, the output is always printed to standard
4394 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4395 same information as @option{-dRS}. For @var{n} greater than one, it
4396 also output basic block probabilities, detailed ready list information
4397 and unit/insn info. For @var{n} greater than two, it includes RTL
4398 at abort point, control-flow and regions info. And for @var{n} over
4399 four, @option{-fsched-verbose} also includes dependence info.
4403 Store the usual ``temporary'' intermediate files permanently; place them
4404 in the current directory and name them based on the source file. Thus,
4405 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4406 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4407 preprocessed @file{foo.i} output file even though the compiler now
4408 normally uses an integrated preprocessor.
4410 When used in combination with the @option{-x} command line option,
4411 @option{-save-temps} is sensible enough to avoid over writing an
4412 input source file with the same extension as an intermediate file.
4413 The corresponding intermediate file may be obtained by renaming the
4414 source file before using @option{-save-temps}.
4418 Report the CPU time taken by each subprocess in the compilation
4419 sequence. For C source files, this is the compiler proper and assembler
4420 (plus the linker if linking is done). The output looks like this:
4427 The first number on each line is the ``user time'', that is time spent
4428 executing the program itself. The second number is ``system time'',
4429 time spent executing operating system routines on behalf of the program.
4430 Both numbers are in seconds.
4432 @item -fvar-tracking
4433 @opindex fvar-tracking
4434 Run variable tracking pass. It computes where variables are stored at each
4435 position in code. Better debugging information is then generated
4436 (if the debugging information format supports this information).
4438 It is enabled by default when compiling with optimization (@option{-Os},
4439 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4440 the debug info format supports it.
4442 @item -print-file-name=@var{library}
4443 @opindex print-file-name
4444 Print the full absolute name of the library file @var{library} that
4445 would be used when linking---and don't do anything else. With this
4446 option, GCC does not compile or link anything; it just prints the
4449 @item -print-multi-directory
4450 @opindex print-multi-directory
4451 Print the directory name corresponding to the multilib selected by any
4452 other switches present in the command line. This directory is supposed
4453 to exist in @env{GCC_EXEC_PREFIX}.
4455 @item -print-multi-lib
4456 @opindex print-multi-lib
4457 Print the mapping from multilib directory names to compiler switches
4458 that enable them. The directory name is separated from the switches by
4459 @samp{;}, and each switch starts with an @samp{@@} instead of the
4460 @samp{-}, without spaces between multiple switches. This is supposed to
4461 ease shell-processing.
4463 @item -print-prog-name=@var{program}
4464 @opindex print-prog-name
4465 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4467 @item -print-libgcc-file-name
4468 @opindex print-libgcc-file-name
4469 Same as @option{-print-file-name=libgcc.a}.
4471 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4472 but you do want to link with @file{libgcc.a}. You can do
4475 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4478 @item -print-search-dirs
4479 @opindex print-search-dirs
4480 Print the name of the configured installation directory and a list of
4481 program and library directories @command{gcc} will search---and don't do anything else.
4483 This is useful when @command{gcc} prints the error message
4484 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4485 To resolve this you either need to put @file{cpp0} and the other compiler
4486 components where @command{gcc} expects to find them, or you can set the environment
4487 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4488 Don't forget the trailing @samp{/}.
4489 @xref{Environment Variables}.
4492 @opindex dumpmachine
4493 Print the compiler's target machine (for example,
4494 @samp{i686-pc-linux-gnu})---and don't do anything else.
4497 @opindex dumpversion
4498 Print the compiler version (for example, @samp{3.0})---and don't do
4503 Print the compiler's built-in specs---and don't do anything else. (This
4504 is used when GCC itself is being built.) @xref{Spec Files}.
4506 @item -feliminate-unused-debug-types
4507 @opindex feliminate-unused-debug-types
4508 Normally, when producing DWARF2 output, GCC will emit debugging
4509 information for all types declared in a compilation
4510 unit, regardless of whether or not they are actually used
4511 in that compilation unit. Sometimes this is useful, such as
4512 if, in the debugger, you want to cast a value to a type that is
4513 not actually used in your program (but is declared). More often,
4514 however, this results in a significant amount of wasted space.
4515 With this option, GCC will avoid producing debug symbol output
4516 for types that are nowhere used in the source file being compiled.
4519 @node Optimize Options
4520 @section Options That Control Optimization
4521 @cindex optimize options
4522 @cindex options, optimization
4524 These options control various sorts of optimizations.
4526 Without any optimization option, the compiler's goal is to reduce the
4527 cost of compilation and to make debugging produce the expected
4528 results. Statements are independent: if you stop the program with a
4529 breakpoint between statements, you can then assign a new value to any
4530 variable or change the program counter to any other statement in the
4531 function and get exactly the results you would expect from the source
4534 Turning on optimization flags makes the compiler attempt to improve
4535 the performance and/or code size at the expense of compilation time
4536 and possibly the ability to debug the program.
4538 The compiler performs optimization based on the knowledge it has of
4539 the program. Optimization levels @option{-O} and above, in
4540 particular, enable @emph{unit-at-a-time} mode, which allows the
4541 compiler to consider information gained from later functions in
4542 the file when compiling a function. Compiling multiple files at
4543 once to a single output file in @emph{unit-at-a-time} mode allows
4544 the compiler to use information gained from all of the files when
4545 compiling each of them.
4547 Not all optimizations are controlled directly by a flag. Only
4548 optimizations that have a flag are listed.
4555 Optimize. Optimizing compilation takes somewhat more time, and a lot
4556 more memory for a large function.
4558 With @option{-O}, the compiler tries to reduce code size and execution
4559 time, without performing any optimizations that take a great deal of
4562 @option{-O} turns on the following optimization flags:
4563 @gccoptlist{-fdefer-pop @gol
4564 -fdelayed-branch @gol
4565 -fguess-branch-probability @gol
4566 -fcprop-registers @gol
4567 -fif-conversion @gol
4568 -fif-conversion2 @gol
4571 -ftree-dominator-opts @gol
4576 -ftree-copyrename @gol
4579 -funit-at-a-time @gol
4582 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4583 where doing so does not interfere with debugging.
4587 Optimize even more. GCC performs nearly all supported optimizations
4588 that do not involve a space-speed tradeoff. The compiler does not
4589 perform loop unrolling or function inlining when you specify @option{-O2}.
4590 As compared to @option{-O}, this option increases both compilation time
4591 and the performance of the generated code.
4593 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4594 also turns on the following optimization flags:
4595 @gccoptlist{-fthread-jumps @gol
4597 -foptimize-sibling-calls @gol
4598 -fcse-follow-jumps -fcse-skip-blocks @gol
4599 -fgcse -fgcse-lm @gol
4600 -fexpensive-optimizations @gol
4601 -frerun-cse-after-loop @gol
4604 -fschedule-insns -fschedule-insns2 @gol
4605 -fsched-interblock -fsched-spec @gol
4607 -fstrict-aliasing @gol
4608 -fdelete-null-pointer-checks @gol
4609 -freorder-blocks -freorder-functions @gol
4610 -falign-functions -falign-jumps @gol
4611 -falign-loops -falign-labels @gol
4615 Please note the warning under @option{-fgcse} about
4616 invoking @option{-O2} on programs that use computed gotos.
4620 Optimize yet more. @option{-O3} turns on all optimizations specified by
4621 @option{-O2} and also turns on the @option{-finline-functions},
4622 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4626 Reduce compilation time and make debugging produce the expected
4627 results. This is the default.
4631 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4632 do not typically increase code size. It also performs further
4633 optimizations designed to reduce code size.
4635 @option{-Os} disables the following optimization flags:
4636 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4637 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4638 -fprefetch-loop-arrays -ftree-vect-loop-version}
4640 If you use multiple @option{-O} options, with or without level numbers,
4641 the last such option is the one that is effective.
4644 Options of the form @option{-f@var{flag}} specify machine-independent
4645 flags. Most flags have both positive and negative forms; the negative
4646 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4647 below, only one of the forms is listed---the one you typically will
4648 use. You can figure out the other form by either removing @samp{no-}
4651 The following options control specific optimizations. They are either
4652 activated by @option{-O} options or are related to ones that are. You
4653 can use the following flags in the rare cases when ``fine-tuning'' of
4654 optimizations to be performed is desired.
4657 @item -fno-default-inline
4658 @opindex fno-default-inline
4659 Do not make member functions inline by default merely because they are
4660 defined inside the class scope (C++ only). Otherwise, when you specify
4661 @w{@option{-O}}, member functions defined inside class scope are compiled
4662 inline by default; i.e., you don't need to add @samp{inline} in front of
4663 the member function name.
4665 @item -fno-defer-pop
4666 @opindex fno-defer-pop
4667 Always pop the arguments to each function call as soon as that function
4668 returns. For machines which must pop arguments after a function call,
4669 the compiler normally lets arguments accumulate on the stack for several
4670 function calls and pops them all at once.
4672 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4675 @opindex fforce-addr
4676 Force memory address constants to be copied into registers before
4677 doing arithmetic on them.
4679 @item -fforward-propagate
4680 @opindex fforward-propagate
4681 Perform a forward propagation pass on RTL. The pass tries to combine two
4682 instructions and checks if the result can be simplified. If loop unrolling
4683 is active, two passes are performed and the second is scheduled after
4686 This option is enabled by default at optimization levels @option{-O2},
4687 @option{-O3}, @option{-Os}.
4689 @item -fomit-frame-pointer
4690 @opindex fomit-frame-pointer
4691 Don't keep the frame pointer in a register for functions that
4692 don't need one. This avoids the instructions to save, set up and
4693 restore frame pointers; it also makes an extra register available
4694 in many functions. @strong{It also makes debugging impossible on
4697 On some machines, such as the VAX, this flag has no effect, because
4698 the standard calling sequence automatically handles the frame pointer
4699 and nothing is saved by pretending it doesn't exist. The
4700 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4701 whether a target machine supports this flag. @xref{Registers,,Register
4702 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4704 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4706 @item -foptimize-sibling-calls
4707 @opindex foptimize-sibling-calls
4708 Optimize sibling and tail recursive calls.
4710 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4714 Don't pay attention to the @code{inline} keyword. Normally this option
4715 is used to keep the compiler from expanding any functions inline.
4716 Note that if you are not optimizing, no functions can be expanded inline.
4718 @item -finline-functions
4719 @opindex finline-functions
4720 Integrate all simple functions into their callers. The compiler
4721 heuristically decides which functions are simple enough to be worth
4722 integrating in this way.
4724 If all calls to a given function are integrated, and the function is
4725 declared @code{static}, then the function is normally not output as
4726 assembler code in its own right.
4728 Enabled at level @option{-O3}.
4730 @item -finline-functions-called-once
4731 @opindex finline-functions-called-once
4732 Consider all @code{static} functions called once for inlining into their
4733 caller even if they are not marked @code{inline}. If a call to a given
4734 function is integrated, then the function is not output as assembler code
4737 Enabled if @option{-funit-at-a-time} is enabled.
4739 @item -fearly-inlining
4740 @opindex fearly-inlining
4741 Inline functions marked by @code{always_inline} and functions whose body seems
4742 smaller than the function call overhead early before doing
4743 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4744 makes profiling significantly cheaper and usually inlining faster on programs
4745 having large chains of nested wrapper functions.
4749 @item -finline-limit=@var{n}
4750 @opindex finline-limit
4751 By default, GCC limits the size of functions that can be inlined. This flag
4752 allows the control of this limit for functions that are explicitly marked as
4753 inline (i.e., marked with the inline keyword or defined within the class
4754 definition in c++). @var{n} is the size of functions that can be inlined in
4755 number of pseudo instructions (not counting parameter handling). The default
4756 value of @var{n} is 600.
4757 Increasing this value can result in more inlined code at
4758 the cost of compilation time and memory consumption. Decreasing usually makes
4759 the compilation faster and less code will be inlined (which presumably
4760 means slower programs). This option is particularly useful for programs that
4761 use inlining heavily such as those based on recursive templates with C++.
4763 Inlining is actually controlled by a number of parameters, which may be
4764 specified individually by using @option{--param @var{name}=@var{value}}.
4765 The @option{-finline-limit=@var{n}} option sets some of these parameters
4769 @item max-inline-insns-single
4770 is set to @var{n}/2.
4771 @item max-inline-insns-auto
4772 is set to @var{n}/2.
4773 @item min-inline-insns
4774 is set to 130 or @var{n}/4, whichever is smaller.
4775 @item max-inline-insns-rtl
4779 See below for a documentation of the individual
4780 parameters controlling inlining.
4782 @emph{Note:} pseudo instruction represents, in this particular context, an
4783 abstract measurement of function's size. In no way does it represent a count
4784 of assembly instructions and as such its exact meaning might change from one
4785 release to an another.
4787 @item -fkeep-inline-functions
4788 @opindex fkeep-inline-functions
4789 In C, emit @code{static} functions that are declared @code{inline}
4790 into the object file, even if the function has been inlined into all
4791 of its callers. This switch does not affect functions using the
4792 @code{extern inline} extension in GNU C89@. In C++, emit any and all
4793 inline functions into the object file.
4795 @item -fkeep-static-consts
4796 @opindex fkeep-static-consts
4797 Emit variables declared @code{static const} when optimization isn't turned
4798 on, even if the variables aren't referenced.
4800 GCC enables this option by default. If you want to force the compiler to
4801 check if the variable was referenced, regardless of whether or not
4802 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4804 @item -fmerge-constants
4805 Attempt to merge identical constants (string constants and floating point
4806 constants) across compilation units.
4808 This option is the default for optimized compilation if the assembler and
4809 linker support it. Use @option{-fno-merge-constants} to inhibit this
4812 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4814 @item -fmerge-all-constants
4815 Attempt to merge identical constants and identical variables.
4817 This option implies @option{-fmerge-constants}. In addition to
4818 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4819 arrays or initialized constant variables with integral or floating point
4820 types. Languages like C or C++ require each non-automatic variable to
4821 have distinct location, so using this option will result in non-conforming
4824 @item -fmodulo-sched
4825 @opindex fmodulo-sched
4826 Perform swing modulo scheduling immediately before the first scheduling
4827 pass. This pass looks at innermost loops and reorders their
4828 instructions by overlapping different iterations.
4830 @item -fno-branch-count-reg
4831 @opindex fno-branch-count-reg
4832 Do not use ``decrement and branch'' instructions on a count register,
4833 but instead generate a sequence of instructions that decrement a
4834 register, compare it against zero, then branch based upon the result.
4835 This option is only meaningful on architectures that support such
4836 instructions, which include x86, PowerPC, IA-64 and S/390.
4838 The default is @option{-fbranch-count-reg}.
4840 @item -fno-function-cse
4841 @opindex fno-function-cse
4842 Do not put function addresses in registers; make each instruction that
4843 calls a constant function contain the function's address explicitly.
4845 This option results in less efficient code, but some strange hacks
4846 that alter the assembler output may be confused by the optimizations
4847 performed when this option is not used.
4849 The default is @option{-ffunction-cse}
4851 @item -fno-zero-initialized-in-bss
4852 @opindex fno-zero-initialized-in-bss
4853 If the target supports a BSS section, GCC by default puts variables that
4854 are initialized to zero into BSS@. This can save space in the resulting
4857 This option turns off this behavior because some programs explicitly
4858 rely on variables going to the data section. E.g., so that the
4859 resulting executable can find the beginning of that section and/or make
4860 assumptions based on that.
4862 The default is @option{-fzero-initialized-in-bss}.
4864 @item -fbounds-check
4865 @opindex fbounds-check
4866 For front-ends that support it, generate additional code to check that
4867 indices used to access arrays are within the declared range. This is
4868 currently only supported by the Java and Fortran front-ends, where
4869 this option defaults to true and false respectively.
4871 @item -fmudflap -fmudflapth -fmudflapir
4875 @cindex bounds checking
4877 For front-ends that support it (C and C++), instrument all risky
4878 pointer/array dereferencing operations, some standard library
4879 string/heap functions, and some other associated constructs with
4880 range/validity tests. Modules so instrumented should be immune to
4881 buffer overflows, invalid heap use, and some other classes of C/C++
4882 programming errors. The instrumentation relies on a separate runtime
4883 library (@file{libmudflap}), which will be linked into a program if
4884 @option{-fmudflap} is given at link time. Run-time behavior of the
4885 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4886 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4889 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4890 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4891 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4892 instrumentation should ignore pointer reads. This produces less
4893 instrumentation (and therefore faster execution) and still provides
4894 some protection against outright memory corrupting writes, but allows
4895 erroneously read data to propagate within a program.
4897 @item -fthread-jumps
4898 @opindex fthread-jumps
4899 Perform optimizations where we check to see if a jump branches to a
4900 location where another comparison subsumed by the first is found. If
4901 so, the first branch is redirected to either the destination of the
4902 second branch or a point immediately following it, depending on whether
4903 the condition is known to be true or false.
4905 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4907 @item -fcse-follow-jumps
4908 @opindex fcse-follow-jumps
4909 In common subexpression elimination, scan through jump instructions
4910 when the target of the jump is not reached by any other path. For
4911 example, when CSE encounters an @code{if} statement with an
4912 @code{else} clause, CSE will follow the jump when the condition
4915 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4917 @item -fcse-skip-blocks
4918 @opindex fcse-skip-blocks
4919 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4920 follow jumps which conditionally skip over blocks. When CSE
4921 encounters a simple @code{if} statement with no else clause,
4922 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4923 body of the @code{if}.
4925 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4927 @item -frerun-cse-after-loop
4928 @opindex frerun-cse-after-loop
4929 Re-run common subexpression elimination after loop optimizations has been
4932 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4936 Perform a global common subexpression elimination pass.
4937 This pass also performs global constant and copy propagation.
4939 @emph{Note:} When compiling a program using computed gotos, a GCC
4940 extension, you may get better runtime performance if you disable
4941 the global common subexpression elimination pass by adding
4942 @option{-fno-gcse} to the command line.
4944 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4948 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4949 attempt to move loads which are only killed by stores into themselves. This
4950 allows a loop containing a load/store sequence to be changed to a load outside
4951 the loop, and a copy/store within the loop.
4953 Enabled by default when gcse is enabled.
4957 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4958 global common subexpression elimination. This pass will attempt to move
4959 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4960 loops containing a load/store sequence can be changed to a load before
4961 the loop and a store after the loop.
4963 Not enabled at any optimization level.
4967 When @option{-fgcse-las} is enabled, the global common subexpression
4968 elimination pass eliminates redundant loads that come after stores to the
4969 same memory location (both partial and full redundancies).
4971 Not enabled at any optimization level.
4973 @item -fgcse-after-reload
4974 @opindex fgcse-after-reload
4975 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4976 pass is performed after reload. The purpose of this pass is to cleanup
4979 @item -funsafe-loop-optimizations
4980 @opindex funsafe-loop-optimizations
4981 If given, the loop optimizer will assume that loop indices do not
4982 overflow, and that the loops with nontrivial exit condition are not
4983 infinite. This enables a wider range of loop optimizations even if
4984 the loop optimizer itself cannot prove that these assumptions are valid.
4985 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4986 if it finds this kind of loop.
4988 @item -fcrossjumping
4989 @opindex crossjumping
4990 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4991 resulting code may or may not perform better than without cross-jumping.
4993 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4995 @item -fif-conversion
4996 @opindex if-conversion
4997 Attempt to transform conditional jumps into branch-less equivalents. This
4998 include use of conditional moves, min, max, set flags and abs instructions, and
4999 some tricks doable by standard arithmetics. The use of conditional execution
5000 on chips where it is available is controlled by @code{if-conversion2}.
5002 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5004 @item -fif-conversion2
5005 @opindex if-conversion2
5006 Use conditional execution (where available) to transform conditional jumps into
5007 branch-less equivalents.
5009 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5011 @item -fdelete-null-pointer-checks
5012 @opindex fdelete-null-pointer-checks
5013 Use global dataflow analysis to identify and eliminate useless checks
5014 for null pointers. The compiler assumes that dereferencing a null
5015 pointer would have halted the program. If a pointer is checked after
5016 it has already been dereferenced, it cannot be null.
5018 In some environments, this assumption is not true, and programs can
5019 safely dereference null pointers. Use
5020 @option{-fno-delete-null-pointer-checks} to disable this optimization
5021 for programs which depend on that behavior.
5023 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5025 @item -fexpensive-optimizations
5026 @opindex fexpensive-optimizations
5027 Perform a number of minor optimizations that are relatively expensive.
5029 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5031 @item -foptimize-register-move
5033 @opindex foptimize-register-move
5035 Attempt to reassign register numbers in move instructions and as
5036 operands of other simple instructions in order to maximize the amount of
5037 register tying. This is especially helpful on machines with two-operand
5040 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
5043 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5045 @item -fdelayed-branch
5046 @opindex fdelayed-branch
5047 If supported for the target machine, attempt to reorder instructions
5048 to exploit instruction slots available after delayed branch
5051 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5053 @item -fschedule-insns
5054 @opindex fschedule-insns
5055 If supported for the target machine, attempt to reorder instructions to
5056 eliminate execution stalls due to required data being unavailable. This
5057 helps machines that have slow floating point or memory load instructions
5058 by allowing other instructions to be issued until the result of the load
5059 or floating point instruction is required.
5061 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5063 @item -fschedule-insns2
5064 @opindex fschedule-insns2
5065 Similar to @option{-fschedule-insns}, but requests an additional pass of
5066 instruction scheduling after register allocation has been done. This is
5067 especially useful on machines with a relatively small number of
5068 registers and where memory load instructions take more than one cycle.
5070 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5072 @item -fno-sched-interblock
5073 @opindex fno-sched-interblock
5074 Don't schedule instructions across basic blocks. This is normally
5075 enabled by default when scheduling before register allocation, i.e.@:
5076 with @option{-fschedule-insns} or at @option{-O2} or higher.
5078 @item -fno-sched-spec
5079 @opindex fno-sched-spec
5080 Don't allow speculative motion of non-load instructions. This is normally
5081 enabled by default when scheduling before register allocation, i.e.@:
5082 with @option{-fschedule-insns} or at @option{-O2} or higher.
5084 @item -fsched-spec-load
5085 @opindex fsched-spec-load
5086 Allow speculative motion of some load instructions. This only makes
5087 sense when scheduling before register allocation, i.e.@: with
5088 @option{-fschedule-insns} or at @option{-O2} or higher.
5090 @item -fsched-spec-load-dangerous
5091 @opindex fsched-spec-load-dangerous
5092 Allow speculative motion of more load instructions. This only makes
5093 sense when scheduling before register allocation, i.e.@: with
5094 @option{-fschedule-insns} or at @option{-O2} or higher.
5096 @item -fsched-stalled-insns=@var{n}
5097 @opindex fsched-stalled-insns
5098 Define how many insns (if any) can be moved prematurely from the queue
5099 of stalled insns into the ready list, during the second scheduling pass.
5101 @item -fsched-stalled-insns-dep=@var{n}
5102 @opindex fsched-stalled-insns-dep
5103 Define how many insn groups (cycles) will be examined for a dependency
5104 on a stalled insn that is candidate for premature removal from the queue
5105 of stalled insns. Has an effect only during the second scheduling pass,
5106 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5108 @item -fsched2-use-superblocks
5109 @opindex fsched2-use-superblocks
5110 When scheduling after register allocation, do use superblock scheduling
5111 algorithm. Superblock scheduling allows motion across basic block boundaries
5112 resulting on faster schedules. This option is experimental, as not all machine
5113 descriptions used by GCC model the CPU closely enough to avoid unreliable
5114 results from the algorithm.
5116 This only makes sense when scheduling after register allocation, i.e.@: with
5117 @option{-fschedule-insns2} or at @option{-O2} or higher.
5119 @item -fsched2-use-traces
5120 @opindex fsched2-use-traces
5121 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5122 allocation and additionally perform code duplication in order to increase the
5123 size of superblocks using tracer pass. See @option{-ftracer} for details on
5126 This mode should produce faster but significantly longer programs. Also
5127 without @option{-fbranch-probabilities} the traces constructed may not
5128 match the reality and hurt the performance. This only makes
5129 sense when scheduling after register allocation, i.e.@: with
5130 @option{-fschedule-insns2} or at @option{-O2} or higher.
5134 Eliminates redundant extension instructions and move the non redundant
5135 ones to optimal placement using LCM.
5137 @item -freschedule-modulo-scheduled-loops
5138 @opindex fscheduling-in-modulo-scheduled-loops
5139 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5140 we may want to prevent the later scheduling passes from changing its schedule, we use this
5141 option to control that.
5143 @item -fcaller-saves
5144 @opindex fcaller-saves
5145 Enable values to be allocated in registers that will be clobbered by
5146 function calls, by emitting extra instructions to save and restore the
5147 registers around such calls. Such allocation is done only when it
5148 seems to result in better code than would otherwise be produced.
5150 This option is always enabled by default on certain machines, usually
5151 those which have no call-preserved registers to use instead.
5153 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5156 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5157 enabled by default at @option{-O2} and @option{-O3}.
5160 Perform Full Redundancy Elimination (FRE) on trees. The difference
5161 between FRE and PRE is that FRE only considers expressions
5162 that are computed on all paths leading to the redundant computation.
5163 This analysis faster than PRE, though it exposes fewer redundancies.
5164 This flag is enabled by default at @option{-O} and higher.
5166 @item -ftree-copy-prop
5167 Perform copy propagation on trees. This pass eliminates unnecessary
5168 copy operations. This flag is enabled by default at @option{-O} and
5171 @item -ftree-store-copy-prop
5172 Perform copy propagation of memory loads and stores. This pass
5173 eliminates unnecessary copy operations in memory references
5174 (structures, global variables, arrays, etc). This flag is enabled by
5175 default at @option{-O2} and higher.
5178 Perform structural alias analysis on trees. This flag
5179 is enabled by default at @option{-O} and higher.
5182 Perform interprocedural pointer analysis.
5185 Perform forward store motion on trees. This flag is
5186 enabled by default at @option{-O} and higher.
5189 Perform sparse conditional constant propagation (CCP) on trees. This
5190 pass only operates on local scalar variables and is enabled by default
5191 at @option{-O} and higher.
5193 @item -ftree-store-ccp
5194 Perform sparse conditional constant propagation (CCP) on trees. This
5195 pass operates on both local scalar variables and memory stores and
5196 loads (global variables, structures, arrays, etc). This flag is
5197 enabled by default at @option{-O2} and higher.
5200 Perform dead code elimination (DCE) on trees. This flag is enabled by
5201 default at @option{-O} and higher.
5203 @item -ftree-dominator-opts
5204 Perform a variety of simple scalar cleanups (constant/copy
5205 propagation, redundancy elimination, range propagation and expression
5206 simplification) based on a dominator tree traversal. This also
5207 performs jump threading (to reduce jumps to jumps). This flag is
5208 enabled by default at @option{-O} and higher.
5211 Perform loop header copying on trees. This is beneficial since it increases
5212 effectiveness of code motion optimizations. It also saves one jump. This flag
5213 is enabled by default at @option{-O} and higher. It is not enabled
5214 for @option{-Os}, since it usually increases code size.
5216 @item -ftree-loop-optimize
5217 Perform loop optimizations on trees. This flag is enabled by default
5218 at @option{-O} and higher.
5220 @item -ftree-loop-linear
5221 Perform linear loop transformations on tree. This flag can improve cache
5222 performance and allow further loop optimizations to take place.
5224 @item -ftree-loop-im
5225 Perform loop invariant motion on trees. This pass moves only invariants that
5226 would be hard to handle at RTL level (function calls, operations that expand to
5227 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5228 operands of conditions that are invariant out of the loop, so that we can use
5229 just trivial invariantness analysis in loop unswitching. The pass also includes
5232 @item -ftree-loop-ivcanon
5233 Create a canonical counter for number of iterations in the loop for that
5234 determining number of iterations requires complicated analysis. Later
5235 optimizations then may determine the number easily. Useful especially
5236 in connection with unrolling.
5239 Perform induction variable optimizations (strength reduction, induction
5240 variable merging and induction variable elimination) on trees.
5243 Perform scalar replacement of aggregates. This pass replaces structure
5244 references with scalars to prevent committing structures to memory too
5245 early. This flag is enabled by default at @option{-O} and higher.
5247 @item -ftree-copyrename
5248 Perform copy renaming on trees. This pass attempts to rename compiler
5249 temporaries to other variables at copy locations, usually resulting in
5250 variable names which more closely resemble the original variables. This flag
5251 is enabled by default at @option{-O} and higher.
5254 Perform temporary expression replacement during the SSA->normal phase. Single
5255 use/single def temporaries are replaced at their use location with their
5256 defining expression. This results in non-GIMPLE code, but gives the expanders
5257 much more complex trees to work on resulting in better RTL generation. This is
5258 enabled by default at @option{-O} and higher.
5261 Perform live range splitting during the SSA->normal phase. Distinct live
5262 ranges of a variable are split into unique variables, allowing for better
5263 optimization later. This is enabled by default at @option{-O} and higher.
5265 @item -ftree-vectorize
5266 Perform loop vectorization on trees.
5268 @item -ftree-vect-loop-version
5269 @opindex ftree-vect-loop-version
5270 Perform loop versioning when doing loop vectorization on trees. When a loop
5271 appears to be vectorizable except that data alignment or data dependence cannot
5272 be determined at compile time then vectorized and non-vectorized versions of
5273 the loop are generated along with runtime checks for alignment or dependence
5274 to control which version is executed. This option is enabled by default
5275 except at level @option{-Os} where it is disabled.
5278 Perform Value Range Propagation on trees. This is similar to the
5279 constant propagation pass, but instead of values, ranges of values are
5280 propagated. This allows the optimizers to remove unnecessary range
5281 checks like array bound checks and null pointer checks. This is
5282 enabled by default at @option{-O2} and higher. Null pointer check
5283 elimination is only done if @option{-fdelete-null-pointer-checks} is
5288 Perform tail duplication to enlarge superblock size. This transformation
5289 simplifies the control flow of the function allowing other optimizations to do
5292 @item -funroll-loops
5293 @opindex funroll-loops
5294 Unroll loops whose number of iterations can be determined at compile
5295 time or upon entry to the loop. @option{-funroll-loops} implies
5296 @option{-frerun-cse-after-loop}. This option makes code larger,
5297 and may or may not make it run faster.
5299 @item -funroll-all-loops
5300 @opindex funroll-all-loops
5301 Unroll all loops, even if their number of iterations is uncertain when
5302 the loop is entered. This usually makes programs run more slowly.
5303 @option{-funroll-all-loops} implies the same options as
5304 @option{-funroll-loops},
5306 @item -fsplit-ivs-in-unroller
5307 @opindex -fsplit-ivs-in-unroller
5308 Enables expressing of values of induction variables in later iterations
5309 of the unrolled loop using the value in the first iteration. This breaks
5310 long dependency chains, thus improving efficiency of the scheduling passes.
5312 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5313 same effect. However in cases the loop body is more complicated than
5314 a single basic block, this is not reliable. It also does not work at all
5315 on some of the architectures due to restrictions in the CSE pass.
5317 This optimization is enabled by default.
5319 @item -fvariable-expansion-in-unroller
5320 @opindex -fvariable-expansion-in-unroller
5321 With this option, the compiler will create multiple copies of some
5322 local variables when unrolling a loop which can result in superior code.
5324 @item -fprefetch-loop-arrays
5325 @opindex fprefetch-loop-arrays
5326 If supported by the target machine, generate instructions to prefetch
5327 memory to improve the performance of loops that access large arrays.
5329 This option may generate better or worse code; results are highly
5330 dependent on the structure of loops within the source code.
5332 Disabled at level @option{-Os}.
5335 @itemx -fno-peephole2
5336 @opindex fno-peephole
5337 @opindex fno-peephole2
5338 Disable any machine-specific peephole optimizations. The difference
5339 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5340 are implemented in the compiler; some targets use one, some use the
5341 other, a few use both.
5343 @option{-fpeephole} is enabled by default.
5344 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5346 @item -fno-guess-branch-probability
5347 @opindex fno-guess-branch-probability
5348 Do not guess branch probabilities using heuristics.
5350 GCC will use heuristics to guess branch probabilities if they are
5351 not provided by profiling feedback (@option{-fprofile-arcs}). These
5352 heuristics are based on the control flow graph. If some branch probabilities
5353 are specified by @samp{__builtin_expect}, then the heuristics will be
5354 used to guess branch probabilities for the rest of the control flow graph,
5355 taking the @samp{__builtin_expect} info into account. The interactions
5356 between the heuristics and @samp{__builtin_expect} can be complex, and in
5357 some cases, it may be useful to disable the heuristics so that the effects
5358 of @samp{__builtin_expect} are easier to understand.
5360 The default is @option{-fguess-branch-probability} at levels
5361 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5363 @item -freorder-blocks
5364 @opindex freorder-blocks
5365 Reorder basic blocks in the compiled function in order to reduce number of
5366 taken branches and improve code locality.
5368 Enabled at levels @option{-O2}, @option{-O3}.
5370 @item -freorder-blocks-and-partition
5371 @opindex freorder-blocks-and-partition
5372 In addition to reordering basic blocks in the compiled function, in order
5373 to reduce number of taken branches, partitions hot and cold basic blocks
5374 into separate sections of the assembly and .o files, to improve
5375 paging and cache locality performance.
5377 This optimization is automatically turned off in the presence of
5378 exception handling, for linkonce sections, for functions with a user-defined
5379 section attribute and on any architecture that does not support named
5382 @item -freorder-functions
5383 @opindex freorder-functions
5384 Reorder functions in the object file in order to
5385 improve code locality. This is implemented by using special
5386 subsections @code{.text.hot} for most frequently executed functions and
5387 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5388 the linker so object file format must support named sections and linker must
5389 place them in a reasonable way.
5391 Also profile feedback must be available in to make this option effective. See
5392 @option{-fprofile-arcs} for details.
5394 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5396 @item -fstrict-aliasing
5397 @opindex fstrict-aliasing
5398 Allows the compiler to assume the strictest aliasing rules applicable to
5399 the language being compiled. For C (and C++), this activates
5400 optimizations based on the type of expressions. In particular, an
5401 object of one type is assumed never to reside at the same address as an
5402 object of a different type, unless the types are almost the same. For
5403 example, an @code{unsigned int} can alias an @code{int}, but not a
5404 @code{void*} or a @code{double}. A character type may alias any other
5407 Pay special attention to code like this:
5420 The practice of reading from a different union member than the one most
5421 recently written to (called ``type-punning'') is common. Even with
5422 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5423 is accessed through the union type. So, the code above will work as
5424 expected. However, this code might not:
5435 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5437 @item -falign-functions
5438 @itemx -falign-functions=@var{n}
5439 @opindex falign-functions
5440 Align the start of functions to the next power-of-two greater than
5441 @var{n}, skipping up to @var{n} bytes. For instance,
5442 @option{-falign-functions=32} aligns functions to the next 32-byte
5443 boundary, but @option{-falign-functions=24} would align to the next
5444 32-byte boundary only if this can be done by skipping 23 bytes or less.
5446 @option{-fno-align-functions} and @option{-falign-functions=1} are
5447 equivalent and mean that functions will not be aligned.
5449 Some assemblers only support this flag when @var{n} is a power of two;
5450 in that case, it is rounded up.
5452 If @var{n} is not specified or is zero, use a machine-dependent default.
5454 Enabled at levels @option{-O2}, @option{-O3}.
5456 @item -falign-labels
5457 @itemx -falign-labels=@var{n}
5458 @opindex falign-labels
5459 Align all branch targets to a power-of-two boundary, skipping up to
5460 @var{n} bytes like @option{-falign-functions}. This option can easily
5461 make code slower, because it must insert dummy operations for when the
5462 branch target is reached in the usual flow of the code.
5464 @option{-fno-align-labels} and @option{-falign-labels=1} are
5465 equivalent and mean that labels will not be aligned.
5467 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5468 are greater than this value, then their values are used instead.
5470 If @var{n} is not specified or is zero, use a machine-dependent default
5471 which is very likely to be @samp{1}, meaning no alignment.
5473 Enabled at levels @option{-O2}, @option{-O3}.
5476 @itemx -falign-loops=@var{n}
5477 @opindex falign-loops
5478 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5479 like @option{-falign-functions}. The hope is that the loop will be
5480 executed many times, which will make up for any execution of the dummy
5483 @option{-fno-align-loops} and @option{-falign-loops=1} are
5484 equivalent and mean that loops will not be aligned.
5486 If @var{n} is not specified or is zero, use a machine-dependent default.
5488 Enabled at levels @option{-O2}, @option{-O3}.
5491 @itemx -falign-jumps=@var{n}
5492 @opindex falign-jumps
5493 Align branch targets to a power-of-two boundary, for branch targets
5494 where the targets can only be reached by jumping, skipping up to @var{n}
5495 bytes like @option{-falign-functions}. In this case, no dummy operations
5498 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5499 equivalent and mean that loops will not be aligned.
5501 If @var{n} is not specified or is zero, use a machine-dependent default.
5503 Enabled at levels @option{-O2}, @option{-O3}.
5505 @item -funit-at-a-time
5506 @opindex funit-at-a-time
5507 Parse the whole compilation unit before starting to produce code.
5508 This allows some extra optimizations to take place but consumes
5509 more memory (in general). There are some compatibility issues
5510 with @emph{unit-at-a-time} mode:
5513 enabling @emph{unit-at-a-time} mode may change the order
5514 in which functions, variables, and top-level @code{asm} statements
5515 are emitted, and will likely break code relying on some particular
5516 ordering. The majority of such top-level @code{asm} statements,
5517 though, can be replaced by @code{section} attributes. The
5518 @option{fno-toplevel-reorder} option may be used to keep the ordering
5519 used in the input file, at the cost of some optimizations.
5522 @emph{unit-at-a-time} mode removes unreferenced static variables
5523 and functions. This may result in undefined references
5524 when an @code{asm} statement refers directly to variables or functions
5525 that are otherwise unused. In that case either the variable/function
5526 shall be listed as an operand of the @code{asm} statement operand or,
5527 in the case of top-level @code{asm} statements the attribute @code{used}
5528 shall be used on the declaration.
5531 Static functions now can use non-standard passing conventions that
5532 may break @code{asm} statements calling functions directly. Again,
5533 attribute @code{used} will prevent this behavior.
5536 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5537 but this scheme may not be supported by future releases of GCC@.
5539 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5541 @item -fno-toplevel-reorder
5542 Do not reorder top-level functions, variables, and @code{asm}
5543 statements. Output them in the same order that they appear in the
5544 input file. When this option is used, unreferenced static variables
5545 will not be removed. This option is intended to support existing code
5546 which relies on a particular ordering. For new code, it is better to
5551 Constructs webs as commonly used for register allocation purposes and assign
5552 each web individual pseudo register. This allows the register allocation pass
5553 to operate on pseudos directly, but also strengthens several other optimization
5554 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5555 however, make debugging impossible, since variables will no longer stay in a
5558 Enabled by default with @option{-funroll-loops}.
5560 @item -fwhole-program
5561 @opindex fwhole-program
5562 Assume that the current compilation unit represents whole program being
5563 compiled. All public functions and variables with the exception of @code{main}
5564 and those merged by attribute @code{externally_visible} become static functions
5565 and in a affect gets more aggressively optimized by interprocedural optimizers.
5566 While this option is equivalent to proper use of @code{static} keyword for
5567 programs consisting of single file, in combination with option
5568 @option{--combine} this flag can be used to compile most of smaller scale C
5569 programs since the functions and variables become local for the whole combined
5570 compilation unit, not for the single source file itself.
5573 @item -fno-cprop-registers
5574 @opindex fno-cprop-registers
5575 After register allocation and post-register allocation instruction splitting,
5576 we perform a copy-propagation pass to try to reduce scheduling dependencies
5577 and occasionally eliminate the copy.
5579 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5581 @item -fprofile-generate
5582 @opindex fprofile-generate
5584 Enable options usually used for instrumenting application to produce
5585 profile useful for later recompilation with profile feedback based
5586 optimization. You must use @option{-fprofile-generate} both when
5587 compiling and when linking your program.
5589 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5592 @opindex fprofile-use
5593 Enable profile feedback directed optimizations, and optimizations
5594 generally profitable only with profile feedback available.
5596 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5597 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5601 The following options control compiler behavior regarding floating
5602 point arithmetic. These options trade off between speed and
5603 correctness. All must be specifically enabled.
5607 @opindex ffloat-store
5608 Do not store floating point variables in registers, and inhibit other
5609 options that might change whether a floating point value is taken from a
5612 @cindex floating point precision
5613 This option prevents undesirable excess precision on machines such as
5614 the 68000 where the floating registers (of the 68881) keep more
5615 precision than a @code{double} is supposed to have. Similarly for the
5616 x86 architecture. For most programs, the excess precision does only
5617 good, but a few programs rely on the precise definition of IEEE floating
5618 point. Use @option{-ffloat-store} for such programs, after modifying
5619 them to store all pertinent intermediate computations into variables.
5623 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5624 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5625 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5626 and @option{fcx-limited-range}.
5628 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5630 This option should never be turned on by any @option{-O} option since
5631 it can result in incorrect output for programs which depend on
5632 an exact implementation of IEEE or ISO rules/specifications for
5635 @item -fno-math-errno
5636 @opindex fno-math-errno
5637 Do not set ERRNO after calling math functions that are executed
5638 with a single instruction, e.g., sqrt. A program that relies on
5639 IEEE exceptions for math error handling may want to use this flag
5640 for speed while maintaining IEEE arithmetic compatibility.
5642 This option should never be turned on by any @option{-O} option since
5643 it can result in incorrect output for programs which depend on
5644 an exact implementation of IEEE or ISO rules/specifications for
5647 The default is @option{-fmath-errno}.
5649 On Darwin systems, the math library never sets @code{errno}. There is therefore
5650 no reason for the compiler to consider the possibility that it might,
5651 and @option{-fno-math-errno} is the default.
5653 @item -funsafe-math-optimizations
5654 @opindex funsafe-math-optimizations
5655 Allow optimizations for floating-point arithmetic that (a) assume
5656 that arguments and results are valid and (b) may violate IEEE or
5657 ANSI standards. When used at link-time, it may include libraries
5658 or startup files that change the default FPU control word or other
5659 similar optimizations.
5661 This option should never be turned on by any @option{-O} option since
5662 it can result in incorrect output for programs which depend on
5663 an exact implementation of IEEE or ISO rules/specifications for
5666 The default is @option{-fno-unsafe-math-optimizations}.
5668 @item -ffinite-math-only
5669 @opindex ffinite-math-only
5670 Allow optimizations for floating-point arithmetic that assume
5671 that arguments and results are not NaNs or +-Infs.
5673 This option should never be turned on by any @option{-O} option since
5674 it can result in incorrect output for programs which depend on
5675 an exact implementation of IEEE or ISO rules/specifications.
5677 The default is @option{-fno-finite-math-only}.
5679 @item -fno-trapping-math
5680 @opindex fno-trapping-math
5681 Compile code assuming that floating-point operations cannot generate
5682 user-visible traps. These traps include division by zero, overflow,
5683 underflow, inexact result and invalid operation. This option implies
5684 @option{-fno-signaling-nans}. Setting this option may allow faster
5685 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5687 This option should never be turned on by any @option{-O} option since
5688 it can result in incorrect output for programs which depend on
5689 an exact implementation of IEEE or ISO rules/specifications for
5692 The default is @option{-ftrapping-math}.
5694 @item -frounding-math
5695 @opindex frounding-math
5696 Disable transformations and optimizations that assume default floating
5697 point rounding behavior. This is round-to-zero for all floating point
5698 to integer conversions, and round-to-nearest for all other arithmetic
5699 truncations. This option should be specified for programs that change
5700 the FP rounding mode dynamically, or that may be executed with a
5701 non-default rounding mode. This option disables constant folding of
5702 floating point expressions at compile-time (which may be affected by
5703 rounding mode) and arithmetic transformations that are unsafe in the
5704 presence of sign-dependent rounding modes.
5706 The default is @option{-fno-rounding-math}.
5708 This option is experimental and does not currently guarantee to
5709 disable all GCC optimizations that are affected by rounding mode.
5710 Future versions of GCC may provide finer control of this setting
5711 using C99's @code{FENV_ACCESS} pragma. This command line option
5712 will be used to specify the default state for @code{FENV_ACCESS}.
5714 @item -frtl-abstract-sequences
5715 @opindex frtl-abstract-sequences
5716 It is a size optimization method. This option is to find identical
5717 sequences of code, which can be turned into pseudo-procedures and
5718 then replace all occurrences with calls to the newly created
5719 subroutine. It is kind of an opposite of @option{-finline-functions}.
5720 This optimization runs at RTL level.
5722 @item -fsignaling-nans
5723 @opindex fsignaling-nans
5724 Compile code assuming that IEEE signaling NaNs may generate user-visible
5725 traps during floating-point operations. Setting this option disables
5726 optimizations that may change the number of exceptions visible with
5727 signaling NaNs. This option implies @option{-ftrapping-math}.
5729 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5732 The default is @option{-fno-signaling-nans}.
5734 This option is experimental and does not currently guarantee to
5735 disable all GCC optimizations that affect signaling NaN behavior.
5737 @item -fsingle-precision-constant
5738 @opindex fsingle-precision-constant
5739 Treat floating point constant as single precision constant instead of
5740 implicitly converting it to double precision constant.
5742 @item -fcx-limited-range
5743 @itemx -fno-cx-limited-range
5744 @opindex fcx-limited-range
5745 @opindex fno-cx-limited-range
5746 When enabled, this option states that a range reduction step is not
5747 needed when performing complex division. The default is
5748 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5750 This option controls the default setting of the ISO C99
5751 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5756 The following options control optimizations that may improve
5757 performance, but are not enabled by any @option{-O} options. This
5758 section includes experimental options that may produce broken code.
5761 @item -fbranch-probabilities
5762 @opindex fbranch-probabilities
5763 After running a program compiled with @option{-fprofile-arcs}
5764 (@pxref{Debugging Options,, Options for Debugging Your Program or
5765 @command{gcc}}), you can compile it a second time using
5766 @option{-fbranch-probabilities}, to improve optimizations based on
5767 the number of times each branch was taken. When the program
5768 compiled with @option{-fprofile-arcs} exits it saves arc execution
5769 counts to a file called @file{@var{sourcename}.gcda} for each source
5770 file. The information in this data file is very dependent on the
5771 structure of the generated code, so you must use the same source code
5772 and the same optimization options for both compilations.
5774 With @option{-fbranch-probabilities}, GCC puts a
5775 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5776 These can be used to improve optimization. Currently, they are only
5777 used in one place: in @file{reorg.c}, instead of guessing which path a
5778 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5779 exactly determine which path is taken more often.
5781 @item -fprofile-values
5782 @opindex fprofile-values
5783 If combined with @option{-fprofile-arcs}, it adds code so that some
5784 data about values of expressions in the program is gathered.
5786 With @option{-fbranch-probabilities}, it reads back the data gathered
5787 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5788 notes to instructions for their later usage in optimizations.
5790 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5794 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5795 a code to gather information about values of expressions.
5797 With @option{-fbranch-probabilities}, it reads back the data gathered
5798 and actually performs the optimizations based on them.
5799 Currently the optimizations include specialization of division operation
5800 using the knowledge about the value of the denominator.
5802 @item -frename-registers
5803 @opindex frename-registers
5804 Attempt to avoid false dependencies in scheduled code by making use
5805 of registers left over after register allocation. This optimization
5806 will most benefit processors with lots of registers. Depending on the
5807 debug information format adopted by the target, however, it can
5808 make debugging impossible, since variables will no longer stay in
5809 a ``home register''.
5811 Enabled by default with @option{-funroll-loops}.
5815 Perform tail duplication to enlarge superblock size. This transformation
5816 simplifies the control flow of the function allowing other optimizations to do
5819 Enabled with @option{-fprofile-use}.
5821 @item -funroll-loops
5822 @opindex funroll-loops
5823 Unroll loops whose number of iterations can be determined at compile time or
5824 upon entry to the loop. @option{-funroll-loops} implies
5825 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5826 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5827 small constant number of iterations). This option makes code larger, and may
5828 or may not make it run faster.
5830 Enabled with @option{-fprofile-use}.
5832 @item -funroll-all-loops
5833 @opindex funroll-all-loops
5834 Unroll all loops, even if their number of iterations is uncertain when
5835 the loop is entered. This usually makes programs run more slowly.
5836 @option{-funroll-all-loops} implies the same options as
5837 @option{-funroll-loops}.
5840 @opindex fpeel-loops
5841 Peels the loops for that there is enough information that they do not
5842 roll much (from profile feedback). It also turns on complete loop peeling
5843 (i.e.@: complete removal of loops with small constant number of iterations).
5845 Enabled with @option{-fprofile-use}.
5847 @item -fmove-loop-invariants
5848 @opindex fmove-loop-invariants
5849 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
5850 at level @option{-O1}
5852 @item -funswitch-loops
5853 @opindex funswitch-loops
5854 Move branches with loop invariant conditions out of the loop, with duplicates
5855 of the loop on both branches (modified according to result of the condition).
5857 @item -ffunction-sections
5858 @itemx -fdata-sections
5859 @opindex ffunction-sections
5860 @opindex fdata-sections
5861 Place each function or data item into its own section in the output
5862 file if the target supports arbitrary sections. The name of the
5863 function or the name of the data item determines the section's name
5866 Use these options on systems where the linker can perform optimizations
5867 to improve locality of reference in the instruction space. Most systems
5868 using the ELF object format and SPARC processors running Solaris 2 have
5869 linkers with such optimizations. AIX may have these optimizations in
5872 Only use these options when there are significant benefits from doing
5873 so. When you specify these options, the assembler and linker will
5874 create larger object and executable files and will also be slower.
5875 You will not be able to use @code{gprof} on all systems if you
5876 specify this option and you may have problems with debugging if
5877 you specify both this option and @option{-g}.
5879 @item -fbranch-target-load-optimize
5880 @opindex fbranch-target-load-optimize
5881 Perform branch target register load optimization before prologue / epilogue
5883 The use of target registers can typically be exposed only during reload,
5884 thus hoisting loads out of loops and doing inter-block scheduling needs
5885 a separate optimization pass.
5887 @item -fbranch-target-load-optimize2
5888 @opindex fbranch-target-load-optimize2
5889 Perform branch target register load optimization after prologue / epilogue
5892 @item -fbtr-bb-exclusive
5893 @opindex fbtr-bb-exclusive
5894 When performing branch target register load optimization, don't reuse
5895 branch target registers in within any basic block.
5897 @item -fstack-protector
5898 Emit extra code to check for buffer overflows, such as stack smashing
5899 attacks. This is done by adding a guard variable to functions with
5900 vulnerable objects. This includes functions that call alloca, and
5901 functions with buffers larger than 8 bytes. The guards are initialized
5902 when a function is entered and then checked when the function exits.
5903 If a guard check fails, an error message is printed and the program exits.
5905 @item -fstack-protector-all
5906 Like @option{-fstack-protector} except that all functions are protected.
5908 @item -fsection-anchors
5909 @opindex fsection-anchors
5910 Try to reduce the number of symbolic address calculations by using
5911 shared ``anchor'' symbols to address nearby objects. This transformation
5912 can help to reduce the number of GOT entries and GOT accesses on some
5915 For example, the implementation of the following function @code{foo}:
5919 int foo (void) @{ return a + b + c; @}
5922 would usually calculate the addresses of all three variables, but if you
5923 compile it with @option{-fsection-anchors}, it will access the variables
5924 from a common anchor point instead. The effect is similar to the
5925 following pseudocode (which isn't valid C):
5930 register int *xr = &x;
5931 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5935 Not all targets support this option.
5937 @item --param @var{name}=@var{value}
5939 In some places, GCC uses various constants to control the amount of
5940 optimization that is done. For example, GCC will not inline functions
5941 that contain more that a certain number of instructions. You can
5942 control some of these constants on the command-line using the
5943 @option{--param} option.
5945 The names of specific parameters, and the meaning of the values, are
5946 tied to the internals of the compiler, and are subject to change
5947 without notice in future releases.
5949 In each case, the @var{value} is an integer. The allowable choices for
5950 @var{name} are given in the following table:
5953 @item salias-max-implicit-fields
5954 The maximum number of fields in a variable without direct
5955 structure accesses for which structure aliasing will consider trying
5956 to track each field. The default is 5
5958 @item salias-max-array-elements
5959 The maximum number of elements an array can have and its elements
5960 still be tracked individually by structure aliasing. The default is 4
5962 @item sra-max-structure-size
5963 The maximum structure size, in bytes, at which the scalar replacement
5964 of aggregates (SRA) optimization will perform block copies. The
5965 default value, 0, implies that GCC will select the most appropriate
5968 @item sra-field-structure-ratio
5969 The threshold ratio (as a percentage) between instantiated fields and
5970 the complete structure size. We say that if the ratio of the number
5971 of bytes in instantiated fields to the number of bytes in the complete
5972 structure exceeds this parameter, then block copies are not used. The
5975 @item max-crossjump-edges
5976 The maximum number of incoming edges to consider for crossjumping.
5977 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5978 the number of edges incoming to each block. Increasing values mean
5979 more aggressive optimization, making the compile time increase with
5980 probably small improvement in executable size.
5982 @item min-crossjump-insns
5983 The minimum number of instructions which must be matched at the end
5984 of two blocks before crossjumping will be performed on them. This
5985 value is ignored in the case where all instructions in the block being
5986 crossjumped from are matched. The default value is 5.
5988 @item max-grow-copy-bb-insns
5989 The maximum code size expansion factor when copying basic blocks
5990 instead of jumping. The expansion is relative to a jump instruction.
5991 The default value is 8.
5993 @item max-goto-duplication-insns
5994 The maximum number of instructions to duplicate to a block that jumps
5995 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5996 passes, GCC factors computed gotos early in the compilation process,
5997 and unfactors them as late as possible. Only computed jumps at the
5998 end of a basic blocks with no more than max-goto-duplication-insns are
5999 unfactored. The default value is 8.
6001 @item max-delay-slot-insn-search
6002 The maximum number of instructions to consider when looking for an
6003 instruction to fill a delay slot. If more than this arbitrary number of
6004 instructions is searched, the time savings from filling the delay slot
6005 will be minimal so stop searching. Increasing values mean more
6006 aggressive optimization, making the compile time increase with probably
6007 small improvement in executable run time.
6009 @item max-delay-slot-live-search
6010 When trying to fill delay slots, the maximum number of instructions to
6011 consider when searching for a block with valid live register
6012 information. Increasing this arbitrarily chosen value means more
6013 aggressive optimization, increasing the compile time. This parameter
6014 should be removed when the delay slot code is rewritten to maintain the
6017 @item max-gcse-memory
6018 The approximate maximum amount of memory that will be allocated in
6019 order to perform the global common subexpression elimination
6020 optimization. If more memory than specified is required, the
6021 optimization will not be done.
6023 @item max-gcse-passes
6024 The maximum number of passes of GCSE to run. The default is 1.
6026 @item max-pending-list-length
6027 The maximum number of pending dependencies scheduling will allow
6028 before flushing the current state and starting over. Large functions
6029 with few branches or calls can create excessively large lists which
6030 needlessly consume memory and resources.
6032 @item max-inline-insns-single
6033 Several parameters control the tree inliner used in gcc.
6034 This number sets the maximum number of instructions (counted in GCC's
6035 internal representation) in a single function that the tree inliner
6036 will consider for inlining. This only affects functions declared
6037 inline and methods implemented in a class declaration (C++).
6038 The default value is 450.
6040 @item max-inline-insns-auto
6041 When you use @option{-finline-functions} (included in @option{-O3}),
6042 a lot of functions that would otherwise not be considered for inlining
6043 by the compiler will be investigated. To those functions, a different
6044 (more restrictive) limit compared to functions declared inline can
6046 The default value is 90.
6048 @item large-function-insns
6049 The limit specifying really large functions. For functions larger than this
6050 limit after inlining inlining is constrained by
6051 @option{--param large-function-growth}. This parameter is useful primarily
6052 to avoid extreme compilation time caused by non-linear algorithms used by the
6054 This parameter is ignored when @option{-funit-at-a-time} is not used.
6055 The default value is 2700.
6057 @item large-function-growth
6058 Specifies maximal growth of large function caused by inlining in percents.
6059 This parameter is ignored when @option{-funit-at-a-time} is not used.
6060 The default value is 100 which limits large function growth to 2.0 times
6063 @item large-unit-insns
6064 The limit specifying large translation unit. Growth caused by inlining of
6065 units larger than this limit is limited by @option{--param inline-unit-growth}.
6066 For small units this might be too tight (consider unit consisting of function A
6067 that is inline and B that just calls A three time. If B is small relative to
6068 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6069 large units consisting of small inlininable functions however the overall unit
6070 growth limit is needed to avoid exponential explosion of code size. Thus for
6071 smaller units, the size is increased to @option{--param large-unit-insns}
6072 before applying @option{--param inline-unit-growth}. The default is 10000
6074 @item inline-unit-growth
6075 Specifies maximal overall growth of the compilation unit caused by inlining.
6076 This parameter is ignored when @option{-funit-at-a-time} is not used.
6077 The default value is 50 which limits unit growth to 1.5 times the original
6080 @item large-stack-frame
6081 The limit specifying large stack frames. While inlining the algorithm is trying
6082 to not grow past this limit too much. Default value is 256 bytes.
6084 @item large-stack-frame-growth
6085 Specifies maximal growth of large stack frames caused by inlining in percents.
6086 The default value is 1000 which limits large stack frame growth to 11 times
6089 @item max-inline-insns-recursive
6090 @itemx max-inline-insns-recursive-auto
6091 Specifies maximum number of instructions out-of-line copy of self recursive inline
6092 function can grow into by performing recursive inlining.
6094 For functions declared inline @option{--param max-inline-insns-recursive} is
6095 taken into account. For function not declared inline, recursive inlining
6096 happens only when @option{-finline-functions} (included in @option{-O3}) is
6097 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6098 default value is 450.
6100 @item max-inline-recursive-depth
6101 @itemx max-inline-recursive-depth-auto
6102 Specifies maximum recursion depth used by the recursive inlining.
6104 For functions declared inline @option{--param max-inline-recursive-depth} is
6105 taken into account. For function not declared inline, recursive inlining
6106 happens only when @option{-finline-functions} (included in @option{-O3}) is
6107 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6108 default value is 450.
6110 @item min-inline-recursive-probability
6111 Recursive inlining is profitable only for function having deep recursion
6112 in average and can hurt for function having little recursion depth by
6113 increasing the prologue size or complexity of function body to other
6116 When profile feedback is available (see @option{-fprofile-generate}) the actual
6117 recursion depth can be guessed from probability that function will recurse via
6118 given call expression. This parameter limits inlining only to call expression
6119 whose probability exceeds given threshold (in percents). The default value is
6122 @item inline-call-cost
6123 Specify cost of call instruction relative to simple arithmetics operations
6124 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6125 functions and at the same time increases size of leaf function that is believed to
6126 reduce function size by being inlined. In effect it increases amount of
6127 inlining for code having large abstraction penalty (many functions that just
6128 pass the arguments to other functions) and decrease inlining for code with low
6129 abstraction penalty. The default value is 16.
6131 @item max-unrolled-insns
6132 The maximum number of instructions that a loop should have if that loop
6133 is unrolled, and if the loop is unrolled, it determines how many times
6134 the loop code is unrolled.
6136 @item max-average-unrolled-insns
6137 The maximum number of instructions biased by probabilities of their execution
6138 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6139 it determines how many times the loop code is unrolled.
6141 @item max-unroll-times
6142 The maximum number of unrollings of a single loop.
6144 @item max-peeled-insns
6145 The maximum number of instructions that a loop should have if that loop
6146 is peeled, and if the loop is peeled, it determines how many times
6147 the loop code is peeled.
6149 @item max-peel-times
6150 The maximum number of peelings of a single loop.
6152 @item max-completely-peeled-insns
6153 The maximum number of insns of a completely peeled loop.
6155 @item max-completely-peel-times
6156 The maximum number of iterations of a loop to be suitable for complete peeling.
6158 @item max-unswitch-insns
6159 The maximum number of insns of an unswitched loop.
6161 @item max-unswitch-level
6162 The maximum number of branches unswitched in a single loop.
6165 The minimum cost of an expensive expression in the loop invariant motion.
6167 @item iv-consider-all-candidates-bound
6168 Bound on number of candidates for induction variables below that
6169 all candidates are considered for each use in induction variable
6170 optimizations. Only the most relevant candidates are considered
6171 if there are more candidates, to avoid quadratic time complexity.
6173 @item iv-max-considered-uses
6174 The induction variable optimizations give up on loops that contain more
6175 induction variable uses.
6177 @item iv-always-prune-cand-set-bound
6178 If number of candidates in the set is smaller than this value,
6179 we always try to remove unnecessary ivs from the set during its
6180 optimization when a new iv is added to the set.
6182 @item scev-max-expr-size
6183 Bound on size of expressions used in the scalar evolutions analyzer.
6184 Large expressions slow the analyzer.
6186 @item vect-max-version-checks
6187 The maximum number of runtime checks that can be performed when doing
6188 loop versioning in the vectorizer. See option ftree-vect-loop-version
6189 for more information.
6191 @item max-iterations-to-track
6193 The maximum number of iterations of a loop the brute force algorithm
6194 for analysis of # of iterations of the loop tries to evaluate.
6196 @item hot-bb-count-fraction
6197 Select fraction of the maximal count of repetitions of basic block in program
6198 given basic block needs to have to be considered hot.
6200 @item hot-bb-frequency-fraction
6201 Select fraction of the maximal frequency of executions of basic block in
6202 function given basic block needs to have to be considered hot
6204 @item max-predicted-iterations
6205 The maximum number of loop iterations we predict statically. This is useful
6206 in cases where function contain single loop with known bound and other loop
6207 with unknown. We predict the known number of iterations correctly, while
6208 the unknown number of iterations average to roughly 10. This means that the
6209 loop without bounds would appear artificially cold relative to the other one.
6211 @item tracer-dynamic-coverage
6212 @itemx tracer-dynamic-coverage-feedback
6214 This value is used to limit superblock formation once the given percentage of
6215 executed instructions is covered. This limits unnecessary code size
6218 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6219 feedback is available. The real profiles (as opposed to statically estimated
6220 ones) are much less balanced allowing the threshold to be larger value.
6222 @item tracer-max-code-growth
6223 Stop tail duplication once code growth has reached given percentage. This is
6224 rather hokey argument, as most of the duplicates will be eliminated later in
6225 cross jumping, so it may be set to much higher values than is the desired code
6228 @item tracer-min-branch-ratio
6230 Stop reverse growth when the reverse probability of best edge is less than this
6231 threshold (in percent).
6233 @item tracer-min-branch-ratio
6234 @itemx tracer-min-branch-ratio-feedback
6236 Stop forward growth if the best edge do have probability lower than this
6239 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6240 compilation for profile feedback and one for compilation without. The value
6241 for compilation with profile feedback needs to be more conservative (higher) in
6242 order to make tracer effective.
6244 @item max-cse-path-length
6246 Maximum number of basic blocks on path that cse considers. The default is 10.
6249 The maximum instructions CSE process before flushing. The default is 1000.
6251 @item max-aliased-vops
6253 Maximum number of virtual operands per statement allowed to represent
6254 aliases before triggering the alias grouping heuristic. Alias
6255 grouping reduces compile times and memory consumption needed for
6256 aliasing at the expense of precision loss in alias information.
6258 @item ggc-min-expand
6260 GCC uses a garbage collector to manage its own memory allocation. This
6261 parameter specifies the minimum percentage by which the garbage
6262 collector's heap should be allowed to expand between collections.
6263 Tuning this may improve compilation speed; it has no effect on code
6266 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6267 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6268 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6269 GCC is not able to calculate RAM on a particular platform, the lower
6270 bound of 30% is used. Setting this parameter and
6271 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6272 every opportunity. This is extremely slow, but can be useful for
6275 @item ggc-min-heapsize
6277 Minimum size of the garbage collector's heap before it begins bothering
6278 to collect garbage. The first collection occurs after the heap expands
6279 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6280 tuning this may improve compilation speed, and has no effect on code
6283 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6284 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6285 with a lower bound of 4096 (four megabytes) and an upper bound of
6286 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6287 particular platform, the lower bound is used. Setting this parameter
6288 very large effectively disables garbage collection. Setting this
6289 parameter and @option{ggc-min-expand} to zero causes a full collection
6290 to occur at every opportunity.
6292 @item max-reload-search-insns
6293 The maximum number of instruction reload should look backward for equivalent
6294 register. Increasing values mean more aggressive optimization, making the
6295 compile time increase with probably slightly better performance. The default
6298 @item max-cselib-memory-locations
6299 The maximum number of memory locations cselib should take into account.
6300 Increasing values mean more aggressive optimization, making the compile time
6301 increase with probably slightly better performance. The default value is 500.
6303 @item max-flow-memory-locations
6304 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6305 The default value is 100.
6307 @item reorder-blocks-duplicate
6308 @itemx reorder-blocks-duplicate-feedback
6310 Used by basic block reordering pass to decide whether to use unconditional
6311 branch or duplicate the code on its destination. Code is duplicated when its
6312 estimated size is smaller than this value multiplied by the estimated size of
6313 unconditional jump in the hot spots of the program.
6315 The @option{reorder-block-duplicate-feedback} is used only when profile
6316 feedback is available and may be set to higher values than
6317 @option{reorder-block-duplicate} since information about the hot spots is more
6320 @item max-sched-ready-insns
6321 The maximum number of instructions ready to be issued the scheduler should
6322 consider at any given time during the first scheduling pass. Increasing
6323 values mean more thorough searches, making the compilation time increase
6324 with probably little benefit. The default value is 100.
6326 @item max-sched-region-blocks
6327 The maximum number of blocks in a region to be considered for
6328 interblock scheduling. The default value is 10.
6330 @item max-sched-region-insns
6331 The maximum number of insns in a region to be considered for
6332 interblock scheduling. The default value is 100.
6335 The minimum probability (in percents) of reaching a source block
6336 for interblock speculative scheduling. The default value is 40.
6338 @item max-sched-extend-regions-iters
6339 The maximum number of iterations through CFG to extend regions.
6340 0 - disable region extension,
6341 N - do at most N iterations.
6342 The default value is 0.
6344 @item max-sched-insn-conflict-delay
6345 The maximum conflict delay for an insn to be considered for speculative motion.
6346 The default value is 3.
6348 @item sched-spec-prob-cutoff
6349 The minimal probability of speculation success (in percents), so that
6350 speculative insn will be scheduled.
6351 The default value is 40.
6353 @item max-last-value-rtl
6355 The maximum size measured as number of RTLs that can be recorded in an expression
6356 in combiner for a pseudo register as last known value of that register. The default
6359 @item integer-share-limit
6360 Small integer constants can use a shared data structure, reducing the
6361 compiler's memory usage and increasing its speed. This sets the maximum
6362 value of a shared integer constant's. The default value is 256.
6364 @item min-virtual-mappings
6365 Specifies the minimum number of virtual mappings in the incremental
6366 SSA updater that should be registered to trigger the virtual mappings
6367 heuristic defined by virtual-mappings-ratio. The default value is
6370 @item virtual-mappings-ratio
6371 If the number of virtual mappings is virtual-mappings-ratio bigger
6372 than the number of virtual symbols to be updated, then the incremental
6373 SSA updater switches to a full update for those symbols. The default
6376 @item ssp-buffer-size
6377 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6378 protection when @option{-fstack-protection} is used.
6380 @item max-jump-thread-duplication-stmts
6381 Maximum number of statements allowed in a block that needs to be
6382 duplicated when threading jumps.
6384 @item max-fields-for-field-sensitive
6385 Maximum number of fields in a structure we will treat in
6386 a field sensitive manner during pointer analysis.
6388 @item prefetch-latency
6389 Estimate on average number of instructions that are executed before
6390 prefetch finishes. The distance we prefetch ahead is proportional
6391 to this constant. Increasing this number may also lead to less
6392 streams being prefetched (see @option{simultaneous-prefetches}).
6394 @item simultaneous-prefetches
6395 Maximum number of prefetches that can run at the same time.
6397 @item l1-cache-line-size
6398 The size of cache line in L1 cache, in bytes.
6401 The number of cache lines in L1 cache.
6403 @item verify-canonical-types
6404 Whether the compiler should verify the ``canonical'' types used for
6405 type equality comparisons within the C++ and Objective-C++ front
6406 ends. Set to 1 (the default when GCC is configured with
6407 --enable-checking) to enable verification, 0 to disable verification
6408 (the default when GCC is configured with --disable-checking).
6413 @node Preprocessor Options
6414 @section Options Controlling the Preprocessor
6415 @cindex preprocessor options
6416 @cindex options, preprocessor
6418 These options control the C preprocessor, which is run on each C source
6419 file before actual compilation.
6421 If you use the @option{-E} option, nothing is done except preprocessing.
6422 Some of these options make sense only together with @option{-E} because
6423 they cause the preprocessor output to be unsuitable for actual
6428 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6429 and pass @var{option} directly through to the preprocessor. If
6430 @var{option} contains commas, it is split into multiple options at the
6431 commas. However, many options are modified, translated or interpreted
6432 by the compiler driver before being passed to the preprocessor, and
6433 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6434 interface is undocumented and subject to change, so whenever possible
6435 you should avoid using @option{-Wp} and let the driver handle the
6438 @item -Xpreprocessor @var{option}
6439 @opindex preprocessor
6440 Pass @var{option} as an option to the preprocessor. You can use this to
6441 supply system-specific preprocessor options which GCC does not know how to
6444 If you want to pass an option that takes an argument, you must use
6445 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6448 @include cppopts.texi
6450 @node Assembler Options
6451 @section Passing Options to the Assembler
6453 @c prevent bad page break with this line
6454 You can pass options to the assembler.
6457 @item -Wa,@var{option}
6459 Pass @var{option} as an option to the assembler. If @var{option}
6460 contains commas, it is split into multiple options at the commas.
6462 @item -Xassembler @var{option}
6464 Pass @var{option} as an option to the assembler. You can use this to
6465 supply system-specific assembler options which GCC does not know how to
6468 If you want to pass an option that takes an argument, you must use
6469 @option{-Xassembler} twice, once for the option and once for the argument.
6474 @section Options for Linking
6475 @cindex link options
6476 @cindex options, linking
6478 These options come into play when the compiler links object files into
6479 an executable output file. They are meaningless if the compiler is
6480 not doing a link step.
6484 @item @var{object-file-name}
6485 A file name that does not end in a special recognized suffix is
6486 considered to name an object file or library. (Object files are
6487 distinguished from libraries by the linker according to the file
6488 contents.) If linking is done, these object files are used as input
6497 If any of these options is used, then the linker is not run, and
6498 object file names should not be used as arguments. @xref{Overall
6502 @item -l@var{library}
6503 @itemx -l @var{library}
6505 Search the library named @var{library} when linking. (The second
6506 alternative with the library as a separate argument is only for
6507 POSIX compliance and is not recommended.)
6509 It makes a difference where in the command you write this option; the
6510 linker searches and processes libraries and object files in the order they
6511 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6512 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6513 to functions in @samp{z}, those functions may not be loaded.
6515 The linker searches a standard list of directories for the library,
6516 which is actually a file named @file{lib@var{library}.a}. The linker
6517 then uses this file as if it had been specified precisely by name.
6519 The directories searched include several standard system directories
6520 plus any that you specify with @option{-L}.
6522 Normally the files found this way are library files---archive files
6523 whose members are object files. The linker handles an archive file by
6524 scanning through it for members which define symbols that have so far
6525 been referenced but not defined. But if the file that is found is an
6526 ordinary object file, it is linked in the usual fashion. The only
6527 difference between using an @option{-l} option and specifying a file name
6528 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6529 and searches several directories.
6533 You need this special case of the @option{-l} option in order to
6534 link an Objective-C or Objective-C++ program.
6537 @opindex nostartfiles
6538 Do not use the standard system startup files when linking.
6539 The standard system libraries are used normally, unless @option{-nostdlib}
6540 or @option{-nodefaultlibs} is used.
6542 @item -nodefaultlibs
6543 @opindex nodefaultlibs
6544 Do not use the standard system libraries when linking.
6545 Only the libraries you specify will be passed to the linker.
6546 The standard startup files are used normally, unless @option{-nostartfiles}
6547 is used. The compiler may generate calls to @code{memcmp},
6548 @code{memset}, @code{memcpy} and @code{memmove}.
6549 These entries are usually resolved by entries in
6550 libc. These entry points should be supplied through some other
6551 mechanism when this option is specified.
6555 Do not use the standard system startup files or libraries when linking.
6556 No startup files and only the libraries you specify will be passed to
6557 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6558 @code{memcpy} and @code{memmove}.
6559 These entries are usually resolved by entries in
6560 libc. These entry points should be supplied through some other
6561 mechanism when this option is specified.
6563 @cindex @option{-lgcc}, use with @option{-nostdlib}
6564 @cindex @option{-nostdlib} and unresolved references
6565 @cindex unresolved references and @option{-nostdlib}
6566 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6567 @cindex @option{-nodefaultlibs} and unresolved references
6568 @cindex unresolved references and @option{-nodefaultlibs}
6569 One of the standard libraries bypassed by @option{-nostdlib} and
6570 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6571 that GCC uses to overcome shortcomings of particular machines, or special
6572 needs for some languages.
6573 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6574 Collection (GCC) Internals},
6575 for more discussion of @file{libgcc.a}.)
6576 In most cases, you need @file{libgcc.a} even when you want to avoid
6577 other standard libraries. In other words, when you specify @option{-nostdlib}
6578 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6579 This ensures that you have no unresolved references to internal GCC
6580 library subroutines. (For example, @samp{__main}, used to ensure C++
6581 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6582 GNU Compiler Collection (GCC) Internals}.)
6586 Produce a position independent executable on targets which support it.
6587 For predictable results, you must also specify the same set of options
6588 that were used to generate code (@option{-fpie}, @option{-fPIE},
6589 or model suboptions) when you specify this option.
6593 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6594 that support it. This instructs the linker to add all symbols, not
6595 only used ones, to the dynamic symbol table. This option is needed
6596 for some uses of @code{dlopen} or to allow obtaining backtraces
6597 from within a program.
6601 Remove all symbol table and relocation information from the executable.
6605 On systems that support dynamic linking, this prevents linking with the shared
6606 libraries. On other systems, this option has no effect.
6610 Produce a shared object which can then be linked with other objects to
6611 form an executable. Not all systems support this option. For predictable
6612 results, you must also specify the same set of options that were used to
6613 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6614 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6615 needs to build supplementary stub code for constructors to work. On
6616 multi-libbed systems, @samp{gcc -shared} must select the correct support
6617 libraries to link against. Failing to supply the correct flags may lead
6618 to subtle defects. Supplying them in cases where they are not necessary
6621 @item -shared-libgcc
6622 @itemx -static-libgcc
6623 @opindex shared-libgcc
6624 @opindex static-libgcc
6625 On systems that provide @file{libgcc} as a shared library, these options
6626 force the use of either the shared or static version respectively.
6627 If no shared version of @file{libgcc} was built when the compiler was
6628 configured, these options have no effect.
6630 There are several situations in which an application should use the
6631 shared @file{libgcc} instead of the static version. The most common
6632 of these is when the application wishes to throw and catch exceptions
6633 across different shared libraries. In that case, each of the libraries
6634 as well as the application itself should use the shared @file{libgcc}.
6636 Therefore, the G++ and GCJ drivers automatically add
6637 @option{-shared-libgcc} whenever you build a shared library or a main
6638 executable, because C++ and Java programs typically use exceptions, so
6639 this is the right thing to do.
6641 If, instead, you use the GCC driver to create shared libraries, you may
6642 find that they will not always be linked with the shared @file{libgcc}.
6643 If GCC finds, at its configuration time, that you have a non-GNU linker
6644 or a GNU linker that does not support option @option{--eh-frame-hdr},
6645 it will link the shared version of @file{libgcc} into shared libraries
6646 by default. Otherwise, it will take advantage of the linker and optimize
6647 away the linking with the shared version of @file{libgcc}, linking with
6648 the static version of libgcc by default. This allows exceptions to
6649 propagate through such shared libraries, without incurring relocation
6650 costs at library load time.
6652 However, if a library or main executable is supposed to throw or catch
6653 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6654 for the languages used in the program, or using the option
6655 @option{-shared-libgcc}, such that it is linked with the shared
6660 Bind references to global symbols when building a shared object. Warn
6661 about any unresolved references (unless overridden by the link editor
6662 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6665 @item -Xlinker @var{option}
6667 Pass @var{option} as an option to the linker. You can use this to
6668 supply system-specific linker options which GCC does not know how to
6671 If you want to pass an option that takes an argument, you must use
6672 @option{-Xlinker} twice, once for the option and once for the argument.
6673 For example, to pass @option{-assert definitions}, you must write
6674 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6675 @option{-Xlinker "-assert definitions"}, because this passes the entire
6676 string as a single argument, which is not what the linker expects.
6678 @item -Wl,@var{option}
6680 Pass @var{option} as an option to the linker. If @var{option} contains
6681 commas, it is split into multiple options at the commas.
6683 @item -u @var{symbol}
6685 Pretend the symbol @var{symbol} is undefined, to force linking of
6686 library modules to define it. You can use @option{-u} multiple times with
6687 different symbols to force loading of additional library modules.
6690 @node Directory Options
6691 @section Options for Directory Search
6692 @cindex directory options
6693 @cindex options, directory search
6696 These options specify directories to search for header files, for
6697 libraries and for parts of the compiler:
6702 Add the directory @var{dir} to the head of the list of directories to be
6703 searched for header files. This can be used to override a system header
6704 file, substituting your own version, since these directories are
6705 searched before the system header file directories. However, you should
6706 not use this option to add directories that contain vendor-supplied
6707 system header files (use @option{-isystem} for that). If you use more than
6708 one @option{-I} option, the directories are scanned in left-to-right
6709 order; the standard system directories come after.
6711 If a standard system include directory, or a directory specified with
6712 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6713 option will be ignored. The directory will still be searched but as a
6714 system directory at its normal position in the system include chain.
6715 This is to ensure that GCC's procedure to fix buggy system headers and
6716 the ordering for the include_next directive are not inadvertently changed.
6717 If you really need to change the search order for system directories,
6718 use the @option{-nostdinc} and/or @option{-isystem} options.
6720 @item -iquote@var{dir}
6722 Add the directory @var{dir} to the head of the list of directories to
6723 be searched for header files only for the case of @samp{#include
6724 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6725 otherwise just like @option{-I}.
6729 Add directory @var{dir} to the list of directories to be searched
6732 @item -B@var{prefix}
6734 This option specifies where to find the executables, libraries,
6735 include files, and data files of the compiler itself.
6737 The compiler driver program runs one or more of the subprograms
6738 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6739 @var{prefix} as a prefix for each program it tries to run, both with and
6740 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6742 For each subprogram to be run, the compiler driver first tries the
6743 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6744 was not specified, the driver tries two standard prefixes, which are
6745 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6746 those results in a file name that is found, the unmodified program
6747 name is searched for using the directories specified in your
6748 @env{PATH} environment variable.
6750 The compiler will check to see if the path provided by the @option{-B}
6751 refers to a directory, and if necessary it will add a directory
6752 separator character at the end of the path.
6754 @option{-B} prefixes that effectively specify directory names also apply
6755 to libraries in the linker, because the compiler translates these
6756 options into @option{-L} options for the linker. They also apply to
6757 includes files in the preprocessor, because the compiler translates these
6758 options into @option{-isystem} options for the preprocessor. In this case,
6759 the compiler appends @samp{include} to the prefix.
6761 The run-time support file @file{libgcc.a} can also be searched for using
6762 the @option{-B} prefix, if needed. If it is not found there, the two
6763 standard prefixes above are tried, and that is all. The file is left
6764 out of the link if it is not found by those means.
6766 Another way to specify a prefix much like the @option{-B} prefix is to use
6767 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6770 As a special kludge, if the path provided by @option{-B} is
6771 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6772 9, then it will be replaced by @file{[dir/]include}. This is to help
6773 with boot-strapping the compiler.
6775 @item -specs=@var{file}
6777 Process @var{file} after the compiler reads in the standard @file{specs}
6778 file, in order to override the defaults that the @file{gcc} driver
6779 program uses when determining what switches to pass to @file{cc1},
6780 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6781 @option{-specs=@var{file}} can be specified on the command line, and they
6782 are processed in order, from left to right.
6784 @item --sysroot=@var{dir}
6786 Use @var{dir} as the logical root directory for headers and libraries.
6787 For example, if the compiler would normally search for headers in
6788 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6789 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6791 If you use both this option and the @option{-isysroot} option, then
6792 the @option{--sysroot} option will apply to libraries, but the
6793 @option{-isysroot} option will apply to header files.
6795 The GNU linker (beginning with version 2.16) has the necessary support
6796 for this option. If your linker does not support this option, the
6797 header file aspect of @option{--sysroot} will still work, but the
6798 library aspect will not.
6802 This option has been deprecated. Please use @option{-iquote} instead for
6803 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6804 Any directories you specify with @option{-I} options before the @option{-I-}
6805 option are searched only for the case of @samp{#include "@var{file}"};
6806 they are not searched for @samp{#include <@var{file}>}.
6808 If additional directories are specified with @option{-I} options after
6809 the @option{-I-}, these directories are searched for all @samp{#include}
6810 directives. (Ordinarily @emph{all} @option{-I} directories are used
6813 In addition, the @option{-I-} option inhibits the use of the current
6814 directory (where the current input file came from) as the first search
6815 directory for @samp{#include "@var{file}"}. There is no way to
6816 override this effect of @option{-I-}. With @option{-I.} you can specify
6817 searching the directory which was current when the compiler was
6818 invoked. That is not exactly the same as what the preprocessor does
6819 by default, but it is often satisfactory.
6821 @option{-I-} does not inhibit the use of the standard system directories
6822 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6829 @section Specifying subprocesses and the switches to pass to them
6832 @command{gcc} is a driver program. It performs its job by invoking a
6833 sequence of other programs to do the work of compiling, assembling and
6834 linking. GCC interprets its command-line parameters and uses these to
6835 deduce which programs it should invoke, and which command-line options
6836 it ought to place on their command lines. This behavior is controlled
6837 by @dfn{spec strings}. In most cases there is one spec string for each
6838 program that GCC can invoke, but a few programs have multiple spec
6839 strings to control their behavior. The spec strings built into GCC can
6840 be overridden by using the @option{-specs=} command-line switch to specify
6843 @dfn{Spec files} are plaintext files that are used to construct spec
6844 strings. They consist of a sequence of directives separated by blank
6845 lines. The type of directive is determined by the first non-whitespace
6846 character on the line and it can be one of the following:
6849 @item %@var{command}
6850 Issues a @var{command} to the spec file processor. The commands that can
6854 @item %include <@var{file}>
6856 Search for @var{file} and insert its text at the current point in the
6859 @item %include_noerr <@var{file}>
6860 @cindex %include_noerr
6861 Just like @samp{%include}, but do not generate an error message if the include
6862 file cannot be found.
6864 @item %rename @var{old_name} @var{new_name}
6866 Rename the spec string @var{old_name} to @var{new_name}.
6870 @item *[@var{spec_name}]:
6871 This tells the compiler to create, override or delete the named spec
6872 string. All lines after this directive up to the next directive or
6873 blank line are considered to be the text for the spec string. If this
6874 results in an empty string then the spec will be deleted. (Or, if the
6875 spec did not exist, then nothing will happened.) Otherwise, if the spec
6876 does not currently exist a new spec will be created. If the spec does
6877 exist then its contents will be overridden by the text of this
6878 directive, unless the first character of that text is the @samp{+}
6879 character, in which case the text will be appended to the spec.
6881 @item [@var{suffix}]:
6882 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6883 and up to the next directive or blank line are considered to make up the
6884 spec string for the indicated suffix. When the compiler encounters an
6885 input file with the named suffix, it will processes the spec string in
6886 order to work out how to compile that file. For example:
6893 This says that any input file whose name ends in @samp{.ZZ} should be
6894 passed to the program @samp{z-compile}, which should be invoked with the
6895 command-line switch @option{-input} and with the result of performing the
6896 @samp{%i} substitution. (See below.)
6898 As an alternative to providing a spec string, the text that follows a
6899 suffix directive can be one of the following:
6902 @item @@@var{language}
6903 This says that the suffix is an alias for a known @var{language}. This is
6904 similar to using the @option{-x} command-line switch to GCC to specify a
6905 language explicitly. For example:
6912 Says that .ZZ files are, in fact, C++ source files.
6915 This causes an error messages saying:
6918 @var{name} compiler not installed on this system.
6922 GCC already has an extensive list of suffixes built into it.
6923 This directive will add an entry to the end of the list of suffixes, but
6924 since the list is searched from the end backwards, it is effectively
6925 possible to override earlier entries using this technique.
6929 GCC has the following spec strings built into it. Spec files can
6930 override these strings or create their own. Note that individual
6931 targets can also add their own spec strings to this list.
6934 asm Options to pass to the assembler
6935 asm_final Options to pass to the assembler post-processor
6936 cpp Options to pass to the C preprocessor
6937 cc1 Options to pass to the C compiler
6938 cc1plus Options to pass to the C++ compiler
6939 endfile Object files to include at the end of the link
6940 link Options to pass to the linker
6941 lib Libraries to include on the command line to the linker
6942 libgcc Decides which GCC support library to pass to the linker
6943 linker Sets the name of the linker
6944 predefines Defines to be passed to the C preprocessor
6945 signed_char Defines to pass to CPP to say whether @code{char} is signed
6947 startfile Object files to include at the start of the link
6950 Here is a small example of a spec file:
6956 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6959 This example renames the spec called @samp{lib} to @samp{old_lib} and
6960 then overrides the previous definition of @samp{lib} with a new one.
6961 The new definition adds in some extra command-line options before
6962 including the text of the old definition.
6964 @dfn{Spec strings} are a list of command-line options to be passed to their
6965 corresponding program. In addition, the spec strings can contain
6966 @samp{%}-prefixed sequences to substitute variable text or to
6967 conditionally insert text into the command line. Using these constructs
6968 it is possible to generate quite complex command lines.
6970 Here is a table of all defined @samp{%}-sequences for spec
6971 strings. Note that spaces are not generated automatically around the
6972 results of expanding these sequences. Therefore you can concatenate them
6973 together or combine them with constant text in a single argument.
6977 Substitute one @samp{%} into the program name or argument.
6980 Substitute the name of the input file being processed.
6983 Substitute the basename of the input file being processed.
6984 This is the substring up to (and not including) the last period
6985 and not including the directory.
6988 This is the same as @samp{%b}, but include the file suffix (text after
6992 Marks the argument containing or following the @samp{%d} as a
6993 temporary file name, so that that file will be deleted if GCC exits
6994 successfully. Unlike @samp{%g}, this contributes no text to the
6997 @item %g@var{suffix}
6998 Substitute a file name that has suffix @var{suffix} and is chosen
6999 once per compilation, and mark the argument in the same way as
7000 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
7001 name is now chosen in a way that is hard to predict even when previously
7002 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
7003 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
7004 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
7005 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
7006 was simply substituted with a file name chosen once per compilation,
7007 without regard to any appended suffix (which was therefore treated
7008 just like ordinary text), making such attacks more likely to succeed.
7010 @item %u@var{suffix}
7011 Like @samp{%g}, but generates a new temporary file name even if
7012 @samp{%u@var{suffix}} was already seen.
7014 @item %U@var{suffix}
7015 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
7016 new one if there is no such last file name. In the absence of any
7017 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
7018 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
7019 would involve the generation of two distinct file names, one
7020 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
7021 simply substituted with a file name chosen for the previous @samp{%u},
7022 without regard to any appended suffix.
7024 @item %j@var{suffix}
7025 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
7026 writable, and if save-temps is off; otherwise, substitute the name
7027 of a temporary file, just like @samp{%u}. This temporary file is not
7028 meant for communication between processes, but rather as a junk
7031 @item %|@var{suffix}
7032 @itemx %m@var{suffix}
7033 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
7034 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
7035 all. These are the two most common ways to instruct a program that it
7036 should read from standard input or write to standard output. If you
7037 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
7038 construct: see for example @file{f/lang-specs.h}.
7040 @item %.@var{SUFFIX}
7041 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
7042 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
7043 terminated by the next space or %.
7046 Marks the argument containing or following the @samp{%w} as the
7047 designated output file of this compilation. This puts the argument
7048 into the sequence of arguments that @samp{%o} will substitute later.
7051 Substitutes the names of all the output files, with spaces
7052 automatically placed around them. You should write spaces
7053 around the @samp{%o} as well or the results are undefined.
7054 @samp{%o} is for use in the specs for running the linker.
7055 Input files whose names have no recognized suffix are not compiled
7056 at all, but they are included among the output files, so they will
7060 Substitutes the suffix for object files. Note that this is
7061 handled specially when it immediately follows @samp{%g, %u, or %U},
7062 because of the need for those to form complete file names. The
7063 handling is such that @samp{%O} is treated exactly as if it had already
7064 been substituted, except that @samp{%g, %u, and %U} do not currently
7065 support additional @var{suffix} characters following @samp{%O} as they would
7066 following, for example, @samp{.o}.
7069 Substitutes the standard macro predefinitions for the
7070 current target machine. Use this when running @code{cpp}.
7073 Like @samp{%p}, but puts @samp{__} before and after the name of each
7074 predefined macro, except for macros that start with @samp{__} or with
7075 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7079 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7080 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7081 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7082 and @option{-imultilib} as necessary.
7085 Current argument is the name of a library or startup file of some sort.
7086 Search for that file in a standard list of directories and substitute
7087 the full name found.
7090 Print @var{str} as an error message. @var{str} is terminated by a newline.
7091 Use this when inconsistent options are detected.
7094 Substitute the contents of spec string @var{name} at this point.
7097 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7099 @item %x@{@var{option}@}
7100 Accumulate an option for @samp{%X}.
7103 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7107 Output the accumulated assembler options specified by @option{-Wa}.
7110 Output the accumulated preprocessor options specified by @option{-Wp}.
7113 Process the @code{asm} spec. This is used to compute the
7114 switches to be passed to the assembler.
7117 Process the @code{asm_final} spec. This is a spec string for
7118 passing switches to an assembler post-processor, if such a program is
7122 Process the @code{link} spec. This is the spec for computing the
7123 command line passed to the linker. Typically it will make use of the
7124 @samp{%L %G %S %D and %E} sequences.
7127 Dump out a @option{-L} option for each directory that GCC believes might
7128 contain startup files. If the target supports multilibs then the
7129 current multilib directory will be prepended to each of these paths.
7132 Process the @code{lib} spec. This is a spec string for deciding which
7133 libraries should be included on the command line to the linker.
7136 Process the @code{libgcc} spec. This is a spec string for deciding
7137 which GCC support library should be included on the command line to the linker.
7140 Process the @code{startfile} spec. This is a spec for deciding which
7141 object files should be the first ones passed to the linker. Typically
7142 this might be a file named @file{crt0.o}.
7145 Process the @code{endfile} spec. This is a spec string that specifies
7146 the last object files that will be passed to the linker.
7149 Process the @code{cpp} spec. This is used to construct the arguments
7150 to be passed to the C preprocessor.
7153 Process the @code{cc1} spec. This is used to construct the options to be
7154 passed to the actual C compiler (@samp{cc1}).
7157 Process the @code{cc1plus} spec. This is used to construct the options to be
7158 passed to the actual C++ compiler (@samp{cc1plus}).
7161 Substitute the variable part of a matched option. See below.
7162 Note that each comma in the substituted string is replaced by
7166 Remove all occurrences of @code{-S} from the command line. Note---this
7167 command is position dependent. @samp{%} commands in the spec string
7168 before this one will see @code{-S}, @samp{%} commands in the spec string
7169 after this one will not.
7171 @item %:@var{function}(@var{args})
7172 Call the named function @var{function}, passing it @var{args}.
7173 @var{args} is first processed as a nested spec string, then split
7174 into an argument vector in the usual fashion. The function returns
7175 a string which is processed as if it had appeared literally as part
7176 of the current spec.
7178 The following built-in spec functions are provided:
7181 @item @code{if-exists}
7182 The @code{if-exists} spec function takes one argument, an absolute
7183 pathname to a file. If the file exists, @code{if-exists} returns the
7184 pathname. Here is a small example of its usage:
7188 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7191 @item @code{if-exists-else}
7192 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7193 spec function, except that it takes two arguments. The first argument is
7194 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7195 returns the pathname. If it does not exist, it returns the second argument.
7196 This way, @code{if-exists-else} can be used to select one file or another,
7197 based on the existence of the first. Here is a small example of its usage:
7201 crt0%O%s %:if-exists(crti%O%s) \
7202 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7205 @item @code{replace-outfile}
7206 The @code{replace-outfile} spec function takes two arguments. It looks for the
7207 first argument in the outfiles array and replaces it with the second argument. Here
7208 is a small example of its usage:
7211 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7217 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7218 If that switch was not specified, this substitutes nothing. Note that
7219 the leading dash is omitted when specifying this option, and it is
7220 automatically inserted if the substitution is performed. Thus the spec
7221 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7222 and would output the command line option @option{-foo}.
7224 @item %W@{@code{S}@}
7225 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7228 @item %@{@code{S}*@}
7229 Substitutes all the switches specified to GCC whose names start
7230 with @code{-S}, but which also take an argument. This is used for
7231 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7232 GCC considers @option{-o foo} as being
7233 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7234 text, including the space. Thus two arguments would be generated.
7236 @item %@{@code{S}*&@code{T}*@}
7237 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7238 (the order of @code{S} and @code{T} in the spec is not significant).
7239 There can be any number of ampersand-separated variables; for each the
7240 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7242 @item %@{@code{S}:@code{X}@}
7243 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7245 @item %@{!@code{S}:@code{X}@}
7246 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7248 @item %@{@code{S}*:@code{X}@}
7249 Substitutes @code{X} if one or more switches whose names start with
7250 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7251 once, no matter how many such switches appeared. However, if @code{%*}
7252 appears somewhere in @code{X}, then @code{X} will be substituted once
7253 for each matching switch, with the @code{%*} replaced by the part of
7254 that switch that matched the @code{*}.
7256 @item %@{.@code{S}:@code{X}@}
7257 Substitutes @code{X}, if processing a file with suffix @code{S}.
7259 @item %@{!.@code{S}:@code{X}@}
7260 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7262 @item %@{@code{S}|@code{P}:@code{X}@}
7263 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7264 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7265 although they have a stronger binding than the @samp{|}. If @code{%*}
7266 appears in @code{X}, all of the alternatives must be starred, and only
7267 the first matching alternative is substituted.
7269 For example, a spec string like this:
7272 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7275 will output the following command-line options from the following input
7276 command-line options:
7281 -d fred.c -foo -baz -boggle
7282 -d jim.d -bar -baz -boggle
7285 @item %@{S:X; T:Y; :D@}
7287 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7288 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7289 be as many clauses as you need. This may be combined with @code{.},
7290 @code{!}, @code{|}, and @code{*} as needed.
7295 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7296 construct may contain other nested @samp{%} constructs or spaces, or
7297 even newlines. They are processed as usual, as described above.
7298 Trailing white space in @code{X} is ignored. White space may also
7299 appear anywhere on the left side of the colon in these constructs,
7300 except between @code{.} or @code{*} and the corresponding word.
7302 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7303 handled specifically in these constructs. If another value of
7304 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7305 @option{-W} switch is found later in the command line, the earlier
7306 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7307 just one letter, which passes all matching options.
7309 The character @samp{|} at the beginning of the predicate text is used to
7310 indicate that a command should be piped to the following command, but
7311 only if @option{-pipe} is specified.
7313 It is built into GCC which switches take arguments and which do not.
7314 (You might think it would be useful to generalize this to allow each
7315 compiler's spec to say which switches take arguments. But this cannot
7316 be done in a consistent fashion. GCC cannot even decide which input
7317 files have been specified without knowing which switches take arguments,
7318 and it must know which input files to compile in order to tell which
7321 GCC also knows implicitly that arguments starting in @option{-l} are to be
7322 treated as compiler output files, and passed to the linker in their
7323 proper position among the other output files.
7325 @c man begin OPTIONS
7327 @node Target Options
7328 @section Specifying Target Machine and Compiler Version
7329 @cindex target options
7330 @cindex cross compiling
7331 @cindex specifying machine version
7332 @cindex specifying compiler version and target machine
7333 @cindex compiler version, specifying
7334 @cindex target machine, specifying
7336 The usual way to run GCC is to run the executable called @file{gcc}, or
7337 @file{<machine>-gcc} when cross-compiling, or
7338 @file{<machine>-gcc-<version>} to run a version other than the one that
7339 was installed last. Sometimes this is inconvenient, so GCC provides
7340 options that will switch to another cross-compiler or version.
7343 @item -b @var{machine}
7345 The argument @var{machine} specifies the target machine for compilation.
7347 The value to use for @var{machine} is the same as was specified as the
7348 machine type when configuring GCC as a cross-compiler. For
7349 example, if a cross-compiler was configured with @samp{configure
7350 arm-elf}, meaning to compile for an arm processor with elf binaries,
7351 then you would specify @option{-b arm-elf} to run that cross compiler.
7352 Because there are other options beginning with @option{-b}, the
7353 configuration must contain a hyphen.
7355 @item -V @var{version}
7357 The argument @var{version} specifies which version of GCC to run.
7358 This is useful when multiple versions are installed. For example,
7359 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7362 The @option{-V} and @option{-b} options work by running the
7363 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7364 use them if you can just run that directly.
7366 @node Submodel Options
7367 @section Hardware Models and Configurations
7368 @cindex submodel options
7369 @cindex specifying hardware config
7370 @cindex hardware models and configurations, specifying
7371 @cindex machine dependent options
7373 Earlier we discussed the standard option @option{-b} which chooses among
7374 different installed compilers for completely different target
7375 machines, such as VAX vs.@: 68000 vs.@: 80386.
7377 In addition, each of these target machine types can have its own
7378 special options, starting with @samp{-m}, to choose among various
7379 hardware models or configurations---for example, 68010 vs 68020,
7380 floating coprocessor or none. A single installed version of the
7381 compiler can compile for any model or configuration, according to the
7384 Some configurations of the compiler also support additional special
7385 options, usually for compatibility with other compilers on the same
7388 @c This list is ordered alphanumerically by subsection name.
7389 @c It should be the same order and spelling as these options are listed
7390 @c in Machine Dependent Options
7396 * Blackfin Options::
7400 * DEC Alpha Options::
7401 * DEC Alpha/VMS Options::
7403 * GNU/Linux Options::
7406 * i386 and x86-64 Options::
7419 * RS/6000 and PowerPC Options::
7420 * S/390 and zSeries Options::
7425 * System V Options::
7426 * TMS320C3x/C4x Options::
7430 * Xstormy16 Options::
7436 @subsection ARC Options
7439 These options are defined for ARC implementations:
7444 Compile code for little endian mode. This is the default.
7448 Compile code for big endian mode.
7451 @opindex mmangle-cpu
7452 Prepend the name of the cpu to all public symbol names.
7453 In multiple-processor systems, there are many ARC variants with different
7454 instruction and register set characteristics. This flag prevents code
7455 compiled for one cpu to be linked with code compiled for another.
7456 No facility exists for handling variants that are ``almost identical''.
7457 This is an all or nothing option.
7459 @item -mcpu=@var{cpu}
7461 Compile code for ARC variant @var{cpu}.
7462 Which variants are supported depend on the configuration.
7463 All variants support @option{-mcpu=base}, this is the default.
7465 @item -mtext=@var{text-section}
7466 @itemx -mdata=@var{data-section}
7467 @itemx -mrodata=@var{readonly-data-section}
7471 Put functions, data, and readonly data in @var{text-section},
7472 @var{data-section}, and @var{readonly-data-section} respectively
7473 by default. This can be overridden with the @code{section} attribute.
7474 @xref{Variable Attributes}.
7479 @subsection ARM Options
7482 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7486 @item -mabi=@var{name}
7488 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7489 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7492 @opindex mapcs-frame
7493 Generate a stack frame that is compliant with the ARM Procedure Call
7494 Standard for all functions, even if this is not strictly necessary for
7495 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7496 with this option will cause the stack frames not to be generated for
7497 leaf functions. The default is @option{-mno-apcs-frame}.
7501 This is a synonym for @option{-mapcs-frame}.
7504 @c not currently implemented
7505 @item -mapcs-stack-check
7506 @opindex mapcs-stack-check
7507 Generate code to check the amount of stack space available upon entry to
7508 every function (that actually uses some stack space). If there is
7509 insufficient space available then either the function
7510 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7511 called, depending upon the amount of stack space required. The run time
7512 system is required to provide these functions. The default is
7513 @option{-mno-apcs-stack-check}, since this produces smaller code.
7515 @c not currently implemented
7517 @opindex mapcs-float
7518 Pass floating point arguments using the float point registers. This is
7519 one of the variants of the APCS@. This option is recommended if the
7520 target hardware has a floating point unit or if a lot of floating point
7521 arithmetic is going to be performed by the code. The default is
7522 @option{-mno-apcs-float}, since integer only code is slightly increased in
7523 size if @option{-mapcs-float} is used.
7525 @c not currently implemented
7526 @item -mapcs-reentrant
7527 @opindex mapcs-reentrant
7528 Generate reentrant, position independent code. The default is
7529 @option{-mno-apcs-reentrant}.
7532 @item -mthumb-interwork
7533 @opindex mthumb-interwork
7534 Generate code which supports calling between the ARM and Thumb
7535 instruction sets. Without this option the two instruction sets cannot
7536 be reliably used inside one program. The default is
7537 @option{-mno-thumb-interwork}, since slightly larger code is generated
7538 when @option{-mthumb-interwork} is specified.
7540 @item -mno-sched-prolog
7541 @opindex mno-sched-prolog
7542 Prevent the reordering of instructions in the function prolog, or the
7543 merging of those instruction with the instructions in the function's
7544 body. This means that all functions will start with a recognizable set
7545 of instructions (or in fact one of a choice from a small set of
7546 different function prologues), and this information can be used to
7547 locate the start if functions inside an executable piece of code. The
7548 default is @option{-msched-prolog}.
7551 @opindex mhard-float
7552 Generate output containing floating point instructions. This is the
7556 @opindex msoft-float
7557 Generate output containing library calls for floating point.
7558 @strong{Warning:} the requisite libraries are not available for all ARM
7559 targets. Normally the facilities of the machine's usual C compiler are
7560 used, but this cannot be done directly in cross-compilation. You must make
7561 your own arrangements to provide suitable library functions for
7564 @option{-msoft-float} changes the calling convention in the output file;
7565 therefore, it is only useful if you compile @emph{all} of a program with
7566 this option. In particular, you need to compile @file{libgcc.a}, the
7567 library that comes with GCC, with @option{-msoft-float} in order for
7570 @item -mfloat-abi=@var{name}
7572 Specifies which ABI to use for floating point values. Permissible values
7573 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7575 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7576 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7577 of floating point instructions, but still uses the soft-float calling
7580 @item -mlittle-endian
7581 @opindex mlittle-endian
7582 Generate code for a processor running in little-endian mode. This is
7583 the default for all standard configurations.
7586 @opindex mbig-endian
7587 Generate code for a processor running in big-endian mode; the default is
7588 to compile code for a little-endian processor.
7590 @item -mwords-little-endian
7591 @opindex mwords-little-endian
7592 This option only applies when generating code for big-endian processors.
7593 Generate code for a little-endian word order but a big-endian byte
7594 order. That is, a byte order of the form @samp{32107654}. Note: this
7595 option should only be used if you require compatibility with code for
7596 big-endian ARM processors generated by versions of the compiler prior to
7599 @item -mcpu=@var{name}
7601 This specifies the name of the target ARM processor. GCC uses this name
7602 to determine what kind of instructions it can emit when generating
7603 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7604 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7605 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7606 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7607 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7608 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7609 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7610 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7611 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7612 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7613 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7614 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7615 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7616 @samp{arm1156t2-s}, @samp{arm1176jz-s}, @samp{arm1176jzf-s},
7617 @samp{cortex-a8}, @samp{cortex-r4}, @samp{cortex-m3},
7618 @samp{xscale}, @samp{iwmmxt}, @samp{ep9312}.
7620 @itemx -mtune=@var{name}
7622 This option is very similar to the @option{-mcpu=} option, except that
7623 instead of specifying the actual target processor type, and hence
7624 restricting which instructions can be used, it specifies that GCC should
7625 tune the performance of the code as if the target were of the type
7626 specified in this option, but still choosing the instructions that it
7627 will generate based on the cpu specified by a @option{-mcpu=} option.
7628 For some ARM implementations better performance can be obtained by using
7631 @item -march=@var{name}
7633 This specifies the name of the target ARM architecture. GCC uses this
7634 name to determine what kind of instructions it can emit when generating
7635 assembly code. This option can be used in conjunction with or instead
7636 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7637 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7638 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7639 @samp{armv6t2}, @samp{armv6z}, @samp{armv6zk}, @samp{armv7}, @samp{armv7-a},
7640 @samp{armv7-r}, @samp{armv7-m}, @samp{iwmmxt}, @samp{ep9312}.
7642 @item -mfpu=@var{name}
7643 @itemx -mfpe=@var{number}
7644 @itemx -mfp=@var{number}
7648 This specifies what floating point hardware (or hardware emulation) is
7649 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7650 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7651 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7652 with older versions of GCC@.
7654 If @option{-msoft-float} is specified this specifies the format of
7655 floating point values.
7657 @item -mstructure-size-boundary=@var{n}
7658 @opindex mstructure-size-boundary
7659 The size of all structures and unions will be rounded up to a multiple
7660 of the number of bits set by this option. Permissible values are 8, 32
7661 and 64. The default value varies for different toolchains. For the COFF
7662 targeted toolchain the default value is 8. A value of 64 is only allowed
7663 if the underlying ABI supports it.
7665 Specifying the larger number can produce faster, more efficient code, but
7666 can also increase the size of the program. Different values are potentially
7667 incompatible. Code compiled with one value cannot necessarily expect to
7668 work with code or libraries compiled with another value, if they exchange
7669 information using structures or unions.
7671 @item -mabort-on-noreturn
7672 @opindex mabort-on-noreturn
7673 Generate a call to the function @code{abort} at the end of a
7674 @code{noreturn} function. It will be executed if the function tries to
7678 @itemx -mno-long-calls
7679 @opindex mlong-calls
7680 @opindex mno-long-calls
7681 Tells the compiler to perform function calls by first loading the
7682 address of the function into a register and then performing a subroutine
7683 call on this register. This switch is needed if the target function
7684 will lie outside of the 64 megabyte addressing range of the offset based
7685 version of subroutine call instruction.
7687 Even if this switch is enabled, not all function calls will be turned
7688 into long calls. The heuristic is that static functions, functions
7689 which have the @samp{short-call} attribute, functions that are inside
7690 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7691 definitions have already been compiled within the current compilation
7692 unit, will not be turned into long calls. The exception to this rule is
7693 that weak function definitions, functions with the @samp{long-call}
7694 attribute or the @samp{section} attribute, and functions that are within
7695 the scope of a @samp{#pragma long_calls} directive, will always be
7696 turned into long calls.
7698 This feature is not enabled by default. Specifying
7699 @option{-mno-long-calls} will restore the default behavior, as will
7700 placing the function calls within the scope of a @samp{#pragma
7701 long_calls_off} directive. Note these switches have no effect on how
7702 the compiler generates code to handle function calls via function
7705 @item -mnop-fun-dllimport
7706 @opindex mnop-fun-dllimport
7707 Disable support for the @code{dllimport} attribute.
7709 @item -msingle-pic-base
7710 @opindex msingle-pic-base
7711 Treat the register used for PIC addressing as read-only, rather than
7712 loading it in the prologue for each function. The run-time system is
7713 responsible for initializing this register with an appropriate value
7714 before execution begins.
7716 @item -mpic-register=@var{reg}
7717 @opindex mpic-register
7718 Specify the register to be used for PIC addressing. The default is R10
7719 unless stack-checking is enabled, when R9 is used.
7721 @item -mcirrus-fix-invalid-insns
7722 @opindex mcirrus-fix-invalid-insns
7723 @opindex mno-cirrus-fix-invalid-insns
7724 Insert NOPs into the instruction stream to in order to work around
7725 problems with invalid Maverick instruction combinations. This option
7726 is only valid if the @option{-mcpu=ep9312} option has been used to
7727 enable generation of instructions for the Cirrus Maverick floating
7728 point co-processor. This option is not enabled by default, since the
7729 problem is only present in older Maverick implementations. The default
7730 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7733 @item -mpoke-function-name
7734 @opindex mpoke-function-name
7735 Write the name of each function into the text section, directly
7736 preceding the function prologue. The generated code is similar to this:
7740 .ascii "arm_poke_function_name", 0
7743 .word 0xff000000 + (t1 - t0)
7744 arm_poke_function_name
7746 stmfd sp!, @{fp, ip, lr, pc@}
7750 When performing a stack backtrace, code can inspect the value of
7751 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7752 location @code{pc - 12} and the top 8 bits are set, then we know that
7753 there is a function name embedded immediately preceding this location
7754 and has length @code{((pc[-3]) & 0xff000000)}.
7758 Generate code for the Thumb instruction set. The default is to
7759 use the 32-bit ARM instruction set.
7760 This option automatically enables either 16-bit Thumb-1 or
7761 mixed 16/32-bit Thumb-2 instructions based on the @option{-mcpu=@var{name}}
7762 and @option{-march=@var{name}} options.
7765 @opindex mtpcs-frame
7766 Generate a stack frame that is compliant with the Thumb Procedure Call
7767 Standard for all non-leaf functions. (A leaf function is one that does
7768 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7770 @item -mtpcs-leaf-frame
7771 @opindex mtpcs-leaf-frame
7772 Generate a stack frame that is compliant with the Thumb Procedure Call
7773 Standard for all leaf functions. (A leaf function is one that does
7774 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7776 @item -mcallee-super-interworking
7777 @opindex mcallee-super-interworking
7778 Gives all externally visible functions in the file being compiled an ARM
7779 instruction set header which switches to Thumb mode before executing the
7780 rest of the function. This allows these functions to be called from
7781 non-interworking code.
7783 @item -mcaller-super-interworking
7784 @opindex mcaller-super-interworking
7785 Allows calls via function pointers (including virtual functions) to
7786 execute correctly regardless of whether the target code has been
7787 compiled for interworking or not. There is a small overhead in the cost
7788 of executing a function pointer if this option is enabled.
7790 @item -mtp=@var{name}
7792 Specify the access model for the thread local storage pointer. The valid
7793 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7794 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7795 (supported in the arm6k architecture), and @option{auto}, which uses the
7796 best available method for the selected processor. The default setting is
7802 @subsection AVR Options
7805 These options are defined for AVR implementations:
7808 @item -mmcu=@var{mcu}
7810 Specify ATMEL AVR instruction set or MCU type.
7812 Instruction set avr1 is for the minimal AVR core, not supported by the C
7813 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7814 attiny11, attiny12, attiny15, attiny28).
7816 Instruction set avr2 (default) is for the classic AVR core with up to
7817 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7818 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7819 at90c8534, at90s8535).
7821 Instruction set avr3 is for the classic AVR core with up to 128K program
7822 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7824 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7825 memory space (MCU types: atmega8, atmega83, atmega85).
7827 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7828 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7829 atmega64, atmega128, at43usb355, at94k).
7833 Output instruction sizes to the asm file.
7835 @item -minit-stack=@var{N}
7836 @opindex minit-stack
7837 Specify the initial stack address, which may be a symbol or numeric value,
7838 @samp{__stack} is the default.
7840 @item -mno-interrupts
7841 @opindex mno-interrupts
7842 Generated code is not compatible with hardware interrupts.
7843 Code size will be smaller.
7845 @item -mcall-prologues
7846 @opindex mcall-prologues
7847 Functions prologues/epilogues expanded as call to appropriate
7848 subroutines. Code size will be smaller.
7850 @item -mno-tablejump
7851 @opindex mno-tablejump
7852 Do not generate tablejump insns which sometimes increase code size.
7855 @opindex mtiny-stack
7856 Change only the low 8 bits of the stack pointer.
7860 Assume int to be 8 bit integer. This affects the sizes of all types: A
7861 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7862 and long long will be 4 bytes. Please note that this option does not
7863 comply to the C standards, but it will provide you with smaller code
7867 @node Blackfin Options
7868 @subsection Blackfin Options
7869 @cindex Blackfin Options
7872 @item -momit-leaf-frame-pointer
7873 @opindex momit-leaf-frame-pointer
7874 Don't keep the frame pointer in a register for leaf functions. This
7875 avoids the instructions to save, set up and restore frame pointers and
7876 makes an extra register available in leaf functions. The option
7877 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7878 which might make debugging harder.
7880 @item -mspecld-anomaly
7881 @opindex mspecld-anomaly
7882 When enabled, the compiler will ensure that the generated code does not
7883 contain speculative loads after jump instructions. This option is enabled
7886 @item -mno-specld-anomaly
7887 @opindex mno-specld-anomaly
7888 Don't generate extra code to prevent speculative loads from occurring.
7890 @item -mcsync-anomaly
7891 @opindex mcsync-anomaly
7892 When enabled, the compiler will ensure that the generated code does not
7893 contain CSYNC or SSYNC instructions too soon after conditional branches.
7894 This option is enabled by default.
7896 @item -mno-csync-anomaly
7897 @opindex mno-csync-anomaly
7898 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7899 occurring too soon after a conditional branch.
7903 When enabled, the compiler is free to take advantage of the knowledge that
7904 the entire program fits into the low 64k of memory.
7907 @opindex mno-low-64k
7908 Assume that the program is arbitrarily large. This is the default.
7910 @item -mstack-check-l1
7911 @opindex mstack-check-l1
7912 Do stack checking using information placed into L1 scratchpad memory by the
7915 @item -mid-shared-library
7916 @opindex mid-shared-library
7917 Generate code that supports shared libraries via the library ID method.
7918 This allows for execute in place and shared libraries in an environment
7919 without virtual memory management. This option implies @option{-fPIC}.
7921 @item -mno-id-shared-library
7922 @opindex mno-id-shared-library
7923 Generate code that doesn't assume ID based shared libraries are being used.
7924 This is the default.
7926 @item -mleaf-id-shared-library
7927 @opindex mleaf-id-shared-library
7928 Generate code that supports shared libraries via the library ID method,
7929 but assumes that this library or executable won't link against any other
7930 ID shared libraries. That allows the compiler to use faster code for jumps
7933 @item -mno-leaf-id-shared-library
7934 @opindex mno-leaf-id-shared-library
7935 Do not assume that the code being compiled won't link against any ID shared
7936 libraries. Slower code will be generated for jump and call insns.
7938 @item -mshared-library-id=n
7939 @opindex mshared-library-id
7940 Specified the identification number of the ID based shared library being
7941 compiled. Specifying a value of 0 will generate more compact code, specifying
7942 other values will force the allocation of that number to the current
7943 library but is no more space or time efficient than omitting this option.
7947 Generate code that allows the data segment to be located in a different
7948 area of memory from the text segment. This allows for execute in place in
7949 an environment without virtual memory management by eliminating relocations
7950 against the text section.
7953 @opindex mno-sep-data
7954 Generate code that assumes that the data segment follows the text segment.
7955 This is the default.
7958 @itemx -mno-long-calls
7959 @opindex mlong-calls
7960 @opindex mno-long-calls
7961 Tells the compiler to perform function calls by first loading the
7962 address of the function into a register and then performing a subroutine
7963 call on this register. This switch is needed if the target function
7964 will lie outside of the 24 bit addressing range of the offset based
7965 version of subroutine call instruction.
7967 This feature is not enabled by default. Specifying
7968 @option{-mno-long-calls} will restore the default behavior. Note these
7969 switches have no effect on how the compiler generates code to handle
7970 function calls via function pointers.
7974 @subsection CRIS Options
7975 @cindex CRIS Options
7977 These options are defined specifically for the CRIS ports.
7980 @item -march=@var{architecture-type}
7981 @itemx -mcpu=@var{architecture-type}
7984 Generate code for the specified architecture. The choices for
7985 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7986 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7987 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7990 @item -mtune=@var{architecture-type}
7992 Tune to @var{architecture-type} everything applicable about the generated
7993 code, except for the ABI and the set of available instructions. The
7994 choices for @var{architecture-type} are the same as for
7995 @option{-march=@var{architecture-type}}.
7997 @item -mmax-stack-frame=@var{n}
7998 @opindex mmax-stack-frame
7999 Warn when the stack frame of a function exceeds @var{n} bytes.
8001 @item -melinux-stacksize=@var{n}
8002 @opindex melinux-stacksize
8003 Only available with the @samp{cris-axis-aout} target. Arranges for
8004 indications in the program to the kernel loader that the stack of the
8005 program should be set to @var{n} bytes.
8011 The options @option{-metrax4} and @option{-metrax100} are synonyms for
8012 @option{-march=v3} and @option{-march=v8} respectively.
8014 @item -mmul-bug-workaround
8015 @itemx -mno-mul-bug-workaround
8016 @opindex mmul-bug-workaround
8017 @opindex mno-mul-bug-workaround
8018 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
8019 models where it applies. This option is active by default.
8023 Enable CRIS-specific verbose debug-related information in the assembly
8024 code. This option also has the effect to turn off the @samp{#NO_APP}
8025 formatted-code indicator to the assembler at the beginning of the
8030 Do not use condition-code results from previous instruction; always emit
8031 compare and test instructions before use of condition codes.
8033 @item -mno-side-effects
8034 @opindex mno-side-effects
8035 Do not emit instructions with side-effects in addressing modes other than
8039 @itemx -mno-stack-align
8041 @itemx -mno-data-align
8042 @itemx -mconst-align
8043 @itemx -mno-const-align
8044 @opindex mstack-align
8045 @opindex mno-stack-align
8046 @opindex mdata-align
8047 @opindex mno-data-align
8048 @opindex mconst-align
8049 @opindex mno-const-align
8050 These options (no-options) arranges (eliminate arrangements) for the
8051 stack-frame, individual data and constants to be aligned for the maximum
8052 single data access size for the chosen CPU model. The default is to
8053 arrange for 32-bit alignment. ABI details such as structure layout are
8054 not affected by these options.
8062 Similar to the stack- data- and const-align options above, these options
8063 arrange for stack-frame, writable data and constants to all be 32-bit,
8064 16-bit or 8-bit aligned. The default is 32-bit alignment.
8066 @item -mno-prologue-epilogue
8067 @itemx -mprologue-epilogue
8068 @opindex mno-prologue-epilogue
8069 @opindex mprologue-epilogue
8070 With @option{-mno-prologue-epilogue}, the normal function prologue and
8071 epilogue that sets up the stack-frame are omitted and no return
8072 instructions or return sequences are generated in the code. Use this
8073 option only together with visual inspection of the compiled code: no
8074 warnings or errors are generated when call-saved registers must be saved,
8075 or storage for local variable needs to be allocated.
8081 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8082 instruction sequences that load addresses for functions from the PLT part
8083 of the GOT rather than (traditional on other architectures) calls to the
8084 PLT@. The default is @option{-mgotplt}.
8088 Legacy no-op option only recognized with the cris-axis-aout target.
8092 Legacy no-op option only recognized with the cris-axis-elf and
8093 cris-axis-linux-gnu targets.
8097 Only recognized with the cris-axis-aout target, where it selects a
8098 GNU/linux-like multilib, include files and instruction set for
8103 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8107 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8108 to link with input-output functions from a simulator library. Code,
8109 initialized data and zero-initialized data are allocated consecutively.
8113 Like @option{-sim}, but pass linker options to locate initialized data at
8114 0x40000000 and zero-initialized data at 0x80000000.
8118 @subsection CRX Options
8121 These options are defined specifically for the CRX ports.
8127 Enable the use of multiply-accumulate instructions. Disabled by default.
8131 Push instructions will be used to pass outgoing arguments when functions
8132 are called. Enabled by default.
8135 @node Darwin Options
8136 @subsection Darwin Options
8137 @cindex Darwin options
8139 These options are defined for all architectures running the Darwin operating
8142 FSF GCC on Darwin does not create ``fat'' object files; it will create
8143 an object file for the single architecture that it was built to
8144 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8145 @option{-arch} options are used; it does so by running the compiler or
8146 linker multiple times and joining the results together with
8149 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8150 @samp{i686}) is determined by the flags that specify the ISA
8151 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8152 @option{-force_cpusubtype_ALL} option can be used to override this.
8154 The Darwin tools vary in their behavior when presented with an ISA
8155 mismatch. The assembler, @file{as}, will only permit instructions to
8156 be used that are valid for the subtype of the file it is generating,
8157 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8158 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8159 and print an error if asked to create a shared library with a less
8160 restrictive subtype than its input files (for instance, trying to put
8161 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8162 for executables, @file{ld}, will quietly give the executable the most
8163 restrictive subtype of any of its input files.
8168 Add the framework directory @var{dir} to the head of the list of
8169 directories to be searched for header files. These directories are
8170 interleaved with those specified by @option{-I} options and are
8171 scanned in a left-to-right order.
8173 A framework directory is a directory with frameworks in it. A
8174 framework is a directory with a @samp{"Headers"} and/or
8175 @samp{"PrivateHeaders"} directory contained directly in it that ends
8176 in @samp{".framework"}. The name of a framework is the name of this
8177 directory excluding the @samp{".framework"}. Headers associated with
8178 the framework are found in one of those two directories, with
8179 @samp{"Headers"} being searched first. A subframework is a framework
8180 directory that is in a framework's @samp{"Frameworks"} directory.
8181 Includes of subframework headers can only appear in a header of a
8182 framework that contains the subframework, or in a sibling subframework
8183 header. Two subframeworks are siblings if they occur in the same
8184 framework. A subframework should not have the same name as a
8185 framework, a warning will be issued if this is violated. Currently a
8186 subframework cannot have subframeworks, in the future, the mechanism
8187 may be extended to support this. The standard frameworks can be found
8188 in @samp{"/System/Library/Frameworks"} and
8189 @samp{"/Library/Frameworks"}. An example include looks like
8190 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8191 the name of the framework and header.h is found in the
8192 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8196 Emit debugging information for symbols that are used. For STABS
8197 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8198 This is by default ON@.
8202 Emit debugging information for all symbols and types.
8204 @item -mmacosx-version-min=@var{version}
8205 The earliest version of MacOS X that this executable will run on
8206 is @var{version}. Typical values of @var{version} include @code{10.1},
8207 @code{10.2}, and @code{10.3.9}.
8209 The default for this option is to make choices that seem to be most
8214 Enable kernel development mode. The @option{-mkernel} option sets
8215 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8216 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8217 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8218 applicable. This mode also sets @option{-mno-altivec},
8219 @option{-msoft-float}, @option{-fno-builtin} and
8220 @option{-mlong-branch} for PowerPC targets.
8222 @item -mone-byte-bool
8223 @opindex -mone-byte-bool
8224 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8225 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8226 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8227 option has no effect on x86.
8229 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8230 to generate code that is not binary compatible with code generated
8231 without that switch. Using this switch may require recompiling all
8232 other modules in a program, including system libraries. Use this
8233 switch to conform to a non-default data model.
8235 @item -mfix-and-continue
8236 @itemx -ffix-and-continue
8237 @itemx -findirect-data
8238 @opindex mfix-and-continue
8239 @opindex ffix-and-continue
8240 @opindex findirect-data
8241 Generate code suitable for fast turn around development. Needed to
8242 enable gdb to dynamically load @code{.o} files into already running
8243 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8244 are provided for backwards compatibility.
8248 Loads all members of static archive libraries.
8249 See man ld(1) for more information.
8251 @item -arch_errors_fatal
8252 @opindex arch_errors_fatal
8253 Cause the errors having to do with files that have the wrong architecture
8257 @opindex bind_at_load
8258 Causes the output file to be marked such that the dynamic linker will
8259 bind all undefined references when the file is loaded or launched.
8263 Produce a Mach-o bundle format file.
8264 See man ld(1) for more information.
8266 @item -bundle_loader @var{executable}
8267 @opindex bundle_loader
8268 This option specifies the @var{executable} that will be loading the build
8269 output file being linked. See man ld(1) for more information.
8272 @opindex -dynamiclib
8273 When passed this option, GCC will produce a dynamic library instead of
8274 an executable when linking, using the Darwin @file{libtool} command.
8276 @item -force_cpusubtype_ALL
8277 @opindex -force_cpusubtype_ALL
8278 This causes GCC's output file to have the @var{ALL} subtype, instead of
8279 one controlled by the @option{-mcpu} or @option{-march} option.
8281 @item -allowable_client @var{client_name}
8283 @itemx -compatibility_version
8284 @itemx -current_version
8286 @itemx -dependency-file
8288 @itemx -dylinker_install_name
8290 @itemx -exported_symbols_list
8292 @itemx -flat_namespace
8293 @itemx -force_flat_namespace
8294 @itemx -headerpad_max_install_names
8297 @itemx -install_name
8298 @itemx -keep_private_externs
8299 @itemx -multi_module
8300 @itemx -multiply_defined
8301 @itemx -multiply_defined_unused
8303 @itemx -no_dead_strip_inits_and_terms
8304 @itemx -nofixprebinding
8307 @itemx -noseglinkedit
8308 @itemx -pagezero_size
8310 @itemx -prebind_all_twolevel_modules
8311 @itemx -private_bundle
8312 @itemx -read_only_relocs
8314 @itemx -sectobjectsymbols
8318 @itemx -sectobjectsymbols
8321 @itemx -segs_read_only_addr
8322 @itemx -segs_read_write_addr
8323 @itemx -seg_addr_table
8324 @itemx -seg_addr_table_filename
8327 @itemx -segs_read_only_addr
8328 @itemx -segs_read_write_addr
8329 @itemx -single_module
8332 @itemx -sub_umbrella
8333 @itemx -twolevel_namespace
8336 @itemx -unexported_symbols_list
8337 @itemx -weak_reference_mismatches
8340 @opindex allowable_client
8341 @opindex client_name
8342 @opindex compatibility_version
8343 @opindex current_version
8345 @opindex dependency-file
8347 @opindex dylinker_install_name
8349 @opindex exported_symbols_list
8351 @opindex flat_namespace
8352 @opindex force_flat_namespace
8353 @opindex headerpad_max_install_names
8356 @opindex install_name
8357 @opindex keep_private_externs
8358 @opindex multi_module
8359 @opindex multiply_defined
8360 @opindex multiply_defined_unused
8362 @opindex no_dead_strip_inits_and_terms
8363 @opindex nofixprebinding
8364 @opindex nomultidefs
8366 @opindex noseglinkedit
8367 @opindex pagezero_size
8369 @opindex prebind_all_twolevel_modules
8370 @opindex private_bundle
8371 @opindex read_only_relocs
8373 @opindex sectobjectsymbols
8377 @opindex sectobjectsymbols
8380 @opindex segs_read_only_addr
8381 @opindex segs_read_write_addr
8382 @opindex seg_addr_table
8383 @opindex seg_addr_table_filename
8384 @opindex seglinkedit
8386 @opindex segs_read_only_addr
8387 @opindex segs_read_write_addr
8388 @opindex single_module
8390 @opindex sub_library
8391 @opindex sub_umbrella
8392 @opindex twolevel_namespace
8395 @opindex unexported_symbols_list
8396 @opindex weak_reference_mismatches
8397 @opindex whatsloaded
8399 These options are passed to the Darwin linker. The Darwin linker man page
8400 describes them in detail.
8403 @node DEC Alpha Options
8404 @subsection DEC Alpha Options
8406 These @samp{-m} options are defined for the DEC Alpha implementations:
8409 @item -mno-soft-float
8411 @opindex mno-soft-float
8412 @opindex msoft-float
8413 Use (do not use) the hardware floating-point instructions for
8414 floating-point operations. When @option{-msoft-float} is specified,
8415 functions in @file{libgcc.a} will be used to perform floating-point
8416 operations. Unless they are replaced by routines that emulate the
8417 floating-point operations, or compiled in such a way as to call such
8418 emulations routines, these routines will issue floating-point
8419 operations. If you are compiling for an Alpha without floating-point
8420 operations, you must ensure that the library is built so as not to call
8423 Note that Alpha implementations without floating-point operations are
8424 required to have floating-point registers.
8429 @opindex mno-fp-regs
8430 Generate code that uses (does not use) the floating-point register set.
8431 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8432 register set is not used, floating point operands are passed in integer
8433 registers as if they were integers and floating-point results are passed
8434 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8435 so any function with a floating-point argument or return value called by code
8436 compiled with @option{-mno-fp-regs} must also be compiled with that
8439 A typical use of this option is building a kernel that does not use,
8440 and hence need not save and restore, any floating-point registers.
8444 The Alpha architecture implements floating-point hardware optimized for
8445 maximum performance. It is mostly compliant with the IEEE floating
8446 point standard. However, for full compliance, software assistance is
8447 required. This option generates code fully IEEE compliant code
8448 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8449 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8450 defined during compilation. The resulting code is less efficient but is
8451 able to correctly support denormalized numbers and exceptional IEEE
8452 values such as not-a-number and plus/minus infinity. Other Alpha
8453 compilers call this option @option{-ieee_with_no_inexact}.
8455 @item -mieee-with-inexact
8456 @opindex mieee-with-inexact
8457 This is like @option{-mieee} except the generated code also maintains
8458 the IEEE @var{inexact-flag}. Turning on this option causes the
8459 generated code to implement fully-compliant IEEE math. In addition to
8460 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8461 macro. On some Alpha implementations the resulting code may execute
8462 significantly slower than the code generated by default. Since there is
8463 very little code that depends on the @var{inexact-flag}, you should
8464 normally not specify this option. Other Alpha compilers call this
8465 option @option{-ieee_with_inexact}.
8467 @item -mfp-trap-mode=@var{trap-mode}
8468 @opindex mfp-trap-mode
8469 This option controls what floating-point related traps are enabled.
8470 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8471 The trap mode can be set to one of four values:
8475 This is the default (normal) setting. The only traps that are enabled
8476 are the ones that cannot be disabled in software (e.g., division by zero
8480 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8484 Like @samp{u}, but the instructions are marked to be safe for software
8485 completion (see Alpha architecture manual for details).
8488 Like @samp{su}, but inexact traps are enabled as well.
8491 @item -mfp-rounding-mode=@var{rounding-mode}
8492 @opindex mfp-rounding-mode
8493 Selects the IEEE rounding mode. Other Alpha compilers call this option
8494 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8499 Normal IEEE rounding mode. Floating point numbers are rounded towards
8500 the nearest machine number or towards the even machine number in case
8504 Round towards minus infinity.
8507 Chopped rounding mode. Floating point numbers are rounded towards zero.
8510 Dynamic rounding mode. A field in the floating point control register
8511 (@var{fpcr}, see Alpha architecture reference manual) controls the
8512 rounding mode in effect. The C library initializes this register for
8513 rounding towards plus infinity. Thus, unless your program modifies the
8514 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8517 @item -mtrap-precision=@var{trap-precision}
8518 @opindex mtrap-precision
8519 In the Alpha architecture, floating point traps are imprecise. This
8520 means without software assistance it is impossible to recover from a
8521 floating trap and program execution normally needs to be terminated.
8522 GCC can generate code that can assist operating system trap handlers
8523 in determining the exact location that caused a floating point trap.
8524 Depending on the requirements of an application, different levels of
8525 precisions can be selected:
8529 Program precision. This option is the default and means a trap handler
8530 can only identify which program caused a floating point exception.
8533 Function precision. The trap handler can determine the function that
8534 caused a floating point exception.
8537 Instruction precision. The trap handler can determine the exact
8538 instruction that caused a floating point exception.
8541 Other Alpha compilers provide the equivalent options called
8542 @option{-scope_safe} and @option{-resumption_safe}.
8544 @item -mieee-conformant
8545 @opindex mieee-conformant
8546 This option marks the generated code as IEEE conformant. You must not
8547 use this option unless you also specify @option{-mtrap-precision=i} and either
8548 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8549 is to emit the line @samp{.eflag 48} in the function prologue of the
8550 generated assembly file. Under DEC Unix, this has the effect that
8551 IEEE-conformant math library routines will be linked in.
8553 @item -mbuild-constants
8554 @opindex mbuild-constants
8555 Normally GCC examines a 32- or 64-bit integer constant to
8556 see if it can construct it from smaller constants in two or three
8557 instructions. If it cannot, it will output the constant as a literal and
8558 generate code to load it from the data segment at runtime.
8560 Use this option to require GCC to construct @emph{all} integer constants
8561 using code, even if it takes more instructions (the maximum is six).
8563 You would typically use this option to build a shared library dynamic
8564 loader. Itself a shared library, it must relocate itself in memory
8565 before it can find the variables and constants in its own data segment.
8571 Select whether to generate code to be assembled by the vendor-supplied
8572 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8590 Indicate whether GCC should generate code to use the optional BWX,
8591 CIX, FIX and MAX instruction sets. The default is to use the instruction
8592 sets supported by the CPU type specified via @option{-mcpu=} option or that
8593 of the CPU on which GCC was built if none was specified.
8598 @opindex mfloat-ieee
8599 Generate code that uses (does not use) VAX F and G floating point
8600 arithmetic instead of IEEE single and double precision.
8602 @item -mexplicit-relocs
8603 @itemx -mno-explicit-relocs
8604 @opindex mexplicit-relocs
8605 @opindex mno-explicit-relocs
8606 Older Alpha assemblers provided no way to generate symbol relocations
8607 except via assembler macros. Use of these macros does not allow
8608 optimal instruction scheduling. GNU binutils as of version 2.12
8609 supports a new syntax that allows the compiler to explicitly mark
8610 which relocations should apply to which instructions. This option
8611 is mostly useful for debugging, as GCC detects the capabilities of
8612 the assembler when it is built and sets the default accordingly.
8616 @opindex msmall-data
8617 @opindex mlarge-data
8618 When @option{-mexplicit-relocs} is in effect, static data is
8619 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8620 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8621 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8622 16-bit relocations off of the @code{$gp} register. This limits the
8623 size of the small data area to 64KB, but allows the variables to be
8624 directly accessed via a single instruction.
8626 The default is @option{-mlarge-data}. With this option the data area
8627 is limited to just below 2GB@. Programs that require more than 2GB of
8628 data must use @code{malloc} or @code{mmap} to allocate the data in the
8629 heap instead of in the program's data segment.
8631 When generating code for shared libraries, @option{-fpic} implies
8632 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8636 @opindex msmall-text
8637 @opindex mlarge-text
8638 When @option{-msmall-text} is used, the compiler assumes that the
8639 code of the entire program (or shared library) fits in 4MB, and is
8640 thus reachable with a branch instruction. When @option{-msmall-data}
8641 is used, the compiler can assume that all local symbols share the
8642 same @code{$gp} value, and thus reduce the number of instructions
8643 required for a function call from 4 to 1.
8645 The default is @option{-mlarge-text}.
8647 @item -mcpu=@var{cpu_type}
8649 Set the instruction set and instruction scheduling parameters for
8650 machine type @var{cpu_type}. You can specify either the @samp{EV}
8651 style name or the corresponding chip number. GCC supports scheduling
8652 parameters for the EV4, EV5 and EV6 family of processors and will
8653 choose the default values for the instruction set from the processor
8654 you specify. If you do not specify a processor type, GCC will default
8655 to the processor on which the compiler was built.
8657 Supported values for @var{cpu_type} are
8663 Schedules as an EV4 and has no instruction set extensions.
8667 Schedules as an EV5 and has no instruction set extensions.
8671 Schedules as an EV5 and supports the BWX extension.
8676 Schedules as an EV5 and supports the BWX and MAX extensions.
8680 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8684 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8687 @item -mtune=@var{cpu_type}
8689 Set only the instruction scheduling parameters for machine type
8690 @var{cpu_type}. The instruction set is not changed.
8692 @item -mmemory-latency=@var{time}
8693 @opindex mmemory-latency
8694 Sets the latency the scheduler should assume for typical memory
8695 references as seen by the application. This number is highly
8696 dependent on the memory access patterns used by the application
8697 and the size of the external cache on the machine.
8699 Valid options for @var{time} are
8703 A decimal number representing clock cycles.
8709 The compiler contains estimates of the number of clock cycles for
8710 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8711 (also called Dcache, Scache, and Bcache), as well as to main memory.
8712 Note that L3 is only valid for EV5.
8717 @node DEC Alpha/VMS Options
8718 @subsection DEC Alpha/VMS Options
8720 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8723 @item -mvms-return-codes
8724 @opindex mvms-return-codes
8725 Return VMS condition codes from main. The default is to return POSIX
8726 style condition (e.g.@ error) codes.
8730 @subsection FRV Options
8737 Only use the first 32 general purpose registers.
8742 Use all 64 general purpose registers.
8747 Use only the first 32 floating point registers.
8752 Use all 64 floating point registers
8755 @opindex mhard-float
8757 Use hardware instructions for floating point operations.
8760 @opindex msoft-float
8762 Use library routines for floating point operations.
8767 Dynamically allocate condition code registers.
8772 Do not try to dynamically allocate condition code registers, only
8773 use @code{icc0} and @code{fcc0}.
8778 Change ABI to use double word insns.
8783 Do not use double word instructions.
8788 Use floating point double instructions.
8793 Do not use floating point double instructions.
8798 Use media instructions.
8803 Do not use media instructions.
8808 Use multiply and add/subtract instructions.
8813 Do not use multiply and add/subtract instructions.
8818 Select the FDPIC ABI, that uses function descriptors to represent
8819 pointers to functions. Without any PIC/PIE-related options, it
8820 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8821 assumes GOT entries and small data are within a 12-bit range from the
8822 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8823 are computed with 32 bits.
8826 @opindex minline-plt
8828 Enable inlining of PLT entries in function calls to functions that are
8829 not known to bind locally. It has no effect without @option{-mfdpic}.
8830 It's enabled by default if optimizing for speed and compiling for
8831 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8832 optimization option such as @option{-O3} or above is present in the
8838 Assume a large TLS segment when generating thread-local code.
8843 Do not assume a large TLS segment when generating thread-local code.
8848 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8849 that is known to be in read-only sections. It's enabled by default,
8850 except for @option{-fpic} or @option{-fpie}: even though it may help
8851 make the global offset table smaller, it trades 1 instruction for 4.
8852 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8853 one of which may be shared by multiple symbols, and it avoids the need
8854 for a GOT entry for the referenced symbol, so it's more likely to be a
8855 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8857 @item -multilib-library-pic
8858 @opindex multilib-library-pic
8860 Link with the (library, not FD) pic libraries. It's implied by
8861 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8862 @option{-fpic} without @option{-mfdpic}. You should never have to use
8868 Follow the EABI requirement of always creating a frame pointer whenever
8869 a stack frame is allocated. This option is enabled by default and can
8870 be disabled with @option{-mno-linked-fp}.
8873 @opindex mlong-calls
8875 Use indirect addressing to call functions outside the current
8876 compilation unit. This allows the functions to be placed anywhere
8877 within the 32-bit address space.
8879 @item -malign-labels
8880 @opindex malign-labels
8882 Try to align labels to an 8-byte boundary by inserting nops into the
8883 previous packet. This option only has an effect when VLIW packing
8884 is enabled. It doesn't create new packets; it merely adds nops to
8888 @opindex mlibrary-pic
8890 Generate position-independent EABI code.
8895 Use only the first four media accumulator registers.
8900 Use all eight media accumulator registers.
8905 Pack VLIW instructions.
8910 Do not pack VLIW instructions.
8915 Do not mark ABI switches in e_flags.
8920 Enable the use of conditional-move instructions (default).
8922 This switch is mainly for debugging the compiler and will likely be removed
8923 in a future version.
8925 @item -mno-cond-move
8926 @opindex mno-cond-move
8928 Disable the use of conditional-move instructions.
8930 This switch is mainly for debugging the compiler and will likely be removed
8931 in a future version.
8936 Enable the use of conditional set instructions (default).
8938 This switch is mainly for debugging the compiler and will likely be removed
8939 in a future version.
8944 Disable the use of conditional set instructions.
8946 This switch is mainly for debugging the compiler and will likely be removed
8947 in a future version.
8952 Enable the use of conditional execution (default).
8954 This switch is mainly for debugging the compiler and will likely be removed
8955 in a future version.
8957 @item -mno-cond-exec
8958 @opindex mno-cond-exec
8960 Disable the use of conditional execution.
8962 This switch is mainly for debugging the compiler and will likely be removed
8963 in a future version.
8966 @opindex mvliw-branch
8968 Run a pass to pack branches into VLIW instructions (default).
8970 This switch is mainly for debugging the compiler and will likely be removed
8971 in a future version.
8973 @item -mno-vliw-branch
8974 @opindex mno-vliw-branch
8976 Do not run a pass to pack branches into VLIW instructions.
8978 This switch is mainly for debugging the compiler and will likely be removed
8979 in a future version.
8981 @item -mmulti-cond-exec
8982 @opindex mmulti-cond-exec
8984 Enable optimization of @code{&&} and @code{||} in conditional execution
8987 This switch is mainly for debugging the compiler and will likely be removed
8988 in a future version.
8990 @item -mno-multi-cond-exec
8991 @opindex mno-multi-cond-exec
8993 Disable optimization of @code{&&} and @code{||} in conditional execution.
8995 This switch is mainly for debugging the compiler and will likely be removed
8996 in a future version.
8998 @item -mnested-cond-exec
8999 @opindex mnested-cond-exec
9001 Enable nested conditional execution optimizations (default).
9003 This switch is mainly for debugging the compiler and will likely be removed
9004 in a future version.
9006 @item -mno-nested-cond-exec
9007 @opindex mno-nested-cond-exec
9009 Disable nested conditional execution optimizations.
9011 This switch is mainly for debugging the compiler and will likely be removed
9012 in a future version.
9014 @item -moptimize-membar
9015 @opindex moptimize-membar
9017 This switch removes redundant @code{membar} instructions from the
9018 compiler generated code. It is enabled by default.
9020 @item -mno-optimize-membar
9021 @opindex mno-optimize-membar
9023 This switch disables the automatic removal of redundant @code{membar}
9024 instructions from the generated code.
9026 @item -mtomcat-stats
9027 @opindex mtomcat-stats
9029 Cause gas to print out tomcat statistics.
9031 @item -mcpu=@var{cpu}
9034 Select the processor type for which to generate code. Possible values are
9035 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
9036 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
9040 @node GNU/Linux Options
9041 @subsection GNU/Linux Options
9043 These @samp{-m} options are defined for GNU/Linux targets:
9048 Use the GNU C library instead of uClibc. This is the default except
9049 on @samp{*-*-linux-*uclibc*} targets.
9053 Use uClibc instead of the GNU C library. This is the default on
9054 @samp{*-*-linux-*uclibc*} targets.
9057 @node H8/300 Options
9058 @subsection H8/300 Options
9060 These @samp{-m} options are defined for the H8/300 implementations:
9065 Shorten some address references at link time, when possible; uses the
9066 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9067 ld, Using ld}, for a fuller description.
9071 Generate code for the H8/300H@.
9075 Generate code for the H8S@.
9079 Generate code for the H8S and H8/300H in the normal mode. This switch
9080 must be used either with @option{-mh} or @option{-ms}.
9084 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9088 Make @code{int} data 32 bits by default.
9092 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9093 The default for the H8/300H and H8S is to align longs and floats on 4
9095 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9096 This option has no effect on the H8/300.
9100 @subsection HPPA Options
9101 @cindex HPPA Options
9103 These @samp{-m} options are defined for the HPPA family of computers:
9106 @item -march=@var{architecture-type}
9108 Generate code for the specified architecture. The choices for
9109 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9110 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9111 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9112 architecture option for your machine. Code compiled for lower numbered
9113 architectures will run on higher numbered architectures, but not the
9117 @itemx -mpa-risc-1-1
9118 @itemx -mpa-risc-2-0
9119 @opindex mpa-risc-1-0
9120 @opindex mpa-risc-1-1
9121 @opindex mpa-risc-2-0
9122 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9125 @opindex mbig-switch
9126 Generate code suitable for big switch tables. Use this option only if
9127 the assembler/linker complain about out of range branches within a switch
9130 @item -mjump-in-delay
9131 @opindex mjump-in-delay
9132 Fill delay slots of function calls with unconditional jump instructions
9133 by modifying the return pointer for the function call to be the target
9134 of the conditional jump.
9136 @item -mdisable-fpregs
9137 @opindex mdisable-fpregs
9138 Prevent floating point registers from being used in any manner. This is
9139 necessary for compiling kernels which perform lazy context switching of
9140 floating point registers. If you use this option and attempt to perform
9141 floating point operations, the compiler will abort.
9143 @item -mdisable-indexing
9144 @opindex mdisable-indexing
9145 Prevent the compiler from using indexing address modes. This avoids some
9146 rather obscure problems when compiling MIG generated code under MACH@.
9148 @item -mno-space-regs
9149 @opindex mno-space-regs
9150 Generate code that assumes the target has no space registers. This allows
9151 GCC to generate faster indirect calls and use unscaled index address modes.
9153 Such code is suitable for level 0 PA systems and kernels.
9155 @item -mfast-indirect-calls
9156 @opindex mfast-indirect-calls
9157 Generate code that assumes calls never cross space boundaries. This
9158 allows GCC to emit code which performs faster indirect calls.
9160 This option will not work in the presence of shared libraries or nested
9163 @item -mfixed-range=@var{register-range}
9164 @opindex mfixed-range
9165 Generate code treating the given register range as fixed registers.
9166 A fixed register is one that the register allocator can not use. This is
9167 useful when compiling kernel code. A register range is specified as
9168 two registers separated by a dash. Multiple register ranges can be
9169 specified separated by a comma.
9171 @item -mlong-load-store
9172 @opindex mlong-load-store
9173 Generate 3-instruction load and store sequences as sometimes required by
9174 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9177 @item -mportable-runtime
9178 @opindex mportable-runtime
9179 Use the portable calling conventions proposed by HP for ELF systems.
9183 Enable the use of assembler directives only GAS understands.
9185 @item -mschedule=@var{cpu-type}
9187 Schedule code according to the constraints for the machine type
9188 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9189 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9190 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9191 proper scheduling option for your machine. The default scheduling is
9195 @opindex mlinker-opt
9196 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9197 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9198 linkers in which they give bogus error messages when linking some programs.
9201 @opindex msoft-float
9202 Generate output containing library calls for floating point.
9203 @strong{Warning:} the requisite libraries are not available for all HPPA
9204 targets. Normally the facilities of the machine's usual C compiler are
9205 used, but this cannot be done directly in cross-compilation. You must make
9206 your own arrangements to provide suitable library functions for
9207 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9208 does provide software floating point support.
9210 @option{-msoft-float} changes the calling convention in the output file;
9211 therefore, it is only useful if you compile @emph{all} of a program with
9212 this option. In particular, you need to compile @file{libgcc.a}, the
9213 library that comes with GCC, with @option{-msoft-float} in order for
9218 Generate the predefine, @code{_SIO}, for server IO@. The default is
9219 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9220 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9221 options are available under HP-UX and HI-UX@.
9225 Use GNU ld specific options. This passes @option{-shared} to ld when
9226 building a shared library. It is the default when GCC is configured,
9227 explicitly or implicitly, with the GNU linker. This option does not
9228 have any affect on which ld is called, it only changes what parameters
9229 are passed to that ld. The ld that is called is determined by the
9230 @option{--with-ld} configure option, GCC's program search path, and
9231 finally by the user's @env{PATH}. The linker used by GCC can be printed
9232 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9233 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9237 Use HP ld specific options. This passes @option{-b} to ld when building
9238 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9239 links. It is the default when GCC is configured, explicitly or
9240 implicitly, with the HP linker. This option does not have any affect on
9241 which ld is called, it only changes what parameters are passed to that
9242 ld. The ld that is called is determined by the @option{--with-ld}
9243 configure option, GCC's program search path, and finally by the user's
9244 @env{PATH}. The linker used by GCC can be printed using @samp{which
9245 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9246 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9249 @opindex mno-long-calls
9250 Generate code that uses long call sequences. This ensures that a call
9251 is always able to reach linker generated stubs. The default is to generate
9252 long calls only when the distance from the call site to the beginning
9253 of the function or translation unit, as the case may be, exceeds a
9254 predefined limit set by the branch type being used. The limits for
9255 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9256 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9259 Distances are measured from the beginning of functions when using the
9260 @option{-ffunction-sections} option, or when using the @option{-mgas}
9261 and @option{-mno-portable-runtime} options together under HP-UX with
9264 It is normally not desirable to use this option as it will degrade
9265 performance. However, it may be useful in large applications,
9266 particularly when partial linking is used to build the application.
9268 The types of long calls used depends on the capabilities of the
9269 assembler and linker, and the type of code being generated. The
9270 impact on systems that support long absolute calls, and long pic
9271 symbol-difference or pc-relative calls should be relatively small.
9272 However, an indirect call is used on 32-bit ELF systems in pic code
9273 and it is quite long.
9275 @item -munix=@var{unix-std}
9277 Generate compiler predefines and select a startfile for the specified
9278 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9279 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9280 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9281 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9282 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9285 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9286 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9287 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9288 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9289 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9290 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9292 It is @emph{important} to note that this option changes the interfaces
9293 for various library routines. It also affects the operational behavior
9294 of the C library. Thus, @emph{extreme} care is needed in using this
9297 Library code that is intended to operate with more than one UNIX
9298 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9299 as appropriate. Most GNU software doesn't provide this capability.
9303 Suppress the generation of link options to search libdld.sl when the
9304 @option{-static} option is specified on HP-UX 10 and later.
9308 The HP-UX implementation of setlocale in libc has a dependency on
9309 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9310 when the @option{-static} option is specified, special link options
9311 are needed to resolve this dependency.
9313 On HP-UX 10 and later, the GCC driver adds the necessary options to
9314 link with libdld.sl when the @option{-static} option is specified.
9315 This causes the resulting binary to be dynamic. On the 64-bit port,
9316 the linkers generate dynamic binaries by default in any case. The
9317 @option{-nolibdld} option can be used to prevent the GCC driver from
9318 adding these link options.
9322 Add support for multithreading with the @dfn{dce thread} library
9323 under HP-UX@. This option sets flags for both the preprocessor and
9327 @node i386 and x86-64 Options
9328 @subsection Intel 386 and AMD x86-64 Options
9329 @cindex i386 Options
9330 @cindex x86-64 Options
9331 @cindex Intel 386 Options
9332 @cindex AMD x86-64 Options
9334 These @samp{-m} options are defined for the i386 and x86-64 family of
9338 @item -mtune=@var{cpu-type}
9340 Tune to @var{cpu-type} everything applicable about the generated code, except
9341 for the ABI and the set of available instructions. The choices for
9345 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9346 If you know the CPU on which your code will run, then you should use
9347 the corresponding @option{-mtune} option instead of
9348 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9349 of your application will have, then you should use this option.
9351 As new processors are deployed in the marketplace, the behavior of this
9352 option will change. Therefore, if you upgrade to a newer version of
9353 GCC, the code generated option will change to reflect the processors
9354 that were most common when that version of GCC was released.
9356 There is no @option{-march=generic} option because @option{-march}
9357 indicates the instruction set the compiler can use, and there is no
9358 generic instruction set applicable to all processors. In contrast,
9359 @option{-mtune} indicates the processor (or, in this case, collection of
9360 processors) for which the code is optimized.
9362 This selects the CPU to tune for at compilation time by determining
9363 the processor type of the compiling machine. Using @option{-mtune=native}
9364 will produce code optimized for the local machine under the constraints
9365 of the selected instruction set. Using @option{-march=native} will
9366 enable all instruction subsets supported by the local machine (hence
9367 the result might not run on different machines).
9369 Original Intel's i386 CPU@.
9371 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9373 Intel Pentium CPU with no MMX support.
9375 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9377 Intel PentiumPro CPU@.
9379 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9380 instruction set will be used, so the code will run on all i686 family chips.
9382 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9383 @item pentium3, pentium3m
9384 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9387 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9388 support. Used by Centrino notebooks.
9389 @item pentium4, pentium4m
9390 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9392 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9395 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9396 SSE2 and SSE3 instruction set support.
9398 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9399 instruction set support.
9401 AMD K6 CPU with MMX instruction set support.
9403 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9404 @item athlon, athlon-tbird
9405 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9407 @item athlon-4, athlon-xp, athlon-mp
9408 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9409 instruction set support.
9410 @item k8, opteron, athlon64, athlon-fx
9411 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9412 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9414 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9417 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9418 instruction set support.
9420 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9421 implemented for this chip.)
9423 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9424 implemented for this chip.)
9426 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9429 While picking a specific @var{cpu-type} will schedule things appropriately
9430 for that particular chip, the compiler will not generate any code that
9431 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9434 @item -march=@var{cpu-type}
9436 Generate instructions for the machine type @var{cpu-type}. The choices
9437 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9438 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9440 @item -mcpu=@var{cpu-type}
9442 A deprecated synonym for @option{-mtune}.
9451 @opindex mpentiumpro
9452 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9453 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9454 These synonyms are deprecated.
9456 @item -mfpmath=@var{unit}
9458 Generate floating point arithmetics for selected unit @var{unit}. The choices
9463 Use the standard 387 floating point coprocessor present majority of chips and
9464 emulated otherwise. Code compiled with this option will run almost everywhere.
9465 The temporary results are computed in 80bit precision instead of precision
9466 specified by the type resulting in slightly different results compared to most
9467 of other chips. See @option{-ffloat-store} for more detailed description.
9469 This is the default choice for i386 compiler.
9472 Use scalar floating point instructions present in the SSE instruction set.
9473 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9474 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9475 instruction set supports only single precision arithmetics, thus the double and
9476 extended precision arithmetics is still done using 387. Later version, present
9477 only in Pentium4 and the future AMD x86-64 chips supports double precision
9480 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9481 or @option{-msse2} switches to enable SSE extensions and make this option
9482 effective. For the x86-64 compiler, these extensions are enabled by default.
9484 The resulting code should be considerably faster in the majority of cases and avoid
9485 the numerical instability problems of 387 code, but may break some existing
9486 code that expects temporaries to be 80bit.
9488 This is the default choice for the x86-64 compiler.
9491 Attempt to utilize both instruction sets at once. This effectively double the
9492 amount of available registers and on chips with separate execution units for
9493 387 and SSE the execution resources too. Use this option with care, as it is
9494 still experimental, because the GCC register allocator does not model separate
9495 functional units well resulting in instable performance.
9498 @item -masm=@var{dialect}
9499 @opindex masm=@var{dialect}
9500 Output asm instructions using selected @var{dialect}. Supported
9501 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9502 not support @samp{intel}.
9507 @opindex mno-ieee-fp
9508 Control whether or not the compiler uses IEEE floating point
9509 comparisons. These handle correctly the case where the result of a
9510 comparison is unordered.
9513 @opindex msoft-float
9514 Generate output containing library calls for floating point.
9515 @strong{Warning:} the requisite libraries are not part of GCC@.
9516 Normally the facilities of the machine's usual C compiler are used, but
9517 this can't be done directly in cross-compilation. You must make your
9518 own arrangements to provide suitable library functions for
9521 On machines where a function returns floating point results in the 80387
9522 register stack, some floating point opcodes may be emitted even if
9523 @option{-msoft-float} is used.
9525 @item -mno-fp-ret-in-387
9526 @opindex mno-fp-ret-in-387
9527 Do not use the FPU registers for return values of functions.
9529 The usual calling convention has functions return values of types
9530 @code{float} and @code{double} in an FPU register, even if there
9531 is no FPU@. The idea is that the operating system should emulate
9534 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9535 in ordinary CPU registers instead.
9537 @item -mno-fancy-math-387
9538 @opindex mno-fancy-math-387
9539 Some 387 emulators do not support the @code{sin}, @code{cos} and
9540 @code{sqrt} instructions for the 387. Specify this option to avoid
9541 generating those instructions. This option is the default on FreeBSD,
9542 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9543 indicates that the target cpu will always have an FPU and so the
9544 instruction will not need emulation. As of revision 2.6.1, these
9545 instructions are not generated unless you also use the
9546 @option{-funsafe-math-optimizations} switch.
9548 @item -malign-double
9549 @itemx -mno-align-double
9550 @opindex malign-double
9551 @opindex mno-align-double
9552 Control whether GCC aligns @code{double}, @code{long double}, and
9553 @code{long long} variables on a two word boundary or a one word
9554 boundary. Aligning @code{double} variables on a two word boundary will
9555 produce code that runs somewhat faster on a @samp{Pentium} at the
9556 expense of more memory.
9558 On x86-64, @option{-malign-double} is enabled by default.
9560 @strong{Warning:} if you use the @option{-malign-double} switch,
9561 structures containing the above types will be aligned differently than
9562 the published application binary interface specifications for the 386
9563 and will not be binary compatible with structures in code compiled
9564 without that switch.
9566 @item -m96bit-long-double
9567 @itemx -m128bit-long-double
9568 @opindex m96bit-long-double
9569 @opindex m128bit-long-double
9570 These switches control the size of @code{long double} type. The i386
9571 application binary interface specifies the size to be 96 bits,
9572 so @option{-m96bit-long-double} is the default in 32 bit mode.
9574 Modern architectures (Pentium and newer) would prefer @code{long double}
9575 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9576 conforming to the ABI, this would not be possible. So specifying a
9577 @option{-m128bit-long-double} will align @code{long double}
9578 to a 16 byte boundary by padding the @code{long double} with an additional
9581 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9582 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9584 Notice that neither of these options enable any extra precision over the x87
9585 standard of 80 bits for a @code{long double}.
9587 @strong{Warning:} if you override the default value for your target ABI, the
9588 structures and arrays containing @code{long double} variables will change
9589 their size as well as function calling convention for function taking
9590 @code{long double} will be modified. Hence they will not be binary
9591 compatible with arrays or structures in code compiled without that switch.
9593 @item -mmlarge-data-threshold=@var{number}
9594 @opindex mlarge-data-threshold=@var{number}
9595 When @option{-mcmodel=medium} is specified, the data greater than
9596 @var{threshold} are placed in large data section. This value must be the
9597 same across all object linked into the binary and defaults to 65535.
9600 @itemx -mno-svr3-shlib
9601 @opindex msvr3-shlib
9602 @opindex mno-svr3-shlib
9603 Control whether GCC places uninitialized local variables into the
9604 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9605 into @code{bss}. These options are meaningful only on System V Release 3.
9609 Use a different function-calling convention, in which functions that
9610 take a fixed number of arguments return with the @code{ret} @var{num}
9611 instruction, which pops their arguments while returning. This saves one
9612 instruction in the caller since there is no need to pop the arguments
9615 You can specify that an individual function is called with this calling
9616 sequence with the function attribute @samp{stdcall}. You can also
9617 override the @option{-mrtd} option by using the function attribute
9618 @samp{cdecl}. @xref{Function Attributes}.
9620 @strong{Warning:} this calling convention is incompatible with the one
9621 normally used on Unix, so you cannot use it if you need to call
9622 libraries compiled with the Unix compiler.
9624 Also, you must provide function prototypes for all functions that
9625 take variable numbers of arguments (including @code{printf});
9626 otherwise incorrect code will be generated for calls to those
9629 In addition, seriously incorrect code will result if you call a
9630 function with too many arguments. (Normally, extra arguments are
9631 harmlessly ignored.)
9633 @item -mregparm=@var{num}
9635 Control how many registers are used to pass integer arguments. By
9636 default, no registers are used to pass arguments, and at most 3
9637 registers can be used. You can control this behavior for a specific
9638 function by using the function attribute @samp{regparm}.
9639 @xref{Function Attributes}.
9641 @strong{Warning:} if you use this switch, and
9642 @var{num} is nonzero, then you must build all modules with the same
9643 value, including any libraries. This includes the system libraries and
9647 @opindex msseregparm
9648 Use SSE register passing conventions for float and double arguments
9649 and return values. You can control this behavior for a specific
9650 function by using the function attribute @samp{sseregparm}.
9651 @xref{Function Attributes}.
9653 @strong{Warning:} if you use this switch then you must build all
9654 modules with the same value, including any libraries. This includes
9655 the system libraries and startup modules.
9657 @item -mstackrealign
9658 @opindex mstackrealign
9659 Realign the stack at entry. On the Intel x86, the
9660 @option{-mstackrealign} option will generate an alternate prologue and
9661 epilogue that realigns the runtime stack. This supports mixing legacy
9662 codes that keep a 4-byte aligned stack with modern codes that keep a
9663 16-byte stack for SSE compatibility. The alternate prologue and
9664 epilogue are slower and bigger than the regular ones, and the
9665 alternate prologue requires an extra scratch register; this lowers the
9666 number of registers available if used in conjunction with the
9667 @code{regparm} attribute. The @option{-mstackrealign} option is
9668 incompatible with the nested function prologue; this is considered a
9669 hard error. See also the attribute @code{force_align_arg_pointer},
9670 applicable to individual functions.
9672 @item -mpreferred-stack-boundary=@var{num}
9673 @opindex mpreferred-stack-boundary
9674 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9675 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9676 the default is 4 (16 bytes or 128 bits).
9678 On Pentium and PentiumPro, @code{double} and @code{long double} values
9679 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9680 suffer significant run time performance penalties. On Pentium III, the
9681 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9682 properly if it is not 16 byte aligned.
9684 To ensure proper alignment of this values on the stack, the stack boundary
9685 must be as aligned as that required by any value stored on the stack.
9686 Further, every function must be generated such that it keeps the stack
9687 aligned. Thus calling a function compiled with a higher preferred
9688 stack boundary from a function compiled with a lower preferred stack
9689 boundary will most likely misalign the stack. It is recommended that
9690 libraries that use callbacks always use the default setting.
9692 This extra alignment does consume extra stack space, and generally
9693 increases code size. Code that is sensitive to stack space usage, such
9694 as embedded systems and operating system kernels, may want to reduce the
9695 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9715 These switches enable or disable the use of instructions in the MMX,
9716 SSE, SSE2, SSE3, SSSE3 or 3DNow! extended instruction sets.
9717 These extensions are also available as built-in functions: see
9718 @ref{X86 Built-in Functions}, for details of the functions enabled and
9719 disabled by these switches.
9721 To have SSE/SSE2 instructions generated automatically from floating-point
9722 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9724 These options will enable GCC to use these extended instructions in
9725 generated code, even without @option{-mfpmath=sse}. Applications which
9726 perform runtime CPU detection must compile separate files for each
9727 supported architecture, using the appropriate flags. In particular,
9728 the file containing the CPU detection code should be compiled without
9732 @itemx -mno-push-args
9734 @opindex mno-push-args
9735 Use PUSH operations to store outgoing parameters. This method is shorter
9736 and usually equally fast as method using SUB/MOV operations and is enabled
9737 by default. In some cases disabling it may improve performance because of
9738 improved scheduling and reduced dependencies.
9740 @item -maccumulate-outgoing-args
9741 @opindex maccumulate-outgoing-args
9742 If enabled, the maximum amount of space required for outgoing arguments will be
9743 computed in the function prologue. This is faster on most modern CPUs
9744 because of reduced dependencies, improved scheduling and reduced stack usage
9745 when preferred stack boundary is not equal to 2. The drawback is a notable
9746 increase in code size. This switch implies @option{-mno-push-args}.
9750 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9751 on thread-safe exception handling must compile and link all code with the
9752 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9753 @option{-D_MT}; when linking, it links in a special thread helper library
9754 @option{-lmingwthrd} which cleans up per thread exception handling data.
9756 @item -mno-align-stringops
9757 @opindex mno-align-stringops
9758 Do not align destination of inlined string operations. This switch reduces
9759 code size and improves performance in case the destination is already aligned,
9760 but GCC doesn't know about it.
9762 @item -minline-all-stringops
9763 @opindex minline-all-stringops
9764 By default GCC inlines string operations only when destination is known to be
9765 aligned at least to 4 byte boundary. This enables more inlining, increase code
9766 size, but may improve performance of code that depends on fast memcpy, strlen
9767 and memset for short lengths.
9769 @item -minline-stringops-dynamically
9770 @opindex minline-stringops-dynamically
9771 For string operation of unknown size, inline runtime checks so for small
9772 blocks inline code is used, while for large blocks library call is used.
9774 @item -mstringop-strategy=@var{alg}
9775 @opindex mstringop-strategy=@var{alg}
9776 Overwrite internal decision heuristic about particular algorithm to inline
9777 string operation with. The allowed values are @code{rep_byte},
9778 @code{rep_4byte}, @code{rep_8byte} for expanding using i386 @code{rep} prefix
9779 of specified size, @code{byte_loop}, @code{loop}, @code{unrolled_loop} for
9780 expanding inline loop, @code{libcall} for always expanding library call.
9782 @item -momit-leaf-frame-pointer
9783 @opindex momit-leaf-frame-pointer
9784 Don't keep the frame pointer in a register for leaf functions. This
9785 avoids the instructions to save, set up and restore frame pointers and
9786 makes an extra register available in leaf functions. The option
9787 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9788 which might make debugging harder.
9790 @item -mtls-direct-seg-refs
9791 @itemx -mno-tls-direct-seg-refs
9792 @opindex mtls-direct-seg-refs
9793 Controls whether TLS variables may be accessed with offsets from the
9794 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9795 or whether the thread base pointer must be added. Whether or not this
9796 is legal depends on the operating system, and whether it maps the
9797 segment to cover the entire TLS area.
9799 For systems that use GNU libc, the default is on.
9802 These @samp{-m} switches are supported in addition to the above
9803 on AMD x86-64 processors in 64-bit environments.
9810 Generate code for a 32-bit or 64-bit environment.
9811 The 32-bit environment sets int, long and pointer to 32 bits and
9812 generates code that runs on any i386 system.
9813 The 64-bit environment sets int to 32 bits and long and pointer
9814 to 64 bits and generates code for AMD's x86-64 architecture.
9817 @opindex no-red-zone
9818 Do not use a so called red zone for x86-64 code. The red zone is mandated
9819 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9820 stack pointer that will not be modified by signal or interrupt handlers
9821 and therefore can be used for temporary data without adjusting the stack
9822 pointer. The flag @option{-mno-red-zone} disables this red zone.
9824 @item -mcmodel=small
9825 @opindex mcmodel=small
9826 Generate code for the small code model: the program and its symbols must
9827 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9828 Programs can be statically or dynamically linked. This is the default
9831 @item -mcmodel=kernel
9832 @opindex mcmodel=kernel
9833 Generate code for the kernel code model. The kernel runs in the
9834 negative 2 GB of the address space.
9835 This model has to be used for Linux kernel code.
9837 @item -mcmodel=medium
9838 @opindex mcmodel=medium
9839 Generate code for the medium model: The program is linked in the lower 2
9840 GB of the address space but symbols can be located anywhere in the
9841 address space. Programs can be statically or dynamically linked, but
9842 building of shared libraries are not supported with the medium model.
9844 @item -mcmodel=large
9845 @opindex mcmodel=large
9846 Generate code for the large model: This model makes no assumptions
9847 about addresses and sizes of sections. Currently GCC does not implement
9852 @subsection IA-64 Options
9853 @cindex IA-64 Options
9855 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9859 @opindex mbig-endian
9860 Generate code for a big endian target. This is the default for HP-UX@.
9862 @item -mlittle-endian
9863 @opindex mlittle-endian
9864 Generate code for a little endian target. This is the default for AIX5
9871 Generate (or don't) code for the GNU assembler. This is the default.
9872 @c Also, this is the default if the configure option @option{--with-gnu-as}
9879 Generate (or don't) code for the GNU linker. This is the default.
9880 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9885 Generate code that does not use a global pointer register. The result
9886 is not position independent code, and violates the IA-64 ABI@.
9888 @item -mvolatile-asm-stop
9889 @itemx -mno-volatile-asm-stop
9890 @opindex mvolatile-asm-stop
9891 @opindex mno-volatile-asm-stop
9892 Generate (or don't) a stop bit immediately before and after volatile asm
9895 @item -mregister-names
9896 @itemx -mno-register-names
9897 @opindex mregister-names
9898 @opindex mno-register-names
9899 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9900 the stacked registers. This may make assembler output more readable.
9906 Disable (or enable) optimizations that use the small data section. This may
9907 be useful for working around optimizer bugs.
9910 @opindex mconstant-gp
9911 Generate code that uses a single constant global pointer value. This is
9912 useful when compiling kernel code.
9916 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9917 This is useful when compiling firmware code.
9919 @item -minline-float-divide-min-latency
9920 @opindex minline-float-divide-min-latency
9921 Generate code for inline divides of floating point values
9922 using the minimum latency algorithm.
9924 @item -minline-float-divide-max-throughput
9925 @opindex minline-float-divide-max-throughput
9926 Generate code for inline divides of floating point values
9927 using the maximum throughput algorithm.
9929 @item -minline-int-divide-min-latency
9930 @opindex minline-int-divide-min-latency
9931 Generate code for inline divides of integer values
9932 using the minimum latency algorithm.
9934 @item -minline-int-divide-max-throughput
9935 @opindex minline-int-divide-max-throughput
9936 Generate code for inline divides of integer values
9937 using the maximum throughput algorithm.
9939 @item -minline-sqrt-min-latency
9940 @opindex minline-sqrt-min-latency
9941 Generate code for inline square roots
9942 using the minimum latency algorithm.
9944 @item -minline-sqrt-max-throughput
9945 @opindex minline-sqrt-max-throughput
9946 Generate code for inline square roots
9947 using the maximum throughput algorithm.
9949 @item -mno-dwarf2-asm
9951 @opindex mno-dwarf2-asm
9952 @opindex mdwarf2-asm
9953 Don't (or do) generate assembler code for the DWARF2 line number debugging
9954 info. This may be useful when not using the GNU assembler.
9956 @item -mearly-stop-bits
9957 @itemx -mno-early-stop-bits
9958 @opindex mearly-stop-bits
9959 @opindex mno-early-stop-bits
9960 Allow stop bits to be placed earlier than immediately preceding the
9961 instruction that triggered the stop bit. This can improve instruction
9962 scheduling, but does not always do so.
9964 @item -mfixed-range=@var{register-range}
9965 @opindex mfixed-range
9966 Generate code treating the given register range as fixed registers.
9967 A fixed register is one that the register allocator can not use. This is
9968 useful when compiling kernel code. A register range is specified as
9969 two registers separated by a dash. Multiple register ranges can be
9970 specified separated by a comma.
9972 @item -mtls-size=@var{tls-size}
9974 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9977 @item -mtune=@var{cpu-type}
9979 Tune the instruction scheduling for a particular CPU, Valid values are
9980 itanium, itanium1, merced, itanium2, and mckinley.
9986 Add support for multithreading using the POSIX threads library. This
9987 option sets flags for both the preprocessor and linker. It does
9988 not affect the thread safety of object code produced by the compiler or
9989 that of libraries supplied with it. These are HP-UX specific flags.
9995 Generate code for a 32-bit or 64-bit environment.
9996 The 32-bit environment sets int, long and pointer to 32 bits.
9997 The 64-bit environment sets int to 32 bits and long and pointer
9998 to 64 bits. These are HP-UX specific flags.
10000 @item -mno-sched-br-data-spec
10001 @itemx -msched-br-data-spec
10002 @opindex -mno-sched-br-data-spec
10003 @opindex -msched-br-data-spec
10004 (Dis/En)able data speculative scheduling before reload.
10005 This will result in generation of the ld.a instructions and
10006 the corresponding check instructions (ld.c / chk.a).
10007 The default is 'disable'.
10009 @item -msched-ar-data-spec
10010 @itemx -mno-sched-ar-data-spec
10011 @opindex -msched-ar-data-spec
10012 @opindex -mno-sched-ar-data-spec
10013 (En/Dis)able data speculative scheduling after reload.
10014 This will result in generation of the ld.a instructions and
10015 the corresponding check instructions (ld.c / chk.a).
10016 The default is 'enable'.
10018 @item -mno-sched-control-spec
10019 @itemx -msched-control-spec
10020 @opindex -mno-sched-control-spec
10021 @opindex -msched-control-spec
10022 (Dis/En)able control speculative scheduling. This feature is
10023 available only during region scheduling (i.e. before reload).
10024 This will result in generation of the ld.s instructions and
10025 the corresponding check instructions chk.s .
10026 The default is 'disable'.
10028 @item -msched-br-in-data-spec
10029 @itemx -mno-sched-br-in-data-spec
10030 @opindex -msched-br-in-data-spec
10031 @opindex -mno-sched-br-in-data-spec
10032 (En/Dis)able speculative scheduling of the instructions that
10033 are dependent on the data speculative loads before reload.
10034 This is effective only with @option{-msched-br-data-spec} enabled.
10035 The default is 'enable'.
10037 @item -msched-ar-in-data-spec
10038 @itemx -mno-sched-ar-in-data-spec
10039 @opindex -msched-ar-in-data-spec
10040 @opindex -mno-sched-ar-in-data-spec
10041 (En/Dis)able speculative scheduling of the instructions that
10042 are dependent on the data speculative loads after reload.
10043 This is effective only with @option{-msched-ar-data-spec} enabled.
10044 The default is 'enable'.
10046 @item -msched-in-control-spec
10047 @itemx -mno-sched-in-control-spec
10048 @opindex -msched-in-control-spec
10049 @opindex -mno-sched-in-control-spec
10050 (En/Dis)able speculative scheduling of the instructions that
10051 are dependent on the control speculative loads.
10052 This is effective only with @option{-msched-control-spec} enabled.
10053 The default is 'enable'.
10056 @itemx -mno-sched-ldc
10057 @opindex -msched-ldc
10058 @opindex -mno-sched-ldc
10059 (En/Dis)able use of simple data speculation checks ld.c .
10060 If disabled, only chk.a instructions will be emitted to check
10061 data speculative loads.
10062 The default is 'enable'.
10064 @item -mno-sched-control-ldc
10065 @itemx -msched-control-ldc
10066 @opindex -mno-sched-control-ldc
10067 @opindex -msched-control-ldc
10068 (Dis/En)able use of ld.c instructions to check control speculative loads.
10069 If enabled, in case of control speculative load with no speculatively
10070 scheduled dependent instructions this load will be emitted as ld.sa and
10071 ld.c will be used to check it.
10072 The default is 'disable'.
10074 @item -mno-sched-spec-verbose
10075 @itemx -msched-spec-verbose
10076 @opindex -mno-sched-spec-verbose
10077 @opindex -msched-spec-verbose
10078 (Dis/En)able printing of the information about speculative motions.
10080 @item -mno-sched-prefer-non-data-spec-insns
10081 @itemx -msched-prefer-non-data-spec-insns
10082 @opindex -mno-sched-prefer-non-data-spec-insns
10083 @opindex -msched-prefer-non-data-spec-insns
10084 If enabled, data speculative instructions will be chosen for schedule
10085 only if there are no other choices at the moment. This will make
10086 the use of the data speculation much more conservative.
10087 The default is 'disable'.
10089 @item -mno-sched-prefer-non-control-spec-insns
10090 @itemx -msched-prefer-non-control-spec-insns
10091 @opindex -mno-sched-prefer-non-control-spec-insns
10092 @opindex -msched-prefer-non-control-spec-insns
10093 If enabled, control speculative instructions will be chosen for schedule
10094 only if there are no other choices at the moment. This will make
10095 the use of the control speculation much more conservative.
10096 The default is 'disable'.
10098 @item -mno-sched-count-spec-in-critical-path
10099 @itemx -msched-count-spec-in-critical-path
10100 @opindex -mno-sched-count-spec-in-critical-path
10101 @opindex -msched-count-spec-in-critical-path
10102 If enabled, speculative dependencies will be considered during
10103 computation of the instructions priorities. This will make the use of the
10104 speculation a bit more conservative.
10105 The default is 'disable'.
10110 @subsection M32C Options
10111 @cindex M32C options
10114 @item -mcpu=@var{name}
10116 Select the CPU for which code is generated. @var{name} may be one of
10117 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10118 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10119 the M32C/80 series.
10123 Specifies that the program will be run on the simulator. This causes
10124 an alternate runtime library to be linked in which supports, for
10125 example, file I/O. You must not use this option when generating
10126 programs that will run on real hardware; you must provide your own
10127 runtime library for whatever I/O functions are needed.
10129 @item -memregs=@var{number}
10131 Specifies the number of memory-based pseudo-registers GCC will use
10132 during code generation. These pseudo-registers will be used like real
10133 registers, so there is a tradeoff between GCC's ability to fit the
10134 code into available registers, and the performance penalty of using
10135 memory instead of registers. Note that all modules in a program must
10136 be compiled with the same value for this option. Because of that, you
10137 must not use this option with the default runtime libraries gcc
10142 @node M32R/D Options
10143 @subsection M32R/D Options
10144 @cindex M32R/D options
10146 These @option{-m} options are defined for Renesas M32R/D architectures:
10151 Generate code for the M32R/2@.
10155 Generate code for the M32R/X@.
10159 Generate code for the M32R@. This is the default.
10161 @item -mmodel=small
10162 @opindex mmodel=small
10163 Assume all objects live in the lower 16MB of memory (so that their addresses
10164 can be loaded with the @code{ld24} instruction), and assume all subroutines
10165 are reachable with the @code{bl} instruction.
10166 This is the default.
10168 The addressability of a particular object can be set with the
10169 @code{model} attribute.
10171 @item -mmodel=medium
10172 @opindex mmodel=medium
10173 Assume objects may be anywhere in the 32-bit address space (the compiler
10174 will generate @code{seth/add3} instructions to load their addresses), and
10175 assume all subroutines are reachable with the @code{bl} instruction.
10177 @item -mmodel=large
10178 @opindex mmodel=large
10179 Assume objects may be anywhere in the 32-bit address space (the compiler
10180 will generate @code{seth/add3} instructions to load their addresses), and
10181 assume subroutines may not be reachable with the @code{bl} instruction
10182 (the compiler will generate the much slower @code{seth/add3/jl}
10183 instruction sequence).
10186 @opindex msdata=none
10187 Disable use of the small data area. Variables will be put into
10188 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10189 @code{section} attribute has been specified).
10190 This is the default.
10192 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10193 Objects may be explicitly put in the small data area with the
10194 @code{section} attribute using one of these sections.
10196 @item -msdata=sdata
10197 @opindex msdata=sdata
10198 Put small global and static data in the small data area, but do not
10199 generate special code to reference them.
10202 @opindex msdata=use
10203 Put small global and static data in the small data area, and generate
10204 special instructions to reference them.
10208 @cindex smaller data references
10209 Put global and static objects less than or equal to @var{num} bytes
10210 into the small data or bss sections instead of the normal data or bss
10211 sections. The default value of @var{num} is 8.
10212 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10213 for this option to have any effect.
10215 All modules should be compiled with the same @option{-G @var{num}} value.
10216 Compiling with different values of @var{num} may or may not work; if it
10217 doesn't the linker will give an error message---incorrect code will not be
10222 Makes the M32R specific code in the compiler display some statistics
10223 that might help in debugging programs.
10225 @item -malign-loops
10226 @opindex malign-loops
10227 Align all loops to a 32-byte boundary.
10229 @item -mno-align-loops
10230 @opindex mno-align-loops
10231 Do not enforce a 32-byte alignment for loops. This is the default.
10233 @item -missue-rate=@var{number}
10234 @opindex missue-rate=@var{number}
10235 Issue @var{number} instructions per cycle. @var{number} can only be 1
10238 @item -mbranch-cost=@var{number}
10239 @opindex mbranch-cost=@var{number}
10240 @var{number} can only be 1 or 2. If it is 1 then branches will be
10241 preferred over conditional code, if it is 2, then the opposite will
10244 @item -mflush-trap=@var{number}
10245 @opindex mflush-trap=@var{number}
10246 Specifies the trap number to use to flush the cache. The default is
10247 12. Valid numbers are between 0 and 15 inclusive.
10249 @item -mno-flush-trap
10250 @opindex mno-flush-trap
10251 Specifies that the cache cannot be flushed by using a trap.
10253 @item -mflush-func=@var{name}
10254 @opindex mflush-func=@var{name}
10255 Specifies the name of the operating system function to call to flush
10256 the cache. The default is @emph{_flush_cache}, but a function call
10257 will only be used if a trap is not available.
10259 @item -mno-flush-func
10260 @opindex mno-flush-func
10261 Indicates that there is no OS function for flushing the cache.
10265 @node M680x0 Options
10266 @subsection M680x0 Options
10267 @cindex M680x0 options
10269 These are the @samp{-m} options defined for the 68000 series. The default
10270 values for these options depends on which style of 68000 was selected when
10271 the compiler was configured; the defaults for the most common choices are
10279 Generate output for a 68000. This is the default
10280 when the compiler is configured for 68000-based systems.
10282 Use this option for microcontrollers with a 68000 or EC000 core,
10283 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10289 Generate output for a 68020. This is the default
10290 when the compiler is configured for 68020-based systems.
10294 Generate output containing 68881 instructions for floating point.
10295 This is the default for most 68020 systems unless @option{--nfp} was
10296 specified when the compiler was configured.
10300 Generate output for a 68030. This is the default when the compiler is
10301 configured for 68030-based systems.
10305 Generate output for a 68040. This is the default when the compiler is
10306 configured for 68040-based systems.
10308 This option inhibits the use of 68881/68882 instructions that have to be
10309 emulated by software on the 68040. Use this option if your 68040 does not
10310 have code to emulate those instructions.
10314 Generate output for a 68060. This is the default when the compiler is
10315 configured for 68060-based systems.
10317 This option inhibits the use of 68020 and 68881/68882 instructions that
10318 have to be emulated by software on the 68060. Use this option if your 68060
10319 does not have code to emulate those instructions.
10323 Generate output for a CPU32. This is the default
10324 when the compiler is configured for CPU32-based systems.
10326 Use this option for microcontrollers with a
10327 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10328 68336, 68340, 68341, 68349 and 68360.
10332 Generate output for a 520X ``coldfire'' family cpu. This is the default
10333 when the compiler is configured for 520X-based systems.
10335 Use this option for microcontroller with a 5200 core, including
10336 the MCF5202, MCF5203, MCF5204 and MCF5202.
10340 Generate output for a ColdFire V4e family cpu (e.g.@: 547x/548x).
10341 This includes use of hardware floating point instructions.
10345 Generate output for a 68040, without using any of the new instructions.
10346 This results in code which can run relatively efficiently on either a
10347 68020/68881 or a 68030 or a 68040. The generated code does use the
10348 68881 instructions that are emulated on the 68040.
10352 Generate output for a 68060, without using any of the new instructions.
10353 This results in code which can run relatively efficiently on either a
10354 68020/68881 or a 68030 or a 68040. The generated code does use the
10355 68881 instructions that are emulated on the 68060.
10358 @opindex msoft-float
10359 Generate output containing library calls for floating point.
10360 @strong{Warning:} the requisite libraries are not available for all m68k
10361 targets. Normally the facilities of the machine's usual C compiler are
10362 used, but this can't be done directly in cross-compilation. You must
10363 make your own arrangements to provide suitable library functions for
10364 cross-compilation. The embedded targets @samp{m68k-*-aout} and
10365 @samp{m68k-*-coff} do provide software floating point support.
10369 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10370 Additionally, parameters passed on the stack are also aligned to a
10371 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10374 @opindex mnobitfield
10375 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10376 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10380 Do use the bit-field instructions. The @option{-m68020} option implies
10381 @option{-mbitfield}. This is the default if you use a configuration
10382 designed for a 68020.
10386 Use a different function-calling convention, in which functions
10387 that take a fixed number of arguments return with the @code{rtd}
10388 instruction, which pops their arguments while returning. This
10389 saves one instruction in the caller since there is no need to pop
10390 the arguments there.
10392 This calling convention is incompatible with the one normally
10393 used on Unix, so you cannot use it if you need to call libraries
10394 compiled with the Unix compiler.
10396 Also, you must provide function prototypes for all functions that
10397 take variable numbers of arguments (including @code{printf});
10398 otherwise incorrect code will be generated for calls to those
10401 In addition, seriously incorrect code will result if you call a
10402 function with too many arguments. (Normally, extra arguments are
10403 harmlessly ignored.)
10405 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10406 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10409 @itemx -mno-align-int
10410 @opindex malign-int
10411 @opindex mno-align-int
10412 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10413 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10414 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10415 Aligning variables on 32-bit boundaries produces code that runs somewhat
10416 faster on processors with 32-bit busses at the expense of more memory.
10418 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10419 align structures containing the above types differently than
10420 most published application binary interface specifications for the m68k.
10424 Use the pc-relative addressing mode of the 68000 directly, instead of
10425 using a global offset table. At present, this option implies @option{-fpic},
10426 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10427 not presently supported with @option{-mpcrel}, though this could be supported for
10428 68020 and higher processors.
10430 @item -mno-strict-align
10431 @itemx -mstrict-align
10432 @opindex mno-strict-align
10433 @opindex mstrict-align
10434 Do not (do) assume that unaligned memory references will be handled by
10438 Generate code that allows the data segment to be located in a different
10439 area of memory from the text segment. This allows for execute in place in
10440 an environment without virtual memory management. This option implies
10443 @item -mno-sep-data
10444 Generate code that assumes that the data segment follows the text segment.
10445 This is the default.
10447 @item -mid-shared-library
10448 Generate code that supports shared libraries via the library ID method.
10449 This allows for execute in place and shared libraries in an environment
10450 without virtual memory management. This option implies @option{-fPIC}.
10452 @item -mno-id-shared-library
10453 Generate code that doesn't assume ID based shared libraries are being used.
10454 This is the default.
10456 @item -mshared-library-id=n
10457 Specified the identification number of the ID based shared library being
10458 compiled. Specifying a value of 0 will generate more compact code, specifying
10459 other values will force the allocation of that number to the current
10460 library but is no more space or time efficient than omitting this option.
10464 @node M68hc1x Options
10465 @subsection M68hc1x Options
10466 @cindex M68hc1x options
10468 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10469 microcontrollers. The default values for these options depends on
10470 which style of microcontroller was selected when the compiler was configured;
10471 the defaults for the most common choices are given below.
10478 Generate output for a 68HC11. This is the default
10479 when the compiler is configured for 68HC11-based systems.
10485 Generate output for a 68HC12. This is the default
10486 when the compiler is configured for 68HC12-based systems.
10492 Generate output for a 68HCS12.
10494 @item -mauto-incdec
10495 @opindex mauto-incdec
10496 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10503 Enable the use of 68HC12 min and max instructions.
10506 @itemx -mno-long-calls
10507 @opindex mlong-calls
10508 @opindex mno-long-calls
10509 Treat all calls as being far away (near). If calls are assumed to be
10510 far away, the compiler will use the @code{call} instruction to
10511 call a function and the @code{rtc} instruction for returning.
10515 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10517 @item -msoft-reg-count=@var{count}
10518 @opindex msoft-reg-count
10519 Specify the number of pseudo-soft registers which are used for the
10520 code generation. The maximum number is 32. Using more pseudo-soft
10521 register may or may not result in better code depending on the program.
10522 The default is 4 for 68HC11 and 2 for 68HC12.
10526 @node MCore Options
10527 @subsection MCore Options
10528 @cindex MCore options
10530 These are the @samp{-m} options defined for the Motorola M*Core
10536 @itemx -mno-hardlit
10538 @opindex mno-hardlit
10539 Inline constants into the code stream if it can be done in two
10540 instructions or less.
10546 Use the divide instruction. (Enabled by default).
10548 @item -mrelax-immediate
10549 @itemx -mno-relax-immediate
10550 @opindex mrelax-immediate
10551 @opindex mno-relax-immediate
10552 Allow arbitrary sized immediates in bit operations.
10554 @item -mwide-bitfields
10555 @itemx -mno-wide-bitfields
10556 @opindex mwide-bitfields
10557 @opindex mno-wide-bitfields
10558 Always treat bit-fields as int-sized.
10560 @item -m4byte-functions
10561 @itemx -mno-4byte-functions
10562 @opindex m4byte-functions
10563 @opindex mno-4byte-functions
10564 Force all functions to be aligned to a four byte boundary.
10566 @item -mcallgraph-data
10567 @itemx -mno-callgraph-data
10568 @opindex mcallgraph-data
10569 @opindex mno-callgraph-data
10570 Emit callgraph information.
10573 @itemx -mno-slow-bytes
10574 @opindex mslow-bytes
10575 @opindex mno-slow-bytes
10576 Prefer word access when reading byte quantities.
10578 @item -mlittle-endian
10579 @itemx -mbig-endian
10580 @opindex mlittle-endian
10581 @opindex mbig-endian
10582 Generate code for a little endian target.
10588 Generate code for the 210 processor.
10592 @subsection MIPS Options
10593 @cindex MIPS options
10599 Generate big-endian code.
10603 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10606 @item -march=@var{arch}
10608 Generate code that will run on @var{arch}, which can be the name of a
10609 generic MIPS ISA, or the name of a particular processor.
10611 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10612 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10613 The processor names are:
10614 @samp{4kc}, @samp{4km}, @samp{4kp},
10615 @samp{4kec}, @samp{4kem}, @samp{4kep},
10616 @samp{5kc}, @samp{5kf},
10618 @samp{24kc}, @samp{24kf}, @samp{24kx},
10619 @samp{24kec}, @samp{24kef}, @samp{24kex},
10620 @samp{34kc}, @samp{34kf}, @samp{34kx},
10623 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10624 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10625 @samp{rm7000}, @samp{rm9000},
10628 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10629 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10630 The special value @samp{from-abi} selects the
10631 most compatible architecture for the selected ABI (that is,
10632 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10634 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10635 (for example, @samp{-march=r2k}). Prefixes are optional, and
10636 @samp{vr} may be written @samp{r}.
10638 GCC defines two macros based on the value of this option. The first
10639 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10640 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10641 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10642 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10643 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10645 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10646 above. In other words, it will have the full prefix and will not
10647 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10648 the macro names the resolved architecture (either @samp{"mips1"} or
10649 @samp{"mips3"}). It names the default architecture when no
10650 @option{-march} option is given.
10652 @item -mtune=@var{arch}
10654 Optimize for @var{arch}. Among other things, this option controls
10655 the way instructions are scheduled, and the perceived cost of arithmetic
10656 operations. The list of @var{arch} values is the same as for
10659 When this option is not used, GCC will optimize for the processor
10660 specified by @option{-march}. By using @option{-march} and
10661 @option{-mtune} together, it is possible to generate code that will
10662 run on a family of processors, but optimize the code for one
10663 particular member of that family.
10665 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10666 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10667 @samp{-march} ones described above.
10671 Equivalent to @samp{-march=mips1}.
10675 Equivalent to @samp{-march=mips2}.
10679 Equivalent to @samp{-march=mips3}.
10683 Equivalent to @samp{-march=mips4}.
10687 Equivalent to @samp{-march=mips32}.
10691 Equivalent to @samp{-march=mips32r2}.
10695 Equivalent to @samp{-march=mips64}.
10700 @opindex mno-mips16
10701 Generate (do not generate) MIPS16 code. If GCC is targetting a
10702 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10714 Generate code for the given ABI@.
10716 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10717 generates 64-bit code when you select a 64-bit architecture, but you
10718 can use @option{-mgp32} to get 32-bit code instead.
10720 For information about the O64 ABI, see
10721 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10723 GCC supports a variant of the o32 ABI in which floating-point registers
10724 are 64 rather than 32 bits wide. You can select this combination with
10725 @option{-mabi=32} @option{-mfp64}. This ABI relies on the @samp{mthc1}
10726 and @samp{mfhc1} instructions and is therefore only supported for
10727 MIPS32R2 processors.
10729 The register assignments for arguments and return values remain the
10730 same, but each scalar value is passed in a single 64-bit register
10731 rather than a pair of 32-bit registers. For example, scalar
10732 floating-point values are returned in @samp{$f0} only, not a
10733 @samp{$f0}/@samp{$f1} pair. The set of call-saved registers also
10734 remains the same, but all 64 bits are saved.
10737 @itemx -mno-abicalls
10739 @opindex mno-abicalls
10740 Generate (do not generate) code that is suitable for SVR4-style
10741 dynamic objects. @option{-mabicalls} is the default for SVR4-based
10746 Generate (do not generate) code that is fully position-independent,
10747 and that can therefore be linked into shared libraries. This option
10748 only affects @option{-mabicalls}.
10750 All @option{-mabicalls} code has traditionally been position-independent,
10751 regardless of options like @option{-fPIC} and @option{-fpic}. However,
10752 as an extension, the GNU toolchain allows executables to use absolute
10753 accesses for locally-binding symbols. It can also use shorter GP
10754 initialization sequences and generate direct calls to locally-defined
10755 functions. This mode is selected by @option{-mno-shared}.
10757 @option{-mno-shared} depends on binutils 2.16 or higher and generates
10758 objects that can only be linked by the GNU linker. However, the option
10759 does not affect the ABI of the final executable; it only affects the ABI
10760 of relocatable objects. Using @option{-mno-shared} will generally make
10761 executables both smaller and quicker.
10763 @option{-mshared} is the default.
10769 Lift (do not lift) the usual restrictions on the size of the global
10772 GCC normally uses a single instruction to load values from the GOT@.
10773 While this is relatively efficient, it will only work if the GOT
10774 is smaller than about 64k. Anything larger will cause the linker
10775 to report an error such as:
10777 @cindex relocation truncated to fit (MIPS)
10779 relocation truncated to fit: R_MIPS_GOT16 foobar
10782 If this happens, you should recompile your code with @option{-mxgot}.
10783 It should then work with very large GOTs, although it will also be
10784 less efficient, since it will take three instructions to fetch the
10785 value of a global symbol.
10787 Note that some linkers can create multiple GOTs. If you have such a
10788 linker, you should only need to use @option{-mxgot} when a single object
10789 file accesses more than 64k's worth of GOT entries. Very few do.
10791 These options have no effect unless GCC is generating position
10796 Assume that general-purpose registers are 32 bits wide.
10800 Assume that general-purpose registers are 64 bits wide.
10804 Assume that floating-point registers are 32 bits wide.
10808 Assume that floating-point registers are 64 bits wide.
10811 @opindex mhard-float
10812 Use floating-point coprocessor instructions.
10815 @opindex msoft-float
10816 Do not use floating-point coprocessor instructions. Implement
10817 floating-point calculations using library calls instead.
10819 @item -msingle-float
10820 @opindex msingle-float
10821 Assume that the floating-point coprocessor only supports single-precision
10824 @itemx -mdouble-float
10825 @opindex mdouble-float
10826 Assume that the floating-point coprocessor supports double-precision
10827 operations. This is the default.
10833 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10835 @itemx -mpaired-single
10836 @itemx -mno-paired-single
10837 @opindex mpaired-single
10838 @opindex mno-paired-single
10839 Use (do not use) paired-single floating-point instructions.
10840 @xref{MIPS Paired-Single Support}. This option can only be used
10841 when generating 64-bit code and requires hardware floating-point
10842 support to be enabled.
10847 @opindex mno-mips3d
10848 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10849 The option @option{-mips3d} implies @option{-mpaired-single}.
10853 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10854 an explanation of the default and the way that the pointer size is
10859 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10861 The default size of @code{int}s, @code{long}s and pointers depends on
10862 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10863 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10864 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10865 or the same size as integer registers, whichever is smaller.
10871 Assume (do not assume) that all symbols have 32-bit values, regardless
10872 of the selected ABI@. This option is useful in combination with
10873 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10874 to generate shorter and faster references to symbolic addresses.
10878 @cindex smaller data references (MIPS)
10879 @cindex gp-relative references (MIPS)
10880 Put global and static items less than or equal to @var{num} bytes into
10881 the small data or bss section instead of the normal data or bss section.
10882 This allows the data to be accessed using a single instruction.
10884 All modules should be compiled with the same @option{-G @var{num}}
10887 @item -membedded-data
10888 @itemx -mno-embedded-data
10889 @opindex membedded-data
10890 @opindex mno-embedded-data
10891 Allocate variables to the read-only data section first if possible, then
10892 next in the small data section if possible, otherwise in data. This gives
10893 slightly slower code than the default, but reduces the amount of RAM required
10894 when executing, and thus may be preferred for some embedded systems.
10896 @item -muninit-const-in-rodata
10897 @itemx -mno-uninit-const-in-rodata
10898 @opindex muninit-const-in-rodata
10899 @opindex mno-uninit-const-in-rodata
10900 Put uninitialized @code{const} variables in the read-only data section.
10901 This option is only meaningful in conjunction with @option{-membedded-data}.
10903 @item -msplit-addresses
10904 @itemx -mno-split-addresses
10905 @opindex msplit-addresses
10906 @opindex mno-split-addresses
10907 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10908 relocation operators. This option has been superseded by
10909 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10911 @item -mexplicit-relocs
10912 @itemx -mno-explicit-relocs
10913 @opindex mexplicit-relocs
10914 @opindex mno-explicit-relocs
10915 Use (do not use) assembler relocation operators when dealing with symbolic
10916 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10917 is to use assembler macros instead.
10919 @option{-mexplicit-relocs} is the default if GCC was configured
10920 to use an assembler that supports relocation operators.
10922 @item -mcheck-zero-division
10923 @itemx -mno-check-zero-division
10924 @opindex mcheck-zero-division
10925 @opindex mno-check-zero-division
10926 Trap (do not trap) on integer division by zero. The default is
10927 @option{-mcheck-zero-division}.
10929 @item -mdivide-traps
10930 @itemx -mdivide-breaks
10931 @opindex mdivide-traps
10932 @opindex mdivide-breaks
10933 MIPS systems check for division by zero by generating either a
10934 conditional trap or a break instruction. Using traps results in
10935 smaller code, but is only supported on MIPS II and later. Also, some
10936 versions of the Linux kernel have a bug that prevents trap from
10937 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10938 allow conditional traps on architectures that support them and
10939 @option{-mdivide-breaks} to force the use of breaks.
10941 The default is usually @option{-mdivide-traps}, but this can be
10942 overridden at configure time using @option{--with-divide=breaks}.
10943 Divide-by-zero checks can be completely disabled using
10944 @option{-mno-check-zero-division}.
10949 @opindex mno-memcpy
10950 Force (do not force) the use of @code{memcpy()} for non-trivial block
10951 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10952 most constant-sized copies.
10955 @itemx -mno-long-calls
10956 @opindex mlong-calls
10957 @opindex mno-long-calls
10958 Disable (do not disable) use of the @code{jal} instruction. Calling
10959 functions using @code{jal} is more efficient but requires the caller
10960 and callee to be in the same 256 megabyte segment.
10962 This option has no effect on abicalls code. The default is
10963 @option{-mno-long-calls}.
10969 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10970 instructions, as provided by the R4650 ISA@.
10973 @itemx -mno-fused-madd
10974 @opindex mfused-madd
10975 @opindex mno-fused-madd
10976 Enable (disable) use of the floating point multiply-accumulate
10977 instructions, when they are available. The default is
10978 @option{-mfused-madd}.
10980 When multiply-accumulate instructions are used, the intermediate
10981 product is calculated to infinite precision and is not subject to
10982 the FCSR Flush to Zero bit. This may be undesirable in some
10987 Tell the MIPS assembler to not run its preprocessor over user
10988 assembler files (with a @samp{.s} suffix) when assembling them.
10991 @itemx -mno-fix-r4000
10992 @opindex mfix-r4000
10993 @opindex mno-fix-r4000
10994 Work around certain R4000 CPU errata:
10997 A double-word or a variable shift may give an incorrect result if executed
10998 immediately after starting an integer division.
11000 A double-word or a variable shift may give an incorrect result if executed
11001 while an integer multiplication is in progress.
11003 An integer division may give an incorrect result if started in a delay slot
11004 of a taken branch or a jump.
11008 @itemx -mno-fix-r4400
11009 @opindex mfix-r4400
11010 @opindex mno-fix-r4400
11011 Work around certain R4400 CPU errata:
11014 A double-word or a variable shift may give an incorrect result if executed
11015 immediately after starting an integer division.
11019 @itemx -mno-fix-vr4120
11020 @opindex mfix-vr4120
11021 Work around certain VR4120 errata:
11024 @code{dmultu} does not always produce the correct result.
11026 @code{div} and @code{ddiv} do not always produce the correct result if one
11027 of the operands is negative.
11029 The workarounds for the division errata rely on special functions in
11030 @file{libgcc.a}. At present, these functions are only provided by
11031 the @code{mips64vr*-elf} configurations.
11033 Other VR4120 errata require a nop to be inserted between certain pairs of
11034 instructions. These errata are handled by the assembler, not by GCC itself.
11037 @opindex mfix-vr4130
11038 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
11039 workarounds are implemented by the assembler rather than by GCC,
11040 although GCC will avoid using @code{mflo} and @code{mfhi} if the
11041 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
11042 instructions are available instead.
11045 @itemx -mno-fix-sb1
11047 Work around certain SB-1 CPU core errata.
11048 (This flag currently works around the SB-1 revision 2
11049 ``F1'' and ``F2'' floating point errata.)
11051 @item -mflush-func=@var{func}
11052 @itemx -mno-flush-func
11053 @opindex mflush-func
11054 Specifies the function to call to flush the I and D caches, or to not
11055 call any such function. If called, the function must take the same
11056 arguments as the common @code{_flush_func()}, that is, the address of the
11057 memory range for which the cache is being flushed, the size of the
11058 memory range, and the number 3 (to flush both caches). The default
11059 depends on the target GCC was configured for, but commonly is either
11060 @samp{_flush_func} or @samp{__cpu_flush}.
11062 @item -mbranch-likely
11063 @itemx -mno-branch-likely
11064 @opindex mbranch-likely
11065 @opindex mno-branch-likely
11066 Enable or disable use of Branch Likely instructions, regardless of the
11067 default for the selected architecture. By default, Branch Likely
11068 instructions may be generated if they are supported by the selected
11069 architecture. An exception is for the MIPS32 and MIPS64 architectures
11070 and processors which implement those architectures; for those, Branch
11071 Likely instructions will not be generated by default because the MIPS32
11072 and MIPS64 architectures specifically deprecate their use.
11074 @item -mfp-exceptions
11075 @itemx -mno-fp-exceptions
11076 @opindex mfp-exceptions
11077 Specifies whether FP exceptions are enabled. This affects how we schedule
11078 FP instructions for some processors. The default is that FP exceptions are
11081 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11082 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11085 @item -mvr4130-align
11086 @itemx -mno-vr4130-align
11087 @opindex mvr4130-align
11088 The VR4130 pipeline is two-way superscalar, but can only issue two
11089 instructions together if the first one is 8-byte aligned. When this
11090 option is enabled, GCC will align pairs of instructions that it
11091 thinks should execute in parallel.
11093 This option only has an effect when optimizing for the VR4130.
11094 It normally makes code faster, but at the expense of making it bigger.
11095 It is enabled by default at optimization level @option{-O3}.
11099 @subsection MMIX Options
11100 @cindex MMIX Options
11102 These options are defined for the MMIX:
11106 @itemx -mno-libfuncs
11108 @opindex mno-libfuncs
11109 Specify that intrinsic library functions are being compiled, passing all
11110 values in registers, no matter the size.
11113 @itemx -mno-epsilon
11115 @opindex mno-epsilon
11116 Generate floating-point comparison instructions that compare with respect
11117 to the @code{rE} epsilon register.
11119 @item -mabi=mmixware
11121 @opindex mabi-mmixware
11123 Generate code that passes function parameters and return values that (in
11124 the called function) are seen as registers @code{$0} and up, as opposed to
11125 the GNU ABI which uses global registers @code{$231} and up.
11127 @item -mzero-extend
11128 @itemx -mno-zero-extend
11129 @opindex mzero-extend
11130 @opindex mno-zero-extend
11131 When reading data from memory in sizes shorter than 64 bits, use (do not
11132 use) zero-extending load instructions by default, rather than
11133 sign-extending ones.
11136 @itemx -mno-knuthdiv
11138 @opindex mno-knuthdiv
11139 Make the result of a division yielding a remainder have the same sign as
11140 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
11141 remainder follows the sign of the dividend. Both methods are
11142 arithmetically valid, the latter being almost exclusively used.
11144 @item -mtoplevel-symbols
11145 @itemx -mno-toplevel-symbols
11146 @opindex mtoplevel-symbols
11147 @opindex mno-toplevel-symbols
11148 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
11149 code can be used with the @code{PREFIX} assembly directive.
11153 Generate an executable in the ELF format, rather than the default
11154 @samp{mmo} format used by the @command{mmix} simulator.
11156 @item -mbranch-predict
11157 @itemx -mno-branch-predict
11158 @opindex mbranch-predict
11159 @opindex mno-branch-predict
11160 Use (do not use) the probable-branch instructions, when static branch
11161 prediction indicates a probable branch.
11163 @item -mbase-addresses
11164 @itemx -mno-base-addresses
11165 @opindex mbase-addresses
11166 @opindex mno-base-addresses
11167 Generate (do not generate) code that uses @emph{base addresses}. Using a
11168 base address automatically generates a request (handled by the assembler
11169 and the linker) for a constant to be set up in a global register. The
11170 register is used for one or more base address requests within the range 0
11171 to 255 from the value held in the register. The generally leads to short
11172 and fast code, but the number of different data items that can be
11173 addressed is limited. This means that a program that uses lots of static
11174 data may require @option{-mno-base-addresses}.
11176 @item -msingle-exit
11177 @itemx -mno-single-exit
11178 @opindex msingle-exit
11179 @opindex mno-single-exit
11180 Force (do not force) generated code to have a single exit point in each
11184 @node MN10300 Options
11185 @subsection MN10300 Options
11186 @cindex MN10300 options
11188 These @option{-m} options are defined for Matsushita MN10300 architectures:
11193 Generate code to avoid bugs in the multiply instructions for the MN10300
11194 processors. This is the default.
11196 @item -mno-mult-bug
11197 @opindex mno-mult-bug
11198 Do not generate code to avoid bugs in the multiply instructions for the
11199 MN10300 processors.
11203 Generate code which uses features specific to the AM33 processor.
11207 Do not generate code which uses features specific to the AM33 processor. This
11210 @item -mreturn-pointer-on-d0
11211 @opindex mreturn-pointer-on-d0
11212 When generating a function which returns a pointer, return the pointer
11213 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11214 only in a0, and attempts to call such functions without a prototype
11215 would result in errors. Note that this option is on by default; use
11216 @option{-mno-return-pointer-on-d0} to disable it.
11220 Do not link in the C run-time initialization object file.
11224 Indicate to the linker that it should perform a relaxation optimization pass
11225 to shorten branches, calls and absolute memory addresses. This option only
11226 has an effect when used on the command line for the final link step.
11228 This option makes symbolic debugging impossible.
11232 @subsection MT Options
11235 These @option{-m} options are defined for Morpho MT architectures:
11239 @item -march=@var{cpu-type}
11241 Generate code that will run on @var{cpu-type}, which is the name of a system
11242 representing a certain processor type. Possible values for
11243 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11244 @samp{ms1-16-003} and @samp{ms2}.
11246 When this option is not used, the default is @option{-march=ms1-16-002}.
11250 Use byte loads and stores when generating code.
11254 Do not use byte loads and stores when generating code.
11258 Use simulator runtime
11262 Do not link in the C run-time initialization object file
11263 @file{crti.o}. Other run-time initialization and termination files
11264 such as @file{startup.o} and @file{exit.o} are still included on the
11265 linker command line.
11269 @node PDP-11 Options
11270 @subsection PDP-11 Options
11271 @cindex PDP-11 Options
11273 These options are defined for the PDP-11:
11278 Use hardware FPP floating point. This is the default. (FIS floating
11279 point on the PDP-11/40 is not supported.)
11282 @opindex msoft-float
11283 Do not use hardware floating point.
11287 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11291 Return floating-point results in memory. This is the default.
11295 Generate code for a PDP-11/40.
11299 Generate code for a PDP-11/45. This is the default.
11303 Generate code for a PDP-11/10.
11305 @item -mbcopy-builtin
11306 @opindex bcopy-builtin
11307 Use inline @code{movmemhi} patterns for copying memory. This is the
11312 Do not use inline @code{movmemhi} patterns for copying memory.
11318 Use 16-bit @code{int}. This is the default.
11324 Use 32-bit @code{int}.
11327 @itemx -mno-float32
11329 @opindex mno-float32
11330 Use 64-bit @code{float}. This is the default.
11333 @itemx -mno-float64
11335 @opindex mno-float64
11336 Use 32-bit @code{float}.
11340 Use @code{abshi2} pattern. This is the default.
11344 Do not use @code{abshi2} pattern.
11346 @item -mbranch-expensive
11347 @opindex mbranch-expensive
11348 Pretend that branches are expensive. This is for experimenting with
11349 code generation only.
11351 @item -mbranch-cheap
11352 @opindex mbranch-cheap
11353 Do not pretend that branches are expensive. This is the default.
11357 Generate code for a system with split I&D@.
11361 Generate code for a system without split I&D@. This is the default.
11365 Use Unix assembler syntax. This is the default when configured for
11366 @samp{pdp11-*-bsd}.
11370 Use DEC assembler syntax. This is the default when configured for any
11371 PDP-11 target other than @samp{pdp11-*-bsd}.
11374 @node PowerPC Options
11375 @subsection PowerPC Options
11376 @cindex PowerPC options
11378 These are listed under @xref{RS/6000 and PowerPC Options}.
11380 @node RS/6000 and PowerPC Options
11381 @subsection IBM RS/6000 and PowerPC Options
11382 @cindex RS/6000 and PowerPC Options
11383 @cindex IBM RS/6000 and PowerPC Options
11385 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11392 @itemx -mno-powerpc
11393 @itemx -mpowerpc-gpopt
11394 @itemx -mno-powerpc-gpopt
11395 @itemx -mpowerpc-gfxopt
11396 @itemx -mno-powerpc-gfxopt
11398 @itemx -mno-powerpc64
11402 @itemx -mno-popcntb
11410 @opindex mno-power2
11412 @opindex mno-powerpc
11413 @opindex mpowerpc-gpopt
11414 @opindex mno-powerpc-gpopt
11415 @opindex mpowerpc-gfxopt
11416 @opindex mno-powerpc-gfxopt
11417 @opindex mpowerpc64
11418 @opindex mno-powerpc64
11422 @opindex mno-popcntb
11426 @opindex mno-mfpgpr
11427 GCC supports two related instruction set architectures for the
11428 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11429 instructions supported by the @samp{rios} chip set used in the original
11430 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11431 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11432 the IBM 4xx, 6xx, and follow-on microprocessors.
11434 Neither architecture is a subset of the other. However there is a
11435 large common subset of instructions supported by both. An MQ
11436 register is included in processors supporting the POWER architecture.
11438 You use these options to specify which instructions are available on the
11439 processor you are using. The default value of these options is
11440 determined when configuring GCC@. Specifying the
11441 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11442 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11443 rather than the options listed above.
11445 The @option{-mpower} option allows GCC to generate instructions that
11446 are found only in the POWER architecture and to use the MQ register.
11447 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11448 to generate instructions that are present in the POWER2 architecture but
11449 not the original POWER architecture.
11451 The @option{-mpowerpc} option allows GCC to generate instructions that
11452 are found only in the 32-bit subset of the PowerPC architecture.
11453 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11454 GCC to use the optional PowerPC architecture instructions in the
11455 General Purpose group, including floating-point square root. Specifying
11456 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11457 use the optional PowerPC architecture instructions in the Graphics
11458 group, including floating-point select.
11460 The @option{-mmfcrf} option allows GCC to generate the move from
11461 condition register field instruction implemented on the POWER4
11462 processor and other processors that support the PowerPC V2.01
11464 The @option{-mpopcntb} option allows GCC to generate the popcount and
11465 double precision FP reciprocal estimate instruction implemented on the
11466 POWER5 processor and other processors that support the PowerPC V2.02
11468 The @option{-mfprnd} option allows GCC to generate the FP round to
11469 integer instructions implemented on the POWER5+ processor and other
11470 processors that support the PowerPC V2.03 architecture.
11471 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
11472 general purpose register instructions implemented on the POWER6X
11473 processor and other processors that support the extended PowerPC V2.05
11476 The @option{-mpowerpc64} option allows GCC to generate the additional
11477 64-bit instructions that are found in the full PowerPC64 architecture
11478 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11479 @option{-mno-powerpc64}.
11481 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11482 will use only the instructions in the common subset of both
11483 architectures plus some special AIX common-mode calls, and will not use
11484 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11485 permits GCC to use any instruction from either architecture and to
11486 allow use of the MQ register; specify this for the Motorola MPC601.
11488 @item -mnew-mnemonics
11489 @itemx -mold-mnemonics
11490 @opindex mnew-mnemonics
11491 @opindex mold-mnemonics
11492 Select which mnemonics to use in the generated assembler code. With
11493 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11494 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11495 assembler mnemonics defined for the POWER architecture. Instructions
11496 defined in only one architecture have only one mnemonic; GCC uses that
11497 mnemonic irrespective of which of these options is specified.
11499 GCC defaults to the mnemonics appropriate for the architecture in
11500 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11501 value of these option. Unless you are building a cross-compiler, you
11502 should normally not specify either @option{-mnew-mnemonics} or
11503 @option{-mold-mnemonics}, but should instead accept the default.
11505 @item -mcpu=@var{cpu_type}
11507 Set architecture type, register usage, choice of mnemonics, and
11508 instruction scheduling parameters for machine type @var{cpu_type}.
11509 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11510 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11511 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11512 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11513 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11514 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11515 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11516 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11517 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
11518 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11520 @option{-mcpu=common} selects a completely generic processor. Code
11521 generated under this option will run on any POWER or PowerPC processor.
11522 GCC will use only the instructions in the common subset of both
11523 architectures, and will not use the MQ register. GCC assumes a generic
11524 processor model for scheduling purposes.
11526 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11527 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11528 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11529 types, with an appropriate, generic processor model assumed for
11530 scheduling purposes.
11532 The other options specify a specific processor. Code generated under
11533 those options will run best on that processor, and may not run at all on
11536 The @option{-mcpu} options automatically enable or disable the
11537 following options: @option{-maltivec}, @option{-mfprnd},
11538 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11539 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11540 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11541 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw},
11542 @option{-mdlmzb}, @option{-mmfpgpr}.
11543 The particular options set for any particular CPU will vary between
11544 compiler versions, depending on what setting seems to produce optimal
11545 code for that CPU; it doesn't necessarily reflect the actual hardware's
11546 capabilities. If you wish to set an individual option to a particular
11547 value, you may specify it after the @option{-mcpu} option, like
11548 @samp{-mcpu=970 -mno-altivec}.
11550 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11551 not enabled or disabled by the @option{-mcpu} option at present because
11552 AIX does not have full support for these options. You may still
11553 enable or disable them individually if you're sure it'll work in your
11556 @item -mtune=@var{cpu_type}
11558 Set the instruction scheduling parameters for machine type
11559 @var{cpu_type}, but do not set the architecture type, register usage, or
11560 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11561 values for @var{cpu_type} are used for @option{-mtune} as for
11562 @option{-mcpu}. If both are specified, the code generated will use the
11563 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11564 scheduling parameters set by @option{-mtune}.
11570 Generate code to compute division as reciprocal estimate and iterative
11571 refinement, creating opportunities for increased throughput. This
11572 feature requires: optional PowerPC Graphics instruction set for single
11573 precision and FRE instruction for double precision, assuming divides
11574 cannot generate user-visible traps, and the domain values not include
11575 Infinities, denormals or zero denominator.
11578 @itemx -mno-altivec
11580 @opindex mno-altivec
11581 Generate code that uses (does not use) AltiVec instructions, and also
11582 enable the use of built-in functions that allow more direct access to
11583 the AltiVec instruction set. You may also need to set
11584 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11590 @opindex mno-vrsave
11591 Generate VRSAVE instructions when generating AltiVec code.
11594 @opindex msecure-plt
11595 Generate code that allows ld and ld.so to build executables and shared
11596 libraries with non-exec .plt and .got sections. This is a PowerPC
11597 32-bit SYSV ABI option.
11601 Generate code that uses a BSS .plt section that ld.so fills in, and
11602 requires .plt and .got sections that are both writable and executable.
11603 This is a PowerPC 32-bit SYSV ABI option.
11609 This switch enables or disables the generation of ISEL instructions.
11611 @item -misel=@var{yes/no}
11612 This switch has been deprecated. Use @option{-misel} and
11613 @option{-mno-isel} instead.
11619 This switch enables or disables the generation of SPE simd
11622 @item -mspe=@var{yes/no}
11623 This option has been deprecated. Use @option{-mspe} and
11624 @option{-mno-spe} instead.
11626 @item -mfloat-gprs=@var{yes/single/double/no}
11627 @itemx -mfloat-gprs
11628 @opindex mfloat-gprs
11629 This switch enables or disables the generation of floating point
11630 operations on the general purpose registers for architectures that
11633 The argument @var{yes} or @var{single} enables the use of
11634 single-precision floating point operations.
11636 The argument @var{double} enables the use of single and
11637 double-precision floating point operations.
11639 The argument @var{no} disables floating point operations on the
11640 general purpose registers.
11642 This option is currently only available on the MPC854x.
11648 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11649 targets (including GNU/Linux). The 32-bit environment sets int, long
11650 and pointer to 32 bits and generates code that runs on any PowerPC
11651 variant. The 64-bit environment sets int to 32 bits and long and
11652 pointer to 64 bits, and generates code for PowerPC64, as for
11653 @option{-mpowerpc64}.
11656 @itemx -mno-fp-in-toc
11657 @itemx -mno-sum-in-toc
11658 @itemx -mminimal-toc
11660 @opindex mno-fp-in-toc
11661 @opindex mno-sum-in-toc
11662 @opindex mminimal-toc
11663 Modify generation of the TOC (Table Of Contents), which is created for
11664 every executable file. The @option{-mfull-toc} option is selected by
11665 default. In that case, GCC will allocate at least one TOC entry for
11666 each unique non-automatic variable reference in your program. GCC
11667 will also place floating-point constants in the TOC@. However, only
11668 16,384 entries are available in the TOC@.
11670 If you receive a linker error message that saying you have overflowed
11671 the available TOC space, you can reduce the amount of TOC space used
11672 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11673 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11674 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11675 generate code to calculate the sum of an address and a constant at
11676 run-time instead of putting that sum into the TOC@. You may specify one
11677 or both of these options. Each causes GCC to produce very slightly
11678 slower and larger code at the expense of conserving TOC space.
11680 If you still run out of space in the TOC even when you specify both of
11681 these options, specify @option{-mminimal-toc} instead. This option causes
11682 GCC to make only one TOC entry for every file. When you specify this
11683 option, GCC will produce code that is slower and larger but which
11684 uses extremely little TOC space. You may wish to use this option
11685 only on files that contain less frequently executed code.
11691 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11692 @code{long} type, and the infrastructure needed to support them.
11693 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11694 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11695 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11698 @itemx -mno-xl-compat
11699 @opindex mxl-compat
11700 @opindex mno-xl-compat
11701 Produce code that conforms more closely to IBM XL compiler semantics
11702 when using AIX-compatible ABI. Pass floating-point arguments to
11703 prototyped functions beyond the register save area (RSA) on the stack
11704 in addition to argument FPRs. Do not assume that most significant
11705 double in 128-bit long double value is properly rounded when comparing
11706 values and converting to double. Use XL symbol names for long double
11709 The AIX calling convention was extended but not initially documented to
11710 handle an obscure K&R C case of calling a function that takes the
11711 address of its arguments with fewer arguments than declared. IBM XL
11712 compilers access floating point arguments which do not fit in the
11713 RSA from the stack when a subroutine is compiled without
11714 optimization. Because always storing floating-point arguments on the
11715 stack is inefficient and rarely needed, this option is not enabled by
11716 default and only is necessary when calling subroutines compiled by IBM
11717 XL compilers without optimization.
11721 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11722 application written to use message passing with special startup code to
11723 enable the application to run. The system must have PE installed in the
11724 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11725 must be overridden with the @option{-specs=} option to specify the
11726 appropriate directory location. The Parallel Environment does not
11727 support threads, so the @option{-mpe} option and the @option{-pthread}
11728 option are incompatible.
11730 @item -malign-natural
11731 @itemx -malign-power
11732 @opindex malign-natural
11733 @opindex malign-power
11734 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11735 @option{-malign-natural} overrides the ABI-defined alignment of larger
11736 types, such as floating-point doubles, on their natural size-based boundary.
11737 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11738 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11740 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11744 @itemx -mhard-float
11745 @opindex msoft-float
11746 @opindex mhard-float
11747 Generate code that does not use (uses) the floating-point register set.
11748 Software floating point emulation is provided if you use the
11749 @option{-msoft-float} option, and pass the option to GCC when linking.
11752 @itemx -mno-multiple
11754 @opindex mno-multiple
11755 Generate code that uses (does not use) the load multiple word
11756 instructions and the store multiple word instructions. These
11757 instructions are generated by default on POWER systems, and not
11758 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11759 endian PowerPC systems, since those instructions do not work when the
11760 processor is in little endian mode. The exceptions are PPC740 and
11761 PPC750 which permit the instructions usage in little endian mode.
11766 @opindex mno-string
11767 Generate code that uses (does not use) the load string instructions
11768 and the store string word instructions to save multiple registers and
11769 do small block moves. These instructions are generated by default on
11770 POWER systems, and not generated on PowerPC systems. Do not use
11771 @option{-mstring} on little endian PowerPC systems, since those
11772 instructions do not work when the processor is in little endian mode.
11773 The exceptions are PPC740 and PPC750 which permit the instructions
11774 usage in little endian mode.
11779 @opindex mno-update
11780 Generate code that uses (does not use) the load or store instructions
11781 that update the base register to the address of the calculated memory
11782 location. These instructions are generated by default. If you use
11783 @option{-mno-update}, there is a small window between the time that the
11784 stack pointer is updated and the address of the previous frame is
11785 stored, which means code that walks the stack frame across interrupts or
11786 signals may get corrupted data.
11789 @itemx -mno-fused-madd
11790 @opindex mfused-madd
11791 @opindex mno-fused-madd
11792 Generate code that uses (does not use) the floating point multiply and
11793 accumulate instructions. These instructions are generated by default if
11794 hardware floating is used.
11800 Generate code that uses (does not use) the half-word multiply and
11801 multiply-accumulate instructions on the IBM 405 and 440 processors.
11802 These instructions are generated by default when targetting those
11809 Generate code that uses (does not use) the string-search @samp{dlmzb}
11810 instruction on the IBM 405 and 440 processors. This instruction is
11811 generated by default when targetting those processors.
11813 @item -mno-bit-align
11815 @opindex mno-bit-align
11816 @opindex mbit-align
11817 On System V.4 and embedded PowerPC systems do not (do) force structures
11818 and unions that contain bit-fields to be aligned to the base type of the
11821 For example, by default a structure containing nothing but 8
11822 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11823 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11824 the structure would be aligned to a 1 byte boundary and be one byte in
11827 @item -mno-strict-align
11828 @itemx -mstrict-align
11829 @opindex mno-strict-align
11830 @opindex mstrict-align
11831 On System V.4 and embedded PowerPC systems do not (do) assume that
11832 unaligned memory references will be handled by the system.
11834 @item -mrelocatable
11835 @itemx -mno-relocatable
11836 @opindex mrelocatable
11837 @opindex mno-relocatable
11838 On embedded PowerPC systems generate code that allows (does not allow)
11839 the program to be relocated to a different address at runtime. If you
11840 use @option{-mrelocatable} on any module, all objects linked together must
11841 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11843 @item -mrelocatable-lib
11844 @itemx -mno-relocatable-lib
11845 @opindex mrelocatable-lib
11846 @opindex mno-relocatable-lib
11847 On embedded PowerPC systems generate code that allows (does not allow)
11848 the program to be relocated to a different address at runtime. Modules
11849 compiled with @option{-mrelocatable-lib} can be linked with either modules
11850 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11851 with modules compiled with the @option{-mrelocatable} options.
11857 On System V.4 and embedded PowerPC systems do not (do) assume that
11858 register 2 contains a pointer to a global area pointing to the addresses
11859 used in the program.
11862 @itemx -mlittle-endian
11864 @opindex mlittle-endian
11865 On System V.4 and embedded PowerPC systems compile code for the
11866 processor in little endian mode. The @option{-mlittle-endian} option is
11867 the same as @option{-mlittle}.
11870 @itemx -mbig-endian
11872 @opindex mbig-endian
11873 On System V.4 and embedded PowerPC systems compile code for the
11874 processor in big endian mode. The @option{-mbig-endian} option is
11875 the same as @option{-mbig}.
11877 @item -mdynamic-no-pic
11878 @opindex mdynamic-no-pic
11879 On Darwin and Mac OS X systems, compile code so that it is not
11880 relocatable, but that its external references are relocatable. The
11881 resulting code is suitable for applications, but not shared
11884 @item -mprioritize-restricted-insns=@var{priority}
11885 @opindex mprioritize-restricted-insns
11886 This option controls the priority that is assigned to
11887 dispatch-slot restricted instructions during the second scheduling
11888 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11889 @var{no/highest/second-highest} priority to dispatch slot restricted
11892 @item -msched-costly-dep=@var{dependence_type}
11893 @opindex msched-costly-dep
11894 This option controls which dependences are considered costly
11895 by the target during instruction scheduling. The argument
11896 @var{dependence_type} takes one of the following values:
11897 @var{no}: no dependence is costly,
11898 @var{all}: all dependences are costly,
11899 @var{true_store_to_load}: a true dependence from store to load is costly,
11900 @var{store_to_load}: any dependence from store to load is costly,
11901 @var{number}: any dependence which latency >= @var{number} is costly.
11903 @item -minsert-sched-nops=@var{scheme}
11904 @opindex minsert-sched-nops
11905 This option controls which nop insertion scheme will be used during
11906 the second scheduling pass. The argument @var{scheme} takes one of the
11908 @var{no}: Don't insert nops.
11909 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11910 according to the scheduler's grouping.
11911 @var{regroup_exact}: Insert nops to force costly dependent insns into
11912 separate groups. Insert exactly as many nops as needed to force an insn
11913 to a new group, according to the estimated processor grouping.
11914 @var{number}: Insert nops to force costly dependent insns into
11915 separate groups. Insert @var{number} nops to force an insn to a new group.
11918 @opindex mcall-sysv
11919 On System V.4 and embedded PowerPC systems compile code using calling
11920 conventions that adheres to the March 1995 draft of the System V
11921 Application Binary Interface, PowerPC processor supplement. This is the
11922 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11924 @item -mcall-sysv-eabi
11925 @opindex mcall-sysv-eabi
11926 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11928 @item -mcall-sysv-noeabi
11929 @opindex mcall-sysv-noeabi
11930 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11932 @item -mcall-solaris
11933 @opindex mcall-solaris
11934 On System V.4 and embedded PowerPC systems compile code for the Solaris
11938 @opindex mcall-linux
11939 On System V.4 and embedded PowerPC systems compile code for the
11940 Linux-based GNU system.
11944 On System V.4 and embedded PowerPC systems compile code for the
11945 Hurd-based GNU system.
11947 @item -mcall-netbsd
11948 @opindex mcall-netbsd
11949 On System V.4 and embedded PowerPC systems compile code for the
11950 NetBSD operating system.
11952 @item -maix-struct-return
11953 @opindex maix-struct-return
11954 Return all structures in memory (as specified by the AIX ABI)@.
11956 @item -msvr4-struct-return
11957 @opindex msvr4-struct-return
11958 Return structures smaller than 8 bytes in registers (as specified by the
11961 @item -mabi=@var{abi-type}
11963 Extend the current ABI with a particular extension, or remove such extension.
11964 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11965 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11969 Extend the current ABI with SPE ABI extensions. This does not change
11970 the default ABI, instead it adds the SPE ABI extensions to the current
11974 @opindex mabi=no-spe
11975 Disable Booke SPE ABI extensions for the current ABI@.
11977 @item -mabi=ibmlongdouble
11978 @opindex mabi=ibmlongdouble
11979 Change the current ABI to use IBM extended precision long double.
11980 This is a PowerPC 32-bit SYSV ABI option.
11982 @item -mabi=ieeelongdouble
11983 @opindex mabi=ieeelongdouble
11984 Change the current ABI to use IEEE extended precision long double.
11985 This is a PowerPC 32-bit Linux ABI option.
11988 @itemx -mno-prototype
11989 @opindex mprototype
11990 @opindex mno-prototype
11991 On System V.4 and embedded PowerPC systems assume that all calls to
11992 variable argument functions are properly prototyped. Otherwise, the
11993 compiler must insert an instruction before every non prototyped call to
11994 set or clear bit 6 of the condition code register (@var{CR}) to
11995 indicate whether floating point values were passed in the floating point
11996 registers in case the function takes a variable arguments. With
11997 @option{-mprototype}, only calls to prototyped variable argument functions
11998 will set or clear the bit.
12002 On embedded PowerPC systems, assume that the startup module is called
12003 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
12004 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
12009 On embedded PowerPC systems, assume that the startup module is called
12010 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
12015 On embedded PowerPC systems, assume that the startup module is called
12016 @file{crt0.o} and the standard C libraries are @file{libads.a} and
12019 @item -myellowknife
12020 @opindex myellowknife
12021 On embedded PowerPC systems, assume that the startup module is called
12022 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
12027 On System V.4 and embedded PowerPC systems, specify that you are
12028 compiling for a VxWorks system.
12032 Specify that you are compiling for the WindISS simulation environment.
12036 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
12037 header to indicate that @samp{eabi} extended relocations are used.
12043 On System V.4 and embedded PowerPC systems do (do not) adhere to the
12044 Embedded Applications Binary Interface (eabi) which is a set of
12045 modifications to the System V.4 specifications. Selecting @option{-meabi}
12046 means that the stack is aligned to an 8 byte boundary, a function
12047 @code{__eabi} is called to from @code{main} to set up the eabi
12048 environment, and the @option{-msdata} option can use both @code{r2} and
12049 @code{r13} to point to two separate small data areas. Selecting
12050 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
12051 do not call an initialization function from @code{main}, and the
12052 @option{-msdata} option will only use @code{r13} to point to a single
12053 small data area. The @option{-meabi} option is on by default if you
12054 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
12057 @opindex msdata=eabi
12058 On System V.4 and embedded PowerPC systems, put small initialized
12059 @code{const} global and static data in the @samp{.sdata2} section, which
12060 is pointed to by register @code{r2}. Put small initialized
12061 non-@code{const} global and static data in the @samp{.sdata} section,
12062 which is pointed to by register @code{r13}. Put small uninitialized
12063 global and static data in the @samp{.sbss} section, which is adjacent to
12064 the @samp{.sdata} section. The @option{-msdata=eabi} option is
12065 incompatible with the @option{-mrelocatable} option. The
12066 @option{-msdata=eabi} option also sets the @option{-memb} option.
12069 @opindex msdata=sysv
12070 On System V.4 and embedded PowerPC systems, put small global and static
12071 data in the @samp{.sdata} section, which is pointed to by register
12072 @code{r13}. Put small uninitialized global and static data in the
12073 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
12074 The @option{-msdata=sysv} option is incompatible with the
12075 @option{-mrelocatable} option.
12077 @item -msdata=default
12079 @opindex msdata=default
12081 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12082 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12083 same as @option{-msdata=sysv}.
12086 @opindex msdata-data
12087 On System V.4 and embedded PowerPC systems, put small global
12088 data in the @samp{.sdata} section. Put small uninitialized global
12089 data in the @samp{.sbss} section. Do not use register @code{r13}
12090 to address small data however. This is the default behavior unless
12091 other @option{-msdata} options are used.
12095 @opindex msdata=none
12097 On embedded PowerPC systems, put all initialized global and static data
12098 in the @samp{.data} section, and all uninitialized data in the
12099 @samp{.bss} section.
12103 @cindex smaller data references (PowerPC)
12104 @cindex .sdata/.sdata2 references (PowerPC)
12105 On embedded PowerPC systems, put global and static items less than or
12106 equal to @var{num} bytes into the small data or bss sections instead of
12107 the normal data or bss section. By default, @var{num} is 8. The
12108 @option{-G @var{num}} switch is also passed to the linker.
12109 All modules should be compiled with the same @option{-G @var{num}} value.
12112 @itemx -mno-regnames
12114 @opindex mno-regnames
12115 On System V.4 and embedded PowerPC systems do (do not) emit register
12116 names in the assembly language output using symbolic forms.
12119 @itemx -mno-longcall
12121 @opindex mno-longcall
12122 By default assume that all calls are far away so that a longer more
12123 expensive calling sequence is required. This is required for calls
12124 further than 32 megabytes (33,554,432 bytes) from the current location.
12125 A short call will be generated if the compiler knows
12126 the call cannot be that far away. This setting can be overridden by
12127 the @code{shortcall} function attribute, or by @code{#pragma
12130 Some linkers are capable of detecting out-of-range calls and generating
12131 glue code on the fly. On these systems, long calls are unnecessary and
12132 generate slower code. As of this writing, the AIX linker can do this,
12133 as can the GNU linker for PowerPC/64. It is planned to add this feature
12134 to the GNU linker for 32-bit PowerPC systems as well.
12136 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
12137 callee, L42'', plus a ``branch island'' (glue code). The two target
12138 addresses represent the callee and the ``branch island''. The
12139 Darwin/PPC linker will prefer the first address and generate a ``bl
12140 callee'' if the PPC ``bl'' instruction will reach the callee directly;
12141 otherwise, the linker will generate ``bl L42'' to call the ``branch
12142 island''. The ``branch island'' is appended to the body of the
12143 calling function; it computes the full 32-bit address of the callee
12146 On Mach-O (Darwin) systems, this option directs the compiler emit to
12147 the glue for every direct call, and the Darwin linker decides whether
12148 to use or discard it.
12150 In the future, we may cause GCC to ignore all longcall specifications
12151 when the linker is known to generate glue.
12155 Adds support for multithreading with the @dfn{pthreads} library.
12156 This option sets flags for both the preprocessor and linker.
12160 @node S/390 and zSeries Options
12161 @subsection S/390 and zSeries Options
12162 @cindex S/390 and zSeries Options
12164 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
12168 @itemx -msoft-float
12169 @opindex mhard-float
12170 @opindex msoft-float
12171 Use (do not use) the hardware floating-point instructions and registers
12172 for floating-point operations. When @option{-msoft-float} is specified,
12173 functions in @file{libgcc.a} will be used to perform floating-point
12174 operations. When @option{-mhard-float} is specified, the compiler
12175 generates IEEE floating-point instructions. This is the default.
12177 @item -mlong-double-64
12178 @itemx -mlong-double-128
12179 @opindex mlong-double-64
12180 @opindex mlong-double-128
12181 These switches control the size of @code{long double} type. A size
12182 of 64bit makes the @code{long double} type equivalent to the @code{double}
12183 type. This is the default.
12186 @itemx -mno-backchain
12187 @opindex mbackchain
12188 @opindex mno-backchain
12189 Store (do not store) the address of the caller's frame as backchain pointer
12190 into the callee's stack frame.
12191 A backchain may be needed to allow debugging using tools that do not understand
12192 DWARF-2 call frame information.
12193 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12194 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12195 the backchain is placed into the topmost word of the 96/160 byte register
12198 In general, code compiled with @option{-mbackchain} is call-compatible with
12199 code compiled with @option{-mmo-backchain}; however, use of the backchain
12200 for debugging purposes usually requires that the whole binary is built with
12201 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12202 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12203 to build a linux kernel use @option{-msoft-float}.
12205 The default is to not maintain the backchain.
12207 @item -mpacked-stack
12208 @item -mno-packed-stack
12209 @opindex mpacked-stack
12210 @opindex mno-packed-stack
12211 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12212 specified, the compiler uses the all fields of the 96/160 byte register save
12213 area only for their default purpose; unused fields still take up stack space.
12214 When @option{-mpacked-stack} is specified, register save slots are densely
12215 packed at the top of the register save area; unused space is reused for other
12216 purposes, allowing for more efficient use of the available stack space.
12217 However, when @option{-mbackchain} is also in effect, the topmost word of
12218 the save area is always used to store the backchain, and the return address
12219 register is always saved two words below the backchain.
12221 As long as the stack frame backchain is not used, code generated with
12222 @option{-mpacked-stack} is call-compatible with code generated with
12223 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12224 S/390 or zSeries generated code that uses the stack frame backchain at run
12225 time, not just for debugging purposes. Such code is not call-compatible
12226 with code compiled with @option{-mpacked-stack}. Also, note that the
12227 combination of @option{-mbackchain},
12228 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12229 to build a linux kernel use @option{-msoft-float}.
12231 The default is to not use the packed stack layout.
12234 @itemx -mno-small-exec
12235 @opindex msmall-exec
12236 @opindex mno-small-exec
12237 Generate (or do not generate) code using the @code{bras} instruction
12238 to do subroutine calls.
12239 This only works reliably if the total executable size does not
12240 exceed 64k. The default is to use the @code{basr} instruction instead,
12241 which does not have this limitation.
12247 When @option{-m31} is specified, generate code compliant to the
12248 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12249 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12250 particular to generate 64-bit instructions. For the @samp{s390}
12251 targets, the default is @option{-m31}, while the @samp{s390x}
12252 targets default to @option{-m64}.
12258 When @option{-mzarch} is specified, generate code using the
12259 instructions available on z/Architecture.
12260 When @option{-mesa} is specified, generate code using the
12261 instructions available on ESA/390. Note that @option{-mesa} is
12262 not possible with @option{-m64}.
12263 When generating code compliant to the GNU/Linux for S/390 ABI,
12264 the default is @option{-mesa}. When generating code compliant
12265 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12271 Generate (or do not generate) code using the @code{mvcle} instruction
12272 to perform block moves. When @option{-mno-mvcle} is specified,
12273 use a @code{mvc} loop instead. This is the default unless optimizing for
12280 Print (or do not print) additional debug information when compiling.
12281 The default is to not print debug information.
12283 @item -march=@var{cpu-type}
12285 Generate code that will run on @var{cpu-type}, which is the name of a system
12286 representing a certain processor type. Possible values for
12287 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12288 When generating code using the instructions available on z/Architecture,
12289 the default is @option{-march=z900}. Otherwise, the default is
12290 @option{-march=g5}.
12292 @item -mtune=@var{cpu-type}
12294 Tune to @var{cpu-type} everything applicable about the generated code,
12295 except for the ABI and the set of available instructions.
12296 The list of @var{cpu-type} values is the same as for @option{-march}.
12297 The default is the value used for @option{-march}.
12300 @itemx -mno-tpf-trace
12301 @opindex mtpf-trace
12302 @opindex mno-tpf-trace
12303 Generate code that adds (does not add) in TPF OS specific branches to trace
12304 routines in the operating system. This option is off by default, even
12305 when compiling for the TPF OS@.
12308 @itemx -mno-fused-madd
12309 @opindex mfused-madd
12310 @opindex mno-fused-madd
12311 Generate code that uses (does not use) the floating point multiply and
12312 accumulate instructions. These instructions are generated by default if
12313 hardware floating point is used.
12315 @item -mwarn-framesize=@var{framesize}
12316 @opindex mwarn-framesize
12317 Emit a warning if the current function exceeds the given frame size. Because
12318 this is a compile time check it doesn't need to be a real problem when the program
12319 runs. It is intended to identify functions which most probably cause
12320 a stack overflow. It is useful to be used in an environment with limited stack
12321 size e.g.@: the linux kernel.
12323 @item -mwarn-dynamicstack
12324 @opindex mwarn-dynamicstack
12325 Emit a warning if the function calls alloca or uses dynamically
12326 sized arrays. This is generally a bad idea with a limited stack size.
12328 @item -mstack-guard=@var{stack-guard}
12329 @item -mstack-size=@var{stack-size}
12330 @opindex mstack-guard
12331 @opindex mstack-size
12332 These arguments always have to be used in conjunction. If they are present the s390
12333 back end emits additional instructions in the function prologue which trigger a trap
12334 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12335 (remember that the stack on s390 grows downward). These options are intended to
12336 be used to help debugging stack overflow problems. The additionally emitted code
12337 causes only little overhead and hence can also be used in production like systems
12338 without greater performance degradation. The given values have to be exact
12339 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12341 In order to be efficient the extra code makes the assumption that the stack starts
12342 at an address aligned to the value given by @var{stack-size}.
12345 @node Score Options
12346 @subsection Score Options
12347 @cindex Score Options
12349 These options are defined for Score implementations:
12354 Compile code for little endian mode.
12358 Compile code for big endian mode. This is the default.
12362 Enable the use of multiply-accumulate instructions. Disabled by default.
12366 Specify the SCORE5U of the target architecture.
12370 Specify the SCORE7 of the target architecture. This is the default.
12374 @subsection SH Options
12376 These @samp{-m} options are defined for the SH implementations:
12381 Generate code for the SH1.
12385 Generate code for the SH2.
12388 Generate code for the SH2e.
12392 Generate code for the SH3.
12396 Generate code for the SH3e.
12400 Generate code for the SH4 without a floating-point unit.
12402 @item -m4-single-only
12403 @opindex m4-single-only
12404 Generate code for the SH4 with a floating-point unit that only
12405 supports single-precision arithmetic.
12409 Generate code for the SH4 assuming the floating-point unit is in
12410 single-precision mode by default.
12414 Generate code for the SH4.
12418 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12419 floating-point unit is not used.
12421 @item -m4a-single-only
12422 @opindex m4a-single-only
12423 Generate code for the SH4a, in such a way that no double-precision
12424 floating point operations are used.
12427 @opindex m4a-single
12428 Generate code for the SH4a assuming the floating-point unit is in
12429 single-precision mode by default.
12433 Generate code for the SH4a.
12437 Same as @option{-m4a-nofpu}, except that it implicitly passes
12438 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12439 instructions at the moment.
12443 Compile code for the processor in big endian mode.
12447 Compile code for the processor in little endian mode.
12451 Align doubles at 64-bit boundaries. Note that this changes the calling
12452 conventions, and thus some functions from the standard C library will
12453 not work unless you recompile it first with @option{-mdalign}.
12457 Shorten some address references at link time, when possible; uses the
12458 linker option @option{-relax}.
12462 Use 32-bit offsets in @code{switch} tables. The default is to use
12467 Enable the use of the instruction @code{fmovd}.
12471 Comply with the calling conventions defined by Renesas.
12475 Comply with the calling conventions defined by Renesas.
12479 Comply with the calling conventions defined for GCC before the Renesas
12480 conventions were available. This option is the default for all
12481 targets of the SH toolchain except for @samp{sh-symbianelf}.
12484 @opindex mnomacsave
12485 Mark the @code{MAC} register as call-clobbered, even if
12486 @option{-mhitachi} is given.
12490 Increase IEEE-compliance of floating-point code.
12491 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12492 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12493 comparisons of NANs / infinities incurs extra overhead in every
12494 floating point comparison, therefore the default is set to
12495 @option{-ffinite-math-only}.
12497 @item -minline-ic_invalidate
12498 @opindex minline-ic_invalidate
12499 Inline code to invalidate instruction cache entries after setting up
12500 nested function trampolines.
12501 This option has no effect if -musermode is in effect and the selected
12502 code generation option (e.g. -m4) does not allow the use of the icbi
12504 If the selected code generation option does not allow the use of the icbi
12505 instruction, and -musermode is not in effect, the inlined code will
12506 manipulate the instruction cache address array directly with an associative
12507 write. This not only requires privileged mode, but it will also
12508 fail if the cache line had been mapped via the TLB and has become unmapped.
12512 Dump instruction size and location in the assembly code.
12515 @opindex mpadstruct
12516 This option is deprecated. It pads structures to multiple of 4 bytes,
12517 which is incompatible with the SH ABI@.
12521 Optimize for space instead of speed. Implied by @option{-Os}.
12524 @opindex mprefergot
12525 When generating position-independent code, emit function calls using
12526 the Global Offset Table instead of the Procedure Linkage Table.
12530 Don't generate privileged mode only code; implies -mno-inline-ic_invalidate
12531 if the inlined code would not work in user mode.
12532 This is the default when the target is @code{sh-*-linux*}.
12534 @item -multcost=@var{number}
12535 @opindex multcost=@var{number}
12536 Set the cost to assume for a multiply insn.
12538 @item -mdiv=@var{strategy}
12539 @opindex mdiv=@var{strategy}
12540 Set the division strategy to use for SHmedia code. @var{strategy} must be
12541 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12542 inv:call2, inv:fp .
12543 "fp" performs the operation in floating point. This has a very high latency,
12544 but needs only a few instructions, so it might be a good choice if
12545 your code has enough easily exploitable ILP to allow the compiler to
12546 schedule the floating point instructions together with other instructions.
12547 Division by zero causes a floating point exception.
12548 "inv" uses integer operations to calculate the inverse of the divisor,
12549 and then multiplies the dividend with the inverse. This strategy allows
12550 cse and hoisting of the inverse calculation. Division by zero calculates
12551 an unspecified result, but does not trap.
12552 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12553 have been found, or if the entire operation has been hoisted to the same
12554 place, the last stages of the inverse calculation are intertwined with the
12555 final multiply to reduce the overall latency, at the expense of using a few
12556 more instructions, and thus offering fewer scheduling opportunities with
12558 "call" calls a library function that usually implements the inv:minlat
12560 This gives high code density for m5-*media-nofpu compilations.
12561 "call2" uses a different entry point of the same library function, where it
12562 assumes that a pointer to a lookup table has already been set up, which
12563 exposes the pointer load to cse / code hoisting optimizations.
12564 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12565 code generation, but if the code stays unoptimized, revert to the "call",
12566 "call2", or "fp" strategies, respectively. Note that the
12567 potentially-trapping side effect of division by zero is carried by a
12568 separate instruction, so it is possible that all the integer instructions
12569 are hoisted out, but the marker for the side effect stays where it is.
12570 A recombination to fp operations or a call is not possible in that case.
12571 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12572 that the inverse calculation was nor separated from the multiply, they speed
12573 up division where the dividend fits into 20 bits (plus sign where applicable),
12574 by inserting a test to skip a number of operations in this case; this test
12575 slows down the case of larger dividends. inv20u assumes the case of a such
12576 a small dividend to be unlikely, and inv20l assumes it to be likely.
12578 @item -mdivsi3_libfunc=@var{name}
12579 @opindex mdivsi3_libfunc=@var{name}
12580 Set the name of the library function used for 32 bit signed division to
12581 @var{name}. This only affect the name used in the call and inv:call
12582 division strategies, and the compiler will still expect the same
12583 sets of input/output/clobbered registers as if this option was not present.
12585 @item -madjust-unroll
12586 @opindex madjust-unroll
12587 Throttle unrolling to avoid thrashing target registers.
12588 This option only has an effect if the gcc code base supports the
12589 TARGET_ADJUST_UNROLL_MAX target hook.
12591 @item -mindexed-addressing
12592 @opindex mindexed-addressing
12593 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12594 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12595 semantics for the indexed addressing mode. The architecture allows the
12596 implementation of processors with 64 bit MMU, which the OS could use to
12597 get 32 bit addressing, but since no current hardware implementation supports
12598 this or any other way to make the indexed addressing mode safe to use in
12599 the 32 bit ABI, the default is -mno-indexed-addressing.
12601 @item -mgettrcost=@var{number}
12602 @opindex mgettrcost=@var{number}
12603 Set the cost assumed for the gettr instruction to @var{number}.
12604 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12608 Assume pt* instructions won't trap. This will generally generate better
12609 scheduled code, but is unsafe on current hardware. The current architecture
12610 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12611 This has the unintentional effect of making it unsafe to schedule ptabs /
12612 ptrel before a branch, or hoist it out of a loop. For example,
12613 __do_global_ctors, a part of libgcc that runs constructors at program
12614 startup, calls functions in a list which is delimited by -1. With the
12615 -mpt-fixed option, the ptabs will be done before testing against -1.
12616 That means that all the constructors will be run a bit quicker, but when
12617 the loop comes to the end of the list, the program crashes because ptabs
12618 loads -1 into a target register. Since this option is unsafe for any
12619 hardware implementing the current architecture specification, the default
12620 is -mno-pt-fixed. Unless the user specifies a specific cost with
12621 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12622 this deters register allocation using target registers for storing
12625 @item -minvalid-symbols
12626 @opindex minvalid-symbols
12627 Assume symbols might be invalid. Ordinary function symbols generated by
12628 the compiler will always be valid to load with movi/shori/ptabs or
12629 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12630 to generate symbols that will cause ptabs / ptrel to trap.
12631 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12632 It will then prevent cross-basic-block cse, hoisting and most scheduling
12633 of symbol loads. The default is @option{-mno-invalid-symbols}.
12636 @node SPARC Options
12637 @subsection SPARC Options
12638 @cindex SPARC options
12640 These @samp{-m} options are supported on the SPARC:
12643 @item -mno-app-regs
12645 @opindex mno-app-regs
12647 Specify @option{-mapp-regs} to generate output using the global registers
12648 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12651 To be fully SVR4 ABI compliant at the cost of some performance loss,
12652 specify @option{-mno-app-regs}. You should compile libraries and system
12653 software with this option.
12656 @itemx -mhard-float
12658 @opindex mhard-float
12659 Generate output containing floating point instructions. This is the
12663 @itemx -msoft-float
12665 @opindex msoft-float
12666 Generate output containing library calls for floating point.
12667 @strong{Warning:} the requisite libraries are not available for all SPARC
12668 targets. Normally the facilities of the machine's usual C compiler are
12669 used, but this cannot be done directly in cross-compilation. You must make
12670 your own arrangements to provide suitable library functions for
12671 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12672 @samp{sparclite-*-*} do provide software floating point support.
12674 @option{-msoft-float} changes the calling convention in the output file;
12675 therefore, it is only useful if you compile @emph{all} of a program with
12676 this option. In particular, you need to compile @file{libgcc.a}, the
12677 library that comes with GCC, with @option{-msoft-float} in order for
12680 @item -mhard-quad-float
12681 @opindex mhard-quad-float
12682 Generate output containing quad-word (long double) floating point
12685 @item -msoft-quad-float
12686 @opindex msoft-quad-float
12687 Generate output containing library calls for quad-word (long double)
12688 floating point instructions. The functions called are those specified
12689 in the SPARC ABI@. This is the default.
12691 As of this writing, there are no SPARC implementations that have hardware
12692 support for the quad-word floating point instructions. They all invoke
12693 a trap handler for one of these instructions, and then the trap handler
12694 emulates the effect of the instruction. Because of the trap handler overhead,
12695 this is much slower than calling the ABI library routines. Thus the
12696 @option{-msoft-quad-float} option is the default.
12698 @item -mno-unaligned-doubles
12699 @itemx -munaligned-doubles
12700 @opindex mno-unaligned-doubles
12701 @opindex munaligned-doubles
12702 Assume that doubles have 8 byte alignment. This is the default.
12704 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12705 alignment only if they are contained in another type, or if they have an
12706 absolute address. Otherwise, it assumes they have 4 byte alignment.
12707 Specifying this option avoids some rare compatibility problems with code
12708 generated by other compilers. It is not the default because it results
12709 in a performance loss, especially for floating point code.
12711 @item -mno-faster-structs
12712 @itemx -mfaster-structs
12713 @opindex mno-faster-structs
12714 @opindex mfaster-structs
12715 With @option{-mfaster-structs}, the compiler assumes that structures
12716 should have 8 byte alignment. This enables the use of pairs of
12717 @code{ldd} and @code{std} instructions for copies in structure
12718 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12719 However, the use of this changed alignment directly violates the SPARC
12720 ABI@. Thus, it's intended only for use on targets where the developer
12721 acknowledges that their resulting code will not be directly in line with
12722 the rules of the ABI@.
12724 @item -mimpure-text
12725 @opindex mimpure-text
12726 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12727 the compiler to not pass @option{-z text} to the linker when linking a
12728 shared object. Using this option, you can link position-dependent
12729 code into a shared object.
12731 @option{-mimpure-text} suppresses the ``relocations remain against
12732 allocatable but non-writable sections'' linker error message.
12733 However, the necessary relocations will trigger copy-on-write, and the
12734 shared object is not actually shared across processes. Instead of
12735 using @option{-mimpure-text}, you should compile all source code with
12736 @option{-fpic} or @option{-fPIC}.
12738 This option is only available on SunOS and Solaris.
12740 @item -mcpu=@var{cpu_type}
12742 Set the instruction set, register set, and instruction scheduling parameters
12743 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12744 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12745 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12746 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
12747 @samp{ultrasparc3}, and @samp{niagara}.
12749 Default instruction scheduling parameters are used for values that select
12750 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12751 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12753 Here is a list of each supported architecture and their supported
12758 v8: supersparc, hypersparc
12759 sparclite: f930, f934, sparclite86x
12761 v9: ultrasparc, ultrasparc3, niagara
12764 By default (unless configured otherwise), GCC generates code for the V7
12765 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12766 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12767 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12768 SPARCStation 1, 2, IPX etc.
12770 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12771 architecture. The only difference from V7 code is that the compiler emits
12772 the integer multiply and integer divide instructions which exist in SPARC-V8
12773 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12774 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12777 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12778 the SPARC architecture. This adds the integer multiply, integer divide step
12779 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12780 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12781 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12782 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12783 MB86934 chip, which is the more recent SPARClite with FPU@.
12785 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12786 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12787 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12788 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12789 optimizes it for the TEMIC SPARClet chip.
12791 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12792 architecture. This adds 64-bit integer and floating-point move instructions,
12793 3 additional floating-point condition code registers and conditional move
12794 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12795 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
12796 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12797 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
12798 @option{-mcpu=niagara}, the compiler additionally optimizes it for
12799 Sun UltraSPARC T1 chips.
12801 @item -mtune=@var{cpu_type}
12803 Set the instruction scheduling parameters for machine type
12804 @var{cpu_type}, but do not set the instruction set or register set that the
12805 option @option{-mcpu=@var{cpu_type}} would.
12807 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12808 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12809 that select a particular cpu implementation. Those are @samp{cypress},
12810 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12811 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
12812 @samp{ultrasparc3}, and @samp{niagara}.
12817 @opindex mno-v8plus
12818 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12819 difference from the V8 ABI is that the global and out registers are
12820 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12821 mode for all SPARC-V9 processors.
12827 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12828 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12831 These @samp{-m} options are supported in addition to the above
12832 on SPARC-V9 processors in 64-bit environments:
12835 @item -mlittle-endian
12836 @opindex mlittle-endian
12837 Generate code for a processor running in little-endian mode. It is only
12838 available for a few configurations and most notably not on Solaris and Linux.
12844 Generate code for a 32-bit or 64-bit environment.
12845 The 32-bit environment sets int, long and pointer to 32 bits.
12846 The 64-bit environment sets int to 32 bits and long and pointer
12849 @item -mcmodel=medlow
12850 @opindex mcmodel=medlow
12851 Generate code for the Medium/Low code model: 64-bit addresses, programs
12852 must be linked in the low 32 bits of memory. Programs can be statically
12853 or dynamically linked.
12855 @item -mcmodel=medmid
12856 @opindex mcmodel=medmid
12857 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12858 must be linked in the low 44 bits of memory, the text and data segments must
12859 be less than 2GB in size and the data segment must be located within 2GB of
12862 @item -mcmodel=medany
12863 @opindex mcmodel=medany
12864 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12865 may be linked anywhere in memory, the text and data segments must be less
12866 than 2GB in size and the data segment must be located within 2GB of the
12869 @item -mcmodel=embmedany
12870 @opindex mcmodel=embmedany
12871 Generate code for the Medium/Anywhere code model for embedded systems:
12872 64-bit addresses, the text and data segments must be less than 2GB in
12873 size, both starting anywhere in memory (determined at link time). The
12874 global register %g4 points to the base of the data segment. Programs
12875 are statically linked and PIC is not supported.
12878 @itemx -mno-stack-bias
12879 @opindex mstack-bias
12880 @opindex mno-stack-bias
12881 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12882 frame pointer if present, are offset by @minus{}2047 which must be added back
12883 when making stack frame references. This is the default in 64-bit mode.
12884 Otherwise, assume no such offset is present.
12887 These switches are supported in addition to the above on Solaris:
12892 Add support for multithreading using the Solaris threads library. This
12893 option sets flags for both the preprocessor and linker. This option does
12894 not affect the thread safety of object code produced by the compiler or
12895 that of libraries supplied with it.
12899 Add support for multithreading using the POSIX threads library. This
12900 option sets flags for both the preprocessor and linker. This option does
12901 not affect the thread safety of object code produced by the compiler or
12902 that of libraries supplied with it.
12906 This is a synonym for @option{-pthreads}.
12910 @subsection SPU Options
12911 @cindex SPU options
12913 These @samp{-m} options are supported on the SPU:
12917 @itemx -merror-reloc
12918 @opindex mwarn-reloc
12919 @opindex merror-reloc
12921 The loader for SPU does not handle dynamic relocations. By default, GCC
12922 will give an error when it generates code that requires a dynamic
12923 relocation. @option{-mno-error-reloc} disables the error,
12924 @option{-mwarn-reloc} will generate a warning instead.
12927 @itemx -munsafe-dma
12929 @opindex munsafe-dma
12931 Instructions which initiate or test completion of DMA must not be
12932 reordered with respect to loads and stores of the memory which is being
12933 accessed. Users typically address this problem using the volatile
12934 keyword, but that can lead to inefficient code in places where the
12935 memory is known to not change. Rather than mark the memory as volatile
12936 we treat the DMA instructions as potentially effecting all memory. With
12937 @option{-munsafe-dma} users must use the volatile keyword to protect
12940 @item -mbranch-hints
12941 @opindex mbranch-hints
12943 By default, GCC will generate a branch hint instruction to avoid
12944 pipeline stalls for always taken or probably taken branches. A hint
12945 will not be generated closer than 8 instructions away from its branch.
12946 There is little reason to disable them, except for debugging purposes,
12947 or to make an object a little bit smaller.
12951 @opindex msmall-mem
12952 @opindex mlarge-mem
12954 By default, GCC generates code assuming that addresses are never larger
12955 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
12956 a full 32 bit address.
12961 By default, GCC links against startup code that assumes the SPU-style
12962 main function interface (which has an unconventional parameter list).
12963 With @option{-mstdmain}, GCC will link your program against startup
12964 code that assumes a C99-style interface to @code{main}, including a
12965 local copy of @code{argv} strings.
12967 @item -mfixed-range=@var{register-range}
12968 @opindex mfixed-range
12969 Generate code treating the given register range as fixed registers.
12970 A fixed register is one that the register allocator can not use. This is
12971 useful when compiling kernel code. A register range is specified as
12972 two registers separated by a dash. Multiple register ranges can be
12973 specified separated by a comma.
12977 @node System V Options
12978 @subsection Options for System V
12980 These additional options are available on System V Release 4 for
12981 compatibility with other compilers on those systems:
12986 Create a shared object.
12987 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12991 Identify the versions of each tool used by the compiler, in a
12992 @code{.ident} assembler directive in the output.
12996 Refrain from adding @code{.ident} directives to the output file (this is
12999 @item -YP,@var{dirs}
13001 Search the directories @var{dirs}, and no others, for libraries
13002 specified with @option{-l}.
13004 @item -Ym,@var{dir}
13006 Look in the directory @var{dir} to find the M4 preprocessor.
13007 The assembler uses this option.
13008 @c This is supposed to go with a -Yd for predefined M4 macro files, but
13009 @c the generic assembler that comes with Solaris takes just -Ym.
13012 @node TMS320C3x/C4x Options
13013 @subsection TMS320C3x/C4x Options
13014 @cindex TMS320C3x/C4x Options
13016 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
13020 @item -mcpu=@var{cpu_type}
13022 Set the instruction set, register set, and instruction scheduling
13023 parameters for machine type @var{cpu_type}. Supported values for
13024 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
13025 @samp{c44}. The default is @samp{c40} to generate code for the
13030 @itemx -msmall-memory
13032 @opindex mbig-memory
13034 @opindex msmall-memory
13036 Generates code for the big or small memory model. The small memory
13037 model assumed that all data fits into one 64K word page. At run-time
13038 the data page (DP) register must be set to point to the 64K page
13039 containing the .bss and .data program sections. The big memory model is
13040 the default and requires reloading of the DP register for every direct
13047 Allow (disallow) allocation of general integer operands into the block
13048 count register BK@.
13054 Enable (disable) generation of code using decrement and branch,
13055 DBcond(D), instructions. This is enabled by default for the C4x. To be
13056 on the safe side, this is disabled for the C3x, since the maximum
13057 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
13058 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
13059 that it can utilize the decrement and branch instruction, but will give
13060 up if there is more than one memory reference in the loop. Thus a loop
13061 where the loop counter is decremented can generate slightly more
13062 efficient code, in cases where the RPTB instruction cannot be utilized.
13064 @item -mdp-isr-reload
13066 @opindex mdp-isr-reload
13068 Force the DP register to be saved on entry to an interrupt service
13069 routine (ISR), reloaded to point to the data section, and restored on
13070 exit from the ISR@. This should not be required unless someone has
13071 violated the small memory model by modifying the DP register, say within
13078 For the C3x use the 24-bit MPYI instruction for integer multiplies
13079 instead of a library call to guarantee 32-bit results. Note that if one
13080 of the operands is a constant, then the multiplication will be performed
13081 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
13082 then squaring operations are performed inline instead of a library call.
13085 @itemx -mno-fast-fix
13087 @opindex mno-fast-fix
13088 The C3x/C4x FIX instruction to convert a floating point value to an
13089 integer value chooses the nearest integer less than or equal to the
13090 floating point value rather than to the nearest integer. Thus if the
13091 floating point number is negative, the result will be incorrectly
13092 truncated an additional code is necessary to detect and correct this
13093 case. This option can be used to disable generation of the additional
13094 code required to correct the result.
13100 Enable (disable) generation of repeat block sequences using the RPTB
13101 instruction for zero overhead looping. The RPTB construct is only used
13102 for innermost loops that do not call functions or jump across the loop
13103 boundaries. There is no advantage having nested RPTB loops due to the
13104 overhead required to save and restore the RC, RS, and RE registers.
13105 This is enabled by default with @option{-O2}.
13107 @item -mrpts=@var{count}
13111 Enable (disable) the use of the single instruction repeat instruction
13112 RPTS@. If a repeat block contains a single instruction, and the loop
13113 count can be guaranteed to be less than the value @var{count}, GCC will
13114 emit a RPTS instruction instead of a RPTB@. If no value is specified,
13115 then a RPTS will be emitted even if the loop count cannot be determined
13116 at compile time. Note that the repeated instruction following RPTS does
13117 not have to be reloaded from memory each iteration, thus freeing up the
13118 CPU buses for operands. However, since interrupts are blocked by this
13119 instruction, it is disabled by default.
13121 @item -mloop-unsigned
13122 @itemx -mno-loop-unsigned
13123 @opindex mloop-unsigned
13124 @opindex mno-loop-unsigned
13125 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
13126 is @math{2^{31} + 1} since these instructions test if the iteration count is
13127 negative to terminate the loop. If the iteration count is unsigned
13128 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
13129 exceeded. This switch allows an unsigned iteration count.
13133 Try to emit an assembler syntax that the TI assembler (asm30) is happy
13134 with. This also enforces compatibility with the API employed by the TI
13135 C3x C compiler. For example, long doubles are passed as structures
13136 rather than in floating point registers.
13142 Generate code that uses registers (stack) for passing arguments to functions.
13143 By default, arguments are passed in registers where possible rather
13144 than by pushing arguments on to the stack.
13146 @item -mparallel-insns
13147 @itemx -mno-parallel-insns
13148 @opindex mparallel-insns
13149 @opindex mno-parallel-insns
13150 Allow the generation of parallel instructions. This is enabled by
13151 default with @option{-O2}.
13153 @item -mparallel-mpy
13154 @itemx -mno-parallel-mpy
13155 @opindex mparallel-mpy
13156 @opindex mno-parallel-mpy
13157 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
13158 provided @option{-mparallel-insns} is also specified. These instructions have
13159 tight register constraints which can pessimize the code generation
13160 of large functions.
13165 @subsection V850 Options
13166 @cindex V850 Options
13168 These @samp{-m} options are defined for V850 implementations:
13172 @itemx -mno-long-calls
13173 @opindex mlong-calls
13174 @opindex mno-long-calls
13175 Treat all calls as being far away (near). If calls are assumed to be
13176 far away, the compiler will always load the functions address up into a
13177 register, and call indirect through the pointer.
13183 Do not optimize (do optimize) basic blocks that use the same index
13184 pointer 4 or more times to copy pointer into the @code{ep} register, and
13185 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
13186 option is on by default if you optimize.
13188 @item -mno-prolog-function
13189 @itemx -mprolog-function
13190 @opindex mno-prolog-function
13191 @opindex mprolog-function
13192 Do not use (do use) external functions to save and restore registers
13193 at the prologue and epilogue of a function. The external functions
13194 are slower, but use less code space if more than one function saves
13195 the same number of registers. The @option{-mprolog-function} option
13196 is on by default if you optimize.
13200 Try to make the code as small as possible. At present, this just turns
13201 on the @option{-mep} and @option{-mprolog-function} options.
13203 @item -mtda=@var{n}
13205 Put static or global variables whose size is @var{n} bytes or less into
13206 the tiny data area that register @code{ep} points to. The tiny data
13207 area can hold up to 256 bytes in total (128 bytes for byte references).
13209 @item -msda=@var{n}
13211 Put static or global variables whose size is @var{n} bytes or less into
13212 the small data area that register @code{gp} points to. The small data
13213 area can hold up to 64 kilobytes.
13215 @item -mzda=@var{n}
13217 Put static or global variables whose size is @var{n} bytes or less into
13218 the first 32 kilobytes of memory.
13222 Specify that the target processor is the V850.
13225 @opindex mbig-switch
13226 Generate code suitable for big switch tables. Use this option only if
13227 the assembler/linker complain about out of range branches within a switch
13232 This option will cause r2 and r5 to be used in the code generated by
13233 the compiler. This setting is the default.
13235 @item -mno-app-regs
13236 @opindex mno-app-regs
13237 This option will cause r2 and r5 to be treated as fixed registers.
13241 Specify that the target processor is the V850E1. The preprocessor
13242 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13243 this option is used.
13247 Specify that the target processor is the V850E@. The preprocessor
13248 constant @samp{__v850e__} will be defined if this option is used.
13250 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13251 are defined then a default target processor will be chosen and the
13252 relevant @samp{__v850*__} preprocessor constant will be defined.
13254 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13255 defined, regardless of which processor variant is the target.
13257 @item -mdisable-callt
13258 @opindex mdisable-callt
13259 This option will suppress generation of the CALLT instruction for the
13260 v850e and v850e1 flavors of the v850 architecture. The default is
13261 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13266 @subsection VAX Options
13267 @cindex VAX options
13269 These @samp{-m} options are defined for the VAX:
13274 Do not output certain jump instructions (@code{aobleq} and so on)
13275 that the Unix assembler for the VAX cannot handle across long
13280 Do output those jump instructions, on the assumption that you
13281 will assemble with the GNU assembler.
13285 Output code for g-format floating point numbers instead of d-format.
13288 @node x86-64 Options
13289 @subsection x86-64 Options
13290 @cindex x86-64 options
13292 These are listed under @xref{i386 and x86-64 Options}.
13294 @node Xstormy16 Options
13295 @subsection Xstormy16 Options
13296 @cindex Xstormy16 Options
13298 These options are defined for Xstormy16:
13303 Choose startup files and linker script suitable for the simulator.
13306 @node Xtensa Options
13307 @subsection Xtensa Options
13308 @cindex Xtensa Options
13310 These options are supported for Xtensa targets:
13314 @itemx -mno-const16
13316 @opindex mno-const16
13317 Enable or disable use of @code{CONST16} instructions for loading
13318 constant values. The @code{CONST16} instruction is currently not a
13319 standard option from Tensilica. When enabled, @code{CONST16}
13320 instructions are always used in place of the standard @code{L32R}
13321 instructions. The use of @code{CONST16} is enabled by default only if
13322 the @code{L32R} instruction is not available.
13325 @itemx -mno-fused-madd
13326 @opindex mfused-madd
13327 @opindex mno-fused-madd
13328 Enable or disable use of fused multiply/add and multiply/subtract
13329 instructions in the floating-point option. This has no effect if the
13330 floating-point option is not also enabled. Disabling fused multiply/add
13331 and multiply/subtract instructions forces the compiler to use separate
13332 instructions for the multiply and add/subtract operations. This may be
13333 desirable in some cases where strict IEEE 754-compliant results are
13334 required: the fused multiply add/subtract instructions do not round the
13335 intermediate result, thereby producing results with @emph{more} bits of
13336 precision than specified by the IEEE standard. Disabling fused multiply
13337 add/subtract instructions also ensures that the program output is not
13338 sensitive to the compiler's ability to combine multiply and add/subtract
13341 @item -mtext-section-literals
13342 @itemx -mno-text-section-literals
13343 @opindex mtext-section-literals
13344 @opindex mno-text-section-literals
13345 Control the treatment of literal pools. The default is
13346 @option{-mno-text-section-literals}, which places literals in a separate
13347 section in the output file. This allows the literal pool to be placed
13348 in a data RAM/ROM, and it also allows the linker to combine literal
13349 pools from separate object files to remove redundant literals and
13350 improve code size. With @option{-mtext-section-literals}, the literals
13351 are interspersed in the text section in order to keep them as close as
13352 possible to their references. This may be necessary for large assembly
13355 @item -mtarget-align
13356 @itemx -mno-target-align
13357 @opindex mtarget-align
13358 @opindex mno-target-align
13359 When this option is enabled, GCC instructs the assembler to
13360 automatically align instructions to reduce branch penalties at the
13361 expense of some code density. The assembler attempts to widen density
13362 instructions to align branch targets and the instructions following call
13363 instructions. If there are not enough preceding safe density
13364 instructions to align a target, no widening will be performed. The
13365 default is @option{-mtarget-align}. These options do not affect the
13366 treatment of auto-aligned instructions like @code{LOOP}, which the
13367 assembler will always align, either by widening density instructions or
13368 by inserting no-op instructions.
13371 @itemx -mno-longcalls
13372 @opindex mlongcalls
13373 @opindex mno-longcalls
13374 When this option is enabled, GCC instructs the assembler to translate
13375 direct calls to indirect calls unless it can determine that the target
13376 of a direct call is in the range allowed by the call instruction. This
13377 translation typically occurs for calls to functions in other source
13378 files. Specifically, the assembler translates a direct @code{CALL}
13379 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13380 The default is @option{-mno-longcalls}. This option should be used in
13381 programs where the call target can potentially be out of range. This
13382 option is implemented in the assembler, not the compiler, so the
13383 assembly code generated by GCC will still show direct call
13384 instructions---look at the disassembled object code to see the actual
13385 instructions. Note that the assembler will use an indirect call for
13386 every cross-file call, not just those that really will be out of range.
13389 @node zSeries Options
13390 @subsection zSeries Options
13391 @cindex zSeries options
13393 These are listed under @xref{S/390 and zSeries Options}.
13395 @node Code Gen Options
13396 @section Options for Code Generation Conventions
13397 @cindex code generation conventions
13398 @cindex options, code generation
13399 @cindex run-time options
13401 These machine-independent options control the interface conventions
13402 used in code generation.
13404 Most of them have both positive and negative forms; the negative form
13405 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13406 one of the forms is listed---the one which is not the default. You
13407 can figure out the other form by either removing @samp{no-} or adding
13411 @item -fbounds-check
13412 @opindex fbounds-check
13413 For front-ends that support it, generate additional code to check that
13414 indices used to access arrays are within the declared range. This is
13415 currently only supported by the Java and Fortran front-ends, where
13416 this option defaults to true and false respectively.
13420 This option generates traps for signed overflow on addition, subtraction,
13421 multiplication operations.
13425 This option instructs the compiler to assume that signed arithmetic
13426 overflow of addition, subtraction and multiplication wraps around
13427 using twos-complement representation. This flag enables some optimizations
13428 and disables others. This option is enabled by default for the Java
13429 front-end, as required by the Java language specification.
13432 @opindex fexceptions
13433 Enable exception handling. Generates extra code needed to propagate
13434 exceptions. For some targets, this implies GCC will generate frame
13435 unwind information for all functions, which can produce significant data
13436 size overhead, although it does not affect execution. If you do not
13437 specify this option, GCC will enable it by default for languages like
13438 C++ which normally require exception handling, and disable it for
13439 languages like C that do not normally require it. However, you may need
13440 to enable this option when compiling C code that needs to interoperate
13441 properly with exception handlers written in C++. You may also wish to
13442 disable this option if you are compiling older C++ programs that don't
13443 use exception handling.
13445 @item -fnon-call-exceptions
13446 @opindex fnon-call-exceptions
13447 Generate code that allows trapping instructions to throw exceptions.
13448 Note that this requires platform-specific runtime support that does
13449 not exist everywhere. Moreover, it only allows @emph{trapping}
13450 instructions to throw exceptions, i.e.@: memory references or floating
13451 point instructions. It does not allow exceptions to be thrown from
13452 arbitrary signal handlers such as @code{SIGALRM}.
13454 @item -funwind-tables
13455 @opindex funwind-tables
13456 Similar to @option{-fexceptions}, except that it will just generate any needed
13457 static data, but will not affect the generated code in any other way.
13458 You will normally not enable this option; instead, a language processor
13459 that needs this handling would enable it on your behalf.
13461 @item -fasynchronous-unwind-tables
13462 @opindex fasynchronous-unwind-tables
13463 Generate unwind table in dwarf2 format, if supported by target machine. The
13464 table is exact at each instruction boundary, so it can be used for stack
13465 unwinding from asynchronous events (such as debugger or garbage collector).
13467 @item -fpcc-struct-return
13468 @opindex fpcc-struct-return
13469 Return ``short'' @code{struct} and @code{union} values in memory like
13470 longer ones, rather than in registers. This convention is less
13471 efficient, but it has the advantage of allowing intercallability between
13472 GCC-compiled files and files compiled with other compilers, particularly
13473 the Portable C Compiler (pcc).
13475 The precise convention for returning structures in memory depends
13476 on the target configuration macros.
13478 Short structures and unions are those whose size and alignment match
13479 that of some integer type.
13481 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13482 switch is not binary compatible with code compiled with the
13483 @option{-freg-struct-return} switch.
13484 Use it to conform to a non-default application binary interface.
13486 @item -freg-struct-return
13487 @opindex freg-struct-return
13488 Return @code{struct} and @code{union} values in registers when possible.
13489 This is more efficient for small structures than
13490 @option{-fpcc-struct-return}.
13492 If you specify neither @option{-fpcc-struct-return} nor
13493 @option{-freg-struct-return}, GCC defaults to whichever convention is
13494 standard for the target. If there is no standard convention, GCC
13495 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13496 the principal compiler. In those cases, we can choose the standard, and
13497 we chose the more efficient register return alternative.
13499 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13500 switch is not binary compatible with code compiled with the
13501 @option{-fpcc-struct-return} switch.
13502 Use it to conform to a non-default application binary interface.
13504 @item -fshort-enums
13505 @opindex fshort-enums
13506 Allocate to an @code{enum} type only as many bytes as it needs for the
13507 declared range of possible values. Specifically, the @code{enum} type
13508 will be equivalent to the smallest integer type which has enough room.
13510 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13511 code that is not binary compatible with code generated without that switch.
13512 Use it to conform to a non-default application binary interface.
13514 @item -fshort-double
13515 @opindex fshort-double
13516 Use the same size for @code{double} as for @code{float}.
13518 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13519 code that is not binary compatible with code generated without that switch.
13520 Use it to conform to a non-default application binary interface.
13522 @item -fshort-wchar
13523 @opindex fshort-wchar
13524 Override the underlying type for @samp{wchar_t} to be @samp{short
13525 unsigned int} instead of the default for the target. This option is
13526 useful for building programs to run under WINE@.
13528 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13529 code that is not binary compatible with code generated without that switch.
13530 Use it to conform to a non-default application binary interface.
13533 @opindex fno-common
13534 In C, allocate even uninitialized global variables in the data section of the
13535 object file, rather than generating them as common blocks. This has the
13536 effect that if the same variable is declared (without @code{extern}) in
13537 two different compilations, you will get an error when you link them.
13538 The only reason this might be useful is if you wish to verify that the
13539 program will work on other systems which always work this way.
13543 Ignore the @samp{#ident} directive.
13545 @item -finhibit-size-directive
13546 @opindex finhibit-size-directive
13547 Don't output a @code{.size} assembler directive, or anything else that
13548 would cause trouble if the function is split in the middle, and the
13549 two halves are placed at locations far apart in memory. This option is
13550 used when compiling @file{crtstuff.c}; you should not need to use it
13553 @item -fverbose-asm
13554 @opindex fverbose-asm
13555 Put extra commentary information in the generated assembly code to
13556 make it more readable. This option is generally only of use to those
13557 who actually need to read the generated assembly code (perhaps while
13558 debugging the compiler itself).
13560 @option{-fno-verbose-asm}, the default, causes the
13561 extra information to be omitted and is useful when comparing two assembler
13564 @item -frecord-gcc-switches
13565 @opindex frecord-gcc-switches
13566 This switch causes the command line that was used to invoke the
13567 compiler to be recorded into the object file that is being created.
13568 This switch is only implemented on some targets and the exact format
13569 of the recording is target and binary file format dependent, but it
13570 usually takes the form of a section containing ASCII text. This
13571 switch is related to the @option{-fverbose-asm} switch, but that
13572 switch only records information in the assembler output file as
13573 comments, so it never reaches the object file.
13577 @cindex global offset table
13579 Generate position-independent code (PIC) suitable for use in a shared
13580 library, if supported for the target machine. Such code accesses all
13581 constant addresses through a global offset table (GOT)@. The dynamic
13582 loader resolves the GOT entries when the program starts (the dynamic
13583 loader is not part of GCC; it is part of the operating system). If
13584 the GOT size for the linked executable exceeds a machine-specific
13585 maximum size, you get an error message from the linker indicating that
13586 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13587 instead. (These maximums are 8k on the SPARC and 32k
13588 on the m68k and RS/6000. The 386 has no such limit.)
13590 Position-independent code requires special support, and therefore works
13591 only on certain machines. For the 386, GCC supports PIC for System V
13592 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13593 position-independent.
13595 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13600 If supported for the target machine, emit position-independent code,
13601 suitable for dynamic linking and avoiding any limit on the size of the
13602 global offset table. This option makes a difference on the m68k,
13603 PowerPC and SPARC@.
13605 Position-independent code requires special support, and therefore works
13606 only on certain machines.
13608 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13615 These options are similar to @option{-fpic} and @option{-fPIC}, but
13616 generated position independent code can be only linked into executables.
13617 Usually these options are used when @option{-pie} GCC option will be
13618 used during linking.
13620 @item -fno-jump-tables
13621 @opindex fno-jump-tables
13622 Do not use jump tables for switch statements even where it would be
13623 more efficient than other code generation strategies. This option is
13624 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13625 building code which forms part of a dynamic linker and cannot
13626 reference the address of a jump table. On some targets, jump tables
13627 do not require a GOT and this option is not needed.
13629 @item -ffixed-@var{reg}
13631 Treat the register named @var{reg} as a fixed register; generated code
13632 should never refer to it (except perhaps as a stack pointer, frame
13633 pointer or in some other fixed role).
13635 @var{reg} must be the name of a register. The register names accepted
13636 are machine-specific and are defined in the @code{REGISTER_NAMES}
13637 macro in the machine description macro file.
13639 This flag does not have a negative form, because it specifies a
13642 @item -fcall-used-@var{reg}
13643 @opindex fcall-used
13644 Treat the register named @var{reg} as an allocable register that is
13645 clobbered by function calls. It may be allocated for temporaries or
13646 variables that do not live across a call. Functions compiled this way
13647 will not save and restore the register @var{reg}.
13649 It is an error to used this flag with the frame pointer or stack pointer.
13650 Use of this flag for other registers that have fixed pervasive roles in
13651 the machine's execution model will produce disastrous results.
13653 This flag does not have a negative form, because it specifies a
13656 @item -fcall-saved-@var{reg}
13657 @opindex fcall-saved
13658 Treat the register named @var{reg} as an allocable register saved by
13659 functions. It may be allocated even for temporaries or variables that
13660 live across a call. Functions compiled this way will save and restore
13661 the register @var{reg} if they use it.
13663 It is an error to used this flag with the frame pointer or stack pointer.
13664 Use of this flag for other registers that have fixed pervasive roles in
13665 the machine's execution model will produce disastrous results.
13667 A different sort of disaster will result from the use of this flag for
13668 a register in which function values may be returned.
13670 This flag does not have a negative form, because it specifies a
13673 @item -fpack-struct[=@var{n}]
13674 @opindex fpack-struct
13675 Without a value specified, pack all structure members together without
13676 holes. When a value is specified (which must be a small power of two), pack
13677 structure members according to this value, representing the maximum
13678 alignment (that is, objects with default alignment requirements larger than
13679 this will be output potentially unaligned at the next fitting location.
13681 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13682 code that is not binary compatible with code generated without that switch.
13683 Additionally, it makes the code suboptimal.
13684 Use it to conform to a non-default application binary interface.
13686 @item -finstrument-functions
13687 @opindex finstrument-functions
13688 Generate instrumentation calls for entry and exit to functions. Just
13689 after function entry and just before function exit, the following
13690 profiling functions will be called with the address of the current
13691 function and its call site. (On some platforms,
13692 @code{__builtin_return_address} does not work beyond the current
13693 function, so the call site information may not be available to the
13694 profiling functions otherwise.)
13697 void __cyg_profile_func_enter (void *this_fn,
13699 void __cyg_profile_func_exit (void *this_fn,
13703 The first argument is the address of the start of the current function,
13704 which may be looked up exactly in the symbol table.
13706 This instrumentation is also done for functions expanded inline in other
13707 functions. The profiling calls will indicate where, conceptually, the
13708 inline function is entered and exited. This means that addressable
13709 versions of such functions must be available. If all your uses of a
13710 function are expanded inline, this may mean an additional expansion of
13711 code size. If you use @samp{extern inline} in your C code, an
13712 addressable version of such functions must be provided. (This is
13713 normally the case anyways, but if you get lucky and the optimizer always
13714 expands the functions inline, you might have gotten away without
13715 providing static copies.)
13717 A function may be given the attribute @code{no_instrument_function}, in
13718 which case this instrumentation will not be done. This can be used, for
13719 example, for the profiling functions listed above, high-priority
13720 interrupt routines, and any functions from which the profiling functions
13721 cannot safely be called (perhaps signal handlers, if the profiling
13722 routines generate output or allocate memory).
13724 @item -fstack-check
13725 @opindex fstack-check
13726 Generate code to verify that you do not go beyond the boundary of the
13727 stack. You should specify this flag if you are running in an
13728 environment with multiple threads, but only rarely need to specify it in
13729 a single-threaded environment since stack overflow is automatically
13730 detected on nearly all systems if there is only one stack.
13732 Note that this switch does not actually cause checking to be done; the
13733 operating system must do that. The switch causes generation of code
13734 to ensure that the operating system sees the stack being extended.
13736 @item -fstack-limit-register=@var{reg}
13737 @itemx -fstack-limit-symbol=@var{sym}
13738 @itemx -fno-stack-limit
13739 @opindex fstack-limit-register
13740 @opindex fstack-limit-symbol
13741 @opindex fno-stack-limit
13742 Generate code to ensure that the stack does not grow beyond a certain value,
13743 either the value of a register or the address of a symbol. If the stack
13744 would grow beyond the value, a signal is raised. For most targets,
13745 the signal is raised before the stack overruns the boundary, so
13746 it is possible to catch the signal without taking special precautions.
13748 For instance, if the stack starts at absolute address @samp{0x80000000}
13749 and grows downwards, you can use the flags
13750 @option{-fstack-limit-symbol=__stack_limit} and
13751 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13752 of 128KB@. Note that this may only work with the GNU linker.
13754 @cindex aliasing of parameters
13755 @cindex parameters, aliased
13756 @item -fargument-alias
13757 @itemx -fargument-noalias
13758 @itemx -fargument-noalias-global
13759 @itemx -fargument-noalias-anything
13760 @opindex fargument-alias
13761 @opindex fargument-noalias
13762 @opindex fargument-noalias-global
13763 @opindex fargument-noalias-anything
13764 Specify the possible relationships among parameters and between
13765 parameters and global data.
13767 @option{-fargument-alias} specifies that arguments (parameters) may
13768 alias each other and may alias global storage.@*
13769 @option{-fargument-noalias} specifies that arguments do not alias
13770 each other, but may alias global storage.@*
13771 @option{-fargument-noalias-global} specifies that arguments do not
13772 alias each other and do not alias global storage.
13773 @option{-fargument-noalias-anything} specifies that arguments do not
13774 alias any other storage.
13776 Each language will automatically use whatever option is required by
13777 the language standard. You should not need to use these options yourself.
13779 @item -fleading-underscore
13780 @opindex fleading-underscore
13781 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13782 change the way C symbols are represented in the object file. One use
13783 is to help link with legacy assembly code.
13785 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13786 generate code that is not binary compatible with code generated without that
13787 switch. Use it to conform to a non-default application binary interface.
13788 Not all targets provide complete support for this switch.
13790 @item -ftls-model=@var{model}
13791 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13792 The @var{model} argument should be one of @code{global-dynamic},
13793 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13795 The default without @option{-fpic} is @code{initial-exec}; with
13796 @option{-fpic} the default is @code{global-dynamic}.
13798 @item -fvisibility=@var{default|internal|hidden|protected}
13799 @opindex fvisibility
13800 Set the default ELF image symbol visibility to the specified option---all
13801 symbols will be marked with this unless overridden within the code.
13802 Using this feature can very substantially improve linking and
13803 load times of shared object libraries, produce more optimized
13804 code, provide near-perfect API export and prevent symbol clashes.
13805 It is @strong{strongly} recommended that you use this in any shared objects
13808 Despite the nomenclature, @code{default} always means public ie;
13809 available to be linked against from outside the shared object.
13810 @code{protected} and @code{internal} are pretty useless in real-world
13811 usage so the only other commonly used option will be @code{hidden}.
13812 The default if @option{-fvisibility} isn't specified is
13813 @code{default}, i.e., make every
13814 symbol public---this causes the same behavior as previous versions of
13817 A good explanation of the benefits offered by ensuring ELF
13818 symbols have the correct visibility is given by ``How To Write
13819 Shared Libraries'' by Ulrich Drepper (which can be found at
13820 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13821 solution made possible by this option to marking things hidden when
13822 the default is public is to make the default hidden and mark things
13823 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13824 and @code{__attribute__ ((visibility("default")))} instead of
13825 @code{__declspec(dllexport)} you get almost identical semantics with
13826 identical syntax. This is a great boon to those working with
13827 cross-platform projects.
13829 For those adding visibility support to existing code, you may find
13830 @samp{#pragma GCC visibility} of use. This works by you enclosing
13831 the declarations you wish to set visibility for with (for example)
13832 @samp{#pragma GCC visibility push(hidden)} and
13833 @samp{#pragma GCC visibility pop}.
13834 Bear in mind that symbol visibility should be viewed @strong{as
13835 part of the API interface contract} and thus all new code should
13836 always specify visibility when it is not the default ie; declarations
13837 only for use within the local DSO should @strong{always} be marked explicitly
13838 as hidden as so to avoid PLT indirection overheads---making this
13839 abundantly clear also aids readability and self-documentation of the code.
13840 Note that due to ISO C++ specification requirements, operator new and
13841 operator delete must always be of default visibility.
13843 Be aware that headers from outside your project, in particular system
13844 headers and headers from any other library you use, may not be
13845 expecting to be compiled with visibility other than the default. You
13846 may need to explicitly say @samp{#pragma GCC visibility push(default)}
13847 before including any such headers.
13849 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
13850 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
13851 no modifications. However, this means that calls to @samp{extern}
13852 functions with no explicit visibility will use the PLT, so it is more
13853 effective to use @samp{__attribute ((visibility))} and/or
13854 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
13855 declarations should be treated as hidden.
13857 Note that @samp{-fvisibility} does affect C++ vague linkage
13858 entities. This means that, for instance, an exception class that will
13859 be thrown between DSOs must be explicitly marked with default
13860 visibility so that the @samp{type_info} nodes will be unified between
13863 An overview of these techniques, their benefits and how to use them
13864 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13870 @node Environment Variables
13871 @section Environment Variables Affecting GCC
13872 @cindex environment variables
13874 @c man begin ENVIRONMENT
13875 This section describes several environment variables that affect how GCC
13876 operates. Some of them work by specifying directories or prefixes to use
13877 when searching for various kinds of files. Some are used to specify other
13878 aspects of the compilation environment.
13880 Note that you can also specify places to search using options such as
13881 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13882 take precedence over places specified using environment variables, which
13883 in turn take precedence over those specified by the configuration of GCC@.
13884 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13885 GNU Compiler Collection (GCC) Internals}.
13890 @c @itemx LC_COLLATE
13892 @c @itemx LC_MONETARY
13893 @c @itemx LC_NUMERIC
13898 @c @findex LC_COLLATE
13899 @findex LC_MESSAGES
13900 @c @findex LC_MONETARY
13901 @c @findex LC_NUMERIC
13905 These environment variables control the way that GCC uses
13906 localization information that allow GCC to work with different
13907 national conventions. GCC inspects the locale categories
13908 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13909 so. These locale categories can be set to any value supported by your
13910 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13911 Kingdom encoded in UTF-8.
13913 The @env{LC_CTYPE} environment variable specifies character
13914 classification. GCC uses it to determine the character boundaries in
13915 a string; this is needed for some multibyte encodings that contain quote
13916 and escape characters that would otherwise be interpreted as a string
13919 The @env{LC_MESSAGES} environment variable specifies the language to
13920 use in diagnostic messages.
13922 If the @env{LC_ALL} environment variable is set, it overrides the value
13923 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13924 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13925 environment variable. If none of these variables are set, GCC
13926 defaults to traditional C English behavior.
13930 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13931 files. GCC uses temporary files to hold the output of one stage of
13932 compilation which is to be used as input to the next stage: for example,
13933 the output of the preprocessor, which is the input to the compiler
13936 @item GCC_EXEC_PREFIX
13937 @findex GCC_EXEC_PREFIX
13938 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13939 names of the subprograms executed by the compiler. No slash is added
13940 when this prefix is combined with the name of a subprogram, but you can
13941 specify a prefix that ends with a slash if you wish.
13943 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13944 an appropriate prefix to use based on the pathname it was invoked with.
13946 If GCC cannot find the subprogram using the specified prefix, it
13947 tries looking in the usual places for the subprogram.
13949 The default value of @env{GCC_EXEC_PREFIX} is
13950 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13951 of @code{prefix} when you ran the @file{configure} script.
13953 Other prefixes specified with @option{-B} take precedence over this prefix.
13955 This prefix is also used for finding files such as @file{crt0.o} that are
13958 In addition, the prefix is used in an unusual way in finding the
13959 directories to search for header files. For each of the standard
13960 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13961 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13962 replacing that beginning with the specified prefix to produce an
13963 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13964 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13965 These alternate directories are searched first; the standard directories
13968 @item COMPILER_PATH
13969 @findex COMPILER_PATH
13970 The value of @env{COMPILER_PATH} is a colon-separated list of
13971 directories, much like @env{PATH}. GCC tries the directories thus
13972 specified when searching for subprograms, if it can't find the
13973 subprograms using @env{GCC_EXEC_PREFIX}.
13976 @findex LIBRARY_PATH
13977 The value of @env{LIBRARY_PATH} is a colon-separated list of
13978 directories, much like @env{PATH}. When configured as a native compiler,
13979 GCC tries the directories thus specified when searching for special
13980 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13981 using GCC also uses these directories when searching for ordinary
13982 libraries for the @option{-l} option (but directories specified with
13983 @option{-L} come first).
13987 @cindex locale definition
13988 This variable is used to pass locale information to the compiler. One way in
13989 which this information is used is to determine the character set to be used
13990 when character literals, string literals and comments are parsed in C and C++.
13991 When the compiler is configured to allow multibyte characters,
13992 the following values for @env{LANG} are recognized:
13996 Recognize JIS characters.
13998 Recognize SJIS characters.
14000 Recognize EUCJP characters.
14003 If @env{LANG} is not defined, or if it has some other value, then the
14004 compiler will use mblen and mbtowc as defined by the default locale to
14005 recognize and translate multibyte characters.
14009 Some additional environments variables affect the behavior of the
14012 @include cppenv.texi
14016 @node Precompiled Headers
14017 @section Using Precompiled Headers
14018 @cindex precompiled headers
14019 @cindex speed of compilation
14021 Often large projects have many header files that are included in every
14022 source file. The time the compiler takes to process these header files
14023 over and over again can account for nearly all of the time required to
14024 build the project. To make builds faster, GCC allows users to
14025 `precompile' a header file; then, if builds can use the precompiled
14026 header file they will be much faster.
14028 To create a precompiled header file, simply compile it as you would any
14029 other file, if necessary using the @option{-x} option to make the driver
14030 treat it as a C or C++ header file. You will probably want to use a
14031 tool like @command{make} to keep the precompiled header up-to-date when
14032 the headers it contains change.
14034 A precompiled header file will be searched for when @code{#include} is
14035 seen in the compilation. As it searches for the included file
14036 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
14037 compiler looks for a precompiled header in each directory just before it
14038 looks for the include file in that directory. The name searched for is
14039 the name specified in the @code{#include} with @samp{.gch} appended. If
14040 the precompiled header file can't be used, it is ignored.
14042 For instance, if you have @code{#include "all.h"}, and you have
14043 @file{all.h.gch} in the same directory as @file{all.h}, then the
14044 precompiled header file will be used if possible, and the original
14045 header will be used otherwise.
14047 Alternatively, you might decide to put the precompiled header file in a
14048 directory and use @option{-I} to ensure that directory is searched
14049 before (or instead of) the directory containing the original header.
14050 Then, if you want to check that the precompiled header file is always
14051 used, you can put a file of the same name as the original header in this
14052 directory containing an @code{#error} command.
14054 This also works with @option{-include}. So yet another way to use
14055 precompiled headers, good for projects not designed with precompiled
14056 header files in mind, is to simply take most of the header files used by
14057 a project, include them from another header file, precompile that header
14058 file, and @option{-include} the precompiled header. If the header files
14059 have guards against multiple inclusion, they will be skipped because
14060 they've already been included (in the precompiled header).
14062 If you need to precompile the same header file for different
14063 languages, targets, or compiler options, you can instead make a
14064 @emph{directory} named like @file{all.h.gch}, and put each precompiled
14065 header in the directory, perhaps using @option{-o}. It doesn't matter
14066 what you call the files in the directory, every precompiled header in
14067 the directory will be considered. The first precompiled header
14068 encountered in the directory that is valid for this compilation will
14069 be used; they're searched in no particular order.
14071 There are many other possibilities, limited only by your imagination,
14072 good sense, and the constraints of your build system.
14074 A precompiled header file can be used only when these conditions apply:
14078 Only one precompiled header can be used in a particular compilation.
14081 A precompiled header can't be used once the first C token is seen. You
14082 can have preprocessor directives before a precompiled header; you can
14083 even include a precompiled header from inside another header, so long as
14084 there are no C tokens before the @code{#include}.
14087 The precompiled header file must be produced for the same language as
14088 the current compilation. You can't use a C precompiled header for a C++
14092 The precompiled header file must have been produced by the same compiler
14093 binary as the current compilation is using.
14096 Any macros defined before the precompiled header is included must
14097 either be defined in the same way as when the precompiled header was
14098 generated, or must not affect the precompiled header, which usually
14099 means that they don't appear in the precompiled header at all.
14101 The @option{-D} option is one way to define a macro before a
14102 precompiled header is included; using a @code{#define} can also do it.
14103 There are also some options that define macros implicitly, like
14104 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
14107 @item If debugging information is output when using the precompiled
14108 header, using @option{-g} or similar, the same kind of debugging information
14109 must have been output when building the precompiled header. However,
14110 a precompiled header built using @option{-g} can be used in a compilation
14111 when no debugging information is being output.
14113 @item The same @option{-m} options must generally be used when building
14114 and using the precompiled header. @xref{Submodel Options},
14115 for any cases where this rule is relaxed.
14117 @item Each of the following options must be the same when building and using
14118 the precompiled header:
14120 @gccoptlist{-fexceptions -funit-at-a-time}
14123 Some other command-line options starting with @option{-f},
14124 @option{-p}, or @option{-O} must be defined in the same way as when
14125 the precompiled header was generated. At present, it's not clear
14126 which options are safe to change and which are not; the safest choice
14127 is to use exactly the same options when generating and using the
14128 precompiled header. The following are known to be safe:
14130 @gccoptlist{-fmessage-length= -fpreprocessed
14131 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
14132 -fsched-verbose=<number> -fschedule-insns -fvisibility=
14137 For all of these except the last, the compiler will automatically
14138 ignore the precompiled header if the conditions aren't met. If you
14139 find an option combination that doesn't work and doesn't cause the
14140 precompiled header to be ignored, please consider filing a bug report,
14143 If you do use differing options when generating and using the
14144 precompiled header, the actual behavior will be a mixture of the
14145 behavior for the options. For instance, if you use @option{-g} to
14146 generate the precompiled header but not when using it, you may or may
14147 not get debugging information for routines in the precompiled header.
14149 @node Running Protoize
14150 @section Running Protoize
14152 The program @code{protoize} is an optional part of GCC@. You can use
14153 it to add prototypes to a program, thus converting the program to ISO
14154 C in one respect. The companion program @code{unprotoize} does the
14155 reverse: it removes argument types from any prototypes that are found.
14157 When you run these programs, you must specify a set of source files as
14158 command line arguments. The conversion programs start out by compiling
14159 these files to see what functions they define. The information gathered
14160 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
14162 After scanning comes actual conversion. The specified files are all
14163 eligible to be converted; any files they include (whether sources or
14164 just headers) are eligible as well.
14166 But not all the eligible files are converted. By default,
14167 @code{protoize} and @code{unprotoize} convert only source and header
14168 files in the current directory. You can specify additional directories
14169 whose files should be converted with the @option{-d @var{directory}}
14170 option. You can also specify particular files to exclude with the
14171 @option{-x @var{file}} option. A file is converted if it is eligible, its
14172 directory name matches one of the specified directory names, and its
14173 name within the directory has not been excluded.
14175 Basic conversion with @code{protoize} consists of rewriting most
14176 function definitions and function declarations to specify the types of
14177 the arguments. The only ones not rewritten are those for varargs
14180 @code{protoize} optionally inserts prototype declarations at the
14181 beginning of the source file, to make them available for any calls that
14182 precede the function's definition. Or it can insert prototype
14183 declarations with block scope in the blocks where undeclared functions
14186 Basic conversion with @code{unprotoize} consists of rewriting most
14187 function declarations to remove any argument types, and rewriting
14188 function definitions to the old-style pre-ISO form.
14190 Both conversion programs print a warning for any function declaration or
14191 definition that they can't convert. You can suppress these warnings
14194 The output from @code{protoize} or @code{unprotoize} replaces the
14195 original source file. The original file is renamed to a name ending
14196 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
14197 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
14198 for DOS) file already exists, then the source file is simply discarded.
14200 @code{protoize} and @code{unprotoize} both depend on GCC itself to
14201 scan the program and collect information about the functions it uses.
14202 So neither of these programs will work until GCC is installed.
14204 Here is a table of the options you can use with @code{protoize} and
14205 @code{unprotoize}. Each option works with both programs unless
14209 @item -B @var{directory}
14210 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
14211 usual directory (normally @file{/usr/local/lib}). This file contains
14212 prototype information about standard system functions. This option
14213 applies only to @code{protoize}.
14215 @item -c @var{compilation-options}
14216 Use @var{compilation-options} as the options when running @command{gcc} to
14217 produce the @samp{.X} files. The special option @option{-aux-info} is
14218 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
14220 Note that the compilation options must be given as a single argument to
14221 @code{protoize} or @code{unprotoize}. If you want to specify several
14222 @command{gcc} options, you must quote the entire set of compilation options
14223 to make them a single word in the shell.
14225 There are certain @command{gcc} arguments that you cannot use, because they
14226 would produce the wrong kind of output. These include @option{-g},
14227 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14228 the @var{compilation-options}, they are ignored.
14231 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14232 systems) instead of @samp{.c}. This is convenient if you are converting
14233 a C program to C++. This option applies only to @code{protoize}.
14236 Add explicit global declarations. This means inserting explicit
14237 declarations at the beginning of each source file for each function
14238 that is called in the file and was not declared. These declarations
14239 precede the first function definition that contains a call to an
14240 undeclared function. This option applies only to @code{protoize}.
14242 @item -i @var{string}
14243 Indent old-style parameter declarations with the string @var{string}.
14244 This option applies only to @code{protoize}.
14246 @code{unprotoize} converts prototyped function definitions to old-style
14247 function definitions, where the arguments are declared between the
14248 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14249 uses five spaces as the indentation. If you want to indent with just
14250 one space instead, use @option{-i " "}.
14253 Keep the @samp{.X} files. Normally, they are deleted after conversion
14257 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14258 a prototype declaration for each function in each block which calls the
14259 function without any declaration. This option applies only to
14263 Make no real changes. This mode just prints information about the conversions
14264 that would have been done without @option{-n}.
14267 Make no @samp{.save} files. The original files are simply deleted.
14268 Use this option with caution.
14270 @item -p @var{program}
14271 Use the program @var{program} as the compiler. Normally, the name
14272 @file{gcc} is used.
14275 Work quietly. Most warnings are suppressed.
14278 Print the version number, just like @option{-v} for @command{gcc}.
14281 If you need special compiler options to compile one of your program's
14282 source files, then you should generate that file's @samp{.X} file
14283 specially, by running @command{gcc} on that source file with the
14284 appropriate options and the option @option{-aux-info}. Then run
14285 @code{protoize} on the entire set of files. @code{protoize} will use
14286 the existing @samp{.X} file because it is newer than the source file.
14290 gcc -Dfoo=bar file1.c -aux-info file1.X
14295 You need to include the special files along with the rest in the
14296 @code{protoize} command, even though their @samp{.X} files already
14297 exist, because otherwise they won't get converted.
14299 @xref{Protoize Caveats}, for more information on how to use
14300 @code{protoize} successfully.