1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
11 @c man begin COPYRIGHT
12 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
13 1999, 2000, 2001, 2002, 2003, 2004, 2005, 2006 Free Software Foundation, Inc.
15 Permission is granted to copy, distribute and/or modify this document
16 under the terms of the GNU Free Documentation License, Version 1.2 or
17 any later version published by the Free Software Foundation; with the
18 Invariant Sections being ``GNU General Public License'' and ``Funding
19 Free Software'', the Front-Cover texts being (a) (see below), and with
20 the Back-Cover Texts being (b) (see below). A copy of the license is
21 included in the gfdl(7) man page.
23 (a) The FSF's Front-Cover Text is:
27 (b) The FSF's Back-Cover Text is:
29 You have freedom to copy and modify this GNU Manual, like GNU
30 software. Copies published by the Free Software Foundation raise
31 funds for GNU development.
33 @c Set file name and title for the man page.
35 @settitle GNU project C and C++ compiler
37 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
38 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
39 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
40 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
41 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
42 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
43 [@option{-o} @var{outfile}] [@@@var{file}] @var{infile}@dots{}
45 Only the most useful options are listed here; see below for the
46 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
49 gpl(7), gfdl(7), fsf-funding(7),
50 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
51 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
52 @file{ld}, @file{binutils} and @file{gdb}.
55 For instructions on reporting bugs, see
56 @w{@uref{http://gcc.gnu.org/bugs.html}}.
59 See the Info entry for @command{gcc}, or
60 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
61 for contributors to GCC@.
66 @chapter GCC Command Options
67 @cindex GCC command options
68 @cindex command options
69 @cindex options, GCC command
71 @c man begin DESCRIPTION
72 When you invoke GCC, it normally does preprocessing, compilation,
73 assembly and linking. The ``overall options'' allow you to stop this
74 process at an intermediate stage. For example, the @option{-c} option
75 says not to run the linker. Then the output consists of object files
76 output by the assembler.
78 Other options are passed on to one stage of processing. Some options
79 control the preprocessor and others the compiler itself. Yet other
80 options control the assembler and linker; most of these are not
81 documented here, since you rarely need to use any of them.
83 @cindex C compilation options
84 Most of the command line options that you can use with GCC are useful
85 for C programs; when an option is only useful with another language
86 (usually C++), the explanation says so explicitly. If the description
87 for a particular option does not mention a source language, you can use
88 that option with all supported languages.
90 @cindex C++ compilation options
91 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
92 options for compiling C++ programs.
94 @cindex grouping options
95 @cindex options, grouping
96 The @command{gcc} program accepts options and file names as operands. Many
97 options have multi-letter names; therefore multiple single-letter options
98 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
101 @cindex order of options
102 @cindex options, order
103 You can mix options and other arguments. For the most part, the order
104 you use doesn't matter. Order does matter when you use several options
105 of the same kind; for example, if you specify @option{-L} more than once,
106 the directories are searched in the order specified.
108 Many options have long names starting with @samp{-f} or with
109 @samp{-W}---for example,
110 @option{-fmove-loop-invariants}, @option{-Wformat} and so on. Most of
111 these have both positive and negative forms; the negative form of
112 @option{-ffoo} would be @option{-fno-foo}. This manual documents
113 only one of these two forms, whichever one is not the default.
117 @xref{Option Index}, for an index to GCC's options.
120 * Option Summary:: Brief list of all options, without explanations.
121 * Overall Options:: Controlling the kind of output:
122 an executable, object files, assembler files,
123 or preprocessed source.
124 * Invoking G++:: Compiling C++ programs.
125 * C Dialect Options:: Controlling the variant of C language compiled.
126 * C++ Dialect Options:: Variations on C++.
127 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
129 * Language Independent Options:: Controlling how diagnostics should be
131 * Warning Options:: How picky should the compiler be?
132 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
133 * Optimize Options:: How much optimization?
134 * Preprocessor Options:: Controlling header files and macro definitions.
135 Also, getting dependency information for Make.
136 * Assembler Options:: Passing options to the assembler.
137 * Link Options:: Specifying libraries and so on.
138 * Directory Options:: Where to find header files and libraries.
139 Where to find the compiler executable files.
140 * Spec Files:: How to pass switches to sub-processes.
141 * Target Options:: Running a cross-compiler, or an old version of GCC.
142 * Submodel Options:: Specifying minor hardware or convention variations,
143 such as 68010 vs 68020.
144 * Code Gen Options:: Specifying conventions for function calls, data layout
146 * Environment Variables:: Env vars that affect GCC.
147 * Precompiled Headers:: Compiling a header once, and using it many times.
148 * Running Protoize:: Automatically adding or removing function prototypes.
154 @section Option Summary
156 Here is a summary of all the options, grouped by type. Explanations are
157 in the following sections.
160 @item Overall Options
161 @xref{Overall Options,,Options Controlling the Kind of Output}.
162 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
163 -x @var{language} -v -### --help --target-help --version @@@var{file}}
165 @item C Language Options
166 @xref{C Dialect Options,,Options Controlling C Dialect}.
167 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
168 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
169 -fhosted -ffreestanding -fopenmp -fms-extensions @gol
170 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
171 -fallow-single-precision -fcond-mismatch @gol
172 -fsigned-bitfields -fsigned-char @gol
173 -funsigned-bitfields -funsigned-char}
175 @item C++ Language Options
176 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
177 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
178 -fconserve-space -ffriend-injection @gol
179 -fno-elide-constructors @gol
180 -fno-enforce-eh-specs @gol
181 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
182 -fno-implicit-templates @gol
183 -fno-implicit-inline-templates @gol
184 -fno-implement-inlines -fms-extensions @gol
185 -fno-nonansi-builtins -fno-operator-names @gol
186 -fno-optional-diags -fpermissive @gol
187 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
188 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
189 -fno-default-inline -fvisibility-inlines-hidden @gol
190 -Wabi -Wctor-dtor-privacy @gol
191 -Wnon-virtual-dtor -Wreorder @gol
192 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
193 -Wno-non-template-friend -Wold-style-cast @gol
194 -Woverloaded-virtual -Wno-pmf-conversions @gol
197 @item Objective-C and Objective-C++ Language Options
198 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
199 Objective-C and Objective-C++ Dialects}.
200 @gccoptlist{-fconstant-string-class=@var{class-name} @gol
201 -fgnu-runtime -fnext-runtime @gol
202 -fno-nil-receivers @gol
203 -fobjc-call-cxx-cdtors @gol
204 -fobjc-direct-dispatch @gol
205 -fobjc-exceptions @gol
207 -freplace-objc-classes @gol
210 -Wassign-intercept @gol
211 -Wno-protocol -Wselector @gol
212 -Wstrict-selector-match @gol
213 -Wundeclared-selector}
215 @item Language Independent Options
216 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
217 @gccoptlist{-fmessage-length=@var{n} @gol
218 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]} @gol
219 -fdiagnostics-show-option}
221 @item Warning Options
222 @xref{Warning Options,,Options to Request or Suppress Warnings}.
223 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
224 -w -Wextra -Wall -Waggregate-return -Walways-true -Wno-attributes @gol
225 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
226 -Wconversion -Wno-deprecated-declarations @gol
227 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
228 -Werror -Werror-* -Werror-implicit-function-declaration @gol
229 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
230 -Wno-format-extra-args -Wformat-nonliteral @gol
231 -Wformat-security -Wformat-y2k @gol
232 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
233 -Wimport -Wno-import -Winit-self -Winline @gol
234 -Wno-int-to-pointer-cast @gol
235 -Wno-invalid-offsetof -Winvalid-pch @gol
236 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
237 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
238 -Wmissing-format-attribute -Wmissing-include-dirs @gol
239 -Wmissing-noreturn @gol
240 -Wno-multichar -Wnonnull -Wno-overflow @gol
241 -Woverlength-strings -Wpacked -Wpadded @gol
242 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
243 -Wredundant-decls @gol
244 -Wreturn-type -Wsequence-point -Wshadow @gol
245 -Wsign-compare -Wstack-protector @gol
246 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
247 -Wstring-literal-comparison @gol
248 -Wswitch -Wswitch-default -Wswitch-enum @gol
249 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
250 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
251 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
252 -Wunused-value -Wunused-variable -Wvariadic-macros @gol
253 -Wvolatile-register-var -Wwrite-strings}
255 @item C-only Warning Options
256 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
257 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
258 -Wstrict-prototypes -Wtraditional @gol
259 -Wdeclaration-after-statement -Wpointer-sign}
261 @item Debugging Options
262 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
263 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
264 -fdump-noaddr -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
265 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
266 -fdump-ipa-all -fdump-ipa-cgraph @gol
268 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
273 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
277 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
278 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
280 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-nrv -fdump-tree-vect @gol
282 -fdump-tree-sink @gol
283 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
284 -fdump-tree-salias @gol
285 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
286 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
287 -ftree-vectorizer-verbose=@var{n} @gol
288 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
289 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
290 -feliminate-unused-debug-symbols -femit-class-debug-always @gol
291 -fmem-report -fprofile-arcs @gol
292 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
293 -ftest-coverage -ftime-report -fvar-tracking @gol
294 -g -g@var{level} -gcoff -gdwarf-2 @gol
295 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
296 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
297 -print-multi-directory -print-multi-lib @gol
298 -print-prog-name=@var{program} -print-search-dirs -Q @gol
301 @item Optimization Options
302 @xref{Optimize Options,,Options that Control Optimization}.
303 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
304 -falign-labels=@var{n} -falign-loops=@var{n} @gol
305 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
306 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
307 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
308 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
309 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
310 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
311 -fexpensive-optimizations -ffast-math -ffloat-store @gol
312 -fforce-addr -fforward-propagate -ffunction-sections @gol
313 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
314 -fcrossjumping -fif-conversion -fif-conversion2 @gol
315 -finline-functions -finline-functions-called-once @gol
316 -finline-limit=@var{n} -fkeep-inline-functions @gol
317 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
318 -fmodulo-sched -fno-branch-count-reg @gol
319 -fno-default-inline -fno-defer-pop -fmove-loop-invariants @gol
320 -fno-function-cse -fno-guess-branch-probability @gol
321 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
322 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
323 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss @gol
324 -fomit-frame-pointer -foptimize-register-move @gol
325 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
326 -fprofile-generate -fprofile-use @gol
327 -fregmove -frename-registers @gol
328 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
329 -frerun-cse-after-loop @gol
330 -frounding-math -frtl-abstract-sequences @gol
331 -fschedule-insns -fschedule-insns2 @gol
332 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
333 -fsched-spec-load-dangerous @gol
334 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
335 -fsched2-use-superblocks @gol
336 -fsched2-use-traces -fsee -freschedule-modulo-scheduled-loops @gol
337 -fsection-anchors -fsignaling-nans -fsingle-precision-constant @gol
338 -fstack-protector -fstack-protector-all @gol
339 -fstrict-aliasing -ftracer -fthread-jumps @gol
340 -funroll-all-loops -funroll-loops -fpeel-loops @gol
341 -fsplit-ivs-in-unroller -funswitch-loops @gol
342 -fvariable-expansion-in-unroller @gol
343 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
344 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
345 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
346 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
347 -ftree-vect-loop-version -ftree-salias -fipa-pta -fweb @gol
348 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
349 --param @var{name}=@var{value}
350 -O -O0 -O1 -O2 -O3 -Os}
352 @item Preprocessor Options
353 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
354 @gccoptlist{-A@var{question}=@var{answer} @gol
355 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
356 -C -dD -dI -dM -dN @gol
357 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
358 -idirafter @var{dir} @gol
359 -include @var{file} -imacros @var{file} @gol
360 -iprefix @var{file} -iwithprefix @var{dir} @gol
361 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
362 -imultilib @var{dir} -isysroot @var{dir} @gol
363 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
364 -P -fworking-directory -remap @gol
365 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
366 -Xpreprocessor @var{option}}
368 @item Assembler Option
369 @xref{Assembler Options,,Passing Options to the Assembler}.
370 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
373 @xref{Link Options,,Options for Linking}.
374 @gccoptlist{@var{object-file-name} -l@var{library} @gol
375 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
376 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
377 -Wl,@var{option} -Xlinker @var{option} @gol
380 @item Directory Options
381 @xref{Directory Options,,Options for Directory Search}.
382 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
383 -specs=@var{file} -I- --sysroot=@var{dir}}
386 @c I wrote this xref this way to avoid overfull hbox. -- rms
387 @xref{Target Options}.
388 @gccoptlist{-V @var{version} -b @var{machine}}
390 @item Machine Dependent Options
391 @xref{Submodel Options,,Hardware Models and Configurations}.
392 @c This list is ordered alphanumerically by subsection name.
393 @c Try and put the significant identifier (CPU or system) first,
394 @c so users have a clue at guessing where the ones they want will be.
397 @gccoptlist{-EB -EL @gol
398 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
399 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
402 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
403 -mabi=@var{name} @gol
404 -mapcs-stack-check -mno-apcs-stack-check @gol
405 -mapcs-float -mno-apcs-float @gol
406 -mapcs-reentrant -mno-apcs-reentrant @gol
407 -msched-prolog -mno-sched-prolog @gol
408 -mlittle-endian -mbig-endian -mwords-little-endian @gol
409 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
410 -mthumb-interwork -mno-thumb-interwork @gol
411 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
412 -mstructure-size-boundary=@var{n} @gol
413 -mabort-on-noreturn @gol
414 -mlong-calls -mno-long-calls @gol
415 -msingle-pic-base -mno-single-pic-base @gol
416 -mpic-register=@var{reg} @gol
417 -mnop-fun-dllimport @gol
418 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
419 -mpoke-function-name @gol
421 -mtpcs-frame -mtpcs-leaf-frame @gol
422 -mcaller-super-interworking -mcallee-super-interworking @gol
426 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
427 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
429 @emph{Blackfin Options}
430 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
431 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
432 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library @gol
433 -mno-id-shared-library -mshared-library-id=@var{n} @gol
434 -mleaf-id-shared-library -mno-leaf-id-shared-library @gol
435 -msep-data -mno-sep-data -mlong-calls -mno-long-calls}
438 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
439 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
440 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
441 -mstack-align -mdata-align -mconst-align @gol
442 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
443 -melf -maout -melinux -mlinux -sim -sim2 @gol
444 -mmul-bug-workaround -mno-mul-bug-workaround}
447 @gccoptlist{-mmac -mpush-args}
449 @emph{Darwin Options}
450 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
451 -arch_only -bind_at_load -bundle -bundle_loader @gol
452 -client_name -compatibility_version -current_version @gol
454 -dependency-file -dylib_file -dylinker_install_name @gol
455 -dynamic -dynamiclib -exported_symbols_list @gol
456 -filelist -flat_namespace -force_cpusubtype_ALL @gol
457 -force_flat_namespace -headerpad_max_install_names @gol
458 -image_base -init -install_name -keep_private_externs @gol
459 -multi_module -multiply_defined -multiply_defined_unused @gol
460 -noall_load -no_dead_strip_inits_and_terms @gol
461 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
462 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
463 -private_bundle -read_only_relocs -sectalign @gol
464 -sectobjectsymbols -whyload -seg1addr @gol
465 -sectcreate -sectobjectsymbols -sectorder @gol
466 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
467 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
468 -segprot -segs_read_only_addr -segs_read_write_addr @gol
469 -single_module -static -sub_library -sub_umbrella @gol
470 -twolevel_namespace -umbrella -undefined @gol
471 -unexported_symbols_list -weak_reference_mismatches @gol
472 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
473 -mkernel -mone-byte-bool}
475 @emph{DEC Alpha Options}
476 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
477 -mieee -mieee-with-inexact -mieee-conformant @gol
478 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
479 -mtrap-precision=@var{mode} -mbuild-constants @gol
480 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
481 -mbwx -mmax -mfix -mcix @gol
482 -mfloat-vax -mfloat-ieee @gol
483 -mexplicit-relocs -msmall-data -mlarge-data @gol
484 -msmall-text -mlarge-text @gol
485 -mmemory-latency=@var{time}}
487 @emph{DEC Alpha/VMS Options}
488 @gccoptlist{-mvms-return-codes}
491 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
492 -mhard-float -msoft-float @gol
493 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
494 -mdouble -mno-double @gol
495 -mmedia -mno-media -mmuladd -mno-muladd @gol
496 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
497 -mlinked-fp -mlong-calls -malign-labels @gol
498 -mlibrary-pic -macc-4 -macc-8 @gol
499 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
500 -moptimize-membar -mno-optimize-membar @gol
501 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
502 -mvliw-branch -mno-vliw-branch @gol
503 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
504 -mno-nested-cond-exec -mtomcat-stats @gol
508 @emph{GNU/Linux Options}
509 @gccoptlist{-muclibc}
511 @emph{H8/300 Options}
512 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
515 @gccoptlist{-march=@var{architecture-type} @gol
516 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
517 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
518 -mfixed-range=@var{register-range} @gol
519 -mjump-in-delay -mlinker-opt -mlong-calls @gol
520 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
521 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
522 -mno-jump-in-delay -mno-long-load-store @gol
523 -mno-portable-runtime -mno-soft-float @gol
524 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
525 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
526 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
527 -munix=@var{unix-std} -nolibdld -static -threads}
529 @emph{i386 and x86-64 Options}
530 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
531 -mfpmath=@var{unit} @gol
532 -masm=@var{dialect} -mno-fancy-math-387 @gol
533 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
534 -mno-wide-multiply -mrtd -malign-double @gol
535 -mpreferred-stack-boundary=@var{num} @gol
536 -mmmx -msse -msse2 -msse3 -mssse3 -m3dnow @gol
537 -mthreads -mno-align-stringops -minline-all-stringops @gol
538 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
539 -m96bit-long-double -mregparm=@var{num} -mx87regparm @gol
540 -msseregparm @gol -mstackrealign @gol
541 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
542 -mcmodel=@var{code-model} @gol
543 -m32 -m64 -mlarge-data-threshold=@var{num}}
546 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
547 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
548 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
549 -minline-float-divide-max-throughput @gol
550 -minline-int-divide-min-latency @gol
551 -minline-int-divide-max-throughput @gol
552 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
553 -mno-dwarf2-asm -mearly-stop-bits @gol
554 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
555 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64 @gol
556 -mno-sched-br-data-spec -msched-ar-data-spec -mno-sched-control-spec @gol
557 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec @gol
558 -msched-ldc -mno-sched-control-ldc -mno-sched-spec-verbose @gol
559 -mno-sched-prefer-non-data-spec-insns @gol
560 -mno-sched-prefer-non-control-spec-insns @gol
561 -mno-sched-count-spec-in-critical-path}
563 @emph{M32R/D Options}
564 @gccoptlist{-m32r2 -m32rx -m32r @gol
566 -malign-loops -mno-align-loops @gol
567 -missue-rate=@var{number} @gol
568 -mbranch-cost=@var{number} @gol
569 -mmodel=@var{code-size-model-type} @gol
570 -msdata=@var{sdata-type} @gol
571 -mno-flush-func -mflush-func=@var{name} @gol
572 -mno-flush-trap -mflush-trap=@var{number} @gol
576 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
578 @emph{M680x0 Options}
579 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
580 -m68060 -mcpu32 -m5200 -mcfv4e -m68881 -mbitfield @gol
581 -mc68000 -mc68020 @gol
582 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
583 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
584 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
586 @emph{M68hc1x Options}
587 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
588 -mauto-incdec -minmax -mlong-calls -mshort @gol
589 -msoft-reg-count=@var{count}}
592 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
593 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
594 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
595 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
596 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
599 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
600 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
601 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
602 -mshared -mno-shared -mxgot -mno-xgot -mgp32 -mgp64 @gol
603 -mfp32 -mfp64 -mhard-float -msoft-float @gol
604 -msingle-float -mdouble-float -mdsp -mpaired-single -mips3d @gol
605 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
606 -G@var{num} -membedded-data -mno-embedded-data @gol
607 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
608 -msplit-addresses -mno-split-addresses @gol
609 -mexplicit-relocs -mno-explicit-relocs @gol
610 -mcheck-zero-division -mno-check-zero-division @gol
611 -mdivide-traps -mdivide-breaks @gol
612 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
613 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
614 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
615 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
616 -mfix-sb1 -mno-fix-sb1 @gol
617 -mflush-func=@var{func} -mno-flush-func @gol
618 -mbranch-likely -mno-branch-likely @gol
619 -mfp-exceptions -mno-fp-exceptions @gol
620 -mvr4130-align -mno-vr4130-align}
623 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
624 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
625 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
626 -mno-base-addresses -msingle-exit -mno-single-exit}
628 @emph{MN10300 Options}
629 @gccoptlist{-mmult-bug -mno-mult-bug @gol
630 -mam33 -mno-am33 @gol
631 -mam33-2 -mno-am33-2 @gol
632 -mreturn-pointer-on-d0 @gol
636 @gccoptlist{-mno-crt0 -mbacc -msim @gol
637 -march=@var{cpu-type} }
639 @emph{PDP-11 Options}
640 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
641 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
642 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
643 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
644 -mbranch-expensive -mbranch-cheap @gol
645 -msplit -mno-split -munix-asm -mdec-asm}
647 @emph{PowerPC Options}
648 See RS/6000 and PowerPC Options.
650 @emph{RS/6000 and PowerPC Options}
651 @gccoptlist{-mcpu=@var{cpu-type} @gol
652 -mtune=@var{cpu-type} @gol
653 -mpower -mno-power -mpower2 -mno-power2 @gol
654 -mpowerpc -mpowerpc64 -mno-powerpc @gol
655 -maltivec -mno-altivec @gol
656 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
657 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
658 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
659 -mmfpgpr -mno-mfpgpr @gol
660 -mnew-mnemonics -mold-mnemonics @gol
661 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
662 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
663 -malign-power -malign-natural @gol
664 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
665 -mstring -mno-string -mupdate -mno-update @gol
666 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
667 -mstrict-align -mno-strict-align -mrelocatable @gol
668 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
669 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
670 -mdynamic-no-pic -maltivec -mswdiv @gol
671 -mprioritize-restricted-insns=@var{priority} @gol
672 -msched-costly-dep=@var{dependence_type} @gol
673 -minsert-sched-nops=@var{scheme} @gol
674 -mcall-sysv -mcall-netbsd @gol
675 -maix-struct-return -msvr4-struct-return @gol
676 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
677 -misel -mno-isel @gol
678 -misel=yes -misel=no @gol
680 -mspe=yes -mspe=no @gol
681 -mvrsave -mno-vrsave @gol
682 -mmulhw -mno-mulhw @gol
683 -mdlmzb -mno-dlmzb @gol
684 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
685 -mprototype -mno-prototype @gol
686 -msim -mmvme -mads -myellowknife -memb -msdata @gol
687 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
689 @emph{S/390 and zSeries Options}
690 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
691 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
692 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
693 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
694 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
695 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
696 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
699 @gccoptlist{-mel -mel @gol
704 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
705 -m4-nofpu -m4-single-only -m4-single -m4 @gol
706 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
707 -m5-64media -m5-64media-nofpu @gol
708 -m5-32media -m5-32media-nofpu @gol
709 -m5-compact -m5-compact-nofpu @gol
710 -mb -ml -mdalign -mrelax @gol
711 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
712 -mieee -misize -mpadstruct -mspace @gol
713 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
714 -mdivsi3_libfunc=@var{name} @gol
715 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
719 @gccoptlist{-mcpu=@var{cpu-type} @gol
720 -mtune=@var{cpu-type} @gol
721 -mcmodel=@var{code-model} @gol
722 -m32 -m64 -mapp-regs -mno-app-regs @gol
723 -mfaster-structs -mno-faster-structs @gol
724 -mfpu -mno-fpu -mhard-float -msoft-float @gol
725 -mhard-quad-float -msoft-quad-float @gol
726 -mimpure-text -mno-impure-text -mlittle-endian @gol
727 -mstack-bias -mno-stack-bias @gol
728 -munaligned-doubles -mno-unaligned-doubles @gol
729 -mv8plus -mno-v8plus -mvis -mno-vis
730 -threads -pthreads -pthread}
733 @gccoptlist{-mwarn-reloc -merror-reloc @gol
734 -msafe-dma -munsafe-dma @gol
736 -msmall-mem -mlarge-mem}
738 @emph{System V Options}
739 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
741 @emph{TMS320C3x/C4x Options}
742 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
743 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
744 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
745 -mparallel-insns -mparallel-mpy -mpreserve-float}
748 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
749 -mprolog-function -mno-prolog-function -mspace @gol
750 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
751 -mapp-regs -mno-app-regs @gol
752 -mdisable-callt -mno-disable-callt @gol
758 @gccoptlist{-mg -mgnu -munix}
760 @emph{x86-64 Options}
761 See i386 and x86-64 Options.
763 @emph{Xstormy16 Options}
766 @emph{Xtensa Options}
767 @gccoptlist{-mconst16 -mno-const16 @gol
768 -mfused-madd -mno-fused-madd @gol
769 -mtext-section-literals -mno-text-section-literals @gol
770 -mtarget-align -mno-target-align @gol
771 -mlongcalls -mno-longcalls}
773 @emph{zSeries Options}
774 See S/390 and zSeries Options.
776 @item Code Generation Options
777 @xref{Code Gen Options,,Options for Code Generation Conventions}.
778 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
779 -ffixed-@var{reg} -fexceptions @gol
780 -fnon-call-exceptions -funwind-tables @gol
781 -fasynchronous-unwind-tables @gol
782 -finhibit-size-directive -finstrument-functions @gol
783 -fno-common -fno-ident @gol
784 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
785 -fno-jump-tables @gol
786 -freg-struct-return -fshort-enums @gol
787 -fshort-double -fshort-wchar @gol
788 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
789 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
790 -fargument-alias -fargument-noalias @gol
791 -fargument-noalias-global -fargument-noalias-anything
792 -fleading-underscore -ftls-model=@var{model} @gol
793 -ftrapv -fwrapv -fbounds-check @gol
798 * Overall Options:: Controlling the kind of output:
799 an executable, object files, assembler files,
800 or preprocessed source.
801 * C Dialect Options:: Controlling the variant of C language compiled.
802 * C++ Dialect Options:: Variations on C++.
803 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
805 * Language Independent Options:: Controlling how diagnostics should be
807 * Warning Options:: How picky should the compiler be?
808 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
809 * Optimize Options:: How much optimization?
810 * Preprocessor Options:: Controlling header files and macro definitions.
811 Also, getting dependency information for Make.
812 * Assembler Options:: Passing options to the assembler.
813 * Link Options:: Specifying libraries and so on.
814 * Directory Options:: Where to find header files and libraries.
815 Where to find the compiler executable files.
816 * Spec Files:: How to pass switches to sub-processes.
817 * Target Options:: Running a cross-compiler, or an old version of GCC.
820 @node Overall Options
821 @section Options Controlling the Kind of Output
823 Compilation can involve up to four stages: preprocessing, compilation
824 proper, assembly and linking, always in that order. GCC is capable of
825 preprocessing and compiling several files either into several
826 assembler input files, or into one assembler input file; then each
827 assembler input file produces an object file, and linking combines all
828 the object files (those newly compiled, and those specified as input)
829 into an executable file.
831 @cindex file name suffix
832 For any given input file, the file name suffix determines what kind of
837 C source code which must be preprocessed.
840 C source code which should not be preprocessed.
843 C++ source code which should not be preprocessed.
846 Objective-C source code. Note that you must link with the @file{libobjc}
847 library to make an Objective-C program work.
850 Objective-C source code which should not be preprocessed.
854 Objective-C++ source code. Note that you must link with the @file{libobjc}
855 library to make an Objective-C++ program work. Note that @samp{.M} refers
856 to a literal capital M@.
859 Objective-C++ source code which should not be preprocessed.
862 C, C++, Objective-C or Objective-C++ header file to be turned into a
867 @itemx @var{file}.cxx
868 @itemx @var{file}.cpp
869 @itemx @var{file}.CPP
870 @itemx @var{file}.c++
872 C++ source code which must be preprocessed. Note that in @samp{.cxx},
873 the last two letters must both be literally @samp{x}. Likewise,
874 @samp{.C} refers to a literal capital C@.
878 Objective-C++ source code which must be preprocessed.
881 Objective-C++ source code which should not be preprocessed.
885 C++ header file to be turned into a precompiled header.
888 @itemx @var{file}.for
889 @itemx @var{file}.FOR
890 Fixed form Fortran source code which should not be preprocessed.
893 @itemx @var{file}.fpp
894 @itemx @var{file}.FPP
895 Fixed form Fortran source code which must be preprocessed (with the traditional
899 @itemx @var{file}.f95
900 Free form Fortran source code which should not be preprocessed.
903 @itemx @var{file}.F95
904 Free form Fortran source code which must be preprocessed (with the
905 traditional preprocessor).
907 @c FIXME: Descriptions of Java file types.
914 Ada source code file which contains a library unit declaration (a
915 declaration of a package, subprogram, or generic, or a generic
916 instantiation), or a library unit renaming declaration (a package,
917 generic, or subprogram renaming declaration). Such files are also
920 @itemx @var{file}.adb
921 Ada source code file containing a library unit body (a subprogram or
922 package body). Such files are also called @dfn{bodies}.
924 @c GCC also knows about some suffixes for languages not yet included:
935 Assembler code which must be preprocessed.
938 An object file to be fed straight into linking.
939 Any file name with no recognized suffix is treated this way.
943 You can specify the input language explicitly with the @option{-x} option:
946 @item -x @var{language}
947 Specify explicitly the @var{language} for the following input files
948 (rather than letting the compiler choose a default based on the file
949 name suffix). This option applies to all following input files until
950 the next @option{-x} option. Possible values for @var{language} are:
952 c c-header c-cpp-output
953 c++ c++-header c++-cpp-output
954 objective-c objective-c-header objective-c-cpp-output
955 objective-c++ objective-c++-header objective-c++-cpp-output
956 assembler assembler-with-cpp
964 Turn off any specification of a language, so that subsequent files are
965 handled according to their file name suffixes (as they are if @option{-x}
966 has not been used at all).
968 @item -pass-exit-codes
969 @opindex pass-exit-codes
970 Normally the @command{gcc} program will exit with the code of 1 if any
971 phase of the compiler returns a non-success return code. If you specify
972 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
973 numerically highest error produced by any phase that returned an error
974 indication. The C, C++, and Fortran frontends return 4, if an internal
975 compiler error is encountered.
978 If you only want some of the stages of compilation, you can use
979 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
980 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
981 @command{gcc} is to stop. Note that some combinations (for example,
982 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
987 Compile or assemble the source files, but do not link. The linking
988 stage simply is not done. The ultimate output is in the form of an
989 object file for each source file.
991 By default, the object file name for a source file is made by replacing
992 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
994 Unrecognized input files, not requiring compilation or assembly, are
999 Stop after the stage of compilation proper; do not assemble. The output
1000 is in the form of an assembler code file for each non-assembler input
1003 By default, the assembler file name for a source file is made by
1004 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
1006 Input files that don't require compilation are ignored.
1010 Stop after the preprocessing stage; do not run the compiler proper. The
1011 output is in the form of preprocessed source code, which is sent to the
1014 Input files which don't require preprocessing are ignored.
1016 @cindex output file option
1019 Place output in file @var{file}. This applies regardless to whatever
1020 sort of output is being produced, whether it be an executable file,
1021 an object file, an assembler file or preprocessed C code.
1023 If @option{-o} is not specified, the default is to put an executable
1024 file in @file{a.out}, the object file for
1025 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
1026 assembler file in @file{@var{source}.s}, a precompiled header file in
1027 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
1032 Print (on standard error output) the commands executed to run the stages
1033 of compilation. Also print the version number of the compiler driver
1034 program and of the preprocessor and the compiler proper.
1038 Like @option{-v} except the commands are not executed and all command
1039 arguments are quoted. This is useful for shell scripts to capture the
1040 driver-generated command lines.
1044 Use pipes rather than temporary files for communication between the
1045 various stages of compilation. This fails to work on some systems where
1046 the assembler is unable to read from a pipe; but the GNU assembler has
1051 If you are compiling multiple source files, this option tells the driver
1052 to pass all the source files to the compiler at once (for those
1053 languages for which the compiler can handle this). This will allow
1054 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1055 language for which this is supported is C@. If you pass source files for
1056 multiple languages to the driver, using this option, the driver will invoke
1057 the compiler(s) that support IMA once each, passing each compiler all the
1058 source files appropriate for it. For those languages that do not support
1059 IMA this option will be ignored, and the compiler will be invoked once for
1060 each source file in that language. If you use this option in conjunction
1061 with @option{-save-temps}, the compiler will generate multiple
1063 (one for each source file), but only one (combined) @file{.o} or
1068 Print (on the standard output) a description of the command line options
1069 understood by @command{gcc}. If the @option{-v} option is also specified
1070 then @option{--help} will also be passed on to the various processes
1071 invoked by @command{gcc}, so that they can display the command line options
1072 they accept. If the @option{-Wextra} option is also specified then command
1073 line options which have no documentation associated with them will also
1077 @opindex target-help
1078 Print (on the standard output) a description of target specific command
1079 line options for each tool.
1083 Display the version number and copyrights of the invoked GCC@.
1085 @include @value{srcdir}/../libiberty/at-file.texi
1089 @section Compiling C++ Programs
1091 @cindex suffixes for C++ source
1092 @cindex C++ source file suffixes
1093 C++ source files conventionally use one of the suffixes @samp{.C},
1094 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1095 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1096 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1097 files with these names and compiles them as C++ programs even if you
1098 call the compiler the same way as for compiling C programs (usually
1099 with the name @command{gcc}).
1103 However, the use of @command{gcc} does not add the C++ library.
1104 @command{g++} is a program that calls GCC and treats @samp{.c},
1105 @samp{.h} and @samp{.i} files as C++ source files instead of C source
1106 files unless @option{-x} is used, and automatically specifies linking
1107 against the C++ library. This program is also useful when
1108 precompiling a C header file with a @samp{.h} extension for use in C++
1109 compilations. On many systems, @command{g++} is also installed with
1110 the name @command{c++}.
1112 @cindex invoking @command{g++}
1113 When you compile C++ programs, you may specify many of the same
1114 command-line options that you use for compiling programs in any
1115 language; or command-line options meaningful for C and related
1116 languages; or options that are meaningful only for C++ programs.
1117 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1118 explanations of options for languages related to C@.
1119 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1120 explanations of options that are meaningful only for C++ programs.
1122 @node C Dialect Options
1123 @section Options Controlling C Dialect
1124 @cindex dialect options
1125 @cindex language dialect options
1126 @cindex options, dialect
1128 The following options control the dialect of C (or languages derived
1129 from C, such as C++, Objective-C and Objective-C++) that the compiler
1133 @cindex ANSI support
1137 In C mode, support all ISO C90 programs. In C++ mode,
1138 remove GNU extensions that conflict with ISO C++.
1140 This turns off certain features of GCC that are incompatible with ISO
1141 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1142 such as the @code{asm} and @code{typeof} keywords, and
1143 predefined macros such as @code{unix} and @code{vax} that identify the
1144 type of system you are using. It also enables the undesirable and
1145 rarely used ISO trigraph feature. For the C compiler,
1146 it disables recognition of C++ style @samp{//} comments as well as
1147 the @code{inline} keyword.
1149 The alternate keywords @code{__asm__}, @code{__extension__},
1150 @code{__inline__} and @code{__typeof__} continue to work despite
1151 @option{-ansi}. You would not want to use them in an ISO C program, of
1152 course, but it is useful to put them in header files that might be included
1153 in compilations done with @option{-ansi}. Alternate predefined macros
1154 such as @code{__unix__} and @code{__vax__} are also available, with or
1155 without @option{-ansi}.
1157 The @option{-ansi} option does not cause non-ISO programs to be
1158 rejected gratuitously. For that, @option{-pedantic} is required in
1159 addition to @option{-ansi}. @xref{Warning Options}.
1161 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1162 option is used. Some header files may notice this macro and refrain
1163 from declaring certain functions or defining certain macros that the
1164 ISO standard doesn't call for; this is to avoid interfering with any
1165 programs that might use these names for other things.
1167 Functions which would normally be built in but do not have semantics
1168 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1169 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1170 built-in functions provided by GCC}, for details of the functions
1175 Determine the language standard. This option is currently only
1176 supported when compiling C or C++. A value for this option must be
1177 provided; possible values are
1182 ISO C90 (same as @option{-ansi}).
1184 @item iso9899:199409
1185 ISO C90 as modified in amendment 1.
1191 ISO C99. Note that this standard is not yet fully supported; see
1192 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1193 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1196 Default, ISO C90 plus GNU extensions (including some C99 features).
1200 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1201 this will become the default. The name @samp{gnu9x} is deprecated.
1204 The 1998 ISO C++ standard plus amendments.
1207 The same as @option{-std=c++98} plus GNU extensions. This is the
1208 default for C++ code.
1211 The working draft of the upcoming ISO C++0x standard. This option
1212 enables experimental features that are likely to be included in
1213 C++0x. The working draft is constantly changing, and any feature that is
1214 enabled by this flag may be removed from future versions of GCC if it is
1215 not part of the C++0x standard.
1218 The same as @option{-std=c++0x} plus GNU extensions. As with
1219 @option{-std=c++0x}, this option enables experimental features that may
1220 be removed in future versions of GCC.
1223 Even when this option is not specified, you can still use some of the
1224 features of newer standards in so far as they do not conflict with
1225 previous C standards. For example, you may use @code{__restrict__} even
1226 when @option{-std=c99} is not specified.
1228 The @option{-std} options specifying some version of ISO C have the same
1229 effects as @option{-ansi}, except that features that were not in ISO C90
1230 but are in the specified version (for example, @samp{//} comments and
1231 the @code{inline} keyword in ISO C99) are not disabled.
1233 @xref{Standards,,Language Standards Supported by GCC}, for details of
1234 these standard versions.
1236 @item -aux-info @var{filename}
1238 Output to the given filename prototyped declarations for all functions
1239 declared and/or defined in a translation unit, including those in header
1240 files. This option is silently ignored in any language other than C@.
1242 Besides declarations, the file indicates, in comments, the origin of
1243 each declaration (source file and line), whether the declaration was
1244 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1245 @samp{O} for old, respectively, in the first character after the line
1246 number and the colon), and whether it came from a declaration or a
1247 definition (@samp{C} or @samp{F}, respectively, in the following
1248 character). In the case of function definitions, a K&R-style list of
1249 arguments followed by their declarations is also provided, inside
1250 comments, after the declaration.
1254 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1255 keyword, so that code can use these words as identifiers. You can use
1256 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1257 instead. @option{-ansi} implies @option{-fno-asm}.
1259 In C++, this switch only affects the @code{typeof} keyword, since
1260 @code{asm} and @code{inline} are standard keywords. You may want to
1261 use the @option{-fno-gnu-keywords} flag instead, which has the same
1262 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1263 switch only affects the @code{asm} and @code{typeof} keywords, since
1264 @code{inline} is a standard keyword in ISO C99.
1267 @itemx -fno-builtin-@var{function}
1268 @opindex fno-builtin
1269 @cindex built-in functions
1270 Don't recognize built-in functions that do not begin with
1271 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1272 functions provided by GCC}, for details of the functions affected,
1273 including those which are not built-in functions when @option{-ansi} or
1274 @option{-std} options for strict ISO C conformance are used because they
1275 do not have an ISO standard meaning.
1277 GCC normally generates special code to handle certain built-in functions
1278 more efficiently; for instance, calls to @code{alloca} may become single
1279 instructions that adjust the stack directly, and calls to @code{memcpy}
1280 may become inline copy loops. The resulting code is often both smaller
1281 and faster, but since the function calls no longer appear as such, you
1282 cannot set a breakpoint on those calls, nor can you change the behavior
1283 of the functions by linking with a different library. In addition,
1284 when a function is recognized as a built-in function, GCC may use
1285 information about that function to warn about problems with calls to
1286 that function, or to generate more efficient code, even if the
1287 resulting code still contains calls to that function. For example,
1288 warnings are given with @option{-Wformat} for bad calls to
1289 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1290 known not to modify global memory.
1292 With the @option{-fno-builtin-@var{function}} option
1293 only the built-in function @var{function} is
1294 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1295 function is named this is not built-in in this version of GCC, this
1296 option is ignored. There is no corresponding
1297 @option{-fbuiltin-@var{function}} option; if you wish to enable
1298 built-in functions selectively when using @option{-fno-builtin} or
1299 @option{-ffreestanding}, you may define macros such as:
1302 #define abs(n) __builtin_abs ((n))
1303 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1308 @cindex hosted environment
1310 Assert that compilation takes place in a hosted environment. This implies
1311 @option{-fbuiltin}. A hosted environment is one in which the
1312 entire standard library is available, and in which @code{main} has a return
1313 type of @code{int}. Examples are nearly everything except a kernel.
1314 This is equivalent to @option{-fno-freestanding}.
1316 @item -ffreestanding
1317 @opindex ffreestanding
1318 @cindex hosted environment
1320 Assert that compilation takes place in a freestanding environment. This
1321 implies @option{-fno-builtin}. A freestanding environment
1322 is one in which the standard library may not exist, and program startup may
1323 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1324 This is equivalent to @option{-fno-hosted}.
1326 @xref{Standards,,Language Standards Supported by GCC}, for details of
1327 freestanding and hosted environments.
1331 @cindex openmp parallel
1332 Enable handling of OpenMP directives @code{#pragma omp} in C/C++ and
1333 @code{!$omp} in Fortran. When @option{-fopenmp} is specified, the
1334 compiler generates parallel code according to the OpenMP Application
1335 Program Interface v2.5 @w{@uref{http://www.openmp.org/}}.
1337 @item -fms-extensions
1338 @opindex fms-extensions
1339 Accept some non-standard constructs used in Microsoft header files.
1341 Some cases of unnamed fields in structures and unions are only
1342 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1343 fields within structs/unions}, for details.
1347 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1348 options for strict ISO C conformance) implies @option{-trigraphs}.
1350 @item -no-integrated-cpp
1351 @opindex no-integrated-cpp
1352 Performs a compilation in two passes: preprocessing and compiling. This
1353 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1354 @option{-B} option. The user supplied compilation step can then add in
1355 an additional preprocessing step after normal preprocessing but before
1356 compiling. The default is to use the integrated cpp (internal cpp)
1358 The semantics of this option will change if "cc1", "cc1plus", and
1359 "cc1obj" are merged.
1361 @cindex traditional C language
1362 @cindex C language, traditional
1364 @itemx -traditional-cpp
1365 @opindex traditional-cpp
1366 @opindex traditional
1367 Formerly, these options caused GCC to attempt to emulate a pre-standard
1368 C compiler. They are now only supported with the @option{-E} switch.
1369 The preprocessor continues to support a pre-standard mode. See the GNU
1370 CPP manual for details.
1372 @item -fcond-mismatch
1373 @opindex fcond-mismatch
1374 Allow conditional expressions with mismatched types in the second and
1375 third arguments. The value of such an expression is void. This option
1376 is not supported for C++.
1378 @item -funsigned-char
1379 @opindex funsigned-char
1380 Let the type @code{char} be unsigned, like @code{unsigned char}.
1382 Each kind of machine has a default for what @code{char} should
1383 be. It is either like @code{unsigned char} by default or like
1384 @code{signed char} by default.
1386 Ideally, a portable program should always use @code{signed char} or
1387 @code{unsigned char} when it depends on the signedness of an object.
1388 But many programs have been written to use plain @code{char} and
1389 expect it to be signed, or expect it to be unsigned, depending on the
1390 machines they were written for. This option, and its inverse, let you
1391 make such a program work with the opposite default.
1393 The type @code{char} is always a distinct type from each of
1394 @code{signed char} or @code{unsigned char}, even though its behavior
1395 is always just like one of those two.
1398 @opindex fsigned-char
1399 Let the type @code{char} be signed, like @code{signed char}.
1401 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1402 the negative form of @option{-funsigned-char}. Likewise, the option
1403 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1405 @item -fsigned-bitfields
1406 @itemx -funsigned-bitfields
1407 @itemx -fno-signed-bitfields
1408 @itemx -fno-unsigned-bitfields
1409 @opindex fsigned-bitfields
1410 @opindex funsigned-bitfields
1411 @opindex fno-signed-bitfields
1412 @opindex fno-unsigned-bitfields
1413 These options control whether a bit-field is signed or unsigned, when the
1414 declaration does not use either @code{signed} or @code{unsigned}. By
1415 default, such a bit-field is signed, because this is consistent: the
1416 basic integer types such as @code{int} are signed types.
1419 @node C++ Dialect Options
1420 @section Options Controlling C++ Dialect
1422 @cindex compiler options, C++
1423 @cindex C++ options, command line
1424 @cindex options, C++
1425 This section describes the command-line options that are only meaningful
1426 for C++ programs; but you can also use most of the GNU compiler options
1427 regardless of what language your program is in. For example, you
1428 might compile a file @code{firstClass.C} like this:
1431 g++ -g -frepo -O -c firstClass.C
1435 In this example, only @option{-frepo} is an option meant
1436 only for C++ programs; you can use the other options with any
1437 language supported by GCC@.
1439 Here is a list of options that are @emph{only} for compiling C++ programs:
1443 @item -fabi-version=@var{n}
1444 @opindex fabi-version
1445 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1446 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1447 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1448 the version that conforms most closely to the C++ ABI specification.
1449 Therefore, the ABI obtained using version 0 will change as ABI bugs
1452 The default is version 2.
1454 @item -fno-access-control
1455 @opindex fno-access-control
1456 Turn off all access checking. This switch is mainly useful for working
1457 around bugs in the access control code.
1461 Check that the pointer returned by @code{operator new} is non-null
1462 before attempting to modify the storage allocated. This check is
1463 normally unnecessary because the C++ standard specifies that
1464 @code{operator new} will only return @code{0} if it is declared
1465 @samp{throw()}, in which case the compiler will always check the
1466 return value even without this option. In all other cases, when
1467 @code{operator new} has a non-empty exception specification, memory
1468 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1469 @samp{new (nothrow)}.
1471 @item -fconserve-space
1472 @opindex fconserve-space
1473 Put uninitialized or runtime-initialized global variables into the
1474 common segment, as C does. This saves space in the executable at the
1475 cost of not diagnosing duplicate definitions. If you compile with this
1476 flag and your program mysteriously crashes after @code{main()} has
1477 completed, you may have an object that is being destroyed twice because
1478 two definitions were merged.
1480 This option is no longer useful on most targets, now that support has
1481 been added for putting variables into BSS without making them common.
1483 @item -ffriend-injection
1484 @opindex ffriend-injection
1485 Inject friend functions into the enclosing namespace, so that they are
1486 visible outside the scope of the class in which they are declared.
1487 Friend functions were documented to work this way in the old Annotated
1488 C++ Reference Manual, and versions of G++ before 4.1 always worked
1489 that way. However, in ISO C++ a friend function which is not declared
1490 in an enclosing scope can only be found using argument dependent
1491 lookup. This option causes friends to be injected as they were in
1494 This option is for compatibility, and may be removed in a future
1497 @item -fno-elide-constructors
1498 @opindex fno-elide-constructors
1499 The C++ standard allows an implementation to omit creating a temporary
1500 which is only used to initialize another object of the same type.
1501 Specifying this option disables that optimization, and forces G++ to
1502 call the copy constructor in all cases.
1504 @item -fno-enforce-eh-specs
1505 @opindex fno-enforce-eh-specs
1506 Don't generate code to check for violation of exception specifications
1507 at runtime. This option violates the C++ standard, but may be useful
1508 for reducing code size in production builds, much like defining
1509 @samp{NDEBUG}. This does not give user code permission to throw
1510 exceptions in violation of the exception specifications; the compiler
1511 will still optimize based on the specifications, so throwing an
1512 unexpected exception will result in undefined behavior.
1515 @itemx -fno-for-scope
1517 @opindex fno-for-scope
1518 If @option{-ffor-scope} is specified, the scope of variables declared in
1519 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1520 as specified by the C++ standard.
1521 If @option{-fno-for-scope} is specified, the scope of variables declared in
1522 a @i{for-init-statement} extends to the end of the enclosing scope,
1523 as was the case in old versions of G++, and other (traditional)
1524 implementations of C++.
1526 The default if neither flag is given to follow the standard,
1527 but to allow and give a warning for old-style code that would
1528 otherwise be invalid, or have different behavior.
1530 @item -fno-gnu-keywords
1531 @opindex fno-gnu-keywords
1532 Do not recognize @code{typeof} as a keyword, so that code can use this
1533 word as an identifier. You can use the keyword @code{__typeof__} instead.
1534 @option{-ansi} implies @option{-fno-gnu-keywords}.
1536 @item -fno-implicit-templates
1537 @opindex fno-implicit-templates
1538 Never emit code for non-inline templates which are instantiated
1539 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1540 @xref{Template Instantiation}, for more information.
1542 @item -fno-implicit-inline-templates
1543 @opindex fno-implicit-inline-templates
1544 Don't emit code for implicit instantiations of inline templates, either.
1545 The default is to handle inlines differently so that compiles with and
1546 without optimization will need the same set of explicit instantiations.
1548 @item -fno-implement-inlines
1549 @opindex fno-implement-inlines
1550 To save space, do not emit out-of-line copies of inline functions
1551 controlled by @samp{#pragma implementation}. This will cause linker
1552 errors if these functions are not inlined everywhere they are called.
1554 @item -fms-extensions
1555 @opindex fms-extensions
1556 Disable pedantic warnings about constructs used in MFC, such as implicit
1557 int and getting a pointer to member function via non-standard syntax.
1559 @item -fno-nonansi-builtins
1560 @opindex fno-nonansi-builtins
1561 Disable built-in declarations of functions that are not mandated by
1562 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1563 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1565 @item -fno-operator-names
1566 @opindex fno-operator-names
1567 Do not treat the operator name keywords @code{and}, @code{bitand},
1568 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1569 synonyms as keywords.
1571 @item -fno-optional-diags
1572 @opindex fno-optional-diags
1573 Disable diagnostics that the standard says a compiler does not need to
1574 issue. Currently, the only such diagnostic issued by G++ is the one for
1575 a name having multiple meanings within a class.
1578 @opindex fpermissive
1579 Downgrade some diagnostics about nonconformant code from errors to
1580 warnings. Thus, using @option{-fpermissive} will allow some
1581 nonconforming code to compile.
1585 Enable automatic template instantiation at link time. This option also
1586 implies @option{-fno-implicit-templates}. @xref{Template
1587 Instantiation}, for more information.
1591 Disable generation of information about every class with virtual
1592 functions for use by the C++ runtime type identification features
1593 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1594 of the language, you can save some space by using this flag. Note that
1595 exception handling uses the same information, but it will generate it as
1596 needed. The @samp{dynamic_cast} operator can still be used for casts that
1597 do not require runtime type information, i.e. casts to @code{void *} or to
1598 unambiguous base classes.
1602 Emit statistics about front-end processing at the end of the compilation.
1603 This information is generally only useful to the G++ development team.
1605 @item -ftemplate-depth-@var{n}
1606 @opindex ftemplate-depth
1607 Set the maximum instantiation depth for template classes to @var{n}.
1608 A limit on the template instantiation depth is needed to detect
1609 endless recursions during template class instantiation. ANSI/ISO C++
1610 conforming programs must not rely on a maximum depth greater than 17.
1612 @item -fno-threadsafe-statics
1613 @opindex fno-threadsafe-statics
1614 Do not emit the extra code to use the routines specified in the C++
1615 ABI for thread-safe initialization of local statics. You can use this
1616 option to reduce code size slightly in code that doesn't need to be
1619 @item -fuse-cxa-atexit
1620 @opindex fuse-cxa-atexit
1621 Register destructors for objects with static storage duration with the
1622 @code{__cxa_atexit} function rather than the @code{atexit} function.
1623 This option is required for fully standards-compliant handling of static
1624 destructors, but will only work if your C library supports
1625 @code{__cxa_atexit}.
1627 @item -fno-use-cxa-get-exception-ptr
1628 @opindex fno-use-cxa-get-exception-ptr
1629 Don't use the @code{__cxa_get_exception_ptr} runtime routine. This
1630 will cause @code{std::uncaught_exception} to be incorrect, but is necessary
1631 if the runtime routine is not available.
1633 @item -fvisibility-inlines-hidden
1634 @opindex fvisibility-inlines-hidden
1635 This switch declares that the user does not attempt to compare
1636 pointers to inline methods where the addresses of the two functions
1637 were taken in different shared objects.
1639 The effect of this is that GCC may, effectively, mark inline methods with
1640 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1641 appear in the export table of a DSO and do not require a PLT indirection
1642 when used within the DSO@. Enabling this option can have a dramatic effect
1643 on load and link times of a DSO as it massively reduces the size of the
1644 dynamic export table when the library makes heavy use of templates.
1646 The behaviour of this switch is not quite the same as marking the
1647 methods as hidden directly, because it does not affect static variables
1648 local to the function or cause the compiler to deduce that
1649 the function is defined in only one shared object.
1651 You may mark a method as having a visibility explicitly to negate the
1652 effect of the switch for that method. For example, if you do want to
1653 compare pointers to a particular inline method, you might mark it as
1654 having default visibility. Marking the enclosing class with explicit
1655 visibility will have no effect.
1657 Explicitly instantiated inline methods are unaffected by this option
1658 as their linkage might otherwise cross a shared library boundary.
1659 @xref{Template Instantiation}.
1663 Do not use weak symbol support, even if it is provided by the linker.
1664 By default, G++ will use weak symbols if they are available. This
1665 option exists only for testing, and should not be used by end-users;
1666 it will result in inferior code and has no benefits. This option may
1667 be removed in a future release of G++.
1671 Do not search for header files in the standard directories specific to
1672 C++, but do still search the other standard directories. (This option
1673 is used when building the C++ library.)
1676 In addition, these optimization, warning, and code generation options
1677 have meanings only for C++ programs:
1680 @item -fno-default-inline
1681 @opindex fno-default-inline
1682 Do not assume @samp{inline} for functions defined inside a class scope.
1683 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1684 functions will have linkage like inline functions; they just won't be
1687 @item -Wabi @r{(C++ only)}
1689 Warn when G++ generates code that is probably not compatible with the
1690 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1691 all such cases, there are probably some cases that are not warned about,
1692 even though G++ is generating incompatible code. There may also be
1693 cases where warnings are emitted even though the code that is generated
1696 You should rewrite your code to avoid these warnings if you are
1697 concerned about the fact that code generated by G++ may not be binary
1698 compatible with code generated by other compilers.
1700 The known incompatibilities at this point include:
1705 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1706 pack data into the same byte as a base class. For example:
1709 struct A @{ virtual void f(); int f1 : 1; @};
1710 struct B : public A @{ int f2 : 1; @};
1714 In this case, G++ will place @code{B::f2} into the same byte
1715 as@code{A::f1}; other compilers will not. You can avoid this problem
1716 by explicitly padding @code{A} so that its size is a multiple of the
1717 byte size on your platform; that will cause G++ and other compilers to
1718 layout @code{B} identically.
1721 Incorrect handling of tail-padding for virtual bases. G++ does not use
1722 tail padding when laying out virtual bases. For example:
1725 struct A @{ virtual void f(); char c1; @};
1726 struct B @{ B(); char c2; @};
1727 struct C : public A, public virtual B @{@};
1731 In this case, G++ will not place @code{B} into the tail-padding for
1732 @code{A}; other compilers will. You can avoid this problem by
1733 explicitly padding @code{A} so that its size is a multiple of its
1734 alignment (ignoring virtual base classes); that will cause G++ and other
1735 compilers to layout @code{C} identically.
1738 Incorrect handling of bit-fields with declared widths greater than that
1739 of their underlying types, when the bit-fields appear in a union. For
1743 union U @{ int i : 4096; @};
1747 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1748 union too small by the number of bits in an @code{int}.
1751 Empty classes can be placed at incorrect offsets. For example:
1761 struct C : public B, public A @{@};
1765 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1766 it should be placed at offset zero. G++ mistakenly believes that the
1767 @code{A} data member of @code{B} is already at offset zero.
1770 Names of template functions whose types involve @code{typename} or
1771 template template parameters can be mangled incorrectly.
1774 template <typename Q>
1775 void f(typename Q::X) @{@}
1777 template <template <typename> class Q>
1778 void f(typename Q<int>::X) @{@}
1782 Instantiations of these templates may be mangled incorrectly.
1786 @item -Wctor-dtor-privacy @r{(C++ only)}
1787 @opindex Wctor-dtor-privacy
1788 Warn when a class seems unusable because all the constructors or
1789 destructors in that class are private, and it has neither friends nor
1790 public static member functions.
1792 @item -Wnon-virtual-dtor @r{(C++ only)}
1793 @opindex Wnon-virtual-dtor
1794 Warn when a class appears to be polymorphic, thereby requiring a virtual
1795 destructor, yet it declares a non-virtual one. This warning is also
1796 enabled if -Weffc++ is specified.
1798 @item -Wreorder @r{(C++ only)}
1800 @cindex reordering, warning
1801 @cindex warning for reordering of member initializers
1802 Warn when the order of member initializers given in the code does not
1803 match the order in which they must be executed. For instance:
1809 A(): j (0), i (1) @{ @}
1813 The compiler will rearrange the member initializers for @samp{i}
1814 and @samp{j} to match the declaration order of the members, emitting
1815 a warning to that effect. This warning is enabled by @option{-Wall}.
1818 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1821 @item -Weffc++ @r{(C++ only)}
1823 Warn about violations of the following style guidelines from Scott Meyers'
1824 @cite{Effective C++} book:
1828 Item 11: Define a copy constructor and an assignment operator for classes
1829 with dynamically allocated memory.
1832 Item 12: Prefer initialization to assignment in constructors.
1835 Item 14: Make destructors virtual in base classes.
1838 Item 15: Have @code{operator=} return a reference to @code{*this}.
1841 Item 23: Don't try to return a reference when you must return an object.
1845 Also warn about violations of the following style guidelines from
1846 Scott Meyers' @cite{More Effective C++} book:
1850 Item 6: Distinguish between prefix and postfix forms of increment and
1851 decrement operators.
1854 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1858 When selecting this option, be aware that the standard library
1859 headers do not obey all of these guidelines; use @samp{grep -v}
1860 to filter out those warnings.
1862 @item -Wno-deprecated @r{(C++ only)}
1863 @opindex Wno-deprecated
1864 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1866 @item -Wstrict-null-sentinel @r{(C++ only)}
1867 @opindex Wstrict-null-sentinel
1868 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1869 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1870 to @code{__null}. Although it is a null pointer constant not a null pointer,
1871 it is guaranteed to of the same size as a pointer. But this use is
1872 not portable across different compilers.
1874 @item -Wno-non-template-friend @r{(C++ only)}
1875 @opindex Wno-non-template-friend
1876 Disable warnings when non-templatized friend functions are declared
1877 within a template. Since the advent of explicit template specification
1878 support in G++, if the name of the friend is an unqualified-id (i.e.,
1879 @samp{friend foo(int)}), the C++ language specification demands that the
1880 friend declare or define an ordinary, nontemplate function. (Section
1881 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1882 could be interpreted as a particular specialization of a templatized
1883 function. Because this non-conforming behavior is no longer the default
1884 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1885 check existing code for potential trouble spots and is on by default.
1886 This new compiler behavior can be turned off with
1887 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1888 but disables the helpful warning.
1890 @item -Wold-style-cast @r{(C++ only)}
1891 @opindex Wold-style-cast
1892 Warn if an old-style (C-style) cast to a non-void type is used within
1893 a C++ program. The new-style casts (@samp{dynamic_cast},
1894 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1895 less vulnerable to unintended effects and much easier to search for.
1897 @item -Woverloaded-virtual @r{(C++ only)}
1898 @opindex Woverloaded-virtual
1899 @cindex overloaded virtual fn, warning
1900 @cindex warning for overloaded virtual fn
1901 Warn when a function declaration hides virtual functions from a
1902 base class. For example, in:
1909 struct B: public A @{
1914 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1922 will fail to compile.
1924 @item -Wno-pmf-conversions @r{(C++ only)}
1925 @opindex Wno-pmf-conversions
1926 Disable the diagnostic for converting a bound pointer to member function
1929 @item -Wsign-promo @r{(C++ only)}
1930 @opindex Wsign-promo
1931 Warn when overload resolution chooses a promotion from unsigned or
1932 enumerated type to a signed type, over a conversion to an unsigned type of
1933 the same size. Previous versions of G++ would try to preserve
1934 unsignedness, but the standard mandates the current behavior.
1939 A& operator = (int);
1949 In this example, G++ will synthesize a default @samp{A& operator =
1950 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1953 @node Objective-C and Objective-C++ Dialect Options
1954 @section Options Controlling Objective-C and Objective-C++ Dialects
1956 @cindex compiler options, Objective-C and Objective-C++
1957 @cindex Objective-C and Objective-C++ options, command line
1958 @cindex options, Objective-C and Objective-C++
1959 (NOTE: This manual does not describe the Objective-C and Objective-C++
1960 languages themselves. See @xref{Standards,,Language Standards
1961 Supported by GCC}, for references.)
1963 This section describes the command-line options that are only meaningful
1964 for Objective-C and Objective-C++ programs, but you can also use most of
1965 the language-independent GNU compiler options.
1966 For example, you might compile a file @code{some_class.m} like this:
1969 gcc -g -fgnu-runtime -O -c some_class.m
1973 In this example, @option{-fgnu-runtime} is an option meant only for
1974 Objective-C and Objective-C++ programs; you can use the other options with
1975 any language supported by GCC@.
1977 Note that since Objective-C is an extension of the C language, Objective-C
1978 compilations may also use options specific to the C front-end (e.g.,
1979 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1980 C++-specific options (e.g., @option{-Wabi}).
1982 Here is a list of options that are @emph{only} for compiling Objective-C
1983 and Objective-C++ programs:
1986 @item -fconstant-string-class=@var{class-name}
1987 @opindex fconstant-string-class
1988 Use @var{class-name} as the name of the class to instantiate for each
1989 literal string specified with the syntax @code{@@"@dots{}"}. The default
1990 class name is @code{NXConstantString} if the GNU runtime is being used, and
1991 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1992 @option{-fconstant-cfstrings} option, if also present, will override the
1993 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1994 to be laid out as constant CoreFoundation strings.
1997 @opindex fgnu-runtime
1998 Generate object code compatible with the standard GNU Objective-C
1999 runtime. This is the default for most types of systems.
2001 @item -fnext-runtime
2002 @opindex fnext-runtime
2003 Generate output compatible with the NeXT runtime. This is the default
2004 for NeXT-based systems, including Darwin and Mac OS X@. The macro
2005 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
2008 @item -fno-nil-receivers
2009 @opindex fno-nil-receivers
2010 Assume that all Objective-C message dispatches (e.g.,
2011 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
2012 is not @code{nil}. This allows for more efficient entry points in the runtime
2013 to be used. Currently, this option is only available in conjunction with
2014 the NeXT runtime on Mac OS X 10.3 and later.
2016 @item -fobjc-call-cxx-cdtors
2017 @opindex fobjc-call-cxx-cdtors
2018 For each Objective-C class, check if any of its instance variables is a
2019 C++ object with a non-trivial default constructor. If so, synthesize a
2020 special @code{- (id) .cxx_construct} instance method that will run
2021 non-trivial default constructors on any such instance variables, in order,
2022 and then return @code{self}. Similarly, check if any instance variable
2023 is a C++ object with a non-trivial destructor, and if so, synthesize a
2024 special @code{- (void) .cxx_destruct} method that will run
2025 all such default destructors, in reverse order.
2027 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
2028 thusly generated will only operate on instance variables declared in the
2029 current Objective-C class, and not those inherited from superclasses. It
2030 is the responsibility of the Objective-C runtime to invoke all such methods
2031 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
2032 will be invoked by the runtime immediately after a new object
2033 instance is allocated; the @code{- (void) .cxx_destruct} methods will
2034 be invoked immediately before the runtime deallocates an object instance.
2036 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
2037 support for invoking the @code{- (id) .cxx_construct} and
2038 @code{- (void) .cxx_destruct} methods.
2040 @item -fobjc-direct-dispatch
2041 @opindex fobjc-direct-dispatch
2042 Allow fast jumps to the message dispatcher. On Darwin this is
2043 accomplished via the comm page.
2045 @item -fobjc-exceptions
2046 @opindex fobjc-exceptions
2047 Enable syntactic support for structured exception handling in Objective-C,
2048 similar to what is offered by C++ and Java. This option is
2049 unavailable in conjunction with the NeXT runtime on Mac OS X 10.2 and
2058 @@catch (AnObjCClass *exc) @{
2065 @@catch (AnotherClass *exc) @{
2068 @@catch (id allOthers) @{
2078 The @code{@@throw} statement may appear anywhere in an Objective-C or
2079 Objective-C++ program; when used inside of a @code{@@catch} block, the
2080 @code{@@throw} may appear without an argument (as shown above), in which case
2081 the object caught by the @code{@@catch} will be rethrown.
2083 Note that only (pointers to) Objective-C objects may be thrown and
2084 caught using this scheme. When an object is thrown, it will be caught
2085 by the nearest @code{@@catch} clause capable of handling objects of that type,
2086 analogously to how @code{catch} blocks work in C++ and Java. A
2087 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2088 any and all Objective-C exceptions not caught by previous @code{@@catch}
2091 The @code{@@finally} clause, if present, will be executed upon exit from the
2092 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2093 regardless of whether any exceptions are thrown, caught or rethrown
2094 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2095 of the @code{finally} clause in Java.
2097 There are several caveats to using the new exception mechanism:
2101 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2102 idioms provided by the @code{NSException} class, the new
2103 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2104 systems, due to additional functionality needed in the (NeXT) Objective-C
2108 As mentioned above, the new exceptions do not support handling
2109 types other than Objective-C objects. Furthermore, when used from
2110 Objective-C++, the Objective-C exception model does not interoperate with C++
2111 exceptions at this time. This means you cannot @code{@@throw} an exception
2112 from Objective-C and @code{catch} it in C++, or vice versa
2113 (i.e., @code{throw @dots{} @@catch}).
2116 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2117 blocks for thread-safe execution:
2120 @@synchronized (ObjCClass *guard) @{
2125 Upon entering the @code{@@synchronized} block, a thread of execution shall
2126 first check whether a lock has been placed on the corresponding @code{guard}
2127 object by another thread. If it has, the current thread shall wait until
2128 the other thread relinquishes its lock. Once @code{guard} becomes available,
2129 the current thread will place its own lock on it, execute the code contained in
2130 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2131 making @code{guard} available to other threads).
2133 Unlike Java, Objective-C does not allow for entire methods to be marked
2134 @code{@@synchronized}. Note that throwing exceptions out of
2135 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2136 to be unlocked properly.
2140 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2142 @item -freplace-objc-classes
2143 @opindex freplace-objc-classes
2144 Emit a special marker instructing @command{ld(1)} not to statically link in
2145 the resulting object file, and allow @command{dyld(1)} to load it in at
2146 run time instead. This is used in conjunction with the Fix-and-Continue
2147 debugging mode, where the object file in question may be recompiled and
2148 dynamically reloaded in the course of program execution, without the need
2149 to restart the program itself. Currently, Fix-and-Continue functionality
2150 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2155 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2156 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2157 compile time) with static class references that get initialized at load time,
2158 which improves run-time performance. Specifying the @option{-fzero-link} flag
2159 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2160 to be retained. This is useful in Zero-Link debugging mode, since it allows
2161 for individual class implementations to be modified during program execution.
2165 Dump interface declarations for all classes seen in the source file to a
2166 file named @file{@var{sourcename}.decl}.
2168 @item -Wassign-intercept
2169 @opindex Wassign-intercept
2170 Warn whenever an Objective-C assignment is being intercepted by the
2174 @opindex Wno-protocol
2175 If a class is declared to implement a protocol, a warning is issued for
2176 every method in the protocol that is not implemented by the class. The
2177 default behavior is to issue a warning for every method not explicitly
2178 implemented in the class, even if a method implementation is inherited
2179 from the superclass. If you use the @option{-Wno-protocol} option, then
2180 methods inherited from the superclass are considered to be implemented,
2181 and no warning is issued for them.
2185 Warn if multiple methods of different types for the same selector are
2186 found during compilation. The check is performed on the list of methods
2187 in the final stage of compilation. Additionally, a check is performed
2188 for each selector appearing in a @code{@@selector(@dots{})}
2189 expression, and a corresponding method for that selector has been found
2190 during compilation. Because these checks scan the method table only at
2191 the end of compilation, these warnings are not produced if the final
2192 stage of compilation is not reached, for example because an error is
2193 found during compilation, or because the @option{-fsyntax-only} option is
2196 @item -Wstrict-selector-match
2197 @opindex Wstrict-selector-match
2198 Warn if multiple methods with differing argument and/or return types are
2199 found for a given selector when attempting to send a message using this
2200 selector to a receiver of type @code{id} or @code{Class}. When this flag
2201 is off (which is the default behavior), the compiler will omit such warnings
2202 if any differences found are confined to types which share the same size
2205 @item -Wundeclared-selector
2206 @opindex Wundeclared-selector
2207 Warn if a @code{@@selector(@dots{})} expression referring to an
2208 undeclared selector is found. A selector is considered undeclared if no
2209 method with that name has been declared before the
2210 @code{@@selector(@dots{})} expression, either explicitly in an
2211 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2212 an @code{@@implementation} section. This option always performs its
2213 checks as soon as a @code{@@selector(@dots{})} expression is found,
2214 while @option{-Wselector} only performs its checks in the final stage of
2215 compilation. This also enforces the coding style convention
2216 that methods and selectors must be declared before being used.
2218 @item -print-objc-runtime-info
2219 @opindex print-objc-runtime-info
2220 Generate C header describing the largest structure that is passed by
2225 @node Language Independent Options
2226 @section Options to Control Diagnostic Messages Formatting
2227 @cindex options to control diagnostics formatting
2228 @cindex diagnostic messages
2229 @cindex message formatting
2231 Traditionally, diagnostic messages have been formatted irrespective of
2232 the output device's aspect (e.g.@: its width, @dots{}). The options described
2233 below can be used to control the diagnostic messages formatting
2234 algorithm, e.g.@: how many characters per line, how often source location
2235 information should be reported. Right now, only the C++ front end can
2236 honor these options. However it is expected, in the near future, that
2237 the remaining front ends would be able to digest them correctly.
2240 @item -fmessage-length=@var{n}
2241 @opindex fmessage-length
2242 Try to format error messages so that they fit on lines of about @var{n}
2243 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2244 the front ends supported by GCC@. If @var{n} is zero, then no
2245 line-wrapping will be done; each error message will appear on a single
2248 @opindex fdiagnostics-show-location
2249 @item -fdiagnostics-show-location=once
2250 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2251 reporter to emit @emph{once} source location information; that is, in
2252 case the message is too long to fit on a single physical line and has to
2253 be wrapped, the source location won't be emitted (as prefix) again,
2254 over and over, in subsequent continuation lines. This is the default
2257 @item -fdiagnostics-show-location=every-line
2258 Only meaningful in line-wrapping mode. Instructs the diagnostic
2259 messages reporter to emit the same source location information (as
2260 prefix) for physical lines that result from the process of breaking
2261 a message which is too long to fit on a single line.
2263 @item -fdiagnostics-show-option
2264 @opindex fdiagnostics-show-option
2265 This option instructs the diagnostic machinery to add text to each
2266 diagnostic emitted, which indicates which command line option directly
2267 controls that diagnostic, when such an option is known to the
2268 diagnostic machinery.
2272 @node Warning Options
2273 @section Options to Request or Suppress Warnings
2274 @cindex options to control warnings
2275 @cindex warning messages
2276 @cindex messages, warning
2277 @cindex suppressing warnings
2279 Warnings are diagnostic messages that report constructions which
2280 are not inherently erroneous but which are risky or suggest there
2281 may have been an error.
2283 You can request many specific warnings with options beginning @samp{-W},
2284 for example @option{-Wimplicit} to request warnings on implicit
2285 declarations. Each of these specific warning options also has a
2286 negative form beginning @samp{-Wno-} to turn off warnings;
2287 for example, @option{-Wno-implicit}. This manual lists only one of the
2288 two forms, whichever is not the default.
2290 The following options control the amount and kinds of warnings produced
2291 by GCC; for further, language-specific options also refer to
2292 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2296 @cindex syntax checking
2298 @opindex fsyntax-only
2299 Check the code for syntax errors, but don't do anything beyond that.
2303 Issue all the warnings demanded by strict ISO C and ISO C++;
2304 reject all programs that use forbidden extensions, and some other
2305 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2306 version of the ISO C standard specified by any @option{-std} option used.
2308 Valid ISO C and ISO C++ programs should compile properly with or without
2309 this option (though a rare few will require @option{-ansi} or a
2310 @option{-std} option specifying the required version of ISO C)@. However,
2311 without this option, certain GNU extensions and traditional C and C++
2312 features are supported as well. With this option, they are rejected.
2314 @option{-pedantic} does not cause warning messages for use of the
2315 alternate keywords whose names begin and end with @samp{__}. Pedantic
2316 warnings are also disabled in the expression that follows
2317 @code{__extension__}. However, only system header files should use
2318 these escape routes; application programs should avoid them.
2319 @xref{Alternate Keywords}.
2321 Some users try to use @option{-pedantic} to check programs for strict ISO
2322 C conformance. They soon find that it does not do quite what they want:
2323 it finds some non-ISO practices, but not all---only those for which
2324 ISO C @emph{requires} a diagnostic, and some others for which
2325 diagnostics have been added.
2327 A feature to report any failure to conform to ISO C might be useful in
2328 some instances, but would require considerable additional work and would
2329 be quite different from @option{-pedantic}. We don't have plans to
2330 support such a feature in the near future.
2332 Where the standard specified with @option{-std} represents a GNU
2333 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2334 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2335 extended dialect is based. Warnings from @option{-pedantic} are given
2336 where they are required by the base standard. (It would not make sense
2337 for such warnings to be given only for features not in the specified GNU
2338 C dialect, since by definition the GNU dialects of C include all
2339 features the compiler supports with the given option, and there would be
2340 nothing to warn about.)
2342 @item -pedantic-errors
2343 @opindex pedantic-errors
2344 Like @option{-pedantic}, except that errors are produced rather than
2349 Inhibit all warning messages.
2353 Inhibit warning messages about the use of @samp{#import}.
2355 @item -Wchar-subscripts
2356 @opindex Wchar-subscripts
2357 Warn if an array subscript has type @code{char}. This is a common cause
2358 of error, as programmers often forget that this type is signed on some
2360 This warning is enabled by @option{-Wall}.
2364 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2365 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2366 This warning is enabled by @option{-Wall}.
2368 @item -Wfatal-errors
2369 @opindex Wfatal-errors
2370 This option causes the compiler to abort compilation on the first error
2371 occurred rather than trying to keep going and printing further error
2376 @opindex ffreestanding
2377 @opindex fno-builtin
2378 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2379 the arguments supplied have types appropriate to the format string
2380 specified, and that the conversions specified in the format string make
2381 sense. This includes standard functions, and others specified by format
2382 attributes (@pxref{Function Attributes}), in the @code{printf},
2383 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2384 not in the C standard) families (or other target-specific families).
2385 Which functions are checked without format attributes having been
2386 specified depends on the standard version selected, and such checks of
2387 functions without the attribute specified are disabled by
2388 @option{-ffreestanding} or @option{-fno-builtin}.
2390 The formats are checked against the format features supported by GNU
2391 libc version 2.2. These include all ISO C90 and C99 features, as well
2392 as features from the Single Unix Specification and some BSD and GNU
2393 extensions. Other library implementations may not support all these
2394 features; GCC does not support warning about features that go beyond a
2395 particular library's limitations. However, if @option{-pedantic} is used
2396 with @option{-Wformat}, warnings will be given about format features not
2397 in the selected standard version (but not for @code{strfmon} formats,
2398 since those are not in any version of the C standard). @xref{C Dialect
2399 Options,,Options Controlling C Dialect}.
2401 Since @option{-Wformat} also checks for null format arguments for
2402 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2404 @option{-Wformat} is included in @option{-Wall}. For more control over some
2405 aspects of format checking, the options @option{-Wformat-y2k},
2406 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2407 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2408 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2411 @opindex Wformat-y2k
2412 If @option{-Wformat} is specified, also warn about @code{strftime}
2413 formats which may yield only a two-digit year.
2415 @item -Wno-format-extra-args
2416 @opindex Wno-format-extra-args
2417 If @option{-Wformat} is specified, do not warn about excess arguments to a
2418 @code{printf} or @code{scanf} format function. The C standard specifies
2419 that such arguments are ignored.
2421 Where the unused arguments lie between used arguments that are
2422 specified with @samp{$} operand number specifications, normally
2423 warnings are still given, since the implementation could not know what
2424 type to pass to @code{va_arg} to skip the unused arguments. However,
2425 in the case of @code{scanf} formats, this option will suppress the
2426 warning if the unused arguments are all pointers, since the Single
2427 Unix Specification says that such unused arguments are allowed.
2429 @item -Wno-format-zero-length
2430 @opindex Wno-format-zero-length
2431 If @option{-Wformat} is specified, do not warn about zero-length formats.
2432 The C standard specifies that zero-length formats are allowed.
2434 @item -Wformat-nonliteral
2435 @opindex Wformat-nonliteral
2436 If @option{-Wformat} is specified, also warn if the format string is not a
2437 string literal and so cannot be checked, unless the format function
2438 takes its format arguments as a @code{va_list}.
2440 @item -Wformat-security
2441 @opindex Wformat-security
2442 If @option{-Wformat} is specified, also warn about uses of format
2443 functions that represent possible security problems. At present, this
2444 warns about calls to @code{printf} and @code{scanf} functions where the
2445 format string is not a string literal and there are no format arguments,
2446 as in @code{printf (foo);}. This may be a security hole if the format
2447 string came from untrusted input and contains @samp{%n}. (This is
2448 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2449 in future warnings may be added to @option{-Wformat-security} that are not
2450 included in @option{-Wformat-nonliteral}.)
2454 Enable @option{-Wformat} plus format checks not included in
2455 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2456 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2460 Warn about passing a null pointer for arguments marked as
2461 requiring a non-null value by the @code{nonnull} function attribute.
2463 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2464 can be disabled with the @option{-Wno-nonnull} option.
2466 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2468 Warn about uninitialized variables which are initialized with themselves.
2469 Note this option can only be used with the @option{-Wuninitialized} option,
2470 which in turn only works with @option{-O1} and above.
2472 For example, GCC will warn about @code{i} being uninitialized in the
2473 following snippet only when @option{-Winit-self} has been specified:
2484 @item -Wimplicit-int
2485 @opindex Wimplicit-int
2486 Warn when a declaration does not specify a type.
2487 This warning is enabled by @option{-Wall}.
2489 @item -Wimplicit-function-declaration
2490 @itemx -Werror-implicit-function-declaration
2491 @opindex Wimplicit-function-declaration
2492 @opindex Werror-implicit-function-declaration
2493 Give a warning (or error) whenever a function is used before being
2494 declared. The form @option{-Wno-error-implicit-function-declaration}
2496 This warning is enabled by @option{-Wall} (as a warning, not an error).
2500 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2501 This warning is enabled by @option{-Wall}.
2505 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2506 function with external linkage, returning int, taking either zero
2507 arguments, two, or three arguments of appropriate types.
2508 This warning is enabled by @option{-Wall}.
2510 @item -Wmissing-braces
2511 @opindex Wmissing-braces
2512 Warn if an aggregate or union initializer is not fully bracketed. In
2513 the following example, the initializer for @samp{a} is not fully
2514 bracketed, but that for @samp{b} is fully bracketed.
2517 int a[2][2] = @{ 0, 1, 2, 3 @};
2518 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2521 This warning is enabled by @option{-Wall}.
2523 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2524 @opindex Wmissing-include-dirs
2525 Warn if a user-supplied include directory does not exist.
2528 @opindex Wparentheses
2529 Warn if parentheses are omitted in certain contexts, such
2530 as when there is an assignment in a context where a truth value
2531 is expected, or when operators are nested whose precedence people
2532 often get confused about. Only the warning for an assignment used as
2533 a truth value is supported when compiling C++; the other warnings are
2534 only supported when compiling C@.
2536 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2537 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2538 interpretation from that of ordinary mathematical notation.
2540 Also warn about constructions where there may be confusion to which
2541 @code{if} statement an @code{else} branch belongs. Here is an example of
2556 In C, every @code{else} branch belongs to the innermost possible @code{if}
2557 statement, which in this example is @code{if (b)}. This is often not
2558 what the programmer expected, as illustrated in the above example by
2559 indentation the programmer chose. When there is the potential for this
2560 confusion, GCC will issue a warning when this flag is specified.
2561 To eliminate the warning, add explicit braces around the innermost
2562 @code{if} statement so there is no way the @code{else} could belong to
2563 the enclosing @code{if}. The resulting code would look like this:
2579 This warning is enabled by @option{-Wall}.
2581 @item -Wsequence-point
2582 @opindex Wsequence-point
2583 Warn about code that may have undefined semantics because of violations
2584 of sequence point rules in the C and C++ standards.
2586 The C and C++ standards defines the order in which expressions in a C/C++
2587 program are evaluated in terms of @dfn{sequence points}, which represent
2588 a partial ordering between the execution of parts of the program: those
2589 executed before the sequence point, and those executed after it. These
2590 occur after the evaluation of a full expression (one which is not part
2591 of a larger expression), after the evaluation of the first operand of a
2592 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2593 function is called (but after the evaluation of its arguments and the
2594 expression denoting the called function), and in certain other places.
2595 Other than as expressed by the sequence point rules, the order of
2596 evaluation of subexpressions of an expression is not specified. All
2597 these rules describe only a partial order rather than a total order,
2598 since, for example, if two functions are called within one expression
2599 with no sequence point between them, the order in which the functions
2600 are called is not specified. However, the standards committee have
2601 ruled that function calls do not overlap.
2603 It is not specified when between sequence points modifications to the
2604 values of objects take effect. Programs whose behavior depends on this
2605 have undefined behavior; the C and C++ standards specify that ``Between
2606 the previous and next sequence point an object shall have its stored
2607 value modified at most once by the evaluation of an expression.
2608 Furthermore, the prior value shall be read only to determine the value
2609 to be stored.''. If a program breaks these rules, the results on any
2610 particular implementation are entirely unpredictable.
2612 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2613 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2614 diagnosed by this option, and it may give an occasional false positive
2615 result, but in general it has been found fairly effective at detecting
2616 this sort of problem in programs.
2618 The standard is worded confusingly, therefore there is some debate
2619 over the precise meaning of the sequence point rules in subtle cases.
2620 Links to discussions of the problem, including proposed formal
2621 definitions, may be found on the GCC readings page, at
2622 @w{@uref{http://gcc.gnu.org/readings.html}}.
2624 This warning is enabled by @option{-Wall} for C and C++.
2627 @opindex Wreturn-type
2628 Warn whenever a function is defined with a return-type that defaults to
2629 @code{int}. Also warn about any @code{return} statement with no
2630 return-value in a function whose return-type is not @code{void}.
2632 For C, also warn if the return type of a function has a type qualifier
2633 such as @code{const}. Such a type qualifier has no effect, since the
2634 value returned by a function is not an lvalue. ISO C prohibits
2635 qualified @code{void} return types on function definitions, so such
2636 return types always receive a warning even without this option.
2638 For C++, a function without return type always produces a diagnostic
2639 message, even when @option{-Wno-return-type} is specified. The only
2640 exceptions are @samp{main} and functions defined in system headers.
2642 This warning is enabled by @option{-Wall}.
2646 Warn whenever a @code{switch} statement has an index of enumerated type
2647 and lacks a @code{case} for one or more of the named codes of that
2648 enumeration. (The presence of a @code{default} label prevents this
2649 warning.) @code{case} labels outside the enumeration range also
2650 provoke warnings when this option is used.
2651 This warning is enabled by @option{-Wall}.
2653 @item -Wswitch-default
2654 @opindex Wswitch-switch
2655 Warn whenever a @code{switch} statement does not have a @code{default}
2659 @opindex Wswitch-enum
2660 Warn whenever a @code{switch} statement has an index of enumerated type
2661 and lacks a @code{case} for one or more of the named codes of that
2662 enumeration. @code{case} labels outside the enumeration range also
2663 provoke warnings when this option is used.
2667 Warn if any trigraphs are encountered that might change the meaning of
2668 the program (trigraphs within comments are not warned about).
2669 This warning is enabled by @option{-Wall}.
2671 @item -Wunused-function
2672 @opindex Wunused-function
2673 Warn whenever a static function is declared but not defined or a
2674 non-inline static function is unused.
2675 This warning is enabled by @option{-Wall}.
2677 @item -Wunused-label
2678 @opindex Wunused-label
2679 Warn whenever a label is declared but not used.
2680 This warning is enabled by @option{-Wall}.
2682 To suppress this warning use the @samp{unused} attribute
2683 (@pxref{Variable Attributes}).
2685 @item -Wunused-parameter
2686 @opindex Wunused-parameter
2687 Warn whenever a function parameter is unused aside from its declaration.
2689 To suppress this warning use the @samp{unused} attribute
2690 (@pxref{Variable Attributes}).
2692 @item -Wunused-variable
2693 @opindex Wunused-variable
2694 Warn whenever a local variable or non-constant static variable is unused
2695 aside from its declaration.
2696 This warning is enabled by @option{-Wall}.
2698 To suppress this warning use the @samp{unused} attribute
2699 (@pxref{Variable Attributes}).
2701 @item -Wunused-value
2702 @opindex Wunused-value
2703 Warn whenever a statement computes a result that is explicitly not used.
2704 This warning is enabled by @option{-Wall}.
2706 To suppress this warning cast the expression to @samp{void}.
2710 All the above @option{-Wunused} options combined.
2712 In order to get a warning about an unused function parameter, you must
2713 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2714 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2716 @item -Wuninitialized
2717 @opindex Wuninitialized
2718 Warn if an automatic variable is used without first being initialized or
2719 if a variable may be clobbered by a @code{setjmp} call.
2721 These warnings are possible only in optimizing compilation,
2722 because they require data flow information that is computed only
2723 when optimizing. If you do not specify @option{-O}, you will not get
2724 these warnings. Instead, GCC will issue a warning about @option{-Wuninitialized}
2725 requiring @option{-O}.
2727 If you want to warn about code which uses the uninitialized value of the
2728 variable in its own initializer, use the @option{-Winit-self} option.
2730 These warnings occur for individual uninitialized or clobbered
2731 elements of structure, union or array variables as well as for
2732 variables which are uninitialized or clobbered as a whole. They do
2733 not occur for variables or elements declared @code{volatile}. Because
2734 these warnings depend on optimization, the exact variables or elements
2735 for which there are warnings will depend on the precise optimization
2736 options and version of GCC used.
2738 Note that there may be no warning about a variable that is used only
2739 to compute a value that itself is never used, because such
2740 computations may be deleted by data flow analysis before the warnings
2743 These warnings are made optional because GCC is not smart
2744 enough to see all the reasons why the code might be correct
2745 despite appearing to have an error. Here is one example of how
2766 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2767 always initialized, but GCC doesn't know this. Here is
2768 another common case:
2773 if (change_y) save_y = y, y = new_y;
2775 if (change_y) y = save_y;
2780 This has no bug because @code{save_y} is used only if it is set.
2782 @cindex @code{longjmp} warnings
2783 This option also warns when a non-volatile automatic variable might be
2784 changed by a call to @code{longjmp}. These warnings as well are possible
2785 only in optimizing compilation.
2787 The compiler sees only the calls to @code{setjmp}. It cannot know
2788 where @code{longjmp} will be called; in fact, a signal handler could
2789 call it at any point in the code. As a result, you may get a warning
2790 even when there is in fact no problem because @code{longjmp} cannot
2791 in fact be called at the place which would cause a problem.
2793 Some spurious warnings can be avoided if you declare all the functions
2794 you use that never return as @code{noreturn}. @xref{Function
2797 This warning is enabled by @option{-Wall}.
2799 @item -Wunknown-pragmas
2800 @opindex Wunknown-pragmas
2801 @cindex warning for unknown pragmas
2802 @cindex unknown pragmas, warning
2803 @cindex pragmas, warning of unknown
2804 Warn when a #pragma directive is encountered which is not understood by
2805 GCC@. If this command line option is used, warnings will even be issued
2806 for unknown pragmas in system header files. This is not the case if
2807 the warnings were only enabled by the @option{-Wall} command line option.
2810 @opindex Wno-pragmas
2812 Do not warn about misuses of pragmas, such as incorrect parameters,
2813 invalid syntax, or conflicts between pragmas. See also
2814 @samp{-Wunknown-pragmas}.
2816 @item -Wstrict-aliasing
2817 @opindex Wstrict-aliasing
2818 This option is only active when @option{-fstrict-aliasing} is active.
2819 It warns about code which might break the strict aliasing rules that the
2820 compiler is using for optimization. The warning does not catch all
2821 cases, but does attempt to catch the more common pitfalls. It is
2822 included in @option{-Wall}.
2824 @item -Wstrict-aliasing=2
2825 @opindex Wstrict-aliasing=2
2826 This option is only active when @option{-fstrict-aliasing} is active.
2827 It warns about code which might break the strict aliasing rules that the
2828 compiler is using for optimization. This warning catches more cases than
2829 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2830 cases that are safe.
2834 All of the above @samp{-W} options combined. This enables all the
2835 warnings about constructions that some users consider questionable, and
2836 that are easy to avoid (or modify to prevent the warning), even in
2837 conjunction with macros. This also enables some language-specific
2838 warnings described in @ref{C++ Dialect Options} and
2839 @ref{Objective-C and Objective-C++ Dialect Options}.
2842 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2843 Some of them warn about constructions that users generally do not
2844 consider questionable, but which occasionally you might wish to check
2845 for; others warn about constructions that are necessary or hard to avoid
2846 in some cases, and there is no simple way to modify the code to suppress
2853 (This option used to be called @option{-W}. The older name is still
2854 supported, but the newer name is more descriptive.) Print extra warning
2855 messages for these events:
2859 A function can return either with or without a value. (Falling
2860 off the end of the function body is considered returning without
2861 a value.) For example, this function would evoke such a
2875 An expression-statement or the left-hand side of a comma expression
2876 contains no side effects.
2877 To suppress the warning, cast the unused expression to void.
2878 For example, an expression such as @samp{x[i,j]} will cause a warning,
2879 but @samp{x[(void)i,j]} will not.
2882 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2885 Storage-class specifiers like @code{static} are not the first things in
2886 a declaration. According to the C Standard, this usage is obsolescent.
2889 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2893 A comparison between signed and unsigned values could produce an
2894 incorrect result when the signed value is converted to unsigned.
2895 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2898 An aggregate has an initializer which does not initialize all members.
2899 This warning can be independently controlled by
2900 @option{-Wmissing-field-initializers}.
2903 An initialized field without side effects is overridden when using
2904 designated initializers (@pxref{Designated Inits, , Designated
2905 Initializers}). This warning can be independently controlled by
2906 @option{-Woverride-init}.
2909 A function parameter is declared without a type specifier in K&R-style
2917 An empty body occurs in an @samp{if} or @samp{else} statement.
2920 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2921 @samp{>}, or @samp{>=}.
2924 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2927 Any of several floating-point events that often indicate errors, such as
2928 overflow, underflow, loss of precision, etc.
2930 @item @r{(C++ only)}
2931 An enumerator and a non-enumerator both appear in a conditional expression.
2933 @item @r{(C++ only)}
2934 A non-static reference or non-static @samp{const} member appears in a
2935 class without constructors.
2937 @item @r{(C++ only)}
2938 Ambiguous virtual bases.
2940 @item @r{(C++ only)}
2941 Subscripting an array which has been declared @samp{register}.
2943 @item @r{(C++ only)}
2944 Taking the address of a variable which has been declared @samp{register}.
2946 @item @r{(C++ only)}
2947 A base class is not initialized in a derived class' copy constructor.
2950 @item -Wno-div-by-zero
2951 @opindex Wno-div-by-zero
2952 @opindex Wdiv-by-zero
2953 Do not warn about compile-time integer division by zero. Floating point
2954 division by zero is not warned about, as it can be a legitimate way of
2955 obtaining infinities and NaNs.
2957 @item -Wsystem-headers
2958 @opindex Wsystem-headers
2959 @cindex warnings from system headers
2960 @cindex system headers, warnings from
2961 Print warning messages for constructs found in system header files.
2962 Warnings from system headers are normally suppressed, on the assumption
2963 that they usually do not indicate real problems and would only make the
2964 compiler output harder to read. Using this command line option tells
2965 GCC to emit warnings from system headers as if they occurred in user
2966 code. However, note that using @option{-Wall} in conjunction with this
2967 option will @emph{not} warn about unknown pragmas in system
2968 headers---for that, @option{-Wunknown-pragmas} must also be used.
2971 @opindex Wfloat-equal
2972 Warn if floating point values are used in equality comparisons.
2974 The idea behind this is that sometimes it is convenient (for the
2975 programmer) to consider floating-point values as approximations to
2976 infinitely precise real numbers. If you are doing this, then you need
2977 to compute (by analyzing the code, or in some other way) the maximum or
2978 likely maximum error that the computation introduces, and allow for it
2979 when performing comparisons (and when producing output, but that's a
2980 different problem). In particular, instead of testing for equality, you
2981 would check to see whether the two values have ranges that overlap; and
2982 this is done with the relational operators, so equality comparisons are
2985 @item -Wtraditional @r{(C only)}
2986 @opindex Wtraditional
2987 Warn about certain constructs that behave differently in traditional and
2988 ISO C@. Also warn about ISO C constructs that have no traditional C
2989 equivalent, and/or problematic constructs which should be avoided.
2993 Macro parameters that appear within string literals in the macro body.
2994 In traditional C macro replacement takes place within string literals,
2995 but does not in ISO C@.
2998 In traditional C, some preprocessor directives did not exist.
2999 Traditional preprocessors would only consider a line to be a directive
3000 if the @samp{#} appeared in column 1 on the line. Therefore
3001 @option{-Wtraditional} warns about directives that traditional C
3002 understands but would ignore because the @samp{#} does not appear as the
3003 first character on the line. It also suggests you hide directives like
3004 @samp{#pragma} not understood by traditional C by indenting them. Some
3005 traditional implementations would not recognize @samp{#elif}, so it
3006 suggests avoiding it altogether.
3009 A function-like macro that appears without arguments.
3012 The unary plus operator.
3015 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
3016 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
3017 constants.) Note, these suffixes appear in macros defined in the system
3018 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
3019 Use of these macros in user code might normally lead to spurious
3020 warnings, however GCC's integrated preprocessor has enough context to
3021 avoid warning in these cases.
3024 A function declared external in one block and then used after the end of
3028 A @code{switch} statement has an operand of type @code{long}.
3031 A non-@code{static} function declaration follows a @code{static} one.
3032 This construct is not accepted by some traditional C compilers.
3035 The ISO type of an integer constant has a different width or
3036 signedness from its traditional type. This warning is only issued if
3037 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
3038 typically represent bit patterns, are not warned about.
3041 Usage of ISO string concatenation is detected.
3044 Initialization of automatic aggregates.
3047 Identifier conflicts with labels. Traditional C lacks a separate
3048 namespace for labels.
3051 Initialization of unions. If the initializer is zero, the warning is
3052 omitted. This is done under the assumption that the zero initializer in
3053 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
3054 initializer warnings and relies on default initialization to zero in the
3058 Conversions by prototypes between fixed/floating point values and vice
3059 versa. The absence of these prototypes when compiling with traditional
3060 C would cause serious problems. This is a subset of the possible
3061 conversion warnings, for the full set use @option{-Wconversion}.
3064 Use of ISO C style function definitions. This warning intentionally is
3065 @emph{not} issued for prototype declarations or variadic functions
3066 because these ISO C features will appear in your code when using
3067 libiberty's traditional C compatibility macros, @code{PARAMS} and
3068 @code{VPARAMS}. This warning is also bypassed for nested functions
3069 because that feature is already a GCC extension and thus not relevant to
3070 traditional C compatibility.
3073 @item -Wdeclaration-after-statement @r{(C only)}
3074 @opindex Wdeclaration-after-statement
3075 Warn when a declaration is found after a statement in a block. This
3076 construct, known from C++, was introduced with ISO C99 and is by default
3077 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3078 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3082 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3084 @item -Wno-endif-labels
3085 @opindex Wno-endif-labels
3086 @opindex Wendif-labels
3087 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3091 Warn whenever a local variable shadows another local variable, parameter or
3092 global variable or whenever a built-in function is shadowed.
3094 @item -Wlarger-than-@var{len}
3095 @opindex Wlarger-than
3096 Warn whenever an object of larger than @var{len} bytes is defined.
3098 @item -Wunsafe-loop-optimizations
3099 @opindex Wunsafe-loop-optimizations
3100 Warn if the loop cannot be optimized because the compiler could not
3101 assume anything on the bounds of the loop indices. With
3102 @option{-funsafe-loop-optimizations} warn if the compiler made
3105 @item -Wpointer-arith
3106 @opindex Wpointer-arith
3107 Warn about anything that depends on the ``size of'' a function type or
3108 of @code{void}. GNU C assigns these types a size of 1, for
3109 convenience in calculations with @code{void *} pointers and pointers
3112 @item -Wbad-function-cast @r{(C only)}
3113 @opindex Wbad-function-cast
3114 Warn whenever a function call is cast to a non-matching type.
3115 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3118 Warn about ISO C constructs that are outside of the common subset of
3119 ISO C and ISO C++, e.g.@: request for implicit conversion from
3120 @code{void *} to a pointer to non-@code{void} type.
3124 Warn whenever a pointer is cast so as to remove a type qualifier from
3125 the target type. For example, warn if a @code{const char *} is cast
3126 to an ordinary @code{char *}.
3129 @opindex Wcast-align
3130 Warn whenever a pointer is cast such that the required alignment of the
3131 target is increased. For example, warn if a @code{char *} is cast to
3132 an @code{int *} on machines where integers can only be accessed at
3133 two- or four-byte boundaries.
3135 @item -Wwrite-strings
3136 @opindex Wwrite-strings
3137 When compiling C, give string constants the type @code{const
3138 char[@var{length}]} so that
3139 copying the address of one into a non-@code{const} @code{char *}
3140 pointer will get a warning; when compiling C++, warn about the
3141 deprecated conversion from string literals to @code{char *}. This
3142 warning, by default, is enabled for C++ programs.
3143 These warnings will help you find at
3144 compile time code that can try to write into a string constant, but
3145 only if you have been very careful about using @code{const} in
3146 declarations and prototypes. Otherwise, it will just be a nuisance;
3147 this is why we did not make @option{-Wall} request these warnings.
3150 @opindex Wconversion
3151 Warn if a prototype causes a type conversion that is different from what
3152 would happen to the same argument in the absence of a prototype. This
3153 includes conversions of fixed point to floating and vice versa, and
3154 conversions changing the width or signedness of a fixed point argument
3155 except when the same as the default promotion.
3157 Also, warn if a negative integer constant expression is implicitly
3158 converted to an unsigned type. For example, warn about the assignment
3159 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3160 casts like @code{(unsigned) -1}.
3162 @item -Wsign-compare
3163 @opindex Wsign-compare
3164 @cindex warning for comparison of signed and unsigned values
3165 @cindex comparison of signed and unsigned values, warning
3166 @cindex signed and unsigned values, comparison warning
3167 Warn when a comparison between signed and unsigned values could produce
3168 an incorrect result when the signed value is converted to unsigned.
3169 This warning is also enabled by @option{-Wextra}; to get the other warnings
3170 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3172 @item -Waggregate-return
3173 @opindex Waggregate-return
3174 Warn if any functions that return structures or unions are defined or
3175 called. (In languages where you can return an array, this also elicits
3179 @opindex Walways-true
3180 Warn about comparisons which are always true such as testing if
3181 unsigned values are greater than or equal to zero. This warning is
3182 enabled by @option{-Wall}.
3184 @item -Wno-attributes
3185 @opindex Wno-attributes
3186 @opindex Wattributes
3187 Do not warn if an unexpected @code{__attribute__} is used, such as
3188 unrecognized attributes, function attributes applied to variables,
3189 etc. This will not stop errors for incorrect use of supported
3192 @item -Wstrict-prototypes @r{(C only)}
3193 @opindex Wstrict-prototypes
3194 Warn if a function is declared or defined without specifying the
3195 argument types. (An old-style function definition is permitted without
3196 a warning if preceded by a declaration which specifies the argument
3199 @item -Wold-style-definition @r{(C only)}
3200 @opindex Wold-style-definition
3201 Warn if an old-style function definition is used. A warning is given
3202 even if there is a previous prototype.
3204 @item -Wmissing-prototypes @r{(C only)}
3205 @opindex Wmissing-prototypes
3206 Warn if a global function is defined without a previous prototype
3207 declaration. This warning is issued even if the definition itself
3208 provides a prototype. The aim is to detect global functions that fail
3209 to be declared in header files.
3211 @item -Wmissing-declarations @r{(C only)}
3212 @opindex Wmissing-declarations
3213 Warn if a global function is defined without a previous declaration.
3214 Do so even if the definition itself provides a prototype.
3215 Use this option to detect global functions that are not declared in
3218 @item -Wmissing-field-initializers
3219 @opindex Wmissing-field-initializers
3222 Warn if a structure's initializer has some fields missing. For
3223 example, the following code would cause such a warning, because
3224 @code{x.h} is implicitly zero:
3227 struct s @{ int f, g, h; @};
3228 struct s x = @{ 3, 4 @};
3231 This option does not warn about designated initializers, so the following
3232 modification would not trigger a warning:
3235 struct s @{ int f, g, h; @};
3236 struct s x = @{ .f = 3, .g = 4 @};
3239 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3240 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3242 @item -Wmissing-noreturn
3243 @opindex Wmissing-noreturn
3244 Warn about functions which might be candidates for attribute @code{noreturn}.
3245 Note these are only possible candidates, not absolute ones. Care should
3246 be taken to manually verify functions actually do not ever return before
3247 adding the @code{noreturn} attribute, otherwise subtle code generation
3248 bugs could be introduced. You will not get a warning for @code{main} in
3249 hosted C environments.
3251 @item -Wmissing-format-attribute
3252 @opindex Wmissing-format-attribute
3254 Warn about function pointers which might be candidates for @code{format}
3255 attributes. Note these are only possible candidates, not absolute ones.
3256 GCC will guess that function pointers with @code{format} attributes that
3257 are used in assignment, initialization, parameter passing or return
3258 statements should have a corresponding @code{format} attribute in the
3259 resulting type. I.e.@: the left-hand side of the assignment or
3260 initialization, the type of the parameter variable, or the return type
3261 of the containing function respectively should also have a @code{format}
3262 attribute to avoid the warning.
3264 GCC will also warn about function definitions which might be
3265 candidates for @code{format} attributes. Again, these are only
3266 possible candidates. GCC will guess that @code{format} attributes
3267 might be appropriate for any function that calls a function like
3268 @code{vprintf} or @code{vscanf}, but this might not always be the
3269 case, and some functions for which @code{format} attributes are
3270 appropriate may not be detected.
3272 @item -Wno-multichar
3273 @opindex Wno-multichar
3275 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3276 Usually they indicate a typo in the user's code, as they have
3277 implementation-defined values, and should not be used in portable code.
3279 @item -Wnormalized=<none|id|nfc|nfkc>
3280 @opindex Wnormalized
3283 @cindex character set, input normalization
3284 In ISO C and ISO C++, two identifiers are different if they are
3285 different sequences of characters. However, sometimes when characters
3286 outside the basic ASCII character set are used, you can have two
3287 different character sequences that look the same. To avoid confusion,
3288 the ISO 10646 standard sets out some @dfn{normalization rules} which
3289 when applied ensure that two sequences that look the same are turned into
3290 the same sequence. GCC can warn you if you are using identifiers which
3291 have not been normalized; this option controls that warning.
3293 There are four levels of warning that GCC supports. The default is
3294 @option{-Wnormalized=nfc}, which warns about any identifier which is
3295 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3296 recommended form for most uses.
3298 Unfortunately, there are some characters which ISO C and ISO C++ allow
3299 in identifiers that when turned into NFC aren't allowable as
3300 identifiers. That is, there's no way to use these symbols in portable
3301 ISO C or C++ and have all your identifiers in NFC.
3302 @option{-Wnormalized=id} suppresses the warning for these characters.
3303 It is hoped that future versions of the standards involved will correct
3304 this, which is why this option is not the default.
3306 You can switch the warning off for all characters by writing
3307 @option{-Wnormalized=none}. You would only want to do this if you
3308 were using some other normalization scheme (like ``D''), because
3309 otherwise you can easily create bugs that are literally impossible to see.
3311 Some characters in ISO 10646 have distinct meanings but look identical
3312 in some fonts or display methodologies, especially once formatting has
3313 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3314 LETTER N'', will display just like a regular @code{n} which has been
3315 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3316 normalization scheme to convert all these into a standard form as
3317 well, and GCC will warn if your code is not in NFKC if you use
3318 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3319 about every identifier that contains the letter O because it might be
3320 confused with the digit 0, and so is not the default, but may be
3321 useful as a local coding convention if the programming environment is
3322 unable to be fixed to display these characters distinctly.
3324 @item -Wno-deprecated-declarations
3325 @opindex Wno-deprecated-declarations
3326 Do not warn about uses of functions (@pxref{Function Attributes}),
3327 variables (@pxref{Variable Attributes}), and types (@pxref{Type
3328 Attributes}) marked as deprecated by using the @code{deprecated}
3332 @opindex Wno-overflow
3333 Do not warn about compile-time overflow in constant expressions.
3335 @item -Woverride-init
3336 @opindex Woverride-init
3339 Warn if an initialized field without side effects is overridden when
3340 using designated initializers (@pxref{Designated Inits, , Designated
3343 This warning is included in @option{-Wextra}. To get other
3344 @option{-Wextra} warnings without this one, use @samp{-Wextra
3345 -Wno-override-init}.
3349 Warn if a structure is given the packed attribute, but the packed
3350 attribute has no effect on the layout or size of the structure.
3351 Such structures may be mis-aligned for little benefit. For
3352 instance, in this code, the variable @code{f.x} in @code{struct bar}
3353 will be misaligned even though @code{struct bar} does not itself
3354 have the packed attribute:
3361 @} __attribute__((packed));
3371 Warn if padding is included in a structure, either to align an element
3372 of the structure or to align the whole structure. Sometimes when this
3373 happens it is possible to rearrange the fields of the structure to
3374 reduce the padding and so make the structure smaller.
3376 @item -Wredundant-decls
3377 @opindex Wredundant-decls
3378 Warn if anything is declared more than once in the same scope, even in
3379 cases where multiple declaration is valid and changes nothing.
3381 @item -Wnested-externs @r{(C only)}
3382 @opindex Wnested-externs
3383 Warn if an @code{extern} declaration is encountered within a function.
3385 @item -Wunreachable-code
3386 @opindex Wunreachable-code
3387 Warn if the compiler detects that code will never be executed.
3389 This option is intended to warn when the compiler detects that at
3390 least a whole line of source code will never be executed, because
3391 some condition is never satisfied or because it is after a
3392 procedure that never returns.
3394 It is possible for this option to produce a warning even though there
3395 are circumstances under which part of the affected line can be executed,
3396 so care should be taken when removing apparently-unreachable code.
3398 For instance, when a function is inlined, a warning may mean that the
3399 line is unreachable in only one inlined copy of the function.
3401 This option is not made part of @option{-Wall} because in a debugging
3402 version of a program there is often substantial code which checks
3403 correct functioning of the program and is, hopefully, unreachable
3404 because the program does work. Another common use of unreachable
3405 code is to provide behavior which is selectable at compile-time.
3409 Warn if a function can not be inlined and it was declared as inline.
3410 Even with this option, the compiler will not warn about failures to
3411 inline functions declared in system headers.
3413 The compiler uses a variety of heuristics to determine whether or not
3414 to inline a function. For example, the compiler takes into account
3415 the size of the function being inlined and the amount of inlining
3416 that has already been done in the current function. Therefore,
3417 seemingly insignificant changes in the source program can cause the
3418 warnings produced by @option{-Winline} to appear or disappear.
3420 @item -Wno-invalid-offsetof @r{(C++ only)}
3421 @opindex Wno-invalid-offsetof
3422 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3423 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3424 to a non-POD type is undefined. In existing C++ implementations,
3425 however, @samp{offsetof} typically gives meaningful results even when
3426 applied to certain kinds of non-POD types. (Such as a simple
3427 @samp{struct} that fails to be a POD type only by virtue of having a
3428 constructor.) This flag is for users who are aware that they are
3429 writing nonportable code and who have deliberately chosen to ignore the
3432 The restrictions on @samp{offsetof} may be relaxed in a future version
3433 of the C++ standard.
3435 @item -Wno-int-to-pointer-cast @r{(C only)}
3436 @opindex Wno-int-to-pointer-cast
3437 Suppress warnings from casts to pointer type of an integer of a
3440 @item -Wno-pointer-to-int-cast @r{(C only)}
3441 @opindex Wno-pointer-to-int-cast
3442 Suppress warnings from casts from a pointer to an integer type of a
3446 @opindex Winvalid-pch
3447 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3448 the search path but can't be used.
3452 @opindex Wno-long-long
3453 Warn if @samp{long long} type is used. This is default. To inhibit
3454 the warning messages, use @option{-Wno-long-long}. Flags
3455 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3456 only when @option{-pedantic} flag is used.
3458 @item -Wvariadic-macros
3459 @opindex Wvariadic-macros
3460 @opindex Wno-variadic-macros
3461 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3462 alternate syntax when in pedantic ISO C99 mode. This is default.
3463 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3465 @item -Wvolatile-register-var
3466 @opindex Wvolatile-register-var
3467 @opindex Wno-volatile-register-var
3468 Warn if a register variable is declared volatile. The volatile
3469 modifier does not inhibit all optimizations that may eliminate reads
3470 and/or writes to register variables.
3472 @item -Wdisabled-optimization
3473 @opindex Wdisabled-optimization
3474 Warn if a requested optimization pass is disabled. This warning does
3475 not generally indicate that there is anything wrong with your code; it
3476 merely indicates that GCC's optimizers were unable to handle the code
3477 effectively. Often, the problem is that your code is too big or too
3478 complex; GCC will refuse to optimize programs when the optimization
3479 itself is likely to take inordinate amounts of time.
3481 @item -Wpointer-sign
3482 @opindex Wpointer-sign
3483 @opindex Wno-pointer-sign
3484 Warn for pointer argument passing or assignment with different signedness.
3485 This option is only supported for C and Objective-C@. It is implied by
3486 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3487 @option{-Wno-pointer-sign}.
3491 Make all warnings into errors.
3495 Make the specified warning into an errors. The specifier for a
3496 warning is appended, for example @option{-Werror=switch} turns the
3497 warnings controlled by @option{-Wswitch} into errors. This switch
3498 takes a negative form, to be used to negate @option{-Werror} for
3499 specific warnings, for example @option{-Wno-error=switch} makes
3500 @option{-Wswitch} warnings not be errors, even when @option{-Werror}
3501 is in effect. You can use the @option{-fdiagnostics-show-option}
3502 option to have each controllable warning amended with the option which
3503 controls it, to determine what to use with this option.
3505 Note that specifying @option{-Werror=}@var{foo} automatically implies
3506 @option{-W}@var{foo}. However, @option{-Wno-error=}@var{foo} does not
3509 @item -Wstack-protector
3510 @opindex Wstack-protector
3511 This option is only active when @option{-fstack-protector} is active. It
3512 warns about functions that will not be protected against stack smashing.
3514 @item -Wstring-literal-comparison
3515 @opindex Wstring-literal-comparison
3516 Warn about suspicious comparisons to string literal constants. In C,
3517 direct comparisons against the memory address of a string literal, such
3518 as @code{if (x == "abc")}, typically indicate a programmer error, and
3519 even when intentional, result in unspecified behavior and are not portable.
3520 Usually these warnings alert that the programmer intended to use
3521 @code{strcmp}. This warning is enabled by @option{-Wall}.
3523 @item -Woverlength-strings
3524 @opindex Woverlength-strings
3525 Warn about string constants which are longer than the ``minimum
3526 maximum'' length specified in the C standard. Modern compilers
3527 generally allow string constants which are much longer than the
3528 standard's minimum limit, but very portable programs should avoid
3529 using longer strings.
3531 The limit applies @emph{after} string constant concatenation, and does
3532 not count the trailing NUL@. In C89, the limit was 509 characters; in
3533 C99, it was raised to 4095. C++98 does not specify a normative
3534 minimum maximum, so we do not diagnose overlength strings in C++@.
3536 This option is implied by @option{-pedantic}, and can be disabled with
3537 @option{-Wno-overlength-strings}.
3540 @node Debugging Options
3541 @section Options for Debugging Your Program or GCC
3542 @cindex options, debugging
3543 @cindex debugging information options
3545 GCC has various special options that are used for debugging
3546 either your program or GCC:
3551 Produce debugging information in the operating system's native format
3552 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3555 On most systems that use stabs format, @option{-g} enables use of extra
3556 debugging information that only GDB can use; this extra information
3557 makes debugging work better in GDB but will probably make other debuggers
3559 refuse to read the program. If you want to control for certain whether
3560 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3561 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3563 GCC allows you to use @option{-g} with
3564 @option{-O}. The shortcuts taken by optimized code may occasionally
3565 produce surprising results: some variables you declared may not exist
3566 at all; flow of control may briefly move where you did not expect it;
3567 some statements may not be executed because they compute constant
3568 results or their values were already at hand; some statements may
3569 execute in different places because they were moved out of loops.
3571 Nevertheless it proves possible to debug optimized output. This makes
3572 it reasonable to use the optimizer for programs that might have bugs.
3574 The following options are useful when GCC is generated with the
3575 capability for more than one debugging format.
3579 Produce debugging information for use by GDB@. This means to use the
3580 most expressive format available (DWARF 2, stabs, or the native format
3581 if neither of those are supported), including GDB extensions if at all
3586 Produce debugging information in stabs format (if that is supported),
3587 without GDB extensions. This is the format used by DBX on most BSD
3588 systems. On MIPS, Alpha and System V Release 4 systems this option
3589 produces stabs debugging output which is not understood by DBX or SDB@.
3590 On System V Release 4 systems this option requires the GNU assembler.
3592 @item -feliminate-unused-debug-symbols
3593 @opindex feliminate-unused-debug-symbols
3594 Produce debugging information in stabs format (if that is supported),
3595 for only symbols that are actually used.
3597 @item -femit-class-debug-always
3598 Instead of emitting debugging information for a C++ class in only one
3599 object file, emit it in all object files using the class. This option
3600 should be used only with debuggers that are unable to handle the way GCC
3601 normally emits debugging information for classes because using this
3602 option will increase the size of debugging information by as much as a
3607 Produce debugging information in stabs format (if that is supported),
3608 using GNU extensions understood only by the GNU debugger (GDB)@. The
3609 use of these extensions is likely to make other debuggers crash or
3610 refuse to read the program.
3614 Produce debugging information in COFF format (if that is supported).
3615 This is the format used by SDB on most System V systems prior to
3620 Produce debugging information in XCOFF format (if that is supported).
3621 This is the format used by the DBX debugger on IBM RS/6000 systems.
3625 Produce debugging information in XCOFF format (if that is supported),
3626 using GNU extensions understood only by the GNU debugger (GDB)@. The
3627 use of these extensions is likely to make other debuggers crash or
3628 refuse to read the program, and may cause assemblers other than the GNU
3629 assembler (GAS) to fail with an error.
3633 Produce debugging information in DWARF version 2 format (if that is
3634 supported). This is the format used by DBX on IRIX 6. With this
3635 option, GCC uses features of DWARF version 3 when they are useful;
3636 version 3 is upward compatible with version 2, but may still cause
3637 problems for older debuggers.
3641 Produce debugging information in VMS debug format (if that is
3642 supported). This is the format used by DEBUG on VMS systems.
3645 @itemx -ggdb@var{level}
3646 @itemx -gstabs@var{level}
3647 @itemx -gcoff@var{level}
3648 @itemx -gxcoff@var{level}
3649 @itemx -gvms@var{level}
3650 Request debugging information and also use @var{level} to specify how
3651 much information. The default level is 2.
3653 Level 1 produces minimal information, enough for making backtraces in
3654 parts of the program that you don't plan to debug. This includes
3655 descriptions of functions and external variables, but no information
3656 about local variables and no line numbers.
3658 Level 3 includes extra information, such as all the macro definitions
3659 present in the program. Some debuggers support macro expansion when
3660 you use @option{-g3}.
3662 @option{-gdwarf-2} does not accept a concatenated debug level, because
3663 GCC used to support an option @option{-gdwarf} that meant to generate
3664 debug information in version 1 of the DWARF format (which is very
3665 different from version 2), and it would have been too confusing. That
3666 debug format is long obsolete, but the option cannot be changed now.
3667 Instead use an additional @option{-g@var{level}} option to change the
3668 debug level for DWARF2.
3670 @item -feliminate-dwarf2-dups
3671 @opindex feliminate-dwarf2-dups
3672 Compress DWARF2 debugging information by eliminating duplicated
3673 information about each symbol. This option only makes sense when
3674 generating DWARF2 debugging information with @option{-gdwarf-2}.
3676 @cindex @command{prof}
3679 Generate extra code to write profile information suitable for the
3680 analysis program @command{prof}. You must use this option when compiling
3681 the source files you want data about, and you must also use it when
3684 @cindex @command{gprof}
3687 Generate extra code to write profile information suitable for the
3688 analysis program @command{gprof}. You must use this option when compiling
3689 the source files you want data about, and you must also use it when
3694 Makes the compiler print out each function name as it is compiled, and
3695 print some statistics about each pass when it finishes.
3698 @opindex ftime-report
3699 Makes the compiler print some statistics about the time consumed by each
3700 pass when it finishes.
3703 @opindex fmem-report
3704 Makes the compiler print some statistics about permanent memory
3705 allocation when it finishes.
3707 @item -fprofile-arcs
3708 @opindex fprofile-arcs
3709 Add code so that program flow @dfn{arcs} are instrumented. During
3710 execution the program records how many times each branch and call is
3711 executed and how many times it is taken or returns. When the compiled
3712 program exits it saves this data to a file called
3713 @file{@var{auxname}.gcda} for each source file. The data may be used for
3714 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3715 test coverage analysis (@option{-ftest-coverage}). Each object file's
3716 @var{auxname} is generated from the name of the output file, if
3717 explicitly specified and it is not the final executable, otherwise it is
3718 the basename of the source file. In both cases any suffix is removed
3719 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3720 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3721 @xref{Cross-profiling}.
3723 @cindex @command{gcov}
3727 This option is used to compile and link code instrumented for coverage
3728 analysis. The option is a synonym for @option{-fprofile-arcs}
3729 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3730 linking). See the documentation for those options for more details.
3735 Compile the source files with @option{-fprofile-arcs} plus optimization
3736 and code generation options. For test coverage analysis, use the
3737 additional @option{-ftest-coverage} option. You do not need to profile
3738 every source file in a program.
3741 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3742 (the latter implies the former).
3745 Run the program on a representative workload to generate the arc profile
3746 information. This may be repeated any number of times. You can run
3747 concurrent instances of your program, and provided that the file system
3748 supports locking, the data files will be correctly updated. Also
3749 @code{fork} calls are detected and correctly handled (double counting
3753 For profile-directed optimizations, compile the source files again with
3754 the same optimization and code generation options plus
3755 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3756 Control Optimization}).
3759 For test coverage analysis, use @command{gcov} to produce human readable
3760 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3761 @command{gcov} documentation for further information.
3765 With @option{-fprofile-arcs}, for each function of your program GCC
3766 creates a program flow graph, then finds a spanning tree for the graph.
3767 Only arcs that are not on the spanning tree have to be instrumented: the
3768 compiler adds code to count the number of times that these arcs are
3769 executed. When an arc is the only exit or only entrance to a block, the
3770 instrumentation code can be added to the block; otherwise, a new basic
3771 block must be created to hold the instrumentation code.
3774 @item -ftest-coverage
3775 @opindex ftest-coverage
3776 Produce a notes file that the @command{gcov} code-coverage utility
3777 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3778 show program coverage. Each source file's note file is called
3779 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3780 above for a description of @var{auxname} and instructions on how to
3781 generate test coverage data. Coverage data will match the source files
3782 more closely, if you do not optimize.
3784 @item -d@var{letters}
3785 @item -fdump-rtl-@var{pass}
3787 Says to make debugging dumps during compilation at times specified by
3788 @var{letters}. This is used for debugging the RTL-based passes of the
3789 compiler. The file names for most of the dumps are made by appending a
3790 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3791 from the name of the output file, if explicitly specified and it is not
3792 an executable, otherwise it is the basename of the source file.
3794 Most debug dumps can be enabled either passing a letter to the @option{-d}
3795 option, or with a long @option{-fdump-rtl} switch; here are the possible
3796 letters for use in @var{letters} and @var{pass}, and their meanings:
3801 Annotate the assembler output with miscellaneous debugging information.
3804 @itemx -fdump-rtl-bbro
3806 @opindex fdump-rtl-bbro
3807 Dump after block reordering, to @file{@var{file}.148r.bbro}.
3810 @itemx -fdump-rtl-combine
3812 @opindex fdump-rtl-combine
3813 Dump after instruction combination, to the file @file{@var{file}.129r.combine}.
3816 @itemx -fdump-rtl-ce1
3817 @itemx -fdump-rtl-ce2
3819 @opindex fdump-rtl-ce1
3820 @opindex fdump-rtl-ce2
3821 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3822 first if conversion, to the file @file{@var{file}.117r.ce1}. @option{-dC}
3823 and @option{-fdump-rtl-ce2} enable dumping after the second if
3824 conversion, to the file @file{@var{file}.130r.ce2}.
3827 @itemx -fdump-rtl-btl
3828 @itemx -fdump-rtl-dbr
3830 @opindex fdump-rtl-btl
3831 @opindex fdump-rtl-dbr
3832 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3833 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3834 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3835 scheduling, to @file{@var{file}.36.dbr}.
3839 Dump all macro definitions, at the end of preprocessing, in addition to
3843 @itemx -fdump-rtl-ce3
3845 @opindex fdump-rtl-ce3
3846 Dump after the third if conversion, to @file{@var{file}.146r.ce3}.
3849 @itemx -fdump-rtl-cfg
3850 @itemx -fdump-rtl-life
3852 @opindex fdump-rtl-cfg
3853 @opindex fdump-rtl-life
3854 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3855 and data flow analysis, to @file{@var{file}.116r.cfg}. @option{-df}
3856 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3857 to @file{@var{file}.128r.life1} and @file{@var{file}.135r.life2}.
3860 @itemx -fdump-rtl-greg
3862 @opindex fdump-rtl-greg
3863 Dump after global register allocation, to @file{@var{file}.139r.greg}.
3866 @itemx -fdump-rtl-gcse
3867 @itemx -fdump-rtl-bypass
3869 @opindex fdump-rtl-gcse
3870 @opindex fdump-rtl-bypass
3871 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3872 @file{@var{file}.114r.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3873 enable dumping after jump bypassing and control flow optimizations, to
3874 @file{@var{file}.115r.bypass}.
3877 @itemx -fdump-rtl-eh
3879 @opindex fdump-rtl-eh
3880 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3883 @itemx -fdump-rtl-sibling
3885 @opindex fdump-rtl-sibling
3886 Dump after sibling call optimizations, to @file{@var{file}.106r.sibling}.
3889 @itemx -fdump-rtl-jump
3891 @opindex fdump-rtl-jump
3892 Dump after the first jump optimization, to @file{@var{file}.112r.jump}.
3895 @itemx -fdump-rtl-stack
3897 @opindex fdump-rtl-stack
3898 Dump after conversion from registers to stack, to @file{@var{file}.152r.stack}.
3901 @itemx -fdump-rtl-lreg
3903 @opindex fdump-rtl-lreg
3904 Dump after local register allocation, to @file{@var{file}.138r.lreg}.
3907 @itemx -fdump-rtl-loop2
3909 @opindex fdump-rtl-loop2
3910 @option{-dL} and @option{-fdump-rtl-loop2} enable dumping after the
3911 loop optimization pass, to @file{@var{file}.119r.loop2},
3912 @file{@var{file}.120r.loop2_init},
3913 @file{@var{file}.121r.loop2_invariant}, and
3914 @file{@var{file}.125r.loop2_done}.
3917 @itemx -fdump-rtl-sms
3919 @opindex fdump-rtl-sms
3920 Dump after modulo scheduling, to @file{@var{file}.136r.sms}.
3923 @itemx -fdump-rtl-mach
3925 @opindex fdump-rtl-mach
3926 Dump after performing the machine dependent reorganization pass, to
3927 @file{@var{file}.155r.mach}.
3930 @itemx -fdump-rtl-rnreg
3932 @opindex fdump-rtl-rnreg
3933 Dump after register renumbering, to @file{@var{file}.147r.rnreg}.
3936 @itemx -fdump-rtl-regmove
3938 @opindex fdump-rtl-regmove
3939 Dump after the register move pass, to @file{@var{file}.132r.regmove}.
3942 @itemx -fdump-rtl-postreload
3944 @opindex fdump-rtl-postreload
3945 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3948 @itemx -fdump-rtl-expand
3950 @opindex fdump-rtl-expand
3951 Dump after RTL generation, to @file{@var{file}.104r.expand}.
3954 @itemx -fdump-rtl-sched2
3956 @opindex fdump-rtl-sched2
3957 Dump after the second scheduling pass, to @file{@var{file}.149r.sched2}.
3960 @itemx -fdump-rtl-cse
3962 @opindex fdump-rtl-cse
3963 Dump after CSE (including the jump optimization that sometimes follows
3964 CSE), to @file{@var{file}.113r.cse}.
3967 @itemx -fdump-rtl-sched1
3969 @opindex fdump-rtl-sched1
3970 Dump after the first scheduling pass, to @file{@var{file}.136r.sched1}.
3973 @itemx -fdump-rtl-cse2
3975 @opindex fdump-rtl-cse2
3976 Dump after the second CSE pass (including the jump optimization that
3977 sometimes follows CSE), to @file{@var{file}.127r.cse2}.
3980 @itemx -fdump-rtl-tracer
3982 @opindex fdump-rtl-tracer
3983 Dump after running tracer, to @file{@var{file}.118r.tracer}.
3986 @itemx -fdump-rtl-vpt
3987 @itemx -fdump-rtl-vartrack
3989 @opindex fdump-rtl-vpt
3990 @opindex fdump-rtl-vartrack
3991 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3992 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3993 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3994 to @file{@var{file}.154r.vartrack}.
3997 @itemx -fdump-rtl-flow2
3999 @opindex fdump-rtl-flow2
4000 Dump after the second flow pass, to @file{@var{file}.142r.flow2}.
4003 @itemx -fdump-rtl-peephole2
4005 @opindex fdump-rtl-peephole2
4006 Dump after the peephole pass, to @file{@var{file}.145r.peephole2}.
4009 @itemx -fdump-rtl-web
4011 @opindex fdump-rtl-web
4012 Dump after live range splitting, to @file{@var{file}.126r.web}.
4015 @itemx -fdump-rtl-all
4017 @opindex fdump-rtl-all
4018 Produce all the dumps listed above.
4022 Produce a core dump whenever an error occurs.
4026 Print statistics on memory usage, at the end of the run, to
4031 Annotate the assembler output with a comment indicating which
4032 pattern and alternative was used. The length of each instruction is
4037 Dump the RTL in the assembler output as a comment before each instruction.
4038 Also turns on @option{-dp} annotation.
4042 For each of the other indicated dump files (either with @option{-d} or
4043 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
4044 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
4048 Just generate RTL for a function instead of compiling it. Usually used
4049 with @samp{r} (@option{-fdump-rtl-expand}).
4053 Dump debugging information during parsing, to standard error.
4057 @opindex fdump-noaddr
4058 When doing debugging dumps (see @option{-d} option above), suppress
4059 address output. This makes it more feasible to use diff on debugging
4060 dumps for compiler invocations with different compiler binaries and/or
4061 different text / bss / data / heap / stack / dso start locations.
4063 @item -fdump-unnumbered
4064 @opindex fdump-unnumbered
4065 When doing debugging dumps (see @option{-d} option above), suppress instruction
4066 numbers, line number note and address output. This makes it more feasible to
4067 use diff on debugging dumps for compiler invocations with different
4068 options, in particular with and without @option{-g}.
4070 @item -fdump-translation-unit @r{(C++ only)}
4071 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
4072 @opindex fdump-translation-unit
4073 Dump a representation of the tree structure for the entire translation
4074 unit to a file. The file name is made by appending @file{.tu} to the
4075 source file name. If the @samp{-@var{options}} form is used, @var{options}
4076 controls the details of the dump as described for the
4077 @option{-fdump-tree} options.
4079 @item -fdump-class-hierarchy @r{(C++ only)}
4080 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
4081 @opindex fdump-class-hierarchy
4082 Dump a representation of each class's hierarchy and virtual function
4083 table layout to a file. The file name is made by appending @file{.class}
4084 to the source file name. If the @samp{-@var{options}} form is used,
4085 @var{options} controls the details of the dump as described for the
4086 @option{-fdump-tree} options.
4088 @item -fdump-ipa-@var{switch}
4090 Control the dumping at various stages of inter-procedural analysis
4091 language tree to a file. The file name is generated by appending a switch
4092 specific suffix to the source file name. The following dumps are possible:
4096 Enables all inter-procedural analysis dumps; currently the only produced
4097 dump is the @samp{cgraph} dump.
4100 Dumps information about call-graph optimization, unused function removal,
4101 and inlining decisions.
4104 @item -fdump-tree-@var{switch}
4105 @itemx -fdump-tree-@var{switch}-@var{options}
4107 Control the dumping at various stages of processing the intermediate
4108 language tree to a file. The file name is generated by appending a switch
4109 specific suffix to the source file name. If the @samp{-@var{options}}
4110 form is used, @var{options} is a list of @samp{-} separated options that
4111 control the details of the dump. Not all options are applicable to all
4112 dumps, those which are not meaningful will be ignored. The following
4113 options are available
4117 Print the address of each node. Usually this is not meaningful as it
4118 changes according to the environment and source file. Its primary use
4119 is for tying up a dump file with a debug environment.
4121 Inhibit dumping of members of a scope or body of a function merely
4122 because that scope has been reached. Only dump such items when they
4123 are directly reachable by some other path. When dumping pretty-printed
4124 trees, this option inhibits dumping the bodies of control structures.
4126 Print a raw representation of the tree. By default, trees are
4127 pretty-printed into a C-like representation.
4129 Enable more detailed dumps (not honored by every dump option).
4131 Enable dumping various statistics about the pass (not honored by every dump
4134 Enable showing basic block boundaries (disabled in raw dumps).
4136 Enable showing virtual operands for every statement.
4138 Enable showing line numbers for statements.
4140 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4142 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4145 The following tree dumps are possible:
4149 Dump before any tree based optimization, to @file{@var{file}.original}.
4152 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4155 Dump after function inlining, to @file{@var{file}.inlined}.
4158 @opindex fdump-tree-gimple
4159 Dump each function before and after the gimplification pass to a file. The
4160 file name is made by appending @file{.gimple} to the source file name.
4163 @opindex fdump-tree-cfg
4164 Dump the control flow graph of each function to a file. The file name is
4165 made by appending @file{.cfg} to the source file name.
4168 @opindex fdump-tree-vcg
4169 Dump the control flow graph of each function to a file in VCG format. The
4170 file name is made by appending @file{.vcg} to the source file name. Note
4171 that if the file contains more than one function, the generated file cannot
4172 be used directly by VCG@. You will need to cut and paste each function's
4173 graph into its own separate file first.
4176 @opindex fdump-tree-ch
4177 Dump each function after copying loop headers. The file name is made by
4178 appending @file{.ch} to the source file name.
4181 @opindex fdump-tree-ssa
4182 Dump SSA related information to a file. The file name is made by appending
4183 @file{.ssa} to the source file name.
4186 @opindex fdump-tree-salias
4187 Dump structure aliasing variable information to a file. This file name
4188 is made by appending @file{.salias} to the source file name.
4191 @opindex fdump-tree-alias
4192 Dump aliasing information for each function. The file name is made by
4193 appending @file{.alias} to the source file name.
4196 @opindex fdump-tree-ccp
4197 Dump each function after CCP@. The file name is made by appending
4198 @file{.ccp} to the source file name.
4201 @opindex fdump-tree-storeccp
4202 Dump each function after STORE-CCP. The file name is made by appending
4203 @file{.storeccp} to the source file name.
4206 @opindex fdump-tree-pre
4207 Dump trees after partial redundancy elimination. The file name is made
4208 by appending @file{.pre} to the source file name.
4211 @opindex fdump-tree-fre
4212 Dump trees after full redundancy elimination. The file name is made
4213 by appending @file{.fre} to the source file name.
4216 @opindex fdump-tree-copyprop
4217 Dump trees after copy propagation. The file name is made
4218 by appending @file{.copyprop} to the source file name.
4220 @item store_copyprop
4221 @opindex fdump-tree-store_copyprop
4222 Dump trees after store copy-propagation. The file name is made
4223 by appending @file{.store_copyprop} to the source file name.
4226 @opindex fdump-tree-dce
4227 Dump each function after dead code elimination. The file name is made by
4228 appending @file{.dce} to the source file name.
4231 @opindex fdump-tree-mudflap
4232 Dump each function after adding mudflap instrumentation. The file name is
4233 made by appending @file{.mudflap} to the source file name.
4236 @opindex fdump-tree-sra
4237 Dump each function after performing scalar replacement of aggregates. The
4238 file name is made by appending @file{.sra} to the source file name.
4241 @opindex fdump-tree-sink
4242 Dump each function after performing code sinking. The file name is made
4243 by appending @file{.sink} to the source file name.
4246 @opindex fdump-tree-dom
4247 Dump each function after applying dominator tree optimizations. The file
4248 name is made by appending @file{.dom} to the source file name.
4251 @opindex fdump-tree-dse
4252 Dump each function after applying dead store elimination. The file
4253 name is made by appending @file{.dse} to the source file name.
4256 @opindex fdump-tree-phiopt
4257 Dump each function after optimizing PHI nodes into straightline code. The file
4258 name is made by appending @file{.phiopt} to the source file name.
4261 @opindex fdump-tree-forwprop
4262 Dump each function after forward propagating single use variables. The file
4263 name is made by appending @file{.forwprop} to the source file name.
4266 @opindex fdump-tree-copyrename
4267 Dump each function after applying the copy rename optimization. The file
4268 name is made by appending @file{.copyrename} to the source file name.
4271 @opindex fdump-tree-nrv
4272 Dump each function after applying the named return value optimization on
4273 generic trees. The file name is made by appending @file{.nrv} to the source
4277 @opindex fdump-tree-vect
4278 Dump each function after applying vectorization of loops. The file name is
4279 made by appending @file{.vect} to the source file name.
4282 @opindex fdump-tree-vrp
4283 Dump each function after Value Range Propagation (VRP). The file name
4284 is made by appending @file{.vrp} to the source file name.
4287 @opindex fdump-tree-all
4288 Enable all the available tree dumps with the flags provided in this option.
4291 @item -ftree-vectorizer-verbose=@var{n}
4292 @opindex ftree-vectorizer-verbose
4293 This option controls the amount of debugging output the vectorizer prints.
4294 This information is written to standard error, unless
4295 @option{-fdump-tree-all} or @option{-fdump-tree-vect} is specified,
4296 in which case it is output to the usual dump listing file, @file{.vect}.
4297 For @var{n}=0 no diagnostic information is reported.
4298 If @var{n}=1 the vectorizer reports each loop that got vectorized,
4299 and the total number of loops that got vectorized.
4300 If @var{n}=2 the vectorizer also reports non-vectorized loops that passed
4301 the first analysis phase (vect_analyze_loop_form) - i.e. countable,
4302 inner-most, single-bb, single-entry/exit loops. This is the same verbosity
4303 level that @option{-fdump-tree-vect-stats} uses.
4304 Higher verbosity levels mean either more information dumped for each
4305 reported loop, or same amount of information reported for more loops:
4306 If @var{n}=3, alignment related information is added to the reports.
4307 If @var{n}=4, data-references related information (e.g. memory dependences,
4308 memory access-patterns) is added to the reports.
4309 If @var{n}=5, the vectorizer reports also non-vectorized inner-most loops
4310 that did not pass the first analysis phase (i.e. may not be countable, or
4311 may have complicated control-flow).
4312 If @var{n}=6, the vectorizer reports also non-vectorized nested loops.
4313 For @var{n}=7, all the information the vectorizer generates during its
4314 analysis and transformation is reported. This is the same verbosity level
4315 that @option{-fdump-tree-vect-details} uses.
4317 @item -frandom-seed=@var{string}
4318 @opindex frandom-string
4319 This option provides a seed that GCC uses when it would otherwise use
4320 random numbers. It is used to generate certain symbol names
4321 that have to be different in every compiled file. It is also used to
4322 place unique stamps in coverage data files and the object files that
4323 produce them. You can use the @option{-frandom-seed} option to produce
4324 reproducibly identical object files.
4326 The @var{string} should be different for every file you compile.
4328 @item -fsched-verbose=@var{n}
4329 @opindex fsched-verbose
4330 On targets that use instruction scheduling, this option controls the
4331 amount of debugging output the scheduler prints. This information is
4332 written to standard error, unless @option{-dS} or @option{-dR} is
4333 specified, in which case it is output to the usual dump
4334 listing file, @file{.sched} or @file{.sched2} respectively. However
4335 for @var{n} greater than nine, the output is always printed to standard
4338 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4339 same information as @option{-dRS}. For @var{n} greater than one, it
4340 also output basic block probabilities, detailed ready list information
4341 and unit/insn info. For @var{n} greater than two, it includes RTL
4342 at abort point, control-flow and regions info. And for @var{n} over
4343 four, @option{-fsched-verbose} also includes dependence info.
4347 Store the usual ``temporary'' intermediate files permanently; place them
4348 in the current directory and name them based on the source file. Thus,
4349 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4350 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4351 preprocessed @file{foo.i} output file even though the compiler now
4352 normally uses an integrated preprocessor.
4354 When used in combination with the @option{-x} command line option,
4355 @option{-save-temps} is sensible enough to avoid over writing an
4356 input source file with the same extension as an intermediate file.
4357 The corresponding intermediate file may be obtained by renaming the
4358 source file before using @option{-save-temps}.
4362 Report the CPU time taken by each subprocess in the compilation
4363 sequence. For C source files, this is the compiler proper and assembler
4364 (plus the linker if linking is done). The output looks like this:
4371 The first number on each line is the ``user time'', that is time spent
4372 executing the program itself. The second number is ``system time'',
4373 time spent executing operating system routines on behalf of the program.
4374 Both numbers are in seconds.
4376 @item -fvar-tracking
4377 @opindex fvar-tracking
4378 Run variable tracking pass. It computes where variables are stored at each
4379 position in code. Better debugging information is then generated
4380 (if the debugging information format supports this information).
4382 It is enabled by default when compiling with optimization (@option{-Os},
4383 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4384 the debug info format supports it.
4386 @item -print-file-name=@var{library}
4387 @opindex print-file-name
4388 Print the full absolute name of the library file @var{library} that
4389 would be used when linking---and don't do anything else. With this
4390 option, GCC does not compile or link anything; it just prints the
4393 @item -print-multi-directory
4394 @opindex print-multi-directory
4395 Print the directory name corresponding to the multilib selected by any
4396 other switches present in the command line. This directory is supposed
4397 to exist in @env{GCC_EXEC_PREFIX}.
4399 @item -print-multi-lib
4400 @opindex print-multi-lib
4401 Print the mapping from multilib directory names to compiler switches
4402 that enable them. The directory name is separated from the switches by
4403 @samp{;}, and each switch starts with an @samp{@@} instead of the
4404 @samp{-}, without spaces between multiple switches. This is supposed to
4405 ease shell-processing.
4407 @item -print-prog-name=@var{program}
4408 @opindex print-prog-name
4409 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4411 @item -print-libgcc-file-name
4412 @opindex print-libgcc-file-name
4413 Same as @option{-print-file-name=libgcc.a}.
4415 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4416 but you do want to link with @file{libgcc.a}. You can do
4419 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4422 @item -print-search-dirs
4423 @opindex print-search-dirs
4424 Print the name of the configured installation directory and a list of
4425 program and library directories @command{gcc} will search---and don't do anything else.
4427 This is useful when @command{gcc} prints the error message
4428 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4429 To resolve this you either need to put @file{cpp0} and the other compiler
4430 components where @command{gcc} expects to find them, or you can set the environment
4431 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4432 Don't forget the trailing @samp{/}.
4433 @xref{Environment Variables}.
4436 @opindex dumpmachine
4437 Print the compiler's target machine (for example,
4438 @samp{i686-pc-linux-gnu})---and don't do anything else.
4441 @opindex dumpversion
4442 Print the compiler version (for example, @samp{3.0})---and don't do
4447 Print the compiler's built-in specs---and don't do anything else. (This
4448 is used when GCC itself is being built.) @xref{Spec Files}.
4450 @item -feliminate-unused-debug-types
4451 @opindex feliminate-unused-debug-types
4452 Normally, when producing DWARF2 output, GCC will emit debugging
4453 information for all types declared in a compilation
4454 unit, regardless of whether or not they are actually used
4455 in that compilation unit. Sometimes this is useful, such as
4456 if, in the debugger, you want to cast a value to a type that is
4457 not actually used in your program (but is declared). More often,
4458 however, this results in a significant amount of wasted space.
4459 With this option, GCC will avoid producing debug symbol output
4460 for types that are nowhere used in the source file being compiled.
4463 @node Optimize Options
4464 @section Options That Control Optimization
4465 @cindex optimize options
4466 @cindex options, optimization
4468 These options control various sorts of optimizations.
4470 Without any optimization option, the compiler's goal is to reduce the
4471 cost of compilation and to make debugging produce the expected
4472 results. Statements are independent: if you stop the program with a
4473 breakpoint between statements, you can then assign a new value to any
4474 variable or change the program counter to any other statement in the
4475 function and get exactly the results you would expect from the source
4478 Turning on optimization flags makes the compiler attempt to improve
4479 the performance and/or code size at the expense of compilation time
4480 and possibly the ability to debug the program.
4482 The compiler performs optimization based on the knowledge it has of
4483 the program. Optimization levels @option{-O} and above, in
4484 particular, enable @emph{unit-at-a-time} mode, which allows the
4485 compiler to consider information gained from later functions in
4486 the file when compiling a function. Compiling multiple files at
4487 once to a single output file in @emph{unit-at-a-time} mode allows
4488 the compiler to use information gained from all of the files when
4489 compiling each of them.
4491 Not all optimizations are controlled directly by a flag. Only
4492 optimizations that have a flag are listed.
4499 Optimize. Optimizing compilation takes somewhat more time, and a lot
4500 more memory for a large function.
4502 With @option{-O}, the compiler tries to reduce code size and execution
4503 time, without performing any optimizations that take a great deal of
4506 @option{-O} turns on the following optimization flags:
4507 @gccoptlist{-fdefer-pop @gol
4508 -fdelayed-branch @gol
4509 -fguess-branch-probability @gol
4510 -fcprop-registers @gol
4511 -fif-conversion @gol
4512 -fif-conversion2 @gol
4515 -ftree-dominator-opts @gol
4520 -ftree-copyrename @gol
4523 -funit-at-a-time @gol
4526 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4527 where doing so does not interfere with debugging.
4531 Optimize even more. GCC performs nearly all supported optimizations
4532 that do not involve a space-speed tradeoff. The compiler does not
4533 perform loop unrolling or function inlining when you specify @option{-O2}.
4534 As compared to @option{-O}, this option increases both compilation time
4535 and the performance of the generated code.
4537 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4538 also turns on the following optimization flags:
4539 @gccoptlist{-fthread-jumps @gol
4541 -foptimize-sibling-calls @gol
4542 -fcse-follow-jumps -fcse-skip-blocks @gol
4543 -fgcse -fgcse-lm @gol
4544 -fexpensive-optimizations @gol
4545 -frerun-cse-after-loop @gol
4548 -fschedule-insns -fschedule-insns2 @gol
4549 -fsched-interblock -fsched-spec @gol
4551 -fstrict-aliasing @gol
4552 -fdelete-null-pointer-checks @gol
4553 -freorder-blocks -freorder-functions @gol
4554 -falign-functions -falign-jumps @gol
4555 -falign-loops -falign-labels @gol
4559 Please note the warning under @option{-fgcse} about
4560 invoking @option{-O2} on programs that use computed gotos.
4564 Optimize yet more. @option{-O3} turns on all optimizations specified by
4565 @option{-O2} and also turns on the @option{-finline-functions},
4566 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4570 Do not optimize. This is the default.
4574 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4575 do not typically increase code size. It also performs further
4576 optimizations designed to reduce code size.
4578 @option{-Os} disables the following optimization flags:
4579 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4580 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4581 -fprefetch-loop-arrays -ftree-vect-loop-version}
4583 If you use multiple @option{-O} options, with or without level numbers,
4584 the last such option is the one that is effective.
4587 Options of the form @option{-f@var{flag}} specify machine-independent
4588 flags. Most flags have both positive and negative forms; the negative
4589 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4590 below, only one of the forms is listed---the one you typically will
4591 use. You can figure out the other form by either removing @samp{no-}
4594 The following options control specific optimizations. They are either
4595 activated by @option{-O} options or are related to ones that are. You
4596 can use the following flags in the rare cases when ``fine-tuning'' of
4597 optimizations to be performed is desired.
4600 @item -fno-default-inline
4601 @opindex fno-default-inline
4602 Do not make member functions inline by default merely because they are
4603 defined inside the class scope (C++ only). Otherwise, when you specify
4604 @w{@option{-O}}, member functions defined inside class scope are compiled
4605 inline by default; i.e., you don't need to add @samp{inline} in front of
4606 the member function name.
4608 @item -fno-defer-pop
4609 @opindex fno-defer-pop
4610 Always pop the arguments to each function call as soon as that function
4611 returns. For machines which must pop arguments after a function call,
4612 the compiler normally lets arguments accumulate on the stack for several
4613 function calls and pops them all at once.
4615 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4619 Force memory operands to be copied into registers before doing
4620 arithmetic on them. This produces better code by making all memory
4621 references potential common subexpressions. When they are not common
4622 subexpressions, instruction combination should eliminate the separate
4623 register-load. This option is now a nop and will be removed in 4.2.
4626 @opindex fforce-addr
4627 Force memory address constants to be copied into registers before
4628 doing arithmetic on them.
4630 @item -fforward-propagate
4631 @opindex fforward-propagate
4632 Perform a forward propagation pass on RTL. The pass tries to combine two
4633 instructions and checks if the result can be simplified. If loop unrolling
4634 is active, two passes are performed and the second is scheduled after
4637 This option is enabled by default at optimization levels @option{-O2},
4638 @option{-O3}, @option{-Os}.
4640 @item -fomit-frame-pointer
4641 @opindex fomit-frame-pointer
4642 Don't keep the frame pointer in a register for functions that
4643 don't need one. This avoids the instructions to save, set up and
4644 restore frame pointers; it also makes an extra register available
4645 in many functions. @strong{It also makes debugging impossible on
4648 On some machines, such as the VAX, this flag has no effect, because
4649 the standard calling sequence automatically handles the frame pointer
4650 and nothing is saved by pretending it doesn't exist. The
4651 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4652 whether a target machine supports this flag. @xref{Registers,,Register
4653 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4655 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4657 @item -foptimize-sibling-calls
4658 @opindex foptimize-sibling-calls
4659 Optimize sibling and tail recursive calls.
4661 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4665 Don't pay attention to the @code{inline} keyword. Normally this option
4666 is used to keep the compiler from expanding any functions inline.
4667 Note that if you are not optimizing, no functions can be expanded inline.
4669 @item -finline-functions
4670 @opindex finline-functions
4671 Integrate all simple functions into their callers. The compiler
4672 heuristically decides which functions are simple enough to be worth
4673 integrating in this way.
4675 If all calls to a given function are integrated, and the function is
4676 declared @code{static}, then the function is normally not output as
4677 assembler code in its own right.
4679 Enabled at level @option{-O3}.
4681 @item -finline-functions-called-once
4682 @opindex finline-functions-called-once
4683 Consider all @code{static} functions called once for inlining into their
4684 caller even if they are not marked @code{inline}. If a call to a given
4685 function is integrated, then the function is not output as assembler code
4688 Enabled if @option{-funit-at-a-time} is enabled.
4690 @item -fearly-inlining
4691 @opindex fearly-inlining
4692 Inline functions marked by @code{always_inline} and functions whose body seems
4693 smaller than the function call overhead early before doing
4694 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4695 makes profiling significantly cheaper and usually inlining faster on programs
4696 having large chains of nested wrapper functions.
4700 @item -finline-limit=@var{n}
4701 @opindex finline-limit
4702 By default, GCC limits the size of functions that can be inlined. This flag
4703 allows the control of this limit for functions that are explicitly marked as
4704 inline (i.e., marked with the inline keyword or defined within the class
4705 definition in c++). @var{n} is the size of functions that can be inlined in
4706 number of pseudo instructions (not counting parameter handling). The default
4707 value of @var{n} is 600.
4708 Increasing this value can result in more inlined code at
4709 the cost of compilation time and memory consumption. Decreasing usually makes
4710 the compilation faster and less code will be inlined (which presumably
4711 means slower programs). This option is particularly useful for programs that
4712 use inlining heavily such as those based on recursive templates with C++.
4714 Inlining is actually controlled by a number of parameters, which may be
4715 specified individually by using @option{--param @var{name}=@var{value}}.
4716 The @option{-finline-limit=@var{n}} option sets some of these parameters
4720 @item max-inline-insns-single
4721 is set to @var{n}/2.
4722 @item max-inline-insns-auto
4723 is set to @var{n}/2.
4724 @item min-inline-insns
4725 is set to 130 or @var{n}/4, whichever is smaller.
4726 @item max-inline-insns-rtl
4730 See below for a documentation of the individual
4731 parameters controlling inlining.
4733 @emph{Note:} pseudo instruction represents, in this particular context, an
4734 abstract measurement of function's size. In no way does it represent a count
4735 of assembly instructions and as such its exact meaning might change from one
4736 release to an another.
4738 @item -fkeep-inline-functions
4739 @opindex fkeep-inline-functions
4740 In C, emit @code{static} functions that are declared @code{inline}
4741 into the object file, even if the function has been inlined into all
4742 of its callers. This switch does not affect functions using the
4743 @code{extern inline} extension in GNU C89@. In C++, emit any and all
4744 inline functions into the object file.
4746 @item -fkeep-static-consts
4747 @opindex fkeep-static-consts
4748 Emit variables declared @code{static const} when optimization isn't turned
4749 on, even if the variables aren't referenced.
4751 GCC enables this option by default. If you want to force the compiler to
4752 check if the variable was referenced, regardless of whether or not
4753 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4755 @item -fmerge-constants
4756 Attempt to merge identical constants (string constants and floating point
4757 constants) across compilation units.
4759 This option is the default for optimized compilation if the assembler and
4760 linker support it. Use @option{-fno-merge-constants} to inhibit this
4763 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4765 @item -fmerge-all-constants
4766 Attempt to merge identical constants and identical variables.
4768 This option implies @option{-fmerge-constants}. In addition to
4769 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4770 arrays or initialized constant variables with integral or floating point
4771 types. Languages like C or C++ require each non-automatic variable to
4772 have distinct location, so using this option will result in non-conforming
4775 @item -fmodulo-sched
4776 @opindex fmodulo-sched
4777 Perform swing modulo scheduling immediately before the first scheduling
4778 pass. This pass looks at innermost loops and reorders their
4779 instructions by overlapping different iterations.
4781 @item -fno-branch-count-reg
4782 @opindex fno-branch-count-reg
4783 Do not use ``decrement and branch'' instructions on a count register,
4784 but instead generate a sequence of instructions that decrement a
4785 register, compare it against zero, then branch based upon the result.
4786 This option is only meaningful on architectures that support such
4787 instructions, which include x86, PowerPC, IA-64 and S/390.
4789 The default is @option{-fbranch-count-reg}.
4791 @item -fno-function-cse
4792 @opindex fno-function-cse
4793 Do not put function addresses in registers; make each instruction that
4794 calls a constant function contain the function's address explicitly.
4796 This option results in less efficient code, but some strange hacks
4797 that alter the assembler output may be confused by the optimizations
4798 performed when this option is not used.
4800 The default is @option{-ffunction-cse}
4802 @item -fno-zero-initialized-in-bss
4803 @opindex fno-zero-initialized-in-bss
4804 If the target supports a BSS section, GCC by default puts variables that
4805 are initialized to zero into BSS@. This can save space in the resulting
4808 This option turns off this behavior because some programs explicitly
4809 rely on variables going to the data section. E.g., so that the
4810 resulting executable can find the beginning of that section and/or make
4811 assumptions based on that.
4813 The default is @option{-fzero-initialized-in-bss}.
4815 @item -fbounds-check
4816 @opindex fbounds-check
4817 For front-ends that support it, generate additional code to check that
4818 indices used to access arrays are within the declared range. This is
4819 currently only supported by the Java and Fortran front-ends, where
4820 this option defaults to true and false respectively.
4822 @item -fmudflap -fmudflapth -fmudflapir
4826 @cindex bounds checking
4828 For front-ends that support it (C and C++), instrument all risky
4829 pointer/array dereferencing operations, some standard library
4830 string/heap functions, and some other associated constructs with
4831 range/validity tests. Modules so instrumented should be immune to
4832 buffer overflows, invalid heap use, and some other classes of C/C++
4833 programming errors. The instrumentation relies on a separate runtime
4834 library (@file{libmudflap}), which will be linked into a program if
4835 @option{-fmudflap} is given at link time. Run-time behavior of the
4836 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4837 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4840 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4841 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4842 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4843 instrumentation should ignore pointer reads. This produces less
4844 instrumentation (and therefore faster execution) and still provides
4845 some protection against outright memory corrupting writes, but allows
4846 erroneously read data to propagate within a program.
4848 @item -fthread-jumps
4849 @opindex fthread-jumps
4850 Perform optimizations where we check to see if a jump branches to a
4851 location where another comparison subsumed by the first is found. If
4852 so, the first branch is redirected to either the destination of the
4853 second branch or a point immediately following it, depending on whether
4854 the condition is known to be true or false.
4856 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4858 @item -fcse-follow-jumps
4859 @opindex fcse-follow-jumps
4860 In common subexpression elimination, scan through jump instructions
4861 when the target of the jump is not reached by any other path. For
4862 example, when CSE encounters an @code{if} statement with an
4863 @code{else} clause, CSE will follow the jump when the condition
4866 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4868 @item -fcse-skip-blocks
4869 @opindex fcse-skip-blocks
4870 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4871 follow jumps which conditionally skip over blocks. When CSE
4872 encounters a simple @code{if} statement with no else clause,
4873 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4874 body of the @code{if}.
4876 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4878 @item -frerun-cse-after-loop
4879 @opindex frerun-cse-after-loop
4880 Re-run common subexpression elimination after loop optimizations has been
4883 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4887 Perform a global common subexpression elimination pass.
4888 This pass also performs global constant and copy propagation.
4890 @emph{Note:} When compiling a program using computed gotos, a GCC
4891 extension, you may get better runtime performance if you disable
4892 the global common subexpression elimination pass by adding
4893 @option{-fno-gcse} to the command line.
4895 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4899 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4900 attempt to move loads which are only killed by stores into themselves. This
4901 allows a loop containing a load/store sequence to be changed to a load outside
4902 the loop, and a copy/store within the loop.
4904 Enabled by default when gcse is enabled.
4908 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4909 global common subexpression elimination. This pass will attempt to move
4910 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4911 loops containing a load/store sequence can be changed to a load before
4912 the loop and a store after the loop.
4914 Not enabled at any optimization level.
4918 When @option{-fgcse-las} is enabled, the global common subexpression
4919 elimination pass eliminates redundant loads that come after stores to the
4920 same memory location (both partial and full redundancies).
4922 Not enabled at any optimization level.
4924 @item -fgcse-after-reload
4925 @opindex fgcse-after-reload
4926 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4927 pass is performed after reload. The purpose of this pass is to cleanup
4930 @item -funsafe-loop-optimizations
4931 @opindex funsafe-loop-optimizations
4932 If given, the loop optimizer will assume that loop indices do not
4933 overflow, and that the loops with nontrivial exit condition are not
4934 infinite. This enables a wider range of loop optimizations even if
4935 the loop optimizer itself cannot prove that these assumptions are valid.
4936 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4937 if it finds this kind of loop.
4939 @item -fcrossjumping
4940 @opindex crossjumping
4941 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4942 resulting code may or may not perform better than without cross-jumping.
4944 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4946 @item -fif-conversion
4947 @opindex if-conversion
4948 Attempt to transform conditional jumps into branch-less equivalents. This
4949 include use of conditional moves, min, max, set flags and abs instructions, and
4950 some tricks doable by standard arithmetics. The use of conditional execution
4951 on chips where it is available is controlled by @code{if-conversion2}.
4953 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4955 @item -fif-conversion2
4956 @opindex if-conversion2
4957 Use conditional execution (where available) to transform conditional jumps into
4958 branch-less equivalents.
4960 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4962 @item -fdelete-null-pointer-checks
4963 @opindex fdelete-null-pointer-checks
4964 Use global dataflow analysis to identify and eliminate useless checks
4965 for null pointers. The compiler assumes that dereferencing a null
4966 pointer would have halted the program. If a pointer is checked after
4967 it has already been dereferenced, it cannot be null.
4969 In some environments, this assumption is not true, and programs can
4970 safely dereference null pointers. Use
4971 @option{-fno-delete-null-pointer-checks} to disable this optimization
4972 for programs which depend on that behavior.
4974 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4976 @item -fexpensive-optimizations
4977 @opindex fexpensive-optimizations
4978 Perform a number of minor optimizations that are relatively expensive.
4980 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4982 @item -foptimize-register-move
4984 @opindex foptimize-register-move
4986 Attempt to reassign register numbers in move instructions and as
4987 operands of other simple instructions in order to maximize the amount of
4988 register tying. This is especially helpful on machines with two-operand
4991 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4994 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4996 @item -fdelayed-branch
4997 @opindex fdelayed-branch
4998 If supported for the target machine, attempt to reorder instructions
4999 to exploit instruction slots available after delayed branch
5002 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5004 @item -fschedule-insns
5005 @opindex fschedule-insns
5006 If supported for the target machine, attempt to reorder instructions to
5007 eliminate execution stalls due to required data being unavailable. This
5008 helps machines that have slow floating point or memory load instructions
5009 by allowing other instructions to be issued until the result of the load
5010 or floating point instruction is required.
5012 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5014 @item -fschedule-insns2
5015 @opindex fschedule-insns2
5016 Similar to @option{-fschedule-insns}, but requests an additional pass of
5017 instruction scheduling after register allocation has been done. This is
5018 especially useful on machines with a relatively small number of
5019 registers and where memory load instructions take more than one cycle.
5021 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5023 @item -fno-sched-interblock
5024 @opindex fno-sched-interblock
5025 Don't schedule instructions across basic blocks. This is normally
5026 enabled by default when scheduling before register allocation, i.e.@:
5027 with @option{-fschedule-insns} or at @option{-O2} or higher.
5029 @item -fno-sched-spec
5030 @opindex fno-sched-spec
5031 Don't allow speculative motion of non-load instructions. This is normally
5032 enabled by default when scheduling before register allocation, i.e.@:
5033 with @option{-fschedule-insns} or at @option{-O2} or higher.
5035 @item -fsched-spec-load
5036 @opindex fsched-spec-load
5037 Allow speculative motion of some load instructions. This only makes
5038 sense when scheduling before register allocation, i.e.@: with
5039 @option{-fschedule-insns} or at @option{-O2} or higher.
5041 @item -fsched-spec-load-dangerous
5042 @opindex fsched-spec-load-dangerous
5043 Allow speculative motion of more load instructions. This only makes
5044 sense when scheduling before register allocation, i.e.@: with
5045 @option{-fschedule-insns} or at @option{-O2} or higher.
5047 @item -fsched-stalled-insns=@var{n}
5048 @opindex fsched-stalled-insns
5049 Define how many insns (if any) can be moved prematurely from the queue
5050 of stalled insns into the ready list, during the second scheduling pass.
5052 @item -fsched-stalled-insns-dep=@var{n}
5053 @opindex fsched-stalled-insns-dep
5054 Define how many insn groups (cycles) will be examined for a dependency
5055 on a stalled insn that is candidate for premature removal from the queue
5056 of stalled insns. Has an effect only during the second scheduling pass,
5057 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
5059 @item -fsched2-use-superblocks
5060 @opindex fsched2-use-superblocks
5061 When scheduling after register allocation, do use superblock scheduling
5062 algorithm. Superblock scheduling allows motion across basic block boundaries
5063 resulting on faster schedules. This option is experimental, as not all machine
5064 descriptions used by GCC model the CPU closely enough to avoid unreliable
5065 results from the algorithm.
5067 This only makes sense when scheduling after register allocation, i.e.@: with
5068 @option{-fschedule-insns2} or at @option{-O2} or higher.
5070 @item -fsched2-use-traces
5071 @opindex fsched2-use-traces
5072 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
5073 allocation and additionally perform code duplication in order to increase the
5074 size of superblocks using tracer pass. See @option{-ftracer} for details on
5077 This mode should produce faster but significantly longer programs. Also
5078 without @option{-fbranch-probabilities} the traces constructed may not
5079 match the reality and hurt the performance. This only makes
5080 sense when scheduling after register allocation, i.e.@: with
5081 @option{-fschedule-insns2} or at @option{-O2} or higher.
5085 Eliminates redundant extension instructions and move the non redundant
5086 ones to optimal placement using LCM.
5088 @item -freschedule-modulo-scheduled-loops
5089 @opindex fscheduling-in-modulo-scheduled-loops
5090 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
5091 we may want to prevent the later scheduling passes from changing its schedule, we use this
5092 option to control that.
5094 @item -fcaller-saves
5095 @opindex fcaller-saves
5096 Enable values to be allocated in registers that will be clobbered by
5097 function calls, by emitting extra instructions to save and restore the
5098 registers around such calls. Such allocation is done only when it
5099 seems to result in better code than would otherwise be produced.
5101 This option is always enabled by default on certain machines, usually
5102 those which have no call-preserved registers to use instead.
5104 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5107 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
5108 enabled by default at @option{-O2} and @option{-O3}.
5111 Perform Full Redundancy Elimination (FRE) on trees. The difference
5112 between FRE and PRE is that FRE only considers expressions
5113 that are computed on all paths leading to the redundant computation.
5114 This analysis faster than PRE, though it exposes fewer redundancies.
5115 This flag is enabled by default at @option{-O} and higher.
5117 @item -ftree-copy-prop
5118 Perform copy propagation on trees. This pass eliminates unnecessary
5119 copy operations. This flag is enabled by default at @option{-O} and
5122 @item -ftree-store-copy-prop
5123 Perform copy propagation of memory loads and stores. This pass
5124 eliminates unnecessary copy operations in memory references
5125 (structures, global variables, arrays, etc). This flag is enabled by
5126 default at @option{-O2} and higher.
5129 Perform structural alias analysis on trees. This flag
5130 is enabled by default at @option{-O} and higher.
5133 Perform interprocedural pointer analysis.
5136 Perform forward store motion on trees. This flag is
5137 enabled by default at @option{-O} and higher.
5140 Perform sparse conditional constant propagation (CCP) on trees. This
5141 pass only operates on local scalar variables and is enabled by default
5142 at @option{-O} and higher.
5144 @item -ftree-store-ccp
5145 Perform sparse conditional constant propagation (CCP) on trees. This
5146 pass operates on both local scalar variables and memory stores and
5147 loads (global variables, structures, arrays, etc). This flag is
5148 enabled by default at @option{-O2} and higher.
5151 Perform dead code elimination (DCE) on trees. This flag is enabled by
5152 default at @option{-O} and higher.
5154 @item -ftree-dominator-opts
5155 Perform a variety of simple scalar cleanups (constant/copy
5156 propagation, redundancy elimination, range propagation and expression
5157 simplification) based on a dominator tree traversal. This also
5158 performs jump threading (to reduce jumps to jumps). This flag is
5159 enabled by default at @option{-O} and higher.
5162 Perform loop header copying on trees. This is beneficial since it increases
5163 effectiveness of code motion optimizations. It also saves one jump. This flag
5164 is enabled by default at @option{-O} and higher. It is not enabled
5165 for @option{-Os}, since it usually increases code size.
5167 @item -ftree-loop-optimize
5168 Perform loop optimizations on trees. This flag is enabled by default
5169 at @option{-O} and higher.
5171 @item -ftree-loop-linear
5172 Perform linear loop transformations on tree. This flag can improve cache
5173 performance and allow further loop optimizations to take place.
5175 @item -ftree-loop-im
5176 Perform loop invariant motion on trees. This pass moves only invariants that
5177 would be hard to handle at RTL level (function calls, operations that expand to
5178 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5179 operands of conditions that are invariant out of the loop, so that we can use
5180 just trivial invariantness analysis in loop unswitching. The pass also includes
5183 @item -ftree-loop-ivcanon
5184 Create a canonical counter for number of iterations in the loop for that
5185 determining number of iterations requires complicated analysis. Later
5186 optimizations then may determine the number easily. Useful especially
5187 in connection with unrolling.
5190 Perform induction variable optimizations (strength reduction, induction
5191 variable merging and induction variable elimination) on trees.
5194 Perform scalar replacement of aggregates. This pass replaces structure
5195 references with scalars to prevent committing structures to memory too
5196 early. This flag is enabled by default at @option{-O} and higher.
5198 @item -ftree-copyrename
5199 Perform copy renaming on trees. This pass attempts to rename compiler
5200 temporaries to other variables at copy locations, usually resulting in
5201 variable names which more closely resemble the original variables. This flag
5202 is enabled by default at @option{-O} and higher.
5205 Perform temporary expression replacement during the SSA->normal phase. Single
5206 use/single def temporaries are replaced at their use location with their
5207 defining expression. This results in non-GIMPLE code, but gives the expanders
5208 much more complex trees to work on resulting in better RTL generation. This is
5209 enabled by default at @option{-O} and higher.
5212 Perform live range splitting during the SSA->normal phase. Distinct live
5213 ranges of a variable are split into unique variables, allowing for better
5214 optimization later. This is enabled by default at @option{-O} and higher.
5216 @item -ftree-vectorize
5217 Perform loop vectorization on trees.
5219 @item -ftree-vect-loop-version
5220 @opindex ftree-vect-loop-version
5221 Perform loop versioning when doing loop vectorization on trees. When a loop
5222 appears to be vectorizable except that data alignment or data dependence cannot
5223 be determined at compile time then vectorized and non-vectorized versions of
5224 the loop are generated along with runtime checks for alignment or dependence
5225 to control which version is executed. This option is enabled by default
5226 except at level @option{-Os} where it is disabled.
5229 Perform Value Range Propagation on trees. This is similar to the
5230 constant propagation pass, but instead of values, ranges of values are
5231 propagated. This allows the optimizers to remove unnecessary range
5232 checks like array bound checks and null pointer checks. This is
5233 enabled by default at @option{-O2} and higher. Null pointer check
5234 elimination is only done if @option{-fdelete-null-pointer-checks} is
5239 Perform tail duplication to enlarge superblock size. This transformation
5240 simplifies the control flow of the function allowing other optimizations to do
5243 @item -funroll-loops
5244 @opindex funroll-loops
5245 Unroll loops whose number of iterations can be determined at compile
5246 time or upon entry to the loop. @option{-funroll-loops} implies
5247 @option{-frerun-cse-after-loop}. This option makes code larger,
5248 and may or may not make it run faster.
5250 @item -funroll-all-loops
5251 @opindex funroll-all-loops
5252 Unroll all loops, even if their number of iterations is uncertain when
5253 the loop is entered. This usually makes programs run more slowly.
5254 @option{-funroll-all-loops} implies the same options as
5255 @option{-funroll-loops},
5257 @item -fsplit-ivs-in-unroller
5258 @opindex -fsplit-ivs-in-unroller
5259 Enables expressing of values of induction variables in later iterations
5260 of the unrolled loop using the value in the first iteration. This breaks
5261 long dependency chains, thus improving efficiency of the scheduling passes.
5263 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5264 same effect. However in cases the loop body is more complicated than
5265 a single basic block, this is not reliable. It also does not work at all
5266 on some of the architectures due to restrictions in the CSE pass.
5268 This optimization is enabled by default.
5270 @item -fvariable-expansion-in-unroller
5271 @opindex -fvariable-expansion-in-unroller
5272 With this option, the compiler will create multiple copies of some
5273 local variables when unrolling a loop which can result in superior code.
5275 @item -fprefetch-loop-arrays
5276 @opindex fprefetch-loop-arrays
5277 If supported by the target machine, generate instructions to prefetch
5278 memory to improve the performance of loops that access large arrays.
5280 This option may generate better or worse code; results are highly
5281 dependent on the structure of loops within the source code.
5283 Disabled at level @option{-Os}.
5286 @itemx -fno-peephole2
5287 @opindex fno-peephole
5288 @opindex fno-peephole2
5289 Disable any machine-specific peephole optimizations. The difference
5290 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5291 are implemented in the compiler; some targets use one, some use the
5292 other, a few use both.
5294 @option{-fpeephole} is enabled by default.
5295 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5297 @item -fno-guess-branch-probability
5298 @opindex fno-guess-branch-probability
5299 Do not guess branch probabilities using heuristics.
5301 GCC will use heuristics to guess branch probabilities if they are
5302 not provided by profiling feedback (@option{-fprofile-arcs}). These
5303 heuristics are based on the control flow graph. If some branch probabilities
5304 are specified by @samp{__builtin_expect}, then the heuristics will be
5305 used to guess branch probabilities for the rest of the control flow graph,
5306 taking the @samp{__builtin_expect} info into account. The interactions
5307 between the heuristics and @samp{__builtin_expect} can be complex, and in
5308 some cases, it may be useful to disable the heuristics so that the effects
5309 of @samp{__builtin_expect} are easier to understand.
5311 The default is @option{-fguess-branch-probability} at levels
5312 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5314 @item -freorder-blocks
5315 @opindex freorder-blocks
5316 Reorder basic blocks in the compiled function in order to reduce number of
5317 taken branches and improve code locality.
5319 Enabled at levels @option{-O2}, @option{-O3}.
5321 @item -freorder-blocks-and-partition
5322 @opindex freorder-blocks-and-partition
5323 In addition to reordering basic blocks in the compiled function, in order
5324 to reduce number of taken branches, partitions hot and cold basic blocks
5325 into separate sections of the assembly and .o files, to improve
5326 paging and cache locality performance.
5328 This optimization is automatically turned off in the presence of
5329 exception handling, for linkonce sections, for functions with a user-defined
5330 section attribute and on any architecture that does not support named
5333 @item -freorder-functions
5334 @opindex freorder-functions
5335 Reorder functions in the object file in order to
5336 improve code locality. This is implemented by using special
5337 subsections @code{.text.hot} for most frequently executed functions and
5338 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5339 the linker so object file format must support named sections and linker must
5340 place them in a reasonable way.
5342 Also profile feedback must be available in to make this option effective. See
5343 @option{-fprofile-arcs} for details.
5345 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5347 @item -fstrict-aliasing
5348 @opindex fstrict-aliasing
5349 Allows the compiler to assume the strictest aliasing rules applicable to
5350 the language being compiled. For C (and C++), this activates
5351 optimizations based on the type of expressions. In particular, an
5352 object of one type is assumed never to reside at the same address as an
5353 object of a different type, unless the types are almost the same. For
5354 example, an @code{unsigned int} can alias an @code{int}, but not a
5355 @code{void*} or a @code{double}. A character type may alias any other
5358 Pay special attention to code like this:
5371 The practice of reading from a different union member than the one most
5372 recently written to (called ``type-punning'') is common. Even with
5373 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5374 is accessed through the union type. So, the code above will work as
5375 expected. However, this code might not:
5386 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5388 @item -falign-functions
5389 @itemx -falign-functions=@var{n}
5390 @opindex falign-functions
5391 Align the start of functions to the next power-of-two greater than
5392 @var{n}, skipping up to @var{n} bytes. For instance,
5393 @option{-falign-functions=32} aligns functions to the next 32-byte
5394 boundary, but @option{-falign-functions=24} would align to the next
5395 32-byte boundary only if this can be done by skipping 23 bytes or less.
5397 @option{-fno-align-functions} and @option{-falign-functions=1} are
5398 equivalent and mean that functions will not be aligned.
5400 Some assemblers only support this flag when @var{n} is a power of two;
5401 in that case, it is rounded up.
5403 If @var{n} is not specified or is zero, use a machine-dependent default.
5405 Enabled at levels @option{-O2}, @option{-O3}.
5407 @item -falign-labels
5408 @itemx -falign-labels=@var{n}
5409 @opindex falign-labels
5410 Align all branch targets to a power-of-two boundary, skipping up to
5411 @var{n} bytes like @option{-falign-functions}. This option can easily
5412 make code slower, because it must insert dummy operations for when the
5413 branch target is reached in the usual flow of the code.
5415 @option{-fno-align-labels} and @option{-falign-labels=1} are
5416 equivalent and mean that labels will not be aligned.
5418 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5419 are greater than this value, then their values are used instead.
5421 If @var{n} is not specified or is zero, use a machine-dependent default
5422 which is very likely to be @samp{1}, meaning no alignment.
5424 Enabled at levels @option{-O2}, @option{-O3}.
5427 @itemx -falign-loops=@var{n}
5428 @opindex falign-loops
5429 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5430 like @option{-falign-functions}. The hope is that the loop will be
5431 executed many times, which will make up for any execution of the dummy
5434 @option{-fno-align-loops} and @option{-falign-loops=1} are
5435 equivalent and mean that loops will not be aligned.
5437 If @var{n} is not specified or is zero, use a machine-dependent default.
5439 Enabled at levels @option{-O2}, @option{-O3}.
5442 @itemx -falign-jumps=@var{n}
5443 @opindex falign-jumps
5444 Align branch targets to a power-of-two boundary, for branch targets
5445 where the targets can only be reached by jumping, skipping up to @var{n}
5446 bytes like @option{-falign-functions}. In this case, no dummy operations
5449 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5450 equivalent and mean that loops will not be aligned.
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 -funit-at-a-time
5457 @opindex funit-at-a-time
5458 Parse the whole compilation unit before starting to produce code.
5459 This allows some extra optimizations to take place but consumes
5460 more memory (in general). There are some compatibility issues
5461 with @emph{unit-at-a-time} mode:
5464 enabling @emph{unit-at-a-time} mode may change the order
5465 in which functions, variables, and top-level @code{asm} statements
5466 are emitted, and will likely break code relying on some particular
5467 ordering. The majority of such top-level @code{asm} statements,
5468 though, can be replaced by @code{section} attributes. The
5469 @option{fno-toplevel-reorder} option may be used to keep the ordering
5470 used in the input file, at the cost of some optimizations.
5473 @emph{unit-at-a-time} mode removes unreferenced static variables
5474 and functions. This may result in undefined references
5475 when an @code{asm} statement refers directly to variables or functions
5476 that are otherwise unused. In that case either the variable/function
5477 shall be listed as an operand of the @code{asm} statement operand or,
5478 in the case of top-level @code{asm} statements the attribute @code{used}
5479 shall be used on the declaration.
5482 Static functions now can use non-standard passing conventions that
5483 may break @code{asm} statements calling functions directly. Again,
5484 attribute @code{used} will prevent this behavior.
5487 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5488 but this scheme may not be supported by future releases of GCC@.
5490 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5492 @item -fno-toplevel-reorder
5493 Do not reorder top-level functions, variables, and @code{asm}
5494 statements. Output them in the same order that they appear in the
5495 input file. When this option is used, unreferenced static variables
5496 will not be removed. This option is intended to support existing code
5497 which relies on a particular ordering. For new code, it is better to
5502 Constructs webs as commonly used for register allocation purposes and assign
5503 each web individual pseudo register. This allows the register allocation pass
5504 to operate on pseudos directly, but also strengthens several other optimization
5505 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5506 however, make debugging impossible, since variables will no longer stay in a
5509 Enabled by default with @option{-funroll-loops}.
5511 @item -fwhole-program
5512 @opindex fwhole-program
5513 Assume that the current compilation unit represents whole program being
5514 compiled. All public functions and variables with the exception of @code{main}
5515 and those merged by attribute @code{externally_visible} become static functions
5516 and in a affect gets more aggressively optimized by interprocedural optimizers.
5517 While this option is equivalent to proper use of @code{static} keyword for
5518 programs consisting of single file, in combination with option
5519 @option{--combine} this flag can be used to compile most of smaller scale C
5520 programs since the functions and variables become local for the whole combined
5521 compilation unit, not for the single source file itself.
5524 @item -fno-cprop-registers
5525 @opindex fno-cprop-registers
5526 After register allocation and post-register allocation instruction splitting,
5527 we perform a copy-propagation pass to try to reduce scheduling dependencies
5528 and occasionally eliminate the copy.
5530 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5532 @item -fprofile-generate
5533 @opindex fprofile-generate
5535 Enable options usually used for instrumenting application to produce
5536 profile useful for later recompilation with profile feedback based
5537 optimization. You must use @option{-fprofile-generate} both when
5538 compiling and when linking your program.
5540 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5543 @opindex fprofile-use
5544 Enable profile feedback directed optimizations, and optimizations
5545 generally profitable only with profile feedback available.
5547 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5548 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}
5552 The following options control compiler behavior regarding floating
5553 point arithmetic. These options trade off between speed and
5554 correctness. All must be specifically enabled.
5558 @opindex ffloat-store
5559 Do not store floating point variables in registers, and inhibit other
5560 options that might change whether a floating point value is taken from a
5563 @cindex floating point precision
5564 This option prevents undesirable excess precision on machines such as
5565 the 68000 where the floating registers (of the 68881) keep more
5566 precision than a @code{double} is supposed to have. Similarly for the
5567 x86 architecture. For most programs, the excess precision does only
5568 good, but a few programs rely on the precise definition of IEEE floating
5569 point. Use @option{-ffloat-store} for such programs, after modifying
5570 them to store all pertinent intermediate computations into variables.
5574 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5575 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5576 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5577 and @option{fcx-limited-range}.
5579 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5581 This option should never be turned on by any @option{-O} option since
5582 it can result in incorrect output for programs which depend on
5583 an exact implementation of IEEE or ISO rules/specifications for
5586 @item -fno-math-errno
5587 @opindex fno-math-errno
5588 Do not set ERRNO after calling math functions that are executed
5589 with a single instruction, e.g., sqrt. A program that relies on
5590 IEEE exceptions for math error handling may want to use this flag
5591 for speed while maintaining IEEE arithmetic compatibility.
5593 This option should never be turned on by any @option{-O} option since
5594 it can result in incorrect output for programs which depend on
5595 an exact implementation of IEEE or ISO rules/specifications for
5598 The default is @option{-fmath-errno}.
5600 On Darwin systems, the math library never sets @code{errno}. There is therefore
5601 no reason for the compiler to consider the possibility that it might,
5602 and @option{-fno-math-errno} is the default.
5604 @item -funsafe-math-optimizations
5605 @opindex funsafe-math-optimizations
5606 Allow optimizations for floating-point arithmetic that (a) assume
5607 that arguments and results are valid and (b) may violate IEEE or
5608 ANSI standards. When used at link-time, it may include libraries
5609 or startup files that change the default FPU control word or other
5610 similar optimizations.
5612 This option should never be turned on by any @option{-O} option since
5613 it can result in incorrect output for programs which depend on
5614 an exact implementation of IEEE or ISO rules/specifications for
5617 The default is @option{-fno-unsafe-math-optimizations}.
5619 @item -ffinite-math-only
5620 @opindex ffinite-math-only
5621 Allow optimizations for floating-point arithmetic that assume
5622 that arguments and results are not NaNs or +-Infs.
5624 This option should never be turned on by any @option{-O} option since
5625 it can result in incorrect output for programs which depend on
5626 an exact implementation of IEEE or ISO rules/specifications.
5628 The default is @option{-fno-finite-math-only}.
5630 @item -fno-trapping-math
5631 @opindex fno-trapping-math
5632 Compile code assuming that floating-point operations cannot generate
5633 user-visible traps. These traps include division by zero, overflow,
5634 underflow, inexact result and invalid operation. This option implies
5635 @option{-fno-signaling-nans}. Setting this option may allow faster
5636 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5638 This option should never be turned on by any @option{-O} option since
5639 it can result in incorrect output for programs which depend on
5640 an exact implementation of IEEE or ISO rules/specifications for
5643 The default is @option{-ftrapping-math}.
5645 @item -frounding-math
5646 @opindex frounding-math
5647 Disable transformations and optimizations that assume default floating
5648 point rounding behavior. This is round-to-zero for all floating point
5649 to integer conversions, and round-to-nearest for all other arithmetic
5650 truncations. This option should be specified for programs that change
5651 the FP rounding mode dynamically, or that may be executed with a
5652 non-default rounding mode. This option disables constant folding of
5653 floating point expressions at compile-time (which may be affected by
5654 rounding mode) and arithmetic transformations that are unsafe in the
5655 presence of sign-dependent rounding modes.
5657 The default is @option{-fno-rounding-math}.
5659 This option is experimental and does not currently guarantee to
5660 disable all GCC optimizations that are affected by rounding mode.
5661 Future versions of GCC may provide finer control of this setting
5662 using C99's @code{FENV_ACCESS} pragma. This command line option
5663 will be used to specify the default state for @code{FENV_ACCESS}.
5665 @item -frtl-abstract-sequences
5666 @opindex frtl-abstract-sequences
5667 It is a size optimization method. This option is to find identical
5668 sequences of code, which can be turned into pseudo-procedures and
5669 then replace all occurrences with calls to the newly created
5670 subroutine. It is kind of an opposite of @option{-finline-functions}.
5671 This optimization runs at RTL level.
5673 @item -fsignaling-nans
5674 @opindex fsignaling-nans
5675 Compile code assuming that IEEE signaling NaNs may generate user-visible
5676 traps during floating-point operations. Setting this option disables
5677 optimizations that may change the number of exceptions visible with
5678 signaling NaNs. This option implies @option{-ftrapping-math}.
5680 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5683 The default is @option{-fno-signaling-nans}.
5685 This option is experimental and does not currently guarantee to
5686 disable all GCC optimizations that affect signaling NaN behavior.
5688 @item -fsingle-precision-constant
5689 @opindex fsingle-precision-constant
5690 Treat floating point constant as single precision constant instead of
5691 implicitly converting it to double precision constant.
5693 @item -fcx-limited-range
5694 @itemx -fno-cx-limited-range
5695 @opindex fcx-limited-range
5696 @opindex fno-cx-limited-range
5697 When enabled, this option states that a range reduction step is not
5698 needed when performing complex division. The default is
5699 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5701 This option controls the default setting of the ISO C99
5702 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5707 The following options control optimizations that may improve
5708 performance, but are not enabled by any @option{-O} options. This
5709 section includes experimental options that may produce broken code.
5712 @item -fbranch-probabilities
5713 @opindex fbranch-probabilities
5714 After running a program compiled with @option{-fprofile-arcs}
5715 (@pxref{Debugging Options,, Options for Debugging Your Program or
5716 @command{gcc}}), you can compile it a second time using
5717 @option{-fbranch-probabilities}, to improve optimizations based on
5718 the number of times each branch was taken. When the program
5719 compiled with @option{-fprofile-arcs} exits it saves arc execution
5720 counts to a file called @file{@var{sourcename}.gcda} for each source
5721 file The information in this data file is very dependent on the
5722 structure of the generated code, so you must use the same source code
5723 and the same optimization options for both compilations.
5725 With @option{-fbranch-probabilities}, GCC puts a
5726 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5727 These can be used to improve optimization. Currently, they are only
5728 used in one place: in @file{reorg.c}, instead of guessing which path a
5729 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5730 exactly determine which path is taken more often.
5732 @item -fprofile-values
5733 @opindex fprofile-values
5734 If combined with @option{-fprofile-arcs}, it adds code so that some
5735 data about values of expressions in the program is gathered.
5737 With @option{-fbranch-probabilities}, it reads back the data gathered
5738 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5739 notes to instructions for their later usage in optimizations.
5741 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5745 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5746 a code to gather information about values of expressions.
5748 With @option{-fbranch-probabilities}, it reads back the data gathered
5749 and actually performs the optimizations based on them.
5750 Currently the optimizations include specialization of division operation
5751 using the knowledge about the value of the denominator.
5753 @item -frename-registers
5754 @opindex frename-registers
5755 Attempt to avoid false dependencies in scheduled code by making use
5756 of registers left over after register allocation. This optimization
5757 will most benefit processors with lots of registers. Depending on the
5758 debug information format adopted by the target, however, it can
5759 make debugging impossible, since variables will no longer stay in
5760 a ``home register''.
5762 Enabled by default with @option{-funroll-loops}.
5766 Perform tail duplication to enlarge superblock size. This transformation
5767 simplifies the control flow of the function allowing other optimizations to do
5770 Enabled with @option{-fprofile-use}.
5772 @item -funroll-loops
5773 @opindex funroll-loops
5774 Unroll loops whose number of iterations can be determined at compile time or
5775 upon entry to the loop. @option{-funroll-loops} implies
5776 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5777 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5778 small constant number of iterations). This option makes code larger, and may
5779 or may not make it run faster.
5781 Enabled with @option{-fprofile-use}.
5783 @item -funroll-all-loops
5784 @opindex funroll-all-loops
5785 Unroll all loops, even if their number of iterations is uncertain when
5786 the loop is entered. This usually makes programs run more slowly.
5787 @option{-funroll-all-loops} implies the same options as
5788 @option{-funroll-loops}.
5791 @opindex fpeel-loops
5792 Peels the loops for that there is enough information that they do not
5793 roll much (from profile feedback). It also turns on complete loop peeling
5794 (i.e.@: complete removal of loops with small constant number of iterations).
5796 Enabled with @option{-fprofile-use}.
5798 @item -fmove-loop-invariants
5799 @opindex fmove-loop-invariants
5800 Enables the loop invariant motion pass in the RTL loop optimizer. Enabled
5801 at level @option{-O1}
5803 @item -funswitch-loops
5804 @opindex funswitch-loops
5805 Move branches with loop invariant conditions out of the loop, with duplicates
5806 of the loop on both branches (modified according to result of the condition).
5808 @item -ffunction-sections
5809 @itemx -fdata-sections
5810 @opindex ffunction-sections
5811 @opindex fdata-sections
5812 Place each function or data item into its own section in the output
5813 file if the target supports arbitrary sections. The name of the
5814 function or the name of the data item determines the section's name
5817 Use these options on systems where the linker can perform optimizations
5818 to improve locality of reference in the instruction space. Most systems
5819 using the ELF object format and SPARC processors running Solaris 2 have
5820 linkers with such optimizations. AIX may have these optimizations in
5823 Only use these options when there are significant benefits from doing
5824 so. When you specify these options, the assembler and linker will
5825 create larger object and executable files and will also be slower.
5826 You will not be able to use @code{gprof} on all systems if you
5827 specify this option and you may have problems with debugging if
5828 you specify both this option and @option{-g}.
5830 @item -fbranch-target-load-optimize
5831 @opindex fbranch-target-load-optimize
5832 Perform branch target register load optimization before prologue / epilogue
5834 The use of target registers can typically be exposed only during reload,
5835 thus hoisting loads out of loops and doing inter-block scheduling needs
5836 a separate optimization pass.
5838 @item -fbranch-target-load-optimize2
5839 @opindex fbranch-target-load-optimize2
5840 Perform branch target register load optimization after prologue / epilogue
5843 @item -fbtr-bb-exclusive
5844 @opindex fbtr-bb-exclusive
5845 When performing branch target register load optimization, don't reuse
5846 branch target registers in within any basic block.
5848 @item -fstack-protector
5849 Emit extra code to check for buffer overflows, such as stack smashing
5850 attacks. This is done by adding a guard variable to functions with
5851 vulnerable objects. This includes functions that call alloca, and
5852 functions with buffers larger than 8 bytes. The guards are initialized
5853 when a function is entered and then checked when the function exits.
5854 If a guard check fails, an error message is printed and the program exits.
5856 @item -fstack-protector-all
5857 Like @option{-fstack-protector} except that all functions are protected.
5859 @item -fsection-anchors
5860 @opindex fsection-anchors
5861 Try to reduce the number of symbolic address calculations by using
5862 shared ``anchor'' symbols to address nearby objects. This transformation
5863 can help to reduce the number of GOT entries and GOT accesses on some
5866 For example, the implementation of the following function @code{foo}:
5870 int foo (void) @{ return a + b + c; @}
5873 would usually calculate the addresses of all three variables, but if you
5874 compile it with @option{-fsection-anchors}, it will access the variables
5875 from a common anchor point instead. The effect is similar to the
5876 following pseudocode (which isn't valid C):
5881 register int *xr = &x;
5882 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5886 Not all targets support this option.
5888 @item --param @var{name}=@var{value}
5890 In some places, GCC uses various constants to control the amount of
5891 optimization that is done. For example, GCC will not inline functions
5892 that contain more that a certain number of instructions. You can
5893 control some of these constants on the command-line using the
5894 @option{--param} option.
5896 The names of specific parameters, and the meaning of the values, are
5897 tied to the internals of the compiler, and are subject to change
5898 without notice in future releases.
5900 In each case, the @var{value} is an integer. The allowable choices for
5901 @var{name} are given in the following table:
5904 @item salias-max-implicit-fields
5905 The maximum number of fields in a variable without direct
5906 structure accesses for which structure aliasing will consider trying
5907 to track each field. The default is 5
5909 @item salias-max-array-elements
5910 The maximum number of elements an array can have and its elements
5911 still be tracked individually by structure aliasing. The default is 4
5913 @item sra-max-structure-size
5914 The maximum structure size, in bytes, at which the scalar replacement
5915 of aggregates (SRA) optimization will perform block copies. The
5916 default value, 0, implies that GCC will select the most appropriate
5919 @item sra-field-structure-ratio
5920 The threshold ratio (as a percentage) between instantiated fields and
5921 the complete structure size. We say that if the ratio of the number
5922 of bytes in instantiated fields to the number of bytes in the complete
5923 structure exceeds this parameter, then block copies are not used. The
5926 @item max-crossjump-edges
5927 The maximum number of incoming edges to consider for crossjumping.
5928 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5929 the number of edges incoming to each block. Increasing values mean
5930 more aggressive optimization, making the compile time increase with
5931 probably small improvement in executable size.
5933 @item min-crossjump-insns
5934 The minimum number of instructions which must be matched at the end
5935 of two blocks before crossjumping will be performed on them. This
5936 value is ignored in the case where all instructions in the block being
5937 crossjumped from are matched. The default value is 5.
5939 @item max-grow-copy-bb-insns
5940 The maximum code size expansion factor when copying basic blocks
5941 instead of jumping. The expansion is relative to a jump instruction.
5942 The default value is 8.
5944 @item max-goto-duplication-insns
5945 The maximum number of instructions to duplicate to a block that jumps
5946 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5947 passes, GCC factors computed gotos early in the compilation process,
5948 and unfactors them as late as possible. Only computed jumps at the
5949 end of a basic blocks with no more than max-goto-duplication-insns are
5950 unfactored. The default value is 8.
5952 @item max-delay-slot-insn-search
5953 The maximum number of instructions to consider when looking for an
5954 instruction to fill a delay slot. If more than this arbitrary number of
5955 instructions is searched, the time savings from filling the delay slot
5956 will be minimal so stop searching. Increasing values mean more
5957 aggressive optimization, making the compile time increase with probably
5958 small improvement in executable run time.
5960 @item max-delay-slot-live-search
5961 When trying to fill delay slots, the maximum number of instructions to
5962 consider when searching for a block with valid live register
5963 information. Increasing this arbitrarily chosen value means more
5964 aggressive optimization, increasing the compile time. This parameter
5965 should be removed when the delay slot code is rewritten to maintain the
5968 @item max-gcse-memory
5969 The approximate maximum amount of memory that will be allocated in
5970 order to perform the global common subexpression elimination
5971 optimization. If more memory than specified is required, the
5972 optimization will not be done.
5974 @item max-gcse-passes
5975 The maximum number of passes of GCSE to run. The default is 1.
5977 @item max-pending-list-length
5978 The maximum number of pending dependencies scheduling will allow
5979 before flushing the current state and starting over. Large functions
5980 with few branches or calls can create excessively large lists which
5981 needlessly consume memory and resources.
5983 @item max-inline-insns-single
5984 Several parameters control the tree inliner used in gcc.
5985 This number sets the maximum number of instructions (counted in GCC's
5986 internal representation) in a single function that the tree inliner
5987 will consider for inlining. This only affects functions declared
5988 inline and methods implemented in a class declaration (C++).
5989 The default value is 450.
5991 @item max-inline-insns-auto
5992 When you use @option{-finline-functions} (included in @option{-O3}),
5993 a lot of functions that would otherwise not be considered for inlining
5994 by the compiler will be investigated. To those functions, a different
5995 (more restrictive) limit compared to functions declared inline can
5997 The default value is 90.
5999 @item large-function-insns
6000 The limit specifying really large functions. For functions larger than this
6001 limit after inlining inlining is constrained by
6002 @option{--param large-function-growth}. This parameter is useful primarily
6003 to avoid extreme compilation time caused by non-linear algorithms used by the
6005 This parameter is ignored when @option{-funit-at-a-time} is not used.
6006 The default value is 2700.
6008 @item large-function-growth
6009 Specifies maximal growth of large function caused by inlining in percents.
6010 This parameter is ignored when @option{-funit-at-a-time} is not used.
6011 The default value is 100 which limits large function growth to 2.0 times
6014 @item large-unit-insns
6015 The limit specifying large translation unit. Growth caused by inlining of
6016 units larger than this limit is limited by @option{--param inline-unit-growth}.
6017 For small units this might be too tight (consider unit consisting of function A
6018 that is inline and B that just calls A three time. If B is small relative to
6019 A, the growth of unit is 300\% and yet such inlining is very sane. For very
6020 large units consisting of small inlininable functions however the overall unit
6021 growth limit is needed to avoid exponential explosion of code size. Thus for
6022 smaller units, the size is increased to @option{--param large-unit-insns}
6023 before applying @option{--param inline-unit-growth}. The default is 10000
6025 @item inline-unit-growth
6026 Specifies maximal overall growth of the compilation unit caused by inlining.
6027 This parameter is ignored when @option{-funit-at-a-time} is not used.
6028 The default value is 50 which limits unit growth to 1.5 times the original
6031 @item max-inline-insns-recursive
6032 @itemx max-inline-insns-recursive-auto
6033 Specifies maximum number of instructions out-of-line copy of self recursive inline
6034 function can grow into by performing recursive inlining.
6036 For functions declared inline @option{--param max-inline-insns-recursive} is
6037 taken into account. For function not declared inline, recursive inlining
6038 happens only when @option{-finline-functions} (included in @option{-O3}) is
6039 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
6040 default value is 450.
6042 @item max-inline-recursive-depth
6043 @itemx max-inline-recursive-depth-auto
6044 Specifies maximum recursion depth used by the recursive inlining.
6046 For functions declared inline @option{--param max-inline-recursive-depth} is
6047 taken into account. For function not declared inline, recursive inlining
6048 happens only when @option{-finline-functions} (included in @option{-O3}) is
6049 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
6050 default value is 450.
6052 @item min-inline-recursive-probability
6053 Recursive inlining is profitable only for function having deep recursion
6054 in average and can hurt for function having little recursion depth by
6055 increasing the prologue size or complexity of function body to other
6058 When profile feedback is available (see @option{-fprofile-generate}) the actual
6059 recursion depth can be guessed from probability that function will recurse via
6060 given call expression. This parameter limits inlining only to call expression
6061 whose probability exceeds given threshold (in percents). The default value is
6064 @item inline-call-cost
6065 Specify cost of call instruction relative to simple arithmetics operations
6066 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
6067 functions and at the same time increases size of leaf function that is believed to
6068 reduce function size by being inlined. In effect it increases amount of
6069 inlining for code having large abstraction penalty (many functions that just
6070 pass the arguments to other functions) and decrease inlining for code with low
6071 abstraction penalty. The default value is 16.
6073 @item max-unrolled-insns
6074 The maximum number of instructions that a loop should have if that loop
6075 is unrolled, and if the loop is unrolled, it determines how many times
6076 the loop code is unrolled.
6078 @item max-average-unrolled-insns
6079 The maximum number of instructions biased by probabilities of their execution
6080 that a loop should have if that loop is unrolled, and if the loop is unrolled,
6081 it determines how many times the loop code is unrolled.
6083 @item max-unroll-times
6084 The maximum number of unrollings of a single loop.
6086 @item max-peeled-insns
6087 The maximum number of instructions that a loop should have if that loop
6088 is peeled, and if the loop is peeled, it determines how many times
6089 the loop code is peeled.
6091 @item max-peel-times
6092 The maximum number of peelings of a single loop.
6094 @item max-completely-peeled-insns
6095 The maximum number of insns of a completely peeled loop.
6097 @item max-completely-peel-times
6098 The maximum number of iterations of a loop to be suitable for complete peeling.
6100 @item max-unswitch-insns
6101 The maximum number of insns of an unswitched loop.
6103 @item max-unswitch-level
6104 The maximum number of branches unswitched in a single loop.
6107 The minimum cost of an expensive expression in the loop invariant motion.
6109 @item iv-consider-all-candidates-bound
6110 Bound on number of candidates for induction variables below that
6111 all candidates are considered for each use in induction variable
6112 optimizations. Only the most relevant candidates are considered
6113 if there are more candidates, to avoid quadratic time complexity.
6115 @item iv-max-considered-uses
6116 The induction variable optimizations give up on loops that contain more
6117 induction variable uses.
6119 @item iv-always-prune-cand-set-bound
6120 If number of candidates in the set is smaller than this value,
6121 we always try to remove unnecessary ivs from the set during its
6122 optimization when a new iv is added to the set.
6124 @item scev-max-expr-size
6125 Bound on size of expressions used in the scalar evolutions analyzer.
6126 Large expressions slow the analyzer.
6128 @item vect-max-version-checks
6129 The maximum number of runtime checks that can be performed when doing
6130 loop versioning in the vectorizer. See option ftree-vect-loop-version
6131 for more information.
6133 @item max-iterations-to-track
6135 The maximum number of iterations of a loop the brute force algorithm
6136 for analysis of # of iterations of the loop tries to evaluate.
6138 @item hot-bb-count-fraction
6139 Select fraction of the maximal count of repetitions of basic block in program
6140 given basic block needs to have to be considered hot.
6142 @item hot-bb-frequency-fraction
6143 Select fraction of the maximal frequency of executions of basic block in
6144 function given basic block needs to have to be considered hot
6146 @item max-predicted-iterations
6147 The maximum number of loop iterations we predict statically. This is useful
6148 in cases where function contain single loop with known bound and other loop
6149 with unknown. We predict the known number of iterations correctly, while
6150 the unknown number of iterations average to roughly 10. This means that the
6151 loop without bounds would appear artificially cold relative to the other one.
6153 @item tracer-dynamic-coverage
6154 @itemx tracer-dynamic-coverage-feedback
6156 This value is used to limit superblock formation once the given percentage of
6157 executed instructions is covered. This limits unnecessary code size
6160 The @option{tracer-dynamic-coverage-feedback} is used only when profile
6161 feedback is available. The real profiles (as opposed to statically estimated
6162 ones) are much less balanced allowing the threshold to be larger value.
6164 @item tracer-max-code-growth
6165 Stop tail duplication once code growth has reached given percentage. This is
6166 rather hokey argument, as most of the duplicates will be eliminated later in
6167 cross jumping, so it may be set to much higher values than is the desired code
6170 @item tracer-min-branch-ratio
6172 Stop reverse growth when the reverse probability of best edge is less than this
6173 threshold (in percent).
6175 @item tracer-min-branch-ratio
6176 @itemx tracer-min-branch-ratio-feedback
6178 Stop forward growth if the best edge do have probability lower than this
6181 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
6182 compilation for profile feedback and one for compilation without. The value
6183 for compilation with profile feedback needs to be more conservative (higher) in
6184 order to make tracer effective.
6186 @item max-cse-path-length
6188 Maximum number of basic blocks on path that cse considers. The default is 10.
6191 The maximum instructions CSE process before flushing. The default is 1000.
6193 @item global-var-threshold
6195 Counts the number of function calls (@var{n}) and the number of
6196 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6197 single artificial variable will be created to represent all the
6198 call-clobbered variables at function call sites. This artificial
6199 variable will then be made to alias every call-clobbered variable.
6200 (done as @code{int * size_t} on the host machine; beware overflow).
6202 @item max-aliased-vops
6204 Maximum number of virtual operands allowed to represent aliases
6205 before triggering the alias grouping heuristic. Alias grouping
6206 reduces compile times and memory consumption needed for aliasing at
6207 the expense of precision loss in alias information.
6209 @item ggc-min-expand
6211 GCC uses a garbage collector to manage its own memory allocation. This
6212 parameter specifies the minimum percentage by which the garbage
6213 collector's heap should be allowed to expand between collections.
6214 Tuning this may improve compilation speed; it has no effect on code
6217 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6218 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6219 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6220 GCC is not able to calculate RAM on a particular platform, the lower
6221 bound of 30% is used. Setting this parameter and
6222 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6223 every opportunity. This is extremely slow, but can be useful for
6226 @item ggc-min-heapsize
6228 Minimum size of the garbage collector's heap before it begins bothering
6229 to collect garbage. The first collection occurs after the heap expands
6230 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6231 tuning this may improve compilation speed, and has no effect on code
6234 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6235 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6236 with a lower bound of 4096 (four megabytes) and an upper bound of
6237 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6238 particular platform, the lower bound is used. Setting this parameter
6239 very large effectively disables garbage collection. Setting this
6240 parameter and @option{ggc-min-expand} to zero causes a full collection
6241 to occur at every opportunity.
6243 @item max-reload-search-insns
6244 The maximum number of instruction reload should look backward for equivalent
6245 register. Increasing values mean more aggressive optimization, making the
6246 compile time increase with probably slightly better performance. The default
6249 @item max-cselib-memory-locations
6250 The maximum number of memory locations cselib should take into account.
6251 Increasing values mean more aggressive optimization, making the compile time
6252 increase with probably slightly better performance. The default value is 500.
6254 @item max-flow-memory-locations
6255 Similar as @option{max-cselib-memory-locations} but for dataflow liveness.
6256 The default value is 100.
6258 @item reorder-blocks-duplicate
6259 @itemx reorder-blocks-duplicate-feedback
6261 Used by basic block reordering pass to decide whether to use unconditional
6262 branch or duplicate the code on its destination. Code is duplicated when its
6263 estimated size is smaller than this value multiplied by the estimated size of
6264 unconditional jump in the hot spots of the program.
6266 The @option{reorder-block-duplicate-feedback} is used only when profile
6267 feedback is available and may be set to higher values than
6268 @option{reorder-block-duplicate} since information about the hot spots is more
6271 @item max-sched-ready-insns
6272 The maximum number of instructions ready to be issued the scheduler should
6273 consider at any given time during the first scheduling pass. Increasing
6274 values mean more thorough searches, making the compilation time increase
6275 with probably little benefit. The default value is 100.
6277 @item max-sched-region-blocks
6278 The maximum number of blocks in a region to be considered for
6279 interblock scheduling. The default value is 10.
6281 @item max-sched-region-insns
6282 The maximum number of insns in a region to be considered for
6283 interblock scheduling. The default value is 100.
6286 The minimum probability (in percents) of reaching a source block
6287 for interblock speculative scheduling. The default value is 40.
6289 @item max-sched-extend-regions-iters
6290 The maximum number of iterations through CFG to extend regions.
6291 0 - disable region extension,
6292 N - do at most N iterations.
6293 The default value is 0.
6295 @item max-sched-insn-conflict-delay
6296 The maximum conflict delay for an insn to be considered for speculative motion.
6297 The default value is 3.
6299 @item sched-spec-prob-cutoff
6300 The minimal probability of speculation success (in percents), so that
6301 speculative insn will be scheduled.
6302 The default value is 40.
6304 @item max-last-value-rtl
6306 The maximum size measured as number of RTLs that can be recorded in an expression
6307 in combiner for a pseudo register as last known value of that register. The default
6310 @item integer-share-limit
6311 Small integer constants can use a shared data structure, reducing the
6312 compiler's memory usage and increasing its speed. This sets the maximum
6313 value of a shared integer constant's. The default value is 256.
6315 @item min-virtual-mappings
6316 Specifies the minimum number of virtual mappings in the incremental
6317 SSA updater that should be registered to trigger the virtual mappings
6318 heuristic defined by virtual-mappings-ratio. The default value is
6321 @item virtual-mappings-ratio
6322 If the number of virtual mappings is virtual-mappings-ratio bigger
6323 than the number of virtual symbols to be updated, then the incremental
6324 SSA updater switches to a full update for those symbols. The default
6327 @item ssp-buffer-size
6328 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6329 protection when @option{-fstack-protection} is used.
6331 @item max-jump-thread-duplication-stmts
6332 Maximum number of statements allowed in a block that needs to be
6333 duplicated when threading jumps.
6335 @item max-fields-for-field-sensitive
6336 Maximum number of fields in a structure we will treat in
6337 a field sensitive manner during pointer analysis.
6339 @item prefetch-latency
6340 Estimate on average number of instructions that are executed before
6341 prefetch finishes. The distance we prefetch ahead is proportional
6342 to this constant. Increasing this number may also lead to less
6343 streams being prefetched (see @option{simultaneous-prefetches}).
6345 @item simultaneous-prefetches
6346 Maximum number of prefetches that can run at the same time.
6348 @item l1-cache-line-size
6349 The size of cache line in L1 cache, in bytes.
6352 The number of cache lines in L1 cache.
6357 @node Preprocessor Options
6358 @section Options Controlling the Preprocessor
6359 @cindex preprocessor options
6360 @cindex options, preprocessor
6362 These options control the C preprocessor, which is run on each C source
6363 file before actual compilation.
6365 If you use the @option{-E} option, nothing is done except preprocessing.
6366 Some of these options make sense only together with @option{-E} because
6367 they cause the preprocessor output to be unsuitable for actual
6372 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6373 and pass @var{option} directly through to the preprocessor. If
6374 @var{option} contains commas, it is split into multiple options at the
6375 commas. However, many options are modified, translated or interpreted
6376 by the compiler driver before being passed to the preprocessor, and
6377 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6378 interface is undocumented and subject to change, so whenever possible
6379 you should avoid using @option{-Wp} and let the driver handle the
6382 @item -Xpreprocessor @var{option}
6383 @opindex preprocessor
6384 Pass @var{option} as an option to the preprocessor. You can use this to
6385 supply system-specific preprocessor options which GCC does not know how to
6388 If you want to pass an option that takes an argument, you must use
6389 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6392 @include cppopts.texi
6394 @node Assembler Options
6395 @section Passing Options to the Assembler
6397 @c prevent bad page break with this line
6398 You can pass options to the assembler.
6401 @item -Wa,@var{option}
6403 Pass @var{option} as an option to the assembler. If @var{option}
6404 contains commas, it is split into multiple options at the commas.
6406 @item -Xassembler @var{option}
6408 Pass @var{option} as an option to the assembler. You can use this to
6409 supply system-specific assembler options which GCC does not know how to
6412 If you want to pass an option that takes an argument, you must use
6413 @option{-Xassembler} twice, once for the option and once for the argument.
6418 @section Options for Linking
6419 @cindex link options
6420 @cindex options, linking
6422 These options come into play when the compiler links object files into
6423 an executable output file. They are meaningless if the compiler is
6424 not doing a link step.
6428 @item @var{object-file-name}
6429 A file name that does not end in a special recognized suffix is
6430 considered to name an object file or library. (Object files are
6431 distinguished from libraries by the linker according to the file
6432 contents.) If linking is done, these object files are used as input
6441 If any of these options is used, then the linker is not run, and
6442 object file names should not be used as arguments. @xref{Overall
6446 @item -l@var{library}
6447 @itemx -l @var{library}
6449 Search the library named @var{library} when linking. (The second
6450 alternative with the library as a separate argument is only for
6451 POSIX compliance and is not recommended.)
6453 It makes a difference where in the command you write this option; the
6454 linker searches and processes libraries and object files in the order they
6455 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6456 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6457 to functions in @samp{z}, those functions may not be loaded.
6459 The linker searches a standard list of directories for the library,
6460 which is actually a file named @file{lib@var{library}.a}. The linker
6461 then uses this file as if it had been specified precisely by name.
6463 The directories searched include several standard system directories
6464 plus any that you specify with @option{-L}.
6466 Normally the files found this way are library files---archive files
6467 whose members are object files. The linker handles an archive file by
6468 scanning through it for members which define symbols that have so far
6469 been referenced but not defined. But if the file that is found is an
6470 ordinary object file, it is linked in the usual fashion. The only
6471 difference between using an @option{-l} option and specifying a file name
6472 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6473 and searches several directories.
6477 You need this special case of the @option{-l} option in order to
6478 link an Objective-C or Objective-C++ program.
6481 @opindex nostartfiles
6482 Do not use the standard system startup files when linking.
6483 The standard system libraries are used normally, unless @option{-nostdlib}
6484 or @option{-nodefaultlibs} is used.
6486 @item -nodefaultlibs
6487 @opindex nodefaultlibs
6488 Do not use the standard system libraries when linking.
6489 Only the libraries you specify will be passed to the linker.
6490 The standard startup files are used normally, unless @option{-nostartfiles}
6491 is used. The compiler may generate calls to @code{memcmp},
6492 @code{memset}, @code{memcpy} and @code{memmove}.
6493 These entries are usually resolved by entries in
6494 libc. These entry points should be supplied through some other
6495 mechanism when this option is specified.
6499 Do not use the standard system startup files or libraries when linking.
6500 No startup files and only the libraries you specify will be passed to
6501 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6502 @code{memcpy} and @code{memmove}.
6503 These entries are usually resolved by entries in
6504 libc. These entry points should be supplied through some other
6505 mechanism when this option is specified.
6507 @cindex @option{-lgcc}, use with @option{-nostdlib}
6508 @cindex @option{-nostdlib} and unresolved references
6509 @cindex unresolved references and @option{-nostdlib}
6510 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6511 @cindex @option{-nodefaultlibs} and unresolved references
6512 @cindex unresolved references and @option{-nodefaultlibs}
6513 One of the standard libraries bypassed by @option{-nostdlib} and
6514 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6515 that GCC uses to overcome shortcomings of particular machines, or special
6516 needs for some languages.
6517 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6518 Collection (GCC) Internals},
6519 for more discussion of @file{libgcc.a}.)
6520 In most cases, you need @file{libgcc.a} even when you want to avoid
6521 other standard libraries. In other words, when you specify @option{-nostdlib}
6522 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6523 This ensures that you have no unresolved references to internal GCC
6524 library subroutines. (For example, @samp{__main}, used to ensure C++
6525 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6526 GNU Compiler Collection (GCC) Internals}.)
6530 Produce a position independent executable on targets which support it.
6531 For predictable results, you must also specify the same set of options
6532 that were used to generate code (@option{-fpie}, @option{-fPIE},
6533 or model suboptions) when you specify this option.
6537 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6538 that support it. This instructs the linker to add all symbols, not
6539 only used ones, to the dynamic symbol table. This option is needed
6540 for some uses of @code{dlopen} or to allow obtaining backtraces
6541 from within a program.
6545 Remove all symbol table and relocation information from the executable.
6549 On systems that support dynamic linking, this prevents linking with the shared
6550 libraries. On other systems, this option has no effect.
6554 Produce a shared object which can then be linked with other objects to
6555 form an executable. Not all systems support this option. For predictable
6556 results, you must also specify the same set of options that were used to
6557 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6558 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6559 needs to build supplementary stub code for constructors to work. On
6560 multi-libbed systems, @samp{gcc -shared} must select the correct support
6561 libraries to link against. Failing to supply the correct flags may lead
6562 to subtle defects. Supplying them in cases where they are not necessary
6565 @item -shared-libgcc
6566 @itemx -static-libgcc
6567 @opindex shared-libgcc
6568 @opindex static-libgcc
6569 On systems that provide @file{libgcc} as a shared library, these options
6570 force the use of either the shared or static version respectively.
6571 If no shared version of @file{libgcc} was built when the compiler was
6572 configured, these options have no effect.
6574 There are several situations in which an application should use the
6575 shared @file{libgcc} instead of the static version. The most common
6576 of these is when the application wishes to throw and catch exceptions
6577 across different shared libraries. In that case, each of the libraries
6578 as well as the application itself should use the shared @file{libgcc}.
6580 Therefore, the G++ and GCJ drivers automatically add
6581 @option{-shared-libgcc} whenever you build a shared library or a main
6582 executable, because C++ and Java programs typically use exceptions, so
6583 this is the right thing to do.
6585 If, instead, you use the GCC driver to create shared libraries, you may
6586 find that they will not always be linked with the shared @file{libgcc}.
6587 If GCC finds, at its configuration time, that you have a non-GNU linker
6588 or a GNU linker that does not support option @option{--eh-frame-hdr},
6589 it will link the shared version of @file{libgcc} into shared libraries
6590 by default. Otherwise, it will take advantage of the linker and optimize
6591 away the linking with the shared version of @file{libgcc}, linking with
6592 the static version of libgcc by default. This allows exceptions to
6593 propagate through such shared libraries, without incurring relocation
6594 costs at library load time.
6596 However, if a library or main executable is supposed to throw or catch
6597 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6598 for the languages used in the program, or using the option
6599 @option{-shared-libgcc}, such that it is linked with the shared
6604 Bind references to global symbols when building a shared object. Warn
6605 about any unresolved references (unless overridden by the link editor
6606 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6609 @item -Xlinker @var{option}
6611 Pass @var{option} as an option to the linker. You can use this to
6612 supply system-specific linker options which GCC does not know how to
6615 If you want to pass an option that takes an argument, you must use
6616 @option{-Xlinker} twice, once for the option and once for the argument.
6617 For example, to pass @option{-assert definitions}, you must write
6618 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6619 @option{-Xlinker "-assert definitions"}, because this passes the entire
6620 string as a single argument, which is not what the linker expects.
6622 @item -Wl,@var{option}
6624 Pass @var{option} as an option to the linker. If @var{option} contains
6625 commas, it is split into multiple options at the commas.
6627 @item -u @var{symbol}
6629 Pretend the symbol @var{symbol} is undefined, to force linking of
6630 library modules to define it. You can use @option{-u} multiple times with
6631 different symbols to force loading of additional library modules.
6634 @node Directory Options
6635 @section Options for Directory Search
6636 @cindex directory options
6637 @cindex options, directory search
6640 These options specify directories to search for header files, for
6641 libraries and for parts of the compiler:
6646 Add the directory @var{dir} to the head of the list of directories to be
6647 searched for header files. This can be used to override a system header
6648 file, substituting your own version, since these directories are
6649 searched before the system header file directories. However, you should
6650 not use this option to add directories that contain vendor-supplied
6651 system header files (use @option{-isystem} for that). If you use more than
6652 one @option{-I} option, the directories are scanned in left-to-right
6653 order; the standard system directories come after.
6655 If a standard system include directory, or a directory specified with
6656 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6657 option will be ignored. The directory will still be searched but as a
6658 system directory at its normal position in the system include chain.
6659 This is to ensure that GCC's procedure to fix buggy system headers and
6660 the ordering for the include_next directive are not inadvertently changed.
6661 If you really need to change the search order for system directories,
6662 use the @option{-nostdinc} and/or @option{-isystem} options.
6664 @item -iquote@var{dir}
6666 Add the directory @var{dir} to the head of the list of directories to
6667 be searched for header files only for the case of @samp{#include
6668 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6669 otherwise just like @option{-I}.
6673 Add directory @var{dir} to the list of directories to be searched
6676 @item -B@var{prefix}
6678 This option specifies where to find the executables, libraries,
6679 include files, and data files of the compiler itself.
6681 The compiler driver program runs one or more of the subprograms
6682 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6683 @var{prefix} as a prefix for each program it tries to run, both with and
6684 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6686 For each subprogram to be run, the compiler driver first tries the
6687 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6688 was not specified, the driver tries two standard prefixes, which are
6689 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6690 those results in a file name that is found, the unmodified program
6691 name is searched for using the directories specified in your
6692 @env{PATH} environment variable.
6694 The compiler will check to see if the path provided by the @option{-B}
6695 refers to a directory, and if necessary it will add a directory
6696 separator character at the end of the path.
6698 @option{-B} prefixes that effectively specify directory names also apply
6699 to libraries in the linker, because the compiler translates these
6700 options into @option{-L} options for the linker. They also apply to
6701 includes files in the preprocessor, because the compiler translates these
6702 options into @option{-isystem} options for the preprocessor. In this case,
6703 the compiler appends @samp{include} to the prefix.
6705 The run-time support file @file{libgcc.a} can also be searched for using
6706 the @option{-B} prefix, if needed. If it is not found there, the two
6707 standard prefixes above are tried, and that is all. The file is left
6708 out of the link if it is not found by those means.
6710 Another way to specify a prefix much like the @option{-B} prefix is to use
6711 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6714 As a special kludge, if the path provided by @option{-B} is
6715 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6716 9, then it will be replaced by @file{[dir/]include}. This is to help
6717 with boot-strapping the compiler.
6719 @item -specs=@var{file}
6721 Process @var{file} after the compiler reads in the standard @file{specs}
6722 file, in order to override the defaults that the @file{gcc} driver
6723 program uses when determining what switches to pass to @file{cc1},
6724 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6725 @option{-specs=@var{file}} can be specified on the command line, and they
6726 are processed in order, from left to right.
6728 @item --sysroot=@var{dir}
6730 Use @var{dir} as the logical root directory for headers and libraries.
6731 For example, if the compiler would normally search for headers in
6732 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6733 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6735 If you use both this option and the @option{-isysroot} option, then
6736 the @option{--sysroot} option will apply to libraries, but the
6737 @option{-isysroot} option will apply to header files.
6739 The GNU linker (beginning with version 2.16) has the necessary support
6740 for this option. If your linker does not support this option, the
6741 header file aspect of @option{--sysroot} will still work, but the
6742 library aspect will not.
6746 This option has been deprecated. Please use @option{-iquote} instead for
6747 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6748 Any directories you specify with @option{-I} options before the @option{-I-}
6749 option are searched only for the case of @samp{#include "@var{file}"};
6750 they are not searched for @samp{#include <@var{file}>}.
6752 If additional directories are specified with @option{-I} options after
6753 the @option{-I-}, these directories are searched for all @samp{#include}
6754 directives. (Ordinarily @emph{all} @option{-I} directories are used
6757 In addition, the @option{-I-} option inhibits the use of the current
6758 directory (where the current input file came from) as the first search
6759 directory for @samp{#include "@var{file}"}. There is no way to
6760 override this effect of @option{-I-}. With @option{-I.} you can specify
6761 searching the directory which was current when the compiler was
6762 invoked. That is not exactly the same as what the preprocessor does
6763 by default, but it is often satisfactory.
6765 @option{-I-} does not inhibit the use of the standard system directories
6766 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6773 @section Specifying subprocesses and the switches to pass to them
6776 @command{gcc} is a driver program. It performs its job by invoking a
6777 sequence of other programs to do the work of compiling, assembling and
6778 linking. GCC interprets its command-line parameters and uses these to
6779 deduce which programs it should invoke, and which command-line options
6780 it ought to place on their command lines. This behavior is controlled
6781 by @dfn{spec strings}. In most cases there is one spec string for each
6782 program that GCC can invoke, but a few programs have multiple spec
6783 strings to control their behavior. The spec strings built into GCC can
6784 be overridden by using the @option{-specs=} command-line switch to specify
6787 @dfn{Spec files} are plaintext files that are used to construct spec
6788 strings. They consist of a sequence of directives separated by blank
6789 lines. The type of directive is determined by the first non-whitespace
6790 character on the line and it can be one of the following:
6793 @item %@var{command}
6794 Issues a @var{command} to the spec file processor. The commands that can
6798 @item %include <@var{file}>
6800 Search for @var{file} and insert its text at the current point in the
6803 @item %include_noerr <@var{file}>
6804 @cindex %include_noerr
6805 Just like @samp{%include}, but do not generate an error message if the include
6806 file cannot be found.
6808 @item %rename @var{old_name} @var{new_name}
6810 Rename the spec string @var{old_name} to @var{new_name}.
6814 @item *[@var{spec_name}]:
6815 This tells the compiler to create, override or delete the named spec
6816 string. All lines after this directive up to the next directive or
6817 blank line are considered to be the text for the spec string. If this
6818 results in an empty string then the spec will be deleted. (Or, if the
6819 spec did not exist, then nothing will happened.) Otherwise, if the spec
6820 does not currently exist a new spec will be created. If the spec does
6821 exist then its contents will be overridden by the text of this
6822 directive, unless the first character of that text is the @samp{+}
6823 character, in which case the text will be appended to the spec.
6825 @item [@var{suffix}]:
6826 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6827 and up to the next directive or blank line are considered to make up the
6828 spec string for the indicated suffix. When the compiler encounters an
6829 input file with the named suffix, it will processes the spec string in
6830 order to work out how to compile that file. For example:
6837 This says that any input file whose name ends in @samp{.ZZ} should be
6838 passed to the program @samp{z-compile}, which should be invoked with the
6839 command-line switch @option{-input} and with the result of performing the
6840 @samp{%i} substitution. (See below.)
6842 As an alternative to providing a spec string, the text that follows a
6843 suffix directive can be one of the following:
6846 @item @@@var{language}
6847 This says that the suffix is an alias for a known @var{language}. This is
6848 similar to using the @option{-x} command-line switch to GCC to specify a
6849 language explicitly. For example:
6856 Says that .ZZ files are, in fact, C++ source files.
6859 This causes an error messages saying:
6862 @var{name} compiler not installed on this system.
6866 GCC already has an extensive list of suffixes built into it.
6867 This directive will add an entry to the end of the list of suffixes, but
6868 since the list is searched from the end backwards, it is effectively
6869 possible to override earlier entries using this technique.
6873 GCC has the following spec strings built into it. Spec files can
6874 override these strings or create their own. Note that individual
6875 targets can also add their own spec strings to this list.
6878 asm Options to pass to the assembler
6879 asm_final Options to pass to the assembler post-processor
6880 cpp Options to pass to the C preprocessor
6881 cc1 Options to pass to the C compiler
6882 cc1plus Options to pass to the C++ compiler
6883 endfile Object files to include at the end of the link
6884 link Options to pass to the linker
6885 lib Libraries to include on the command line to the linker
6886 libgcc Decides which GCC support library to pass to the linker
6887 linker Sets the name of the linker
6888 predefines Defines to be passed to the C preprocessor
6889 signed_char Defines to pass to CPP to say whether @code{char} is signed
6891 startfile Object files to include at the start of the link
6894 Here is a small example of a spec file:
6900 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6903 This example renames the spec called @samp{lib} to @samp{old_lib} and
6904 then overrides the previous definition of @samp{lib} with a new one.
6905 The new definition adds in some extra command-line options before
6906 including the text of the old definition.
6908 @dfn{Spec strings} are a list of command-line options to be passed to their
6909 corresponding program. In addition, the spec strings can contain
6910 @samp{%}-prefixed sequences to substitute variable text or to
6911 conditionally insert text into the command line. Using these constructs
6912 it is possible to generate quite complex command lines.
6914 Here is a table of all defined @samp{%}-sequences for spec
6915 strings. Note that spaces are not generated automatically around the
6916 results of expanding these sequences. Therefore you can concatenate them
6917 together or combine them with constant text in a single argument.
6921 Substitute one @samp{%} into the program name or argument.
6924 Substitute the name of the input file being processed.
6927 Substitute the basename of the input file being processed.
6928 This is the substring up to (and not including) the last period
6929 and not including the directory.
6932 This is the same as @samp{%b}, but include the file suffix (text after
6936 Marks the argument containing or following the @samp{%d} as a
6937 temporary file name, so that that file will be deleted if GCC exits
6938 successfully. Unlike @samp{%g}, this contributes no text to the
6941 @item %g@var{suffix}
6942 Substitute a file name that has suffix @var{suffix} and is chosen
6943 once per compilation, and mark the argument in the same way as
6944 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6945 name is now chosen in a way that is hard to predict even when previously
6946 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6947 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6948 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6949 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6950 was simply substituted with a file name chosen once per compilation,
6951 without regard to any appended suffix (which was therefore treated
6952 just like ordinary text), making such attacks more likely to succeed.
6954 @item %u@var{suffix}
6955 Like @samp{%g}, but generates a new temporary file name even if
6956 @samp{%u@var{suffix}} was already seen.
6958 @item %U@var{suffix}
6959 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6960 new one if there is no such last file name. In the absence of any
6961 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6962 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6963 would involve the generation of two distinct file names, one
6964 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6965 simply substituted with a file name chosen for the previous @samp{%u},
6966 without regard to any appended suffix.
6968 @item %j@var{suffix}
6969 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6970 writable, and if save-temps is off; otherwise, substitute the name
6971 of a temporary file, just like @samp{%u}. This temporary file is not
6972 meant for communication between processes, but rather as a junk
6975 @item %|@var{suffix}
6976 @itemx %m@var{suffix}
6977 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6978 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6979 all. These are the two most common ways to instruct a program that it
6980 should read from standard input or write to standard output. If you
6981 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6982 construct: see for example @file{f/lang-specs.h}.
6984 @item %.@var{SUFFIX}
6985 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6986 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6987 terminated by the next space or %.
6990 Marks the argument containing or following the @samp{%w} as the
6991 designated output file of this compilation. This puts the argument
6992 into the sequence of arguments that @samp{%o} will substitute later.
6995 Substitutes the names of all the output files, with spaces
6996 automatically placed around them. You should write spaces
6997 around the @samp{%o} as well or the results are undefined.
6998 @samp{%o} is for use in the specs for running the linker.
6999 Input files whose names have no recognized suffix are not compiled
7000 at all, but they are included among the output files, so they will
7004 Substitutes the suffix for object files. Note that this is
7005 handled specially when it immediately follows @samp{%g, %u, or %U},
7006 because of the need for those to form complete file names. The
7007 handling is such that @samp{%O} is treated exactly as if it had already
7008 been substituted, except that @samp{%g, %u, and %U} do not currently
7009 support additional @var{suffix} characters following @samp{%O} as they would
7010 following, for example, @samp{.o}.
7013 Substitutes the standard macro predefinitions for the
7014 current target machine. Use this when running @code{cpp}.
7017 Like @samp{%p}, but puts @samp{__} before and after the name of each
7018 predefined macro, except for macros that start with @samp{__} or with
7019 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
7023 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
7024 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}),
7025 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
7026 and @option{-imultilib} as necessary.
7029 Current argument is the name of a library or startup file of some sort.
7030 Search for that file in a standard list of directories and substitute
7031 the full name found.
7034 Print @var{str} as an error message. @var{str} is terminated by a newline.
7035 Use this when inconsistent options are detected.
7038 Substitute the contents of spec string @var{name} at this point.
7041 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
7043 @item %x@{@var{option}@}
7044 Accumulate an option for @samp{%X}.
7047 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
7051 Output the accumulated assembler options specified by @option{-Wa}.
7054 Output the accumulated preprocessor options specified by @option{-Wp}.
7057 Process the @code{asm} spec. This is used to compute the
7058 switches to be passed to the assembler.
7061 Process the @code{asm_final} spec. This is a spec string for
7062 passing switches to an assembler post-processor, if such a program is
7066 Process the @code{link} spec. This is the spec for computing the
7067 command line passed to the linker. Typically it will make use of the
7068 @samp{%L %G %S %D and %E} sequences.
7071 Dump out a @option{-L} option for each directory that GCC believes might
7072 contain startup files. If the target supports multilibs then the
7073 current multilib directory will be prepended to each of these paths.
7076 Process the @code{lib} spec. This is a spec string for deciding which
7077 libraries should be included on the command line to the linker.
7080 Process the @code{libgcc} spec. This is a spec string for deciding
7081 which GCC support library should be included on the command line to the linker.
7084 Process the @code{startfile} spec. This is a spec for deciding which
7085 object files should be the first ones passed to the linker. Typically
7086 this might be a file named @file{crt0.o}.
7089 Process the @code{endfile} spec. This is a spec string that specifies
7090 the last object files that will be passed to the linker.
7093 Process the @code{cpp} spec. This is used to construct the arguments
7094 to be passed to the C preprocessor.
7097 Process the @code{cc1} spec. This is used to construct the options to be
7098 passed to the actual C compiler (@samp{cc1}).
7101 Process the @code{cc1plus} spec. This is used to construct the options to be
7102 passed to the actual C++ compiler (@samp{cc1plus}).
7105 Substitute the variable part of a matched option. See below.
7106 Note that each comma in the substituted string is replaced by
7110 Remove all occurrences of @code{-S} from the command line. Note---this
7111 command is position dependent. @samp{%} commands in the spec string
7112 before this one will see @code{-S}, @samp{%} commands in the spec string
7113 after this one will not.
7115 @item %:@var{function}(@var{args})
7116 Call the named function @var{function}, passing it @var{args}.
7117 @var{args} is first processed as a nested spec string, then split
7118 into an argument vector in the usual fashion. The function returns
7119 a string which is processed as if it had appeared literally as part
7120 of the current spec.
7122 The following built-in spec functions are provided:
7125 @item @code{if-exists}
7126 The @code{if-exists} spec function takes one argument, an absolute
7127 pathname to a file. If the file exists, @code{if-exists} returns the
7128 pathname. Here is a small example of its usage:
7132 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
7135 @item @code{if-exists-else}
7136 The @code{if-exists-else} spec function is similar to the @code{if-exists}
7137 spec function, except that it takes two arguments. The first argument is
7138 an absolute pathname to a file. If the file exists, @code{if-exists-else}
7139 returns the pathname. If it does not exist, it returns the second argument.
7140 This way, @code{if-exists-else} can be used to select one file or another,
7141 based on the existence of the first. Here is a small example of its usage:
7145 crt0%O%s %:if-exists(crti%O%s) \
7146 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
7149 @item @code{replace-outfile}
7150 The @code{replace-outfile} spec function takes two arguments. It looks for the
7151 first argument in the outfiles array and replaces it with the second argument. Here
7152 is a small example of its usage:
7155 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
7161 Substitutes the @code{-S} switch, if that switch was given to GCC@.
7162 If that switch was not specified, this substitutes nothing. Note that
7163 the leading dash is omitted when specifying this option, and it is
7164 automatically inserted if the substitution is performed. Thus the spec
7165 string @samp{%@{foo@}} would match the command-line option @option{-foo}
7166 and would output the command line option @option{-foo}.
7168 @item %W@{@code{S}@}
7169 Like %@{@code{S}@} but mark last argument supplied within as a file to be
7172 @item %@{@code{S}*@}
7173 Substitutes all the switches specified to GCC whose names start
7174 with @code{-S}, but which also take an argument. This is used for
7175 switches like @option{-o}, @option{-D}, @option{-I}, etc.
7176 GCC considers @option{-o foo} as being
7177 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
7178 text, including the space. Thus two arguments would be generated.
7180 @item %@{@code{S}*&@code{T}*@}
7181 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
7182 (the order of @code{S} and @code{T} in the spec is not significant).
7183 There can be any number of ampersand-separated variables; for each the
7184 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
7186 @item %@{@code{S}:@code{X}@}
7187 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
7189 @item %@{!@code{S}:@code{X}@}
7190 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
7192 @item %@{@code{S}*:@code{X}@}
7193 Substitutes @code{X} if one or more switches whose names start with
7194 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
7195 once, no matter how many such switches appeared. However, if @code{%*}
7196 appears somewhere in @code{X}, then @code{X} will be substituted once
7197 for each matching switch, with the @code{%*} replaced by the part of
7198 that switch that matched the @code{*}.
7200 @item %@{.@code{S}:@code{X}@}
7201 Substitutes @code{X}, if processing a file with suffix @code{S}.
7203 @item %@{!.@code{S}:@code{X}@}
7204 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
7206 @item %@{@code{S}|@code{P}:@code{X}@}
7207 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
7208 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
7209 although they have a stronger binding than the @samp{|}. If @code{%*}
7210 appears in @code{X}, all of the alternatives must be starred, and only
7211 the first matching alternative is substituted.
7213 For example, a spec string like this:
7216 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
7219 will output the following command-line options from the following input
7220 command-line options:
7225 -d fred.c -foo -baz -boggle
7226 -d jim.d -bar -baz -boggle
7229 @item %@{S:X; T:Y; :D@}
7231 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7232 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7233 be as many clauses as you need. This may be combined with @code{.},
7234 @code{!}, @code{|}, and @code{*} as needed.
7239 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7240 construct may contain other nested @samp{%} constructs or spaces, or
7241 even newlines. They are processed as usual, as described above.
7242 Trailing white space in @code{X} is ignored. White space may also
7243 appear anywhere on the left side of the colon in these constructs,
7244 except between @code{.} or @code{*} and the corresponding word.
7246 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7247 handled specifically in these constructs. If another value of
7248 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7249 @option{-W} switch is found later in the command line, the earlier
7250 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7251 just one letter, which passes all matching options.
7253 The character @samp{|} at the beginning of the predicate text is used to
7254 indicate that a command should be piped to the following command, but
7255 only if @option{-pipe} is specified.
7257 It is built into GCC which switches take arguments and which do not.
7258 (You might think it would be useful to generalize this to allow each
7259 compiler's spec to say which switches take arguments. But this cannot
7260 be done in a consistent fashion. GCC cannot even decide which input
7261 files have been specified without knowing which switches take arguments,
7262 and it must know which input files to compile in order to tell which
7265 GCC also knows implicitly that arguments starting in @option{-l} are to be
7266 treated as compiler output files, and passed to the linker in their
7267 proper position among the other output files.
7269 @c man begin OPTIONS
7271 @node Target Options
7272 @section Specifying Target Machine and Compiler Version
7273 @cindex target options
7274 @cindex cross compiling
7275 @cindex specifying machine version
7276 @cindex specifying compiler version and target machine
7277 @cindex compiler version, specifying
7278 @cindex target machine, specifying
7280 The usual way to run GCC is to run the executable called @file{gcc}, or
7281 @file{<machine>-gcc} when cross-compiling, or
7282 @file{<machine>-gcc-<version>} to run a version other than the one that
7283 was installed last. Sometimes this is inconvenient, so GCC provides
7284 options that will switch to another cross-compiler or version.
7287 @item -b @var{machine}
7289 The argument @var{machine} specifies the target machine for compilation.
7291 The value to use for @var{machine} is the same as was specified as the
7292 machine type when configuring GCC as a cross-compiler. For
7293 example, if a cross-compiler was configured with @samp{configure
7294 arm-elf}, meaning to compile for an arm processor with elf binaries,
7295 then you would specify @option{-b arm-elf} to run that cross compiler.
7296 Because there are other options beginning with @option{-b}, the
7297 configuration must contain a hyphen.
7299 @item -V @var{version}
7301 The argument @var{version} specifies which version of GCC to run.
7302 This is useful when multiple versions are installed. For example,
7303 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7306 The @option{-V} and @option{-b} options work by running the
7307 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7308 use them if you can just run that directly.
7310 @node Submodel Options
7311 @section Hardware Models and Configurations
7312 @cindex submodel options
7313 @cindex specifying hardware config
7314 @cindex hardware models and configurations, specifying
7315 @cindex machine dependent options
7317 Earlier we discussed the standard option @option{-b} which chooses among
7318 different installed compilers for completely different target
7319 machines, such as VAX vs.@: 68000 vs.@: 80386.
7321 In addition, each of these target machine types can have its own
7322 special options, starting with @samp{-m}, to choose among various
7323 hardware models or configurations---for example, 68010 vs 68020,
7324 floating coprocessor or none. A single installed version of the
7325 compiler can compile for any model or configuration, according to the
7328 Some configurations of the compiler also support additional special
7329 options, usually for compatibility with other compilers on the same
7332 @c This list is ordered alphanumerically by subsection name.
7333 @c It should be the same order and spelling as these options are listed
7334 @c in Machine Dependent Options
7340 * Blackfin Options::
7344 * DEC Alpha Options::
7345 * DEC Alpha/VMS Options::
7347 * GNU/Linux Options::
7350 * i386 and x86-64 Options::
7363 * RS/6000 and PowerPC Options::
7364 * S/390 and zSeries Options::
7369 * System V Options::
7370 * TMS320C3x/C4x Options::
7374 * Xstormy16 Options::
7380 @subsection ARC Options
7383 These options are defined for ARC implementations:
7388 Compile code for little endian mode. This is the default.
7392 Compile code for big endian mode.
7395 @opindex mmangle-cpu
7396 Prepend the name of the cpu to all public symbol names.
7397 In multiple-processor systems, there are many ARC variants with different
7398 instruction and register set characteristics. This flag prevents code
7399 compiled for one cpu to be linked with code compiled for another.
7400 No facility exists for handling variants that are ``almost identical''.
7401 This is an all or nothing option.
7403 @item -mcpu=@var{cpu}
7405 Compile code for ARC variant @var{cpu}.
7406 Which variants are supported depend on the configuration.
7407 All variants support @option{-mcpu=base}, this is the default.
7409 @item -mtext=@var{text-section}
7410 @itemx -mdata=@var{data-section}
7411 @itemx -mrodata=@var{readonly-data-section}
7415 Put functions, data, and readonly data in @var{text-section},
7416 @var{data-section}, and @var{readonly-data-section} respectively
7417 by default. This can be overridden with the @code{section} attribute.
7418 @xref{Variable Attributes}.
7423 @subsection ARM Options
7426 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7430 @item -mabi=@var{name}
7432 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7433 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7436 @opindex mapcs-frame
7437 Generate a stack frame that is compliant with the ARM Procedure Call
7438 Standard for all functions, even if this is not strictly necessary for
7439 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7440 with this option will cause the stack frames not to be generated for
7441 leaf functions. The default is @option{-mno-apcs-frame}.
7445 This is a synonym for @option{-mapcs-frame}.
7448 @c not currently implemented
7449 @item -mapcs-stack-check
7450 @opindex mapcs-stack-check
7451 Generate code to check the amount of stack space available upon entry to
7452 every function (that actually uses some stack space). If there is
7453 insufficient space available then either the function
7454 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7455 called, depending upon the amount of stack space required. The run time
7456 system is required to provide these functions. The default is
7457 @option{-mno-apcs-stack-check}, since this produces smaller code.
7459 @c not currently implemented
7461 @opindex mapcs-float
7462 Pass floating point arguments using the float point registers. This is
7463 one of the variants of the APCS@. This option is recommended if the
7464 target hardware has a floating point unit or if a lot of floating point
7465 arithmetic is going to be performed by the code. The default is
7466 @option{-mno-apcs-float}, since integer only code is slightly increased in
7467 size if @option{-mapcs-float} is used.
7469 @c not currently implemented
7470 @item -mapcs-reentrant
7471 @opindex mapcs-reentrant
7472 Generate reentrant, position independent code. The default is
7473 @option{-mno-apcs-reentrant}.
7476 @item -mthumb-interwork
7477 @opindex mthumb-interwork
7478 Generate code which supports calling between the ARM and Thumb
7479 instruction sets. Without this option the two instruction sets cannot
7480 be reliably used inside one program. The default is
7481 @option{-mno-thumb-interwork}, since slightly larger code is generated
7482 when @option{-mthumb-interwork} is specified.
7484 @item -mno-sched-prolog
7485 @opindex mno-sched-prolog
7486 Prevent the reordering of instructions in the function prolog, or the
7487 merging of those instruction with the instructions in the function's
7488 body. This means that all functions will start with a recognizable set
7489 of instructions (or in fact one of a choice from a small set of
7490 different function prologues), and this information can be used to
7491 locate the start if functions inside an executable piece of code. The
7492 default is @option{-msched-prolog}.
7495 @opindex mhard-float
7496 Generate output containing floating point instructions. This is the
7500 @opindex msoft-float
7501 Generate output containing library calls for floating point.
7502 @strong{Warning:} the requisite libraries are not available for all ARM
7503 targets. Normally the facilities of the machine's usual C compiler are
7504 used, but this cannot be done directly in cross-compilation. You must make
7505 your own arrangements to provide suitable library functions for
7508 @option{-msoft-float} changes the calling convention in the output file;
7509 therefore, it is only useful if you compile @emph{all} of a program with
7510 this option. In particular, you need to compile @file{libgcc.a}, the
7511 library that comes with GCC, with @option{-msoft-float} in order for
7514 @item -mfloat-abi=@var{name}
7516 Specifies which ABI to use for floating point values. Permissible values
7517 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7519 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7520 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7521 of floating point instructions, but still uses the soft-float calling
7524 @item -mlittle-endian
7525 @opindex mlittle-endian
7526 Generate code for a processor running in little-endian mode. This is
7527 the default for all standard configurations.
7530 @opindex mbig-endian
7531 Generate code for a processor running in big-endian mode; the default is
7532 to compile code for a little-endian processor.
7534 @item -mwords-little-endian
7535 @opindex mwords-little-endian
7536 This option only applies when generating code for big-endian processors.
7537 Generate code for a little-endian word order but a big-endian byte
7538 order. That is, a byte order of the form @samp{32107654}. Note: this
7539 option should only be used if you require compatibility with code for
7540 big-endian ARM processors generated by versions of the compiler prior to
7543 @item -mcpu=@var{name}
7545 This specifies the name of the target ARM processor. GCC uses this name
7546 to determine what kind of instructions it can emit when generating
7547 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7548 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7549 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7550 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7551 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7552 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7553 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7554 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7555 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7556 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7557 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7558 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7559 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7560 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7563 @itemx -mtune=@var{name}
7565 This option is very similar to the @option{-mcpu=} option, except that
7566 instead of specifying the actual target processor type, and hence
7567 restricting which instructions can be used, it specifies that GCC should
7568 tune the performance of the code as if the target were of the type
7569 specified in this option, but still choosing the instructions that it
7570 will generate based on the cpu specified by a @option{-mcpu=} option.
7571 For some ARM implementations better performance can be obtained by using
7574 @item -march=@var{name}
7576 This specifies the name of the target ARM architecture. GCC uses this
7577 name to determine what kind of instructions it can emit when generating
7578 assembly code. This option can be used in conjunction with or instead
7579 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7580 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7581 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7582 @samp{iwmmxt}, @samp{ep9312}.
7584 @item -mfpu=@var{name}
7585 @itemx -mfpe=@var{number}
7586 @itemx -mfp=@var{number}
7590 This specifies what floating point hardware (or hardware emulation) is
7591 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7592 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7593 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7594 with older versions of GCC@.
7596 If @option{-msoft-float} is specified this specifies the format of
7597 floating point values.
7599 @item -mstructure-size-boundary=@var{n}
7600 @opindex mstructure-size-boundary
7601 The size of all structures and unions will be rounded up to a multiple
7602 of the number of bits set by this option. Permissible values are 8, 32
7603 and 64. The default value varies for different toolchains. For the COFF
7604 targeted toolchain the default value is 8. A value of 64 is only allowed
7605 if the underlying ABI supports it.
7607 Specifying the larger number can produce faster, more efficient code, but
7608 can also increase the size of the program. Different values are potentially
7609 incompatible. Code compiled with one value cannot necessarily expect to
7610 work with code or libraries compiled with another value, if they exchange
7611 information using structures or unions.
7613 @item -mabort-on-noreturn
7614 @opindex mabort-on-noreturn
7615 Generate a call to the function @code{abort} at the end of a
7616 @code{noreturn} function. It will be executed if the function tries to
7620 @itemx -mno-long-calls
7621 @opindex mlong-calls
7622 @opindex mno-long-calls
7623 Tells the compiler to perform function calls by first loading the
7624 address of the function into a register and then performing a subroutine
7625 call on this register. This switch is needed if the target function
7626 will lie outside of the 64 megabyte addressing range of the offset based
7627 version of subroutine call instruction.
7629 Even if this switch is enabled, not all function calls will be turned
7630 into long calls. The heuristic is that static functions, functions
7631 which have the @samp{short-call} attribute, functions that are inside
7632 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7633 definitions have already been compiled within the current compilation
7634 unit, will not be turned into long calls. The exception to this rule is
7635 that weak function definitions, functions with the @samp{long-call}
7636 attribute or the @samp{section} attribute, and functions that are within
7637 the scope of a @samp{#pragma long_calls} directive, will always be
7638 turned into long calls.
7640 This feature is not enabled by default. Specifying
7641 @option{-mno-long-calls} will restore the default behavior, as will
7642 placing the function calls within the scope of a @samp{#pragma
7643 long_calls_off} directive. Note these switches have no effect on how
7644 the compiler generates code to handle function calls via function
7647 @item -mnop-fun-dllimport
7648 @opindex mnop-fun-dllimport
7649 Disable support for the @code{dllimport} attribute.
7651 @item -msingle-pic-base
7652 @opindex msingle-pic-base
7653 Treat the register used for PIC addressing as read-only, rather than
7654 loading it in the prologue for each function. The run-time system is
7655 responsible for initializing this register with an appropriate value
7656 before execution begins.
7658 @item -mpic-register=@var{reg}
7659 @opindex mpic-register
7660 Specify the register to be used for PIC addressing. The default is R10
7661 unless stack-checking is enabled, when R9 is used.
7663 @item -mcirrus-fix-invalid-insns
7664 @opindex mcirrus-fix-invalid-insns
7665 @opindex mno-cirrus-fix-invalid-insns
7666 Insert NOPs into the instruction stream to in order to work around
7667 problems with invalid Maverick instruction combinations. This option
7668 is only valid if the @option{-mcpu=ep9312} option has been used to
7669 enable generation of instructions for the Cirrus Maverick floating
7670 point co-processor. This option is not enabled by default, since the
7671 problem is only present in older Maverick implementations. The default
7672 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7675 @item -mpoke-function-name
7676 @opindex mpoke-function-name
7677 Write the name of each function into the text section, directly
7678 preceding the function prologue. The generated code is similar to this:
7682 .ascii "arm_poke_function_name", 0
7685 .word 0xff000000 + (t1 - t0)
7686 arm_poke_function_name
7688 stmfd sp!, @{fp, ip, lr, pc@}
7692 When performing a stack backtrace, code can inspect the value of
7693 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7694 location @code{pc - 12} and the top 8 bits are set, then we know that
7695 there is a function name embedded immediately preceding this location
7696 and has length @code{((pc[-3]) & 0xff000000)}.
7700 Generate code for the 16-bit Thumb instruction set. The default is to
7701 use the 32-bit ARM instruction set.
7704 @opindex mtpcs-frame
7705 Generate a stack frame that is compliant with the Thumb Procedure Call
7706 Standard for all non-leaf functions. (A leaf function is one that does
7707 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7709 @item -mtpcs-leaf-frame
7710 @opindex mtpcs-leaf-frame
7711 Generate a stack frame that is compliant with the Thumb Procedure Call
7712 Standard for all leaf functions. (A leaf function is one that does
7713 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7715 @item -mcallee-super-interworking
7716 @opindex mcallee-super-interworking
7717 Gives all externally visible functions in the file being compiled an ARM
7718 instruction set header which switches to Thumb mode before executing the
7719 rest of the function. This allows these functions to be called from
7720 non-interworking code.
7722 @item -mcaller-super-interworking
7723 @opindex mcaller-super-interworking
7724 Allows calls via function pointers (including virtual functions) to
7725 execute correctly regardless of whether the target code has been
7726 compiled for interworking or not. There is a small overhead in the cost
7727 of executing a function pointer if this option is enabled.
7729 @item -mtp=@var{name}
7731 Specify the access model for the thread local storage pointer. The valid
7732 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7733 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7734 (supported in the arm6k architecture), and @option{auto}, which uses the
7735 best available method for the selected processor. The default setting is
7741 @subsection AVR Options
7744 These options are defined for AVR implementations:
7747 @item -mmcu=@var{mcu}
7749 Specify ATMEL AVR instruction set or MCU type.
7751 Instruction set avr1 is for the minimal AVR core, not supported by the C
7752 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7753 attiny11, attiny12, attiny15, attiny28).
7755 Instruction set avr2 (default) is for the classic AVR core with up to
7756 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7757 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7758 at90c8534, at90s8535).
7760 Instruction set avr3 is for the classic AVR core with up to 128K program
7761 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7763 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7764 memory space (MCU types: atmega8, atmega83, atmega85).
7766 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7767 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7768 atmega64, atmega128, at43usb355, at94k).
7772 Output instruction sizes to the asm file.
7774 @item -minit-stack=@var{N}
7775 @opindex minit-stack
7776 Specify the initial stack address, which may be a symbol or numeric value,
7777 @samp{__stack} is the default.
7779 @item -mno-interrupts
7780 @opindex mno-interrupts
7781 Generated code is not compatible with hardware interrupts.
7782 Code size will be smaller.
7784 @item -mcall-prologues
7785 @opindex mcall-prologues
7786 Functions prologues/epilogues expanded as call to appropriate
7787 subroutines. Code size will be smaller.
7789 @item -mno-tablejump
7790 @opindex mno-tablejump
7791 Do not generate tablejump insns which sometimes increase code size.
7794 @opindex mtiny-stack
7795 Change only the low 8 bits of the stack pointer.
7799 Assume int to be 8 bit integer. This affects the sizes of all types: A
7800 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7801 and long long will be 4 bytes. Please note that this option does not
7802 comply to the C standards, but it will provide you with smaller code
7806 @node Blackfin Options
7807 @subsection Blackfin Options
7808 @cindex Blackfin Options
7811 @item -momit-leaf-frame-pointer
7812 @opindex momit-leaf-frame-pointer
7813 Don't keep the frame pointer in a register for leaf functions. This
7814 avoids the instructions to save, set up and restore frame pointers and
7815 makes an extra register available in leaf functions. The option
7816 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7817 which might make debugging harder.
7819 @item -mspecld-anomaly
7820 @opindex mspecld-anomaly
7821 When enabled, the compiler will ensure that the generated code does not
7822 contain speculative loads after jump instructions. This option is enabled
7825 @item -mno-specld-anomaly
7826 @opindex mno-specld-anomaly
7827 Don't generate extra code to prevent speculative loads from occurring.
7829 @item -mcsync-anomaly
7830 @opindex mcsync-anomaly
7831 When enabled, the compiler will ensure that the generated code does not
7832 contain CSYNC or SSYNC instructions too soon after conditional branches.
7833 This option is enabled by default.
7835 @item -mno-csync-anomaly
7836 @opindex mno-csync-anomaly
7837 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7838 occurring too soon after a conditional branch.
7842 When enabled, the compiler is free to take advantage of the knowledge that
7843 the entire program fits into the low 64k of memory.
7846 @opindex mno-low-64k
7847 Assume that the program is arbitrarily large. This is the default.
7849 @item -mstack-check-l1
7850 @opindex mstack-check-l1
7851 Do stack checking using information placed into L1 scratchpad memory by the
7854 @item -mid-shared-library
7855 @opindex mid-shared-library
7856 Generate code that supports shared libraries via the library ID method.
7857 This allows for execute in place and shared libraries in an environment
7858 without virtual memory management. This option implies @option{-fPIC}.
7860 @item -mno-id-shared-library
7861 @opindex mno-id-shared-library
7862 Generate code that doesn't assume ID based shared libraries are being used.
7863 This is the default.
7865 @item -mleaf-id-shared-library
7866 @opindex mleaf-id-shared-library
7867 Generate code that supports shared libraries via the library ID method,
7868 but assumes that this library or executable won't link against any other
7869 ID shared libraries. That allows the compiler to use faster code for jumps
7872 @item -mno-leaf-id-shared-library
7873 @opindex mno-leaf-id-shared-library
7874 Do not assume that the code being compiled won't link against any ID shared
7875 libraries. Slower code will be generated for jump and call insns.
7877 @item -mshared-library-id=n
7878 @opindex mshared-library-id
7879 Specified the identification number of the ID based shared library being
7880 compiled. Specifying a value of 0 will generate more compact code, specifying
7881 other values will force the allocation of that number to the current
7882 library but is no more space or time efficient than omitting this option.
7886 Generate code that allows the data segment to be located in a different
7887 area of memory from the text segment. This allows for execute in place in
7888 an environment without virtual memory management by eliminating relocations
7889 against the text section.
7892 @opindex mno-sep-data
7893 Generate code that assumes that the data segment follows the text segment.
7894 This is the default.
7897 @itemx -mno-long-calls
7898 @opindex mlong-calls
7899 @opindex mno-long-calls
7900 Tells the compiler to perform function calls by first loading the
7901 address of the function into a register and then performing a subroutine
7902 call on this register. This switch is needed if the target function
7903 will lie outside of the 24 bit addressing range of the offset based
7904 version of subroutine call instruction.
7906 This feature is not enabled by default. Specifying
7907 @option{-mno-long-calls} will restore the default behavior. Note these
7908 switches have no effect on how the compiler generates code to handle
7909 function calls via function pointers.
7913 @subsection CRIS Options
7914 @cindex CRIS Options
7916 These options are defined specifically for the CRIS ports.
7919 @item -march=@var{architecture-type}
7920 @itemx -mcpu=@var{architecture-type}
7923 Generate code for the specified architecture. The choices for
7924 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7925 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7926 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7929 @item -mtune=@var{architecture-type}
7931 Tune to @var{architecture-type} everything applicable about the generated
7932 code, except for the ABI and the set of available instructions. The
7933 choices for @var{architecture-type} are the same as for
7934 @option{-march=@var{architecture-type}}.
7936 @item -mmax-stack-frame=@var{n}
7937 @opindex mmax-stack-frame
7938 Warn when the stack frame of a function exceeds @var{n} bytes.
7940 @item -melinux-stacksize=@var{n}
7941 @opindex melinux-stacksize
7942 Only available with the @samp{cris-axis-aout} target. Arranges for
7943 indications in the program to the kernel loader that the stack of the
7944 program should be set to @var{n} bytes.
7950 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7951 @option{-march=v3} and @option{-march=v8} respectively.
7953 @item -mmul-bug-workaround
7954 @itemx -mno-mul-bug-workaround
7955 @opindex mmul-bug-workaround
7956 @opindex mno-mul-bug-workaround
7957 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7958 models where it applies. This option is active by default.
7962 Enable CRIS-specific verbose debug-related information in the assembly
7963 code. This option also has the effect to turn off the @samp{#NO_APP}
7964 formatted-code indicator to the assembler at the beginning of the
7969 Do not use condition-code results from previous instruction; always emit
7970 compare and test instructions before use of condition codes.
7972 @item -mno-side-effects
7973 @opindex mno-side-effects
7974 Do not emit instructions with side-effects in addressing modes other than
7978 @itemx -mno-stack-align
7980 @itemx -mno-data-align
7981 @itemx -mconst-align
7982 @itemx -mno-const-align
7983 @opindex mstack-align
7984 @opindex mno-stack-align
7985 @opindex mdata-align
7986 @opindex mno-data-align
7987 @opindex mconst-align
7988 @opindex mno-const-align
7989 These options (no-options) arranges (eliminate arrangements) for the
7990 stack-frame, individual data and constants to be aligned for the maximum
7991 single data access size for the chosen CPU model. The default is to
7992 arrange for 32-bit alignment. ABI details such as structure layout are
7993 not affected by these options.
8001 Similar to the stack- data- and const-align options above, these options
8002 arrange for stack-frame, writable data and constants to all be 32-bit,
8003 16-bit or 8-bit aligned. The default is 32-bit alignment.
8005 @item -mno-prologue-epilogue
8006 @itemx -mprologue-epilogue
8007 @opindex mno-prologue-epilogue
8008 @opindex mprologue-epilogue
8009 With @option{-mno-prologue-epilogue}, the normal function prologue and
8010 epilogue that sets up the stack-frame are omitted and no return
8011 instructions or return sequences are generated in the code. Use this
8012 option only together with visual inspection of the compiled code: no
8013 warnings or errors are generated when call-saved registers must be saved,
8014 or storage for local variable needs to be allocated.
8020 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
8021 instruction sequences that load addresses for functions from the PLT part
8022 of the GOT rather than (traditional on other architectures) calls to the
8023 PLT@. The default is @option{-mgotplt}.
8027 Legacy no-op option only recognized with the cris-axis-aout target.
8031 Legacy no-op option only recognized with the cris-axis-elf and
8032 cris-axis-linux-gnu targets.
8036 Only recognized with the cris-axis-aout target, where it selects a
8037 GNU/linux-like multilib, include files and instruction set for
8042 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
8046 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
8047 to link with input-output functions from a simulator library. Code,
8048 initialized data and zero-initialized data are allocated consecutively.
8052 Like @option{-sim}, but pass linker options to locate initialized data at
8053 0x40000000 and zero-initialized data at 0x80000000.
8057 @subsection CRX Options
8060 These options are defined specifically for the CRX ports.
8066 Enable the use of multiply-accumulate instructions. Disabled by default.
8070 Push instructions will be used to pass outgoing arguments when functions
8071 are called. Enabled by default.
8074 @node Darwin Options
8075 @subsection Darwin Options
8076 @cindex Darwin options
8078 These options are defined for all architectures running the Darwin operating
8081 FSF GCC on Darwin does not create ``fat'' object files; it will create
8082 an object file for the single architecture that it was built to
8083 target. Apple's GCC on Darwin does create ``fat'' files if multiple
8084 @option{-arch} options are used; it does so by running the compiler or
8085 linker multiple times and joining the results together with
8088 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
8089 @samp{i686}) is determined by the flags that specify the ISA
8090 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
8091 @option{-force_cpusubtype_ALL} option can be used to override this.
8093 The Darwin tools vary in their behavior when presented with an ISA
8094 mismatch. The assembler, @file{as}, will only permit instructions to
8095 be used that are valid for the subtype of the file it is generating,
8096 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
8097 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
8098 and print an error if asked to create a shared library with a less
8099 restrictive subtype than its input files (for instance, trying to put
8100 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
8101 for executables, @file{ld}, will quietly give the executable the most
8102 restrictive subtype of any of its input files.
8107 Add the framework directory @var{dir} to the head of the list of
8108 directories to be searched for header files. These directories are
8109 interleaved with those specified by @option{-I} options and are
8110 scanned in a left-to-right order.
8112 A framework directory is a directory with frameworks in it. A
8113 framework is a directory with a @samp{"Headers"} and/or
8114 @samp{"PrivateHeaders"} directory contained directly in it that ends
8115 in @samp{".framework"}. The name of a framework is the name of this
8116 directory excluding the @samp{".framework"}. Headers associated with
8117 the framework are found in one of those two directories, with
8118 @samp{"Headers"} being searched first. A subframework is a framework
8119 directory that is in a framework's @samp{"Frameworks"} directory.
8120 Includes of subframework headers can only appear in a header of a
8121 framework that contains the subframework, or in a sibling subframework
8122 header. Two subframeworks are siblings if they occur in the same
8123 framework. A subframework should not have the same name as a
8124 framework, a warning will be issued if this is violated. Currently a
8125 subframework cannot have subframeworks, in the future, the mechanism
8126 may be extended to support this. The standard frameworks can be found
8127 in @samp{"/System/Library/Frameworks"} and
8128 @samp{"/Library/Frameworks"}. An example include looks like
8129 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
8130 the name of the framework and header.h is found in the
8131 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
8135 Emit debugging information for symbols that are used. For STABS
8136 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
8137 This is by default ON@.
8141 Emit debugging information for all symbols and types.
8143 @item -mmacosx-version-min=@var{version}
8144 The earliest version of MacOS X that this executable will run on
8145 is @var{version}. Typical values of @var{version} include @code{10.1},
8146 @code{10.2}, and @code{10.3.9}.
8148 The default for this option is to make choices that seem to be most
8153 Enable kernel development mode. The @option{-mkernel} option sets
8154 @option{-static}, @option{-fno-common}, @option{-fno-cxa-atexit},
8155 @option{-fno-exceptions}, @option{-fno-non-call-exceptions},
8156 @option{-fapple-kext}, @option{-fno-weak} and @option{-fno-rtti} where
8157 applicable. This mode also sets @option{-mno-altivec},
8158 @option{-msoft-float}, @option{-fno-builtin} and
8159 @option{-mlong-branch} for PowerPC targets.
8161 @item -mone-byte-bool
8162 @opindex -mone-byte-bool
8163 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
8164 By default @samp{sizeof(bool)} is @samp{4} when compiling for
8165 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
8166 option has no effect on x86.
8168 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
8169 to generate code that is not binary compatible with code generated
8170 without that switch. Using this switch may require recompiling all
8171 other modules in a program, including system libraries. Use this
8172 switch to conform to a non-default data model.
8174 @item -mfix-and-continue
8175 @itemx -ffix-and-continue
8176 @itemx -findirect-data
8177 @opindex mfix-and-continue
8178 @opindex ffix-and-continue
8179 @opindex findirect-data
8180 Generate code suitable for fast turn around development. Needed to
8181 enable gdb to dynamically load @code{.o} files into already running
8182 programs. @option{-findirect-data} and @option{-ffix-and-continue}
8183 are provided for backwards compatibility.
8187 Loads all members of static archive libraries.
8188 See man ld(1) for more information.
8190 @item -arch_errors_fatal
8191 @opindex arch_errors_fatal
8192 Cause the errors having to do with files that have the wrong architecture
8196 @opindex bind_at_load
8197 Causes the output file to be marked such that the dynamic linker will
8198 bind all undefined references when the file is loaded or launched.
8202 Produce a Mach-o bundle format file.
8203 See man ld(1) for more information.
8205 @item -bundle_loader @var{executable}
8206 @opindex bundle_loader
8207 This option specifies the @var{executable} that will be loading the build
8208 output file being linked. See man ld(1) for more information.
8211 @opindex -dynamiclib
8212 When passed this option, GCC will produce a dynamic library instead of
8213 an executable when linking, using the Darwin @file{libtool} command.
8215 @item -force_cpusubtype_ALL
8216 @opindex -force_cpusubtype_ALL
8217 This causes GCC's output file to have the @var{ALL} subtype, instead of
8218 one controlled by the @option{-mcpu} or @option{-march} option.
8220 @item -allowable_client @var{client_name}
8222 @itemx -compatibility_version
8223 @itemx -current_version
8225 @itemx -dependency-file
8227 @itemx -dylinker_install_name
8229 @itemx -exported_symbols_list
8231 @itemx -flat_namespace
8232 @itemx -force_flat_namespace
8233 @itemx -headerpad_max_install_names
8236 @itemx -install_name
8237 @itemx -keep_private_externs
8238 @itemx -multi_module
8239 @itemx -multiply_defined
8240 @itemx -multiply_defined_unused
8242 @itemx -no_dead_strip_inits_and_terms
8243 @itemx -nofixprebinding
8246 @itemx -noseglinkedit
8247 @itemx -pagezero_size
8249 @itemx -prebind_all_twolevel_modules
8250 @itemx -private_bundle
8251 @itemx -read_only_relocs
8253 @itemx -sectobjectsymbols
8257 @itemx -sectobjectsymbols
8260 @itemx -segs_read_only_addr
8261 @itemx -segs_read_write_addr
8262 @itemx -seg_addr_table
8263 @itemx -seg_addr_table_filename
8266 @itemx -segs_read_only_addr
8267 @itemx -segs_read_write_addr
8268 @itemx -single_module
8271 @itemx -sub_umbrella
8272 @itemx -twolevel_namespace
8275 @itemx -unexported_symbols_list
8276 @itemx -weak_reference_mismatches
8279 @opindex allowable_client
8280 @opindex client_name
8281 @opindex compatibility_version
8282 @opindex current_version
8284 @opindex dependency-file
8286 @opindex dylinker_install_name
8288 @opindex exported_symbols_list
8290 @opindex flat_namespace
8291 @opindex force_flat_namespace
8292 @opindex headerpad_max_install_names
8295 @opindex install_name
8296 @opindex keep_private_externs
8297 @opindex multi_module
8298 @opindex multiply_defined
8299 @opindex multiply_defined_unused
8301 @opindex no_dead_strip_inits_and_terms
8302 @opindex nofixprebinding
8303 @opindex nomultidefs
8305 @opindex noseglinkedit
8306 @opindex pagezero_size
8308 @opindex prebind_all_twolevel_modules
8309 @opindex private_bundle
8310 @opindex read_only_relocs
8312 @opindex sectobjectsymbols
8316 @opindex sectobjectsymbols
8319 @opindex segs_read_only_addr
8320 @opindex segs_read_write_addr
8321 @opindex seg_addr_table
8322 @opindex seg_addr_table_filename
8323 @opindex seglinkedit
8325 @opindex segs_read_only_addr
8326 @opindex segs_read_write_addr
8327 @opindex single_module
8329 @opindex sub_library
8330 @opindex sub_umbrella
8331 @opindex twolevel_namespace
8334 @opindex unexported_symbols_list
8335 @opindex weak_reference_mismatches
8336 @opindex whatsloaded
8338 These options are passed to the Darwin linker. The Darwin linker man page
8339 describes them in detail.
8342 @node DEC Alpha Options
8343 @subsection DEC Alpha Options
8345 These @samp{-m} options are defined for the DEC Alpha implementations:
8348 @item -mno-soft-float
8350 @opindex mno-soft-float
8351 @opindex msoft-float
8352 Use (do not use) the hardware floating-point instructions for
8353 floating-point operations. When @option{-msoft-float} is specified,
8354 functions in @file{libgcc.a} will be used to perform floating-point
8355 operations. Unless they are replaced by routines that emulate the
8356 floating-point operations, or compiled in such a way as to call such
8357 emulations routines, these routines will issue floating-point
8358 operations. If you are compiling for an Alpha without floating-point
8359 operations, you must ensure that the library is built so as not to call
8362 Note that Alpha implementations without floating-point operations are
8363 required to have floating-point registers.
8368 @opindex mno-fp-regs
8369 Generate code that uses (does not use) the floating-point register set.
8370 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8371 register set is not used, floating point operands are passed in integer
8372 registers as if they were integers and floating-point results are passed
8373 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8374 so any function with a floating-point argument or return value called by code
8375 compiled with @option{-mno-fp-regs} must also be compiled with that
8378 A typical use of this option is building a kernel that does not use,
8379 and hence need not save and restore, any floating-point registers.
8383 The Alpha architecture implements floating-point hardware optimized for
8384 maximum performance. It is mostly compliant with the IEEE floating
8385 point standard. However, for full compliance, software assistance is
8386 required. This option generates code fully IEEE compliant code
8387 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8388 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8389 defined during compilation. The resulting code is less efficient but is
8390 able to correctly support denormalized numbers and exceptional IEEE
8391 values such as not-a-number and plus/minus infinity. Other Alpha
8392 compilers call this option @option{-ieee_with_no_inexact}.
8394 @item -mieee-with-inexact
8395 @opindex mieee-with-inexact
8396 This is like @option{-mieee} except the generated code also maintains
8397 the IEEE @var{inexact-flag}. Turning on this option causes the
8398 generated code to implement fully-compliant IEEE math. In addition to
8399 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8400 macro. On some Alpha implementations the resulting code may execute
8401 significantly slower than the code generated by default. Since there is
8402 very little code that depends on the @var{inexact-flag}, you should
8403 normally not specify this option. Other Alpha compilers call this
8404 option @option{-ieee_with_inexact}.
8406 @item -mfp-trap-mode=@var{trap-mode}
8407 @opindex mfp-trap-mode
8408 This option controls what floating-point related traps are enabled.
8409 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8410 The trap mode can be set to one of four values:
8414 This is the default (normal) setting. The only traps that are enabled
8415 are the ones that cannot be disabled in software (e.g., division by zero
8419 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8423 Like @samp{u}, but the instructions are marked to be safe for software
8424 completion (see Alpha architecture manual for details).
8427 Like @samp{su}, but inexact traps are enabled as well.
8430 @item -mfp-rounding-mode=@var{rounding-mode}
8431 @opindex mfp-rounding-mode
8432 Selects the IEEE rounding mode. Other Alpha compilers call this option
8433 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8438 Normal IEEE rounding mode. Floating point numbers are rounded towards
8439 the nearest machine number or towards the even machine number in case
8443 Round towards minus infinity.
8446 Chopped rounding mode. Floating point numbers are rounded towards zero.
8449 Dynamic rounding mode. A field in the floating point control register
8450 (@var{fpcr}, see Alpha architecture reference manual) controls the
8451 rounding mode in effect. The C library initializes this register for
8452 rounding towards plus infinity. Thus, unless your program modifies the
8453 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8456 @item -mtrap-precision=@var{trap-precision}
8457 @opindex mtrap-precision
8458 In the Alpha architecture, floating point traps are imprecise. This
8459 means without software assistance it is impossible to recover from a
8460 floating trap and program execution normally needs to be terminated.
8461 GCC can generate code that can assist operating system trap handlers
8462 in determining the exact location that caused a floating point trap.
8463 Depending on the requirements of an application, different levels of
8464 precisions can be selected:
8468 Program precision. This option is the default and means a trap handler
8469 can only identify which program caused a floating point exception.
8472 Function precision. The trap handler can determine the function that
8473 caused a floating point exception.
8476 Instruction precision. The trap handler can determine the exact
8477 instruction that caused a floating point exception.
8480 Other Alpha compilers provide the equivalent options called
8481 @option{-scope_safe} and @option{-resumption_safe}.
8483 @item -mieee-conformant
8484 @opindex mieee-conformant
8485 This option marks the generated code as IEEE conformant. You must not
8486 use this option unless you also specify @option{-mtrap-precision=i} and either
8487 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8488 is to emit the line @samp{.eflag 48} in the function prologue of the
8489 generated assembly file. Under DEC Unix, this has the effect that
8490 IEEE-conformant math library routines will be linked in.
8492 @item -mbuild-constants
8493 @opindex mbuild-constants
8494 Normally GCC examines a 32- or 64-bit integer constant to
8495 see if it can construct it from smaller constants in two or three
8496 instructions. If it cannot, it will output the constant as a literal and
8497 generate code to load it from the data segment at runtime.
8499 Use this option to require GCC to construct @emph{all} integer constants
8500 using code, even if it takes more instructions (the maximum is six).
8502 You would typically use this option to build a shared library dynamic
8503 loader. Itself a shared library, it must relocate itself in memory
8504 before it can find the variables and constants in its own data segment.
8510 Select whether to generate code to be assembled by the vendor-supplied
8511 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8529 Indicate whether GCC should generate code to use the optional BWX,
8530 CIX, FIX and MAX instruction sets. The default is to use the instruction
8531 sets supported by the CPU type specified via @option{-mcpu=} option or that
8532 of the CPU on which GCC was built if none was specified.
8537 @opindex mfloat-ieee
8538 Generate code that uses (does not use) VAX F and G floating point
8539 arithmetic instead of IEEE single and double precision.
8541 @item -mexplicit-relocs
8542 @itemx -mno-explicit-relocs
8543 @opindex mexplicit-relocs
8544 @opindex mno-explicit-relocs
8545 Older Alpha assemblers provided no way to generate symbol relocations
8546 except via assembler macros. Use of these macros does not allow
8547 optimal instruction scheduling. GNU binutils as of version 2.12
8548 supports a new syntax that allows the compiler to explicitly mark
8549 which relocations should apply to which instructions. This option
8550 is mostly useful for debugging, as GCC detects the capabilities of
8551 the assembler when it is built and sets the default accordingly.
8555 @opindex msmall-data
8556 @opindex mlarge-data
8557 When @option{-mexplicit-relocs} is in effect, static data is
8558 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8559 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8560 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8561 16-bit relocations off of the @code{$gp} register. This limits the
8562 size of the small data area to 64KB, but allows the variables to be
8563 directly accessed via a single instruction.
8565 The default is @option{-mlarge-data}. With this option the data area
8566 is limited to just below 2GB@. Programs that require more than 2GB of
8567 data must use @code{malloc} or @code{mmap} to allocate the data in the
8568 heap instead of in the program's data segment.
8570 When generating code for shared libraries, @option{-fpic} implies
8571 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8575 @opindex msmall-text
8576 @opindex mlarge-text
8577 When @option{-msmall-text} is used, the compiler assumes that the
8578 code of the entire program (or shared library) fits in 4MB, and is
8579 thus reachable with a branch instruction. When @option{-msmall-data}
8580 is used, the compiler can assume that all local symbols share the
8581 same @code{$gp} value, and thus reduce the number of instructions
8582 required for a function call from 4 to 1.
8584 The default is @option{-mlarge-text}.
8586 @item -mcpu=@var{cpu_type}
8588 Set the instruction set and instruction scheduling parameters for
8589 machine type @var{cpu_type}. You can specify either the @samp{EV}
8590 style name or the corresponding chip number. GCC supports scheduling
8591 parameters for the EV4, EV5 and EV6 family of processors and will
8592 choose the default values for the instruction set from the processor
8593 you specify. If you do not specify a processor type, GCC will default
8594 to the processor on which the compiler was built.
8596 Supported values for @var{cpu_type} are
8602 Schedules as an EV4 and has no instruction set extensions.
8606 Schedules as an EV5 and has no instruction set extensions.
8610 Schedules as an EV5 and supports the BWX extension.
8615 Schedules as an EV5 and supports the BWX and MAX extensions.
8619 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8623 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8626 @item -mtune=@var{cpu_type}
8628 Set only the instruction scheduling parameters for machine type
8629 @var{cpu_type}. The instruction set is not changed.
8631 @item -mmemory-latency=@var{time}
8632 @opindex mmemory-latency
8633 Sets the latency the scheduler should assume for typical memory
8634 references as seen by the application. This number is highly
8635 dependent on the memory access patterns used by the application
8636 and the size of the external cache on the machine.
8638 Valid options for @var{time} are
8642 A decimal number representing clock cycles.
8648 The compiler contains estimates of the number of clock cycles for
8649 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8650 (also called Dcache, Scache, and Bcache), as well as to main memory.
8651 Note that L3 is only valid for EV5.
8656 @node DEC Alpha/VMS Options
8657 @subsection DEC Alpha/VMS Options
8659 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8662 @item -mvms-return-codes
8663 @opindex mvms-return-codes
8664 Return VMS condition codes from main. The default is to return POSIX
8665 style condition (e.g.@ error) codes.
8669 @subsection FRV Options
8676 Only use the first 32 general purpose registers.
8681 Use all 64 general purpose registers.
8686 Use only the first 32 floating point registers.
8691 Use all 64 floating point registers
8694 @opindex mhard-float
8696 Use hardware instructions for floating point operations.
8699 @opindex msoft-float
8701 Use library routines for floating point operations.
8706 Dynamically allocate condition code registers.
8711 Do not try to dynamically allocate condition code registers, only
8712 use @code{icc0} and @code{fcc0}.
8717 Change ABI to use double word insns.
8722 Do not use double word instructions.
8727 Use floating point double instructions.
8732 Do not use floating point double instructions.
8737 Use media instructions.
8742 Do not use media instructions.
8747 Use multiply and add/subtract instructions.
8752 Do not use multiply and add/subtract instructions.
8757 Select the FDPIC ABI, that uses function descriptors to represent
8758 pointers to functions. Without any PIC/PIE-related options, it
8759 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8760 assumes GOT entries and small data are within a 12-bit range from the
8761 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8762 are computed with 32 bits.
8765 @opindex minline-plt
8767 Enable inlining of PLT entries in function calls to functions that are
8768 not known to bind locally. It has no effect without @option{-mfdpic}.
8769 It's enabled by default if optimizing for speed and compiling for
8770 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8771 optimization option such as @option{-O3} or above is present in the
8777 Assume a large TLS segment when generating thread-local code.
8782 Do not assume a large TLS segment when generating thread-local code.
8787 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8788 that is known to be in read-only sections. It's enabled by default,
8789 except for @option{-fpic} or @option{-fpie}: even though it may help
8790 make the global offset table smaller, it trades 1 instruction for 4.
8791 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8792 one of which may be shared by multiple symbols, and it avoids the need
8793 for a GOT entry for the referenced symbol, so it's more likely to be a
8794 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8796 @item -multilib-library-pic
8797 @opindex multilib-library-pic
8799 Link with the (library, not FD) pic libraries. It's implied by
8800 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8801 @option{-fpic} without @option{-mfdpic}. You should never have to use
8807 Follow the EABI requirement of always creating a frame pointer whenever
8808 a stack frame is allocated. This option is enabled by default and can
8809 be disabled with @option{-mno-linked-fp}.
8812 @opindex mlong-calls
8814 Use indirect addressing to call functions outside the current
8815 compilation unit. This allows the functions to be placed anywhere
8816 within the 32-bit address space.
8818 @item -malign-labels
8819 @opindex malign-labels
8821 Try to align labels to an 8-byte boundary by inserting nops into the
8822 previous packet. This option only has an effect when VLIW packing
8823 is enabled. It doesn't create new packets; it merely adds nops to
8827 @opindex mlibrary-pic
8829 Generate position-independent EABI code.
8834 Use only the first four media accumulator registers.
8839 Use all eight media accumulator registers.
8844 Pack VLIW instructions.
8849 Do not pack VLIW instructions.
8854 Do not mark ABI switches in e_flags.
8859 Enable the use of conditional-move instructions (default).
8861 This switch is mainly for debugging the compiler and will likely be removed
8862 in a future version.
8864 @item -mno-cond-move
8865 @opindex mno-cond-move
8867 Disable the use of conditional-move instructions.
8869 This switch is mainly for debugging the compiler and will likely be removed
8870 in a future version.
8875 Enable the use of conditional set instructions (default).
8877 This switch is mainly for debugging the compiler and will likely be removed
8878 in a future version.
8883 Disable the use of conditional set instructions.
8885 This switch is mainly for debugging the compiler and will likely be removed
8886 in a future version.
8891 Enable the use of conditional execution (default).
8893 This switch is mainly for debugging the compiler and will likely be removed
8894 in a future version.
8896 @item -mno-cond-exec
8897 @opindex mno-cond-exec
8899 Disable the use of conditional execution.
8901 This switch is mainly for debugging the compiler and will likely be removed
8902 in a future version.
8905 @opindex mvliw-branch
8907 Run a pass to pack branches into VLIW instructions (default).
8909 This switch is mainly for debugging the compiler and will likely be removed
8910 in a future version.
8912 @item -mno-vliw-branch
8913 @opindex mno-vliw-branch
8915 Do not run a pass to pack branches into VLIW instructions.
8917 This switch is mainly for debugging the compiler and will likely be removed
8918 in a future version.
8920 @item -mmulti-cond-exec
8921 @opindex mmulti-cond-exec
8923 Enable optimization of @code{&&} and @code{||} in conditional execution
8926 This switch is mainly for debugging the compiler and will likely be removed
8927 in a future version.
8929 @item -mno-multi-cond-exec
8930 @opindex mno-multi-cond-exec
8932 Disable optimization of @code{&&} and @code{||} in conditional execution.
8934 This switch is mainly for debugging the compiler and will likely be removed
8935 in a future version.
8937 @item -mnested-cond-exec
8938 @opindex mnested-cond-exec
8940 Enable nested conditional execution optimizations (default).
8942 This switch is mainly for debugging the compiler and will likely be removed
8943 in a future version.
8945 @item -mno-nested-cond-exec
8946 @opindex mno-nested-cond-exec
8948 Disable nested conditional execution optimizations.
8950 This switch is mainly for debugging the compiler and will likely be removed
8951 in a future version.
8953 @item -moptimize-membar
8954 @opindex moptimize-membar
8956 This switch removes redundant @code{membar} instructions from the
8957 compiler generated code. It is enabled by default.
8959 @item -mno-optimize-membar
8960 @opindex mno-optimize-membar
8962 This switch disables the automatic removal of redundant @code{membar}
8963 instructions from the generated code.
8965 @item -mtomcat-stats
8966 @opindex mtomcat-stats
8968 Cause gas to print out tomcat statistics.
8970 @item -mcpu=@var{cpu}
8973 Select the processor type for which to generate code. Possible values are
8974 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8975 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8979 @node GNU/Linux Options
8980 @subsection GNU/Linux Options
8982 These @samp{-m} options are defined for GNU/Linux targets:
8987 Use the GNU C library instead of uClibc. This is the default except
8988 on @samp{*-*-linux-*uclibc*} targets.
8992 Use uClibc instead of the GNU C library. This is the default on
8993 @samp{*-*-linux-*uclibc*} targets.
8996 @node H8/300 Options
8997 @subsection H8/300 Options
8999 These @samp{-m} options are defined for the H8/300 implementations:
9004 Shorten some address references at link time, when possible; uses the
9005 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
9006 ld, Using ld}, for a fuller description.
9010 Generate code for the H8/300H@.
9014 Generate code for the H8S@.
9018 Generate code for the H8S and H8/300H in the normal mode. This switch
9019 must be used either with @option{-mh} or @option{-ms}.
9023 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
9027 Make @code{int} data 32 bits by default.
9031 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
9032 The default for the H8/300H and H8S is to align longs and floats on 4
9034 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
9035 This option has no effect on the H8/300.
9039 @subsection HPPA Options
9040 @cindex HPPA Options
9042 These @samp{-m} options are defined for the HPPA family of computers:
9045 @item -march=@var{architecture-type}
9047 Generate code for the specified architecture. The choices for
9048 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
9049 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
9050 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
9051 architecture option for your machine. Code compiled for lower numbered
9052 architectures will run on higher numbered architectures, but not the
9056 @itemx -mpa-risc-1-1
9057 @itemx -mpa-risc-2-0
9058 @opindex mpa-risc-1-0
9059 @opindex mpa-risc-1-1
9060 @opindex mpa-risc-2-0
9061 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
9064 @opindex mbig-switch
9065 Generate code suitable for big switch tables. Use this option only if
9066 the assembler/linker complain about out of range branches within a switch
9069 @item -mjump-in-delay
9070 @opindex mjump-in-delay
9071 Fill delay slots of function calls with unconditional jump instructions
9072 by modifying the return pointer for the function call to be the target
9073 of the conditional jump.
9075 @item -mdisable-fpregs
9076 @opindex mdisable-fpregs
9077 Prevent floating point registers from being used in any manner. This is
9078 necessary for compiling kernels which perform lazy context switching of
9079 floating point registers. If you use this option and attempt to perform
9080 floating point operations, the compiler will abort.
9082 @item -mdisable-indexing
9083 @opindex mdisable-indexing
9084 Prevent the compiler from using indexing address modes. This avoids some
9085 rather obscure problems when compiling MIG generated code under MACH@.
9087 @item -mno-space-regs
9088 @opindex mno-space-regs
9089 Generate code that assumes the target has no space registers. This allows
9090 GCC to generate faster indirect calls and use unscaled index address modes.
9092 Such code is suitable for level 0 PA systems and kernels.
9094 @item -mfast-indirect-calls
9095 @opindex mfast-indirect-calls
9096 Generate code that assumes calls never cross space boundaries. This
9097 allows GCC to emit code which performs faster indirect calls.
9099 This option will not work in the presence of shared libraries or nested
9102 @item -mfixed-range=@var{register-range}
9103 @opindex mfixed-range
9104 Generate code treating the given register range as fixed registers.
9105 A fixed register is one that the register allocator can not use. This is
9106 useful when compiling kernel code. A register range is specified as
9107 two registers separated by a dash. Multiple register ranges can be
9108 specified separated by a comma.
9110 @item -mlong-load-store
9111 @opindex mlong-load-store
9112 Generate 3-instruction load and store sequences as sometimes required by
9113 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
9116 @item -mportable-runtime
9117 @opindex mportable-runtime
9118 Use the portable calling conventions proposed by HP for ELF systems.
9122 Enable the use of assembler directives only GAS understands.
9124 @item -mschedule=@var{cpu-type}
9126 Schedule code according to the constraints for the machine type
9127 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
9128 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
9129 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
9130 proper scheduling option for your machine. The default scheduling is
9134 @opindex mlinker-opt
9135 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
9136 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
9137 linkers in which they give bogus error messages when linking some programs.
9140 @opindex msoft-float
9141 Generate output containing library calls for floating point.
9142 @strong{Warning:} the requisite libraries are not available for all HPPA
9143 targets. Normally the facilities of the machine's usual C compiler are
9144 used, but this cannot be done directly in cross-compilation. You must make
9145 your own arrangements to provide suitable library functions for
9146 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
9147 does provide software floating point support.
9149 @option{-msoft-float} changes the calling convention in the output file;
9150 therefore, it is only useful if you compile @emph{all} of a program with
9151 this option. In particular, you need to compile @file{libgcc.a}, the
9152 library that comes with GCC, with @option{-msoft-float} in order for
9157 Generate the predefine, @code{_SIO}, for server IO@. The default is
9158 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
9159 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
9160 options are available under HP-UX and HI-UX@.
9164 Use GNU ld specific options. This passes @option{-shared} to ld when
9165 building a shared library. It is the default when GCC is configured,
9166 explicitly or implicitly, with the GNU linker. This option does not
9167 have any affect on which ld is called, it only changes what parameters
9168 are passed to that ld. The ld that is called is determined by the
9169 @option{--with-ld} configure option, GCC's program search path, and
9170 finally by the user's @env{PATH}. The linker used by GCC can be printed
9171 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
9172 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9176 Use HP ld specific options. This passes @option{-b} to ld when building
9177 a shared library and passes @option{+Accept TypeMismatch} to ld on all
9178 links. It is the default when GCC is configured, explicitly or
9179 implicitly, with the HP linker. This option does not have any affect on
9180 which ld is called, it only changes what parameters are passed to that
9181 ld. The ld that is called is determined by the @option{--with-ld}
9182 configure option, GCC's program search path, and finally by the user's
9183 @env{PATH}. The linker used by GCC can be printed using @samp{which
9184 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
9185 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
9188 @opindex mno-long-calls
9189 Generate code that uses long call sequences. This ensures that a call
9190 is always able to reach linker generated stubs. The default is to generate
9191 long calls only when the distance from the call site to the beginning
9192 of the function or translation unit, as the case may be, exceeds a
9193 predefined limit set by the branch type being used. The limits for
9194 normal calls are 7,600,000 and 240,000 bytes, respectively for the
9195 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
9198 Distances are measured from the beginning of functions when using the
9199 @option{-ffunction-sections} option, or when using the @option{-mgas}
9200 and @option{-mno-portable-runtime} options together under HP-UX with
9203 It is normally not desirable to use this option as it will degrade
9204 performance. However, it may be useful in large applications,
9205 particularly when partial linking is used to build the application.
9207 The types of long calls used depends on the capabilities of the
9208 assembler and linker, and the type of code being generated. The
9209 impact on systems that support long absolute calls, and long pic
9210 symbol-difference or pc-relative calls should be relatively small.
9211 However, an indirect call is used on 32-bit ELF systems in pic code
9212 and it is quite long.
9214 @item -munix=@var{unix-std}
9216 Generate compiler predefines and select a startfile for the specified
9217 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
9218 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
9219 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
9220 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
9221 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
9224 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
9225 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
9226 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
9227 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
9228 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
9229 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
9231 It is @emph{important} to note that this option changes the interfaces
9232 for various library routines. It also affects the operational behavior
9233 of the C library. Thus, @emph{extreme} care is needed in using this
9236 Library code that is intended to operate with more than one UNIX
9237 standard must test, set and restore the variable @var{__xpg4_extended_mask}
9238 as appropriate. Most GNU software doesn't provide this capability.
9242 Suppress the generation of link options to search libdld.sl when the
9243 @option{-static} option is specified on HP-UX 10 and later.
9247 The HP-UX implementation of setlocale in libc has a dependency on
9248 libdld.sl. There isn't an archive version of libdld.sl. Thus,
9249 when the @option{-static} option is specified, special link options
9250 are needed to resolve this dependency.
9252 On HP-UX 10 and later, the GCC driver adds the necessary options to
9253 link with libdld.sl when the @option{-static} option is specified.
9254 This causes the resulting binary to be dynamic. On the 64-bit port,
9255 the linkers generate dynamic binaries by default in any case. The
9256 @option{-nolibdld} option can be used to prevent the GCC driver from
9257 adding these link options.
9261 Add support for multithreading with the @dfn{dce thread} library
9262 under HP-UX@. This option sets flags for both the preprocessor and
9266 @node i386 and x86-64 Options
9267 @subsection Intel 386 and AMD x86-64 Options
9268 @cindex i386 Options
9269 @cindex x86-64 Options
9270 @cindex Intel 386 Options
9271 @cindex AMD x86-64 Options
9273 These @samp{-m} options are defined for the i386 and x86-64 family of
9277 @item -mtune=@var{cpu-type}
9279 Tune to @var{cpu-type} everything applicable about the generated code, except
9280 for the ABI and the set of available instructions. The choices for
9284 Produce code optimized for the most common IA32/AMD64/EM64T processors.
9285 If you know the CPU on which your code will run, then you should use
9286 the corresponding @option{-mtune} option instead of
9287 @option{-mtune=generic}. But, if you do not know exactly what CPU users
9288 of your application will have, then you should use this option.
9290 As new processors are deployed in the marketplace, the behavior of this
9291 option will change. Therefore, if you upgrade to a newer version of
9292 GCC, the code generated option will change to reflect the processors
9293 that were most common when that version of GCC was released.
9295 There is no @option{-march=generic} option because @option{-march}
9296 indicates the instruction set the compiler can use, and there is no
9297 generic instruction set applicable to all processors. In contrast,
9298 @option{-mtune} indicates the processor (or, in this case, collection of
9299 processors) for which the code is optimized.
9301 This selects the CPU to tune for at compilation time by determining
9302 the processor type of the compiling machine. Using @option{-mtune=native}
9303 will produce code optimized for the local machine under the constraints
9304 of the selected instruction set. Using @option{-march=native} will
9305 enable all instruction subsets supported by the local machine (hence
9306 the result might not run on different machines).
9308 Original Intel's i386 CPU@.
9310 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9312 Intel Pentium CPU with no MMX support.
9314 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9316 Intel PentiumPro CPU@.
9318 Same as @code{generic}, but when used as @code{march} option, PentiumPro
9319 instruction set will be used, so the code will run on all i686 family chips.
9321 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9322 @item pentium3, pentium3m
9323 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9326 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9327 support. Used by Centrino notebooks.
9328 @item pentium4, pentium4m
9329 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9331 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9334 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9335 SSE2 and SSE3 instruction set support.
9337 Intel Core2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 and SSSE3
9338 instruction set support.
9340 AMD K6 CPU with MMX instruction set support.
9342 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9343 @item athlon, athlon-tbird
9344 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9346 @item athlon-4, athlon-xp, athlon-mp
9347 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9348 instruction set support.
9349 @item k8, opteron, athlon64, athlon-fx
9350 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9351 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9353 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9356 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9357 instruction set support.
9359 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9360 implemented for this chip.)
9362 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9363 implemented for this chip.)
9365 Embedded AMD CPU with MMX and 3dNOW! instruction set support.
9368 While picking a specific @var{cpu-type} will schedule things appropriately
9369 for that particular chip, the compiler will not generate any code that
9370 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9373 @item -march=@var{cpu-type}
9375 Generate instructions for the machine type @var{cpu-type}. The choices
9376 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9377 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9379 @item -mcpu=@var{cpu-type}
9381 A deprecated synonym for @option{-mtune}.
9390 @opindex mpentiumpro
9391 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9392 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9393 These synonyms are deprecated.
9395 @item -mfpmath=@var{unit}
9397 Generate floating point arithmetics for selected unit @var{unit}. The choices
9402 Use the standard 387 floating point coprocessor present majority of chips and
9403 emulated otherwise. Code compiled with this option will run almost everywhere.
9404 The temporary results are computed in 80bit precision instead of precision
9405 specified by the type resulting in slightly different results compared to most
9406 of other chips. See @option{-ffloat-store} for more detailed description.
9408 This is the default choice for i386 compiler.
9411 Use scalar floating point instructions present in the SSE instruction set.
9412 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9413 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9414 instruction set supports only single precision arithmetics, thus the double and
9415 extended precision arithmetics is still done using 387. Later version, present
9416 only in Pentium4 and the future AMD x86-64 chips supports double precision
9419 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9420 or @option{-msse2} switches to enable SSE extensions and make this option
9421 effective. For the x86-64 compiler, these extensions are enabled by default.
9423 The resulting code should be considerably faster in the majority of cases and avoid
9424 the numerical instability problems of 387 code, but may break some existing
9425 code that expects temporaries to be 80bit.
9427 This is the default choice for the x86-64 compiler.
9430 Attempt to utilize both instruction sets at once. This effectively double the
9431 amount of available registers and on chips with separate execution units for
9432 387 and SSE the execution resources too. Use this option with care, as it is
9433 still experimental, because the GCC register allocator does not model separate
9434 functional units well resulting in instable performance.
9437 @item -masm=@var{dialect}
9438 @opindex masm=@var{dialect}
9439 Output asm instructions using selected @var{dialect}. Supported
9440 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9441 not support @samp{intel}.
9446 @opindex mno-ieee-fp
9447 Control whether or not the compiler uses IEEE floating point
9448 comparisons. These handle correctly the case where the result of a
9449 comparison is unordered.
9452 @opindex msoft-float
9453 Generate output containing library calls for floating point.
9454 @strong{Warning:} the requisite libraries are not part of GCC@.
9455 Normally the facilities of the machine's usual C compiler are used, but
9456 this can't be done directly in cross-compilation. You must make your
9457 own arrangements to provide suitable library functions for
9460 On machines where a function returns floating point results in the 80387
9461 register stack, some floating point opcodes may be emitted even if
9462 @option{-msoft-float} is used.
9464 @item -mno-fp-ret-in-387
9465 @opindex mno-fp-ret-in-387
9466 Do not use the FPU registers for return values of functions.
9468 The usual calling convention has functions return values of types
9469 @code{float} and @code{double} in an FPU register, even if there
9470 is no FPU@. The idea is that the operating system should emulate
9473 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9474 in ordinary CPU registers instead.
9476 @item -mno-fancy-math-387
9477 @opindex mno-fancy-math-387
9478 Some 387 emulators do not support the @code{sin}, @code{cos} and
9479 @code{sqrt} instructions for the 387. Specify this option to avoid
9480 generating those instructions. This option is the default on FreeBSD,
9481 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9482 indicates that the target cpu will always have an FPU and so the
9483 instruction will not need emulation. As of revision 2.6.1, these
9484 instructions are not generated unless you also use the
9485 @option{-funsafe-math-optimizations} switch.
9487 @item -malign-double
9488 @itemx -mno-align-double
9489 @opindex malign-double
9490 @opindex mno-align-double
9491 Control whether GCC aligns @code{double}, @code{long double}, and
9492 @code{long long} variables on a two word boundary or a one word
9493 boundary. Aligning @code{double} variables on a two word boundary will
9494 produce code that runs somewhat faster on a @samp{Pentium} at the
9495 expense of more memory.
9497 On x86-64, @option{-malign-double} is enabled by default.
9499 @strong{Warning:} if you use the @option{-malign-double} switch,
9500 structures containing the above types will be aligned differently than
9501 the published application binary interface specifications for the 386
9502 and will not be binary compatible with structures in code compiled
9503 without that switch.
9505 @item -m96bit-long-double
9506 @itemx -m128bit-long-double
9507 @opindex m96bit-long-double
9508 @opindex m128bit-long-double
9509 These switches control the size of @code{long double} type. The i386
9510 application binary interface specifies the size to be 96 bits,
9511 so @option{-m96bit-long-double} is the default in 32 bit mode.
9513 Modern architectures (Pentium and newer) would prefer @code{long double}
9514 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9515 conforming to the ABI, this would not be possible. So specifying a
9516 @option{-m128bit-long-double} will align @code{long double}
9517 to a 16 byte boundary by padding the @code{long double} with an additional
9520 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9521 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9523 Notice that neither of these options enable any extra precision over the x87
9524 standard of 80 bits for a @code{long double}.
9526 @strong{Warning:} if you override the default value for your target ABI, the
9527 structures and arrays containing @code{long double} variables will change
9528 their size as well as function calling convention for function taking
9529 @code{long double} will be modified. Hence they will not be binary
9530 compatible with arrays or structures in code compiled without that switch.
9532 @item -mmlarge-data-threshold=@var{number}
9533 @opindex mlarge-data-threshold=@var{number}
9534 When @option{-mcmodel=medium} is specified, the data greater than
9535 @var{threshold} are placed in large data section. This value must be the
9536 same across all object linked into the binary and defaults to 65535.
9539 @itemx -mno-svr3-shlib
9540 @opindex msvr3-shlib
9541 @opindex mno-svr3-shlib
9542 Control whether GCC places uninitialized local variables into the
9543 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9544 into @code{bss}. These options are meaningful only on System V Release 3.
9548 Use a different function-calling convention, in which functions that
9549 take a fixed number of arguments return with the @code{ret} @var{num}
9550 instruction, which pops their arguments while returning. This saves one
9551 instruction in the caller since there is no need to pop the arguments
9554 You can specify that an individual function is called with this calling
9555 sequence with the function attribute @samp{stdcall}. You can also
9556 override the @option{-mrtd} option by using the function attribute
9557 @samp{cdecl}. @xref{Function Attributes}.
9559 @strong{Warning:} this calling convention is incompatible with the one
9560 normally used on Unix, so you cannot use it if you need to call
9561 libraries compiled with the Unix compiler.
9563 Also, you must provide function prototypes for all functions that
9564 take variable numbers of arguments (including @code{printf});
9565 otherwise incorrect code will be generated for calls to those
9568 In addition, seriously incorrect code will result if you call a
9569 function with too many arguments. (Normally, extra arguments are
9570 harmlessly ignored.)
9572 @item -mregparm=@var{num}
9574 Control how many registers are used to pass integer arguments. By
9575 default, no registers are used to pass arguments, and at most 3
9576 registers can be used. You can control this behavior for a specific
9577 function by using the function attribute @samp{regparm}.
9578 @xref{Function Attributes}.
9580 @strong{Warning:} if you use this switch, and
9581 @var{num} is nonzero, then you must build all modules with the same
9582 value, including any libraries. This includes the system libraries and
9586 @opindex mx87regparm
9587 Use 80387 register passing conventions for floating point arguments.
9588 You can control this behavior for a specific function by using the
9589 function attribute @samp{x87regparm}.
9590 @xref{Function Attributes}.
9592 @strong{Warning:} if you use this switch then you must build all
9593 modules with the same value, including any libraries. This includes
9594 the system libraries and startup modules.
9597 @opindex msseregparm
9598 Use SSE register passing conventions for float and double arguments
9599 and return values. You can control this behavior for a specific
9600 function by using the function attribute @samp{sseregparm}.
9601 @xref{Function Attributes}.
9603 @strong{Warning:} if you use this switch then you must build all
9604 modules with the same value, including any libraries. This includes
9605 the system libraries and startup modules.
9607 @item -mstackrealign
9608 @opindex mstackrealign
9609 Realign the stack at entry. On the Intel x86, the
9610 @option{-mstackrealign} option will generate an alternate prologue and
9611 epilogue that realigns the runtime stack. This supports mixing legacy
9612 codes that keep a 4-byte aligned stack with modern codes that keep a
9613 16-byte stack for SSE compatibility. The alternate prologue and
9614 epilogue are slower and bigger than the regular ones, and the
9615 alternate prologue requires an extra scratch register; this lowers the
9616 number of registers available if used in conjunction with the
9617 @code{regparm} attribute. The @option{-mstackrealign} option is
9618 incompatible with the nested function prologue; this is considered a
9619 hard error. See also the attribute @code{force_align_arg_pointer},
9620 applicable to individual functions.
9622 @item -mpreferred-stack-boundary=@var{num}
9623 @opindex mpreferred-stack-boundary
9624 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9625 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9626 the default is 4 (16 bytes or 128 bits).
9628 On Pentium and PentiumPro, @code{double} and @code{long double} values
9629 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9630 suffer significant run time performance penalties. On Pentium III, the
9631 Streaming SIMD Extension (SSE) data type @code{__m128} may not work
9632 properly if it is not 16 byte aligned.
9634 To ensure proper alignment of this values on the stack, the stack boundary
9635 must be as aligned as that required by any value stored on the stack.
9636 Further, every function must be generated such that it keeps the stack
9637 aligned. Thus calling a function compiled with a higher preferred
9638 stack boundary from a function compiled with a lower preferred stack
9639 boundary will most likely misalign the stack. It is recommended that
9640 libraries that use callbacks always use the default setting.
9642 This extra alignment does consume extra stack space, and generally
9643 increases code size. Code that is sensitive to stack space usage, such
9644 as embedded systems and operating system kernels, may want to reduce the
9645 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9665 These switches enable or disable the use of instructions in the MMX,
9666 SSE, SSE2, SSE3, SSSE3 or 3DNow! extended instruction sets.
9667 These extensions are also available as built-in functions: see
9668 @ref{X86 Built-in Functions}, for details of the functions enabled and
9669 disabled by these switches.
9671 To have SSE/SSE2 instructions generated automatically from floating-point
9672 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9674 These options will enable GCC to use these extended instructions in
9675 generated code, even without @option{-mfpmath=sse}. Applications which
9676 perform runtime CPU detection must compile separate files for each
9677 supported architecture, using the appropriate flags. In particular,
9678 the file containing the CPU detection code should be compiled without
9682 @itemx -mno-push-args
9684 @opindex mno-push-args
9685 Use PUSH operations to store outgoing parameters. This method is shorter
9686 and usually equally fast as method using SUB/MOV operations and is enabled
9687 by default. In some cases disabling it may improve performance because of
9688 improved scheduling and reduced dependencies.
9690 @item -maccumulate-outgoing-args
9691 @opindex maccumulate-outgoing-args
9692 If enabled, the maximum amount of space required for outgoing arguments will be
9693 computed in the function prologue. This is faster on most modern CPUs
9694 because of reduced dependencies, improved scheduling and reduced stack usage
9695 when preferred stack boundary is not equal to 2. The drawback is a notable
9696 increase in code size. This switch implies @option{-mno-push-args}.
9700 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9701 on thread-safe exception handling must compile and link all code with the
9702 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9703 @option{-D_MT}; when linking, it links in a special thread helper library
9704 @option{-lmingwthrd} which cleans up per thread exception handling data.
9706 @item -mno-align-stringops
9707 @opindex mno-align-stringops
9708 Do not align destination of inlined string operations. This switch reduces
9709 code size and improves performance in case the destination is already aligned,
9710 but GCC doesn't know about it.
9712 @item -minline-all-stringops
9713 @opindex minline-all-stringops
9714 By default GCC inlines string operations only when destination is known to be
9715 aligned at least to 4 byte boundary. This enables more inlining, increase code
9716 size, but may improve performance of code that depends on fast memcpy, strlen
9717 and memset for short lengths.
9719 @item -momit-leaf-frame-pointer
9720 @opindex momit-leaf-frame-pointer
9721 Don't keep the frame pointer in a register for leaf functions. This
9722 avoids the instructions to save, set up and restore frame pointers and
9723 makes an extra register available in leaf functions. The option
9724 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9725 which might make debugging harder.
9727 @item -mtls-direct-seg-refs
9728 @itemx -mno-tls-direct-seg-refs
9729 @opindex mtls-direct-seg-refs
9730 Controls whether TLS variables may be accessed with offsets from the
9731 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9732 or whether the thread base pointer must be added. Whether or not this
9733 is legal depends on the operating system, and whether it maps the
9734 segment to cover the entire TLS area.
9736 For systems that use GNU libc, the default is on.
9739 These @samp{-m} switches are supported in addition to the above
9740 on AMD x86-64 processors in 64-bit environments.
9747 Generate code for a 32-bit or 64-bit environment.
9748 The 32-bit environment sets int, long and pointer to 32 bits and
9749 generates code that runs on any i386 system.
9750 The 64-bit environment sets int to 32 bits and long and pointer
9751 to 64 bits and generates code for AMD's x86-64 architecture.
9754 @opindex no-red-zone
9755 Do not use a so called red zone for x86-64 code. The red zone is mandated
9756 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9757 stack pointer that will not be modified by signal or interrupt handlers
9758 and therefore can be used for temporary data without adjusting the stack
9759 pointer. The flag @option{-mno-red-zone} disables this red zone.
9761 @item -mcmodel=small
9762 @opindex mcmodel=small
9763 Generate code for the small code model: the program and its symbols must
9764 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9765 Programs can be statically or dynamically linked. This is the default
9768 @item -mcmodel=kernel
9769 @opindex mcmodel=kernel
9770 Generate code for the kernel code model. The kernel runs in the
9771 negative 2 GB of the address space.
9772 This model has to be used for Linux kernel code.
9774 @item -mcmodel=medium
9775 @opindex mcmodel=medium
9776 Generate code for the medium model: The program is linked in the lower 2
9777 GB of the address space but symbols can be located anywhere in the
9778 address space. Programs can be statically or dynamically linked, but
9779 building of shared libraries are not supported with the medium model.
9781 @item -mcmodel=large
9782 @opindex mcmodel=large
9783 Generate code for the large model: This model makes no assumptions
9784 about addresses and sizes of sections. Currently GCC does not implement
9789 @subsection IA-64 Options
9790 @cindex IA-64 Options
9792 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9796 @opindex mbig-endian
9797 Generate code for a big endian target. This is the default for HP-UX@.
9799 @item -mlittle-endian
9800 @opindex mlittle-endian
9801 Generate code for a little endian target. This is the default for AIX5
9808 Generate (or don't) code for the GNU assembler. This is the default.
9809 @c Also, this is the default if the configure option @option{--with-gnu-as}
9816 Generate (or don't) code for the GNU linker. This is the default.
9817 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9822 Generate code that does not use a global pointer register. The result
9823 is not position independent code, and violates the IA-64 ABI@.
9825 @item -mvolatile-asm-stop
9826 @itemx -mno-volatile-asm-stop
9827 @opindex mvolatile-asm-stop
9828 @opindex mno-volatile-asm-stop
9829 Generate (or don't) a stop bit immediately before and after volatile asm
9832 @item -mregister-names
9833 @itemx -mno-register-names
9834 @opindex mregister-names
9835 @opindex mno-register-names
9836 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9837 the stacked registers. This may make assembler output more readable.
9843 Disable (or enable) optimizations that use the small data section. This may
9844 be useful for working around optimizer bugs.
9847 @opindex mconstant-gp
9848 Generate code that uses a single constant global pointer value. This is
9849 useful when compiling kernel code.
9853 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9854 This is useful when compiling firmware code.
9856 @item -minline-float-divide-min-latency
9857 @opindex minline-float-divide-min-latency
9858 Generate code for inline divides of floating point values
9859 using the minimum latency algorithm.
9861 @item -minline-float-divide-max-throughput
9862 @opindex minline-float-divide-max-throughput
9863 Generate code for inline divides of floating point values
9864 using the maximum throughput algorithm.
9866 @item -minline-int-divide-min-latency
9867 @opindex minline-int-divide-min-latency
9868 Generate code for inline divides of integer values
9869 using the minimum latency algorithm.
9871 @item -minline-int-divide-max-throughput
9872 @opindex minline-int-divide-max-throughput
9873 Generate code for inline divides of integer values
9874 using the maximum throughput algorithm.
9876 @item -minline-sqrt-min-latency
9877 @opindex minline-sqrt-min-latency
9878 Generate code for inline square roots
9879 using the minimum latency algorithm.
9881 @item -minline-sqrt-max-throughput
9882 @opindex minline-sqrt-max-throughput
9883 Generate code for inline square roots
9884 using the maximum throughput algorithm.
9886 @item -mno-dwarf2-asm
9888 @opindex mno-dwarf2-asm
9889 @opindex mdwarf2-asm
9890 Don't (or do) generate assembler code for the DWARF2 line number debugging
9891 info. This may be useful when not using the GNU assembler.
9893 @item -mearly-stop-bits
9894 @itemx -mno-early-stop-bits
9895 @opindex mearly-stop-bits
9896 @opindex mno-early-stop-bits
9897 Allow stop bits to be placed earlier than immediately preceding the
9898 instruction that triggered the stop bit. This can improve instruction
9899 scheduling, but does not always do so.
9901 @item -mfixed-range=@var{register-range}
9902 @opindex mfixed-range
9903 Generate code treating the given register range as fixed registers.
9904 A fixed register is one that the register allocator can not use. This is
9905 useful when compiling kernel code. A register range is specified as
9906 two registers separated by a dash. Multiple register ranges can be
9907 specified separated by a comma.
9909 @item -mtls-size=@var{tls-size}
9911 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9914 @item -mtune=@var{cpu-type}
9916 Tune the instruction scheduling for a particular CPU, Valid values are
9917 itanium, itanium1, merced, itanium2, and mckinley.
9923 Add support for multithreading using the POSIX threads library. This
9924 option sets flags for both the preprocessor and linker. It does
9925 not affect the thread safety of object code produced by the compiler or
9926 that of libraries supplied with it. These are HP-UX specific flags.
9932 Generate code for a 32-bit or 64-bit environment.
9933 The 32-bit environment sets int, long and pointer to 32 bits.
9934 The 64-bit environment sets int to 32 bits and long and pointer
9935 to 64 bits. These are HP-UX specific flags.
9937 @item -mno-sched-br-data-spec
9938 @itemx -msched-br-data-spec
9939 @opindex -mno-sched-br-data-spec
9940 @opindex -msched-br-data-spec
9941 (Dis/En)able data speculative scheduling before reload.
9942 This will result in generation of the ld.a instructions and
9943 the corresponding check instructions (ld.c / chk.a).
9944 The default is 'disable'.
9946 @item -msched-ar-data-spec
9947 @itemx -mno-sched-ar-data-spec
9948 @opindex -msched-ar-data-spec
9949 @opindex -mno-sched-ar-data-spec
9950 (En/Dis)able data speculative scheduling after reload.
9951 This will result in generation of the ld.a instructions and
9952 the corresponding check instructions (ld.c / chk.a).
9953 The default is 'enable'.
9955 @item -mno-sched-control-spec
9956 @itemx -msched-control-spec
9957 @opindex -mno-sched-control-spec
9958 @opindex -msched-control-spec
9959 (Dis/En)able control speculative scheduling. This feature is
9960 available only during region scheduling (i.e. before reload).
9961 This will result in generation of the ld.s instructions and
9962 the corresponding check instructions chk.s .
9963 The default is 'disable'.
9965 @item -msched-br-in-data-spec
9966 @itemx -mno-sched-br-in-data-spec
9967 @opindex -msched-br-in-data-spec
9968 @opindex -mno-sched-br-in-data-spec
9969 (En/Dis)able speculative scheduling of the instructions that
9970 are dependent on the data speculative loads before reload.
9971 This is effective only with @option{-msched-br-data-spec} enabled.
9972 The default is 'enable'.
9974 @item -msched-ar-in-data-spec
9975 @itemx -mno-sched-ar-in-data-spec
9976 @opindex -msched-ar-in-data-spec
9977 @opindex -mno-sched-ar-in-data-spec
9978 (En/Dis)able speculative scheduling of the instructions that
9979 are dependent on the data speculative loads after reload.
9980 This is effective only with @option{-msched-ar-data-spec} enabled.
9981 The default is 'enable'.
9983 @item -msched-in-control-spec
9984 @itemx -mno-sched-in-control-spec
9985 @opindex -msched-in-control-spec
9986 @opindex -mno-sched-in-control-spec
9987 (En/Dis)able speculative scheduling of the instructions that
9988 are dependent on the control speculative loads.
9989 This is effective only with @option{-msched-control-spec} enabled.
9990 The default is 'enable'.
9993 @itemx -mno-sched-ldc
9994 @opindex -msched-ldc
9995 @opindex -mno-sched-ldc
9996 (En/Dis)able use of simple data speculation checks ld.c .
9997 If disabled, only chk.a instructions will be emitted to check
9998 data speculative loads.
9999 The default is 'enable'.
10001 @item -mno-sched-control-ldc
10002 @itemx -msched-control-ldc
10003 @opindex -mno-sched-control-ldc
10004 @opindex -msched-control-ldc
10005 (Dis/En)able use of ld.c instructions to check control speculative loads.
10006 If enabled, in case of control speculative load with no speculatively
10007 scheduled dependent instructions this load will be emitted as ld.sa and
10008 ld.c will be used to check it.
10009 The default is 'disable'.
10011 @item -mno-sched-spec-verbose
10012 @itemx -msched-spec-verbose
10013 @opindex -mno-sched-spec-verbose
10014 @opindex -msched-spec-verbose
10015 (Dis/En)able printing of the information about speculative motions.
10017 @item -mno-sched-prefer-non-data-spec-insns
10018 @itemx -msched-prefer-non-data-spec-insns
10019 @opindex -mno-sched-prefer-non-data-spec-insns
10020 @opindex -msched-prefer-non-data-spec-insns
10021 If enabled, data speculative instructions will be chosen for schedule
10022 only if there are no other choices at the moment. This will make
10023 the use of the data speculation much more conservative.
10024 The default is 'disable'.
10026 @item -mno-sched-prefer-non-control-spec-insns
10027 @itemx -msched-prefer-non-control-spec-insns
10028 @opindex -mno-sched-prefer-non-control-spec-insns
10029 @opindex -msched-prefer-non-control-spec-insns
10030 If enabled, control speculative instructions will be chosen for schedule
10031 only if there are no other choices at the moment. This will make
10032 the use of the control speculation much more conservative.
10033 The default is 'disable'.
10035 @item -mno-sched-count-spec-in-critical-path
10036 @itemx -msched-count-spec-in-critical-path
10037 @opindex -mno-sched-count-spec-in-critical-path
10038 @opindex -msched-count-spec-in-critical-path
10039 If enabled, speculative dependencies will be considered during
10040 computation of the instructions priorities. This will make the use of the
10041 speculation a bit more conservative.
10042 The default is 'disable'.
10047 @subsection M32C Options
10048 @cindex M32C options
10051 @item -mcpu=@var{name}
10053 Select the CPU for which code is generated. @var{name} may be one of
10054 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
10055 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
10056 the M32C/80 series.
10060 Specifies that the program will be run on the simulator. This causes
10061 an alternate runtime library to be linked in which supports, for
10062 example, file I/O. You must not use this option when generating
10063 programs that will run on real hardware; you must provide your own
10064 runtime library for whatever I/O functions are needed.
10066 @item -memregs=@var{number}
10068 Specifies the number of memory-based pseudo-registers GCC will use
10069 during code generation. These pseudo-registers will be used like real
10070 registers, so there is a tradeoff between GCC's ability to fit the
10071 code into available registers, and the performance penalty of using
10072 memory instead of registers. Note that all modules in a program must
10073 be compiled with the same value for this option. Because of that, you
10074 must not use this option with the default runtime libraries gcc
10079 @node M32R/D Options
10080 @subsection M32R/D Options
10081 @cindex M32R/D options
10083 These @option{-m} options are defined for Renesas M32R/D architectures:
10088 Generate code for the M32R/2@.
10092 Generate code for the M32R/X@.
10096 Generate code for the M32R@. This is the default.
10098 @item -mmodel=small
10099 @opindex mmodel=small
10100 Assume all objects live in the lower 16MB of memory (so that their addresses
10101 can be loaded with the @code{ld24} instruction), and assume all subroutines
10102 are reachable with the @code{bl} instruction.
10103 This is the default.
10105 The addressability of a particular object can be set with the
10106 @code{model} attribute.
10108 @item -mmodel=medium
10109 @opindex mmodel=medium
10110 Assume objects may be anywhere in the 32-bit address space (the compiler
10111 will generate @code{seth/add3} instructions to load their addresses), and
10112 assume all subroutines are reachable with the @code{bl} instruction.
10114 @item -mmodel=large
10115 @opindex mmodel=large
10116 Assume objects may be anywhere in the 32-bit address space (the compiler
10117 will generate @code{seth/add3} instructions to load their addresses), and
10118 assume subroutines may not be reachable with the @code{bl} instruction
10119 (the compiler will generate the much slower @code{seth/add3/jl}
10120 instruction sequence).
10123 @opindex msdata=none
10124 Disable use of the small data area. Variables will be put into
10125 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
10126 @code{section} attribute has been specified).
10127 This is the default.
10129 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
10130 Objects may be explicitly put in the small data area with the
10131 @code{section} attribute using one of these sections.
10133 @item -msdata=sdata
10134 @opindex msdata=sdata
10135 Put small global and static data in the small data area, but do not
10136 generate special code to reference them.
10139 @opindex msdata=use
10140 Put small global and static data in the small data area, and generate
10141 special instructions to reference them.
10145 @cindex smaller data references
10146 Put global and static objects less than or equal to @var{num} bytes
10147 into the small data or bss sections instead of the normal data or bss
10148 sections. The default value of @var{num} is 8.
10149 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
10150 for this option to have any effect.
10152 All modules should be compiled with the same @option{-G @var{num}} value.
10153 Compiling with different values of @var{num} may or may not work; if it
10154 doesn't the linker will give an error message---incorrect code will not be
10159 Makes the M32R specific code in the compiler display some statistics
10160 that might help in debugging programs.
10162 @item -malign-loops
10163 @opindex malign-loops
10164 Align all loops to a 32-byte boundary.
10166 @item -mno-align-loops
10167 @opindex mno-align-loops
10168 Do not enforce a 32-byte alignment for loops. This is the default.
10170 @item -missue-rate=@var{number}
10171 @opindex missue-rate=@var{number}
10172 Issue @var{number} instructions per cycle. @var{number} can only be 1
10175 @item -mbranch-cost=@var{number}
10176 @opindex mbranch-cost=@var{number}
10177 @var{number} can only be 1 or 2. If it is 1 then branches will be
10178 preferred over conditional code, if it is 2, then the opposite will
10181 @item -mflush-trap=@var{number}
10182 @opindex mflush-trap=@var{number}
10183 Specifies the trap number to use to flush the cache. The default is
10184 12. Valid numbers are between 0 and 15 inclusive.
10186 @item -mno-flush-trap
10187 @opindex mno-flush-trap
10188 Specifies that the cache cannot be flushed by using a trap.
10190 @item -mflush-func=@var{name}
10191 @opindex mflush-func=@var{name}
10192 Specifies the name of the operating system function to call to flush
10193 the cache. The default is @emph{_flush_cache}, but a function call
10194 will only be used if a trap is not available.
10196 @item -mno-flush-func
10197 @opindex mno-flush-func
10198 Indicates that there is no OS function for flushing the cache.
10202 @node M680x0 Options
10203 @subsection M680x0 Options
10204 @cindex M680x0 options
10206 These are the @samp{-m} options defined for the 68000 series. The default
10207 values for these options depends on which style of 68000 was selected when
10208 the compiler was configured; the defaults for the most common choices are
10216 Generate output for a 68000. This is the default
10217 when the compiler is configured for 68000-based systems.
10219 Use this option for microcontrollers with a 68000 or EC000 core,
10220 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
10226 Generate output for a 68020. This is the default
10227 when the compiler is configured for 68020-based systems.
10231 Generate output containing 68881 instructions for floating point.
10232 This is the default for most 68020 systems unless @option{--nfp} was
10233 specified when the compiler was configured.
10237 Generate output for a 68030. This is the default when the compiler is
10238 configured for 68030-based systems.
10242 Generate output for a 68040. This is the default when the compiler is
10243 configured for 68040-based systems.
10245 This option inhibits the use of 68881/68882 instructions that have to be
10246 emulated by software on the 68040. Use this option if your 68040 does not
10247 have code to emulate those instructions.
10251 Generate output for a 68060. This is the default when the compiler is
10252 configured for 68060-based systems.
10254 This option inhibits the use of 68020 and 68881/68882 instructions that
10255 have to be emulated by software on the 68060. Use this option if your 68060
10256 does not have code to emulate those instructions.
10260 Generate output for a CPU32. This is the default
10261 when the compiler is configured for CPU32-based systems.
10263 Use this option for microcontrollers with a
10264 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
10265 68336, 68340, 68341, 68349 and 68360.
10269 Generate output for a 520X ``coldfire'' family cpu. This is the default
10270 when the compiler is configured for 520X-based systems.
10272 Use this option for microcontroller with a 5200 core, including
10273 the MCF5202, MCF5203, MCF5204 and MCF5202.
10277 Generate output for a ColdFire V4e family cpu (e.g.@: 547x/548x).
10278 This includes use of hardware floating point instructions.
10282 Generate output for a 68040, without using any of the new instructions.
10283 This results in code which can run relatively efficiently on either a
10284 68020/68881 or a 68030 or a 68040. The generated code does use the
10285 68881 instructions that are emulated on the 68040.
10289 Generate output for a 68060, without using any of the new instructions.
10290 This results in code which can run relatively efficiently on either a
10291 68020/68881 or a 68030 or a 68040. The generated code does use the
10292 68881 instructions that are emulated on the 68060.
10295 @opindex msoft-float
10296 Generate output containing library calls for floating point.
10297 @strong{Warning:} the requisite libraries are not available for all m68k
10298 targets. Normally the facilities of the machine's usual C compiler are
10299 used, but this can't be done directly in cross-compilation. You must
10300 make your own arrangements to provide suitable library functions for
10301 cross-compilation. The embedded targets @samp{m68k-*-aout} and
10302 @samp{m68k-*-coff} do provide software floating point support.
10306 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10307 Additionally, parameters passed on the stack are also aligned to a
10308 16-bit boundary even on targets whose API mandates promotion to 32-bit.
10311 @opindex mnobitfield
10312 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
10313 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
10317 Do use the bit-field instructions. The @option{-m68020} option implies
10318 @option{-mbitfield}. This is the default if you use a configuration
10319 designed for a 68020.
10323 Use a different function-calling convention, in which functions
10324 that take a fixed number of arguments return with the @code{rtd}
10325 instruction, which pops their arguments while returning. This
10326 saves one instruction in the caller since there is no need to pop
10327 the arguments there.
10329 This calling convention is incompatible with the one normally
10330 used on Unix, so you cannot use it if you need to call libraries
10331 compiled with the Unix compiler.
10333 Also, you must provide function prototypes for all functions that
10334 take variable numbers of arguments (including @code{printf});
10335 otherwise incorrect code will be generated for calls to those
10338 In addition, seriously incorrect code will result if you call a
10339 function with too many arguments. (Normally, extra arguments are
10340 harmlessly ignored.)
10342 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
10343 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
10346 @itemx -mno-align-int
10347 @opindex malign-int
10348 @opindex mno-align-int
10349 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
10350 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
10351 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
10352 Aligning variables on 32-bit boundaries produces code that runs somewhat
10353 faster on processors with 32-bit busses at the expense of more memory.
10355 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
10356 align structures containing the above types differently than
10357 most published application binary interface specifications for the m68k.
10361 Use the pc-relative addressing mode of the 68000 directly, instead of
10362 using a global offset table. At present, this option implies @option{-fpic},
10363 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
10364 not presently supported with @option{-mpcrel}, though this could be supported for
10365 68020 and higher processors.
10367 @item -mno-strict-align
10368 @itemx -mstrict-align
10369 @opindex mno-strict-align
10370 @opindex mstrict-align
10371 Do not (do) assume that unaligned memory references will be handled by
10375 Generate code that allows the data segment to be located in a different
10376 area of memory from the text segment. This allows for execute in place in
10377 an environment without virtual memory management. This option implies
10380 @item -mno-sep-data
10381 Generate code that assumes that the data segment follows the text segment.
10382 This is the default.
10384 @item -mid-shared-library
10385 Generate code that supports shared libraries via the library ID method.
10386 This allows for execute in place and shared libraries in an environment
10387 without virtual memory management. This option implies @option{-fPIC}.
10389 @item -mno-id-shared-library
10390 Generate code that doesn't assume ID based shared libraries are being used.
10391 This is the default.
10393 @item -mshared-library-id=n
10394 Specified the identification number of the ID based shared library being
10395 compiled. Specifying a value of 0 will generate more compact code, specifying
10396 other values will force the allocation of that number to the current
10397 library but is no more space or time efficient than omitting this option.
10401 @node M68hc1x Options
10402 @subsection M68hc1x Options
10403 @cindex M68hc1x options
10405 These are the @samp{-m} options defined for the 68hc11 and 68hc12
10406 microcontrollers. The default values for these options depends on
10407 which style of microcontroller was selected when the compiler was configured;
10408 the defaults for the most common choices are given below.
10415 Generate output for a 68HC11. This is the default
10416 when the compiler is configured for 68HC11-based systems.
10422 Generate output for a 68HC12. This is the default
10423 when the compiler is configured for 68HC12-based systems.
10429 Generate output for a 68HCS12.
10431 @item -mauto-incdec
10432 @opindex mauto-incdec
10433 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
10440 Enable the use of 68HC12 min and max instructions.
10443 @itemx -mno-long-calls
10444 @opindex mlong-calls
10445 @opindex mno-long-calls
10446 Treat all calls as being far away (near). If calls are assumed to be
10447 far away, the compiler will use the @code{call} instruction to
10448 call a function and the @code{rtc} instruction for returning.
10452 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10454 @item -msoft-reg-count=@var{count}
10455 @opindex msoft-reg-count
10456 Specify the number of pseudo-soft registers which are used for the
10457 code generation. The maximum number is 32. Using more pseudo-soft
10458 register may or may not result in better code depending on the program.
10459 The default is 4 for 68HC11 and 2 for 68HC12.
10463 @node MCore Options
10464 @subsection MCore Options
10465 @cindex MCore options
10467 These are the @samp{-m} options defined for the Motorola M*Core
10473 @itemx -mno-hardlit
10475 @opindex mno-hardlit
10476 Inline constants into the code stream if it can be done in two
10477 instructions or less.
10483 Use the divide instruction. (Enabled by default).
10485 @item -mrelax-immediate
10486 @itemx -mno-relax-immediate
10487 @opindex mrelax-immediate
10488 @opindex mno-relax-immediate
10489 Allow arbitrary sized immediates in bit operations.
10491 @item -mwide-bitfields
10492 @itemx -mno-wide-bitfields
10493 @opindex mwide-bitfields
10494 @opindex mno-wide-bitfields
10495 Always treat bit-fields as int-sized.
10497 @item -m4byte-functions
10498 @itemx -mno-4byte-functions
10499 @opindex m4byte-functions
10500 @opindex mno-4byte-functions
10501 Force all functions to be aligned to a four byte boundary.
10503 @item -mcallgraph-data
10504 @itemx -mno-callgraph-data
10505 @opindex mcallgraph-data
10506 @opindex mno-callgraph-data
10507 Emit callgraph information.
10510 @itemx -mno-slow-bytes
10511 @opindex mslow-bytes
10512 @opindex mno-slow-bytes
10513 Prefer word access when reading byte quantities.
10515 @item -mlittle-endian
10516 @itemx -mbig-endian
10517 @opindex mlittle-endian
10518 @opindex mbig-endian
10519 Generate code for a little endian target.
10525 Generate code for the 210 processor.
10529 @subsection MIPS Options
10530 @cindex MIPS options
10536 Generate big-endian code.
10540 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10543 @item -march=@var{arch}
10545 Generate code that will run on @var{arch}, which can be the name of a
10546 generic MIPS ISA, or the name of a particular processor.
10548 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10549 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10550 The processor names are:
10551 @samp{4kc}, @samp{4km}, @samp{4kp},
10552 @samp{4kec}, @samp{4kem}, @samp{4kep},
10553 @samp{5kc}, @samp{5kf},
10555 @samp{24kc}, @samp{24kf}, @samp{24kx},
10556 @samp{24kec}, @samp{24kef}, @samp{24kex},
10557 @samp{34kc}, @samp{34kf}, @samp{34kx},
10560 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10561 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10562 @samp{rm7000}, @samp{rm9000},
10565 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10566 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10567 The special value @samp{from-abi} selects the
10568 most compatible architecture for the selected ABI (that is,
10569 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10571 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10572 (for example, @samp{-march=r2k}). Prefixes are optional, and
10573 @samp{vr} may be written @samp{r}.
10575 GCC defines two macros based on the value of this option. The first
10576 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10577 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10578 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10579 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10580 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10582 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10583 above. In other words, it will have the full prefix and will not
10584 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10585 the macro names the resolved architecture (either @samp{"mips1"} or
10586 @samp{"mips3"}). It names the default architecture when no
10587 @option{-march} option is given.
10589 @item -mtune=@var{arch}
10591 Optimize for @var{arch}. Among other things, this option controls
10592 the way instructions are scheduled, and the perceived cost of arithmetic
10593 operations. The list of @var{arch} values is the same as for
10596 When this option is not used, GCC will optimize for the processor
10597 specified by @option{-march}. By using @option{-march} and
10598 @option{-mtune} together, it is possible to generate code that will
10599 run on a family of processors, but optimize the code for one
10600 particular member of that family.
10602 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10603 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10604 @samp{-march} ones described above.
10608 Equivalent to @samp{-march=mips1}.
10612 Equivalent to @samp{-march=mips2}.
10616 Equivalent to @samp{-march=mips3}.
10620 Equivalent to @samp{-march=mips4}.
10624 Equivalent to @samp{-march=mips32}.
10628 Equivalent to @samp{-march=mips32r2}.
10632 Equivalent to @samp{-march=mips64}.
10637 @opindex mno-mips16
10638 Generate (do not generate) MIPS16 code. If GCC is targetting a
10639 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10651 Generate code for the given ABI@.
10653 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10654 generates 64-bit code when you select a 64-bit architecture, but you
10655 can use @option{-mgp32} to get 32-bit code instead.
10657 For information about the O64 ABI, see
10658 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10661 @itemx -mno-abicalls
10663 @opindex mno-abicalls
10664 Generate (do not generate) code that is suitable for SVR4-style
10665 dynamic objects. @option{-mabicalls} is the default for SVR4-based
10670 Generate (do not generate) code that is fully position-independent,
10671 and that can therefore be linked into shared libraries. This option
10672 only affects @option{-mabicalls}.
10674 All @option{-mabicalls} code has traditionally been position-independent,
10675 regardless of options like @option{-fPIC} and @option{-fpic}. However,
10676 as an extension, the GNU toolchain allows executables to use absolute
10677 accesses for locally-binding symbols. It can also use shorter GP
10678 initialization sequences and generate direct calls to locally-defined
10679 functions. This mode is selected by @option{-mno-shared}.
10681 @option{-mno-shared} depends on binutils 2.16 or higher and generates
10682 objects that can only be linked by the GNU linker. However, the option
10683 does not affect the ABI of the final executable; it only affects the ABI
10684 of relocatable objects. Using @option{-mno-shared} will generally make
10685 executables both smaller and quicker.
10687 @option{-mshared} is the default.
10693 Lift (do not lift) the usual restrictions on the size of the global
10696 GCC normally uses a single instruction to load values from the GOT@.
10697 While this is relatively efficient, it will only work if the GOT
10698 is smaller than about 64k. Anything larger will cause the linker
10699 to report an error such as:
10701 @cindex relocation truncated to fit (MIPS)
10703 relocation truncated to fit: R_MIPS_GOT16 foobar
10706 If this happens, you should recompile your code with @option{-mxgot}.
10707 It should then work with very large GOTs, although it will also be
10708 less efficient, since it will take three instructions to fetch the
10709 value of a global symbol.
10711 Note that some linkers can create multiple GOTs. If you have such a
10712 linker, you should only need to use @option{-mxgot} when a single object
10713 file accesses more than 64k's worth of GOT entries. Very few do.
10715 These options have no effect unless GCC is generating position
10720 Assume that general-purpose registers are 32 bits wide.
10724 Assume that general-purpose registers are 64 bits wide.
10728 Assume that floating-point registers are 32 bits wide.
10732 Assume that floating-point registers are 64 bits wide.
10735 @opindex mhard-float
10736 Use floating-point coprocessor instructions.
10739 @opindex msoft-float
10740 Do not use floating-point coprocessor instructions. Implement
10741 floating-point calculations using library calls instead.
10743 @item -msingle-float
10744 @opindex msingle-float
10745 Assume that the floating-point coprocessor only supports single-precision
10748 @itemx -mdouble-float
10749 @opindex mdouble-float
10750 Assume that the floating-point coprocessor supports double-precision
10751 operations. This is the default.
10757 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10759 @itemx -mpaired-single
10760 @itemx -mno-paired-single
10761 @opindex mpaired-single
10762 @opindex mno-paired-single
10763 Use (do not use) paired-single floating-point instructions.
10764 @xref{MIPS Paired-Single Support}. This option can only be used
10765 when generating 64-bit code and requires hardware floating-point
10766 support to be enabled.
10771 @opindex mno-mips3d
10772 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10773 The option @option{-mips3d} implies @option{-mpaired-single}.
10777 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10778 an explanation of the default and the way that the pointer size is
10783 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10785 The default size of @code{int}s, @code{long}s and pointers depends on
10786 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10787 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10788 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10789 or the same size as integer registers, whichever is smaller.
10795 Assume (do not assume) that all symbols have 32-bit values, regardless
10796 of the selected ABI@. This option is useful in combination with
10797 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10798 to generate shorter and faster references to symbolic addresses.
10802 @cindex smaller data references (MIPS)
10803 @cindex gp-relative references (MIPS)
10804 Put global and static items less than or equal to @var{num} bytes into
10805 the small data or bss section instead of the normal data or bss section.
10806 This allows the data to be accessed using a single instruction.
10808 All modules should be compiled with the same @option{-G @var{num}}
10811 @item -membedded-data
10812 @itemx -mno-embedded-data
10813 @opindex membedded-data
10814 @opindex mno-embedded-data
10815 Allocate variables to the read-only data section first if possible, then
10816 next in the small data section if possible, otherwise in data. This gives
10817 slightly slower code than the default, but reduces the amount of RAM required
10818 when executing, and thus may be preferred for some embedded systems.
10820 @item -muninit-const-in-rodata
10821 @itemx -mno-uninit-const-in-rodata
10822 @opindex muninit-const-in-rodata
10823 @opindex mno-uninit-const-in-rodata
10824 Put uninitialized @code{const} variables in the read-only data section.
10825 This option is only meaningful in conjunction with @option{-membedded-data}.
10827 @item -msplit-addresses
10828 @itemx -mno-split-addresses
10829 @opindex msplit-addresses
10830 @opindex mno-split-addresses
10831 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10832 relocation operators. This option has been superseded by
10833 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10835 @item -mexplicit-relocs
10836 @itemx -mno-explicit-relocs
10837 @opindex mexplicit-relocs
10838 @opindex mno-explicit-relocs
10839 Use (do not use) assembler relocation operators when dealing with symbolic
10840 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10841 is to use assembler macros instead.
10843 @option{-mexplicit-relocs} is the default if GCC was configured
10844 to use an assembler that supports relocation operators.
10846 @item -mcheck-zero-division
10847 @itemx -mno-check-zero-division
10848 @opindex mcheck-zero-division
10849 @opindex mno-check-zero-division
10850 Trap (do not trap) on integer division by zero. The default is
10851 @option{-mcheck-zero-division}.
10853 @item -mdivide-traps
10854 @itemx -mdivide-breaks
10855 @opindex mdivide-traps
10856 @opindex mdivide-breaks
10857 MIPS systems check for division by zero by generating either a
10858 conditional trap or a break instruction. Using traps results in
10859 smaller code, but is only supported on MIPS II and later. Also, some
10860 versions of the Linux kernel have a bug that prevents trap from
10861 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10862 allow conditional traps on architectures that support them and
10863 @option{-mdivide-breaks} to force the use of breaks.
10865 The default is usually @option{-mdivide-traps}, but this can be
10866 overridden at configure time using @option{--with-divide=breaks}.
10867 Divide-by-zero checks can be completely disabled using
10868 @option{-mno-check-zero-division}.
10873 @opindex mno-memcpy
10874 Force (do not force) the use of @code{memcpy()} for non-trivial block
10875 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10876 most constant-sized copies.
10879 @itemx -mno-long-calls
10880 @opindex mlong-calls
10881 @opindex mno-long-calls
10882 Disable (do not disable) use of the @code{jal} instruction. Calling
10883 functions using @code{jal} is more efficient but requires the caller
10884 and callee to be in the same 256 megabyte segment.
10886 This option has no effect on abicalls code. The default is
10887 @option{-mno-long-calls}.
10893 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10894 instructions, as provided by the R4650 ISA@.
10897 @itemx -mno-fused-madd
10898 @opindex mfused-madd
10899 @opindex mno-fused-madd
10900 Enable (disable) use of the floating point multiply-accumulate
10901 instructions, when they are available. The default is
10902 @option{-mfused-madd}.
10904 When multiply-accumulate instructions are used, the intermediate
10905 product is calculated to infinite precision and is not subject to
10906 the FCSR Flush to Zero bit. This may be undesirable in some
10911 Tell the MIPS assembler to not run its preprocessor over user
10912 assembler files (with a @samp{.s} suffix) when assembling them.
10915 @itemx -mno-fix-r4000
10916 @opindex mfix-r4000
10917 @opindex mno-fix-r4000
10918 Work around certain R4000 CPU errata:
10921 A double-word or a variable shift may give an incorrect result if executed
10922 immediately after starting an integer division.
10924 A double-word or a variable shift may give an incorrect result if executed
10925 while an integer multiplication is in progress.
10927 An integer division may give an incorrect result if started in a delay slot
10928 of a taken branch or a jump.
10932 @itemx -mno-fix-r4400
10933 @opindex mfix-r4400
10934 @opindex mno-fix-r4400
10935 Work around certain R4400 CPU errata:
10938 A double-word or a variable shift may give an incorrect result if executed
10939 immediately after starting an integer division.
10943 @itemx -mno-fix-vr4120
10944 @opindex mfix-vr4120
10945 Work around certain VR4120 errata:
10948 @code{dmultu} does not always produce the correct result.
10950 @code{div} and @code{ddiv} do not always produce the correct result if one
10951 of the operands is negative.
10953 The workarounds for the division errata rely on special functions in
10954 @file{libgcc.a}. At present, these functions are only provided by
10955 the @code{mips64vr*-elf} configurations.
10957 Other VR4120 errata require a nop to be inserted between certain pairs of
10958 instructions. These errata are handled by the assembler, not by GCC itself.
10961 @opindex mfix-vr4130
10962 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10963 workarounds are implemented by the assembler rather than by GCC,
10964 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10965 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10966 instructions are available instead.
10969 @itemx -mno-fix-sb1
10971 Work around certain SB-1 CPU core errata.
10972 (This flag currently works around the SB-1 revision 2
10973 ``F1'' and ``F2'' floating point errata.)
10975 @item -mflush-func=@var{func}
10976 @itemx -mno-flush-func
10977 @opindex mflush-func
10978 Specifies the function to call to flush the I and D caches, or to not
10979 call any such function. If called, the function must take the same
10980 arguments as the common @code{_flush_func()}, that is, the address of the
10981 memory range for which the cache is being flushed, the size of the
10982 memory range, and the number 3 (to flush both caches). The default
10983 depends on the target GCC was configured for, but commonly is either
10984 @samp{_flush_func} or @samp{__cpu_flush}.
10986 @item -mbranch-likely
10987 @itemx -mno-branch-likely
10988 @opindex mbranch-likely
10989 @opindex mno-branch-likely
10990 Enable or disable use of Branch Likely instructions, regardless of the
10991 default for the selected architecture. By default, Branch Likely
10992 instructions may be generated if they are supported by the selected
10993 architecture. An exception is for the MIPS32 and MIPS64 architectures
10994 and processors which implement those architectures; for those, Branch
10995 Likely instructions will not be generated by default because the MIPS32
10996 and MIPS64 architectures specifically deprecate their use.
10998 @item -mfp-exceptions
10999 @itemx -mno-fp-exceptions
11000 @opindex mfp-exceptions
11001 Specifies whether FP exceptions are enabled. This affects how we schedule
11002 FP instructions for some processors. The default is that FP exceptions are
11005 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
11006 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
11009 @item -mvr4130-align
11010 @itemx -mno-vr4130-align
11011 @opindex mvr4130-align
11012 The VR4130 pipeline is two-way superscalar, but can only issue two
11013 instructions together if the first one is 8-byte aligned. When this
11014 option is enabled, GCC will align pairs of instructions that it
11015 thinks should execute in parallel.
11017 This option only has an effect when optimizing for the VR4130.
11018 It normally makes code faster, but at the expense of making it bigger.
11019 It is enabled by default at optimization level @option{-O3}.
11023 @subsection MMIX Options
11024 @cindex MMIX Options
11026 These options are defined for the MMIX:
11030 @itemx -mno-libfuncs
11032 @opindex mno-libfuncs
11033 Specify that intrinsic library functions are being compiled, passing all
11034 values in registers, no matter the size.
11037 @itemx -mno-epsilon
11039 @opindex mno-epsilon
11040 Generate floating-point comparison instructions that compare with respect
11041 to the @code{rE} epsilon register.
11043 @item -mabi=mmixware
11045 @opindex mabi-mmixware
11047 Generate code that passes function parameters and return values that (in
11048 the called function) are seen as registers @code{$0} and up, as opposed to
11049 the GNU ABI which uses global registers @code{$231} and up.
11051 @item -mzero-extend
11052 @itemx -mno-zero-extend
11053 @opindex mzero-extend
11054 @opindex mno-zero-extend
11055 When reading data from memory in sizes shorter than 64 bits, use (do not
11056 use) zero-extending load instructions by default, rather than
11057 sign-extending ones.
11060 @itemx -mno-knuthdiv
11062 @opindex mno-knuthdiv
11063 Make the result of a division yielding a remainder have the same sign as
11064 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
11065 remainder follows the sign of the dividend. Both methods are
11066 arithmetically valid, the latter being almost exclusively used.
11068 @item -mtoplevel-symbols
11069 @itemx -mno-toplevel-symbols
11070 @opindex mtoplevel-symbols
11071 @opindex mno-toplevel-symbols
11072 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
11073 code can be used with the @code{PREFIX} assembly directive.
11077 Generate an executable in the ELF format, rather than the default
11078 @samp{mmo} format used by the @command{mmix} simulator.
11080 @item -mbranch-predict
11081 @itemx -mno-branch-predict
11082 @opindex mbranch-predict
11083 @opindex mno-branch-predict
11084 Use (do not use) the probable-branch instructions, when static branch
11085 prediction indicates a probable branch.
11087 @item -mbase-addresses
11088 @itemx -mno-base-addresses
11089 @opindex mbase-addresses
11090 @opindex mno-base-addresses
11091 Generate (do not generate) code that uses @emph{base addresses}. Using a
11092 base address automatically generates a request (handled by the assembler
11093 and the linker) for a constant to be set up in a global register. The
11094 register is used for one or more base address requests within the range 0
11095 to 255 from the value held in the register. The generally leads to short
11096 and fast code, but the number of different data items that can be
11097 addressed is limited. This means that a program that uses lots of static
11098 data may require @option{-mno-base-addresses}.
11100 @item -msingle-exit
11101 @itemx -mno-single-exit
11102 @opindex msingle-exit
11103 @opindex mno-single-exit
11104 Force (do not force) generated code to have a single exit point in each
11108 @node MN10300 Options
11109 @subsection MN10300 Options
11110 @cindex MN10300 options
11112 These @option{-m} options are defined for Matsushita MN10300 architectures:
11117 Generate code to avoid bugs in the multiply instructions for the MN10300
11118 processors. This is the default.
11120 @item -mno-mult-bug
11121 @opindex mno-mult-bug
11122 Do not generate code to avoid bugs in the multiply instructions for the
11123 MN10300 processors.
11127 Generate code which uses features specific to the AM33 processor.
11131 Do not generate code which uses features specific to the AM33 processor. This
11134 @item -mreturn-pointer-on-d0
11135 @opindex mreturn-pointer-on-d0
11136 When generating a function which returns a pointer, return the pointer
11137 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
11138 only in a0, and attempts to call such functions without a prototype
11139 would result in errors. Note that this option is on by default; use
11140 @option{-mno-return-pointer-on-d0} to disable it.
11144 Do not link in the C run-time initialization object file.
11148 Indicate to the linker that it should perform a relaxation optimization pass
11149 to shorten branches, calls and absolute memory addresses. This option only
11150 has an effect when used on the command line for the final link step.
11152 This option makes symbolic debugging impossible.
11156 @subsection MT Options
11159 These @option{-m} options are defined for Morpho MT architectures:
11163 @item -march=@var{cpu-type}
11165 Generate code that will run on @var{cpu-type}, which is the name of a system
11166 representing a certain processor type. Possible values for
11167 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
11168 @samp{ms1-16-003} and @samp{ms2}.
11170 When this option is not used, the default is @option{-march=ms1-16-002}.
11174 Use byte loads and stores when generating code.
11178 Do not use byte loads and stores when generating code.
11182 Use simulator runtime
11186 Do not link in the C run-time initialization object file
11187 @file{crti.o}. Other run-time initialization and termination files
11188 such as @file{startup.o} and @file{exit.o} are still included on the
11189 linker command line.
11193 @node PDP-11 Options
11194 @subsection PDP-11 Options
11195 @cindex PDP-11 Options
11197 These options are defined for the PDP-11:
11202 Use hardware FPP floating point. This is the default. (FIS floating
11203 point on the PDP-11/40 is not supported.)
11206 @opindex msoft-float
11207 Do not use hardware floating point.
11211 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
11215 Return floating-point results in memory. This is the default.
11219 Generate code for a PDP-11/40.
11223 Generate code for a PDP-11/45. This is the default.
11227 Generate code for a PDP-11/10.
11229 @item -mbcopy-builtin
11230 @opindex bcopy-builtin
11231 Use inline @code{movmemhi} patterns for copying memory. This is the
11236 Do not use inline @code{movmemhi} patterns for copying memory.
11242 Use 16-bit @code{int}. This is the default.
11248 Use 32-bit @code{int}.
11251 @itemx -mno-float32
11253 @opindex mno-float32
11254 Use 64-bit @code{float}. This is the default.
11257 @itemx -mno-float64
11259 @opindex mno-float64
11260 Use 32-bit @code{float}.
11264 Use @code{abshi2} pattern. This is the default.
11268 Do not use @code{abshi2} pattern.
11270 @item -mbranch-expensive
11271 @opindex mbranch-expensive
11272 Pretend that branches are expensive. This is for experimenting with
11273 code generation only.
11275 @item -mbranch-cheap
11276 @opindex mbranch-cheap
11277 Do not pretend that branches are expensive. This is the default.
11281 Generate code for a system with split I&D@.
11285 Generate code for a system without split I&D@. This is the default.
11289 Use Unix assembler syntax. This is the default when configured for
11290 @samp{pdp11-*-bsd}.
11294 Use DEC assembler syntax. This is the default when configured for any
11295 PDP-11 target other than @samp{pdp11-*-bsd}.
11298 @node PowerPC Options
11299 @subsection PowerPC Options
11300 @cindex PowerPC options
11302 These are listed under @xref{RS/6000 and PowerPC Options}.
11304 @node RS/6000 and PowerPC Options
11305 @subsection IBM RS/6000 and PowerPC Options
11306 @cindex RS/6000 and PowerPC Options
11307 @cindex IBM RS/6000 and PowerPC Options
11309 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
11316 @itemx -mno-powerpc
11317 @itemx -mpowerpc-gpopt
11318 @itemx -mno-powerpc-gpopt
11319 @itemx -mpowerpc-gfxopt
11320 @itemx -mno-powerpc-gfxopt
11322 @itemx -mno-powerpc64
11326 @itemx -mno-popcntb
11334 @opindex mno-power2
11336 @opindex mno-powerpc
11337 @opindex mpowerpc-gpopt
11338 @opindex mno-powerpc-gpopt
11339 @opindex mpowerpc-gfxopt
11340 @opindex mno-powerpc-gfxopt
11341 @opindex mpowerpc64
11342 @opindex mno-powerpc64
11346 @opindex mno-popcntb
11350 @opindex mno-mfpgpr
11351 GCC supports two related instruction set architectures for the
11352 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
11353 instructions supported by the @samp{rios} chip set used in the original
11354 RS/6000 systems and the @dfn{PowerPC} instruction set is the
11355 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
11356 the IBM 4xx, 6xx, and follow-on microprocessors.
11358 Neither architecture is a subset of the other. However there is a
11359 large common subset of instructions supported by both. An MQ
11360 register is included in processors supporting the POWER architecture.
11362 You use these options to specify which instructions are available on the
11363 processor you are using. The default value of these options is
11364 determined when configuring GCC@. Specifying the
11365 @option{-mcpu=@var{cpu_type}} overrides the specification of these
11366 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
11367 rather than the options listed above.
11369 The @option{-mpower} option allows GCC to generate instructions that
11370 are found only in the POWER architecture and to use the MQ register.
11371 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
11372 to generate instructions that are present in the POWER2 architecture but
11373 not the original POWER architecture.
11375 The @option{-mpowerpc} option allows GCC to generate instructions that
11376 are found only in the 32-bit subset of the PowerPC architecture.
11377 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
11378 GCC to use the optional PowerPC architecture instructions in the
11379 General Purpose group, including floating-point square root. Specifying
11380 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
11381 use the optional PowerPC architecture instructions in the Graphics
11382 group, including floating-point select.
11384 The @option{-mmfcrf} option allows GCC to generate the move from
11385 condition register field instruction implemented on the POWER4
11386 processor and other processors that support the PowerPC V2.01
11388 The @option{-mpopcntb} option allows GCC to generate the popcount and
11389 double precision FP reciprocal estimate instruction implemented on the
11390 POWER5 processor and other processors that support the PowerPC V2.02
11392 The @option{-mfprnd} option allows GCC to generate the FP round to
11393 integer instructions implemented on the POWER5+ processor and other
11394 processors that support the PowerPC V2.03 architecture.
11395 The @option{-mmfpgpr} option allows GCC to generate the FP move to/from
11396 general purpose register instructions implemented on the POWER6X
11397 processor and other processors that support the extended PowerPC V2.05
11400 The @option{-mpowerpc64} option allows GCC to generate the additional
11401 64-bit instructions that are found in the full PowerPC64 architecture
11402 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
11403 @option{-mno-powerpc64}.
11405 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
11406 will use only the instructions in the common subset of both
11407 architectures plus some special AIX common-mode calls, and will not use
11408 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
11409 permits GCC to use any instruction from either architecture and to
11410 allow use of the MQ register; specify this for the Motorola MPC601.
11412 @item -mnew-mnemonics
11413 @itemx -mold-mnemonics
11414 @opindex mnew-mnemonics
11415 @opindex mold-mnemonics
11416 Select which mnemonics to use in the generated assembler code. With
11417 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
11418 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
11419 assembler mnemonics defined for the POWER architecture. Instructions
11420 defined in only one architecture have only one mnemonic; GCC uses that
11421 mnemonic irrespective of which of these options is specified.
11423 GCC defaults to the mnemonics appropriate for the architecture in
11424 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
11425 value of these option. Unless you are building a cross-compiler, you
11426 should normally not specify either @option{-mnew-mnemonics} or
11427 @option{-mold-mnemonics}, but should instead accept the default.
11429 @item -mcpu=@var{cpu_type}
11431 Set architecture type, register usage, choice of mnemonics, and
11432 instruction scheduling parameters for machine type @var{cpu_type}.
11433 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
11434 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
11435 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
11436 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
11437 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
11438 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
11439 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
11440 @samp{power4}, @samp{power5}, @samp{power5+}, @samp{power6},
11441 @samp{power6x}, @samp{common}, @samp{powerpc}, @samp{powerpc64},
11442 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
11444 @option{-mcpu=common} selects a completely generic processor. Code
11445 generated under this option will run on any POWER or PowerPC processor.
11446 GCC will use only the instructions in the common subset of both
11447 architectures, and will not use the MQ register. GCC assumes a generic
11448 processor model for scheduling purposes.
11450 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
11451 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
11452 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
11453 types, with an appropriate, generic processor model assumed for
11454 scheduling purposes.
11456 The other options specify a specific processor. Code generated under
11457 those options will run best on that processor, and may not run at all on
11460 The @option{-mcpu} options automatically enable or disable the
11461 following options: @option{-maltivec}, @option{-mfprnd},
11462 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
11463 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
11464 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
11465 @option{-mpowerpc-gfxopt}, @option{-mstring}, @option{-mmulhw},
11466 @option{-mdlmzb}, @option{-mmfpgpr}.
11467 The particular options set for any particular CPU will vary between
11468 compiler versions, depending on what setting seems to produce optimal
11469 code for that CPU; it doesn't necessarily reflect the actual hardware's
11470 capabilities. If you wish to set an individual option to a particular
11471 value, you may specify it after the @option{-mcpu} option, like
11472 @samp{-mcpu=970 -mno-altivec}.
11474 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
11475 not enabled or disabled by the @option{-mcpu} option at present because
11476 AIX does not have full support for these options. You may still
11477 enable or disable them individually if you're sure it'll work in your
11480 @item -mtune=@var{cpu_type}
11482 Set the instruction scheduling parameters for machine type
11483 @var{cpu_type}, but do not set the architecture type, register usage, or
11484 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
11485 values for @var{cpu_type} are used for @option{-mtune} as for
11486 @option{-mcpu}. If both are specified, the code generated will use the
11487 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11488 scheduling parameters set by @option{-mtune}.
11494 Generate code to compute division as reciprocal estimate and iterative
11495 refinement, creating opportunities for increased throughput. This
11496 feature requires: optional PowerPC Graphics instruction set for single
11497 precision and FRE instruction for double precision, assuming divides
11498 cannot generate user-visible traps, and the domain values not include
11499 Infinities, denormals or zero denominator.
11502 @itemx -mno-altivec
11504 @opindex mno-altivec
11505 Generate code that uses (does not use) AltiVec instructions, and also
11506 enable the use of built-in functions that allow more direct access to
11507 the AltiVec instruction set. You may also need to set
11508 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11514 @opindex mno-vrsave
11515 Generate VRSAVE instructions when generating AltiVec code.
11518 @opindex msecure-plt
11519 Generate code that allows ld and ld.so to build executables and shared
11520 libraries with non-exec .plt and .got sections. This is a PowerPC
11521 32-bit SYSV ABI option.
11525 Generate code that uses a BSS .plt section that ld.so fills in, and
11526 requires .plt and .got sections that are both writable and executable.
11527 This is a PowerPC 32-bit SYSV ABI option.
11533 This switch enables or disables the generation of ISEL instructions.
11535 @item -misel=@var{yes/no}
11536 This switch has been deprecated. Use @option{-misel} and
11537 @option{-mno-isel} instead.
11543 This switch enables or disables the generation of SPE simd
11546 @item -mspe=@var{yes/no}
11547 This option has been deprecated. Use @option{-mspe} and
11548 @option{-mno-spe} instead.
11550 @item -mfloat-gprs=@var{yes/single/double/no}
11551 @itemx -mfloat-gprs
11552 @opindex mfloat-gprs
11553 This switch enables or disables the generation of floating point
11554 operations on the general purpose registers for architectures that
11557 The argument @var{yes} or @var{single} enables the use of
11558 single-precision floating point operations.
11560 The argument @var{double} enables the use of single and
11561 double-precision floating point operations.
11563 The argument @var{no} disables floating point operations on the
11564 general purpose registers.
11566 This option is currently only available on the MPC854x.
11572 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11573 targets (including GNU/Linux). The 32-bit environment sets int, long
11574 and pointer to 32 bits and generates code that runs on any PowerPC
11575 variant. The 64-bit environment sets int to 32 bits and long and
11576 pointer to 64 bits, and generates code for PowerPC64, as for
11577 @option{-mpowerpc64}.
11580 @itemx -mno-fp-in-toc
11581 @itemx -mno-sum-in-toc
11582 @itemx -mminimal-toc
11584 @opindex mno-fp-in-toc
11585 @opindex mno-sum-in-toc
11586 @opindex mminimal-toc
11587 Modify generation of the TOC (Table Of Contents), which is created for
11588 every executable file. The @option{-mfull-toc} option is selected by
11589 default. In that case, GCC will allocate at least one TOC entry for
11590 each unique non-automatic variable reference in your program. GCC
11591 will also place floating-point constants in the TOC@. However, only
11592 16,384 entries are available in the TOC@.
11594 If you receive a linker error message that saying you have overflowed
11595 the available TOC space, you can reduce the amount of TOC space used
11596 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11597 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11598 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11599 generate code to calculate the sum of an address and a constant at
11600 run-time instead of putting that sum into the TOC@. You may specify one
11601 or both of these options. Each causes GCC to produce very slightly
11602 slower and larger code at the expense of conserving TOC space.
11604 If you still run out of space in the TOC even when you specify both of
11605 these options, specify @option{-mminimal-toc} instead. This option causes
11606 GCC to make only one TOC entry for every file. When you specify this
11607 option, GCC will produce code that is slower and larger but which
11608 uses extremely little TOC space. You may wish to use this option
11609 only on files that contain less frequently executed code.
11615 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11616 @code{long} type, and the infrastructure needed to support them.
11617 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11618 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11619 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11622 @itemx -mno-xl-compat
11623 @opindex mxl-compat
11624 @opindex mno-xl-compat
11625 Produce code that conforms more closely to IBM XL compiler semantics
11626 when using AIX-compatible ABI. Pass floating-point arguments to
11627 prototyped functions beyond the register save area (RSA) on the stack
11628 in addition to argument FPRs. Do not assume that most significant
11629 double in 128-bit long double value is properly rounded when comparing
11630 values and converting to double. Use XL symbol names for long double
11633 The AIX calling convention was extended but not initially documented to
11634 handle an obscure K&R C case of calling a function that takes the
11635 address of its arguments with fewer arguments than declared. IBM XL
11636 compilers access floating point arguments which do not fit in the
11637 RSA from the stack when a subroutine is compiled without
11638 optimization. Because always storing floating-point arguments on the
11639 stack is inefficient and rarely needed, this option is not enabled by
11640 default and only is necessary when calling subroutines compiled by IBM
11641 XL compilers without optimization.
11645 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11646 application written to use message passing with special startup code to
11647 enable the application to run. The system must have PE installed in the
11648 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11649 must be overridden with the @option{-specs=} option to specify the
11650 appropriate directory location. The Parallel Environment does not
11651 support threads, so the @option{-mpe} option and the @option{-pthread}
11652 option are incompatible.
11654 @item -malign-natural
11655 @itemx -malign-power
11656 @opindex malign-natural
11657 @opindex malign-power
11658 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11659 @option{-malign-natural} overrides the ABI-defined alignment of larger
11660 types, such as floating-point doubles, on their natural size-based boundary.
11661 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11662 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11664 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11668 @itemx -mhard-float
11669 @opindex msoft-float
11670 @opindex mhard-float
11671 Generate code that does not use (uses) the floating-point register set.
11672 Software floating point emulation is provided if you use the
11673 @option{-msoft-float} option, and pass the option to GCC when linking.
11676 @itemx -mno-multiple
11678 @opindex mno-multiple
11679 Generate code that uses (does not use) the load multiple word
11680 instructions and the store multiple word instructions. These
11681 instructions are generated by default on POWER systems, and not
11682 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11683 endian PowerPC systems, since those instructions do not work when the
11684 processor is in little endian mode. The exceptions are PPC740 and
11685 PPC750 which permit the instructions usage in little endian mode.
11690 @opindex mno-string
11691 Generate code that uses (does not use) the load string instructions
11692 and the store string word instructions to save multiple registers and
11693 do small block moves. These instructions are generated by default on
11694 POWER systems, and not generated on PowerPC systems. Do not use
11695 @option{-mstring} on little endian PowerPC systems, since those
11696 instructions do not work when the processor is in little endian mode.
11697 The exceptions are PPC740 and PPC750 which permit the instructions
11698 usage in little endian mode.
11703 @opindex mno-update
11704 Generate code that uses (does not use) the load or store instructions
11705 that update the base register to the address of the calculated memory
11706 location. These instructions are generated by default. If you use
11707 @option{-mno-update}, there is a small window between the time that the
11708 stack pointer is updated and the address of the previous frame is
11709 stored, which means code that walks the stack frame across interrupts or
11710 signals may get corrupted data.
11713 @itemx -mno-fused-madd
11714 @opindex mfused-madd
11715 @opindex mno-fused-madd
11716 Generate code that uses (does not use) the floating point multiply and
11717 accumulate instructions. These instructions are generated by default if
11718 hardware floating is used.
11724 Generate code that uses (does not use) the half-word multiply and
11725 multiply-accumulate instructions on the IBM 405 and 440 processors.
11726 These instructions are generated by default when targetting those
11733 Generate code that uses (does not use) the string-search @samp{dlmzb}
11734 instruction on the IBM 405 and 440 processors. This instruction is
11735 generated by default when targetting those processors.
11737 @item -mno-bit-align
11739 @opindex mno-bit-align
11740 @opindex mbit-align
11741 On System V.4 and embedded PowerPC systems do not (do) force structures
11742 and unions that contain bit-fields to be aligned to the base type of the
11745 For example, by default a structure containing nothing but 8
11746 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11747 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11748 the structure would be aligned to a 1 byte boundary and be one byte in
11751 @item -mno-strict-align
11752 @itemx -mstrict-align
11753 @opindex mno-strict-align
11754 @opindex mstrict-align
11755 On System V.4 and embedded PowerPC systems do not (do) assume that
11756 unaligned memory references will be handled by the system.
11758 @item -mrelocatable
11759 @itemx -mno-relocatable
11760 @opindex mrelocatable
11761 @opindex mno-relocatable
11762 On embedded PowerPC systems generate code that allows (does not allow)
11763 the program to be relocated to a different address at runtime. If you
11764 use @option{-mrelocatable} on any module, all objects linked together must
11765 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11767 @item -mrelocatable-lib
11768 @itemx -mno-relocatable-lib
11769 @opindex mrelocatable-lib
11770 @opindex mno-relocatable-lib
11771 On embedded PowerPC systems generate code that allows (does not allow)
11772 the program to be relocated to a different address at runtime. Modules
11773 compiled with @option{-mrelocatable-lib} can be linked with either modules
11774 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11775 with modules compiled with the @option{-mrelocatable} options.
11781 On System V.4 and embedded PowerPC systems do not (do) assume that
11782 register 2 contains a pointer to a global area pointing to the addresses
11783 used in the program.
11786 @itemx -mlittle-endian
11788 @opindex mlittle-endian
11789 On System V.4 and embedded PowerPC systems compile code for the
11790 processor in little endian mode. The @option{-mlittle-endian} option is
11791 the same as @option{-mlittle}.
11794 @itemx -mbig-endian
11796 @opindex mbig-endian
11797 On System V.4 and embedded PowerPC systems compile code for the
11798 processor in big endian mode. The @option{-mbig-endian} option is
11799 the same as @option{-mbig}.
11801 @item -mdynamic-no-pic
11802 @opindex mdynamic-no-pic
11803 On Darwin and Mac OS X systems, compile code so that it is not
11804 relocatable, but that its external references are relocatable. The
11805 resulting code is suitable for applications, but not shared
11808 @item -mprioritize-restricted-insns=@var{priority}
11809 @opindex mprioritize-restricted-insns
11810 This option controls the priority that is assigned to
11811 dispatch-slot restricted instructions during the second scheduling
11812 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11813 @var{no/highest/second-highest} priority to dispatch slot restricted
11816 @item -msched-costly-dep=@var{dependence_type}
11817 @opindex msched-costly-dep
11818 This option controls which dependences are considered costly
11819 by the target during instruction scheduling. The argument
11820 @var{dependence_type} takes one of the following values:
11821 @var{no}: no dependence is costly,
11822 @var{all}: all dependences are costly,
11823 @var{true_store_to_load}: a true dependence from store to load is costly,
11824 @var{store_to_load}: any dependence from store to load is costly,
11825 @var{number}: any dependence which latency >= @var{number} is costly.
11827 @item -minsert-sched-nops=@var{scheme}
11828 @opindex minsert-sched-nops
11829 This option controls which nop insertion scheme will be used during
11830 the second scheduling pass. The argument @var{scheme} takes one of the
11832 @var{no}: Don't insert nops.
11833 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11834 according to the scheduler's grouping.
11835 @var{regroup_exact}: Insert nops to force costly dependent insns into
11836 separate groups. Insert exactly as many nops as needed to force an insn
11837 to a new group, according to the estimated processor grouping.
11838 @var{number}: Insert nops to force costly dependent insns into
11839 separate groups. Insert @var{number} nops to force an insn to a new group.
11842 @opindex mcall-sysv
11843 On System V.4 and embedded PowerPC systems compile code using calling
11844 conventions that adheres to the March 1995 draft of the System V
11845 Application Binary Interface, PowerPC processor supplement. This is the
11846 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11848 @item -mcall-sysv-eabi
11849 @opindex mcall-sysv-eabi
11850 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11852 @item -mcall-sysv-noeabi
11853 @opindex mcall-sysv-noeabi
11854 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11856 @item -mcall-solaris
11857 @opindex mcall-solaris
11858 On System V.4 and embedded PowerPC systems compile code for the Solaris
11862 @opindex mcall-linux
11863 On System V.4 and embedded PowerPC systems compile code for the
11864 Linux-based GNU system.
11868 On System V.4 and embedded PowerPC systems compile code for the
11869 Hurd-based GNU system.
11871 @item -mcall-netbsd
11872 @opindex mcall-netbsd
11873 On System V.4 and embedded PowerPC systems compile code for the
11874 NetBSD operating system.
11876 @item -maix-struct-return
11877 @opindex maix-struct-return
11878 Return all structures in memory (as specified by the AIX ABI)@.
11880 @item -msvr4-struct-return
11881 @opindex msvr4-struct-return
11882 Return structures smaller than 8 bytes in registers (as specified by the
11885 @item -mabi=@var{abi-type}
11887 Extend the current ABI with a particular extension, or remove such extension.
11888 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11889 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11893 Extend the current ABI with SPE ABI extensions. This does not change
11894 the default ABI, instead it adds the SPE ABI extensions to the current
11898 @opindex mabi=no-spe
11899 Disable Booke SPE ABI extensions for the current ABI@.
11901 @item -mabi=ibmlongdouble
11902 @opindex mabi=ibmlongdouble
11903 Change the current ABI to use IBM extended precision long double.
11904 This is a PowerPC 32-bit SYSV ABI option.
11906 @item -mabi=ieeelongdouble
11907 @opindex mabi=ieeelongdouble
11908 Change the current ABI to use IEEE extended precision long double.
11909 This is a PowerPC 32-bit Linux ABI option.
11912 @itemx -mno-prototype
11913 @opindex mprototype
11914 @opindex mno-prototype
11915 On System V.4 and embedded PowerPC systems assume that all calls to
11916 variable argument functions are properly prototyped. Otherwise, the
11917 compiler must insert an instruction before every non prototyped call to
11918 set or clear bit 6 of the condition code register (@var{CR}) to
11919 indicate whether floating point values were passed in the floating point
11920 registers in case the function takes a variable arguments. With
11921 @option{-mprototype}, only calls to prototyped variable argument functions
11922 will set or clear the bit.
11926 On embedded PowerPC systems, assume that the startup module is called
11927 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11928 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11933 On embedded PowerPC systems, assume that the startup module is called
11934 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11939 On embedded PowerPC systems, assume that the startup module is called
11940 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11943 @item -myellowknife
11944 @opindex myellowknife
11945 On embedded PowerPC systems, assume that the startup module is called
11946 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11951 On System V.4 and embedded PowerPC systems, specify that you are
11952 compiling for a VxWorks system.
11956 Specify that you are compiling for the WindISS simulation environment.
11960 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11961 header to indicate that @samp{eabi} extended relocations are used.
11967 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11968 Embedded Applications Binary Interface (eabi) which is a set of
11969 modifications to the System V.4 specifications. Selecting @option{-meabi}
11970 means that the stack is aligned to an 8 byte boundary, a function
11971 @code{__eabi} is called to from @code{main} to set up the eabi
11972 environment, and the @option{-msdata} option can use both @code{r2} and
11973 @code{r13} to point to two separate small data areas. Selecting
11974 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11975 do not call an initialization function from @code{main}, and the
11976 @option{-msdata} option will only use @code{r13} to point to a single
11977 small data area. The @option{-meabi} option is on by default if you
11978 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11981 @opindex msdata=eabi
11982 On System V.4 and embedded PowerPC systems, put small initialized
11983 @code{const} global and static data in the @samp{.sdata2} section, which
11984 is pointed to by register @code{r2}. Put small initialized
11985 non-@code{const} global and static data in the @samp{.sdata} section,
11986 which is pointed to by register @code{r13}. Put small uninitialized
11987 global and static data in the @samp{.sbss} section, which is adjacent to
11988 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11989 incompatible with the @option{-mrelocatable} option. The
11990 @option{-msdata=eabi} option also sets the @option{-memb} option.
11993 @opindex msdata=sysv
11994 On System V.4 and embedded PowerPC systems, put small global and static
11995 data in the @samp{.sdata} section, which is pointed to by register
11996 @code{r13}. Put small uninitialized global and static data in the
11997 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11998 The @option{-msdata=sysv} option is incompatible with the
11999 @option{-mrelocatable} option.
12001 @item -msdata=default
12003 @opindex msdata=default
12005 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
12006 compile code the same as @option{-msdata=eabi}, otherwise compile code the
12007 same as @option{-msdata=sysv}.
12010 @opindex msdata-data
12011 On System V.4 and embedded PowerPC systems, put small global
12012 data in the @samp{.sdata} section. Put small uninitialized global
12013 data in the @samp{.sbss} section. Do not use register @code{r13}
12014 to address small data however. This is the default behavior unless
12015 other @option{-msdata} options are used.
12019 @opindex msdata=none
12021 On embedded PowerPC systems, put all initialized global and static data
12022 in the @samp{.data} section, and all uninitialized data in the
12023 @samp{.bss} section.
12027 @cindex smaller data references (PowerPC)
12028 @cindex .sdata/.sdata2 references (PowerPC)
12029 On embedded PowerPC systems, put global and static items less than or
12030 equal to @var{num} bytes into the small data or bss sections instead of
12031 the normal data or bss section. By default, @var{num} is 8. The
12032 @option{-G @var{num}} switch is also passed to the linker.
12033 All modules should be compiled with the same @option{-G @var{num}} value.
12036 @itemx -mno-regnames
12038 @opindex mno-regnames
12039 On System V.4 and embedded PowerPC systems do (do not) emit register
12040 names in the assembly language output using symbolic forms.
12043 @itemx -mno-longcall
12045 @opindex mno-longcall
12046 By default assume that all calls are far away so that a longer more
12047 expensive calling sequence is required. This is required for calls
12048 further than 32 megabytes (33,554,432 bytes) from the current location.
12049 A short call will be generated if the compiler knows
12050 the call cannot be that far away. This setting can be overridden by
12051 the @code{shortcall} function attribute, or by @code{#pragma
12054 Some linkers are capable of detecting out-of-range calls and generating
12055 glue code on the fly. On these systems, long calls are unnecessary and
12056 generate slower code. As of this writing, the AIX linker can do this,
12057 as can the GNU linker for PowerPC/64. It is planned to add this feature
12058 to the GNU linker for 32-bit PowerPC systems as well.
12060 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
12061 callee, L42'', plus a ``branch island'' (glue code). The two target
12062 addresses represent the callee and the ``branch island''. The
12063 Darwin/PPC linker will prefer the first address and generate a ``bl
12064 callee'' if the PPC ``bl'' instruction will reach the callee directly;
12065 otherwise, the linker will generate ``bl L42'' to call the ``branch
12066 island''. The ``branch island'' is appended to the body of the
12067 calling function; it computes the full 32-bit address of the callee
12070 On Mach-O (Darwin) systems, this option directs the compiler emit to
12071 the glue for every direct call, and the Darwin linker decides whether
12072 to use or discard it.
12074 In the future, we may cause GCC to ignore all longcall specifications
12075 when the linker is known to generate glue.
12079 Adds support for multithreading with the @dfn{pthreads} library.
12080 This option sets flags for both the preprocessor and linker.
12084 @node S/390 and zSeries Options
12085 @subsection S/390 and zSeries Options
12086 @cindex S/390 and zSeries Options
12088 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
12092 @itemx -msoft-float
12093 @opindex mhard-float
12094 @opindex msoft-float
12095 Use (do not use) the hardware floating-point instructions and registers
12096 for floating-point operations. When @option{-msoft-float} is specified,
12097 functions in @file{libgcc.a} will be used to perform floating-point
12098 operations. When @option{-mhard-float} is specified, the compiler
12099 generates IEEE floating-point instructions. This is the default.
12101 @item -mlong-double-64
12102 @itemx -mlong-double-128
12103 @opindex mlong-double-64
12104 @opindex mlong-double-128
12105 These switches control the size of @code{long double} type. A size
12106 of 64bit makes the @code{long double} type equivalent to the @code{double}
12107 type. This is the default.
12110 @itemx -mno-backchain
12111 @opindex mbackchain
12112 @opindex mno-backchain
12113 Store (do not store) the address of the caller's frame as backchain pointer
12114 into the callee's stack frame.
12115 A backchain may be needed to allow debugging using tools that do not understand
12116 DWARF-2 call frame information.
12117 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
12118 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
12119 the backchain is placed into the topmost word of the 96/160 byte register
12122 In general, code compiled with @option{-mbackchain} is call-compatible with
12123 code compiled with @option{-mmo-backchain}; however, use of the backchain
12124 for debugging purposes usually requires that the whole binary is built with
12125 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
12126 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12127 to build a linux kernel use @option{-msoft-float}.
12129 The default is to not maintain the backchain.
12131 @item -mpacked-stack
12132 @item -mno-packed-stack
12133 @opindex mpacked-stack
12134 @opindex mno-packed-stack
12135 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
12136 specified, the compiler uses the all fields of the 96/160 byte register save
12137 area only for their default purpose; unused fields still take up stack space.
12138 When @option{-mpacked-stack} is specified, register save slots are densely
12139 packed at the top of the register save area; unused space is reused for other
12140 purposes, allowing for more efficient use of the available stack space.
12141 However, when @option{-mbackchain} is also in effect, the topmost word of
12142 the save area is always used to store the backchain, and the return address
12143 register is always saved two words below the backchain.
12145 As long as the stack frame backchain is not used, code generated with
12146 @option{-mpacked-stack} is call-compatible with code generated with
12147 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
12148 S/390 or zSeries generated code that uses the stack frame backchain at run
12149 time, not just for debugging purposes. Such code is not call-compatible
12150 with code compiled with @option{-mpacked-stack}. Also, note that the
12151 combination of @option{-mbackchain},
12152 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
12153 to build a linux kernel use @option{-msoft-float}.
12155 The default is to not use the packed stack layout.
12158 @itemx -mno-small-exec
12159 @opindex msmall-exec
12160 @opindex mno-small-exec
12161 Generate (or do not generate) code using the @code{bras} instruction
12162 to do subroutine calls.
12163 This only works reliably if the total executable size does not
12164 exceed 64k. The default is to use the @code{basr} instruction instead,
12165 which does not have this limitation.
12171 When @option{-m31} is specified, generate code compliant to the
12172 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
12173 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
12174 particular to generate 64-bit instructions. For the @samp{s390}
12175 targets, the default is @option{-m31}, while the @samp{s390x}
12176 targets default to @option{-m64}.
12182 When @option{-mzarch} is specified, generate code using the
12183 instructions available on z/Architecture.
12184 When @option{-mesa} is specified, generate code using the
12185 instructions available on ESA/390. Note that @option{-mesa} is
12186 not possible with @option{-m64}.
12187 When generating code compliant to the GNU/Linux for S/390 ABI,
12188 the default is @option{-mesa}. When generating code compliant
12189 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
12195 Generate (or do not generate) code using the @code{mvcle} instruction
12196 to perform block moves. When @option{-mno-mvcle} is specified,
12197 use a @code{mvc} loop instead. This is the default unless optimizing for
12204 Print (or do not print) additional debug information when compiling.
12205 The default is to not print debug information.
12207 @item -march=@var{cpu-type}
12209 Generate code that will run on @var{cpu-type}, which is the name of a system
12210 representing a certain processor type. Possible values for
12211 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
12212 When generating code using the instructions available on z/Architecture,
12213 the default is @option{-march=z900}. Otherwise, the default is
12214 @option{-march=g5}.
12216 @item -mtune=@var{cpu-type}
12218 Tune to @var{cpu-type} everything applicable about the generated code,
12219 except for the ABI and the set of available instructions.
12220 The list of @var{cpu-type} values is the same as for @option{-march}.
12221 The default is the value used for @option{-march}.
12224 @itemx -mno-tpf-trace
12225 @opindex mtpf-trace
12226 @opindex mno-tpf-trace
12227 Generate code that adds (does not add) in TPF OS specific branches to trace
12228 routines in the operating system. This option is off by default, even
12229 when compiling for the TPF OS@.
12232 @itemx -mno-fused-madd
12233 @opindex mfused-madd
12234 @opindex mno-fused-madd
12235 Generate code that uses (does not use) the floating point multiply and
12236 accumulate instructions. These instructions are generated by default if
12237 hardware floating point is used.
12239 @item -mwarn-framesize=@var{framesize}
12240 @opindex mwarn-framesize
12241 Emit a warning if the current function exceeds the given frame size. Because
12242 this is a compile time check it doesn't need to be a real problem when the program
12243 runs. It is intended to identify functions which most probably cause
12244 a stack overflow. It is useful to be used in an environment with limited stack
12245 size e.g.@: the linux kernel.
12247 @item -mwarn-dynamicstack
12248 @opindex mwarn-dynamicstack
12249 Emit a warning if the function calls alloca or uses dynamically
12250 sized arrays. This is generally a bad idea with a limited stack size.
12252 @item -mstack-guard=@var{stack-guard}
12253 @item -mstack-size=@var{stack-size}
12254 @opindex mstack-guard
12255 @opindex mstack-size
12256 These arguments always have to be used in conjunction. If they are present the s390
12257 back end emits additional instructions in the function prologue which trigger a trap
12258 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
12259 (remember that the stack on s390 grows downward). These options are intended to
12260 be used to help debugging stack overflow problems. The additionally emitted code
12261 causes only little overhead and hence can also be used in production like systems
12262 without greater performance degradation. The given values have to be exact
12263 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
12265 In order to be efficient the extra code makes the assumption that the stack starts
12266 at an address aligned to the value given by @var{stack-size}.
12269 @node Score Options
12270 @subsection Score Options
12271 @cindex Score Options
12273 These options are defined for Score implementations:
12278 Compile code for little endian mode.
12282 Compile code for big endian mode. This is the default.
12286 Enable the use of multiply-accumulate instructions. Disabled by default.
12290 Specify the SCORE5U of the target architecture.
12294 Specify the SCORE7 of the target architecture. This is the default.
12298 @subsection SH Options
12300 These @samp{-m} options are defined for the SH implementations:
12305 Generate code for the SH1.
12309 Generate code for the SH2.
12312 Generate code for the SH2e.
12316 Generate code for the SH3.
12320 Generate code for the SH3e.
12324 Generate code for the SH4 without a floating-point unit.
12326 @item -m4-single-only
12327 @opindex m4-single-only
12328 Generate code for the SH4 with a floating-point unit that only
12329 supports single-precision arithmetic.
12333 Generate code for the SH4 assuming the floating-point unit is in
12334 single-precision mode by default.
12338 Generate code for the SH4.
12342 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
12343 floating-point unit is not used.
12345 @item -m4a-single-only
12346 @opindex m4a-single-only
12347 Generate code for the SH4a, in such a way that no double-precision
12348 floating point operations are used.
12351 @opindex m4a-single
12352 Generate code for the SH4a assuming the floating-point unit is in
12353 single-precision mode by default.
12357 Generate code for the SH4a.
12361 Same as @option{-m4a-nofpu}, except that it implicitly passes
12362 @option{-dsp} to the assembler. GCC doesn't generate any DSP
12363 instructions at the moment.
12367 Compile code for the processor in big endian mode.
12371 Compile code for the processor in little endian mode.
12375 Align doubles at 64-bit boundaries. Note that this changes the calling
12376 conventions, and thus some functions from the standard C library will
12377 not work unless you recompile it first with @option{-mdalign}.
12381 Shorten some address references at link time, when possible; uses the
12382 linker option @option{-relax}.
12386 Use 32-bit offsets in @code{switch} tables. The default is to use
12391 Enable the use of the instruction @code{fmovd}.
12395 Comply with the calling conventions defined by Renesas.
12399 Comply with the calling conventions defined by Renesas.
12403 Comply with the calling conventions defined for GCC before the Renesas
12404 conventions were available. This option is the default for all
12405 targets of the SH toolchain except for @samp{sh-symbianelf}.
12408 @opindex mnomacsave
12409 Mark the @code{MAC} register as call-clobbered, even if
12410 @option{-mhitachi} is given.
12414 Increase IEEE-compliance of floating-point code.
12415 At the moment, this is equivalent to @option{-fno-finite-math-only}.
12416 When generating 16 bit SH opcodes, getting IEEE-conforming results for
12417 comparisons of NANs / infinities incurs extra overhead in every
12418 floating point comparison, therefore the default is set to
12419 @option{-ffinite-math-only}.
12423 Dump instruction size and location in the assembly code.
12426 @opindex mpadstruct
12427 This option is deprecated. It pads structures to multiple of 4 bytes,
12428 which is incompatible with the SH ABI@.
12432 Optimize for space instead of speed. Implied by @option{-Os}.
12435 @opindex mprefergot
12436 When generating position-independent code, emit function calls using
12437 the Global Offset Table instead of the Procedure Linkage Table.
12441 Generate a library function call to invalidate instruction cache
12442 entries, after fixing up a trampoline. This library function call
12443 doesn't assume it can write to the whole memory address space. This
12444 is the default when the target is @code{sh-*-linux*}.
12446 @item -multcost=@var{number}
12447 @opindex multcost=@var{number}
12448 Set the cost to assume for a multiply insn.
12450 @item -mdiv=@var{strategy}
12451 @opindex mdiv=@var{strategy}
12452 Set the division strategy to use for SHmedia code. @var{strategy} must be
12453 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
12454 inv:call2, inv:fp .
12455 "fp" performs the operation in floating point. This has a very high latency,
12456 but needs only a few instructions, so it might be a good choice if
12457 your code has enough easily exploitable ILP to allow the compiler to
12458 schedule the floating point instructions together with other instructions.
12459 Division by zero causes a floating point exception.
12460 "inv" uses integer operations to calculate the inverse of the divisor,
12461 and then multiplies the dividend with the inverse. This strategy allows
12462 cse and hoisting of the inverse calculation. Division by zero calculates
12463 an unspecified result, but does not trap.
12464 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
12465 have been found, or if the entire operation has been hoisted to the same
12466 place, the last stages of the inverse calculation are intertwined with the
12467 final multiply to reduce the overall latency, at the expense of using a few
12468 more instructions, and thus offering fewer scheduling opportunities with
12470 "call" calls a library function that usually implements the inv:minlat
12472 This gives high code density for m5-*media-nofpu compilations.
12473 "call2" uses a different entry point of the same library function, where it
12474 assumes that a pointer to a lookup table has already been set up, which
12475 exposes the pointer load to cse / code hoisting optimizations.
12476 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
12477 code generation, but if the code stays unoptimized, revert to the "call",
12478 "call2", or "fp" strategies, respectively. Note that the
12479 potentially-trapping side effect of division by zero is carried by a
12480 separate instruction, so it is possible that all the integer instructions
12481 are hoisted out, but the marker for the side effect stays where it is.
12482 A recombination to fp operations or a call is not possible in that case.
12483 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
12484 that the inverse calculation was nor separated from the multiply, they speed
12485 up division where the dividend fits into 20 bits (plus sign where applicable),
12486 by inserting a test to skip a number of operations in this case; this test
12487 slows down the case of larger dividends. inv20u assumes the case of a such
12488 a small dividend to be unlikely, and inv20l assumes it to be likely.
12490 @item -mdivsi3_libfunc=@var{name}
12491 @opindex mdivsi3_libfunc=@var{name}
12492 Set the name of the library function used for 32 bit signed division to
12493 @var{name}. This only affect the name used in the call and inv:call
12494 division strategies, and the compiler will still expect the same
12495 sets of input/output/clobbered registers as if this option was not present.
12497 @item -madjust-unroll
12498 @opindex madjust-unroll
12499 Throttle unrolling to avoid thrashing target registers.
12500 This option only has an effect if the gcc code base supports the
12501 TARGET_ADJUST_UNROLL_MAX target hook.
12503 @item -mindexed-addressing
12504 @opindex mindexed-addressing
12505 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
12506 This is only safe if the hardware and/or OS implement 32 bit wrap-around
12507 semantics for the indexed addressing mode. The architecture allows the
12508 implementation of processors with 64 bit MMU, which the OS could use to
12509 get 32 bit addressing, but since no current hardware implementation supports
12510 this or any other way to make the indexed addressing mode safe to use in
12511 the 32 bit ABI, the default is -mno-indexed-addressing.
12513 @item -mgettrcost=@var{number}
12514 @opindex mgettrcost=@var{number}
12515 Set the cost assumed for the gettr instruction to @var{number}.
12516 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
12520 Assume pt* instructions won't trap. This will generally generate better
12521 scheduled code, but is unsafe on current hardware. The current architecture
12522 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
12523 This has the unintentional effect of making it unsafe to schedule ptabs /
12524 ptrel before a branch, or hoist it out of a loop. For example,
12525 __do_global_ctors, a part of libgcc that runs constructors at program
12526 startup, calls functions in a list which is delimited by -1. With the
12527 -mpt-fixed option, the ptabs will be done before testing against -1.
12528 That means that all the constructors will be run a bit quicker, but when
12529 the loop comes to the end of the list, the program crashes because ptabs
12530 loads -1 into a target register. Since this option is unsafe for any
12531 hardware implementing the current architecture specification, the default
12532 is -mno-pt-fixed. Unless the user specifies a specific cost with
12533 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12534 this deters register allocation using target registers for storing
12537 @item -minvalid-symbols
12538 @opindex minvalid-symbols
12539 Assume symbols might be invalid. Ordinary function symbols generated by
12540 the compiler will always be valid to load with movi/shori/ptabs or
12541 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12542 to generate symbols that will cause ptabs / ptrel to trap.
12543 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12544 It will then prevent cross-basic-block cse, hoisting and most scheduling
12545 of symbol loads. The default is @option{-mno-invalid-symbols}.
12548 @node SPARC Options
12549 @subsection SPARC Options
12550 @cindex SPARC options
12552 These @samp{-m} options are supported on the SPARC:
12555 @item -mno-app-regs
12557 @opindex mno-app-regs
12559 Specify @option{-mapp-regs} to generate output using the global registers
12560 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12563 To be fully SVR4 ABI compliant at the cost of some performance loss,
12564 specify @option{-mno-app-regs}. You should compile libraries and system
12565 software with this option.
12568 @itemx -mhard-float
12570 @opindex mhard-float
12571 Generate output containing floating point instructions. This is the
12575 @itemx -msoft-float
12577 @opindex msoft-float
12578 Generate output containing library calls for floating point.
12579 @strong{Warning:} the requisite libraries are not available for all SPARC
12580 targets. Normally the facilities of the machine's usual C compiler are
12581 used, but this cannot be done directly in cross-compilation. You must make
12582 your own arrangements to provide suitable library functions for
12583 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12584 @samp{sparclite-*-*} do provide software floating point support.
12586 @option{-msoft-float} changes the calling convention in the output file;
12587 therefore, it is only useful if you compile @emph{all} of a program with
12588 this option. In particular, you need to compile @file{libgcc.a}, the
12589 library that comes with GCC, with @option{-msoft-float} in order for
12592 @item -mhard-quad-float
12593 @opindex mhard-quad-float
12594 Generate output containing quad-word (long double) floating point
12597 @item -msoft-quad-float
12598 @opindex msoft-quad-float
12599 Generate output containing library calls for quad-word (long double)
12600 floating point instructions. The functions called are those specified
12601 in the SPARC ABI@. This is the default.
12603 As of this writing, there are no SPARC implementations that have hardware
12604 support for the quad-word floating point instructions. They all invoke
12605 a trap handler for one of these instructions, and then the trap handler
12606 emulates the effect of the instruction. Because of the trap handler overhead,
12607 this is much slower than calling the ABI library routines. Thus the
12608 @option{-msoft-quad-float} option is the default.
12610 @item -mno-unaligned-doubles
12611 @itemx -munaligned-doubles
12612 @opindex mno-unaligned-doubles
12613 @opindex munaligned-doubles
12614 Assume that doubles have 8 byte alignment. This is the default.
12616 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12617 alignment only if they are contained in another type, or if they have an
12618 absolute address. Otherwise, it assumes they have 4 byte alignment.
12619 Specifying this option avoids some rare compatibility problems with code
12620 generated by other compilers. It is not the default because it results
12621 in a performance loss, especially for floating point code.
12623 @item -mno-faster-structs
12624 @itemx -mfaster-structs
12625 @opindex mno-faster-structs
12626 @opindex mfaster-structs
12627 With @option{-mfaster-structs}, the compiler assumes that structures
12628 should have 8 byte alignment. This enables the use of pairs of
12629 @code{ldd} and @code{std} instructions for copies in structure
12630 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12631 However, the use of this changed alignment directly violates the SPARC
12632 ABI@. Thus, it's intended only for use on targets where the developer
12633 acknowledges that their resulting code will not be directly in line with
12634 the rules of the ABI@.
12636 @item -mimpure-text
12637 @opindex mimpure-text
12638 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12639 the compiler to not pass @option{-z text} to the linker when linking a
12640 shared object. Using this option, you can link position-dependent
12641 code into a shared object.
12643 @option{-mimpure-text} suppresses the ``relocations remain against
12644 allocatable but non-writable sections'' linker error message.
12645 However, the necessary relocations will trigger copy-on-write, and the
12646 shared object is not actually shared across processes. Instead of
12647 using @option{-mimpure-text}, you should compile all source code with
12648 @option{-fpic} or @option{-fPIC}.
12650 This option is only available on SunOS and Solaris.
12652 @item -mcpu=@var{cpu_type}
12654 Set the instruction set, register set, and instruction scheduling parameters
12655 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12656 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12657 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12658 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc},
12659 @samp{ultrasparc3}, and @samp{niagara}.
12661 Default instruction scheduling parameters are used for values that select
12662 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12663 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12665 Here is a list of each supported architecture and their supported
12670 v8: supersparc, hypersparc
12671 sparclite: f930, f934, sparclite86x
12673 v9: ultrasparc, ultrasparc3, niagara
12676 By default (unless configured otherwise), GCC generates code for the V7
12677 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12678 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12679 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12680 SPARCStation 1, 2, IPX etc.
12682 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12683 architecture. The only difference from V7 code is that the compiler emits
12684 the integer multiply and integer divide instructions which exist in SPARC-V8
12685 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12686 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12689 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12690 the SPARC architecture. This adds the integer multiply, integer divide step
12691 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12692 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12693 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12694 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12695 MB86934 chip, which is the more recent SPARClite with FPU@.
12697 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12698 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12699 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12700 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12701 optimizes it for the TEMIC SPARClet chip.
12703 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12704 architecture. This adds 64-bit integer and floating-point move instructions,
12705 3 additional floating-point condition code registers and conditional move
12706 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12707 optimizes it for the Sun UltraSPARC I/II/IIi chips. With
12708 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12709 Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ chips. With
12710 @option{-mcpu=niagara}, the compiler additionally optimizes it for
12711 Sun UltraSPARC T1 chips.
12713 @item -mtune=@var{cpu_type}
12715 Set the instruction scheduling parameters for machine type
12716 @var{cpu_type}, but do not set the instruction set or register set that the
12717 option @option{-mcpu=@var{cpu_type}} would.
12719 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12720 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12721 that select a particular cpu implementation. Those are @samp{cypress},
12722 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12723 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc},
12724 @samp{ultrasparc3}, and @samp{niagara}.
12729 @opindex mno-v8plus
12730 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12731 difference from the V8 ABI is that the global and out registers are
12732 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12733 mode for all SPARC-V9 processors.
12739 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12740 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12743 These @samp{-m} options are supported in addition to the above
12744 on SPARC-V9 processors in 64-bit environments:
12747 @item -mlittle-endian
12748 @opindex mlittle-endian
12749 Generate code for a processor running in little-endian mode. It is only
12750 available for a few configurations and most notably not on Solaris and Linux.
12756 Generate code for a 32-bit or 64-bit environment.
12757 The 32-bit environment sets int, long and pointer to 32 bits.
12758 The 64-bit environment sets int to 32 bits and long and pointer
12761 @item -mcmodel=medlow
12762 @opindex mcmodel=medlow
12763 Generate code for the Medium/Low code model: 64-bit addresses, programs
12764 must be linked in the low 32 bits of memory. Programs can be statically
12765 or dynamically linked.
12767 @item -mcmodel=medmid
12768 @opindex mcmodel=medmid
12769 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12770 must be linked in the low 44 bits of memory, the text and data segments must
12771 be less than 2GB in size and the data segment must be located within 2GB of
12774 @item -mcmodel=medany
12775 @opindex mcmodel=medany
12776 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12777 may be linked anywhere in memory, the text and data segments must be less
12778 than 2GB in size and the data segment must be located within 2GB of the
12781 @item -mcmodel=embmedany
12782 @opindex mcmodel=embmedany
12783 Generate code for the Medium/Anywhere code model for embedded systems:
12784 64-bit addresses, the text and data segments must be less than 2GB in
12785 size, both starting anywhere in memory (determined at link time). The
12786 global register %g4 points to the base of the data segment. Programs
12787 are statically linked and PIC is not supported.
12790 @itemx -mno-stack-bias
12791 @opindex mstack-bias
12792 @opindex mno-stack-bias
12793 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12794 frame pointer if present, are offset by @minus{}2047 which must be added back
12795 when making stack frame references. This is the default in 64-bit mode.
12796 Otherwise, assume no such offset is present.
12799 These switches are supported in addition to the above on Solaris:
12804 Add support for multithreading using the Solaris threads library. This
12805 option sets flags for both the preprocessor and linker. This option does
12806 not affect the thread safety of object code produced by the compiler or
12807 that of libraries supplied with it.
12811 Add support for multithreading using the POSIX threads library. This
12812 option sets flags for both the preprocessor and linker. This option does
12813 not affect the thread safety of object code produced by the compiler or
12814 that of libraries supplied with it.
12818 This is a synonym for @option{-pthreads}.
12822 @subsection SPU Options
12823 @cindex SPU options
12825 These @samp{-m} options are supported on the SPU:
12829 @itemx -merror-reloc
12830 @opindex mwarn-reloc
12831 @opindex merror-reloc
12833 The loader for SPU does not handle dynamic relocations. By default, GCC
12834 will give an error when it generates code that requires a dynamic
12835 relocation. @option{-mno-error-reloc} disables the error,
12836 @option{-mwarn-reloc} will generate a warning instead.
12839 @itemx -munsafe-dma
12841 @opindex munsafe-dma
12843 Instructions which initiate or test completion of DMA must not be
12844 reordered with respect to loads and stores of the memory which is being
12845 accessed. Users typically address this problem using the volatile
12846 keyword, but that can lead to inefficient code in places where the
12847 memory is known to not change. Rather than mark the memory as volatile
12848 we treat the DMA instructions as potentially effecting all memory. With
12849 @option{-munsafe-dma} users must use the volatile keyword to protect
12852 @item -mbranch-hints
12853 @opindex mbranch-hints
12855 By default, GCC will generate a branch hint instruction to avoid
12856 pipeline stalls for always taken or probably taken branches. A hint
12857 will not be generated closer than 8 instructions away from its branch.
12858 There is little reason to disable them, except for debugging purposes,
12859 or to make an object a little bit smaller.
12863 @opindex msmall-mem
12864 @opindex mlarge-mem
12866 By default, GCC generates code assuming that addresses are never larger
12867 than 18 bits. With @option{-mlarge-mem} code is generated that assumes
12868 a full 32 bit address.
12872 @node System V Options
12873 @subsection Options for System V
12875 These additional options are available on System V Release 4 for
12876 compatibility with other compilers on those systems:
12881 Create a shared object.
12882 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12886 Identify the versions of each tool used by the compiler, in a
12887 @code{.ident} assembler directive in the output.
12891 Refrain from adding @code{.ident} directives to the output file (this is
12894 @item -YP,@var{dirs}
12896 Search the directories @var{dirs}, and no others, for libraries
12897 specified with @option{-l}.
12899 @item -Ym,@var{dir}
12901 Look in the directory @var{dir} to find the M4 preprocessor.
12902 The assembler uses this option.
12903 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12904 @c the generic assembler that comes with Solaris takes just -Ym.
12907 @node TMS320C3x/C4x Options
12908 @subsection TMS320C3x/C4x Options
12909 @cindex TMS320C3x/C4x Options
12911 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12915 @item -mcpu=@var{cpu_type}
12917 Set the instruction set, register set, and instruction scheduling
12918 parameters for machine type @var{cpu_type}. Supported values for
12919 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12920 @samp{c44}. The default is @samp{c40} to generate code for the
12925 @itemx -msmall-memory
12927 @opindex mbig-memory
12929 @opindex msmall-memory
12931 Generates code for the big or small memory model. The small memory
12932 model assumed that all data fits into one 64K word page. At run-time
12933 the data page (DP) register must be set to point to the 64K page
12934 containing the .bss and .data program sections. The big memory model is
12935 the default and requires reloading of the DP register for every direct
12942 Allow (disallow) allocation of general integer operands into the block
12943 count register BK@.
12949 Enable (disable) generation of code using decrement and branch,
12950 DBcond(D), instructions. This is enabled by default for the C4x. To be
12951 on the safe side, this is disabled for the C3x, since the maximum
12952 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12953 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12954 that it can utilize the decrement and branch instruction, but will give
12955 up if there is more than one memory reference in the loop. Thus a loop
12956 where the loop counter is decremented can generate slightly more
12957 efficient code, in cases where the RPTB instruction cannot be utilized.
12959 @item -mdp-isr-reload
12961 @opindex mdp-isr-reload
12963 Force the DP register to be saved on entry to an interrupt service
12964 routine (ISR), reloaded to point to the data section, and restored on
12965 exit from the ISR@. This should not be required unless someone has
12966 violated the small memory model by modifying the DP register, say within
12973 For the C3x use the 24-bit MPYI instruction for integer multiplies
12974 instead of a library call to guarantee 32-bit results. Note that if one
12975 of the operands is a constant, then the multiplication will be performed
12976 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12977 then squaring operations are performed inline instead of a library call.
12980 @itemx -mno-fast-fix
12982 @opindex mno-fast-fix
12983 The C3x/C4x FIX instruction to convert a floating point value to an
12984 integer value chooses the nearest integer less than or equal to the
12985 floating point value rather than to the nearest integer. Thus if the
12986 floating point number is negative, the result will be incorrectly
12987 truncated an additional code is necessary to detect and correct this
12988 case. This option can be used to disable generation of the additional
12989 code required to correct the result.
12995 Enable (disable) generation of repeat block sequences using the RPTB
12996 instruction for zero overhead looping. The RPTB construct is only used
12997 for innermost loops that do not call functions or jump across the loop
12998 boundaries. There is no advantage having nested RPTB loops due to the
12999 overhead required to save and restore the RC, RS, and RE registers.
13000 This is enabled by default with @option{-O2}.
13002 @item -mrpts=@var{count}
13006 Enable (disable) the use of the single instruction repeat instruction
13007 RPTS@. If a repeat block contains a single instruction, and the loop
13008 count can be guaranteed to be less than the value @var{count}, GCC will
13009 emit a RPTS instruction instead of a RPTB@. If no value is specified,
13010 then a RPTS will be emitted even if the loop count cannot be determined
13011 at compile time. Note that the repeated instruction following RPTS does
13012 not have to be reloaded from memory each iteration, thus freeing up the
13013 CPU buses for operands. However, since interrupts are blocked by this
13014 instruction, it is disabled by default.
13016 @item -mloop-unsigned
13017 @itemx -mno-loop-unsigned
13018 @opindex mloop-unsigned
13019 @opindex mno-loop-unsigned
13020 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
13021 is @math{2^{31} + 1} since these instructions test if the iteration count is
13022 negative to terminate the loop. If the iteration count is unsigned
13023 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
13024 exceeded. This switch allows an unsigned iteration count.
13028 Try to emit an assembler syntax that the TI assembler (asm30) is happy
13029 with. This also enforces compatibility with the API employed by the TI
13030 C3x C compiler. For example, long doubles are passed as structures
13031 rather than in floating point registers.
13037 Generate code that uses registers (stack) for passing arguments to functions.
13038 By default, arguments are passed in registers where possible rather
13039 than by pushing arguments on to the stack.
13041 @item -mparallel-insns
13042 @itemx -mno-parallel-insns
13043 @opindex mparallel-insns
13044 @opindex mno-parallel-insns
13045 Allow the generation of parallel instructions. This is enabled by
13046 default with @option{-O2}.
13048 @item -mparallel-mpy
13049 @itemx -mno-parallel-mpy
13050 @opindex mparallel-mpy
13051 @opindex mno-parallel-mpy
13052 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
13053 provided @option{-mparallel-insns} is also specified. These instructions have
13054 tight register constraints which can pessimize the code generation
13055 of large functions.
13060 @subsection V850 Options
13061 @cindex V850 Options
13063 These @samp{-m} options are defined for V850 implementations:
13067 @itemx -mno-long-calls
13068 @opindex mlong-calls
13069 @opindex mno-long-calls
13070 Treat all calls as being far away (near). If calls are assumed to be
13071 far away, the compiler will always load the functions address up into a
13072 register, and call indirect through the pointer.
13078 Do not optimize (do optimize) basic blocks that use the same index
13079 pointer 4 or more times to copy pointer into the @code{ep} register, and
13080 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
13081 option is on by default if you optimize.
13083 @item -mno-prolog-function
13084 @itemx -mprolog-function
13085 @opindex mno-prolog-function
13086 @opindex mprolog-function
13087 Do not use (do use) external functions to save and restore registers
13088 at the prologue and epilogue of a function. The external functions
13089 are slower, but use less code space if more than one function saves
13090 the same number of registers. The @option{-mprolog-function} option
13091 is on by default if you optimize.
13095 Try to make the code as small as possible. At present, this just turns
13096 on the @option{-mep} and @option{-mprolog-function} options.
13098 @item -mtda=@var{n}
13100 Put static or global variables whose size is @var{n} bytes or less into
13101 the tiny data area that register @code{ep} points to. The tiny data
13102 area can hold up to 256 bytes in total (128 bytes for byte references).
13104 @item -msda=@var{n}
13106 Put static or global variables whose size is @var{n} bytes or less into
13107 the small data area that register @code{gp} points to. The small data
13108 area can hold up to 64 kilobytes.
13110 @item -mzda=@var{n}
13112 Put static or global variables whose size is @var{n} bytes or less into
13113 the first 32 kilobytes of memory.
13117 Specify that the target processor is the V850.
13120 @opindex mbig-switch
13121 Generate code suitable for big switch tables. Use this option only if
13122 the assembler/linker complain about out of range branches within a switch
13127 This option will cause r2 and r5 to be used in the code generated by
13128 the compiler. This setting is the default.
13130 @item -mno-app-regs
13131 @opindex mno-app-regs
13132 This option will cause r2 and r5 to be treated as fixed registers.
13136 Specify that the target processor is the V850E1. The preprocessor
13137 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
13138 this option is used.
13142 Specify that the target processor is the V850E@. The preprocessor
13143 constant @samp{__v850e__} will be defined if this option is used.
13145 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
13146 are defined then a default target processor will be chosen and the
13147 relevant @samp{__v850*__} preprocessor constant will be defined.
13149 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
13150 defined, regardless of which processor variant is the target.
13152 @item -mdisable-callt
13153 @opindex mdisable-callt
13154 This option will suppress generation of the CALLT instruction for the
13155 v850e and v850e1 flavors of the v850 architecture. The default is
13156 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
13161 @subsection VAX Options
13162 @cindex VAX options
13164 These @samp{-m} options are defined for the VAX:
13169 Do not output certain jump instructions (@code{aobleq} and so on)
13170 that the Unix assembler for the VAX cannot handle across long
13175 Do output those jump instructions, on the assumption that you
13176 will assemble with the GNU assembler.
13180 Output code for g-format floating point numbers instead of d-format.
13183 @node x86-64 Options
13184 @subsection x86-64 Options
13185 @cindex x86-64 options
13187 These are listed under @xref{i386 and x86-64 Options}.
13189 @node Xstormy16 Options
13190 @subsection Xstormy16 Options
13191 @cindex Xstormy16 Options
13193 These options are defined for Xstormy16:
13198 Choose startup files and linker script suitable for the simulator.
13201 @node Xtensa Options
13202 @subsection Xtensa Options
13203 @cindex Xtensa Options
13205 These options are supported for Xtensa targets:
13209 @itemx -mno-const16
13211 @opindex mno-const16
13212 Enable or disable use of @code{CONST16} instructions for loading
13213 constant values. The @code{CONST16} instruction is currently not a
13214 standard option from Tensilica. When enabled, @code{CONST16}
13215 instructions are always used in place of the standard @code{L32R}
13216 instructions. The use of @code{CONST16} is enabled by default only if
13217 the @code{L32R} instruction is not available.
13220 @itemx -mno-fused-madd
13221 @opindex mfused-madd
13222 @opindex mno-fused-madd
13223 Enable or disable use of fused multiply/add and multiply/subtract
13224 instructions in the floating-point option. This has no effect if the
13225 floating-point option is not also enabled. Disabling fused multiply/add
13226 and multiply/subtract instructions forces the compiler to use separate
13227 instructions for the multiply and add/subtract operations. This may be
13228 desirable in some cases where strict IEEE 754-compliant results are
13229 required: the fused multiply add/subtract instructions do not round the
13230 intermediate result, thereby producing results with @emph{more} bits of
13231 precision than specified by the IEEE standard. Disabling fused multiply
13232 add/subtract instructions also ensures that the program output is not
13233 sensitive to the compiler's ability to combine multiply and add/subtract
13236 @item -mtext-section-literals
13237 @itemx -mno-text-section-literals
13238 @opindex mtext-section-literals
13239 @opindex mno-text-section-literals
13240 Control the treatment of literal pools. The default is
13241 @option{-mno-text-section-literals}, which places literals in a separate
13242 section in the output file. This allows the literal pool to be placed
13243 in a data RAM/ROM, and it also allows the linker to combine literal
13244 pools from separate object files to remove redundant literals and
13245 improve code size. With @option{-mtext-section-literals}, the literals
13246 are interspersed in the text section in order to keep them as close as
13247 possible to their references. This may be necessary for large assembly
13250 @item -mtarget-align
13251 @itemx -mno-target-align
13252 @opindex mtarget-align
13253 @opindex mno-target-align
13254 When this option is enabled, GCC instructs the assembler to
13255 automatically align instructions to reduce branch penalties at the
13256 expense of some code density. The assembler attempts to widen density
13257 instructions to align branch targets and the instructions following call
13258 instructions. If there are not enough preceding safe density
13259 instructions to align a target, no widening will be performed. The
13260 default is @option{-mtarget-align}. These options do not affect the
13261 treatment of auto-aligned instructions like @code{LOOP}, which the
13262 assembler will always align, either by widening density instructions or
13263 by inserting no-op instructions.
13266 @itemx -mno-longcalls
13267 @opindex mlongcalls
13268 @opindex mno-longcalls
13269 When this option is enabled, GCC instructs the assembler to translate
13270 direct calls to indirect calls unless it can determine that the target
13271 of a direct call is in the range allowed by the call instruction. This
13272 translation typically occurs for calls to functions in other source
13273 files. Specifically, the assembler translates a direct @code{CALL}
13274 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
13275 The default is @option{-mno-longcalls}. This option should be used in
13276 programs where the call target can potentially be out of range. This
13277 option is implemented in the assembler, not the compiler, so the
13278 assembly code generated by GCC will still show direct call
13279 instructions---look at the disassembled object code to see the actual
13280 instructions. Note that the assembler will use an indirect call for
13281 every cross-file call, not just those that really will be out of range.
13284 @node zSeries Options
13285 @subsection zSeries Options
13286 @cindex zSeries options
13288 These are listed under @xref{S/390 and zSeries Options}.
13290 @node Code Gen Options
13291 @section Options for Code Generation Conventions
13292 @cindex code generation conventions
13293 @cindex options, code generation
13294 @cindex run-time options
13296 These machine-independent options control the interface conventions
13297 used in code generation.
13299 Most of them have both positive and negative forms; the negative form
13300 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
13301 one of the forms is listed---the one which is not the default. You
13302 can figure out the other form by either removing @samp{no-} or adding
13306 @item -fbounds-check
13307 @opindex fbounds-check
13308 For front-ends that support it, generate additional code to check that
13309 indices used to access arrays are within the declared range. This is
13310 currently only supported by the Java and Fortran front-ends, where
13311 this option defaults to true and false respectively.
13315 This option generates traps for signed overflow on addition, subtraction,
13316 multiplication operations.
13320 This option instructs the compiler to assume that signed arithmetic
13321 overflow of addition, subtraction and multiplication wraps around
13322 using twos-complement representation. This flag enables some optimizations
13323 and disables others. This option is enabled by default for the Java
13324 front-end, as required by the Java language specification.
13327 @opindex fexceptions
13328 Enable exception handling. Generates extra code needed to propagate
13329 exceptions. For some targets, this implies GCC will generate frame
13330 unwind information for all functions, which can produce significant data
13331 size overhead, although it does not affect execution. If you do not
13332 specify this option, GCC will enable it by default for languages like
13333 C++ which normally require exception handling, and disable it for
13334 languages like C that do not normally require it. However, you may need
13335 to enable this option when compiling C code that needs to interoperate
13336 properly with exception handlers written in C++. You may also wish to
13337 disable this option if you are compiling older C++ programs that don't
13338 use exception handling.
13340 @item -fnon-call-exceptions
13341 @opindex fnon-call-exceptions
13342 Generate code that allows trapping instructions to throw exceptions.
13343 Note that this requires platform-specific runtime support that does
13344 not exist everywhere. Moreover, it only allows @emph{trapping}
13345 instructions to throw exceptions, i.e.@: memory references or floating
13346 point instructions. It does not allow exceptions to be thrown from
13347 arbitrary signal handlers such as @code{SIGALRM}.
13349 @item -funwind-tables
13350 @opindex funwind-tables
13351 Similar to @option{-fexceptions}, except that it will just generate any needed
13352 static data, but will not affect the generated code in any other way.
13353 You will normally not enable this option; instead, a language processor
13354 that needs this handling would enable it on your behalf.
13356 @item -fasynchronous-unwind-tables
13357 @opindex fasynchronous-unwind-tables
13358 Generate unwind table in dwarf2 format, if supported by target machine. The
13359 table is exact at each instruction boundary, so it can be used for stack
13360 unwinding from asynchronous events (such as debugger or garbage collector).
13362 @item -fpcc-struct-return
13363 @opindex fpcc-struct-return
13364 Return ``short'' @code{struct} and @code{union} values in memory like
13365 longer ones, rather than in registers. This convention is less
13366 efficient, but it has the advantage of allowing intercallability between
13367 GCC-compiled files and files compiled with other compilers, particularly
13368 the Portable C Compiler (pcc).
13370 The precise convention for returning structures in memory depends
13371 on the target configuration macros.
13373 Short structures and unions are those whose size and alignment match
13374 that of some integer type.
13376 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
13377 switch is not binary compatible with code compiled with the
13378 @option{-freg-struct-return} switch.
13379 Use it to conform to a non-default application binary interface.
13381 @item -freg-struct-return
13382 @opindex freg-struct-return
13383 Return @code{struct} and @code{union} values in registers when possible.
13384 This is more efficient for small structures than
13385 @option{-fpcc-struct-return}.
13387 If you specify neither @option{-fpcc-struct-return} nor
13388 @option{-freg-struct-return}, GCC defaults to whichever convention is
13389 standard for the target. If there is no standard convention, GCC
13390 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
13391 the principal compiler. In those cases, we can choose the standard, and
13392 we chose the more efficient register return alternative.
13394 @strong{Warning:} code compiled with the @option{-freg-struct-return}
13395 switch is not binary compatible with code compiled with the
13396 @option{-fpcc-struct-return} switch.
13397 Use it to conform to a non-default application binary interface.
13399 @item -fshort-enums
13400 @opindex fshort-enums
13401 Allocate to an @code{enum} type only as many bytes as it needs for the
13402 declared range of possible values. Specifically, the @code{enum} type
13403 will be equivalent to the smallest integer type which has enough room.
13405 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
13406 code that is not binary compatible with code generated without that switch.
13407 Use it to conform to a non-default application binary interface.
13409 @item -fshort-double
13410 @opindex fshort-double
13411 Use the same size for @code{double} as for @code{float}.
13413 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
13414 code that is not binary compatible with code generated without that switch.
13415 Use it to conform to a non-default application binary interface.
13417 @item -fshort-wchar
13418 @opindex fshort-wchar
13419 Override the underlying type for @samp{wchar_t} to be @samp{short
13420 unsigned int} instead of the default for the target. This option is
13421 useful for building programs to run under WINE@.
13423 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
13424 code that is not binary compatible with code generated without that switch.
13425 Use it to conform to a non-default application binary interface.
13428 @opindex fno-common
13429 In C, allocate even uninitialized global variables in the data section of the
13430 object file, rather than generating them as common blocks. This has the
13431 effect that if the same variable is declared (without @code{extern}) in
13432 two different compilations, you will get an error when you link them.
13433 The only reason this might be useful is if you wish to verify that the
13434 program will work on other systems which always work this way.
13438 Ignore the @samp{#ident} directive.
13440 @item -finhibit-size-directive
13441 @opindex finhibit-size-directive
13442 Don't output a @code{.size} assembler directive, or anything else that
13443 would cause trouble if the function is split in the middle, and the
13444 two halves are placed at locations far apart in memory. This option is
13445 used when compiling @file{crtstuff.c}; you should not need to use it
13448 @item -fverbose-asm
13449 @opindex fverbose-asm
13450 Put extra commentary information in the generated assembly code to
13451 make it more readable. This option is generally only of use to those
13452 who actually need to read the generated assembly code (perhaps while
13453 debugging the compiler itself).
13455 @option{-fno-verbose-asm}, the default, causes the
13456 extra information to be omitted and is useful when comparing two assembler
13461 @cindex global offset table
13463 Generate position-independent code (PIC) suitable for use in a shared
13464 library, if supported for the target machine. Such code accesses all
13465 constant addresses through a global offset table (GOT)@. The dynamic
13466 loader resolves the GOT entries when the program starts (the dynamic
13467 loader is not part of GCC; it is part of the operating system). If
13468 the GOT size for the linked executable exceeds a machine-specific
13469 maximum size, you get an error message from the linker indicating that
13470 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
13471 instead. (These maximums are 8k on the SPARC and 32k
13472 on the m68k and RS/6000. The 386 has no such limit.)
13474 Position-independent code requires special support, and therefore works
13475 only on certain machines. For the 386, GCC supports PIC for System V
13476 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
13477 position-independent.
13479 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13484 If supported for the target machine, emit position-independent code,
13485 suitable for dynamic linking and avoiding any limit on the size of the
13486 global offset table. This option makes a difference on the m68k,
13487 PowerPC and SPARC@.
13489 Position-independent code requires special support, and therefore works
13490 only on certain machines.
13492 When this flag is set, the macros @code{__pic__} and @code{__PIC__}
13499 These options are similar to @option{-fpic} and @option{-fPIC}, but
13500 generated position independent code can be only linked into executables.
13501 Usually these options are used when @option{-pie} GCC option will be
13502 used during linking.
13504 @item -fno-jump-tables
13505 @opindex fno-jump-tables
13506 Do not use jump tables for switch statements even where it would be
13507 more efficient than other code generation strategies. This option is
13508 of use in conjunction with @option{-fpic} or @option{-fPIC} for
13509 building code which forms part of a dynamic linker and cannot
13510 reference the address of a jump table. On some targets, jump tables
13511 do not require a GOT and this option is not needed.
13513 @item -ffixed-@var{reg}
13515 Treat the register named @var{reg} as a fixed register; generated code
13516 should never refer to it (except perhaps as a stack pointer, frame
13517 pointer or in some other fixed role).
13519 @var{reg} must be the name of a register. The register names accepted
13520 are machine-specific and are defined in the @code{REGISTER_NAMES}
13521 macro in the machine description macro file.
13523 This flag does not have a negative form, because it specifies a
13526 @item -fcall-used-@var{reg}
13527 @opindex fcall-used
13528 Treat the register named @var{reg} as an allocable register that is
13529 clobbered by function calls. It may be allocated for temporaries or
13530 variables that do not live across a call. Functions compiled this way
13531 will not save and restore the register @var{reg}.
13533 It is an error to used this flag with the frame pointer or stack pointer.
13534 Use of this flag for other registers that have fixed pervasive roles in
13535 the machine's execution model will produce disastrous results.
13537 This flag does not have a negative form, because it specifies a
13540 @item -fcall-saved-@var{reg}
13541 @opindex fcall-saved
13542 Treat the register named @var{reg} as an allocable register saved by
13543 functions. It may be allocated even for temporaries or variables that
13544 live across a call. Functions compiled this way will save and restore
13545 the register @var{reg} if they use it.
13547 It is an error to used this flag with the frame pointer or stack pointer.
13548 Use of this flag for other registers that have fixed pervasive roles in
13549 the machine's execution model will produce disastrous results.
13551 A different sort of disaster will result from the use of this flag for
13552 a register in which function values may be returned.
13554 This flag does not have a negative form, because it specifies a
13557 @item -fpack-struct[=@var{n}]
13558 @opindex fpack-struct
13559 Without a value specified, pack all structure members together without
13560 holes. When a value is specified (which must be a small power of two), pack
13561 structure members according to this value, representing the maximum
13562 alignment (that is, objects with default alignment requirements larger than
13563 this will be output potentially unaligned at the next fitting location.
13565 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
13566 code that is not binary compatible with code generated without that switch.
13567 Additionally, it makes the code suboptimal.
13568 Use it to conform to a non-default application binary interface.
13570 @item -finstrument-functions
13571 @opindex finstrument-functions
13572 Generate instrumentation calls for entry and exit to functions. Just
13573 after function entry and just before function exit, the following
13574 profiling functions will be called with the address of the current
13575 function and its call site. (On some platforms,
13576 @code{__builtin_return_address} does not work beyond the current
13577 function, so the call site information may not be available to the
13578 profiling functions otherwise.)
13581 void __cyg_profile_func_enter (void *this_fn,
13583 void __cyg_profile_func_exit (void *this_fn,
13587 The first argument is the address of the start of the current function,
13588 which may be looked up exactly in the symbol table.
13590 This instrumentation is also done for functions expanded inline in other
13591 functions. The profiling calls will indicate where, conceptually, the
13592 inline function is entered and exited. This means that addressable
13593 versions of such functions must be available. If all your uses of a
13594 function are expanded inline, this may mean an additional expansion of
13595 code size. If you use @samp{extern inline} in your C code, an
13596 addressable version of such functions must be provided. (This is
13597 normally the case anyways, but if you get lucky and the optimizer always
13598 expands the functions inline, you might have gotten away without
13599 providing static copies.)
13601 A function may be given the attribute @code{no_instrument_function}, in
13602 which case this instrumentation will not be done. This can be used, for
13603 example, for the profiling functions listed above, high-priority
13604 interrupt routines, and any functions from which the profiling functions
13605 cannot safely be called (perhaps signal handlers, if the profiling
13606 routines generate output or allocate memory).
13608 @item -fstack-check
13609 @opindex fstack-check
13610 Generate code to verify that you do not go beyond the boundary of the
13611 stack. You should specify this flag if you are running in an
13612 environment with multiple threads, but only rarely need to specify it in
13613 a single-threaded environment since stack overflow is automatically
13614 detected on nearly all systems if there is only one stack.
13616 Note that this switch does not actually cause checking to be done; the
13617 operating system must do that. The switch causes generation of code
13618 to ensure that the operating system sees the stack being extended.
13620 @item -fstack-limit-register=@var{reg}
13621 @itemx -fstack-limit-symbol=@var{sym}
13622 @itemx -fno-stack-limit
13623 @opindex fstack-limit-register
13624 @opindex fstack-limit-symbol
13625 @opindex fno-stack-limit
13626 Generate code to ensure that the stack does not grow beyond a certain value,
13627 either the value of a register or the address of a symbol. If the stack
13628 would grow beyond the value, a signal is raised. For most targets,
13629 the signal is raised before the stack overruns the boundary, so
13630 it is possible to catch the signal without taking special precautions.
13632 For instance, if the stack starts at absolute address @samp{0x80000000}
13633 and grows downwards, you can use the flags
13634 @option{-fstack-limit-symbol=__stack_limit} and
13635 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13636 of 128KB@. Note that this may only work with the GNU linker.
13638 @cindex aliasing of parameters
13639 @cindex parameters, aliased
13640 @item -fargument-alias
13641 @itemx -fargument-noalias
13642 @itemx -fargument-noalias-global
13643 @itemx -fargument-noalias-anything
13644 @opindex fargument-alias
13645 @opindex fargument-noalias
13646 @opindex fargument-noalias-global
13647 @opindex fargument-noalias-anything
13648 Specify the possible relationships among parameters and between
13649 parameters and global data.
13651 @option{-fargument-alias} specifies that arguments (parameters) may
13652 alias each other and may alias global storage.@*
13653 @option{-fargument-noalias} specifies that arguments do not alias
13654 each other, but may alias global storage.@*
13655 @option{-fargument-noalias-global} specifies that arguments do not
13656 alias each other and do not alias global storage.
13657 @option{-fargument-noalias-anything} specifies that arguments do not
13658 alias any other storage.
13660 Each language will automatically use whatever option is required by
13661 the language standard. You should not need to use these options yourself.
13663 @item -fleading-underscore
13664 @opindex fleading-underscore
13665 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13666 change the way C symbols are represented in the object file. One use
13667 is to help link with legacy assembly code.
13669 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13670 generate code that is not binary compatible with code generated without that
13671 switch. Use it to conform to a non-default application binary interface.
13672 Not all targets provide complete support for this switch.
13674 @item -ftls-model=@var{model}
13675 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13676 The @var{model} argument should be one of @code{global-dynamic},
13677 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13679 The default without @option{-fpic} is @code{initial-exec}; with
13680 @option{-fpic} the default is @code{global-dynamic}.
13682 @item -fvisibility=@var{default|internal|hidden|protected}
13683 @opindex fvisibility
13684 Set the default ELF image symbol visibility to the specified option---all
13685 symbols will be marked with this unless overridden within the code.
13686 Using this feature can very substantially improve linking and
13687 load times of shared object libraries, produce more optimized
13688 code, provide near-perfect API export and prevent symbol clashes.
13689 It is @strong{strongly} recommended that you use this in any shared objects
13692 Despite the nomenclature, @code{default} always means public ie;
13693 available to be linked against from outside the shared object.
13694 @code{protected} and @code{internal} are pretty useless in real-world
13695 usage so the only other commonly used option will be @code{hidden}.
13696 The default if @option{-fvisibility} isn't specified is
13697 @code{default}, i.e., make every
13698 symbol public---this causes the same behavior as previous versions of
13701 A good explanation of the benefits offered by ensuring ELF
13702 symbols have the correct visibility is given by ``How To Write
13703 Shared Libraries'' by Ulrich Drepper (which can be found at
13704 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13705 solution made possible by this option to marking things hidden when
13706 the default is public is to make the default hidden and mark things
13707 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13708 and @code{__attribute__ ((visibility("default")))} instead of
13709 @code{__declspec(dllexport)} you get almost identical semantics with
13710 identical syntax. This is a great boon to those working with
13711 cross-platform projects.
13713 For those adding visibility support to existing code, you may find
13714 @samp{#pragma GCC visibility} of use. This works by you enclosing
13715 the declarations you wish to set visibility for with (for example)
13716 @samp{#pragma GCC visibility push(hidden)} and
13717 @samp{#pragma GCC visibility pop}.
13718 Bear in mind that symbol visibility should be viewed @strong{as
13719 part of the API interface contract} and thus all new code should
13720 always specify visibility when it is not the default ie; declarations
13721 only for use within the local DSO should @strong{always} be marked explicitly
13722 as hidden as so to avoid PLT indirection overheads---making this
13723 abundantly clear also aids readability and self-documentation of the code.
13724 Note that due to ISO C++ specification requirements, operator new and
13725 operator delete must always be of default visibility.
13727 Be aware that headers from outside your project, in particular system
13728 headers and headers from any other library you use, may not be
13729 expecting to be compiled with visibility other than the default. You
13730 may need to explicitly say @samp{#pragma GCC visibility push(default)}
13731 before including any such headers.
13733 @samp{extern} declarations are not affected by @samp{-fvisibility}, so
13734 a lot of code can be recompiled with @samp{-fvisibility=hidden} with
13735 no modifications. However, this means that calls to @samp{extern}
13736 functions with no explicit visibility will use the PLT, so it is more
13737 effective to use @samp{__attribute ((visibility))} and/or
13738 @samp{#pragma GCC visibility} to tell the compiler which @samp{extern}
13739 declarations should be treated as hidden.
13741 Note that @samp{-fvisibility} does affect C++ vague linkage
13742 entities. This means that, for instance, an exception class that will
13743 be thrown between DSOs must be explicitly marked with default
13744 visibility so that the @samp{type_info} nodes will be unified between
13747 An overview of these techniques, their benefits and how to use them
13748 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13754 @node Environment Variables
13755 @section Environment Variables Affecting GCC
13756 @cindex environment variables
13758 @c man begin ENVIRONMENT
13759 This section describes several environment variables that affect how GCC
13760 operates. Some of them work by specifying directories or prefixes to use
13761 when searching for various kinds of files. Some are used to specify other
13762 aspects of the compilation environment.
13764 Note that you can also specify places to search using options such as
13765 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13766 take precedence over places specified using environment variables, which
13767 in turn take precedence over those specified by the configuration of GCC@.
13768 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13769 GNU Compiler Collection (GCC) Internals}.
13774 @c @itemx LC_COLLATE
13776 @c @itemx LC_MONETARY
13777 @c @itemx LC_NUMERIC
13782 @c @findex LC_COLLATE
13783 @findex LC_MESSAGES
13784 @c @findex LC_MONETARY
13785 @c @findex LC_NUMERIC
13789 These environment variables control the way that GCC uses
13790 localization information that allow GCC to work with different
13791 national conventions. GCC inspects the locale categories
13792 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13793 so. These locale categories can be set to any value supported by your
13794 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13795 Kingdom encoded in UTF-8.
13797 The @env{LC_CTYPE} environment variable specifies character
13798 classification. GCC uses it to determine the character boundaries in
13799 a string; this is needed for some multibyte encodings that contain quote
13800 and escape characters that would otherwise be interpreted as a string
13803 The @env{LC_MESSAGES} environment variable specifies the language to
13804 use in diagnostic messages.
13806 If the @env{LC_ALL} environment variable is set, it overrides the value
13807 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13808 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13809 environment variable. If none of these variables are set, GCC
13810 defaults to traditional C English behavior.
13814 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13815 files. GCC uses temporary files to hold the output of one stage of
13816 compilation which is to be used as input to the next stage: for example,
13817 the output of the preprocessor, which is the input to the compiler
13820 @item GCC_EXEC_PREFIX
13821 @findex GCC_EXEC_PREFIX
13822 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13823 names of the subprograms executed by the compiler. No slash is added
13824 when this prefix is combined with the name of a subprogram, but you can
13825 specify a prefix that ends with a slash if you wish.
13827 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13828 an appropriate prefix to use based on the pathname it was invoked with.
13830 If GCC cannot find the subprogram using the specified prefix, it
13831 tries looking in the usual places for the subprogram.
13833 The default value of @env{GCC_EXEC_PREFIX} is
13834 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13835 of @code{prefix} when you ran the @file{configure} script.
13837 Other prefixes specified with @option{-B} take precedence over this prefix.
13839 This prefix is also used for finding files such as @file{crt0.o} that are
13842 In addition, the prefix is used in an unusual way in finding the
13843 directories to search for header files. For each of the standard
13844 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13845 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13846 replacing that beginning with the specified prefix to produce an
13847 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13848 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13849 These alternate directories are searched first; the standard directories
13852 @item COMPILER_PATH
13853 @findex COMPILER_PATH
13854 The value of @env{COMPILER_PATH} is a colon-separated list of
13855 directories, much like @env{PATH}. GCC tries the directories thus
13856 specified when searching for subprograms, if it can't find the
13857 subprograms using @env{GCC_EXEC_PREFIX}.
13860 @findex LIBRARY_PATH
13861 The value of @env{LIBRARY_PATH} is a colon-separated list of
13862 directories, much like @env{PATH}. When configured as a native compiler,
13863 GCC tries the directories thus specified when searching for special
13864 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13865 using GCC also uses these directories when searching for ordinary
13866 libraries for the @option{-l} option (but directories specified with
13867 @option{-L} come first).
13871 @cindex locale definition
13872 This variable is used to pass locale information to the compiler. One way in
13873 which this information is used is to determine the character set to be used
13874 when character literals, string literals and comments are parsed in C and C++.
13875 When the compiler is configured to allow multibyte characters,
13876 the following values for @env{LANG} are recognized:
13880 Recognize JIS characters.
13882 Recognize SJIS characters.
13884 Recognize EUCJP characters.
13887 If @env{LANG} is not defined, or if it has some other value, then the
13888 compiler will use mblen and mbtowc as defined by the default locale to
13889 recognize and translate multibyte characters.
13893 Some additional environments variables affect the behavior of the
13896 @include cppenv.texi
13900 @node Precompiled Headers
13901 @section Using Precompiled Headers
13902 @cindex precompiled headers
13903 @cindex speed of compilation
13905 Often large projects have many header files that are included in every
13906 source file. The time the compiler takes to process these header files
13907 over and over again can account for nearly all of the time required to
13908 build the project. To make builds faster, GCC allows users to
13909 `precompile' a header file; then, if builds can use the precompiled
13910 header file they will be much faster.
13912 To create a precompiled header file, simply compile it as you would any
13913 other file, if necessary using the @option{-x} option to make the driver
13914 treat it as a C or C++ header file. You will probably want to use a
13915 tool like @command{make} to keep the precompiled header up-to-date when
13916 the headers it contains change.
13918 A precompiled header file will be searched for when @code{#include} is
13919 seen in the compilation. As it searches for the included file
13920 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13921 compiler looks for a precompiled header in each directory just before it
13922 looks for the include file in that directory. The name searched for is
13923 the name specified in the @code{#include} with @samp{.gch} appended. If
13924 the precompiled header file can't be used, it is ignored.
13926 For instance, if you have @code{#include "all.h"}, and you have
13927 @file{all.h.gch} in the same directory as @file{all.h}, then the
13928 precompiled header file will be used if possible, and the original
13929 header will be used otherwise.
13931 Alternatively, you might decide to put the precompiled header file in a
13932 directory and use @option{-I} to ensure that directory is searched
13933 before (or instead of) the directory containing the original header.
13934 Then, if you want to check that the precompiled header file is always
13935 used, you can put a file of the same name as the original header in this
13936 directory containing an @code{#error} command.
13938 This also works with @option{-include}. So yet another way to use
13939 precompiled headers, good for projects not designed with precompiled
13940 header files in mind, is to simply take most of the header files used by
13941 a project, include them from another header file, precompile that header
13942 file, and @option{-include} the precompiled header. If the header files
13943 have guards against multiple inclusion, they will be skipped because
13944 they've already been included (in the precompiled header).
13946 If you need to precompile the same header file for different
13947 languages, targets, or compiler options, you can instead make a
13948 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13949 header in the directory, perhaps using @option{-o}. It doesn't matter
13950 what you call the files in the directory, every precompiled header in
13951 the directory will be considered. The first precompiled header
13952 encountered in the directory that is valid for this compilation will
13953 be used; they're searched in no particular order.
13955 There are many other possibilities, limited only by your imagination,
13956 good sense, and the constraints of your build system.
13958 A precompiled header file can be used only when these conditions apply:
13962 Only one precompiled header can be used in a particular compilation.
13965 A precompiled header can't be used once the first C token is seen. You
13966 can have preprocessor directives before a precompiled header; you can
13967 even include a precompiled header from inside another header, so long as
13968 there are no C tokens before the @code{#include}.
13971 The precompiled header file must be produced for the same language as
13972 the current compilation. You can't use a C precompiled header for a C++
13976 The precompiled header file must have been produced by the same compiler
13977 binary as the current compilation is using.
13980 Any macros defined before the precompiled header is included must
13981 either be defined in the same way as when the precompiled header was
13982 generated, or must not affect the precompiled header, which usually
13983 means that they don't appear in the precompiled header at all.
13985 The @option{-D} option is one way to define a macro before a
13986 precompiled header is included; using a @code{#define} can also do it.
13987 There are also some options that define macros implicitly, like
13988 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13991 @item If debugging information is output when using the precompiled
13992 header, using @option{-g} or similar, the same kind of debugging information
13993 must have been output when building the precompiled header. However,
13994 a precompiled header built using @option{-g} can be used in a compilation
13995 when no debugging information is being output.
13997 @item The same @option{-m} options must generally be used when building
13998 and using the precompiled header. @xref{Submodel Options},
13999 for any cases where this rule is relaxed.
14001 @item Each of the following options must be the same when building and using
14002 the precompiled header:
14004 @gccoptlist{-fexceptions -funit-at-a-time}
14007 Some other command-line options starting with @option{-f},
14008 @option{-p}, or @option{-O} must be defined in the same way as when
14009 the precompiled header was generated. At present, it's not clear
14010 which options are safe to change and which are not; the safest choice
14011 is to use exactly the same options when generating and using the
14012 precompiled header. The following are known to be safe:
14014 @gccoptlist{-fmessage-length= -fpreprocessed
14015 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
14016 -fsched-verbose=<number> -fschedule-insns -fvisibility=
14021 For all of these except the last, the compiler will automatically
14022 ignore the precompiled header if the conditions aren't met. If you
14023 find an option combination that doesn't work and doesn't cause the
14024 precompiled header to be ignored, please consider filing a bug report,
14027 If you do use differing options when generating and using the
14028 precompiled header, the actual behavior will be a mixture of the
14029 behavior for the options. For instance, if you use @option{-g} to
14030 generate the precompiled header but not when using it, you may or may
14031 not get debugging information for routines in the precompiled header.
14033 @node Running Protoize
14034 @section Running Protoize
14036 The program @code{protoize} is an optional part of GCC@. You can use
14037 it to add prototypes to a program, thus converting the program to ISO
14038 C in one respect. The companion program @code{unprotoize} does the
14039 reverse: it removes argument types from any prototypes that are found.
14041 When you run these programs, you must specify a set of source files as
14042 command line arguments. The conversion programs start out by compiling
14043 these files to see what functions they define. The information gathered
14044 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
14046 After scanning comes actual conversion. The specified files are all
14047 eligible to be converted; any files they include (whether sources or
14048 just headers) are eligible as well.
14050 But not all the eligible files are converted. By default,
14051 @code{protoize} and @code{unprotoize} convert only source and header
14052 files in the current directory. You can specify additional directories
14053 whose files should be converted with the @option{-d @var{directory}}
14054 option. You can also specify particular files to exclude with the
14055 @option{-x @var{file}} option. A file is converted if it is eligible, its
14056 directory name matches one of the specified directory names, and its
14057 name within the directory has not been excluded.
14059 Basic conversion with @code{protoize} consists of rewriting most
14060 function definitions and function declarations to specify the types of
14061 the arguments. The only ones not rewritten are those for varargs
14064 @code{protoize} optionally inserts prototype declarations at the
14065 beginning of the source file, to make them available for any calls that
14066 precede the function's definition. Or it can insert prototype
14067 declarations with block scope in the blocks where undeclared functions
14070 Basic conversion with @code{unprotoize} consists of rewriting most
14071 function declarations to remove any argument types, and rewriting
14072 function definitions to the old-style pre-ISO form.
14074 Both conversion programs print a warning for any function declaration or
14075 definition that they can't convert. You can suppress these warnings
14078 The output from @code{protoize} or @code{unprotoize} replaces the
14079 original source file. The original file is renamed to a name ending
14080 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
14081 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
14082 for DOS) file already exists, then the source file is simply discarded.
14084 @code{protoize} and @code{unprotoize} both depend on GCC itself to
14085 scan the program and collect information about the functions it uses.
14086 So neither of these programs will work until GCC is installed.
14088 Here is a table of the options you can use with @code{protoize} and
14089 @code{unprotoize}. Each option works with both programs unless
14093 @item -B @var{directory}
14094 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
14095 usual directory (normally @file{/usr/local/lib}). This file contains
14096 prototype information about standard system functions. This option
14097 applies only to @code{protoize}.
14099 @item -c @var{compilation-options}
14100 Use @var{compilation-options} as the options when running @command{gcc} to
14101 produce the @samp{.X} files. The special option @option{-aux-info} is
14102 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
14104 Note that the compilation options must be given as a single argument to
14105 @code{protoize} or @code{unprotoize}. If you want to specify several
14106 @command{gcc} options, you must quote the entire set of compilation options
14107 to make them a single word in the shell.
14109 There are certain @command{gcc} arguments that you cannot use, because they
14110 would produce the wrong kind of output. These include @option{-g},
14111 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
14112 the @var{compilation-options}, they are ignored.
14115 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
14116 systems) instead of @samp{.c}. This is convenient if you are converting
14117 a C program to C++. This option applies only to @code{protoize}.
14120 Add explicit global declarations. This means inserting explicit
14121 declarations at the beginning of each source file for each function
14122 that is called in the file and was not declared. These declarations
14123 precede the first function definition that contains a call to an
14124 undeclared function. This option applies only to @code{protoize}.
14126 @item -i @var{string}
14127 Indent old-style parameter declarations with the string @var{string}.
14128 This option applies only to @code{protoize}.
14130 @code{unprotoize} converts prototyped function definitions to old-style
14131 function definitions, where the arguments are declared between the
14132 argument list and the initial @samp{@{}. By default, @code{unprotoize}
14133 uses five spaces as the indentation. If you want to indent with just
14134 one space instead, use @option{-i " "}.
14137 Keep the @samp{.X} files. Normally, they are deleted after conversion
14141 Add explicit local declarations. @code{protoize} with @option{-l} inserts
14142 a prototype declaration for each function in each block which calls the
14143 function without any declaration. This option applies only to
14147 Make no real changes. This mode just prints information about the conversions
14148 that would have been done without @option{-n}.
14151 Make no @samp{.save} files. The original files are simply deleted.
14152 Use this option with caution.
14154 @item -p @var{program}
14155 Use the program @var{program} as the compiler. Normally, the name
14156 @file{gcc} is used.
14159 Work quietly. Most warnings are suppressed.
14162 Print the version number, just like @option{-v} for @command{gcc}.
14165 If you need special compiler options to compile one of your program's
14166 source files, then you should generate that file's @samp{.X} file
14167 specially, by running @command{gcc} on that source file with the
14168 appropriate options and the option @option{-aux-info}. Then run
14169 @code{protoize} on the entire set of files. @code{protoize} will use
14170 the existing @samp{.X} file because it is newer than the source file.
14174 gcc -Dfoo=bar file1.c -aux-info file1.X
14179 You need to include the special files along with the rest in the
14180 @code{protoize} command, even though their @samp{.X} files already
14181 exist, because otherwise they won't get converted.
14183 @xref{Protoize Caveats}, for more information on how to use
14184 @code{protoize} successfully.