1 @c Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
2 @c 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
3 @c This is part of the GCC manual.
4 @c For copying conditions, see the file gcc.texi.
8 Copyright @copyright{} 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997,
9 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
11 Permission is granted to copy, distribute and/or modify this document
12 under the terms of the GNU Free Documentation License, Version 1.2 or
13 any later version published by the Free Software Foundation; with the
14 Invariant Sections being ``GNU General Public License'' and ``Funding
15 Free Software'', the Front-Cover texts being (a) (see below), and with
16 the Back-Cover Texts being (b) (see below). A copy of the license is
17 included in the gfdl(7) man page.
19 (a) The FSF's Front-Cover Text is:
23 (b) The FSF's Back-Cover Text is:
25 You have freedom to copy and modify this GNU Manual, like GNU
26 software. Copies published by the Free Software Foundation raise
27 funds for GNU development.
29 @c Set file name and title for the man page.
31 @settitle GNU project C and C++ compiler
33 gcc [@option{-c}|@option{-S}|@option{-E}] [@option{-std=}@var{standard}]
34 [@option{-g}] [@option{-pg}] [@option{-O}@var{level}]
35 [@option{-W}@var{warn}@dots{}] [@option{-pedantic}]
36 [@option{-I}@var{dir}@dots{}] [@option{-L}@var{dir}@dots{}]
37 [@option{-D}@var{macro}[=@var{defn}]@dots{}] [@option{-U}@var{macro}]
38 [@option{-f}@var{option}@dots{}] [@option{-m}@var{machine-option}@dots{}]
39 [@option{-o} @var{outfile}] @var{infile}@dots{}
41 Only the most useful options are listed here; see below for the
42 remainder. @samp{g++} accepts mostly the same options as @samp{gcc}.
45 gpl(7), gfdl(7), fsf-funding(7),
46 cpp(1), gcov(1), as(1), ld(1), gdb(1), adb(1), dbx(1), sdb(1)
47 and the Info entries for @file{gcc}, @file{cpp}, @file{as},
48 @file{ld}, @file{binutils} and @file{gdb}.
51 For instructions on reporting bugs, see
52 @w{@uref{http://gcc.gnu.org/bugs.html}}.
55 See the Info entry for @command{gcc}, or
56 @w{@uref{http://gcc.gnu.org/onlinedocs/gcc/Contributors.html}},
57 for contributors to GCC@.
62 @chapter GCC Command Options
63 @cindex GCC command options
64 @cindex command options
65 @cindex options, GCC command
67 @c man begin DESCRIPTION
68 When you invoke GCC, it normally does preprocessing, compilation,
69 assembly and linking. The ``overall options'' allow you to stop this
70 process at an intermediate stage. For example, the @option{-c} option
71 says not to run the linker. Then the output consists of object files
72 output by the assembler.
74 Other options are passed on to one stage of processing. Some options
75 control the preprocessor and others the compiler itself. Yet other
76 options control the assembler and linker; most of these are not
77 documented here, since you rarely need to use any of them.
79 @cindex C compilation options
80 Most of the command line options that you can use with GCC are useful
81 for C programs; when an option is only useful with another language
82 (usually C++), the explanation says so explicitly. If the description
83 for a particular option does not mention a source language, you can use
84 that option with all supported languages.
86 @cindex C++ compilation options
87 @xref{Invoking G++,,Compiling C++ Programs}, for a summary of special
88 options for compiling C++ programs.
90 @cindex grouping options
91 @cindex options, grouping
92 The @command{gcc} program accepts options and file names as operands. Many
93 options have multi-letter names; therefore multiple single-letter options
94 may @emph{not} be grouped: @option{-dr} is very different from @w{@samp{-d
97 @cindex order of options
98 @cindex options, order
99 You can mix options and other arguments. For the most part, the order
100 you use doesn't matter. Order does matter when you use several options
101 of the same kind; for example, if you specify @option{-L} more than once,
102 the directories are searched in the order specified.
104 Many options have long names starting with @samp{-f} or with
105 @samp{-W}---for example,
106 @option{-fstrength-reduce}, @option{-Wformat} and so on. Most of
107 these have both positive and negative forms; the negative form of
108 @option{-ffoo} would be @option{-fno-foo}. This manual documents
109 only one of these two forms, whichever one is not the default.
113 @xref{Option Index}, for an index to GCC's options.
116 * Option Summary:: Brief list of all options, without explanations.
117 * Overall Options:: Controlling the kind of output:
118 an executable, object files, assembler files,
119 or preprocessed source.
120 * Invoking G++:: Compiling C++ programs.
121 * C Dialect Options:: Controlling the variant of C language compiled.
122 * C++ Dialect Options:: Variations on C++.
123 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
125 * Language Independent Options:: Controlling how diagnostics should be
127 * Warning Options:: How picky should the compiler be?
128 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
129 * Optimize Options:: How much optimization?
130 * Preprocessor Options:: Controlling header files and macro definitions.
131 Also, getting dependency information for Make.
132 * Assembler Options:: Passing options to the assembler.
133 * Link Options:: Specifying libraries and so on.
134 * Directory Options:: Where to find header files and libraries.
135 Where to find the compiler executable files.
136 * Spec Files:: How to pass switches to sub-processes.
137 * Target Options:: Running a cross-compiler, or an old version of GCC.
138 * Submodel Options:: Specifying minor hardware or convention variations,
139 such as 68010 vs 68020.
140 * Code Gen Options:: Specifying conventions for function calls, data layout
142 * Environment Variables:: Env vars that affect GCC.
143 * Precompiled Headers:: Compiling a header once, and using it many times.
144 * Running Protoize:: Automatically adding or removing function prototypes.
150 @section Option Summary
152 Here is a summary of all the options, grouped by type. Explanations are
153 in the following sections.
156 @item Overall Options
157 @xref{Overall Options,,Options Controlling the Kind of Output}.
158 @gccoptlist{-c -S -E -o @var{file} -combine -pipe -pass-exit-codes @gol
159 -x @var{language} -v -### --help --target-help --version}
161 @item C Language Options
162 @xref{C Dialect Options,,Options Controlling C Dialect}.
163 @gccoptlist{-ansi -std=@var{standard} -aux-info @var{filename} @gol
164 -fno-asm -fno-builtin -fno-builtin-@var{function} @gol
165 -fhosted -ffreestanding -fms-extensions @gol
166 -trigraphs -no-integrated-cpp -traditional -traditional-cpp @gol
167 -fallow-single-precision -fcond-mismatch @gol
168 -fsigned-bitfields -fsigned-char @gol
169 -funsigned-bitfields -funsigned-char}
171 @item C++ Language Options
172 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}.
173 @gccoptlist{-fabi-version=@var{n} -fno-access-control -fcheck-new @gol
174 -fconserve-space -ffriend-injection -fno-const-strings @gol
175 -fno-elide-constructors @gol
176 -fno-enforce-eh-specs @gol
177 -ffor-scope -fno-for-scope -fno-gnu-keywords @gol
178 -fno-implicit-templates @gol
179 -fno-implicit-inline-templates @gol
180 -fno-implement-inlines -fms-extensions @gol
181 -fno-nonansi-builtins -fno-operator-names @gol
182 -fno-optional-diags -fpermissive @gol
183 -frepo -fno-rtti -fstats -ftemplate-depth-@var{n} @gol
184 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++ @gol
185 -fno-default-inline -fvisibility-inlines-hidden @gol
186 -Wabi -Wctor-dtor-privacy @gol
187 -Wnon-virtual-dtor -Wreorder @gol
188 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel @gol
189 -Wno-non-template-friend -Wold-style-cast @gol
190 -Woverloaded-virtual -Wno-pmf-conversions @gol
193 @item Objective-C and Objective-C++ Language Options
194 @xref{Objective-C and Objective-C++ Dialect Options,,Options Controlling
195 Objective-C and Objective-C++ Dialects}.
197 -fconstant-string-class=@var{class-name} @gol
198 -fgnu-runtime -fnext-runtime @gol
199 -fno-nil-receivers @gol
200 -fobjc-call-cxx-cdtors @gol
201 -fobjc-direct-dispatch @gol
202 -fobjc-exceptions @gol
204 -freplace-objc-classes @gol
207 -Wassign-intercept @gol
208 -Wno-protocol -Wselector @gol
209 -Wstrict-selector-match @gol
210 -Wundeclared-selector}
212 @item Language Independent Options
213 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
214 @gccoptlist{-fmessage-length=@var{n} @gol
215 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}} @gol
216 -fdiagnostics-show-options
218 @item Warning Options
219 @xref{Warning Options,,Options to Request or Suppress Warnings}.
220 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
221 -w -Wextra -Wall -Waggregate-return -Wno-attributes @gol
222 -Wc++-compat -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
223 -Wconversion -Wno-deprecated-declarations @gol
224 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
225 -Werror -Werror-implicit-function-declaration @gol
226 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
227 -Wno-format-extra-args -Wformat-nonliteral @gol
228 -Wformat-security -Wformat-y2k @gol
229 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
230 -Wimport -Wno-import -Winit-self -Winline @gol
231 -Wno-int-to-pointer-cast @gol
232 -Wno-invalid-offsetof -Winvalid-pch @gol
233 -Wlarger-than-@var{len} -Wunsafe-loop-optimizations -Wlong-long @gol
234 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
235 -Wmissing-format-attribute -Wmissing-include-dirs @gol
236 -Wmissing-noreturn @gol
237 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
238 -Wparentheses -Wpointer-arith -Wno-pointer-to-int-cast @gol
239 -Wredundant-decls @gol
240 -Wreturn-type -Wsequence-point -Wshadow @gol
241 -Wsign-compare -Wstack-protector @gol
242 -Wstrict-aliasing -Wstrict-aliasing=2 @gol
243 -Wswitch -Wswitch-default -Wswitch-enum @gol
244 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
245 -Wunknown-pragmas -Wno-pragmas -Wunreachable-code @gol
246 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
247 -Wunused-value -Wunused-variable -Wvariadic-macros @gol
248 -Wvolatile-register-var -Wwrite-strings}
250 @item C-only Warning Options
251 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
252 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
253 -Wstrict-prototypes -Wtraditional @gol
254 -Wdeclaration-after-statement -Wpointer-sign}
256 @item Debugging Options
257 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
258 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
259 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
260 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
261 -fdump-ipa-all -fdump-ipa-cgraph @gol
263 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
268 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
270 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
271 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
272 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
273 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
274 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
275 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
276 -fdump-tree-nrv -fdump-tree-vect @gol
277 -fdump-tree-sink @gol
278 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
279 -fdump-tree-salias @gol
280 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
281 -fdump-tree-vrp@r{[}-@var{n}@r{]} @gol
282 -ftree-vectorizer-verbose=@var{n} @gol
283 -fdump-tree-storeccp@r{[}-@var{n}@r{]} @gol
284 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
285 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs @gol
286 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
287 -ftest-coverage -ftime-report -fvar-tracking @gol
288 -g -g@var{level} -gcoff -gdwarf-2 @gol
289 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
290 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
291 -print-multi-directory -print-multi-lib @gol
292 -print-prog-name=@var{program} -print-search-dirs -Q @gol
295 @item Optimization Options
296 @xref{Optimize Options,,Options that Control Optimization}.
297 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
298 -falign-labels=@var{n} -falign-loops=@var{n} @gol
299 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
300 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
301 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
302 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
303 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
304 -fdelayed-branch -fdelete-null-pointer-checks -fearly-inlining @gol
305 -fexpensive-optimizations -ffast-math -ffloat-store @gol
306 -fforce-addr -ffunction-sections @gol
307 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
308 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
309 -finline-functions -finline-functions-called-once @gol
310 -finline-limit=@var{n} -fkeep-inline-functions @gol
311 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
312 -fmodulo-sched -fno-branch-count-reg @gol
313 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
314 -fno-function-cse -fno-guess-branch-probability @gol
315 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
316 -funsafe-math-optimizations -funsafe-loop-optimizations -ffinite-math-only @gol
317 -fno-trapping-math -fno-zero-initialized-in-bss @gol
318 -fomit-frame-pointer -foptimize-register-move @gol
319 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
320 -fprofile-generate -fprofile-use @gol
321 -fregmove -frename-registers @gol
322 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
323 -frerun-cse-after-loop -frerun-loop-opt @gol
324 -frounding-math -fschedule-insns -fschedule-insns2 @gol
325 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
326 -fsched-spec-load-dangerous @gol
327 -fsched-stalled-insns=@var{n} -fsched-stalled-insns-dep=@var{n} @gol
328 -fsched2-use-superblocks @gol
329 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
330 -fsignaling-nans -fsingle-precision-constant @gol
331 -fstack-protector -fstack-protector-all @gol
332 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
333 -funroll-all-loops -funroll-loops -fpeel-loops @gol
334 -fsplit-ivs-in-unroller -funswitch-loops @gol
335 -fvariable-expansion-in-unroller @gol
336 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
337 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
338 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
339 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
340 -ftree-vect-loop-version -ftree-salias -fweb @gol
341 -ftree-copy-prop -ftree-store-ccp -ftree-store-copy-prop -fwhole-program @gol
342 --param @var{name}=@var{value}
343 -O -O0 -O1 -O2 -O3 -Os}
345 @item Preprocessor Options
346 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
347 @gccoptlist{-A@var{question}=@var{answer} @gol
348 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
349 -C -dD -dI -dM -dN @gol
350 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
351 -idirafter @var{dir} @gol
352 -include @var{file} -imacros @var{file} @gol
353 -iprefix @var{file} -iwithprefix @var{dir} @gol
354 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
355 -isysroot @var{dir} @gol
356 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
357 -P -fworking-directory -remap @gol
358 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
359 -Xpreprocessor @var{option}}
361 @item Assembler Option
362 @xref{Assembler Options,,Passing Options to the Assembler}.
363 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
366 @xref{Link Options,,Options for Linking}.
367 @gccoptlist{@var{object-file-name} -l@var{library} @gol
368 -nostartfiles -nodefaultlibs -nostdlib -pie -rdynamic @gol
369 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
370 -Wl,@var{option} -Xlinker @var{option} @gol
373 @item Directory Options
374 @xref{Directory Options,,Options for Directory Search}.
375 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir}
376 -specs=@var{file} -I- --sysroot=@var{dir}}
379 @c I wrote this xref this way to avoid overfull hbox. -- rms
380 @xref{Target Options}.
381 @gccoptlist{-V @var{version} -b @var{machine}}
383 @item Machine Dependent Options
384 @xref{Submodel Options,,Hardware Models and Configurations}.
385 @c This list is ordered alphanumerically by subsection name.
386 @c Try and put the significant identifier (CPU or system) first,
387 @c so users have a clue at guessing where the ones they want will be.
390 @gccoptlist{-EB -EL @gol
391 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
392 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
395 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
396 -mabi=@var{name} @gol
397 -mapcs-stack-check -mno-apcs-stack-check @gol
398 -mapcs-float -mno-apcs-float @gol
399 -mapcs-reentrant -mno-apcs-reentrant @gol
400 -msched-prolog -mno-sched-prolog @gol
401 -mlittle-endian -mbig-endian -mwords-little-endian @gol
402 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
403 -mthumb-interwork -mno-thumb-interwork @gol
404 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
405 -mstructure-size-boundary=@var{n} @gol
406 -mabort-on-noreturn @gol
407 -mlong-calls -mno-long-calls @gol
408 -msingle-pic-base -mno-single-pic-base @gol
409 -mpic-register=@var{reg} @gol
410 -mnop-fun-dllimport @gol
411 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
412 -mpoke-function-name @gol
414 -mtpcs-frame -mtpcs-leaf-frame @gol
415 -mcaller-super-interworking -mcallee-super-interworking @gol
419 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
420 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
422 @emph{Blackfin Options}
423 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer @gol
424 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly @gol
425 -mlow-64k -mno-low64k -mid-shared-library @gol
426 -mno-id-shared-library -mshared-library-id=@var{n} @gol
427 -mlong-calls -mno-long-calls}
430 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
431 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
432 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
433 -mstack-align -mdata-align -mconst-align @gol
434 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
435 -melf -maout -melinux -mlinux -sim -sim2 @gol
436 -mmul-bug-workaround -mno-mul-bug-workaround}
439 @gccoptlist{-mmac -mpush-args}
441 @emph{Darwin Options}
442 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
443 -arch_only -bind_at_load -bundle -bundle_loader @gol
444 -client_name -compatibility_version -current_version @gol
446 -dependency-file -dylib_file -dylinker_install_name @gol
447 -dynamic -dynamiclib -exported_symbols_list @gol
448 -filelist -flat_namespace -force_cpusubtype_ALL @gol
449 -force_flat_namespace -headerpad_max_install_names @gol
450 -image_base -init -install_name -keep_private_externs @gol
451 -multi_module -multiply_defined -multiply_defined_unused @gol
452 -noall_load -no_dead_strip_inits_and_terms @gol
453 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
454 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
455 -private_bundle -read_only_relocs -sectalign @gol
456 -sectobjectsymbols -whyload -seg1addr @gol
457 -sectcreate -sectobjectsymbols -sectorder @gol
458 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
459 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
460 -segprot -segs_read_only_addr -segs_read_write_addr @gol
461 -single_module -static -sub_library -sub_umbrella @gol
462 -twolevel_namespace -umbrella -undefined @gol
463 -unexported_symbols_list -weak_reference_mismatches @gol
464 -whatsloaded -F -gused -gfull -mmacosx-version-min=@var{version} @gol
467 @emph{DEC Alpha Options}
468 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
469 -mieee -mieee-with-inexact -mieee-conformant @gol
470 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
471 -mtrap-precision=@var{mode} -mbuild-constants @gol
472 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
473 -mbwx -mmax -mfix -mcix @gol
474 -mfloat-vax -mfloat-ieee @gol
475 -mexplicit-relocs -msmall-data -mlarge-data @gol
476 -msmall-text -mlarge-text @gol
477 -mmemory-latency=@var{time}}
479 @emph{DEC Alpha/VMS Options}
480 @gccoptlist{-mvms-return-codes}
483 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
484 -mhard-float -msoft-float @gol
485 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
486 -mdouble -mno-double @gol
487 -mmedia -mno-media -mmuladd -mno-muladd @gol
488 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
489 -mlinked-fp -mlong-calls -malign-labels @gol
490 -mlibrary-pic -macc-4 -macc-8 @gol
491 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
492 -moptimize-membar -mno-optimize-membar @gol
493 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
494 -mvliw-branch -mno-vliw-branch @gol
495 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
496 -mno-nested-cond-exec -mtomcat-stats @gol
500 @emph{H8/300 Options}
501 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
504 @gccoptlist{-march=@var{architecture-type} @gol
505 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
506 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
507 -mfixed-range=@var{register-range} @gol
508 -mjump-in-delay -mlinker-opt -mlong-calls @gol
509 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
510 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
511 -mno-jump-in-delay -mno-long-load-store @gol
512 -mno-portable-runtime -mno-soft-float @gol
513 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
514 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
515 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
516 -munix=@var{unix-std} -nolibdld -static -threads}
518 @emph{i386 and x86-64 Options}
519 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
520 -mfpmath=@var{unit} @gol
521 -masm=@var{dialect} -mno-fancy-math-387 @gol
522 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
523 -mno-wide-multiply -mrtd -malign-double @gol
524 -mpreferred-stack-boundary=@var{num} @gol
525 -mmmx -msse -msse2 -msse3 -m3dnow @gol
526 -mthreads -mno-align-stringops -minline-all-stringops @gol
527 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
528 -m96bit-long-double -mregparm=@var{num} -msseregparm @gol
529 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs @gol
530 -mcmodel=@var{code-model} @gol
531 -m32 -m64 -mlarge-data-threshold=@var{num}}
534 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
535 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
536 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
537 -minline-float-divide-max-throughput @gol
538 -minline-int-divide-min-latency @gol
539 -minline-int-divide-max-throughput @gol
540 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
541 -mno-dwarf2-asm -mearly-stop-bits @gol
542 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
543 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
545 @emph{M32R/D Options}
546 @gccoptlist{-m32r2 -m32rx -m32r @gol
548 -malign-loops -mno-align-loops @gol
549 -missue-rate=@var{number} @gol
550 -mbranch-cost=@var{number} @gol
551 -mmodel=@var{code-size-model-type} @gol
552 -msdata=@var{sdata-type} @gol
553 -mno-flush-func -mflush-func=@var{name} @gol
554 -mno-flush-trap -mflush-trap=@var{number} @gol
558 @gccoptlist{-mcpu=@var{cpu} -msim -memregs=@var{number}}
560 @emph{M680x0 Options}
561 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
562 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
563 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
564 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
565 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
567 @emph{M68hc1x Options}
568 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
569 -mauto-incdec -minmax -mlong-calls -mshort @gol
570 -msoft-reg-count=@var{count}}
573 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
574 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
575 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
576 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
577 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
580 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
581 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
582 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
583 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
584 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
585 -mdsp -mpaired-single -mips3d @gol
586 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
587 -G@var{num} -membedded-data -mno-embedded-data @gol
588 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
589 -msplit-addresses -mno-split-addresses @gol
590 -mexplicit-relocs -mno-explicit-relocs @gol
591 -mcheck-zero-division -mno-check-zero-division @gol
592 -mdivide-traps -mdivide-breaks @gol
593 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
594 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
595 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
596 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
597 -mfix-sb1 -mno-fix-sb1 @gol
598 -mflush-func=@var{func} -mno-flush-func @gol
599 -mbranch-likely -mno-branch-likely @gol
600 -mfp-exceptions -mno-fp-exceptions @gol
601 -mvr4130-align -mno-vr4130-align}
604 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
605 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
606 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
607 -mno-base-addresses -msingle-exit -mno-single-exit}
609 @emph{MN10300 Options}
610 @gccoptlist{-mmult-bug -mno-mult-bug @gol
611 -mam33 -mno-am33 @gol
612 -mam33-2 -mno-am33-2 @gol
613 -mreturn-pointer-on-d0 @gol
617 @gccoptlist{-mno-crt0 -mbacc -msim @gol
618 -march=@var{cpu-type} }
620 @emph{PDP-11 Options}
621 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
622 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
623 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
624 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
625 -mbranch-expensive -mbranch-cheap @gol
626 -msplit -mno-split -munix-asm -mdec-asm}
628 @emph{PowerPC Options}
629 See RS/6000 and PowerPC Options.
631 @emph{RS/6000 and PowerPC Options}
632 @gccoptlist{-mcpu=@var{cpu-type} @gol
633 -mtune=@var{cpu-type} @gol
634 -mpower -mno-power -mpower2 -mno-power2 @gol
635 -mpowerpc -mpowerpc64 -mno-powerpc @gol
636 -maltivec -mno-altivec @gol
637 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
638 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
639 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mfprnd -mno-fprnd @gol
640 -mnew-mnemonics -mold-mnemonics @gol
641 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
642 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
643 -malign-power -malign-natural @gol
644 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
645 -mstring -mno-string -mupdate -mno-update @gol
646 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
647 -mstrict-align -mno-strict-align -mrelocatable @gol
648 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
649 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
650 -mdynamic-no-pic -maltivec -mswdiv @gol
651 -mprioritize-restricted-insns=@var{priority} @gol
652 -msched-costly-dep=@var{dependence_type} @gol
653 -minsert-sched-nops=@var{scheme} @gol
654 -mcall-sysv -mcall-netbsd @gol
655 -maix-struct-return -msvr4-struct-return @gol
656 -mabi=@var{abi-type} -msecure-plt -mbss-plt @gol
657 -misel -mno-isel @gol
658 -misel=yes -misel=no @gol
660 -mspe=yes -mspe=no @gol
661 -mvrsave -mno-vrsave @gol
662 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
663 -mprototype -mno-prototype @gol
664 -msim -mmvme -mads -myellowknife -memb -msdata @gol
665 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
667 @emph{S/390 and zSeries Options}
668 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
669 -mhard-float -msoft-float -mlong-double-64 -mlong-double-128 @gol
670 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack @gol
671 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
672 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
673 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
674 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
677 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
678 -m4-nofpu -m4-single-only -m4-single -m4 @gol
679 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
680 -m5-64media -m5-64media-nofpu @gol
681 -m5-32media -m5-32media-nofpu @gol
682 -m5-compact -m5-compact-nofpu @gol
683 -mb -ml -mdalign -mrelax @gol
684 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
685 -mieee -misize -mpadstruct -mspace @gol
686 -mprefergot -musermode -multcost=@var{number} -mdiv=@var{strategy} @gol
687 -mdivsi3_libfunc=@var{name} @gol
688 -madjust-unroll -mindexed-addressing -mgettrcost=@var{number} -mpt-fixed @gol
692 @gccoptlist{-mcpu=@var{cpu-type} @gol
693 -mtune=@var{cpu-type} @gol
694 -mcmodel=@var{code-model} @gol
695 -m32 -m64 -mapp-regs -mno-app-regs @gol
696 -mfaster-structs -mno-faster-structs @gol
697 -mfpu -mno-fpu -mhard-float -msoft-float @gol
698 -mhard-quad-float -msoft-quad-float @gol
699 -mimpure-text -mno-impure-text -mlittle-endian @gol
700 -mstack-bias -mno-stack-bias @gol
701 -munaligned-doubles -mno-unaligned-doubles @gol
702 -mv8plus -mno-v8plus -mvis -mno-vis
703 -threads -pthreads -pthread}
705 @emph{System V Options}
706 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
708 @emph{TMS320C3x/C4x Options}
709 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
710 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
711 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
712 -mparallel-insns -mparallel-mpy -mpreserve-float}
715 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
716 -mprolog-function -mno-prolog-function -mspace @gol
717 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
718 -mapp-regs -mno-app-regs @gol
719 -mdisable-callt -mno-disable-callt @gol
725 @gccoptlist{-mg -mgnu -munix}
727 @emph{x86-64 Options}
728 See i386 and x86-64 Options.
730 @emph{Xstormy16 Options}
733 @emph{Xtensa Options}
734 @gccoptlist{-mconst16 -mno-const16 @gol
735 -mfused-madd -mno-fused-madd @gol
736 -mtext-section-literals -mno-text-section-literals @gol
737 -mtarget-align -mno-target-align @gol
738 -mlongcalls -mno-longcalls}
740 @emph{zSeries Options}
741 See S/390 and zSeries Options.
743 @item Code Generation Options
744 @xref{Code Gen Options,,Options for Code Generation Conventions}.
745 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
746 -ffixed-@var{reg} -fexceptions @gol
747 -fnon-call-exceptions -funwind-tables @gol
748 -fasynchronous-unwind-tables @gol
749 -finhibit-size-directive -finstrument-functions @gol
750 -fno-common -fno-ident @gol
751 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
752 -fno-jump-tables @gol
753 -freg-struct-return -fshared-data -fshort-enums @gol
754 -fshort-double -fshort-wchar @gol
755 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
756 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
757 -fargument-alias -fargument-noalias @gol
758 -fargument-noalias-global -fleading-underscore @gol
759 -ftls-model=@var{model} @gol
760 -ftrapv -fwrapv -fbounds-check @gol
765 * Overall Options:: Controlling the kind of output:
766 an executable, object files, assembler files,
767 or preprocessed source.
768 * C Dialect Options:: Controlling the variant of C language compiled.
769 * C++ Dialect Options:: Variations on C++.
770 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
772 * Language Independent Options:: Controlling how diagnostics should be
774 * Warning Options:: How picky should the compiler be?
775 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
776 * Optimize Options:: How much optimization?
777 * Preprocessor Options:: Controlling header files and macro definitions.
778 Also, getting dependency information for Make.
779 * Assembler Options:: Passing options to the assembler.
780 * Link Options:: Specifying libraries and so on.
781 * Directory Options:: Where to find header files and libraries.
782 Where to find the compiler executable files.
783 * Spec Files:: How to pass switches to sub-processes.
784 * Target Options:: Running a cross-compiler, or an old version of GCC.
787 @node Overall Options
788 @section Options Controlling the Kind of Output
790 Compilation can involve up to four stages: preprocessing, compilation
791 proper, assembly and linking, always in that order. GCC is capable of
792 preprocessing and compiling several files either into several
793 assembler input files, or into one assembler input file; then each
794 assembler input file produces an object file, and linking combines all
795 the object files (those newly compiled, and those specified as input)
796 into an executable file.
798 @cindex file name suffix
799 For any given input file, the file name suffix determines what kind of
804 C source code which must be preprocessed.
807 C source code which should not be preprocessed.
810 C++ source code which should not be preprocessed.
813 Objective-C source code. Note that you must link with the @file{libobjc}
814 library to make an Objective-C program work.
817 Objective-C source code which should not be preprocessed.
821 Objective-C++ source code. Note that you must link with the @file{libobjc}
822 library to make an Objective-C++ program work. Note that @samp{.M} refers
823 to a literal capital M@.
826 Objective-C++ source code which should not be preprocessed.
829 C, C++, Objective-C or Objective-C++ header file to be turned into a
834 @itemx @var{file}.cxx
835 @itemx @var{file}.cpp
836 @itemx @var{file}.CPP
837 @itemx @var{file}.c++
839 C++ source code which must be preprocessed. Note that in @samp{.cxx},
840 the last two letters must both be literally @samp{x}. Likewise,
841 @samp{.C} refers to a literal capital C@.
845 Objective-C++ source code which must be preprocessed.
848 Objective-C++ source code which should not be preprocessed.
852 C++ header file to be turned into a precompiled header.
855 @itemx @var{file}.for
856 @itemx @var{file}.FOR
857 Fixed form Fortran source code which should not be preprocessed.
860 @itemx @var{file}.fpp
861 @itemx @var{file}.FPP
862 Fixed form Fortran source code which must be preprocessed (with the traditional
866 @itemx @var{file}.f95
867 Free form Fortran source code which should not be preprocessed.
870 @itemx @var{file}.F95
871 Free form Fortran source code which must be preprocessed (with the
872 traditional preprocessor).
874 @c FIXME: Descriptions of Java file types.
881 Ada source code file which contains a library unit declaration (a
882 declaration of a package, subprogram, or generic, or a generic
883 instantiation), or a library unit renaming declaration (a package,
884 generic, or subprogram renaming declaration). Such files are also
887 @itemx @var{file}.adb
888 Ada source code file containing a library unit body (a subprogram or
889 package body). Such files are also called @dfn{bodies}.
891 @c GCC also knows about some suffixes for languages not yet included:
902 Assembler code which must be preprocessed.
905 An object file to be fed straight into linking.
906 Any file name with no recognized suffix is treated this way.
910 You can specify the input language explicitly with the @option{-x} option:
913 @item -x @var{language}
914 Specify explicitly the @var{language} for the following input files
915 (rather than letting the compiler choose a default based on the file
916 name suffix). This option applies to all following input files until
917 the next @option{-x} option. Possible values for @var{language} are:
919 c c-header c-cpp-output
920 c++ c++-header c++-cpp-output
921 objective-c objective-c-header objective-c-cpp-output
922 objective-c++ objective-c++-header objective-c++-cpp-output
923 assembler assembler-with-cpp
932 Turn off any specification of a language, so that subsequent files are
933 handled according to their file name suffixes (as they are if @option{-x}
934 has not been used at all).
936 @item -pass-exit-codes
937 @opindex pass-exit-codes
938 Normally the @command{gcc} program will exit with the code of 1 if any
939 phase of the compiler returns a non-success return code. If you specify
940 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
941 numerically highest error produced by any phase that returned an error
945 If you only want some of the stages of compilation, you can use
946 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
947 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
948 @command{gcc} is to stop. Note that some combinations (for example,
949 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
954 Compile or assemble the source files, but do not link. The linking
955 stage simply is not done. The ultimate output is in the form of an
956 object file for each source file.
958 By default, the object file name for a source file is made by replacing
959 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
961 Unrecognized input files, not requiring compilation or assembly, are
966 Stop after the stage of compilation proper; do not assemble. The output
967 is in the form of an assembler code file for each non-assembler input
970 By default, the assembler file name for a source file is made by
971 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
973 Input files that don't require compilation are ignored.
977 Stop after the preprocessing stage; do not run the compiler proper. The
978 output is in the form of preprocessed source code, which is sent to the
981 Input files which don't require preprocessing are ignored.
983 @cindex output file option
986 Place output in file @var{file}. This applies regardless to whatever
987 sort of output is being produced, whether it be an executable file,
988 an object file, an assembler file or preprocessed C code.
990 If @option{-o} is not specified, the default is to put an executable
991 file in @file{a.out}, the object file for
992 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
993 assembler file in @file{@var{source}.s}, a precompiled header file in
994 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
999 Print (on standard error output) the commands executed to run the stages
1000 of compilation. Also print the version number of the compiler driver
1001 program and of the preprocessor and the compiler proper.
1005 Like @option{-v} except the commands are not executed and all command
1006 arguments are quoted. This is useful for shell scripts to capture the
1007 driver-generated command lines.
1011 Use pipes rather than temporary files for communication between the
1012 various stages of compilation. This fails to work on some systems where
1013 the assembler is unable to read from a pipe; but the GNU assembler has
1018 If you are compiling multiple source files, this option tells the driver
1019 to pass all the source files to the compiler at once (for those
1020 languages for which the compiler can handle this). This will allow
1021 intermodule analysis (IMA) to be performed by the compiler. Currently the only
1022 language for which this is supported is C@. If you pass source files for
1023 multiple languages to the driver, using this option, the driver will invoke
1024 the compiler(s) that support IMA once each, passing each compiler all the
1025 source files appropriate for it. For those languages that do not support
1026 IMA this option will be ignored, and the compiler will be invoked once for
1027 each source file in that language. If you use this option in conjunction
1028 with @option{-save-temps}, the compiler will generate multiple
1030 (one for each source file), but only one (combined) @file{.o} or
1035 Print (on the standard output) a description of the command line options
1036 understood by @command{gcc}. If the @option{-v} option is also specified
1037 then @option{--help} will also be passed on to the various processes
1038 invoked by @command{gcc}, so that they can display the command line options
1039 they accept. If the @option{-Wextra} option is also specified then command
1040 line options which have no documentation associated with them will also
1044 @opindex target-help
1045 Print (on the standard output) a description of target specific command
1046 line options for each tool.
1050 Display the version number and copyrights of the invoked GCC@.
1054 @section Compiling C++ Programs
1056 @cindex suffixes for C++ source
1057 @cindex C++ source file suffixes
1058 C++ source files conventionally use one of the suffixes @samp{.C},
1059 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1060 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1061 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1062 files with these names and compiles them as C++ programs even if you
1063 call the compiler the same way as for compiling C programs (usually
1064 with the name @command{gcc}).
1068 However, C++ programs often require class libraries as well as a
1069 compiler that understands the C++ language---and under some
1070 circumstances, you might want to compile programs or header files from
1071 standard input, or otherwise without a suffix that flags them as C++
1072 programs. You might also like to precompile a C header file with a
1073 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1074 program that calls GCC with the default language set to C++, and
1075 automatically specifies linking against the C++ library. On many
1076 systems, @command{g++} is also installed with the name @command{c++}.
1078 @cindex invoking @command{g++}
1079 When you compile C++ programs, you may specify many of the same
1080 command-line options that you use for compiling programs in any
1081 language; or command-line options meaningful for C and related
1082 languages; or options that are meaningful only for C++ programs.
1083 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1084 explanations of options for languages related to C@.
1085 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1086 explanations of options that are meaningful only for C++ programs.
1088 @node C Dialect Options
1089 @section Options Controlling C Dialect
1090 @cindex dialect options
1091 @cindex language dialect options
1092 @cindex options, dialect
1094 The following options control the dialect of C (or languages derived
1095 from C, such as C++, Objective-C and Objective-C++) that the compiler
1099 @cindex ANSI support
1103 In C mode, support all ISO C90 programs. In C++ mode,
1104 remove GNU extensions that conflict with ISO C++.
1106 This turns off certain features of GCC that are incompatible with ISO
1107 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1108 such as the @code{asm} and @code{typeof} keywords, and
1109 predefined macros such as @code{unix} and @code{vax} that identify the
1110 type of system you are using. It also enables the undesirable and
1111 rarely used ISO trigraph feature. For the C compiler,
1112 it disables recognition of C++ style @samp{//} comments as well as
1113 the @code{inline} keyword.
1115 The alternate keywords @code{__asm__}, @code{__extension__},
1116 @code{__inline__} and @code{__typeof__} continue to work despite
1117 @option{-ansi}. You would not want to use them in an ISO C program, of
1118 course, but it is useful to put them in header files that might be included
1119 in compilations done with @option{-ansi}. Alternate predefined macros
1120 such as @code{__unix__} and @code{__vax__} are also available, with or
1121 without @option{-ansi}.
1123 The @option{-ansi} option does not cause non-ISO programs to be
1124 rejected gratuitously. For that, @option{-pedantic} is required in
1125 addition to @option{-ansi}. @xref{Warning Options}.
1127 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1128 option is used. Some header files may notice this macro and refrain
1129 from declaring certain functions or defining certain macros that the
1130 ISO standard doesn't call for; this is to avoid interfering with any
1131 programs that might use these names for other things.
1133 Functions which would normally be built in but do not have semantics
1134 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1135 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1136 built-in functions provided by GCC}, for details of the functions
1141 Determine the language standard. This option is currently only
1142 supported when compiling C or C++. A value for this option must be
1143 provided; possible values are
1148 ISO C90 (same as @option{-ansi}).
1150 @item iso9899:199409
1151 ISO C90 as modified in amendment 1.
1157 ISO C99. Note that this standard is not yet fully supported; see
1158 @w{@uref{http://gcc.gnu.org/gcc-4.1/c99status.html}} for more information. The
1159 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1162 Default, ISO C90 plus GNU extensions (including some C99 features).
1166 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1167 this will become the default. The name @samp{gnu9x} is deprecated.
1170 The 1998 ISO C++ standard plus amendments.
1173 The same as @option{-std=c++98} plus GNU extensions. This is the
1174 default for C++ code.
1177 Even when this option is not specified, you can still use some of the
1178 features of newer standards in so far as they do not conflict with
1179 previous C standards. For example, you may use @code{__restrict__} even
1180 when @option{-std=c99} is not specified.
1182 The @option{-std} options specifying some version of ISO C have the same
1183 effects as @option{-ansi}, except that features that were not in ISO C90
1184 but are in the specified version (for example, @samp{//} comments and
1185 the @code{inline} keyword in ISO C99) are not disabled.
1187 @xref{Standards,,Language Standards Supported by GCC}, for details of
1188 these standard versions.
1190 @item -aux-info @var{filename}
1192 Output to the given filename prototyped declarations for all functions
1193 declared and/or defined in a translation unit, including those in header
1194 files. This option is silently ignored in any language other than C@.
1196 Besides declarations, the file indicates, in comments, the origin of
1197 each declaration (source file and line), whether the declaration was
1198 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1199 @samp{O} for old, respectively, in the first character after the line
1200 number and the colon), and whether it came from a declaration or a
1201 definition (@samp{C} or @samp{F}, respectively, in the following
1202 character). In the case of function definitions, a K&R-style list of
1203 arguments followed by their declarations is also provided, inside
1204 comments, after the declaration.
1208 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1209 keyword, so that code can use these words as identifiers. You can use
1210 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1211 instead. @option{-ansi} implies @option{-fno-asm}.
1213 In C++, this switch only affects the @code{typeof} keyword, since
1214 @code{asm} and @code{inline} are standard keywords. You may want to
1215 use the @option{-fno-gnu-keywords} flag instead, which has the same
1216 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1217 switch only affects the @code{asm} and @code{typeof} keywords, since
1218 @code{inline} is a standard keyword in ISO C99.
1221 @itemx -fno-builtin-@var{function}
1222 @opindex fno-builtin
1223 @cindex built-in functions
1224 Don't recognize built-in functions that do not begin with
1225 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1226 functions provided by GCC}, for details of the functions affected,
1227 including those which are not built-in functions when @option{-ansi} or
1228 @option{-std} options for strict ISO C conformance are used because they
1229 do not have an ISO standard meaning.
1231 GCC normally generates special code to handle certain built-in functions
1232 more efficiently; for instance, calls to @code{alloca} may become single
1233 instructions that adjust the stack directly, and calls to @code{memcpy}
1234 may become inline copy loops. The resulting code is often both smaller
1235 and faster, but since the function calls no longer appear as such, you
1236 cannot set a breakpoint on those calls, nor can you change the behavior
1237 of the functions by linking with a different library. In addition,
1238 when a function is recognized as a built-in function, GCC may use
1239 information about that function to warn about problems with calls to
1240 that function, or to generate more efficient code, even if the
1241 resulting code still contains calls to that function. For example,
1242 warnings are given with @option{-Wformat} for bad calls to
1243 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1244 known not to modify global memory.
1246 With the @option{-fno-builtin-@var{function}} option
1247 only the built-in function @var{function} is
1248 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1249 function is named this is not built-in in this version of GCC, this
1250 option is ignored. There is no corresponding
1251 @option{-fbuiltin-@var{function}} option; if you wish to enable
1252 built-in functions selectively when using @option{-fno-builtin} or
1253 @option{-ffreestanding}, you may define macros such as:
1256 #define abs(n) __builtin_abs ((n))
1257 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1262 @cindex hosted environment
1264 Assert that compilation takes place in a hosted environment. This implies
1265 @option{-fbuiltin}. A hosted environment is one in which the
1266 entire standard library is available, and in which @code{main} has a return
1267 type of @code{int}. Examples are nearly everything except a kernel.
1268 This is equivalent to @option{-fno-freestanding}.
1270 @item -ffreestanding
1271 @opindex ffreestanding
1272 @cindex hosted environment
1274 Assert that compilation takes place in a freestanding environment. This
1275 implies @option{-fno-builtin}. A freestanding environment
1276 is one in which the standard library may not exist, and program startup may
1277 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1278 This is equivalent to @option{-fno-hosted}.
1280 @xref{Standards,,Language Standards Supported by GCC}, for details of
1281 freestanding and hosted environments.
1283 @item -fms-extensions
1284 @opindex fms-extensions
1285 Accept some non-standard constructs used in Microsoft header files.
1287 Some cases of unnamed fields in structures and unions are only
1288 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1289 fields within structs/unions}, for details.
1293 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1294 options for strict ISO C conformance) implies @option{-trigraphs}.
1296 @item -no-integrated-cpp
1297 @opindex no-integrated-cpp
1298 Performs a compilation in two passes: preprocessing and compiling. This
1299 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1300 @option{-B} option. The user supplied compilation step can then add in
1301 an additional preprocessing step after normal preprocessing but before
1302 compiling. The default is to use the integrated cpp (internal cpp)
1304 The semantics of this option will change if "cc1", "cc1plus", and
1305 "cc1obj" are merged.
1307 @cindex traditional C language
1308 @cindex C language, traditional
1310 @itemx -traditional-cpp
1311 @opindex traditional-cpp
1312 @opindex traditional
1313 Formerly, these options caused GCC to attempt to emulate a pre-standard
1314 C compiler. They are now only supported with the @option{-E} switch.
1315 The preprocessor continues to support a pre-standard mode. See the GNU
1316 CPP manual for details.
1318 @item -fcond-mismatch
1319 @opindex fcond-mismatch
1320 Allow conditional expressions with mismatched types in the second and
1321 third arguments. The value of such an expression is void. This option
1322 is not supported for C++.
1324 @item -funsigned-char
1325 @opindex funsigned-char
1326 Let the type @code{char} be unsigned, like @code{unsigned char}.
1328 Each kind of machine has a default for what @code{char} should
1329 be. It is either like @code{unsigned char} by default or like
1330 @code{signed char} by default.
1332 Ideally, a portable program should always use @code{signed char} or
1333 @code{unsigned char} when it depends on the signedness of an object.
1334 But many programs have been written to use plain @code{char} and
1335 expect it to be signed, or expect it to be unsigned, depending on the
1336 machines they were written for. This option, and its inverse, let you
1337 make such a program work with the opposite default.
1339 The type @code{char} is always a distinct type from each of
1340 @code{signed char} or @code{unsigned char}, even though its behavior
1341 is always just like one of those two.
1344 @opindex fsigned-char
1345 Let the type @code{char} be signed, like @code{signed char}.
1347 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1348 the negative form of @option{-funsigned-char}. Likewise, the option
1349 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1351 @item -fsigned-bitfields
1352 @itemx -funsigned-bitfields
1353 @itemx -fno-signed-bitfields
1354 @itemx -fno-unsigned-bitfields
1355 @opindex fsigned-bitfields
1356 @opindex funsigned-bitfields
1357 @opindex fno-signed-bitfields
1358 @opindex fno-unsigned-bitfields
1359 These options control whether a bit-field is signed or unsigned, when the
1360 declaration does not use either @code{signed} or @code{unsigned}. By
1361 default, such a bit-field is signed, because this is consistent: the
1362 basic integer types such as @code{int} are signed types.
1365 @node C++ Dialect Options
1366 @section Options Controlling C++ Dialect
1368 @cindex compiler options, C++
1369 @cindex C++ options, command line
1370 @cindex options, C++
1371 This section describes the command-line options that are only meaningful
1372 for C++ programs; but you can also use most of the GNU compiler options
1373 regardless of what language your program is in. For example, you
1374 might compile a file @code{firstClass.C} like this:
1377 g++ -g -frepo -O -c firstClass.C
1381 In this example, only @option{-frepo} is an option meant
1382 only for C++ programs; you can use the other options with any
1383 language supported by GCC@.
1385 Here is a list of options that are @emph{only} for compiling C++ programs:
1389 @item -fabi-version=@var{n}
1390 @opindex fabi-version
1391 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1392 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1393 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1394 the version that conforms most closely to the C++ ABI specification.
1395 Therefore, the ABI obtained using version 0 will change as ABI bugs
1398 The default is version 2.
1400 @item -fno-access-control
1401 @opindex fno-access-control
1402 Turn off all access checking. This switch is mainly useful for working
1403 around bugs in the access control code.
1407 Check that the pointer returned by @code{operator new} is non-null
1408 before attempting to modify the storage allocated. This check is
1409 normally unnecessary because the C++ standard specifies that
1410 @code{operator new} will only return @code{0} if it is declared
1411 @samp{throw()}, in which case the compiler will always check the
1412 return value even without this option. In all other cases, when
1413 @code{operator new} has a non-empty exception specification, memory
1414 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1415 @samp{new (nothrow)}.
1417 @item -fconserve-space
1418 @opindex fconserve-space
1419 Put uninitialized or runtime-initialized global variables into the
1420 common segment, as C does. This saves space in the executable at the
1421 cost of not diagnosing duplicate definitions. If you compile with this
1422 flag and your program mysteriously crashes after @code{main()} has
1423 completed, you may have an object that is being destroyed twice because
1424 two definitions were merged.
1426 This option is no longer useful on most targets, now that support has
1427 been added for putting variables into BSS without making them common.
1429 @item -ffriend-injection
1430 @opindex ffriend-injection
1431 Inject friend functions into the enclosing namespace, so that they are
1432 visible outside the scope of the class in which they are declared.
1433 Friend functions were documented to work this way in the old Annotated
1434 C++ Reference Manual, and versions of G++ before 4.1 always worked
1435 that way. However, in ISO C++ a friend function which is not declared
1436 in an enclosing scope can only be found using argument dependent
1437 lookup. This option causes friends to be injected as they were in
1440 This option is for compatibility, and may be removed in a future
1443 @item -fno-const-strings
1444 @opindex fno-const-strings
1445 Give string constants type @code{char *} instead of type @code{const
1446 char *}. By default, G++ uses type @code{const char *} as required by
1447 the standard. Even if you use @option{-fno-const-strings}, you cannot
1448 actually modify the value of a string constant.
1450 This option might be removed in a future release of G++. For maximum
1451 portability, you should structure your code so that it works with
1452 string constants that have type @code{const char *}.
1454 @item -fno-elide-constructors
1455 @opindex fno-elide-constructors
1456 The C++ standard allows an implementation to omit creating a temporary
1457 which is only used to initialize another object of the same type.
1458 Specifying this option disables that optimization, and forces G++ to
1459 call the copy constructor in all cases.
1461 @item -fno-enforce-eh-specs
1462 @opindex fno-enforce-eh-specs
1463 Don't generate code to check for violation of exception specifications
1464 at runtime. This option violates the C++ standard, but may be useful
1465 for reducing code size in production builds, much like defining
1466 @samp{NDEBUG}. This does not give user code permission to throw
1467 exceptions in violation of the exception specifications; the compiler
1468 will still optimize based on the specifications, so throwing an
1469 unexpected exception will result in undefined behavior.
1472 @itemx -fno-for-scope
1474 @opindex fno-for-scope
1475 If @option{-ffor-scope} is specified, the scope of variables declared in
1476 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1477 as specified by the C++ standard.
1478 If @option{-fno-for-scope} is specified, the scope of variables declared in
1479 a @i{for-init-statement} extends to the end of the enclosing scope,
1480 as was the case in old versions of G++, and other (traditional)
1481 implementations of C++.
1483 The default if neither flag is given to follow the standard,
1484 but to allow and give a warning for old-style code that would
1485 otherwise be invalid, or have different behavior.
1487 @item -fno-gnu-keywords
1488 @opindex fno-gnu-keywords
1489 Do not recognize @code{typeof} as a keyword, so that code can use this
1490 word as an identifier. You can use the keyword @code{__typeof__} instead.
1491 @option{-ansi} implies @option{-fno-gnu-keywords}.
1493 @item -fno-implicit-templates
1494 @opindex fno-implicit-templates
1495 Never emit code for non-inline templates which are instantiated
1496 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1497 @xref{Template Instantiation}, for more information.
1499 @item -fno-implicit-inline-templates
1500 @opindex fno-implicit-inline-templates
1501 Don't emit code for implicit instantiations of inline templates, either.
1502 The default is to handle inlines differently so that compiles with and
1503 without optimization will need the same set of explicit instantiations.
1505 @item -fno-implement-inlines
1506 @opindex fno-implement-inlines
1507 To save space, do not emit out-of-line copies of inline functions
1508 controlled by @samp{#pragma implementation}. This will cause linker
1509 errors if these functions are not inlined everywhere they are called.
1511 @item -fms-extensions
1512 @opindex fms-extensions
1513 Disable pedantic warnings about constructs used in MFC, such as implicit
1514 int and getting a pointer to member function via non-standard syntax.
1516 @item -fno-nonansi-builtins
1517 @opindex fno-nonansi-builtins
1518 Disable built-in declarations of functions that are not mandated by
1519 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1520 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1522 @item -fno-operator-names
1523 @opindex fno-operator-names
1524 Do not treat the operator name keywords @code{and}, @code{bitand},
1525 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1526 synonyms as keywords.
1528 @item -fno-optional-diags
1529 @opindex fno-optional-diags
1530 Disable diagnostics that the standard says a compiler does not need to
1531 issue. Currently, the only such diagnostic issued by G++ is the one for
1532 a name having multiple meanings within a class.
1535 @opindex fpermissive
1536 Downgrade some diagnostics about nonconformant code from errors to
1537 warnings. Thus, using @option{-fpermissive} will allow some
1538 nonconforming code to compile.
1542 Enable automatic template instantiation at link time. This option also
1543 implies @option{-fno-implicit-templates}. @xref{Template
1544 Instantiation}, for more information.
1548 Disable generation of information about every class with virtual
1549 functions for use by the C++ runtime type identification features
1550 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1551 of the language, you can save some space by using this flag. Note that
1552 exception handling uses the same information, but it will generate it as
1557 Emit statistics about front-end processing at the end of the compilation.
1558 This information is generally only useful to the G++ development team.
1560 @item -ftemplate-depth-@var{n}
1561 @opindex ftemplate-depth
1562 Set the maximum instantiation depth for template classes to @var{n}.
1563 A limit on the template instantiation depth is needed to detect
1564 endless recursions during template class instantiation. ANSI/ISO C++
1565 conforming programs must not rely on a maximum depth greater than 17.
1567 @item -fno-threadsafe-statics
1568 @opindex fno-threadsafe-statics
1569 Do not emit the extra code to use the routines specified in the C++
1570 ABI for thread-safe initialization of local statics. You can use this
1571 option to reduce code size slightly in code that doesn't need to be
1574 @item -fuse-cxa-atexit
1575 @opindex fuse-cxa-atexit
1576 Register destructors for objects with static storage duration with the
1577 @code{__cxa_atexit} function rather than the @code{atexit} function.
1578 This option is required for fully standards-compliant handling of static
1579 destructors, but will only work if your C library supports
1580 @code{__cxa_atexit}.
1582 @item -fvisibility-inlines-hidden
1583 @opindex fvisibility-inlines-hidden
1584 Causes all inlined methods to be marked with
1585 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1586 appear in the export table of a DSO and do not require a PLT indirection
1587 when used within the DSO@. Enabling this option can have a dramatic effect
1588 on load and link times of a DSO as it massively reduces the size of the
1589 dynamic export table when the library makes heavy use of templates. While
1590 it can cause bloating through duplication of code within each DSO where
1591 it is used, often the wastage is less than the considerable space occupied
1592 by a long symbol name in the export table which is typical when using
1593 templates and namespaces. For even more savings, combine with the
1594 @option{-fvisibility=hidden} switch.
1598 Do not use weak symbol support, even if it is provided by the linker.
1599 By default, G++ will use weak symbols if they are available. This
1600 option exists only for testing, and should not be used by end-users;
1601 it will result in inferior code and has no benefits. This option may
1602 be removed in a future release of G++.
1606 Do not search for header files in the standard directories specific to
1607 C++, but do still search the other standard directories. (This option
1608 is used when building the C++ library.)
1611 In addition, these optimization, warning, and code generation options
1612 have meanings only for C++ programs:
1615 @item -fno-default-inline
1616 @opindex fno-default-inline
1617 Do not assume @samp{inline} for functions defined inside a class scope.
1618 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1619 functions will have linkage like inline functions; they just won't be
1622 @item -Wabi @r{(C++ only)}
1624 Warn when G++ generates code that is probably not compatible with the
1625 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1626 all such cases, there are probably some cases that are not warned about,
1627 even though G++ is generating incompatible code. There may also be
1628 cases where warnings are emitted even though the code that is generated
1631 You should rewrite your code to avoid these warnings if you are
1632 concerned about the fact that code generated by G++ may not be binary
1633 compatible with code generated by other compilers.
1635 The known incompatibilities at this point include:
1640 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1641 pack data into the same byte as a base class. For example:
1644 struct A @{ virtual void f(); int f1 : 1; @};
1645 struct B : public A @{ int f2 : 1; @};
1649 In this case, G++ will place @code{B::f2} into the same byte
1650 as@code{A::f1}; other compilers will not. You can avoid this problem
1651 by explicitly padding @code{A} so that its size is a multiple of the
1652 byte size on your platform; that will cause G++ and other compilers to
1653 layout @code{B} identically.
1656 Incorrect handling of tail-padding for virtual bases. G++ does not use
1657 tail padding when laying out virtual bases. For example:
1660 struct A @{ virtual void f(); char c1; @};
1661 struct B @{ B(); char c2; @};
1662 struct C : public A, public virtual B @{@};
1666 In this case, G++ will not place @code{B} into the tail-padding for
1667 @code{A}; other compilers will. You can avoid this problem by
1668 explicitly padding @code{A} so that its size is a multiple of its
1669 alignment (ignoring virtual base classes); that will cause G++ and other
1670 compilers to layout @code{C} identically.
1673 Incorrect handling of bit-fields with declared widths greater than that
1674 of their underlying types, when the bit-fields appear in a union. For
1678 union U @{ int i : 4096; @};
1682 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1683 union too small by the number of bits in an @code{int}.
1686 Empty classes can be placed at incorrect offsets. For example:
1696 struct C : public B, public A @{@};
1700 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1701 it should be placed at offset zero. G++ mistakenly believes that the
1702 @code{A} data member of @code{B} is already at offset zero.
1705 Names of template functions whose types involve @code{typename} or
1706 template template parameters can be mangled incorrectly.
1709 template <typename Q>
1710 void f(typename Q::X) @{@}
1712 template <template <typename> class Q>
1713 void f(typename Q<int>::X) @{@}
1717 Instantiations of these templates may be mangled incorrectly.
1721 @item -Wctor-dtor-privacy @r{(C++ only)}
1722 @opindex Wctor-dtor-privacy
1723 Warn when a class seems unusable because all the constructors or
1724 destructors in that class are private, and it has neither friends nor
1725 public static member functions.
1727 @item -Wnon-virtual-dtor @r{(C++ only)}
1728 @opindex Wnon-virtual-dtor
1729 Warn when a class appears to be polymorphic, thereby requiring a virtual
1730 destructor, yet it declares a non-virtual one.
1731 This warning is enabled by @option{-Wall}.
1733 @item -Wreorder @r{(C++ only)}
1735 @cindex reordering, warning
1736 @cindex warning for reordering of member initializers
1737 Warn when the order of member initializers given in the code does not
1738 match the order in which they must be executed. For instance:
1744 A(): j (0), i (1) @{ @}
1748 The compiler will rearrange the member initializers for @samp{i}
1749 and @samp{j} to match the declaration order of the members, emitting
1750 a warning to that effect. This warning is enabled by @option{-Wall}.
1753 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1756 @item -Weffc++ @r{(C++ only)}
1758 Warn about violations of the following style guidelines from Scott Meyers'
1759 @cite{Effective C++} book:
1763 Item 11: Define a copy constructor and an assignment operator for classes
1764 with dynamically allocated memory.
1767 Item 12: Prefer initialization to assignment in constructors.
1770 Item 14: Make destructors virtual in base classes.
1773 Item 15: Have @code{operator=} return a reference to @code{*this}.
1776 Item 23: Don't try to return a reference when you must return an object.
1780 Also warn about violations of the following style guidelines from
1781 Scott Meyers' @cite{More Effective C++} book:
1785 Item 6: Distinguish between prefix and postfix forms of increment and
1786 decrement operators.
1789 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1793 When selecting this option, be aware that the standard library
1794 headers do not obey all of these guidelines; use @samp{grep -v}
1795 to filter out those warnings.
1797 @item -Wno-deprecated @r{(C++ only)}
1798 @opindex Wno-deprecated
1799 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1801 @item -Wstrict-null-sentinel @r{(C++ only)}
1802 @opindex Wstrict-null-sentinel
1803 Warn also about the use of an uncasted @code{NULL} as sentinel. When
1804 compiling only with GCC this is a valid sentinel, as @code{NULL} is defined
1805 to @code{__null}. Although it is a null pointer constant not a null pointer,
1806 it is guaranteed to of the same size as a pointer. But this use is
1807 not portable across different compilers.
1809 @item -Wno-non-template-friend @r{(C++ only)}
1810 @opindex Wno-non-template-friend
1811 Disable warnings when non-templatized friend functions are declared
1812 within a template. Since the advent of explicit template specification
1813 support in G++, if the name of the friend is an unqualified-id (i.e.,
1814 @samp{friend foo(int)}), the C++ language specification demands that the
1815 friend declare or define an ordinary, nontemplate function. (Section
1816 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1817 could be interpreted as a particular specialization of a templatized
1818 function. Because this non-conforming behavior is no longer the default
1819 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1820 check existing code for potential trouble spots and is on by default.
1821 This new compiler behavior can be turned off with
1822 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1823 but disables the helpful warning.
1825 @item -Wold-style-cast @r{(C++ only)}
1826 @opindex Wold-style-cast
1827 Warn if an old-style (C-style) cast to a non-void type is used within
1828 a C++ program. The new-style casts (@samp{dynamic_cast},
1829 @samp{static_cast}, @samp{reinterpret_cast}, and @samp{const_cast}) are
1830 less vulnerable to unintended effects and much easier to search for.
1832 @item -Woverloaded-virtual @r{(C++ only)}
1833 @opindex Woverloaded-virtual
1834 @cindex overloaded virtual fn, warning
1835 @cindex warning for overloaded virtual fn
1836 Warn when a function declaration hides virtual functions from a
1837 base class. For example, in:
1844 struct B: public A @{
1849 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1857 will fail to compile.
1859 @item -Wno-pmf-conversions @r{(C++ only)}
1860 @opindex Wno-pmf-conversions
1861 Disable the diagnostic for converting a bound pointer to member function
1864 @item -Wsign-promo @r{(C++ only)}
1865 @opindex Wsign-promo
1866 Warn when overload resolution chooses a promotion from unsigned or
1867 enumerated type to a signed type, over a conversion to an unsigned type of
1868 the same size. Previous versions of G++ would try to preserve
1869 unsignedness, but the standard mandates the current behavior.
1874 A& operator = (int);
1884 In this example, G++ will synthesize a default @samp{A& operator =
1885 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1888 @node Objective-C and Objective-C++ Dialect Options
1889 @section Options Controlling Objective-C and Objective-C++ Dialects
1891 @cindex compiler options, Objective-C and Objective-C++
1892 @cindex Objective-C and Objective-C++ options, command line
1893 @cindex options, Objective-C and Objective-C++
1894 (NOTE: This manual does not describe the Objective-C and Objective-C++
1895 languages themselves. See @xref{Standards,,Language Standards
1896 Supported by GCC}, for references.)
1898 This section describes the command-line options that are only meaningful
1899 for Objective-C and Objective-C++ programs, but you can also use most of
1900 the language-independent GNU compiler options.
1901 For example, you might compile a file @code{some_class.m} like this:
1904 gcc -g -fgnu-runtime -O -c some_class.m
1908 In this example, @option{-fgnu-runtime} is an option meant only for
1909 Objective-C and Objective-C++ programs; you can use the other options with
1910 any language supported by GCC@.
1912 Note that since Objective-C is an extension of the C language, Objective-C
1913 compilations may also use options specific to the C front-end (e.g.,
1914 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1915 C++-specific options (e.g., @option{-Wabi}).
1917 Here is a list of options that are @emph{only} for compiling Objective-C
1918 and Objective-C++ programs:
1921 @item -fconstant-string-class=@var{class-name}
1922 @opindex fconstant-string-class
1923 Use @var{class-name} as the name of the class to instantiate for each
1924 literal string specified with the syntax @code{@@"@dots{}"}. The default
1925 class name is @code{NXConstantString} if the GNU runtime is being used, and
1926 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1927 @option{-fconstant-cfstrings} option, if also present, will override the
1928 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1929 to be laid out as constant CoreFoundation strings.
1932 @opindex fgnu-runtime
1933 Generate object code compatible with the standard GNU Objective-C
1934 runtime. This is the default for most types of systems.
1936 @item -fnext-runtime
1937 @opindex fnext-runtime
1938 Generate output compatible with the NeXT runtime. This is the default
1939 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1940 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1943 @item -fno-nil-receivers
1944 @opindex fno-nil-receivers
1945 Assume that all Objective-C message dispatches (e.g.,
1946 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1947 is not @code{nil}. This allows for more efficient entry points in the runtime
1948 to be used. Currently, this option is only available in conjunction with
1949 the NeXT runtime on Mac OS X 10.3 and later.
1951 @item -fobjc-call-cxx-cdtors
1952 @opindex fobjc-call-cxx-cdtors
1953 For each Objective-C class, check if any of its instance variables is a
1954 C++ object with a non-trivial default constructor. If so, synthesize a
1955 special @code{- (id) .cxx_construct} instance method that will run
1956 non-trivial default constructors on any such instance variables, in order,
1957 and then return @code{self}. Similarly, check if any instance variable
1958 is a C++ object with a non-trivial destructor, and if so, synthesize a
1959 special @code{- (void) .cxx_destruct} method that will run
1960 all such default destructors, in reverse order.
1962 The @code{- (id) .cxx_construct} and/or @code{- (void) .cxx_destruct} methods
1963 thusly generated will only operate on instance variables declared in the
1964 current Objective-C class, and not those inherited from superclasses. It
1965 is the responsibility of the Objective-C runtime to invoke all such methods
1966 in an object's inheritance hierarchy. The @code{- (id) .cxx_construct} methods
1967 will be invoked by the runtime immediately after a new object
1968 instance is allocated; the @code{- (void) .cxx_destruct} methods will
1969 be invoked immediately before the runtime deallocates an object instance.
1971 As of this writing, only the NeXT runtime on Mac OS X 10.4 and later has
1972 support for invoking the @code{- (id) .cxx_construct} and
1973 @code{- (void) .cxx_destruct} methods.
1975 @item -fobjc-direct-dispatch
1976 @opindex fobjc-direct-dispatch
1977 Allow fast jumps to the message dispatcher. On Darwin this is
1978 accomplished via the comm page.
1980 @item -fobjc-exceptions
1981 @opindex fobjc-exceptions
1982 Enable syntactic support for structured exception handling in Objective-C,
1983 similar to what is offered by C++ and Java. Currently, this option is only
1984 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1992 @@catch (AnObjCClass *exc) @{
1999 @@catch (AnotherClass *exc) @{
2002 @@catch (id allOthers) @{
2012 The @code{@@throw} statement may appear anywhere in an Objective-C or
2013 Objective-C++ program; when used inside of a @code{@@catch} block, the
2014 @code{@@throw} may appear without an argument (as shown above), in which case
2015 the object caught by the @code{@@catch} will be rethrown.
2017 Note that only (pointers to) Objective-C objects may be thrown and
2018 caught using this scheme. When an object is thrown, it will be caught
2019 by the nearest @code{@@catch} clause capable of handling objects of that type,
2020 analogously to how @code{catch} blocks work in C++ and Java. A
2021 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
2022 any and all Objective-C exceptions not caught by previous @code{@@catch}
2025 The @code{@@finally} clause, if present, will be executed upon exit from the
2026 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
2027 regardless of whether any exceptions are thrown, caught or rethrown
2028 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
2029 of the @code{finally} clause in Java.
2031 There are several caveats to using the new exception mechanism:
2035 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
2036 idioms provided by the @code{NSException} class, the new
2037 exceptions can only be used on Mac OS X 10.3 (Panther) and later
2038 systems, due to additional functionality needed in the (NeXT) Objective-C
2042 As mentioned above, the new exceptions do not support handling
2043 types other than Objective-C objects. Furthermore, when used from
2044 Objective-C++, the Objective-C exception model does not interoperate with C++
2045 exceptions at this time. This means you cannot @code{@@throw} an exception
2046 from Objective-C and @code{catch} it in C++, or vice versa
2047 (i.e., @code{throw @dots{} @@catch}).
2050 The @option{-fobjc-exceptions} switch also enables the use of synchronization
2051 blocks for thread-safe execution:
2054 @@synchronized (ObjCClass *guard) @{
2059 Upon entering the @code{@@synchronized} block, a thread of execution shall
2060 first check whether a lock has been placed on the corresponding @code{guard}
2061 object by another thread. If it has, the current thread shall wait until
2062 the other thread relinquishes its lock. Once @code{guard} becomes available,
2063 the current thread will place its own lock on it, execute the code contained in
2064 the @code{@@synchronized} block, and finally relinquish the lock (thereby
2065 making @code{guard} available to other threads).
2067 Unlike Java, Objective-C does not allow for entire methods to be marked
2068 @code{@@synchronized}. Note that throwing exceptions out of
2069 @code{@@synchronized} blocks is allowed, and will cause the guarding object
2070 to be unlocked properly.
2074 Enable garbage collection (GC) in Objective-C and Objective-C++ programs.
2076 @item -freplace-objc-classes
2077 @opindex freplace-objc-classes
2078 Emit a special marker instructing @command{ld(1)} not to statically link in
2079 the resulting object file, and allow @command{dyld(1)} to load it in at
2080 run time instead. This is used in conjunction with the Fix-and-Continue
2081 debugging mode, where the object file in question may be recompiled and
2082 dynamically reloaded in the course of program execution, without the need
2083 to restart the program itself. Currently, Fix-and-Continue functionality
2084 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
2089 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
2090 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
2091 compile time) with static class references that get initialized at load time,
2092 which improves run-time performance. Specifying the @option{-fzero-link} flag
2093 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
2094 to be retained. This is useful in Zero-Link debugging mode, since it allows
2095 for individual class implementations to be modified during program execution.
2099 Dump interface declarations for all classes seen in the source file to a
2100 file named @file{@var{sourcename}.decl}.
2102 @item -Wassign-intercept
2103 @opindex Wassign-intercept
2104 Warn whenever an Objective-C assignment is being intercepted by the
2108 @opindex Wno-protocol
2109 If a class is declared to implement a protocol, a warning is issued for
2110 every method in the protocol that is not implemented by the class. The
2111 default behavior is to issue a warning for every method not explicitly
2112 implemented in the class, even if a method implementation is inherited
2113 from the superclass. If you use the @option{-Wno-protocol} option, then
2114 methods inherited from the superclass are considered to be implemented,
2115 and no warning is issued for them.
2119 Warn if multiple methods of different types for the same selector are
2120 found during compilation. The check is performed on the list of methods
2121 in the final stage of compilation. Additionally, a check is performed
2122 for each selector appearing in a @code{@@selector(@dots{})}
2123 expression, and a corresponding method for that selector has been found
2124 during compilation. Because these checks scan the method table only at
2125 the end of compilation, these warnings are not produced if the final
2126 stage of compilation is not reached, for example because an error is
2127 found during compilation, or because the @option{-fsyntax-only} option is
2130 @item -Wstrict-selector-match
2131 @opindex Wstrict-selector-match
2132 Warn if multiple methods with differing argument and/or return types are
2133 found for a given selector when attempting to send a message using this
2134 selector to a receiver of type @code{id} or @code{Class}. When this flag
2135 is off (which is the default behavior), the compiler will omit such warnings
2136 if any differences found are confined to types which share the same size
2139 @item -Wundeclared-selector
2140 @opindex Wundeclared-selector
2141 Warn if a @code{@@selector(@dots{})} expression referring to an
2142 undeclared selector is found. A selector is considered undeclared if no
2143 method with that name has been declared before the
2144 @code{@@selector(@dots{})} expression, either explicitly in an
2145 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2146 an @code{@@implementation} section. This option always performs its
2147 checks as soon as a @code{@@selector(@dots{})} expression is found,
2148 while @option{-Wselector} only performs its checks in the final stage of
2149 compilation. This also enforces the coding style convention
2150 that methods and selectors must be declared before being used.
2152 @item -print-objc-runtime-info
2153 @opindex print-objc-runtime-info
2154 Generate C header describing the largest structure that is passed by
2159 @node Language Independent Options
2160 @section Options to Control Diagnostic Messages Formatting
2161 @cindex options to control diagnostics formatting
2162 @cindex diagnostic messages
2163 @cindex message formatting
2165 Traditionally, diagnostic messages have been formatted irrespective of
2166 the output device's aspect (e.g.@: its width, @dots{}). The options described
2167 below can be used to control the diagnostic messages formatting
2168 algorithm, e.g.@: how many characters per line, how often source location
2169 information should be reported. Right now, only the C++ front end can
2170 honor these options. However it is expected, in the near future, that
2171 the remaining front ends would be able to digest them correctly.
2174 @item -fmessage-length=@var{n}
2175 @opindex fmessage-length
2176 Try to format error messages so that they fit on lines of about @var{n}
2177 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2178 the front ends supported by GCC@. If @var{n} is zero, then no
2179 line-wrapping will be done; each error message will appear on a single
2182 @opindex fdiagnostics-show-location
2183 @item -fdiagnostics-show-location=once
2184 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2185 reporter to emit @emph{once} source location information; that is, in
2186 case the message is too long to fit on a single physical line and has to
2187 be wrapped, the source location won't be emitted (as prefix) again,
2188 over and over, in subsequent continuation lines. This is the default
2191 @item -fdiagnostics-show-location=every-line
2192 Only meaningful in line-wrapping mode. Instructs the diagnostic
2193 messages reporter to emit the same source location information (as
2194 prefix) for physical lines that result from the process of breaking
2195 a message which is too long to fit on a single line.
2197 @item -fdiagnostics-show-options
2198 @opindex fdiagnostics-show-options
2199 This option instructs the diagnostic machinery to add text to each
2200 diagnostic emitted, which indicates which command line option directly
2201 controls that diagnostic, when such an option is known to the
2202 diagnostic machinery.
2206 @node Warning Options
2207 @section Options to Request or Suppress Warnings
2208 @cindex options to control warnings
2209 @cindex warning messages
2210 @cindex messages, warning
2211 @cindex suppressing warnings
2213 Warnings are diagnostic messages that report constructions which
2214 are not inherently erroneous but which are risky or suggest there
2215 may have been an error.
2217 You can request many specific warnings with options beginning @samp{-W},
2218 for example @option{-Wimplicit} to request warnings on implicit
2219 declarations. Each of these specific warning options also has a
2220 negative form beginning @samp{-Wno-} to turn off warnings;
2221 for example, @option{-Wno-implicit}. This manual lists only one of the
2222 two forms, whichever is not the default.
2224 The following options control the amount and kinds of warnings produced
2225 by GCC; for further, language-specific options also refer to
2226 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2230 @cindex syntax checking
2232 @opindex fsyntax-only
2233 Check the code for syntax errors, but don't do anything beyond that.
2237 Issue all the warnings demanded by strict ISO C and ISO C++;
2238 reject all programs that use forbidden extensions, and some other
2239 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2240 version of the ISO C standard specified by any @option{-std} option used.
2242 Valid ISO C and ISO C++ programs should compile properly with or without
2243 this option (though a rare few will require @option{-ansi} or a
2244 @option{-std} option specifying the required version of ISO C)@. However,
2245 without this option, certain GNU extensions and traditional C and C++
2246 features are supported as well. With this option, they are rejected.
2248 @option{-pedantic} does not cause warning messages for use of the
2249 alternate keywords whose names begin and end with @samp{__}. Pedantic
2250 warnings are also disabled in the expression that follows
2251 @code{__extension__}. However, only system header files should use
2252 these escape routes; application programs should avoid them.
2253 @xref{Alternate Keywords}.
2255 Some users try to use @option{-pedantic} to check programs for strict ISO
2256 C conformance. They soon find that it does not do quite what they want:
2257 it finds some non-ISO practices, but not all---only those for which
2258 ISO C @emph{requires} a diagnostic, and some others for which
2259 diagnostics have been added.
2261 A feature to report any failure to conform to ISO C might be useful in
2262 some instances, but would require considerable additional work and would
2263 be quite different from @option{-pedantic}. We don't have plans to
2264 support such a feature in the near future.
2266 Where the standard specified with @option{-std} represents a GNU
2267 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2268 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2269 extended dialect is based. Warnings from @option{-pedantic} are given
2270 where they are required by the base standard. (It would not make sense
2271 for such warnings to be given only for features not in the specified GNU
2272 C dialect, since by definition the GNU dialects of C include all
2273 features the compiler supports with the given option, and there would be
2274 nothing to warn about.)
2276 @item -pedantic-errors
2277 @opindex pedantic-errors
2278 Like @option{-pedantic}, except that errors are produced rather than
2283 Inhibit all warning messages.
2287 Inhibit warning messages about the use of @samp{#import}.
2289 @item -Wchar-subscripts
2290 @opindex Wchar-subscripts
2291 Warn if an array subscript has type @code{char}. This is a common cause
2292 of error, as programmers often forget that this type is signed on some
2294 This warning is enabled by @option{-Wall}.
2298 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2299 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2300 This warning is enabled by @option{-Wall}.
2302 @item -Wfatal-errors
2303 @opindex Wfatal-errors
2304 This option causes the compiler to abort compilation on the first error
2305 occurred rather than trying to keep going and printing further error
2310 @opindex ffreestanding
2311 @opindex fno-builtin
2312 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2313 the arguments supplied have types appropriate to the format string
2314 specified, and that the conversions specified in the format string make
2315 sense. This includes standard functions, and others specified by format
2316 attributes (@pxref{Function Attributes}), in the @code{printf},
2317 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2318 not in the C standard) families (or other target-specific families).
2319 Which functions are checked without format attributes having been
2320 specified depends on the standard version selected, and such checks of
2321 functions without the attribute specified are disabled by
2322 @option{-ffreestanding} or @option{-fno-builtin}.
2324 The formats are checked against the format features supported by GNU
2325 libc version 2.2. These include all ISO C90 and C99 features, as well
2326 as features from the Single Unix Specification and some BSD and GNU
2327 extensions. Other library implementations may not support all these
2328 features; GCC does not support warning about features that go beyond a
2329 particular library's limitations. However, if @option{-pedantic} is used
2330 with @option{-Wformat}, warnings will be given about format features not
2331 in the selected standard version (but not for @code{strfmon} formats,
2332 since those are not in any version of the C standard). @xref{C Dialect
2333 Options,,Options Controlling C Dialect}.
2335 Since @option{-Wformat} also checks for null format arguments for
2336 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2338 @option{-Wformat} is included in @option{-Wall}. For more control over some
2339 aspects of format checking, the options @option{-Wformat-y2k},
2340 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2341 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2342 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2345 @opindex Wformat-y2k
2346 If @option{-Wformat} is specified, also warn about @code{strftime}
2347 formats which may yield only a two-digit year.
2349 @item -Wno-format-extra-args
2350 @opindex Wno-format-extra-args
2351 If @option{-Wformat} is specified, do not warn about excess arguments to a
2352 @code{printf} or @code{scanf} format function. The C standard specifies
2353 that such arguments are ignored.
2355 Where the unused arguments lie between used arguments that are
2356 specified with @samp{$} operand number specifications, normally
2357 warnings are still given, since the implementation could not know what
2358 type to pass to @code{va_arg} to skip the unused arguments. However,
2359 in the case of @code{scanf} formats, this option will suppress the
2360 warning if the unused arguments are all pointers, since the Single
2361 Unix Specification says that such unused arguments are allowed.
2363 @item -Wno-format-zero-length
2364 @opindex Wno-format-zero-length
2365 If @option{-Wformat} is specified, do not warn about zero-length formats.
2366 The C standard specifies that zero-length formats are allowed.
2368 @item -Wformat-nonliteral
2369 @opindex Wformat-nonliteral
2370 If @option{-Wformat} is specified, also warn if the format string is not a
2371 string literal and so cannot be checked, unless the format function
2372 takes its format arguments as a @code{va_list}.
2374 @item -Wformat-security
2375 @opindex Wformat-security
2376 If @option{-Wformat} is specified, also warn about uses of format
2377 functions that represent possible security problems. At present, this
2378 warns about calls to @code{printf} and @code{scanf} functions where the
2379 format string is not a string literal and there are no format arguments,
2380 as in @code{printf (foo);}. This may be a security hole if the format
2381 string came from untrusted input and contains @samp{%n}. (This is
2382 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2383 in future warnings may be added to @option{-Wformat-security} that are not
2384 included in @option{-Wformat-nonliteral}.)
2388 Enable @option{-Wformat} plus format checks not included in
2389 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2390 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2394 Warn about passing a null pointer for arguments marked as
2395 requiring a non-null value by the @code{nonnull} function attribute.
2397 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2398 can be disabled with the @option{-Wno-nonnull} option.
2400 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2402 Warn about uninitialized variables which are initialized with themselves.
2403 Note this option can only be used with the @option{-Wuninitialized} option,
2404 which in turn only works with @option{-O1} and above.
2406 For example, GCC will warn about @code{i} being uninitialized in the
2407 following snippet only when @option{-Winit-self} has been specified:
2418 @item -Wimplicit-int
2419 @opindex Wimplicit-int
2420 Warn when a declaration does not specify a type.
2421 This warning is enabled by @option{-Wall}.
2423 @item -Wimplicit-function-declaration
2424 @itemx -Werror-implicit-function-declaration
2425 @opindex Wimplicit-function-declaration
2426 @opindex Werror-implicit-function-declaration
2427 Give a warning (or error) whenever a function is used before being
2428 declared. The form @option{-Wno-error-implicit-function-declaration}
2430 This warning is enabled by @option{-Wall} (as a warning, not an error).
2434 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2435 This warning is enabled by @option{-Wall}.
2439 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2440 function with external linkage, returning int, taking either zero
2441 arguments, two, or three arguments of appropriate types.
2442 This warning is enabled by @option{-Wall}.
2444 @item -Wmissing-braces
2445 @opindex Wmissing-braces
2446 Warn if an aggregate or union initializer is not fully bracketed. In
2447 the following example, the initializer for @samp{a} is not fully
2448 bracketed, but that for @samp{b} is fully bracketed.
2451 int a[2][2] = @{ 0, 1, 2, 3 @};
2452 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2455 This warning is enabled by @option{-Wall}.
2457 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2458 @opindex Wmissing-include-dirs
2459 Warn if a user-supplied include directory does not exist.
2462 @opindex Wparentheses
2463 Warn if parentheses are omitted in certain contexts, such
2464 as when there is an assignment in a context where a truth value
2465 is expected, or when operators are nested whose precedence people
2466 often get confused about. Only the warning for an assignment used as
2467 a truth value is supported when compiling C++; the other warnings are
2468 only supported when compiling C@.
2470 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2471 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2472 interpretation from that of ordinary mathematical notation.
2474 Also warn about constructions where there may be confusion to which
2475 @code{if} statement an @code{else} branch belongs. Here is an example of
2490 In C, every @code{else} branch belongs to the innermost possible @code{if}
2491 statement, which in this example is @code{if (b)}. This is often not
2492 what the programmer expected, as illustrated in the above example by
2493 indentation the programmer chose. When there is the potential for this
2494 confusion, GCC will issue a warning when this flag is specified.
2495 To eliminate the warning, add explicit braces around the innermost
2496 @code{if} statement so there is no way the @code{else} could belong to
2497 the enclosing @code{if}. The resulting code would look like this:
2513 This warning is enabled by @option{-Wall}.
2515 @item -Wsequence-point
2516 @opindex Wsequence-point
2517 Warn about code that may have undefined semantics because of violations
2518 of sequence point rules in the C standard.
2520 The C standard defines the order in which expressions in a C program are
2521 evaluated in terms of @dfn{sequence points}, which represent a partial
2522 ordering between the execution of parts of the program: those executed
2523 before the sequence point, and those executed after it. These occur
2524 after the evaluation of a full expression (one which is not part of a
2525 larger expression), after the evaluation of the first operand of a
2526 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2527 function is called (but after the evaluation of its arguments and the
2528 expression denoting the called function), and in certain other places.
2529 Other than as expressed by the sequence point rules, the order of
2530 evaluation of subexpressions of an expression is not specified. All
2531 these rules describe only a partial order rather than a total order,
2532 since, for example, if two functions are called within one expression
2533 with no sequence point between them, the order in which the functions
2534 are called is not specified. However, the standards committee have
2535 ruled that function calls do not overlap.
2537 It is not specified when between sequence points modifications to the
2538 values of objects take effect. Programs whose behavior depends on this
2539 have undefined behavior; the C standard specifies that ``Between the
2540 previous and next sequence point an object shall have its stored value
2541 modified at most once by the evaluation of an expression. Furthermore,
2542 the prior value shall be read only to determine the value to be
2543 stored.''. If a program breaks these rules, the results on any
2544 particular implementation are entirely unpredictable.
2546 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2547 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2548 diagnosed by this option, and it may give an occasional false positive
2549 result, but in general it has been found fairly effective at detecting
2550 this sort of problem in programs.
2552 The present implementation of this option only works for C programs. A
2553 future implementation may also work for C++ programs.
2555 The C standard is worded confusingly, therefore there is some debate
2556 over the precise meaning of the sequence point rules in subtle cases.
2557 Links to discussions of the problem, including proposed formal
2558 definitions, may be found on the GCC readings page, at
2559 @w{@uref{http://gcc.gnu.org/readings.html}}.
2561 This warning is enabled by @option{-Wall}.
2564 @opindex Wreturn-type
2565 Warn whenever a function is defined with a return-type that defaults to
2566 @code{int}. Also warn about any @code{return} statement with no
2567 return-value in a function whose return-type is not @code{void}.
2569 For C, also warn if the return type of a function has a type qualifier
2570 such as @code{const}. Such a type qualifier has no effect, since the
2571 value returned by a function is not an lvalue. ISO C prohibits
2572 qualified @code{void} return types on function definitions, so such
2573 return types always receive a warning even without this option.
2575 For C++, a function without return type always produces a diagnostic
2576 message, even when @option{-Wno-return-type} is specified. The only
2577 exceptions are @samp{main} and functions defined in system headers.
2579 This warning is enabled by @option{-Wall}.
2583 Warn whenever a @code{switch} statement has an index of enumerated type
2584 and lacks a @code{case} for one or more of the named codes of that
2585 enumeration. (The presence of a @code{default} label prevents this
2586 warning.) @code{case} labels outside the enumeration range also
2587 provoke warnings when this option is used.
2588 This warning is enabled by @option{-Wall}.
2590 @item -Wswitch-default
2591 @opindex Wswitch-switch
2592 Warn whenever a @code{switch} statement does not have a @code{default}
2596 @opindex Wswitch-enum
2597 Warn whenever a @code{switch} statement has an index of enumerated type
2598 and lacks a @code{case} for one or more of the named codes of that
2599 enumeration. @code{case} labels outside the enumeration range also
2600 provoke warnings when this option is used.
2604 Warn if any trigraphs are encountered that might change the meaning of
2605 the program (trigraphs within comments are not warned about).
2606 This warning is enabled by @option{-Wall}.
2608 @item -Wunused-function
2609 @opindex Wunused-function
2610 Warn whenever a static function is declared but not defined or a
2611 non-inline static function is unused.
2612 This warning is enabled by @option{-Wall}.
2614 @item -Wunused-label
2615 @opindex Wunused-label
2616 Warn whenever a label is declared but not used.
2617 This warning is enabled by @option{-Wall}.
2619 To suppress this warning use the @samp{unused} attribute
2620 (@pxref{Variable Attributes}).
2622 @item -Wunused-parameter
2623 @opindex Wunused-parameter
2624 Warn whenever a function parameter is unused aside from its declaration.
2626 To suppress this warning use the @samp{unused} attribute
2627 (@pxref{Variable Attributes}).
2629 @item -Wunused-variable
2630 @opindex Wunused-variable
2631 Warn whenever a local variable or non-constant static variable is unused
2632 aside from its declaration
2633 This warning is enabled by @option{-Wall}.
2635 To suppress this warning use the @samp{unused} attribute
2636 (@pxref{Variable Attributes}).
2638 @item -Wunused-value
2639 @opindex Wunused-value
2640 Warn whenever a statement computes a result that is explicitly not used.
2641 This warning is enabled by @option{-Wall}.
2643 To suppress this warning cast the expression to @samp{void}.
2647 All the above @option{-Wunused} options combined.
2649 In order to get a warning about an unused function parameter, you must
2650 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2651 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2653 @item -Wuninitialized
2654 @opindex Wuninitialized
2655 Warn if an automatic variable is used without first being initialized or
2656 if a variable may be clobbered by a @code{setjmp} call.
2658 These warnings are possible only in optimizing compilation,
2659 because they require data flow information that is computed only
2660 when optimizing. If you don't specify @option{-O}, you simply won't
2663 If you want to warn about code which uses the uninitialized value of the
2664 variable in its own initializer, use the @option{-Winit-self} option.
2666 These warnings occur for individual uninitialized or clobbered
2667 elements of structure, union or array variables as well as for
2668 variables which are uninitialized or clobbered as a whole. They do
2669 not occur for variables or elements declared @code{volatile}. Because
2670 these warnings depend on optimization, the exact variables or elements
2671 for which there are warnings will depend on the precise optimization
2672 options and version of GCC used.
2674 Note that there may be no warning about a variable that is used only
2675 to compute a value that itself is never used, because such
2676 computations may be deleted by data flow analysis before the warnings
2679 These warnings are made optional because GCC is not smart
2680 enough to see all the reasons why the code might be correct
2681 despite appearing to have an error. Here is one example of how
2702 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2703 always initialized, but GCC doesn't know this. Here is
2704 another common case:
2709 if (change_y) save_y = y, y = new_y;
2711 if (change_y) y = save_y;
2716 This has no bug because @code{save_y} is used only if it is set.
2718 @cindex @code{longjmp} warnings
2719 This option also warns when a non-volatile automatic variable might be
2720 changed by a call to @code{longjmp}. These warnings as well are possible
2721 only in optimizing compilation.
2723 The compiler sees only the calls to @code{setjmp}. It cannot know
2724 where @code{longjmp} will be called; in fact, a signal handler could
2725 call it at any point in the code. As a result, you may get a warning
2726 even when there is in fact no problem because @code{longjmp} cannot
2727 in fact be called at the place which would cause a problem.
2729 Some spurious warnings can be avoided if you declare all the functions
2730 you use that never return as @code{noreturn}. @xref{Function
2733 This warning is enabled by @option{-Wall}.
2735 @item -Wunknown-pragmas
2736 @opindex Wunknown-pragmas
2737 @cindex warning for unknown pragmas
2738 @cindex unknown pragmas, warning
2739 @cindex pragmas, warning of unknown
2740 Warn when a #pragma directive is encountered which is not understood by
2741 GCC@. If this command line option is used, warnings will even be issued
2742 for unknown pragmas in system header files. This is not the case if
2743 the warnings were only enabled by the @option{-Wall} command line option.
2746 @opindex Wno-pragmas
2748 Do not warn about misuses of pragmas, such as incorrect parameters,
2749 invalid syntax, or conflicts between pragmas. See also
2750 @samp{-Wunknown-pragmas}.
2752 @item -Wstrict-aliasing
2753 @opindex Wstrict-aliasing
2754 This option is only active when @option{-fstrict-aliasing} is active.
2755 It warns about code which might break the strict aliasing rules that the
2756 compiler is using for optimization. The warning does not catch all
2757 cases, but does attempt to catch the more common pitfalls. It is
2758 included in @option{-Wall}.
2760 @item -Wstrict-aliasing=2
2761 @opindex Wstrict-aliasing=2
2762 This option is only active when @option{-fstrict-aliasing} is active.
2763 It warns about code which might break the strict aliasing rules that the
2764 compiler is using for optimization. This warning catches more cases than
2765 @option{-Wstrict-aliasing}, but it will also give a warning for some ambiguous
2766 cases that are safe.
2770 All of the above @samp{-W} options combined. This enables all the
2771 warnings about constructions that some users consider questionable, and
2772 that are easy to avoid (or modify to prevent the warning), even in
2773 conjunction with macros. This also enables some language-specific
2774 warnings described in @ref{C++ Dialect Options} and
2775 @ref{Objective-C and Objective-C++ Dialect Options}.
2778 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2779 Some of them warn about constructions that users generally do not
2780 consider questionable, but which occasionally you might wish to check
2781 for; others warn about constructions that are necessary or hard to avoid
2782 in some cases, and there is no simple way to modify the code to suppress
2789 (This option used to be called @option{-W}. The older name is still
2790 supported, but the newer name is more descriptive.) Print extra warning
2791 messages for these events:
2795 A function can return either with or without a value. (Falling
2796 off the end of the function body is considered returning without
2797 a value.) For example, this function would evoke such a
2811 An expression-statement or the left-hand side of a comma expression
2812 contains no side effects.
2813 To suppress the warning, cast the unused expression to void.
2814 For example, an expression such as @samp{x[i,j]} will cause a warning,
2815 but @samp{x[(void)i,j]} will not.
2818 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2821 Storage-class specifiers like @code{static} are not the first things in
2822 a declaration. According to the C Standard, this usage is obsolescent.
2825 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2829 A comparison between signed and unsigned values could produce an
2830 incorrect result when the signed value is converted to unsigned.
2831 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2834 An aggregate has an initializer which does not initialize all members.
2835 This warning can be independently controlled by
2836 @option{-Wmissing-field-initializers}.
2839 A function parameter is declared without a type specifier in K&R-style
2847 An empty body occurs in an @samp{if} or @samp{else} statement.
2850 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2851 @samp{>}, or @samp{>=}.
2854 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2857 Any of several floating-point events that often indicate errors, such as
2858 overflow, underflow, loss of precision, etc.
2860 @item @r{(C++ only)}
2861 An enumerator and a non-enumerator both appear in a conditional expression.
2863 @item @r{(C++ only)}
2864 A non-static reference or non-static @samp{const} member appears in a
2865 class without constructors.
2867 @item @r{(C++ only)}
2868 Ambiguous virtual bases.
2870 @item @r{(C++ only)}
2871 Subscripting an array which has been declared @samp{register}.
2873 @item @r{(C++ only)}
2874 Taking the address of a variable which has been declared @samp{register}.
2876 @item @r{(C++ only)}
2877 A base class is not initialized in a derived class' copy constructor.
2880 @item -Wno-div-by-zero
2881 @opindex Wno-div-by-zero
2882 @opindex Wdiv-by-zero
2883 Do not warn about compile-time integer division by zero. Floating point
2884 division by zero is not warned about, as it can be a legitimate way of
2885 obtaining infinities and NaNs.
2887 @item -Wsystem-headers
2888 @opindex Wsystem-headers
2889 @cindex warnings from system headers
2890 @cindex system headers, warnings from
2891 Print warning messages for constructs found in system header files.
2892 Warnings from system headers are normally suppressed, on the assumption
2893 that they usually do not indicate real problems and would only make the
2894 compiler output harder to read. Using this command line option tells
2895 GCC to emit warnings from system headers as if they occurred in user
2896 code. However, note that using @option{-Wall} in conjunction with this
2897 option will @emph{not} warn about unknown pragmas in system
2898 headers---for that, @option{-Wunknown-pragmas} must also be used.
2901 @opindex Wfloat-equal
2902 Warn if floating point values are used in equality comparisons.
2904 The idea behind this is that sometimes it is convenient (for the
2905 programmer) to consider floating-point values as approximations to
2906 infinitely precise real numbers. If you are doing this, then you need
2907 to compute (by analyzing the code, or in some other way) the maximum or
2908 likely maximum error that the computation introduces, and allow for it
2909 when performing comparisons (and when producing output, but that's a
2910 different problem). In particular, instead of testing for equality, you
2911 would check to see whether the two values have ranges that overlap; and
2912 this is done with the relational operators, so equality comparisons are
2915 @item -Wtraditional @r{(C only)}
2916 @opindex Wtraditional
2917 Warn about certain constructs that behave differently in traditional and
2918 ISO C@. Also warn about ISO C constructs that have no traditional C
2919 equivalent, and/or problematic constructs which should be avoided.
2923 Macro parameters that appear within string literals in the macro body.
2924 In traditional C macro replacement takes place within string literals,
2925 but does not in ISO C@.
2928 In traditional C, some preprocessor directives did not exist.
2929 Traditional preprocessors would only consider a line to be a directive
2930 if the @samp{#} appeared in column 1 on the line. Therefore
2931 @option{-Wtraditional} warns about directives that traditional C
2932 understands but would ignore because the @samp{#} does not appear as the
2933 first character on the line. It also suggests you hide directives like
2934 @samp{#pragma} not understood by traditional C by indenting them. Some
2935 traditional implementations would not recognize @samp{#elif}, so it
2936 suggests avoiding it altogether.
2939 A function-like macro that appears without arguments.
2942 The unary plus operator.
2945 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2946 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2947 constants.) Note, these suffixes appear in macros defined in the system
2948 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2949 Use of these macros in user code might normally lead to spurious
2950 warnings, however GCC's integrated preprocessor has enough context to
2951 avoid warning in these cases.
2954 A function declared external in one block and then used after the end of
2958 A @code{switch} statement has an operand of type @code{long}.
2961 A non-@code{static} function declaration follows a @code{static} one.
2962 This construct is not accepted by some traditional C compilers.
2965 The ISO type of an integer constant has a different width or
2966 signedness from its traditional type. This warning is only issued if
2967 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2968 typically represent bit patterns, are not warned about.
2971 Usage of ISO string concatenation is detected.
2974 Initialization of automatic aggregates.
2977 Identifier conflicts with labels. Traditional C lacks a separate
2978 namespace for labels.
2981 Initialization of unions. If the initializer is zero, the warning is
2982 omitted. This is done under the assumption that the zero initializer in
2983 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2984 initializer warnings and relies on default initialization to zero in the
2988 Conversions by prototypes between fixed/floating point values and vice
2989 versa. The absence of these prototypes when compiling with traditional
2990 C would cause serious problems. This is a subset of the possible
2991 conversion warnings, for the full set use @option{-Wconversion}.
2994 Use of ISO C style function definitions. This warning intentionally is
2995 @emph{not} issued for prototype declarations or variadic functions
2996 because these ISO C features will appear in your code when using
2997 libiberty's traditional C compatibility macros, @code{PARAMS} and
2998 @code{VPARAMS}. This warning is also bypassed for nested functions
2999 because that feature is already a GCC extension and thus not relevant to
3000 traditional C compatibility.
3003 @item -Wdeclaration-after-statement @r{(C only)}
3004 @opindex Wdeclaration-after-statement
3005 Warn when a declaration is found after a statement in a block. This
3006 construct, known from C++, was introduced with ISO C99 and is by default
3007 allowed in GCC@. It is not supported by ISO C90 and was not supported by
3008 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
3012 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
3014 @item -Wno-endif-labels
3015 @opindex Wno-endif-labels
3016 @opindex Wendif-labels
3017 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
3021 Warn whenever a local variable shadows another local variable, parameter or
3022 global variable or whenever a built-in function is shadowed.
3024 @item -Wlarger-than-@var{len}
3025 @opindex Wlarger-than
3026 Warn whenever an object of larger than @var{len} bytes is defined.
3028 @item -Wunsafe-loop-optimizations
3029 @opindex Wunsafe-loop-optimizations
3030 Warn if the loop cannot be optimized because the compiler could not
3031 assume anything on the bounds of the loop indices. With
3032 @option{-funsafe-loop-optimizations} warn if the compiler made
3035 @item -Wpointer-arith
3036 @opindex Wpointer-arith
3037 Warn about anything that depends on the ``size of'' a function type or
3038 of @code{void}. GNU C assigns these types a size of 1, for
3039 convenience in calculations with @code{void *} pointers and pointers
3042 @item -Wbad-function-cast @r{(C only)}
3043 @opindex Wbad-function-cast
3044 Warn whenever a function call is cast to a non-matching type.
3045 For example, warn if @code{int malloc()} is cast to @code{anything *}.
3048 Warn about ISO C constructs that are outside of the common subset of
3049 ISO C and ISO C++, e.g.@: request for implicit conversion from
3050 @code{void *} to a pointer to non-@code{void} type.
3054 Warn whenever a pointer is cast so as to remove a type qualifier from
3055 the target type. For example, warn if a @code{const char *} is cast
3056 to an ordinary @code{char *}.
3059 @opindex Wcast-align
3060 Warn whenever a pointer is cast such that the required alignment of the
3061 target is increased. For example, warn if a @code{char *} is cast to
3062 an @code{int *} on machines where integers can only be accessed at
3063 two- or four-byte boundaries.
3065 @item -Wwrite-strings
3066 @opindex Wwrite-strings
3067 When compiling C, give string constants the type @code{const
3068 char[@var{length}]} so that
3069 copying the address of one into a non-@code{const} @code{char *}
3070 pointer will get a warning; when compiling C++, warn about the
3071 deprecated conversion from string constants to @code{char *}.
3072 These warnings will help you find at
3073 compile time code that can try to write into a string constant, but
3074 only if you have been very careful about using @code{const} in
3075 declarations and prototypes. Otherwise, it will just be a nuisance;
3076 this is why we did not make @option{-Wall} request these warnings.
3079 @opindex Wconversion
3080 Warn if a prototype causes a type conversion that is different from what
3081 would happen to the same argument in the absence of a prototype. This
3082 includes conversions of fixed point to floating and vice versa, and
3083 conversions changing the width or signedness of a fixed point argument
3084 except when the same as the default promotion.
3086 Also, warn if a negative integer constant expression is implicitly
3087 converted to an unsigned type. For example, warn about the assignment
3088 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
3089 casts like @code{(unsigned) -1}.
3091 @item -Wsign-compare
3092 @opindex Wsign-compare
3093 @cindex warning for comparison of signed and unsigned values
3094 @cindex comparison of signed and unsigned values, warning
3095 @cindex signed and unsigned values, comparison warning
3096 Warn when a comparison between signed and unsigned values could produce
3097 an incorrect result when the signed value is converted to unsigned.
3098 This warning is also enabled by @option{-Wextra}; to get the other warnings
3099 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
3101 @item -Waggregate-return
3102 @opindex Waggregate-return
3103 Warn if any functions that return structures or unions are defined or
3104 called. (In languages where you can return an array, this also elicits
3107 @item -Wno-attributes
3108 @opindex Wno-attributes
3109 @opindex Wattributes
3110 Do not warn if an unexpected @code{__attribute__} is used, such as
3111 unrecognized attributes, function attributes applied to variables,
3112 etc. This will not stop errors for incorrect use of supported
3115 @item -Wstrict-prototypes @r{(C only)}
3116 @opindex Wstrict-prototypes
3117 Warn if a function is declared or defined without specifying the
3118 argument types. (An old-style function definition is permitted without
3119 a warning if preceded by a declaration which specifies the argument
3122 @item -Wold-style-definition @r{(C only)}
3123 @opindex Wold-style-definition
3124 Warn if an old-style function definition is used. A warning is given
3125 even if there is a previous prototype.
3127 @item -Wmissing-prototypes @r{(C only)}
3128 @opindex Wmissing-prototypes
3129 Warn if a global function is defined without a previous prototype
3130 declaration. This warning is issued even if the definition itself
3131 provides a prototype. The aim is to detect global functions that fail
3132 to be declared in header files.
3134 @item -Wmissing-declarations @r{(C only)}
3135 @opindex Wmissing-declarations
3136 Warn if a global function is defined without a previous declaration.
3137 Do so even if the definition itself provides a prototype.
3138 Use this option to detect global functions that are not declared in
3141 @item -Wmissing-field-initializers
3142 @opindex Wmissing-field-initializers
3145 Warn if a structure's initializer has some fields missing. For
3146 example, the following code would cause such a warning, because
3147 @code{x.h} is implicitly zero:
3150 struct s @{ int f, g, h; @};
3151 struct s x = @{ 3, 4 @};
3154 This option does not warn about designated initializers, so the following
3155 modification would not trigger a warning:
3158 struct s @{ int f, g, h; @};
3159 struct s x = @{ .f = 3, .g = 4 @};
3162 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3163 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3165 @item -Wmissing-noreturn
3166 @opindex Wmissing-noreturn
3167 Warn about functions which might be candidates for attribute @code{noreturn}.
3168 Note these are only possible candidates, not absolute ones. Care should
3169 be taken to manually verify functions actually do not ever return before
3170 adding the @code{noreturn} attribute, otherwise subtle code generation
3171 bugs could be introduced. You will not get a warning for @code{main} in
3172 hosted C environments.
3174 @item -Wmissing-format-attribute
3175 @opindex Wmissing-format-attribute
3177 Warn about function pointers which might be candidates for @code{format}
3178 attributes. Note these are only possible candidates, not absolute ones.
3179 GCC will guess that function pointers with @code{format} attributes that
3180 are used in assignment, initialization, parameter passing or return
3181 statements should have a corresponding @code{format} attribute in the
3182 resulting type. I.e.@: the left-hand side of the assignment or
3183 initialization, the type of the parameter variable, or the return type
3184 of the containing function respectively should also have a @code{format}
3185 attribute to avoid the warning.
3187 GCC will also warn about function definitions which might be
3188 candidates for @code{format} attributes. Again, these are only
3189 possible candidates. GCC will guess that @code{format} attributes
3190 might be appropriate for any function that calls a function like
3191 @code{vprintf} or @code{vscanf}, but this might not always be the
3192 case, and some functions for which @code{format} attributes are
3193 appropriate may not be detected.
3195 @item -Wno-multichar
3196 @opindex Wno-multichar
3198 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3199 Usually they indicate a typo in the user's code, as they have
3200 implementation-defined values, and should not be used in portable code.
3202 @item -Wnormalized=<none|id|nfc|nfkc>
3203 @opindex Wnormalized
3206 @cindex character set, input normalization
3207 In ISO C and ISO C++, two identifiers are different if they are
3208 different sequences of characters. However, sometimes when characters
3209 outside the basic ASCII character set are used, you can have two
3210 different character sequences that look the same. To avoid confusion,
3211 the ISO 10646 standard sets out some @dfn{normalization rules} which
3212 when applied ensure that two sequences that look the same are turned into
3213 the same sequence. GCC can warn you if you are using identifiers which
3214 have not been normalized; this option controls that warning.
3216 There are four levels of warning that GCC supports. The default is
3217 @option{-Wnormalized=nfc}, which warns about any identifier which is
3218 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3219 recommended form for most uses.
3221 Unfortunately, there are some characters which ISO C and ISO C++ allow
3222 in identifiers that when turned into NFC aren't allowable as
3223 identifiers. That is, there's no way to use these symbols in portable
3224 ISO C or C++ and have all your identifiers in NFC.
3225 @option{-Wnormalized=id} suppresses the warning for these characters.
3226 It is hoped that future versions of the standards involved will correct
3227 this, which is why this option is not the default.
3229 You can switch the warning off for all characters by writing
3230 @option{-Wnormalized=none}. You would only want to do this if you
3231 were using some other normalization scheme (like ``D''), because
3232 otherwise you can easily create bugs that are literally impossible to see.
3234 Some characters in ISO 10646 have distinct meanings but look identical
3235 in some fonts or display methodologies, especially once formatting has
3236 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3237 LETTER N'', will display just like a regular @code{n} which has been
3238 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3239 normalisation scheme to convert all these into a standard form as
3240 well, and GCC will warn if your code is not in NFKC if you use
3241 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3242 about every identifier that contains the letter O because it might be
3243 confused with the digit 0, and so is not the default, but may be
3244 useful as a local coding convention if the programming environment is
3245 unable to be fixed to display these characters distinctly.
3247 @item -Wno-deprecated-declarations
3248 @opindex Wno-deprecated-declarations
3249 Do not warn about uses of functions, variables, and types marked as
3250 deprecated by using the @code{deprecated} attribute.
3251 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3252 @pxref{Type Attributes}.)
3256 Warn if a structure is given the packed attribute, but the packed
3257 attribute has no effect on the layout or size of the structure.
3258 Such structures may be mis-aligned for little benefit. For
3259 instance, in this code, the variable @code{f.x} in @code{struct bar}
3260 will be misaligned even though @code{struct bar} does not itself
3261 have the packed attribute:
3268 @} __attribute__((packed));
3278 Warn if padding is included in a structure, either to align an element
3279 of the structure or to align the whole structure. Sometimes when this
3280 happens it is possible to rearrange the fields of the structure to
3281 reduce the padding and so make the structure smaller.
3283 @item -Wredundant-decls
3284 @opindex Wredundant-decls
3285 Warn if anything is declared more than once in the same scope, even in
3286 cases where multiple declaration is valid and changes nothing.
3288 @item -Wnested-externs @r{(C only)}
3289 @opindex Wnested-externs
3290 Warn if an @code{extern} declaration is encountered within a function.
3292 @item -Wunreachable-code
3293 @opindex Wunreachable-code
3294 Warn if the compiler detects that code will never be executed.
3296 This option is intended to warn when the compiler detects that at
3297 least a whole line of source code will never be executed, because
3298 some condition is never satisfied or because it is after a
3299 procedure that never returns.
3301 It is possible for this option to produce a warning even though there
3302 are circumstances under which part of the affected line can be executed,
3303 so care should be taken when removing apparently-unreachable code.
3305 For instance, when a function is inlined, a warning may mean that the
3306 line is unreachable in only one inlined copy of the function.
3308 This option is not made part of @option{-Wall} because in a debugging
3309 version of a program there is often substantial code which checks
3310 correct functioning of the program and is, hopefully, unreachable
3311 because the program does work. Another common use of unreachable
3312 code is to provide behavior which is selectable at compile-time.
3316 Warn if a function can not be inlined and it was declared as inline.
3317 Even with this option, the compiler will not warn about failures to
3318 inline functions declared in system headers.
3320 The compiler uses a variety of heuristics to determine whether or not
3321 to inline a function. For example, the compiler takes into account
3322 the size of the function being inlined and the amount of inlining
3323 that has already been done in the current function. Therefore,
3324 seemingly insignificant changes in the source program can cause the
3325 warnings produced by @option{-Winline} to appear or disappear.
3327 @item -Wno-invalid-offsetof @r{(C++ only)}
3328 @opindex Wno-invalid-offsetof
3329 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3330 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3331 to a non-POD type is undefined. In existing C++ implementations,
3332 however, @samp{offsetof} typically gives meaningful results even when
3333 applied to certain kinds of non-POD types. (Such as a simple
3334 @samp{struct} that fails to be a POD type only by virtue of having a
3335 constructor.) This flag is for users who are aware that they are
3336 writing nonportable code and who have deliberately chosen to ignore the
3339 The restrictions on @samp{offsetof} may be relaxed in a future version
3340 of the C++ standard.
3342 @item -Wno-int-to-pointer-cast @r{(C only)}
3343 @opindex Wno-int-to-pointer-cast
3344 Suppress warnings from casts to pointer type of an integer of a
3347 @item -Wno-pointer-to-int-cast @r{(C only)}
3348 @opindex Wno-pointer-to-int-cast
3349 Suppress warnings from casts from a pointer to an integer type of a
3353 @opindex Winvalid-pch
3354 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3355 the search path but can't be used.
3359 @opindex Wno-long-long
3360 Warn if @samp{long long} type is used. This is default. To inhibit
3361 the warning messages, use @option{-Wno-long-long}. Flags
3362 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3363 only when @option{-pedantic} flag is used.
3365 @item -Wvariadic-macros
3366 @opindex Wvariadic-macros
3367 @opindex Wno-variadic-macros
3368 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3369 alternate syntax when in pedantic ISO C99 mode. This is default.
3370 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3372 @item -Wvolatile-register-var
3373 @opindex Wvolatile-register-var
3374 @opindex Wno-volatile-register-var
3375 Warn if a register variable is declared volatile. The volatile
3376 modifier does not inhibit all optimizations that may eliminate reads
3377 and/or writes to register variables.
3379 @item -Wdisabled-optimization
3380 @opindex Wdisabled-optimization
3381 Warn if a requested optimization pass is disabled. This warning does
3382 not generally indicate that there is anything wrong with your code; it
3383 merely indicates that GCC's optimizers were unable to handle the code
3384 effectively. Often, the problem is that your code is too big or too
3385 complex; GCC will refuse to optimize programs when the optimization
3386 itself is likely to take inordinate amounts of time.
3388 @item -Wpointer-sign
3389 @opindex Wpointer-sign
3390 @opindex Wno-pointer-sign
3391 Warn for pointer argument passing or assignment with different signedness.
3392 This option is only supported for C and Objective-C@. It is implied by
3393 @option{-Wall} and by @option{-pedantic}, which can be disabled with
3394 @option{-Wno-pointer-sign}.
3398 Make all warnings into errors.
3400 @item -Wstack-protector
3401 This option is only active when @option{-fstack-protector} is active. It
3402 warns about functions that will not be protected against stack smashing.
3406 @node Debugging Options
3407 @section Options for Debugging Your Program or GCC
3408 @cindex options, debugging
3409 @cindex debugging information options
3411 GCC has various special options that are used for debugging
3412 either your program or GCC:
3417 Produce debugging information in the operating system's native format
3418 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3421 On most systems that use stabs format, @option{-g} enables use of extra
3422 debugging information that only GDB can use; this extra information
3423 makes debugging work better in GDB but will probably make other debuggers
3425 refuse to read the program. If you want to control for certain whether
3426 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3427 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3429 GCC allows you to use @option{-g} with
3430 @option{-O}. The shortcuts taken by optimized code may occasionally
3431 produce surprising results: some variables you declared may not exist
3432 at all; flow of control may briefly move where you did not expect it;
3433 some statements may not be executed because they compute constant
3434 results or their values were already at hand; some statements may
3435 execute in different places because they were moved out of loops.
3437 Nevertheless it proves possible to debug optimized output. This makes
3438 it reasonable to use the optimizer for programs that might have bugs.
3440 The following options are useful when GCC is generated with the
3441 capability for more than one debugging format.
3445 Produce debugging information for use by GDB@. This means to use the
3446 most expressive format available (DWARF 2, stabs, or the native format
3447 if neither of those are supported), including GDB extensions if at all
3452 Produce debugging information in stabs format (if that is supported),
3453 without GDB extensions. This is the format used by DBX on most BSD
3454 systems. On MIPS, Alpha and System V Release 4 systems this option
3455 produces stabs debugging output which is not understood by DBX or SDB@.
3456 On System V Release 4 systems this option requires the GNU assembler.
3458 @item -feliminate-unused-debug-symbols
3459 @opindex feliminate-unused-debug-symbols
3460 Produce debugging information in stabs format (if that is supported),
3461 for only symbols that are actually used.
3465 Produce debugging information in stabs format (if that is supported),
3466 using GNU extensions understood only by the GNU debugger (GDB)@. The
3467 use of these extensions is likely to make other debuggers crash or
3468 refuse to read the program.
3472 Produce debugging information in COFF format (if that is supported).
3473 This is the format used by SDB on most System V systems prior to
3478 Produce debugging information in XCOFF format (if that is supported).
3479 This is the format used by the DBX debugger on IBM RS/6000 systems.
3483 Produce debugging information in XCOFF format (if that is supported),
3484 using GNU extensions understood only by the GNU debugger (GDB)@. The
3485 use of these extensions is likely to make other debuggers crash or
3486 refuse to read the program, and may cause assemblers other than the GNU
3487 assembler (GAS) to fail with an error.
3491 Produce debugging information in DWARF version 2 format (if that is
3492 supported). This is the format used by DBX on IRIX 6. With this
3493 option, GCC uses features of DWARF version 3 when they are useful;
3494 version 3 is upward compatible with version 2, but may still cause
3495 problems for older debuggers.
3499 Produce debugging information in VMS debug format (if that is
3500 supported). This is the format used by DEBUG on VMS systems.
3503 @itemx -ggdb@var{level}
3504 @itemx -gstabs@var{level}
3505 @itemx -gcoff@var{level}
3506 @itemx -gxcoff@var{level}
3507 @itemx -gvms@var{level}
3508 Request debugging information and also use @var{level} to specify how
3509 much information. The default level is 2.
3511 Level 1 produces minimal information, enough for making backtraces in
3512 parts of the program that you don't plan to debug. This includes
3513 descriptions of functions and external variables, but no information
3514 about local variables and no line numbers.
3516 Level 3 includes extra information, such as all the macro definitions
3517 present in the program. Some debuggers support macro expansion when
3518 you use @option{-g3}.
3520 @option{-gdwarf-2} does not accept a concatenated debug level, because
3521 GCC used to support an option @option{-gdwarf} that meant to generate
3522 debug information in version 1 of the DWARF format (which is very
3523 different from version 2), and it would have been too confusing. That
3524 debug format is long obsolete, but the option cannot be changed now.
3525 Instead use an additional @option{-g@var{level}} option to change the
3526 debug level for DWARF2.
3528 @item -feliminate-dwarf2-dups
3529 @opindex feliminate-dwarf2-dups
3530 Compress DWARF2 debugging information by eliminating duplicated
3531 information about each symbol. This option only makes sense when
3532 generating DWARF2 debugging information with @option{-gdwarf-2}.
3534 @cindex @command{prof}
3537 Generate extra code to write profile information suitable for the
3538 analysis program @command{prof}. You must use this option when compiling
3539 the source files you want data about, and you must also use it when
3542 @cindex @command{gprof}
3545 Generate extra code to write profile information suitable for the
3546 analysis program @command{gprof}. You must use this option when compiling
3547 the source files you want data about, and you must also use it when
3552 Makes the compiler print out each function name as it is compiled, and
3553 print some statistics about each pass when it finishes.
3556 @opindex ftime-report
3557 Makes the compiler print some statistics about the time consumed by each
3558 pass when it finishes.
3561 @opindex fmem-report
3562 Makes the compiler print some statistics about permanent memory
3563 allocation when it finishes.
3565 @item -fprofile-arcs
3566 @opindex fprofile-arcs
3567 Add code so that program flow @dfn{arcs} are instrumented. During
3568 execution the program records how many times each branch and call is
3569 executed and how many times it is taken or returns. When the compiled
3570 program exits it saves this data to a file called
3571 @file{@var{auxname}.gcda} for each source file. The data may be used for
3572 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3573 test coverage analysis (@option{-ftest-coverage}). Each object file's
3574 @var{auxname} is generated from the name of the output file, if
3575 explicitly specified and it is not the final executable, otherwise it is
3576 the basename of the source file. In both cases any suffix is removed
3577 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3578 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3579 @xref{Cross-profiling}.
3581 @cindex @command{gcov}
3585 This option is used to compile and link code instrumented for coverage
3586 analysis. The option is a synonym for @option{-fprofile-arcs}
3587 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3588 linking). See the documentation for those options for more details.
3593 Compile the source files with @option{-fprofile-arcs} plus optimization
3594 and code generation options. For test coverage analysis, use the
3595 additional @option{-ftest-coverage} option. You do not need to profile
3596 every source file in a program.
3599 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3600 (the latter implies the former).
3603 Run the program on a representative workload to generate the arc profile
3604 information. This may be repeated any number of times. You can run
3605 concurrent instances of your program, and provided that the file system
3606 supports locking, the data files will be correctly updated. Also
3607 @code{fork} calls are detected and correctly handled (double counting
3611 For profile-directed optimizations, compile the source files again with
3612 the same optimization and code generation options plus
3613 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3614 Control Optimization}).
3617 For test coverage analysis, use @command{gcov} to produce human readable
3618 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3619 @command{gcov} documentation for further information.
3623 With @option{-fprofile-arcs}, for each function of your program GCC
3624 creates a program flow graph, then finds a spanning tree for the graph.
3625 Only arcs that are not on the spanning tree have to be instrumented: the
3626 compiler adds code to count the number of times that these arcs are
3627 executed. When an arc is the only exit or only entrance to a block, the
3628 instrumentation code can be added to the block; otherwise, a new basic
3629 block must be created to hold the instrumentation code.
3632 @item -ftest-coverage
3633 @opindex ftest-coverage
3634 Produce a notes file that the @command{gcov} code-coverage utility
3635 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3636 show program coverage. Each source file's note file is called
3637 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3638 above for a description of @var{auxname} and instructions on how to
3639 generate test coverage data. Coverage data will match the source files
3640 more closely, if you do not optimize.
3642 @item -d@var{letters}
3643 @item -fdump-rtl-@var{pass}
3645 Says to make debugging dumps during compilation at times specified by
3646 @var{letters}. This is used for debugging the RTL-based passes of the
3647 compiler. The file names for most of the dumps are made by appending a
3648 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3649 from the name of the output file, if explicitly specified and it is not
3650 an executable, otherwise it is the basename of the source file.
3652 Most debug dumps can be enabled either passing a letter to the @option{-d}
3653 option, or with a long @option{-fdump-rtl} switch; here are the possible
3654 letters for use in @var{letters} and @var{pass}, and their meanings:
3659 Annotate the assembler output with miscellaneous debugging information.
3662 @itemx -fdump-rtl-bp
3664 @opindex fdump-rtl-bp
3665 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3668 @itemx -fdump-rtl-bbro
3670 @opindex fdump-rtl-bbro
3671 Dump after block reordering, to @file{@var{file}.30.bbro}.
3674 @itemx -fdump-rtl-combine
3676 @opindex fdump-rtl-combine
3677 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3680 @itemx -fdump-rtl-ce1
3681 @itemx -fdump-rtl-ce2
3683 @opindex fdump-rtl-ce1
3684 @opindex fdump-rtl-ce2
3685 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3686 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3687 and @option{-fdump-rtl-ce2} enable dumping after the second if
3688 conversion, to the file @file{@var{file}.18.ce2}.
3691 @itemx -fdump-rtl-btl
3692 @itemx -fdump-rtl-dbr
3694 @opindex fdump-rtl-btl
3695 @opindex fdump-rtl-dbr
3696 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3697 target load optimization, to @file{@var{file}.31.btl}. @option{-dd}
3698 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3699 scheduling, to @file{@var{file}.36.dbr}.
3703 Dump all macro definitions, at the end of preprocessing, in addition to
3707 @itemx -fdump-rtl-ce3
3709 @opindex fdump-rtl-ce3
3710 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3713 @itemx -fdump-rtl-cfg
3714 @itemx -fdump-rtl-life
3716 @opindex fdump-rtl-cfg
3717 @opindex fdump-rtl-life
3718 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3719 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3720 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3721 to @file{@var{file}.16.life}.
3724 @itemx -fdump-rtl-greg
3726 @opindex fdump-rtl-greg
3727 Dump after global register allocation, to @file{@var{file}.23.greg}.
3730 @itemx -fdump-rtl-gcse
3731 @itemx -fdump-rtl-bypass
3733 @opindex fdump-rtl-gcse
3734 @opindex fdump-rtl-bypass
3735 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3736 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3737 enable dumping after jump bypassing and control flow optimizations, to
3738 @file{@var{file}.07.bypass}.
3741 @itemx -fdump-rtl-eh
3743 @opindex fdump-rtl-eh
3744 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3747 @itemx -fdump-rtl-sibling
3749 @opindex fdump-rtl-sibling
3750 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3753 @itemx -fdump-rtl-jump
3755 @opindex fdump-rtl-jump
3756 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3759 @itemx -fdump-rtl-stack
3761 @opindex fdump-rtl-stack
3762 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3765 @itemx -fdump-rtl-lreg
3767 @opindex fdump-rtl-lreg
3768 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3771 @itemx -fdump-rtl-loop
3772 @itemx -fdump-rtl-loop2
3774 @opindex fdump-rtl-loop
3775 @opindex fdump-rtl-loop2
3776 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3777 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3778 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3779 @file{@var{file}.13.loop2}.
3782 @itemx -fdump-rtl-sms
3784 @opindex fdump-rtl-sms
3785 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3788 @itemx -fdump-rtl-mach
3790 @opindex fdump-rtl-mach
3791 Dump after performing the machine dependent reorganization pass, to
3792 @file{@var{file}.35.mach}.
3795 @itemx -fdump-rtl-rnreg
3797 @opindex fdump-rtl-rnreg
3798 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3801 @itemx -fdump-rtl-regmove
3803 @opindex fdump-rtl-regmove
3804 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3807 @itemx -fdump-rtl-postreload
3809 @opindex fdump-rtl-postreload
3810 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3813 @itemx -fdump-rtl-expand
3815 @opindex fdump-rtl-expand
3816 Dump after RTL generation, to @file{@var{file}.00.expand}.
3819 @itemx -fdump-rtl-sched2
3821 @opindex fdump-rtl-sched2
3822 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3825 @itemx -fdump-rtl-cse
3827 @opindex fdump-rtl-cse
3828 Dump after CSE (including the jump optimization that sometimes follows
3829 CSE), to @file{@var{file}.04.cse}.
3832 @itemx -fdump-rtl-sched
3834 @opindex fdump-rtl-sched
3835 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3838 @itemx -fdump-rtl-cse2
3840 @opindex fdump-rtl-cse2
3841 Dump after the second CSE pass (including the jump optimization that
3842 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3845 @itemx -fdump-rtl-tracer
3847 @opindex fdump-rtl-tracer
3848 Dump after running tracer, to @file{@var{file}.12.tracer}.
3851 @itemx -fdump-rtl-vpt
3852 @itemx -fdump-rtl-vartrack
3854 @opindex fdump-rtl-vpt
3855 @opindex fdump-rtl-vartrack
3856 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3857 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3858 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3859 to @file{@var{file}.34.vartrack}.
3862 @itemx -fdump-rtl-flow2
3864 @opindex fdump-rtl-flow2
3865 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3868 @itemx -fdump-rtl-peephole2
3870 @opindex fdump-rtl-peephole2
3871 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3874 @itemx -fdump-rtl-web
3876 @opindex fdump-rtl-web
3877 Dump after live range splitting, to @file{@var{file}.14.web}.
3880 @itemx -fdump-rtl-all
3882 @opindex fdump-rtl-all
3883 Produce all the dumps listed above.
3887 Produce a core dump whenever an error occurs.
3891 Print statistics on memory usage, at the end of the run, to
3896 Annotate the assembler output with a comment indicating which
3897 pattern and alternative was used. The length of each instruction is
3902 Dump the RTL in the assembler output as a comment before each instruction.
3903 Also turns on @option{-dp} annotation.
3907 For each of the other indicated dump files (either with @option{-d} or
3908 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3909 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3913 Just generate RTL for a function instead of compiling it. Usually used
3914 with @samp{r} (@option{-fdump-rtl-expand}).
3918 Dump debugging information during parsing, to standard error.
3921 @item -fdump-unnumbered
3922 @opindex fdump-unnumbered
3923 When doing debugging dumps (see @option{-d} option above), suppress instruction
3924 numbers and line number note output. This makes it more feasible to
3925 use diff on debugging dumps for compiler invocations with different
3926 options, in particular with and without @option{-g}.
3928 @item -fdump-translation-unit @r{(C++ only)}
3929 @itemx -fdump-translation-unit-@var{options} @r{(C++ only)}
3930 @opindex fdump-translation-unit
3931 Dump a representation of the tree structure for the entire translation
3932 unit to a file. The file name is made by appending @file{.tu} to the
3933 source file name. If the @samp{-@var{options}} form is used, @var{options}
3934 controls the details of the dump as described for the
3935 @option{-fdump-tree} options.
3937 @item -fdump-class-hierarchy @r{(C++ only)}
3938 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3939 @opindex fdump-class-hierarchy
3940 Dump a representation of each class's hierarchy and virtual function
3941 table layout to a file. The file name is made by appending @file{.class}
3942 to the source file name. If the @samp{-@var{options}} form is used,
3943 @var{options} controls the details of the dump as described for the
3944 @option{-fdump-tree} options.
3946 @item -fdump-ipa-@var{switch}
3948 Control the dumping at various stages of inter-procedural analysis
3949 language tree to a file. The file name is generated by appending a switch
3950 specific suffix to the source file name. The following dumps are possible:
3954 Enables all inter-procedural analysis dumps; currently the only produced
3955 dump is the @samp{cgraph} dump.
3958 Dumps information about call-graph optimization, unused function removal,
3959 and inlining decisions.
3962 @item -fdump-tree-@var{switch}
3963 @itemx -fdump-tree-@var{switch}-@var{options}
3965 Control the dumping at various stages of processing the intermediate
3966 language tree to a file. The file name is generated by appending a switch
3967 specific suffix to the source file name. If the @samp{-@var{options}}
3968 form is used, @var{options} is a list of @samp{-} separated options that
3969 control the details of the dump. Not all options are applicable to all
3970 dumps, those which are not meaningful will be ignored. The following
3971 options are available
3975 Print the address of each node. Usually this is not meaningful as it
3976 changes according to the environment and source file. Its primary use
3977 is for tying up a dump file with a debug environment.
3979 Inhibit dumping of members of a scope or body of a function merely
3980 because that scope has been reached. Only dump such items when they
3981 are directly reachable by some other path. When dumping pretty-printed
3982 trees, this option inhibits dumping the bodies of control structures.
3984 Print a raw representation of the tree. By default, trees are
3985 pretty-printed into a C-like representation.
3987 Enable more detailed dumps (not honored by every dump option).
3989 Enable dumping various statistics about the pass (not honored by every dump
3992 Enable showing basic block boundaries (disabled in raw dumps).
3994 Enable showing virtual operands for every statement.
3996 Enable showing line numbers for statements.
3998 Enable showing the unique ID (@code{DECL_UID}) for each variable.
4000 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
4003 The following tree dumps are possible:
4007 Dump before any tree based optimization, to @file{@var{file}.original}.
4010 Dump after all tree based optimization, to @file{@var{file}.optimized}.
4013 Dump after function inlining, to @file{@var{file}.inlined}.
4016 @opindex fdump-tree-gimple
4017 Dump each function before and after the gimplification pass to a file. The
4018 file name is made by appending @file{.gimple} to the source file name.
4021 @opindex fdump-tree-cfg
4022 Dump the control flow graph of each function to a file. The file name is
4023 made by appending @file{.cfg} to the source file name.
4026 @opindex fdump-tree-vcg
4027 Dump the control flow graph of each function to a file in VCG format. The
4028 file name is made by appending @file{.vcg} to the source file name. Note
4029 that if the file contains more than one function, the generated file cannot
4030 be used directly by VCG@. You will need to cut and paste each function's
4031 graph into its own separate file first.
4034 @opindex fdump-tree-ch
4035 Dump each function after copying loop headers. The file name is made by
4036 appending @file{.ch} to the source file name.
4039 @opindex fdump-tree-ssa
4040 Dump SSA related information to a file. The file name is made by appending
4041 @file{.ssa} to the source file name.
4044 @opindex fdump-tree-salias
4045 Dump structure aliasing variable information to a file. This file name
4046 is made by appending @file{.salias} to the source file name.
4049 @opindex fdump-tree-alias
4050 Dump aliasing information for each function. The file name is made by
4051 appending @file{.alias} to the source file name.
4054 @opindex fdump-tree-ccp
4055 Dump each function after CCP@. The file name is made by appending
4056 @file{.ccp} to the source file name.
4059 @opindex fdump-tree-storeccp
4060 Dump each function after STORE-CCP. The file name is made by appending
4061 @file{.storeccp} to the source file name.
4064 @opindex fdump-tree-pre
4065 Dump trees after partial redundancy elimination. The file name is made
4066 by appending @file{.pre} to the source file name.
4069 @opindex fdump-tree-fre
4070 Dump trees after full redundancy elimination. The file name is made
4071 by appending @file{.fre} to the source file name.
4074 @opindex fdump-tree-copyprop
4075 Dump trees after copy propagation. The file name is made
4076 by appending @file{.copyprop} to the source file name.
4078 @item store_copyprop
4079 @opindex fdump-tree-store_copyprop
4080 Dump trees after store copy-propagation. The file name is made
4081 by appending @file{.store_copyprop} to the source file name.
4084 @opindex fdump-tree-dce
4085 Dump each function after dead code elimination. The file name is made by
4086 appending @file{.dce} to the source file name.
4089 @opindex fdump-tree-mudflap
4090 Dump each function after adding mudflap instrumentation. The file name is
4091 made by appending @file{.mudflap} to the source file name.
4094 @opindex fdump-tree-sra
4095 Dump each function after performing scalar replacement of aggregates. The
4096 file name is made by appending @file{.sra} to the source file name.
4099 @opindex fdump-tree-sink
4100 Dump each function after performing code sinking. The file name is made
4101 by appending @file{.sink} to the source file name.
4104 @opindex fdump-tree-dom
4105 Dump each function after applying dominator tree optimizations. The file
4106 name is made by appending @file{.dom} to the source file name.
4109 @opindex fdump-tree-dse
4110 Dump each function after applying dead store elimination. The file
4111 name is made by appending @file{.dse} to the source file name.
4114 @opindex fdump-tree-phiopt
4115 Dump each function after optimizing PHI nodes into straightline code. The file
4116 name is made by appending @file{.phiopt} to the source file name.
4119 @opindex fdump-tree-forwprop
4120 Dump each function after forward propagating single use variables. The file
4121 name is made by appending @file{.forwprop} to the source file name.
4124 @opindex fdump-tree-copyrename
4125 Dump each function after applying the copy rename optimization. The file
4126 name is made by appending @file{.copyrename} to the source file name.
4129 @opindex fdump-tree-nrv
4130 Dump each function after applying the named return value optimization on
4131 generic trees. The file name is made by appending @file{.nrv} to the source
4135 @opindex fdump-tree-vect
4136 Dump each function after applying vectorization of loops. The file name is
4137 made by appending @file{.vect} to the source file name.
4140 @opindex fdump-tree-vrp
4141 Dump each function after Value Range Propagation (VRP). The file name
4142 is made by appending @file{.vrp} to the source file name.
4145 @opindex fdump-tree-all
4146 Enable all the available tree dumps with the flags provided in this option.
4149 @item -ftree-vectorizer-verbose=@var{n}
4150 @opindex ftree-vectorizer-verbose
4151 This option controls the amount of debugging output the vectorizer prints.
4152 This information is written to standard error, unless @option{-fdump-tree-all}
4153 or @option{-fdump-tree-vect} is specified, in which case it is output to the
4154 usual dump listing file, @file{.vect}.
4156 @item -frandom-seed=@var{string}
4157 @opindex frandom-string
4158 This option provides a seed that GCC uses when it would otherwise use
4159 random numbers. It is used to generate certain symbol names
4160 that have to be different in every compiled file. It is also used to
4161 place unique stamps in coverage data files and the object files that
4162 produce them. You can use the @option{-frandom-seed} option to produce
4163 reproducibly identical object files.
4165 The @var{string} should be different for every file you compile.
4167 @item -fsched-verbose=@var{n}
4168 @opindex fsched-verbose
4169 On targets that use instruction scheduling, this option controls the
4170 amount of debugging output the scheduler prints. This information is
4171 written to standard error, unless @option{-dS} or @option{-dR} is
4172 specified, in which case it is output to the usual dump
4173 listing file, @file{.sched} or @file{.sched2} respectively. However
4174 for @var{n} greater than nine, the output is always printed to standard
4177 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
4178 same information as @option{-dRS}. For @var{n} greater than one, it
4179 also output basic block probabilities, detailed ready list information
4180 and unit/insn info. For @var{n} greater than two, it includes RTL
4181 at abort point, control-flow and regions info. And for @var{n} over
4182 four, @option{-fsched-verbose} also includes dependence info.
4186 Store the usual ``temporary'' intermediate files permanently; place them
4187 in the current directory and name them based on the source file. Thus,
4188 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
4189 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
4190 preprocessed @file{foo.i} output file even though the compiler now
4191 normally uses an integrated preprocessor.
4193 When used in combination with the @option{-x} command line option,
4194 @option{-save-temps} is sensible enough to avoid over writing an
4195 input source file with the same extension as an intermediate file.
4196 The corresponding intermediate file may be obtained by renaming the
4197 source file before using @option{-save-temps}.
4201 Report the CPU time taken by each subprocess in the compilation
4202 sequence. For C source files, this is the compiler proper and assembler
4203 (plus the linker if linking is done). The output looks like this:
4210 The first number on each line is the ``user time'', that is time spent
4211 executing the program itself. The second number is ``system time'',
4212 time spent executing operating system routines on behalf of the program.
4213 Both numbers are in seconds.
4215 @item -fvar-tracking
4216 @opindex fvar-tracking
4217 Run variable tracking pass. It computes where variables are stored at each
4218 position in code. Better debugging information is then generated
4219 (if the debugging information format supports this information).
4221 It is enabled by default when compiling with optimization (@option{-Os},
4222 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4223 the debug info format supports it.
4225 @item -print-file-name=@var{library}
4226 @opindex print-file-name
4227 Print the full absolute name of the library file @var{library} that
4228 would be used when linking---and don't do anything else. With this
4229 option, GCC does not compile or link anything; it just prints the
4232 @item -print-multi-directory
4233 @opindex print-multi-directory
4234 Print the directory name corresponding to the multilib selected by any
4235 other switches present in the command line. This directory is supposed
4236 to exist in @env{GCC_EXEC_PREFIX}.
4238 @item -print-multi-lib
4239 @opindex print-multi-lib
4240 Print the mapping from multilib directory names to compiler switches
4241 that enable them. The directory name is separated from the switches by
4242 @samp{;}, and each switch starts with an @samp{@@} instead of the
4243 @samp{-}, without spaces between multiple switches. This is supposed to
4244 ease shell-processing.
4246 @item -print-prog-name=@var{program}
4247 @opindex print-prog-name
4248 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4250 @item -print-libgcc-file-name
4251 @opindex print-libgcc-file-name
4252 Same as @option{-print-file-name=libgcc.a}.
4254 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4255 but you do want to link with @file{libgcc.a}. You can do
4258 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4261 @item -print-search-dirs
4262 @opindex print-search-dirs
4263 Print the name of the configured installation directory and a list of
4264 program and library directories @command{gcc} will search---and don't do anything else.
4266 This is useful when @command{gcc} prints the error message
4267 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4268 To resolve this you either need to put @file{cpp0} and the other compiler
4269 components where @command{gcc} expects to find them, or you can set the environment
4270 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4271 Don't forget the trailing @samp{/}.
4272 @xref{Environment Variables}.
4275 @opindex dumpmachine
4276 Print the compiler's target machine (for example,
4277 @samp{i686-pc-linux-gnu})---and don't do anything else.
4280 @opindex dumpversion
4281 Print the compiler version (for example, @samp{3.0})---and don't do
4286 Print the compiler's built-in specs---and don't do anything else. (This
4287 is used when GCC itself is being built.) @xref{Spec Files}.
4289 @item -feliminate-unused-debug-types
4290 @opindex feliminate-unused-debug-types
4291 Normally, when producing DWARF2 output, GCC will emit debugging
4292 information for all types declared in a compilation
4293 unit, regardless of whether or not they are actually used
4294 in that compilation unit. Sometimes this is useful, such as
4295 if, in the debugger, you want to cast a value to a type that is
4296 not actually used in your program (but is declared). More often,
4297 however, this results in a significant amount of wasted space.
4298 With this option, GCC will avoid producing debug symbol output
4299 for types that are nowhere used in the source file being compiled.
4302 @node Optimize Options
4303 @section Options That Control Optimization
4304 @cindex optimize options
4305 @cindex options, optimization
4307 These options control various sorts of optimizations.
4309 Without any optimization option, the compiler's goal is to reduce the
4310 cost of compilation and to make debugging produce the expected
4311 results. Statements are independent: if you stop the program with a
4312 breakpoint between statements, you can then assign a new value to any
4313 variable or change the program counter to any other statement in the
4314 function and get exactly the results you would expect from the source
4317 Turning on optimization flags makes the compiler attempt to improve
4318 the performance and/or code size at the expense of compilation time
4319 and possibly the ability to debug the program.
4321 The compiler performs optimization based on the knowledge it has of
4322 the program. Optimization levels @option{-O2} and above, in
4323 particular, enable @emph{unit-at-a-time} mode, which allows the
4324 compiler to consider information gained from later functions in
4325 the file when compiling a function. Compiling multiple files at
4326 once to a single output file in @emph{unit-at-a-time} mode allows
4327 the compiler to use information gained from all of the files when
4328 compiling each of them.
4330 Not all optimizations are controlled directly by a flag. Only
4331 optimizations that have a flag are listed.
4338 Optimize. Optimizing compilation takes somewhat more time, and a lot
4339 more memory for a large function.
4341 With @option{-O}, the compiler tries to reduce code size and execution
4342 time, without performing any optimizations that take a great deal of
4345 @option{-O} turns on the following optimization flags:
4346 @gccoptlist{-fdefer-pop @gol
4347 -fdelayed-branch @gol
4348 -fguess-branch-probability @gol
4349 -fcprop-registers @gol
4350 -floop-optimize @gol
4351 -fif-conversion @gol
4352 -fif-conversion2 @gol
4355 -ftree-dominator-opts @gol
4360 -ftree-copyrename @gol
4365 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4366 where doing so does not interfere with debugging.
4368 @option{-O} doesn't turn on @option{-ftree-sra} for the Ada compiler.
4369 This option must be explicitly specified on the command line to be
4370 enabled for the Ada compiler.
4374 Optimize even more. GCC performs nearly all supported optimizations
4375 that do not involve a space-speed tradeoff. The compiler does not
4376 perform loop unrolling or function inlining when you specify @option{-O2}.
4377 As compared to @option{-O}, this option increases both compilation time
4378 and the performance of the generated code.
4380 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4381 also turns on the following optimization flags:
4382 @gccoptlist{-fthread-jumps @gol
4384 -foptimize-sibling-calls @gol
4385 -fcse-follow-jumps -fcse-skip-blocks @gol
4386 -fgcse -fgcse-lm @gol
4387 -fexpensive-optimizations @gol
4388 -fstrength-reduce @gol
4389 -frerun-cse-after-loop -frerun-loop-opt @gol
4392 -fschedule-insns -fschedule-insns2 @gol
4393 -fsched-interblock -fsched-spec @gol
4395 -fstrict-aliasing @gol
4396 -fdelete-null-pointer-checks @gol
4397 -freorder-blocks -freorder-functions @gol
4398 -funit-at-a-time @gol
4399 -falign-functions -falign-jumps @gol
4400 -falign-loops -falign-labels @gol
4404 Please note the warning under @option{-fgcse} about
4405 invoking @option{-O2} on programs that use computed gotos.
4409 Optimize yet more. @option{-O3} turns on all optimizations specified by
4410 @option{-O2} and also turns on the @option{-finline-functions},
4411 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4415 Do not optimize. This is the default.
4419 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4420 do not typically increase code size. It also performs further
4421 optimizations designed to reduce code size.
4423 @option{-Os} disables the following optimization flags:
4424 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4425 -falign-labels -freorder-blocks -freorder-blocks-and-partition @gol
4426 -fprefetch-loop-arrays -ftree-vect-loop-version}
4428 If you use multiple @option{-O} options, with or without level numbers,
4429 the last such option is the one that is effective.
4432 Options of the form @option{-f@var{flag}} specify machine-independent
4433 flags. Most flags have both positive and negative forms; the negative
4434 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4435 below, only one of the forms is listed---the one you typically will
4436 use. You can figure out the other form by either removing @samp{no-}
4439 The following options control specific optimizations. They are either
4440 activated by @option{-O} options or are related to ones that are. You
4441 can use the following flags in the rare cases when ``fine-tuning'' of
4442 optimizations to be performed is desired.
4445 @item -fno-default-inline
4446 @opindex fno-default-inline
4447 Do not make member functions inline by default merely because they are
4448 defined inside the class scope (C++ only). Otherwise, when you specify
4449 @w{@option{-O}}, member functions defined inside class scope are compiled
4450 inline by default; i.e., you don't need to add @samp{inline} in front of
4451 the member function name.
4453 @item -fno-defer-pop
4454 @opindex fno-defer-pop
4455 Always pop the arguments to each function call as soon as that function
4456 returns. For machines which must pop arguments after a function call,
4457 the compiler normally lets arguments accumulate on the stack for several
4458 function calls and pops them all at once.
4460 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4464 Force memory operands to be copied into registers before doing
4465 arithmetic on them. This produces better code by making all memory
4466 references potential common subexpressions. When they are not common
4467 subexpressions, instruction combination should eliminate the separate
4468 register-load. This option is now a nop and will be removed in 4.2.
4471 @opindex fforce-addr
4472 Force memory address constants to be copied into registers before
4473 doing arithmetic on them.
4475 @item -fomit-frame-pointer
4476 @opindex fomit-frame-pointer
4477 Don't keep the frame pointer in a register for functions that
4478 don't need one. This avoids the instructions to save, set up and
4479 restore frame pointers; it also makes an extra register available
4480 in many functions. @strong{It also makes debugging impossible on
4483 On some machines, such as the VAX, this flag has no effect, because
4484 the standard calling sequence automatically handles the frame pointer
4485 and nothing is saved by pretending it doesn't exist. The
4486 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4487 whether a target machine supports this flag. @xref{Registers,,Register
4488 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4490 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4492 @item -foptimize-sibling-calls
4493 @opindex foptimize-sibling-calls
4494 Optimize sibling and tail recursive calls.
4496 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4500 Don't pay attention to the @code{inline} keyword. Normally this option
4501 is used to keep the compiler from expanding any functions inline.
4502 Note that if you are not optimizing, no functions can be expanded inline.
4504 @item -finline-functions
4505 @opindex finline-functions
4506 Integrate all simple functions into their callers. The compiler
4507 heuristically decides which functions are simple enough to be worth
4508 integrating in this way.
4510 If all calls to a given function are integrated, and the function is
4511 declared @code{static}, then the function is normally not output as
4512 assembler code in its own right.
4514 Enabled at level @option{-O3}.
4516 @item -finline-functions-called-once
4517 @opindex finline-functions-called-once
4518 Consider all @code{static} functions called once for inlining into their
4519 caller even if they are not marked @code{inline}. If a call to a given
4520 function is integrated, then the function is not output as assembler code
4523 Enabled if @option{-funit-at-a-time} is enabled.
4525 @item -fearly-inlining
4526 @opindex fearly-inlining
4527 Inline functions marked by @code{always_inline} and functions whose body seems
4528 smaller than the function call overhead early before doing
4529 @option{-fprofile-generate} instrumentation and real inlining pass. Doing so
4530 makes profiling significantly cheaper and usually inlining faster on programs
4531 having large chains of nested wrapper functions.
4535 @item -finline-limit=@var{n}
4536 @opindex finline-limit
4537 By default, GCC limits the size of functions that can be inlined. This flag
4538 allows the control of this limit for functions that are explicitly marked as
4539 inline (i.e., marked with the inline keyword or defined within the class
4540 definition in c++). @var{n} is the size of functions that can be inlined in
4541 number of pseudo instructions (not counting parameter handling). The default
4542 value of @var{n} is 600.
4543 Increasing this value can result in more inlined code at
4544 the cost of compilation time and memory consumption. Decreasing usually makes
4545 the compilation faster and less code will be inlined (which presumably
4546 means slower programs). This option is particularly useful for programs that
4547 use inlining heavily such as those based on recursive templates with C++.
4549 Inlining is actually controlled by a number of parameters, which may be
4550 specified individually by using @option{--param @var{name}=@var{value}}.
4551 The @option{-finline-limit=@var{n}} option sets some of these parameters
4555 @item max-inline-insns-single
4556 is set to @var{n}/2.
4557 @item max-inline-insns-auto
4558 is set to @var{n}/2.
4559 @item min-inline-insns
4560 is set to 130 or @var{n}/4, whichever is smaller.
4561 @item max-inline-insns-rtl
4565 See below for a documentation of the individual
4566 parameters controlling inlining.
4568 @emph{Note:} pseudo instruction represents, in this particular context, an
4569 abstract measurement of function's size. In no way does it represent a count
4570 of assembly instructions and as such its exact meaning might change from one
4571 release to an another.
4573 @item -fkeep-inline-functions
4574 @opindex fkeep-inline-functions
4575 In C, emit @code{static} functions that are declared @code{inline}
4576 into the object file, even if the function has been inlined into all
4577 of its callers. This switch does not affect functions using the
4578 @code{extern inline} extension in GNU C@. In C++, emit any and all
4579 inline functions into the object file.
4581 @item -fkeep-static-consts
4582 @opindex fkeep-static-consts
4583 Emit variables declared @code{static const} when optimization isn't turned
4584 on, even if the variables aren't referenced.
4586 GCC enables this option by default. If you want to force the compiler to
4587 check if the variable was referenced, regardless of whether or not
4588 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4590 @item -fmerge-constants
4591 Attempt to merge identical constants (string constants and floating point
4592 constants) across compilation units.
4594 This option is the default for optimized compilation if the assembler and
4595 linker support it. Use @option{-fno-merge-constants} to inhibit this
4598 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4600 @item -fmerge-all-constants
4601 Attempt to merge identical constants and identical variables.
4603 This option implies @option{-fmerge-constants}. In addition to
4604 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4605 arrays or initialized constant variables with integral or floating point
4606 types. Languages like C or C++ require each non-automatic variable to
4607 have distinct location, so using this option will result in non-conforming
4610 @item -fmodulo-sched
4611 @opindex fmodulo-sched
4612 Perform swing modulo scheduling immediately before the first scheduling
4613 pass. This pass looks at innermost loops and reorders their
4614 instructions by overlapping different iterations.
4616 @item -fno-branch-count-reg
4617 @opindex fno-branch-count-reg
4618 Do not use ``decrement and branch'' instructions on a count register,
4619 but instead generate a sequence of instructions that decrement a
4620 register, compare it against zero, then branch based upon the result.
4621 This option is only meaningful on architectures that support such
4622 instructions, which include x86, PowerPC, IA-64 and S/390.
4624 The default is @option{-fbranch-count-reg}, enabled when
4625 @option{-fstrength-reduce} is enabled.
4627 @item -fno-function-cse
4628 @opindex fno-function-cse
4629 Do not put function addresses in registers; make each instruction that
4630 calls a constant function contain the function's address explicitly.
4632 This option results in less efficient code, but some strange hacks
4633 that alter the assembler output may be confused by the optimizations
4634 performed when this option is not used.
4636 The default is @option{-ffunction-cse}
4638 @item -fno-zero-initialized-in-bss
4639 @opindex fno-zero-initialized-in-bss
4640 If the target supports a BSS section, GCC by default puts variables that
4641 are initialized to zero into BSS@. This can save space in the resulting
4644 This option turns off this behavior because some programs explicitly
4645 rely on variables going to the data section. E.g., so that the
4646 resulting executable can find the beginning of that section and/or make
4647 assumptions based on that.
4649 The default is @option{-fzero-initialized-in-bss}.
4651 @item -fbounds-check
4652 @opindex fbounds-check
4653 For front-ends that support it, generate additional code to check that
4654 indices used to access arrays are within the declared range. This is
4655 currently only supported by the Java and Fortran front-ends, where
4656 this option defaults to true and false respectively.
4658 @item -fmudflap -fmudflapth -fmudflapir
4662 @cindex bounds checking
4664 For front-ends that support it (C and C++), instrument all risky
4665 pointer/array dereferencing operations, some standard library
4666 string/heap functions, and some other associated constructs with
4667 range/validity tests. Modules so instrumented should be immune to
4668 buffer overflows, invalid heap use, and some other classes of C/C++
4669 programming errors. The instrumentation relies on a separate runtime
4670 library (@file{libmudflap}), which will be linked into a program if
4671 @option{-fmudflap} is given at link time. Run-time behavior of the
4672 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4673 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4676 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4677 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4678 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4679 instrumentation should ignore pointer reads. This produces less
4680 instrumentation (and therefore faster execution) and still provides
4681 some protection against outright memory corrupting writes, but allows
4682 erroneously read data to propagate within a program.
4684 @item -fstrength-reduce
4685 @opindex fstrength-reduce
4686 Perform the optimizations of loop strength reduction and
4687 elimination of iteration variables.
4689 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4691 @item -fthread-jumps
4692 @opindex fthread-jumps
4693 Perform optimizations where we check to see if a jump branches to a
4694 location where another comparison subsumed by the first is found. If
4695 so, the first branch is redirected to either the destination of the
4696 second branch or a point immediately following it, depending on whether
4697 the condition is known to be true or false.
4699 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4701 @item -fcse-follow-jumps
4702 @opindex fcse-follow-jumps
4703 In common subexpression elimination, scan through jump instructions
4704 when the target of the jump is not reached by any other path. For
4705 example, when CSE encounters an @code{if} statement with an
4706 @code{else} clause, CSE will follow the jump when the condition
4709 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4711 @item -fcse-skip-blocks
4712 @opindex fcse-skip-blocks
4713 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4714 follow jumps which conditionally skip over blocks. When CSE
4715 encounters a simple @code{if} statement with no else clause,
4716 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4717 body of the @code{if}.
4719 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4721 @item -frerun-cse-after-loop
4722 @opindex frerun-cse-after-loop
4723 Re-run common subexpression elimination after loop optimizations has been
4726 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4728 @item -frerun-loop-opt
4729 @opindex frerun-loop-opt
4730 Run the loop optimizer twice.
4732 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4736 Perform a global common subexpression elimination pass.
4737 This pass also performs global constant and copy propagation.
4739 @emph{Note:} When compiling a program using computed gotos, a GCC
4740 extension, you may get better runtime performance if you disable
4741 the global common subexpression elimination pass by adding
4742 @option{-fno-gcse} to the command line.
4744 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4748 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4749 attempt to move loads which are only killed by stores into themselves. This
4750 allows a loop containing a load/store sequence to be changed to a load outside
4751 the loop, and a copy/store within the loop.
4753 Enabled by default when gcse is enabled.
4757 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4758 global common subexpression elimination. This pass will attempt to move
4759 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4760 loops containing a load/store sequence can be changed to a load before
4761 the loop and a store after the loop.
4763 Not enabled at any optimization level.
4767 When @option{-fgcse-las} is enabled, the global common subexpression
4768 elimination pass eliminates redundant loads that come after stores to the
4769 same memory location (both partial and full redundancies).
4771 Not enabled at any optimization level.
4773 @item -fgcse-after-reload
4774 @opindex fgcse-after-reload
4775 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4776 pass is performed after reload. The purpose of this pass is to cleanup
4779 @item -floop-optimize
4780 @opindex floop-optimize
4781 Perform loop optimizations: move constant expressions out of loops, simplify
4782 exit test conditions and optionally do strength-reduction as well.
4784 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4786 @item -floop-optimize2
4787 @opindex floop-optimize2
4788 Perform loop optimizations using the new loop optimizer. The optimizations
4789 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4792 @item -funsafe-loop-optimizations
4793 @opindex funsafe-loop-optimizations
4794 If given, the loop optimizer will assume that loop indices do not
4795 overflow, and that the loops with nontrivial exit condition are not
4796 infinite. This enables a wider range of loop optimizations even if
4797 the loop optimizer itself cannot prove that these assumptions are valid.
4798 Using @option{-Wunsafe-loop-optimizations}, the compiler will warn you
4799 if it finds this kind of loop.
4801 @item -fcrossjumping
4802 @opindex crossjumping
4803 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4804 resulting code may or may not perform better than without cross-jumping.
4806 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4808 @item -fif-conversion
4809 @opindex if-conversion
4810 Attempt to transform conditional jumps into branch-less equivalents. This
4811 include use of conditional moves, min, max, set flags and abs instructions, and
4812 some tricks doable by standard arithmetics. The use of conditional execution
4813 on chips where it is available is controlled by @code{if-conversion2}.
4815 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4817 @item -fif-conversion2
4818 @opindex if-conversion2
4819 Use conditional execution (where available) to transform conditional jumps into
4820 branch-less equivalents.
4822 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4824 @item -fdelete-null-pointer-checks
4825 @opindex fdelete-null-pointer-checks
4826 Use global dataflow analysis to identify and eliminate useless checks
4827 for null pointers. The compiler assumes that dereferencing a null
4828 pointer would have halted the program. If a pointer is checked after
4829 it has already been dereferenced, it cannot be null.
4831 In some environments, this assumption is not true, and programs can
4832 safely dereference null pointers. Use
4833 @option{-fno-delete-null-pointer-checks} to disable this optimization
4834 for programs which depend on that behavior.
4836 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4838 @item -fexpensive-optimizations
4839 @opindex fexpensive-optimizations
4840 Perform a number of minor optimizations that are relatively expensive.
4842 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4844 @item -foptimize-register-move
4846 @opindex foptimize-register-move
4848 Attempt to reassign register numbers in move instructions and as
4849 operands of other simple instructions in order to maximize the amount of
4850 register tying. This is especially helpful on machines with two-operand
4853 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4856 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4858 @item -fdelayed-branch
4859 @opindex fdelayed-branch
4860 If supported for the target machine, attempt to reorder instructions
4861 to exploit instruction slots available after delayed branch
4864 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4866 @item -fschedule-insns
4867 @opindex fschedule-insns
4868 If supported for the target machine, attempt to reorder instructions to
4869 eliminate execution stalls due to required data being unavailable. This
4870 helps machines that have slow floating point or memory load instructions
4871 by allowing other instructions to be issued until the result of the load
4872 or floating point instruction is required.
4874 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4876 @item -fschedule-insns2
4877 @opindex fschedule-insns2
4878 Similar to @option{-fschedule-insns}, but requests an additional pass of
4879 instruction scheduling after register allocation has been done. This is
4880 especially useful on machines with a relatively small number of
4881 registers and where memory load instructions take more than one cycle.
4883 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4885 @item -fno-sched-interblock
4886 @opindex fno-sched-interblock
4887 Don't schedule instructions across basic blocks. This is normally
4888 enabled by default when scheduling before register allocation, i.e.@:
4889 with @option{-fschedule-insns} or at @option{-O2} or higher.
4891 @item -fno-sched-spec
4892 @opindex fno-sched-spec
4893 Don't allow speculative motion of non-load instructions. This is normally
4894 enabled by default when scheduling before register allocation, i.e.@:
4895 with @option{-fschedule-insns} or at @option{-O2} or higher.
4897 @item -fsched-spec-load
4898 @opindex fsched-spec-load
4899 Allow speculative motion of some load instructions. This only makes
4900 sense when scheduling before register allocation, i.e.@: with
4901 @option{-fschedule-insns} or at @option{-O2} or higher.
4903 @item -fsched-spec-load-dangerous
4904 @opindex fsched-spec-load-dangerous
4905 Allow speculative motion of more load instructions. This only makes
4906 sense when scheduling before register allocation, i.e.@: with
4907 @option{-fschedule-insns} or at @option{-O2} or higher.
4909 @item -fsched-stalled-insns=@var{n}
4910 @opindex fsched-stalled-insns
4911 Define how many insns (if any) can be moved prematurely from the queue
4912 of stalled insns into the ready list, during the second scheduling pass.
4914 @item -fsched-stalled-insns-dep=@var{n}
4915 @opindex fsched-stalled-insns-dep
4916 Define how many insn groups (cycles) will be examined for a dependency
4917 on a stalled insn that is candidate for premature removal from the queue
4918 of stalled insns. Has an effect only during the second scheduling pass,
4919 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4921 @item -fsched2-use-superblocks
4922 @opindex fsched2-use-superblocks
4923 When scheduling after register allocation, do use superblock scheduling
4924 algorithm. Superblock scheduling allows motion across basic block boundaries
4925 resulting on faster schedules. This option is experimental, as not all machine
4926 descriptions used by GCC model the CPU closely enough to avoid unreliable
4927 results from the algorithm.
4929 This only makes sense when scheduling after register allocation, i.e.@: with
4930 @option{-fschedule-insns2} or at @option{-O2} or higher.
4932 @item -fsched2-use-traces
4933 @opindex fsched2-use-traces
4934 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4935 allocation and additionally perform code duplication in order to increase the
4936 size of superblocks using tracer pass. See @option{-ftracer} for details on
4939 This mode should produce faster but significantly longer programs. Also
4940 without @option{-fbranch-probabilities} the traces constructed may not
4941 match the reality and hurt the performance. This only makes
4942 sense when scheduling after register allocation, i.e.@: with
4943 @option{-fschedule-insns2} or at @option{-O2} or higher.
4945 @item -freschedule-modulo-scheduled-loops
4946 @opindex fscheduling-in-modulo-scheduled-loops
4947 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4948 we may want to prevent the later scheduling passes from changing its schedule, we use this
4949 option to control that.
4951 @item -fcaller-saves
4952 @opindex fcaller-saves
4953 Enable values to be allocated in registers that will be clobbered by
4954 function calls, by emitting extra instructions to save and restore the
4955 registers around such calls. Such allocation is done only when it
4956 seems to result in better code than would otherwise be produced.
4958 This option is always enabled by default on certain machines, usually
4959 those which have no call-preserved registers to use instead.
4961 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4964 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4965 enabled by default at @option{-O2} and @option{-O3}.
4968 Perform Full Redundancy Elimination (FRE) on trees. The difference
4969 between FRE and PRE is that FRE only considers expressions
4970 that are computed on all paths leading to the redundant computation.
4971 This analysis faster than PRE, though it exposes fewer redundancies.
4972 This flag is enabled by default at @option{-O} and higher.
4974 @item -ftree-copy-prop
4975 Perform copy propagation on trees. This pass eliminates unnecessary
4976 copy operations. This flag is enabled by default at @option{-O} and
4979 @item -ftree-store-copy-prop
4980 Perform copy propagation of memory loads and stores. This pass
4981 eliminates unnecessary copy operations in memory references
4982 (structures, global variables, arrays, etc). This flag is enabled by
4983 default at @option{-O2} and higher.
4986 Perform structural alias analysis on trees. This flag
4987 is enabled by default at @option{-O} and higher.
4990 Perform forward store motion on trees. This flag is
4991 enabled by default at @option{-O} and higher.
4994 Perform sparse conditional constant propagation (CCP) on trees. This
4995 pass only operates on local scalar variables and is enabled by default
4996 at @option{-O} and higher.
4998 @item -ftree-store-ccp
4999 Perform sparse conditional constant propagation (CCP) on trees. This
5000 pass operates on both local scalar variables and memory stores and
5001 loads (global variables, structures, arrays, etc). This flag is
5002 enabled by default at @option{-O2} and higher.
5005 Perform dead code elimination (DCE) on trees. This flag is enabled by
5006 default at @option{-O} and higher.
5008 @item -ftree-dominator-opts
5009 Perform a variety of simple scalar cleanups (constant/copy
5010 propagation, redundancy elimination, range propagation and expression
5011 simplification) based on a dominator tree traversal. This also
5012 performs jump threading (to reduce jumps to jumps). This flag is
5013 enabled by default at @option{-O} and higher.
5016 Perform loop header copying on trees. This is beneficial since it increases
5017 effectiveness of code motion optimizations. It also saves one jump. This flag
5018 is enabled by default at @option{-O} and higher. It is not enabled
5019 for @option{-Os}, since it usually increases code size.
5021 @item -ftree-loop-optimize
5022 Perform loop optimizations on trees. This flag is enabled by default
5023 at @option{-O} and higher.
5025 @item -ftree-loop-linear
5026 Perform linear loop transformations on tree. This flag can improve cache
5027 performance and allow further loop optimizations to take place.
5029 @item -ftree-loop-im
5030 Perform loop invariant motion on trees. This pass moves only invariants that
5031 would be hard to handle at RTL level (function calls, operations that expand to
5032 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
5033 operands of conditions that are invariant out of the loop, so that we can use
5034 just trivial invariantness analysis in loop unswitching. The pass also includes
5037 @item -ftree-loop-ivcanon
5038 Create a canonical counter for number of iterations in the loop for that
5039 determining number of iterations requires complicated analysis. Later
5040 optimizations then may determine the number easily. Useful especially
5041 in connection with unrolling.
5044 Perform induction variable optimizations (strength reduction, induction
5045 variable merging and induction variable elimination) on trees.
5048 Perform scalar replacement of aggregates. This pass replaces structure
5049 references with scalars to prevent committing structures to memory too
5050 early. This flag is enabled by default at @option{-O} and higher.
5052 @item -ftree-copyrename
5053 Perform copy renaming on trees. This pass attempts to rename compiler
5054 temporaries to other variables at copy locations, usually resulting in
5055 variable names which more closely resemble the original variables. This flag
5056 is enabled by default at @option{-O} and higher.
5059 Perform temporary expression replacement during the SSA->normal phase. Single
5060 use/single def temporaries are replaced at their use location with their
5061 defining expression. This results in non-GIMPLE code, but gives the expanders
5062 much more complex trees to work on resulting in better RTL generation. This is
5063 enabled by default at @option{-O} and higher.
5066 Perform live range splitting during the SSA->normal phase. Distinct live
5067 ranges of a variable are split into unique variables, allowing for better
5068 optimization later. This is enabled by default at @option{-O} and higher.
5070 @item -ftree-vectorize
5071 Perform loop vectorization on trees.
5073 @item -ftree-vect-loop-version
5074 @opindex ftree-vect-loop-version
5075 Perform loop versioning when doing loop vectorization on trees. When a loop
5076 appears to be vectorizable except that data alignment or data dependence cannot
5077 be determined at compile time then vectorized and non-vectorized versions of
5078 the loop are generated along with runtime checks for alignment or dependence
5079 to control which version is executed. This option is enabled by default
5080 except at level @option{-Os} where it is disabled.
5083 Perform Value Range Propagation on trees. This is similar to the
5084 constant propagation pass, but instead of values, ranges of values are
5085 propagated. This allows the optimizers to remove unnecessary range
5086 checks like array bound checks and null pointer checks. This is
5087 enabled by default at @option{-O2} and higher. Null pointer check
5088 elimination is only done if @option{-fdelete-null-pointer-checks} is
5093 Perform tail duplication to enlarge superblock size. This transformation
5094 simplifies the control flow of the function allowing other optimizations to do
5097 @item -funroll-loops
5098 @opindex funroll-loops
5099 Unroll loops whose number of iterations can be determined at compile
5100 time or upon entry to the loop. @option{-funroll-loops} implies both
5101 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
5102 option makes code larger, and may or may not make it run faster.
5104 @item -funroll-all-loops
5105 @opindex funroll-all-loops
5106 Unroll all loops, even if their number of iterations is uncertain when
5107 the loop is entered. This usually makes programs run more slowly.
5108 @option{-funroll-all-loops} implies the same options as
5109 @option{-funroll-loops},
5111 @item -fsplit-ivs-in-unroller
5112 @opindex -fsplit-ivs-in-unroller
5113 Enables expressing of values of induction variables in later iterations
5114 of the unrolled loop using the value in the first iteration. This breaks
5115 long dependency chains, thus improving efficiency of the scheduling passes.
5117 Combination of @option{-fweb} and CSE is often sufficient to obtain the
5118 same effect. However in cases the loop body is more complicated than
5119 a single basic block, this is not reliable. It also does not work at all
5120 on some of the architectures due to restrictions in the CSE pass.
5122 This optimization is enabled by default.
5124 @item -fvariable-expansion-in-unroller
5125 @opindex -fvariable-expansion-in-unroller
5126 With this option, the compiler will create multiple copies of some
5127 local variables when unrolling a loop which can result in superior code.
5129 @item -fprefetch-loop-arrays
5130 @opindex fprefetch-loop-arrays
5131 If supported by the target machine, generate instructions to prefetch
5132 memory to improve the performance of loops that access large arrays.
5134 These options may generate better or worse code; results are highly
5135 dependent on the structure of loops within the source code.
5138 @itemx -fno-peephole2
5139 @opindex fno-peephole
5140 @opindex fno-peephole2
5141 Disable any machine-specific peephole optimizations. The difference
5142 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
5143 are implemented in the compiler; some targets use one, some use the
5144 other, a few use both.
5146 @option{-fpeephole} is enabled by default.
5147 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5149 @item -fno-guess-branch-probability
5150 @opindex fno-guess-branch-probability
5151 Do not guess branch probabilities using heuristics.
5153 GCC will use heuristics to guess branch probabilities if they are
5154 not provided by profiling feedback (@option{-fprofile-arcs}). These
5155 heuristics are based on the control flow graph. If some branch probabilities
5156 are specified by @samp{__builtin_expect}, then the heuristics will be
5157 used to guess branch probabilities for the rest of the control flow graph,
5158 taking the @samp{__builtin_expect} info into account. The interactions
5159 between the heuristics and @samp{__builtin_expect} can be complex, and in
5160 some cases, it may be useful to disable the heuristics so that the effects
5161 of @samp{__builtin_expect} are easier to understand.
5163 The default is @option{-fguess-branch-probability} at levels
5164 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5166 @item -freorder-blocks
5167 @opindex freorder-blocks
5168 Reorder basic blocks in the compiled function in order to reduce number of
5169 taken branches and improve code locality.
5171 Enabled at levels @option{-O2}, @option{-O3}.
5173 @item -freorder-blocks-and-partition
5174 @opindex freorder-blocks-and-partition
5175 In addition to reordering basic blocks in the compiled function, in order
5176 to reduce number of taken branches, partitions hot and cold basic blocks
5177 into separate sections of the assembly and .o files, to improve
5178 paging and cache locality performance.
5180 This optimization is automatically turned off in the presence of
5181 exception handling, for linkonce sections, for functions with a user-defined
5182 section attribute and on any architecture that does not support named
5185 @item -freorder-functions
5186 @opindex freorder-functions
5187 Reorder functions in the object file in order to
5188 improve code locality. This is implemented by using special
5189 subsections @code{.text.hot} for most frequently executed functions and
5190 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
5191 the linker so object file format must support named sections and linker must
5192 place them in a reasonable way.
5194 Also profile feedback must be available in to make this option effective. See
5195 @option{-fprofile-arcs} for details.
5197 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5199 @item -fstrict-aliasing
5200 @opindex fstrict-aliasing
5201 Allows the compiler to assume the strictest aliasing rules applicable to
5202 the language being compiled. For C (and C++), this activates
5203 optimizations based on the type of expressions. In particular, an
5204 object of one type is assumed never to reside at the same address as an
5205 object of a different type, unless the types are almost the same. For
5206 example, an @code{unsigned int} can alias an @code{int}, but not a
5207 @code{void*} or a @code{double}. A character type may alias any other
5210 Pay special attention to code like this:
5223 The practice of reading from a different union member than the one most
5224 recently written to (called ``type-punning'') is common. Even with
5225 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
5226 is accessed through the union type. So, the code above will work as
5227 expected. However, this code might not:
5238 Every language that wishes to perform language-specific alias analysis
5239 should define a function that computes, given an @code{tree}
5240 node, an alias set for the node. Nodes in different alias sets are not
5241 allowed to alias. For an example, see the C front-end function
5242 @code{c_get_alias_set}.
5244 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
5246 @item -falign-functions
5247 @itemx -falign-functions=@var{n}
5248 @opindex falign-functions
5249 Align the start of functions to the next power-of-two greater than
5250 @var{n}, skipping up to @var{n} bytes. For instance,
5251 @option{-falign-functions=32} aligns functions to the next 32-byte
5252 boundary, but @option{-falign-functions=24} would align to the next
5253 32-byte boundary only if this can be done by skipping 23 bytes or less.
5255 @option{-fno-align-functions} and @option{-falign-functions=1} are
5256 equivalent and mean that functions will not be aligned.
5258 Some assemblers only support this flag when @var{n} is a power of two;
5259 in that case, it is rounded up.
5261 If @var{n} is not specified or is zero, use a machine-dependent default.
5263 Enabled at levels @option{-O2}, @option{-O3}.
5265 @item -falign-labels
5266 @itemx -falign-labels=@var{n}
5267 @opindex falign-labels
5268 Align all branch targets to a power-of-two boundary, skipping up to
5269 @var{n} bytes like @option{-falign-functions}. This option can easily
5270 make code slower, because it must insert dummy operations for when the
5271 branch target is reached in the usual flow of the code.
5273 @option{-fno-align-labels} and @option{-falign-labels=1} are
5274 equivalent and mean that labels will not be aligned.
5276 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5277 are greater than this value, then their values are used instead.
5279 If @var{n} is not specified or is zero, use a machine-dependent default
5280 which is very likely to be @samp{1}, meaning no alignment.
5282 Enabled at levels @option{-O2}, @option{-O3}.
5285 @itemx -falign-loops=@var{n}
5286 @opindex falign-loops
5287 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5288 like @option{-falign-functions}. The hope is that the loop will be
5289 executed many times, which will make up for any execution of the dummy
5292 @option{-fno-align-loops} and @option{-falign-loops=1} are
5293 equivalent and mean that loops will not be aligned.
5295 If @var{n} is not specified or is zero, use a machine-dependent default.
5297 Enabled at levels @option{-O2}, @option{-O3}.
5300 @itemx -falign-jumps=@var{n}
5301 @opindex falign-jumps
5302 Align branch targets to a power-of-two boundary, for branch targets
5303 where the targets can only be reached by jumping, skipping up to @var{n}
5304 bytes like @option{-falign-functions}. In this case, no dummy operations
5307 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5308 equivalent and mean that loops will not be aligned.
5310 If @var{n} is not specified or is zero, use a machine-dependent default.
5312 Enabled at levels @option{-O2}, @option{-O3}.
5314 @item -funit-at-a-time
5315 @opindex funit-at-a-time
5316 Parse the whole compilation unit before starting to produce code.
5317 This allows some extra optimizations to take place but consumes
5318 more memory (in general). There are some compatibility issues
5319 with @emph{unit-at-at-time} mode:
5322 enabling @emph{unit-at-a-time} mode may change the order
5323 in which functions, variables, and top-level @code{asm} statements
5324 are emitted, and will likely break code relying on some particular
5325 ordering. The majority of such top-level @code{asm} statements,
5326 though, can be replaced by @code{section} attributes.
5329 @emph{unit-at-a-time} mode removes unreferenced static variables
5330 and functions. This may result in undefined references
5331 when an @code{asm} statement refers directly to variables or functions
5332 that are otherwise unused. In that case either the variable/function
5333 shall be listed as an operand of the @code{asm} statement operand or,
5334 in the case of top-level @code{asm} statements the attribute @code{used}
5335 shall be used on the declaration.
5338 Static functions now can use non-standard passing conventions that
5339 may break @code{asm} statements calling functions directly. Again,
5340 attribute @code{used} will prevent this behavior.
5343 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5344 but this scheme may not be supported by future releases of GCC@.
5346 Enabled at levels @option{-O2}, @option{-O3}.
5350 Constructs webs as commonly used for register allocation purposes and assign
5351 each web individual pseudo register. This allows the register allocation pass
5352 to operate on pseudos directly, but also strengthens several other optimization
5353 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5354 however, make debugging impossible, since variables will no longer stay in a
5357 Enabled by default with @option{-funroll-loops}.
5359 @item -fwhole-program
5360 @opindex fwhole-program
5361 Assume that the current compilation unit represents whole program being
5362 compiled. All public functions and variables with the exception of @code{main}
5363 and those merged by attribute @code{externally_visible} become static functions
5364 and in a affect gets more aggressively optimized by interprocedural optimizers.
5365 While this option is equivalent to proper use of @code{static} keyword for
5366 programs consisting of single file, in combination with option
5367 @option{--combine} this flag can be used to compile most of smaller scale C
5368 programs since the functions and variables become local for the whole combined
5369 compilation unit, not for the single source file itself.
5372 @item -fno-cprop-registers
5373 @opindex fno-cprop-registers
5374 After register allocation and post-register allocation instruction splitting,
5375 we perform a copy-propagation pass to try to reduce scheduling dependencies
5376 and occasionally eliminate the copy.
5378 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5380 @item -fprofile-generate
5381 @opindex fprofile-generate
5383 Enable options usually used for instrumenting application to produce
5384 profile useful for later recompilation with profile feedback based
5385 optimization. You must use @option{-fprofile-generate} both when
5386 compiling and when linking your program.
5388 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5391 @opindex fprofile-use
5392 Enable profile feedback directed optimizations, and optimizations
5393 generally profitable only with profile feedback available.
5395 The following options are enabled: @code{-fbranch-probabilities}, @code{-fvpt},
5396 @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer},
5397 @code{-fno-loop-optimize}.
5401 The following options control compiler behavior regarding floating
5402 point arithmetic. These options trade off between speed and
5403 correctness. All must be specifically enabled.
5407 @opindex ffloat-store
5408 Do not store floating point variables in registers, and inhibit other
5409 options that might change whether a floating point value is taken from a
5412 @cindex floating point precision
5413 This option prevents undesirable excess precision on machines such as
5414 the 68000 where the floating registers (of the 68881) keep more
5415 precision than a @code{double} is supposed to have. Similarly for the
5416 x86 architecture. For most programs, the excess precision does only
5417 good, but a few programs rely on the precise definition of IEEE floating
5418 point. Use @option{-ffloat-store} for such programs, after modifying
5419 them to store all pertinent intermediate computations into variables.
5423 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5424 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5425 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5426 and @option{fcx-limited-range}.
5428 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5430 This option should never be turned on by any @option{-O} option since
5431 it can result in incorrect output for programs which depend on
5432 an exact implementation of IEEE or ISO rules/specifications for
5435 @item -fno-math-errno
5436 @opindex fno-math-errno
5437 Do not set ERRNO after calling math functions that are executed
5438 with a single instruction, e.g., sqrt. A program that relies on
5439 IEEE exceptions for math error handling may want to use this flag
5440 for speed while maintaining IEEE arithmetic compatibility.
5442 This option should never be turned on by any @option{-O} option since
5443 it can result in incorrect output for programs which depend on
5444 an exact implementation of IEEE or ISO rules/specifications for
5447 The default is @option{-fmath-errno}.
5449 On Darwin systems, the math library never sets @code{errno}. There is therefore
5450 no reason for the compiler to consider the possibility that it might,
5451 and @option{-fno-math-errno} is the default.
5453 @item -funsafe-math-optimizations
5454 @opindex funsafe-math-optimizations
5455 Allow optimizations for floating-point arithmetic that (a) assume
5456 that arguments and results are valid and (b) may violate IEEE or
5457 ANSI standards. When used at link-time, it may include libraries
5458 or startup files that change the default FPU control word or other
5459 similar optimizations.
5461 This option should never be turned on by any @option{-O} option since
5462 it can result in incorrect output for programs which depend on
5463 an exact implementation of IEEE or ISO rules/specifications for
5466 The default is @option{-fno-unsafe-math-optimizations}.
5468 @item -ffinite-math-only
5469 @opindex ffinite-math-only
5470 Allow optimizations for floating-point arithmetic that assume
5471 that arguments and results are not NaNs or +-Infs.
5473 This option should never be turned on by any @option{-O} option since
5474 it can result in incorrect output for programs which depend on
5475 an exact implementation of IEEE or ISO rules/specifications.
5477 The default is @option{-fno-finite-math-only}.
5479 @item -fno-trapping-math
5480 @opindex fno-trapping-math
5481 Compile code assuming that floating-point operations cannot generate
5482 user-visible traps. These traps include division by zero, overflow,
5483 underflow, inexact result and invalid operation. This option implies
5484 @option{-fno-signaling-nans}. Setting this option may allow faster
5485 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5487 This option should never be turned on by any @option{-O} option since
5488 it can result in incorrect output for programs which depend on
5489 an exact implementation of IEEE or ISO rules/specifications for
5492 The default is @option{-ftrapping-math}.
5494 @item -frounding-math
5495 @opindex frounding-math
5496 Disable transformations and optimizations that assume default floating
5497 point rounding behavior. This is round-to-zero for all floating point
5498 to integer conversions, and round-to-nearest for all other arithmetic
5499 truncations. This option should be specified for programs that change
5500 the FP rounding mode dynamically, or that may be executed with a
5501 non-default rounding mode. This option disables constant folding of
5502 floating point expressions at compile-time (which may be affected by
5503 rounding mode) and arithmetic transformations that are unsafe in the
5504 presence of sign-dependent rounding modes.
5506 The default is @option{-fno-rounding-math}.
5508 This option is experimental and does not currently guarantee to
5509 disable all GCC optimizations that are affected by rounding mode.
5510 Future versions of GCC may provide finer control of this setting
5511 using C99's @code{FENV_ACCESS} pragma. This command line option
5512 will be used to specify the default state for @code{FENV_ACCESS}.
5514 @item -fsignaling-nans
5515 @opindex fsignaling-nans
5516 Compile code assuming that IEEE signaling NaNs may generate user-visible
5517 traps during floating-point operations. Setting this option disables
5518 optimizations that may change the number of exceptions visible with
5519 signaling NaNs. This option implies @option{-ftrapping-math}.
5521 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5524 The default is @option{-fno-signaling-nans}.
5526 This option is experimental and does not currently guarantee to
5527 disable all GCC optimizations that affect signaling NaN behavior.
5529 @item -fsingle-precision-constant
5530 @opindex fsingle-precision-constant
5531 Treat floating point constant as single precision constant instead of
5532 implicitly converting it to double precision constant.
5534 @item -fcx-limited-range
5535 @itemx -fno-cx-limited-range
5536 @opindex fcx-limited-range
5537 @opindex fno-cx-limited-range
5538 When enabled, this option states that a range reduction step is not
5539 needed when performing complex division. The default is
5540 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5542 This option controls the default setting of the ISO C99
5543 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5548 The following options control optimizations that may improve
5549 performance, but are not enabled by any @option{-O} options. This
5550 section includes experimental options that may produce broken code.
5553 @item -fbranch-probabilities
5554 @opindex fbranch-probabilities
5555 After running a program compiled with @option{-fprofile-arcs}
5556 (@pxref{Debugging Options,, Options for Debugging Your Program or
5557 @command{gcc}}), you can compile it a second time using
5558 @option{-fbranch-probabilities}, to improve optimizations based on
5559 the number of times each branch was taken. When the program
5560 compiled with @option{-fprofile-arcs} exits it saves arc execution
5561 counts to a file called @file{@var{sourcename}.gcda} for each source
5562 file The information in this data file is very dependent on the
5563 structure of the generated code, so you must use the same source code
5564 and the same optimization options for both compilations.
5566 With @option{-fbranch-probabilities}, GCC puts a
5567 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5568 These can be used to improve optimization. Currently, they are only
5569 used in one place: in @file{reorg.c}, instead of guessing which path a
5570 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5571 exactly determine which path is taken more often.
5573 @item -fprofile-values
5574 @opindex fprofile-values
5575 If combined with @option{-fprofile-arcs}, it adds code so that some
5576 data about values of expressions in the program is gathered.
5578 With @option{-fbranch-probabilities}, it reads back the data gathered
5579 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5580 notes to instructions for their later usage in optimizations.
5582 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5586 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5587 a code to gather information about values of expressions.
5589 With @option{-fbranch-probabilities}, it reads back the data gathered
5590 and actually performs the optimizations based on them.
5591 Currently the optimizations include specialization of division operation
5592 using the knowledge about the value of the denominator.
5594 @item -frename-registers
5595 @opindex frename-registers
5596 Attempt to avoid false dependencies in scheduled code by making use
5597 of registers left over after register allocation. This optimization
5598 will most benefit processors with lots of registers. Depending on the
5599 debug information format adopted by the target, however, it can
5600 make debugging impossible, since variables will no longer stay in
5601 a ``home register''.
5603 Enabled by default with @option{-funroll-loops}.
5607 Perform tail duplication to enlarge superblock size. This transformation
5608 simplifies the control flow of the function allowing other optimizations to do
5611 Enabled with @option{-fprofile-use}.
5613 @item -funroll-loops
5614 @opindex funroll-loops
5615 Unroll loops whose number of iterations can be determined at compile time or
5616 upon entry to the loop. @option{-funroll-loops} implies
5617 @option{-frerun-cse-after-loop}, @option{-fweb} and @option{-frename-registers}.
5618 It also turns on complete loop peeling (i.e.@: complete removal of loops with
5619 small constant number of iterations). This option makes code larger, and may
5620 or may not make it run faster.
5622 Enabled with @option{-fprofile-use}.
5624 @item -funroll-all-loops
5625 @opindex funroll-all-loops
5626 Unroll all loops, even if their number of iterations is uncertain when
5627 the loop is entered. This usually makes programs run more slowly.
5628 @option{-funroll-all-loops} implies the same options as
5629 @option{-funroll-loops}.
5632 @opindex fpeel-loops
5633 Peels the loops for that there is enough information that they do not
5634 roll much (from profile feedback). It also turns on complete loop peeling
5635 (i.e.@: complete removal of loops with small constant number of iterations).
5637 Enabled with @option{-fprofile-use}.
5639 @item -fmove-loop-invariants
5640 @opindex fmove-loop-invariants
5641 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5642 at level @option{-O1}
5644 @item -funswitch-loops
5645 @opindex funswitch-loops
5646 Move branches with loop invariant conditions out of the loop, with duplicates
5647 of the loop on both branches (modified according to result of the condition).
5649 @item -fprefetch-loop-arrays
5650 @opindex fprefetch-loop-arrays
5651 If supported by the target machine, generate instructions to prefetch
5652 memory to improve the performance of loops that access large arrays.
5654 Disabled at level @option{-Os}.
5656 @item -ffunction-sections
5657 @itemx -fdata-sections
5658 @opindex ffunction-sections
5659 @opindex fdata-sections
5660 Place each function or data item into its own section in the output
5661 file if the target supports arbitrary sections. The name of the
5662 function or the name of the data item determines the section's name
5665 Use these options on systems where the linker can perform optimizations
5666 to improve locality of reference in the instruction space. Most systems
5667 using the ELF object format and SPARC processors running Solaris 2 have
5668 linkers with such optimizations. AIX may have these optimizations in
5671 Only use these options when there are significant benefits from doing
5672 so. When you specify these options, the assembler and linker will
5673 create larger object and executable files and will also be slower.
5674 You will not be able to use @code{gprof} on all systems if you
5675 specify this option and you may have problems with debugging if
5676 you specify both this option and @option{-g}.
5678 @item -fbranch-target-load-optimize
5679 @opindex fbranch-target-load-optimize
5680 Perform branch target register load optimization before prologue / epilogue
5682 The use of target registers can typically be exposed only during reload,
5683 thus hoisting loads out of loops and doing inter-block scheduling needs
5684 a separate optimization pass.
5686 @item -fbranch-target-load-optimize2
5687 @opindex fbranch-target-load-optimize2
5688 Perform branch target register load optimization after prologue / epilogue
5691 @item -fbtr-bb-exclusive
5692 @opindex fbtr-bb-exclusive
5693 When performing branch target register load optimization, don't reuse
5694 branch target registers in within any basic block.
5696 @item -fstack-protector
5697 Emit extra code to check for buffer overflows, such as stack smashing
5698 attacks. This is done by adding a guard variable to functions with
5699 vulnerable objects. This includes functions that call alloca, and
5700 functions with buffers larger than 8 bytes. The guards are initialized
5701 when a function is entered and then checked when the function exits.
5702 If a guard check fails, an error message is printed and the program exits.
5704 @item -fstack-protector-all
5705 Like @option{-fstack-protector} except that all functions are protected.
5707 @item --param @var{name}=@var{value}
5709 In some places, GCC uses various constants to control the amount of
5710 optimization that is done. For example, GCC will not inline functions
5711 that contain more that a certain number of instructions. You can
5712 control some of these constants on the command-line using the
5713 @option{--param} option.
5715 The names of specific parameters, and the meaning of the values, are
5716 tied to the internals of the compiler, and are subject to change
5717 without notice in future releases.
5719 In each case, the @var{value} is an integer. The allowable choices for
5720 @var{name} are given in the following table:
5723 @item salias-max-implicit-fields
5724 The maximum number of fields in a variable without direct
5725 structure accesses for which structure aliasing will consider trying
5726 to track each field. The default is 5
5728 @item sra-max-structure-size
5729 The maximum structure size, in bytes, at which the scalar replacement
5730 of aggregates (SRA) optimization will perform block copies. The
5731 default value, 0, implies that GCC will select the most appropriate
5734 @item sra-field-structure-ratio
5735 The threshold ratio (as a percentage) between instantiated fields and
5736 the complete structure size. We say that if the ratio of the number
5737 of bytes in instantiated fields to the number of bytes in the complete
5738 structure exceeds this parameter, then block copies are not used. The
5741 @item max-crossjump-edges
5742 The maximum number of incoming edges to consider for crossjumping.
5743 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5744 the number of edges incoming to each block. Increasing values mean
5745 more aggressive optimization, making the compile time increase with
5746 probably small improvement in executable size.
5748 @item min-crossjump-insns
5749 The minimum number of instructions which must be matched at the end
5750 of two blocks before crossjumping will be performed on them. This
5751 value is ignored in the case where all instructions in the block being
5752 crossjumped from are matched. The default value is 5.
5754 @item max-grow-copy-bb-insns
5755 The maximum code size expansion factor when copying basic blocks
5756 instead of jumping. The expansion is relative to a jump instruction.
5757 The default value is 8.
5759 @item max-goto-duplication-insns
5760 The maximum number of instructions to duplicate to a block that jumps
5761 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5762 passes, GCC factors computed gotos early in the compilation process,
5763 and unfactors them as late as possible. Only computed jumps at the
5764 end of a basic blocks with no more than max-goto-duplication-insns are
5765 unfactored. The default value is 8.
5767 @item max-delay-slot-insn-search
5768 The maximum number of instructions to consider when looking for an
5769 instruction to fill a delay slot. If more than this arbitrary number of
5770 instructions is searched, the time savings from filling the delay slot
5771 will be minimal so stop searching. Increasing values mean more
5772 aggressive optimization, making the compile time increase with probably
5773 small improvement in executable run time.
5775 @item max-delay-slot-live-search
5776 When trying to fill delay slots, the maximum number of instructions to
5777 consider when searching for a block with valid live register
5778 information. Increasing this arbitrarily chosen value means more
5779 aggressive optimization, increasing the compile time. This parameter
5780 should be removed when the delay slot code is rewritten to maintain the
5783 @item max-gcse-memory
5784 The approximate maximum amount of memory that will be allocated in
5785 order to perform the global common subexpression elimination
5786 optimization. If more memory than specified is required, the
5787 optimization will not be done.
5789 @item max-gcse-passes
5790 The maximum number of passes of GCSE to run. The default is 1.
5792 @item max-pending-list-length
5793 The maximum number of pending dependencies scheduling will allow
5794 before flushing the current state and starting over. Large functions
5795 with few branches or calls can create excessively large lists which
5796 needlessly consume memory and resources.
5798 @item max-inline-insns-single
5799 Several parameters control the tree inliner used in gcc.
5800 This number sets the maximum number of instructions (counted in GCC's
5801 internal representation) in a single function that the tree inliner
5802 will consider for inlining. This only affects functions declared
5803 inline and methods implemented in a class declaration (C++).
5804 The default value is 450.
5806 @item max-inline-insns-auto
5807 When you use @option{-finline-functions} (included in @option{-O3}),
5808 a lot of functions that would otherwise not be considered for inlining
5809 by the compiler will be investigated. To those functions, a different
5810 (more restrictive) limit compared to functions declared inline can
5812 The default value is 90.
5814 @item large-function-insns
5815 The limit specifying really large functions. For functions larger than this
5816 limit after inlining inlining is constrained by
5817 @option{--param large-function-growth}. This parameter is useful primarily
5818 to avoid extreme compilation time caused by non-linear algorithms used by the
5820 This parameter is ignored when @option{-funit-at-a-time} is not used.
5821 The default value is 2700.
5823 @item large-function-growth
5824 Specifies maximal growth of large function caused by inlining in percents.
5825 This parameter is ignored when @option{-funit-at-a-time} is not used.
5826 The default value is 100 which limits large function growth to 2.0 times
5829 @item large-unit-insns
5830 The limit specifying large translation unit. Growth caused by inlining of
5831 units larger than this limit is limited by @option{--param inline-unit-growth}.
5832 For small units this might be too tight (consider unit consisting of function A
5833 that is inline and B that just calls A three time. If B is small relative to
5834 A, the growth of unit is 300\% and yet such inlining is very sane. For very
5835 large units consisting of small inlininable functions however the overall unit
5836 growth limit is needed to avoid exponential explosion of code size. Thus for
5837 smaller units, the size is increased to @option{--param large-unit-insns}
5838 before aplying @option{--param inline-unit-growth}. The default is 10000
5840 @item inline-unit-growth
5841 Specifies maximal overall growth of the compilation unit caused by inlining.
5842 This parameter is ignored when @option{-funit-at-a-time} is not used.
5843 The default value is 50 which limits unit growth to 1.5 times the original
5846 @item max-inline-insns-recursive
5847 @itemx max-inline-insns-recursive-auto
5848 Specifies maximum number of instructions out-of-line copy of self recursive inline
5849 function can grow into by performing recursive inlining.
5851 For functions declared inline @option{--param max-inline-insns-recursive} is
5852 taken into acount. For function not declared inline, recursive inlining
5853 happens only when @option{-finline-functions} (included in @option{-O3}) is
5854 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5855 default value is 450.
5857 @item max-inline-recursive-depth
5858 @itemx max-inline-recursive-depth-auto
5859 Specifies maximum recursion depth used by the recursive inlining.
5861 For functions declared inline @option{--param max-inline-recursive-depth} is
5862 taken into acount. For function not declared inline, recursive inlining
5863 happens only when @option{-finline-functions} (included in @option{-O3}) is
5864 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5865 default value is 450.
5867 @item min-inline-recursive-probability
5868 Recursive inlining is profitable only for function having deep recursion
5869 in average and can hurt for function having little recursion depth by
5870 increasing the prologue size or complexity of function body to other
5873 When profile feedback is available (see @option{-fprofile-generate}) the actual
5874 recursion depth can be guessed from probability that function will recurse via
5875 given call expression. This parameter limits inlining only to call expression
5876 whose probability exceeds given threshold (in percents). The default value is
5879 @item inline-call-cost
5880 Specify cost of call instruction relative to simple arithmetics operations
5881 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5882 functions and at the same time increases size of leaf function that is believed to
5883 reduce function size by being inlined. In effect it increases amount of
5884 inlining for code having large abstraction penalty (many functions that just
5885 pass the arguments to other functions) and decrease inlining for code with low
5886 abstraction penalty. The default value is 16.
5888 @item max-unrolled-insns
5889 The maximum number of instructions that a loop should have if that loop
5890 is unrolled, and if the loop is unrolled, it determines how many times
5891 the loop code is unrolled.
5893 @item max-average-unrolled-insns
5894 The maximum number of instructions biased by probabilities of their execution
5895 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5896 it determines how many times the loop code is unrolled.
5898 @item max-unroll-times
5899 The maximum number of unrollings of a single loop.
5901 @item max-peeled-insns
5902 The maximum number of instructions that a loop should have if that loop
5903 is peeled, and if the loop is peeled, it determines how many times
5904 the loop code is peeled.
5906 @item max-peel-times
5907 The maximum number of peelings of a single loop.
5909 @item max-completely-peeled-insns
5910 The maximum number of insns of a completely peeled loop.
5912 @item max-completely-peel-times
5913 The maximum number of iterations of a loop to be suitable for complete peeling.
5915 @item max-unswitch-insns
5916 The maximum number of insns of an unswitched loop.
5918 @item max-unswitch-level
5919 The maximum number of branches unswitched in a single loop.
5922 The minimum cost of an expensive expression in the loop invariant motion.
5924 @item iv-consider-all-candidates-bound
5925 Bound on number of candidates for induction variables below that
5926 all candidates are considered for each use in induction variable
5927 optimizations. Only the most relevant candidates are considered
5928 if there are more candidates, to avoid quadratic time complexity.
5930 @item iv-max-considered-uses
5931 The induction variable optimizations give up on loops that contain more
5932 induction variable uses.
5934 @item iv-always-prune-cand-set-bound
5935 If number of candidates in the set is smaller than this value,
5936 we always try to remove unnecessary ivs from the set during its
5937 optimization when a new iv is added to the set.
5939 @item scev-max-expr-size
5940 Bound on size of expressions used in the scalar evolutions analyzer.
5941 Large expressions slow the analyzer.
5943 @item vect-max-version-checks
5944 The maximum number of runtime checks that can be performed when doing
5945 loop versioning in the vectorizer. See option ftree-vect-loop-version
5946 for more information.
5948 @item max-iterations-to-track
5950 The maximum number of iterations of a loop the brute force algorithm
5951 for analysis of # of iterations of the loop tries to evaluate.
5953 @item hot-bb-count-fraction
5954 Select fraction of the maximal count of repetitions of basic block in program
5955 given basic block needs to have to be considered hot.
5957 @item hot-bb-frequency-fraction
5958 Select fraction of the maximal frequency of executions of basic block in
5959 function given basic block needs to have to be considered hot
5961 @item max-predicted-iterations
5962 The maximum number of loop iterations we predict statically. This is useful
5963 in cases where function contain single loop with known bound and other loop
5964 with unknown. We predict the known number of iterations correctly, while
5965 the unknown number of iterations average to roughly 10. This means that the
5966 loop without bounds would appear artificially cold relative to the other one.
5968 @item tracer-dynamic-coverage
5969 @itemx tracer-dynamic-coverage-feedback
5971 This value is used to limit superblock formation once the given percentage of
5972 executed instructions is covered. This limits unnecessary code size
5975 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5976 feedback is available. The real profiles (as opposed to statically estimated
5977 ones) are much less balanced allowing the threshold to be larger value.
5979 @item tracer-max-code-growth
5980 Stop tail duplication once code growth has reached given percentage. This is
5981 rather hokey argument, as most of the duplicates will be eliminated later in
5982 cross jumping, so it may be set to much higher values than is the desired code
5985 @item tracer-min-branch-ratio
5987 Stop reverse growth when the reverse probability of best edge is less than this
5988 threshold (in percent).
5990 @item tracer-min-branch-ratio
5991 @itemx tracer-min-branch-ratio-feedback
5993 Stop forward growth if the best edge do have probability lower than this
5996 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5997 compilation for profile feedback and one for compilation without. The value
5998 for compilation with profile feedback needs to be more conservative (higher) in
5999 order to make tracer effective.
6001 @item max-cse-path-length
6003 Maximum number of basic blocks on path that cse considers. The default is 10.
6006 The maximum instructions CSE process before flushing. The default is 1000.
6008 @item global-var-threshold
6010 Counts the number of function calls (@var{n}) and the number of
6011 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
6012 single artificial variable will be created to represent all the
6013 call-clobbered variables at function call sites. This artificial
6014 variable will then be made to alias every call-clobbered variable.
6015 (done as @code{int * size_t} on the host machine; beware overflow).
6017 @item max-aliased-vops
6019 Maximum number of virtual operands allowed to represent aliases
6020 before triggering the alias grouping heuristic. Alias grouping
6021 reduces compile times and memory consumption needed for aliasing at
6022 the expense of precision loss in alias information.
6024 @item ggc-min-expand
6026 GCC uses a garbage collector to manage its own memory allocation. This
6027 parameter specifies the minimum percentage by which the garbage
6028 collector's heap should be allowed to expand between collections.
6029 Tuning this may improve compilation speed; it has no effect on code
6032 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
6033 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
6034 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
6035 GCC is not able to calculate RAM on a particular platform, the lower
6036 bound of 30% is used. Setting this parameter and
6037 @option{ggc-min-heapsize} to zero causes a full collection to occur at
6038 every opportunity. This is extremely slow, but can be useful for
6041 @item ggc-min-heapsize
6043 Minimum size of the garbage collector's heap before it begins bothering
6044 to collect garbage. The first collection occurs after the heap expands
6045 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
6046 tuning this may improve compilation speed, and has no effect on code
6049 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
6050 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
6051 with a lower bound of 4096 (four megabytes) and an upper bound of
6052 131072 (128 megabytes). If GCC is not able to calculate RAM on a
6053 particular platform, the lower bound is used. Setting this parameter
6054 very large effectively disables garbage collection. Setting this
6055 parameter and @option{ggc-min-expand} to zero causes a full collection
6056 to occur at every opportunity.
6058 @item max-reload-search-insns
6059 The maximum number of instruction reload should look backward for equivalent
6060 register. Increasing values mean more aggressive optimization, making the
6061 compile time increase with probably slightly better performance. The default
6064 @item max-cselib-memory-location
6065 The maximum number of memory locations cselib should take into acount.
6066 Increasing values mean more aggressive optimization, making the compile time
6067 increase with probably slightly better performance. The default value is 500.
6069 @item max-flow-memory-location
6070 Similar as @option{max-cselib-memory-location} but for dataflow liveness.
6071 The default value is 100.
6073 @item reorder-blocks-duplicate
6074 @itemx reorder-blocks-duplicate-feedback
6076 Used by basic block reordering pass to decide whether to use unconditional
6077 branch or duplicate the code on its destination. Code is duplicated when its
6078 estimated size is smaller than this value multiplied by the estimated size of
6079 unconditional jump in the hot spots of the program.
6081 The @option{reorder-block-duplicate-feedback} is used only when profile
6082 feedback is available and may be set to higher values than
6083 @option{reorder-block-duplicate} since information about the hot spots is more
6086 @item max-sched-region-blocks
6087 The maximum number of blocks in a region to be considered for
6088 interblock scheduling. The default value is 10.
6090 @item max-sched-region-insns
6091 The maximum number of insns in a region to be considered for
6092 interblock scheduling. The default value is 100.
6094 @item min-sched-prob
6095 The minimum probability of reaching a source block for interblock
6096 speculative scheduling. The default value is 40.
6098 @item max-last-value-rtl
6100 The maximum size measured as number of RTLs that can be recorded in an expression
6101 in combiner for a pseudo register as last known value of that register. The default
6104 @item integer-share-limit
6105 Small integer constants can use a shared data structure, reducing the
6106 compiler's memory usage and increasing its speed. This sets the maximum
6107 value of a shared integer constant's. The default value is 256.
6109 @item min-virtual-mappings
6110 Specifies the minimum number of virtual mappings in the incremental
6111 SSA updater that should be registered to trigger the virtual mappings
6112 heuristic defined by virtual-mappings-ratio. The default value is
6115 @item virtual-mappings-ratio
6116 If the number of virtual mappings is virtual-mappings-ratio bigger
6117 than the number of virtual symbols to be updated, then the incremental
6118 SSA updater switches to a full update for those symbols. The default
6121 @item ssp-buffer-size
6122 The minimum size of buffers (i.e. arrays) that will receive stack smashing
6123 protection when @option{-fstack-protection} is used.
6125 @item max-jump-thread-duplication-stmts
6126 Maximum number of statements allowed in a block that needs to be
6127 duplicated when threading jumps.
6129 @item max-fields-for-field-sensitive
6130 Maximum number of fields in a structure we will treat in
6131 a field sensitive manner during pointer analysis.
6136 @node Preprocessor Options
6137 @section Options Controlling the Preprocessor
6138 @cindex preprocessor options
6139 @cindex options, preprocessor
6141 These options control the C preprocessor, which is run on each C source
6142 file before actual compilation.
6144 If you use the @option{-E} option, nothing is done except preprocessing.
6145 Some of these options make sense only together with @option{-E} because
6146 they cause the preprocessor output to be unsuitable for actual
6151 You can use @option{-Wp,@var{option}} to bypass the compiler driver
6152 and pass @var{option} directly through to the preprocessor. If
6153 @var{option} contains commas, it is split into multiple options at the
6154 commas. However, many options are modified, translated or interpreted
6155 by the compiler driver before being passed to the preprocessor, and
6156 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
6157 interface is undocumented and subject to change, so whenever possible
6158 you should avoid using @option{-Wp} and let the driver handle the
6161 @item -Xpreprocessor @var{option}
6162 @opindex preprocessor
6163 Pass @var{option} as an option to the preprocessor. You can use this to
6164 supply system-specific preprocessor options which GCC does not know how to
6167 If you want to pass an option that takes an argument, you must use
6168 @option{-Xpreprocessor} twice, once for the option and once for the argument.
6171 @include cppopts.texi
6173 @node Assembler Options
6174 @section Passing Options to the Assembler
6176 @c prevent bad page break with this line
6177 You can pass options to the assembler.
6180 @item -Wa,@var{option}
6182 Pass @var{option} as an option to the assembler. If @var{option}
6183 contains commas, it is split into multiple options at the commas.
6185 @item -Xassembler @var{option}
6187 Pass @var{option} as an option to the assembler. You can use this to
6188 supply system-specific assembler options which GCC does not know how to
6191 If you want to pass an option that takes an argument, you must use
6192 @option{-Xassembler} twice, once for the option and once for the argument.
6197 @section Options for Linking
6198 @cindex link options
6199 @cindex options, linking
6201 These options come into play when the compiler links object files into
6202 an executable output file. They are meaningless if the compiler is
6203 not doing a link step.
6207 @item @var{object-file-name}
6208 A file name that does not end in a special recognized suffix is
6209 considered to name an object file or library. (Object files are
6210 distinguished from libraries by the linker according to the file
6211 contents.) If linking is done, these object files are used as input
6220 If any of these options is used, then the linker is not run, and
6221 object file names should not be used as arguments. @xref{Overall
6225 @item -l@var{library}
6226 @itemx -l @var{library}
6228 Search the library named @var{library} when linking. (The second
6229 alternative with the library as a separate argument is only for
6230 POSIX compliance and is not recommended.)
6232 It makes a difference where in the command you write this option; the
6233 linker searches and processes libraries and object files in the order they
6234 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
6235 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
6236 to functions in @samp{z}, those functions may not be loaded.
6238 The linker searches a standard list of directories for the library,
6239 which is actually a file named @file{lib@var{library}.a}. The linker
6240 then uses this file as if it had been specified precisely by name.
6242 The directories searched include several standard system directories
6243 plus any that you specify with @option{-L}.
6245 Normally the files found this way are library files---archive files
6246 whose members are object files. The linker handles an archive file by
6247 scanning through it for members which define symbols that have so far
6248 been referenced but not defined. But if the file that is found is an
6249 ordinary object file, it is linked in the usual fashion. The only
6250 difference between using an @option{-l} option and specifying a file name
6251 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
6252 and searches several directories.
6256 You need this special case of the @option{-l} option in order to
6257 link an Objective-C or Objective-C++ program.
6260 @opindex nostartfiles
6261 Do not use the standard system startup files when linking.
6262 The standard system libraries are used normally, unless @option{-nostdlib}
6263 or @option{-nodefaultlibs} is used.
6265 @item -nodefaultlibs
6266 @opindex nodefaultlibs
6267 Do not use the standard system libraries when linking.
6268 Only the libraries you specify will be passed to the linker.
6269 The standard startup files are used normally, unless @option{-nostartfiles}
6270 is used. The compiler may generate calls to @code{memcmp},
6271 @code{memset}, @code{memcpy} and @code{memmove}.
6272 These entries are usually resolved by entries in
6273 libc. These entry points should be supplied through some other
6274 mechanism when this option is specified.
6278 Do not use the standard system startup files or libraries when linking.
6279 No startup files and only the libraries you specify will be passed to
6280 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
6281 @code{memcpy} and @code{memmove}.
6282 These entries are usually resolved by entries in
6283 libc. These entry points should be supplied through some other
6284 mechanism when this option is specified.
6286 @cindex @option{-lgcc}, use with @option{-nostdlib}
6287 @cindex @option{-nostdlib} and unresolved references
6288 @cindex unresolved references and @option{-nostdlib}
6289 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
6290 @cindex @option{-nodefaultlibs} and unresolved references
6291 @cindex unresolved references and @option{-nodefaultlibs}
6292 One of the standard libraries bypassed by @option{-nostdlib} and
6293 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
6294 that GCC uses to overcome shortcomings of particular machines, or special
6295 needs for some languages.
6296 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
6297 Collection (GCC) Internals},
6298 for more discussion of @file{libgcc.a}.)
6299 In most cases, you need @file{libgcc.a} even when you want to avoid
6300 other standard libraries. In other words, when you specify @option{-nostdlib}
6301 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
6302 This ensures that you have no unresolved references to internal GCC
6303 library subroutines. (For example, @samp{__main}, used to ensure C++
6304 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
6305 GNU Compiler Collection (GCC) Internals}.)
6309 Produce a position independent executable on targets which support it.
6310 For predictable results, you must also specify the same set of options
6311 that were used to generate code (@option{-fpie}, @option{-fPIE},
6312 or model suboptions) when you specify this option.
6316 Pass the flag @option{-export-dynamic} to the ELF linker, on targets
6317 that support it. This instructs the linker to add all symbols, not
6318 only used ones, to the dynamic symbol table. This option is needed
6319 for some uses of @code{dlopen} or to allow obtaining backtraces
6320 from within a program.
6324 Remove all symbol table and relocation information from the executable.
6328 On systems that support dynamic linking, this prevents linking with the shared
6329 libraries. On other systems, this option has no effect.
6333 Produce a shared object which can then be linked with other objects to
6334 form an executable. Not all systems support this option. For predictable
6335 results, you must also specify the same set of options that were used to
6336 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
6337 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
6338 needs to build supplementary stub code for constructors to work. On
6339 multi-libbed systems, @samp{gcc -shared} must select the correct support
6340 libraries to link against. Failing to supply the correct flags may lead
6341 to subtle defects. Supplying them in cases where they are not necessary
6344 @item -shared-libgcc
6345 @itemx -static-libgcc
6346 @opindex shared-libgcc
6347 @opindex static-libgcc
6348 On systems that provide @file{libgcc} as a shared library, these options
6349 force the use of either the shared or static version respectively.
6350 If no shared version of @file{libgcc} was built when the compiler was
6351 configured, these options have no effect.
6353 There are several situations in which an application should use the
6354 shared @file{libgcc} instead of the static version. The most common
6355 of these is when the application wishes to throw and catch exceptions
6356 across different shared libraries. In that case, each of the libraries
6357 as well as the application itself should use the shared @file{libgcc}.
6359 Therefore, the G++ and GCJ drivers automatically add
6360 @option{-shared-libgcc} whenever you build a shared library or a main
6361 executable, because C++ and Java programs typically use exceptions, so
6362 this is the right thing to do.
6364 If, instead, you use the GCC driver to create shared libraries, you may
6365 find that they will not always be linked with the shared @file{libgcc}.
6366 If GCC finds, at its configuration time, that you have a non-GNU linker
6367 or a GNU linker that does not support option @option{--eh-frame-hdr},
6368 it will link the shared version of @file{libgcc} into shared libraries
6369 by default. Otherwise, it will take advantage of the linker and optimize
6370 away the linking with the shared version of @file{libgcc}, linking with
6371 the static version of libgcc by default. This allows exceptions to
6372 propagate through such shared libraries, without incurring relocation
6373 costs at library load time.
6375 However, if a library or main executable is supposed to throw or catch
6376 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6377 for the languages used in the program, or using the option
6378 @option{-shared-libgcc}, such that it is linked with the shared
6383 Bind references to global symbols when building a shared object. Warn
6384 about any unresolved references (unless overridden by the link editor
6385 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6388 @item -Xlinker @var{option}
6390 Pass @var{option} as an option to the linker. You can use this to
6391 supply system-specific linker options which GCC does not know how to
6394 If you want to pass an option that takes an argument, you must use
6395 @option{-Xlinker} twice, once for the option and once for the argument.
6396 For example, to pass @option{-assert definitions}, you must write
6397 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6398 @option{-Xlinker "-assert definitions"}, because this passes the entire
6399 string as a single argument, which is not what the linker expects.
6401 @item -Wl,@var{option}
6403 Pass @var{option} as an option to the linker. If @var{option} contains
6404 commas, it is split into multiple options at the commas.
6406 @item -u @var{symbol}
6408 Pretend the symbol @var{symbol} is undefined, to force linking of
6409 library modules to define it. You can use @option{-u} multiple times with
6410 different symbols to force loading of additional library modules.
6413 @node Directory Options
6414 @section Options for Directory Search
6415 @cindex directory options
6416 @cindex options, directory search
6419 These options specify directories to search for header files, for
6420 libraries and for parts of the compiler:
6425 Add the directory @var{dir} to the head of the list of directories to be
6426 searched for header files. This can be used to override a system header
6427 file, substituting your own version, since these directories are
6428 searched before the system header file directories. However, you should
6429 not use this option to add directories that contain vendor-supplied
6430 system header files (use @option{-isystem} for that). If you use more than
6431 one @option{-I} option, the directories are scanned in left-to-right
6432 order; the standard system directories come after.
6434 If a standard system include directory, or a directory specified with
6435 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6436 option will be ignored. The directory will still be searched but as a
6437 system directory at its normal position in the system include chain.
6438 This is to ensure that GCC's procedure to fix buggy system headers and
6439 the ordering for the include_next directive are not inadvertently changed.
6440 If you really need to change the search order for system directories,
6441 use the @option{-nostdinc} and/or @option{-isystem} options.
6443 @item -iquote@var{dir}
6445 Add the directory @var{dir} to the head of the list of directories to
6446 be searched for header files only for the case of @samp{#include
6447 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6448 otherwise just like @option{-I}.
6452 Add directory @var{dir} to the list of directories to be searched
6455 @item -B@var{prefix}
6457 This option specifies where to find the executables, libraries,
6458 include files, and data files of the compiler itself.
6460 The compiler driver program runs one or more of the subprograms
6461 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6462 @var{prefix} as a prefix for each program it tries to run, both with and
6463 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6465 For each subprogram to be run, the compiler driver first tries the
6466 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6467 was not specified, the driver tries two standard prefixes, which are
6468 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6469 those results in a file name that is found, the unmodified program
6470 name is searched for using the directories specified in your
6471 @env{PATH} environment variable.
6473 The compiler will check to see if the path provided by the @option{-B}
6474 refers to a directory, and if necessary it will add a directory
6475 separator character at the end of the path.
6477 @option{-B} prefixes that effectively specify directory names also apply
6478 to libraries in the linker, because the compiler translates these
6479 options into @option{-L} options for the linker. They also apply to
6480 includes files in the preprocessor, because the compiler translates these
6481 options into @option{-isystem} options for the preprocessor. In this case,
6482 the compiler appends @samp{include} to the prefix.
6484 The run-time support file @file{libgcc.a} can also be searched for using
6485 the @option{-B} prefix, if needed. If it is not found there, the two
6486 standard prefixes above are tried, and that is all. The file is left
6487 out of the link if it is not found by those means.
6489 Another way to specify a prefix much like the @option{-B} prefix is to use
6490 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6493 As a special kludge, if the path provided by @option{-B} is
6494 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6495 9, then it will be replaced by @file{[dir/]include}. This is to help
6496 with boot-strapping the compiler.
6498 @item -specs=@var{file}
6500 Process @var{file} after the compiler reads in the standard @file{specs}
6501 file, in order to override the defaults that the @file{gcc} driver
6502 program uses when determining what switches to pass to @file{cc1},
6503 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6504 @option{-specs=@var{file}} can be specified on the command line, and they
6505 are processed in order, from left to right.
6507 @item --sysroot=@var{dir}
6509 Use @var{dir} as the logical root directory for headers and libraries.
6510 For example, if the compiler would normally search for headers in
6511 @file{/usr/include} and libraries in @file{/usr/lib}, it will instead
6512 search @file{@var{dir}/usr/include} and @file{@var{dir}/usr/lib}.
6514 If you use both this option and the @option{-isysroot} option, then
6515 the @option{--sysroot} option will apply to libraries, but the
6516 @option{-isysroot} option will apply to header files.
6518 The GNU linker (beginning with version 2.16) has the necessary support
6519 for this option. If your linker does not support this option, the
6520 header file aspect of @option{--sysroot} will still work, but the
6521 library aspect will not.
6525 This option has been deprecated. Please use @option{-iquote} instead for
6526 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6527 Any directories you specify with @option{-I} options before the @option{-I-}
6528 option are searched only for the case of @samp{#include "@var{file}"};
6529 they are not searched for @samp{#include <@var{file}>}.
6531 If additional directories are specified with @option{-I} options after
6532 the @option{-I-}, these directories are searched for all @samp{#include}
6533 directives. (Ordinarily @emph{all} @option{-I} directories are used
6536 In addition, the @option{-I-} option inhibits the use of the current
6537 directory (where the current input file came from) as the first search
6538 directory for @samp{#include "@var{file}"}. There is no way to
6539 override this effect of @option{-I-}. With @option{-I.} you can specify
6540 searching the directory which was current when the compiler was
6541 invoked. That is not exactly the same as what the preprocessor does
6542 by default, but it is often satisfactory.
6544 @option{-I-} does not inhibit the use of the standard system directories
6545 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6552 @section Specifying subprocesses and the switches to pass to them
6555 @command{gcc} is a driver program. It performs its job by invoking a
6556 sequence of other programs to do the work of compiling, assembling and
6557 linking. GCC interprets its command-line parameters and uses these to
6558 deduce which programs it should invoke, and which command-line options
6559 it ought to place on their command lines. This behavior is controlled
6560 by @dfn{spec strings}. In most cases there is one spec string for each
6561 program that GCC can invoke, but a few programs have multiple spec
6562 strings to control their behavior. The spec strings built into GCC can
6563 be overridden by using the @option{-specs=} command-line switch to specify
6566 @dfn{Spec files} are plaintext files that are used to construct spec
6567 strings. They consist of a sequence of directives separated by blank
6568 lines. The type of directive is determined by the first non-whitespace
6569 character on the line and it can be one of the following:
6572 @item %@var{command}
6573 Issues a @var{command} to the spec file processor. The commands that can
6577 @item %include <@var{file}>
6579 Search for @var{file} and insert its text at the current point in the
6582 @item %include_noerr <@var{file}>
6583 @cindex %include_noerr
6584 Just like @samp{%include}, but do not generate an error message if the include
6585 file cannot be found.
6587 @item %rename @var{old_name} @var{new_name}
6589 Rename the spec string @var{old_name} to @var{new_name}.
6593 @item *[@var{spec_name}]:
6594 This tells the compiler to create, override or delete the named spec
6595 string. All lines after this directive up to the next directive or
6596 blank line are considered to be the text for the spec string. If this
6597 results in an empty string then the spec will be deleted. (Or, if the
6598 spec did not exist, then nothing will happened.) Otherwise, if the spec
6599 does not currently exist a new spec will be created. If the spec does
6600 exist then its contents will be overridden by the text of this
6601 directive, unless the first character of that text is the @samp{+}
6602 character, in which case the text will be appended to the spec.
6604 @item [@var{suffix}]:
6605 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6606 and up to the next directive or blank line are considered to make up the
6607 spec string for the indicated suffix. When the compiler encounters an
6608 input file with the named suffix, it will processes the spec string in
6609 order to work out how to compile that file. For example:
6616 This says that any input file whose name ends in @samp{.ZZ} should be
6617 passed to the program @samp{z-compile}, which should be invoked with the
6618 command-line switch @option{-input} and with the result of performing the
6619 @samp{%i} substitution. (See below.)
6621 As an alternative to providing a spec string, the text that follows a
6622 suffix directive can be one of the following:
6625 @item @@@var{language}
6626 This says that the suffix is an alias for a known @var{language}. This is
6627 similar to using the @option{-x} command-line switch to GCC to specify a
6628 language explicitly. For example:
6635 Says that .ZZ files are, in fact, C++ source files.
6638 This causes an error messages saying:
6641 @var{name} compiler not installed on this system.
6645 GCC already has an extensive list of suffixes built into it.
6646 This directive will add an entry to the end of the list of suffixes, but
6647 since the list is searched from the end backwards, it is effectively
6648 possible to override earlier entries using this technique.
6652 GCC has the following spec strings built into it. Spec files can
6653 override these strings or create their own. Note that individual
6654 targets can also add their own spec strings to this list.
6657 asm Options to pass to the assembler
6658 asm_final Options to pass to the assembler post-processor
6659 cpp Options to pass to the C preprocessor
6660 cc1 Options to pass to the C compiler
6661 cc1plus Options to pass to the C++ compiler
6662 endfile Object files to include at the end of the link
6663 link Options to pass to the linker
6664 lib Libraries to include on the command line to the linker
6665 libgcc Decides which GCC support library to pass to the linker
6666 linker Sets the name of the linker
6667 predefines Defines to be passed to the C preprocessor
6668 signed_char Defines to pass to CPP to say whether @code{char} is signed
6670 startfile Object files to include at the start of the link
6673 Here is a small example of a spec file:
6679 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6682 This example renames the spec called @samp{lib} to @samp{old_lib} and
6683 then overrides the previous definition of @samp{lib} with a new one.
6684 The new definition adds in some extra command-line options before
6685 including the text of the old definition.
6687 @dfn{Spec strings} are a list of command-line options to be passed to their
6688 corresponding program. In addition, the spec strings can contain
6689 @samp{%}-prefixed sequences to substitute variable text or to
6690 conditionally insert text into the command line. Using these constructs
6691 it is possible to generate quite complex command lines.
6693 Here is a table of all defined @samp{%}-sequences for spec
6694 strings. Note that spaces are not generated automatically around the
6695 results of expanding these sequences. Therefore you can concatenate them
6696 together or combine them with constant text in a single argument.
6700 Substitute one @samp{%} into the program name or argument.
6703 Substitute the name of the input file being processed.
6706 Substitute the basename of the input file being processed.
6707 This is the substring up to (and not including) the last period
6708 and not including the directory.
6711 This is the same as @samp{%b}, but include the file suffix (text after
6715 Marks the argument containing or following the @samp{%d} as a
6716 temporary file name, so that that file will be deleted if GCC exits
6717 successfully. Unlike @samp{%g}, this contributes no text to the
6720 @item %g@var{suffix}
6721 Substitute a file name that has suffix @var{suffix} and is chosen
6722 once per compilation, and mark the argument in the same way as
6723 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6724 name is now chosen in a way that is hard to predict even when previously
6725 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6726 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6727 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6728 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6729 was simply substituted with a file name chosen once per compilation,
6730 without regard to any appended suffix (which was therefore treated
6731 just like ordinary text), making such attacks more likely to succeed.
6733 @item %u@var{suffix}
6734 Like @samp{%g}, but generates a new temporary file name even if
6735 @samp{%u@var{suffix}} was already seen.
6737 @item %U@var{suffix}
6738 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6739 new one if there is no such last file name. In the absence of any
6740 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6741 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6742 would involve the generation of two distinct file names, one
6743 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6744 simply substituted with a file name chosen for the previous @samp{%u},
6745 without regard to any appended suffix.
6747 @item %j@var{suffix}
6748 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6749 writable, and if save-temps is off; otherwise, substitute the name
6750 of a temporary file, just like @samp{%u}. This temporary file is not
6751 meant for communication between processes, but rather as a junk
6754 @item %|@var{suffix}
6755 @itemx %m@var{suffix}
6756 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6757 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6758 all. These are the two most common ways to instruct a program that it
6759 should read from standard input or write to standard output. If you
6760 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6761 construct: see for example @file{f/lang-specs.h}.
6763 @item %.@var{SUFFIX}
6764 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6765 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6766 terminated by the next space or %.
6769 Marks the argument containing or following the @samp{%w} as the
6770 designated output file of this compilation. This puts the argument
6771 into the sequence of arguments that @samp{%o} will substitute later.
6774 Substitutes the names of all the output files, with spaces
6775 automatically placed around them. You should write spaces
6776 around the @samp{%o} as well or the results are undefined.
6777 @samp{%o} is for use in the specs for running the linker.
6778 Input files whose names have no recognized suffix are not compiled
6779 at all, but they are included among the output files, so they will
6783 Substitutes the suffix for object files. Note that this is
6784 handled specially when it immediately follows @samp{%g, %u, or %U},
6785 because of the need for those to form complete file names. The
6786 handling is such that @samp{%O} is treated exactly as if it had already
6787 been substituted, except that @samp{%g, %u, and %U} do not currently
6788 support additional @var{suffix} characters following @samp{%O} as they would
6789 following, for example, @samp{.o}.
6792 Substitutes the standard macro predefinitions for the
6793 current target machine. Use this when running @code{cpp}.
6796 Like @samp{%p}, but puts @samp{__} before and after the name of each
6797 predefined macro, except for macros that start with @samp{__} or with
6798 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6802 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6803 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6804 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6808 Current argument is the name of a library or startup file of some sort.
6809 Search for that file in a standard list of directories and substitute
6810 the full name found.
6813 Print @var{str} as an error message. @var{str} is terminated by a newline.
6814 Use this when inconsistent options are detected.
6817 Substitute the contents of spec string @var{name} at this point.
6820 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6822 @item %x@{@var{option}@}
6823 Accumulate an option for @samp{%X}.
6826 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6830 Output the accumulated assembler options specified by @option{-Wa}.
6833 Output the accumulated preprocessor options specified by @option{-Wp}.
6836 Process the @code{asm} spec. This is used to compute the
6837 switches to be passed to the assembler.
6840 Process the @code{asm_final} spec. This is a spec string for
6841 passing switches to an assembler post-processor, if such a program is
6845 Process the @code{link} spec. This is the spec for computing the
6846 command line passed to the linker. Typically it will make use of the
6847 @samp{%L %G %S %D and %E} sequences.
6850 Dump out a @option{-L} option for each directory that GCC believes might
6851 contain startup files. If the target supports multilibs then the
6852 current multilib directory will be prepended to each of these paths.
6855 Process the @code{lib} spec. This is a spec string for deciding which
6856 libraries should be included on the command line to the linker.
6859 Process the @code{libgcc} spec. This is a spec string for deciding
6860 which GCC support library should be included on the command line to the linker.
6863 Process the @code{startfile} spec. This is a spec for deciding which
6864 object files should be the first ones passed to the linker. Typically
6865 this might be a file named @file{crt0.o}.
6868 Process the @code{endfile} spec. This is a spec string that specifies
6869 the last object files that will be passed to the linker.
6872 Process the @code{cpp} spec. This is used to construct the arguments
6873 to be passed to the C preprocessor.
6876 Process the @code{cc1} spec. This is used to construct the options to be
6877 passed to the actual C compiler (@samp{cc1}).
6880 Process the @code{cc1plus} spec. This is used to construct the options to be
6881 passed to the actual C++ compiler (@samp{cc1plus}).
6884 Substitute the variable part of a matched option. See below.
6885 Note that each comma in the substituted string is replaced by
6889 Remove all occurrences of @code{-S} from the command line. Note---this
6890 command is position dependent. @samp{%} commands in the spec string
6891 before this one will see @code{-S}, @samp{%} commands in the spec string
6892 after this one will not.
6894 @item %:@var{function}(@var{args})
6895 Call the named function @var{function}, passing it @var{args}.
6896 @var{args} is first processed as a nested spec string, then split
6897 into an argument vector in the usual fashion. The function returns
6898 a string which is processed as if it had appeared literally as part
6899 of the current spec.
6901 The following built-in spec functions are provided:
6904 @item @code{if-exists}
6905 The @code{if-exists} spec function takes one argument, an absolute
6906 pathname to a file. If the file exists, @code{if-exists} returns the
6907 pathname. Here is a small example of its usage:
6911 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6914 @item @code{if-exists-else}
6915 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6916 spec function, except that it takes two arguments. The first argument is
6917 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6918 returns the pathname. If it does not exist, it returns the second argument.
6919 This way, @code{if-exists-else} can be used to select one file or another,
6920 based on the existence of the first. Here is a small example of its usage:
6924 crt0%O%s %:if-exists(crti%O%s) \
6925 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6928 @item @code{replace-outfile}
6929 The @code{replace-outfile} spec function takes two arguments. It looks for the
6930 first argument in the outfiles array and replaces it with the second argument. Here
6931 is a small example of its usage:
6934 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6940 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6941 If that switch was not specified, this substitutes nothing. Note that
6942 the leading dash is omitted when specifying this option, and it is
6943 automatically inserted if the substitution is performed. Thus the spec
6944 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6945 and would output the command line option @option{-foo}.
6947 @item %W@{@code{S}@}
6948 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6951 @item %@{@code{S}*@}
6952 Substitutes all the switches specified to GCC whose names start
6953 with @code{-S}, but which also take an argument. This is used for
6954 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6955 GCC considers @option{-o foo} as being
6956 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6957 text, including the space. Thus two arguments would be generated.
6959 @item %@{@code{S}*&@code{T}*@}
6960 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6961 (the order of @code{S} and @code{T} in the spec is not significant).
6962 There can be any number of ampersand-separated variables; for each the
6963 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6965 @item %@{@code{S}:@code{X}@}
6966 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6968 @item %@{!@code{S}:@code{X}@}
6969 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6971 @item %@{@code{S}*:@code{X}@}
6972 Substitutes @code{X} if one or more switches whose names start with
6973 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6974 once, no matter how many such switches appeared. However, if @code{%*}
6975 appears somewhere in @code{X}, then @code{X} will be substituted once
6976 for each matching switch, with the @code{%*} replaced by the part of
6977 that switch that matched the @code{*}.
6979 @item %@{.@code{S}:@code{X}@}
6980 Substitutes @code{X}, if processing a file with suffix @code{S}.
6982 @item %@{!.@code{S}:@code{X}@}
6983 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6985 @item %@{@code{S}|@code{P}:@code{X}@}
6986 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6987 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6988 although they have a stronger binding than the @samp{|}. If @code{%*}
6989 appears in @code{X}, all of the alternatives must be starred, and only
6990 the first matching alternative is substituted.
6992 For example, a spec string like this:
6995 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6998 will output the following command-line options from the following input
6999 command-line options:
7004 -d fred.c -foo -baz -boggle
7005 -d jim.d -bar -baz -boggle
7008 @item %@{S:X; T:Y; :D@}
7010 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
7011 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
7012 be as many clauses as you need. This may be combined with @code{.},
7013 @code{!}, @code{|}, and @code{*} as needed.
7018 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
7019 construct may contain other nested @samp{%} constructs or spaces, or
7020 even newlines. They are processed as usual, as described above.
7021 Trailing white space in @code{X} is ignored. White space may also
7022 appear anywhere on the left side of the colon in these constructs,
7023 except between @code{.} or @code{*} and the corresponding word.
7025 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
7026 handled specifically in these constructs. If another value of
7027 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
7028 @option{-W} switch is found later in the command line, the earlier
7029 switch value is ignored, except with @{@code{S}*@} where @code{S} is
7030 just one letter, which passes all matching options.
7032 The character @samp{|} at the beginning of the predicate text is used to
7033 indicate that a command should be piped to the following command, but
7034 only if @option{-pipe} is specified.
7036 It is built into GCC which switches take arguments and which do not.
7037 (You might think it would be useful to generalize this to allow each
7038 compiler's spec to say which switches take arguments. But this cannot
7039 be done in a consistent fashion. GCC cannot even decide which input
7040 files have been specified without knowing which switches take arguments,
7041 and it must know which input files to compile in order to tell which
7044 GCC also knows implicitly that arguments starting in @option{-l} are to be
7045 treated as compiler output files, and passed to the linker in their
7046 proper position among the other output files.
7048 @c man begin OPTIONS
7050 @node Target Options
7051 @section Specifying Target Machine and Compiler Version
7052 @cindex target options
7053 @cindex cross compiling
7054 @cindex specifying machine version
7055 @cindex specifying compiler version and target machine
7056 @cindex compiler version, specifying
7057 @cindex target machine, specifying
7059 The usual way to run GCC is to run the executable called @file{gcc}, or
7060 @file{<machine>-gcc} when cross-compiling, or
7061 @file{<machine>-gcc-<version>} to run a version other than the one that
7062 was installed last. Sometimes this is inconvenient, so GCC provides
7063 options that will switch to another cross-compiler or version.
7066 @item -b @var{machine}
7068 The argument @var{machine} specifies the target machine for compilation.
7070 The value to use for @var{machine} is the same as was specified as the
7071 machine type when configuring GCC as a cross-compiler. For
7072 example, if a cross-compiler was configured with @samp{configure
7073 arm-elf}, meaning to compile for an arm processor with elf binaries,
7074 then you would specify @option{-b arm-elf} to run that cross compiler.
7075 Because there are other options beginning with @option{-b}, the
7076 configuration must contain a hyphen.
7078 @item -V @var{version}
7080 The argument @var{version} specifies which version of GCC to run.
7081 This is useful when multiple versions are installed. For example,
7082 @var{version} might be @samp{4.0}, meaning to run GCC version 4.0.
7085 The @option{-V} and @option{-b} options work by running the
7086 @file{<machine>-gcc-<version>} executable, so there's no real reason to
7087 use them if you can just run that directly.
7089 @node Submodel Options
7090 @section Hardware Models and Configurations
7091 @cindex submodel options
7092 @cindex specifying hardware config
7093 @cindex hardware models and configurations, specifying
7094 @cindex machine dependent options
7096 Earlier we discussed the standard option @option{-b} which chooses among
7097 different installed compilers for completely different target
7098 machines, such as VAX vs.@: 68000 vs.@: 80386.
7100 In addition, each of these target machine types can have its own
7101 special options, starting with @samp{-m}, to choose among various
7102 hardware models or configurations---for example, 68010 vs 68020,
7103 floating coprocessor or none. A single installed version of the
7104 compiler can compile for any model or configuration, according to the
7107 Some configurations of the compiler also support additional special
7108 options, usually for compatibility with other compilers on the same
7111 @c This list is ordered alphanumerically by subsection name.
7112 @c It should be the same order and spelling as these options are listed
7113 @c in Machine Dependent Options
7119 * Blackfin Options::
7123 * DEC Alpha Options::
7124 * DEC Alpha/VMS Options::
7128 * i386 and x86-64 Options::
7141 * RS/6000 and PowerPC Options::
7142 * S/390 and zSeries Options::
7145 * System V Options::
7146 * TMS320C3x/C4x Options::
7150 * Xstormy16 Options::
7156 @subsection ARC Options
7159 These options are defined for ARC implementations:
7164 Compile code for little endian mode. This is the default.
7168 Compile code for big endian mode.
7171 @opindex mmangle-cpu
7172 Prepend the name of the cpu to all public symbol names.
7173 In multiple-processor systems, there are many ARC variants with different
7174 instruction and register set characteristics. This flag prevents code
7175 compiled for one cpu to be linked with code compiled for another.
7176 No facility exists for handling variants that are ``almost identical''.
7177 This is an all or nothing option.
7179 @item -mcpu=@var{cpu}
7181 Compile code for ARC variant @var{cpu}.
7182 Which variants are supported depend on the configuration.
7183 All variants support @option{-mcpu=base}, this is the default.
7185 @item -mtext=@var{text-section}
7186 @itemx -mdata=@var{data-section}
7187 @itemx -mrodata=@var{readonly-data-section}
7191 Put functions, data, and readonly data in @var{text-section},
7192 @var{data-section}, and @var{readonly-data-section} respectively
7193 by default. This can be overridden with the @code{section} attribute.
7194 @xref{Variable Attributes}.
7199 @subsection ARM Options
7202 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
7206 @item -mabi=@var{name}
7208 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
7209 @samp{atpcs}, @samp{aapcs}, @samp{aapcs-linux} and @samp{iwmmxt}.
7212 @opindex mapcs-frame
7213 Generate a stack frame that is compliant with the ARM Procedure Call
7214 Standard for all functions, even if this is not strictly necessary for
7215 correct execution of the code. Specifying @option{-fomit-frame-pointer}
7216 with this option will cause the stack frames not to be generated for
7217 leaf functions. The default is @option{-mno-apcs-frame}.
7221 This is a synonym for @option{-mapcs-frame}.
7224 @c not currently implemented
7225 @item -mapcs-stack-check
7226 @opindex mapcs-stack-check
7227 Generate code to check the amount of stack space available upon entry to
7228 every function (that actually uses some stack space). If there is
7229 insufficient space available then either the function
7230 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
7231 called, depending upon the amount of stack space required. The run time
7232 system is required to provide these functions. The default is
7233 @option{-mno-apcs-stack-check}, since this produces smaller code.
7235 @c not currently implemented
7237 @opindex mapcs-float
7238 Pass floating point arguments using the float point registers. This is
7239 one of the variants of the APCS@. This option is recommended if the
7240 target hardware has a floating point unit or if a lot of floating point
7241 arithmetic is going to be performed by the code. The default is
7242 @option{-mno-apcs-float}, since integer only code is slightly increased in
7243 size if @option{-mapcs-float} is used.
7245 @c not currently implemented
7246 @item -mapcs-reentrant
7247 @opindex mapcs-reentrant
7248 Generate reentrant, position independent code. The default is
7249 @option{-mno-apcs-reentrant}.
7252 @item -mthumb-interwork
7253 @opindex mthumb-interwork
7254 Generate code which supports calling between the ARM and Thumb
7255 instruction sets. Without this option the two instruction sets cannot
7256 be reliably used inside one program. The default is
7257 @option{-mno-thumb-interwork}, since slightly larger code is generated
7258 when @option{-mthumb-interwork} is specified.
7260 @item -mno-sched-prolog
7261 @opindex mno-sched-prolog
7262 Prevent the reordering of instructions in the function prolog, or the
7263 merging of those instruction with the instructions in the function's
7264 body. This means that all functions will start with a recognizable set
7265 of instructions (or in fact one of a choice from a small set of
7266 different function prologues), and this information can be used to
7267 locate the start if functions inside an executable piece of code. The
7268 default is @option{-msched-prolog}.
7271 @opindex mhard-float
7272 Generate output containing floating point instructions. This is the
7276 @opindex msoft-float
7277 Generate output containing library calls for floating point.
7278 @strong{Warning:} the requisite libraries are not available for all ARM
7279 targets. Normally the facilities of the machine's usual C compiler are
7280 used, but this cannot be done directly in cross-compilation. You must make
7281 your own arrangements to provide suitable library functions for
7284 @option{-msoft-float} changes the calling convention in the output file;
7285 therefore, it is only useful if you compile @emph{all} of a program with
7286 this option. In particular, you need to compile @file{libgcc.a}, the
7287 library that comes with GCC, with @option{-msoft-float} in order for
7290 @item -mfloat-abi=@var{name}
7292 Specifies which ABI to use for floating point values. Permissible values
7293 are: @samp{soft}, @samp{softfp} and @samp{hard}.
7295 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
7296 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
7297 of floating point instructions, but still uses the soft-float calling
7300 @item -mlittle-endian
7301 @opindex mlittle-endian
7302 Generate code for a processor running in little-endian mode. This is
7303 the default for all standard configurations.
7306 @opindex mbig-endian
7307 Generate code for a processor running in big-endian mode; the default is
7308 to compile code for a little-endian processor.
7310 @item -mwords-little-endian
7311 @opindex mwords-little-endian
7312 This option only applies when generating code for big-endian processors.
7313 Generate code for a little-endian word order but a big-endian byte
7314 order. That is, a byte order of the form @samp{32107654}. Note: this
7315 option should only be used if you require compatibility with code for
7316 big-endian ARM processors generated by versions of the compiler prior to
7319 @item -mcpu=@var{name}
7321 This specifies the name of the target ARM processor. GCC uses this name
7322 to determine what kind of instructions it can emit when generating
7323 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
7324 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
7325 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
7326 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
7327 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
7328 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
7329 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
7330 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
7331 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
7332 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
7333 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
7334 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
7335 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
7336 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
7339 @itemx -mtune=@var{name}
7341 This option is very similar to the @option{-mcpu=} option, except that
7342 instead of specifying the actual target processor type, and hence
7343 restricting which instructions can be used, it specifies that GCC should
7344 tune the performance of the code as if the target were of the type
7345 specified in this option, but still choosing the instructions that it
7346 will generate based on the cpu specified by a @option{-mcpu=} option.
7347 For some ARM implementations better performance can be obtained by using
7350 @item -march=@var{name}
7352 This specifies the name of the target ARM architecture. GCC uses this
7353 name to determine what kind of instructions it can emit when generating
7354 assembly code. This option can be used in conjunction with or instead
7355 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
7356 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
7357 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
7358 @samp{iwmmxt}, @samp{ep9312}.
7360 @item -mfpu=@var{name}
7361 @itemx -mfpe=@var{number}
7362 @itemx -mfp=@var{number}
7366 This specifies what floating point hardware (or hardware emulation) is
7367 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
7368 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
7369 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
7370 with older versions of GCC@.
7372 If @option{-msoft-float} is specified this specifies the format of
7373 floating point values.
7375 @item -mstructure-size-boundary=@var{n}
7376 @opindex mstructure-size-boundary
7377 The size of all structures and unions will be rounded up to a multiple
7378 of the number of bits set by this option. Permissible values are 8, 32
7379 and 64. The default value varies for different toolchains. For the COFF
7380 targeted toolchain the default value is 8. A value of 64 is only allowed
7381 if the underlying ABI supports it.
7383 Specifying the larger number can produce faster, more efficient code, but
7384 can also increase the size of the program. Different values are potentially
7385 incompatible. Code compiled with one value cannot necessarily expect to
7386 work with code or libraries compiled with another value, if they exchange
7387 information using structures or unions.
7389 @item -mabort-on-noreturn
7390 @opindex mabort-on-noreturn
7391 Generate a call to the function @code{abort} at the end of a
7392 @code{noreturn} function. It will be executed if the function tries to
7396 @itemx -mno-long-calls
7397 @opindex mlong-calls
7398 @opindex mno-long-calls
7399 Tells the compiler to perform function calls by first loading the
7400 address of the function into a register and then performing a subroutine
7401 call on this register. This switch is needed if the target function
7402 will lie outside of the 64 megabyte addressing range of the offset based
7403 version of subroutine call instruction.
7405 Even if this switch is enabled, not all function calls will be turned
7406 into long calls. The heuristic is that static functions, functions
7407 which have the @samp{short-call} attribute, functions that are inside
7408 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7409 definitions have already been compiled within the current compilation
7410 unit, will not be turned into long calls. The exception to this rule is
7411 that weak function definitions, functions with the @samp{long-call}
7412 attribute or the @samp{section} attribute, and functions that are within
7413 the scope of a @samp{#pragma long_calls} directive, will always be
7414 turned into long calls.
7416 This feature is not enabled by default. Specifying
7417 @option{-mno-long-calls} will restore the default behavior, as will
7418 placing the function calls within the scope of a @samp{#pragma
7419 long_calls_off} directive. Note these switches have no effect on how
7420 the compiler generates code to handle function calls via function
7423 @item -mnop-fun-dllimport
7424 @opindex mnop-fun-dllimport
7425 Disable support for the @code{dllimport} attribute.
7427 @item -msingle-pic-base
7428 @opindex msingle-pic-base
7429 Treat the register used for PIC addressing as read-only, rather than
7430 loading it in the prologue for each function. The run-time system is
7431 responsible for initializing this register with an appropriate value
7432 before execution begins.
7434 @item -mpic-register=@var{reg}
7435 @opindex mpic-register
7436 Specify the register to be used for PIC addressing. The default is R10
7437 unless stack-checking is enabled, when R9 is used.
7439 @item -mcirrus-fix-invalid-insns
7440 @opindex mcirrus-fix-invalid-insns
7441 @opindex mno-cirrus-fix-invalid-insns
7442 Insert NOPs into the instruction stream to in order to work around
7443 problems with invalid Maverick instruction combinations. This option
7444 is only valid if the @option{-mcpu=ep9312} option has been used to
7445 enable generation of instructions for the Cirrus Maverick floating
7446 point co-processor. This option is not enabled by default, since the
7447 problem is only present in older Maverick implementations. The default
7448 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7451 @item -mpoke-function-name
7452 @opindex mpoke-function-name
7453 Write the name of each function into the text section, directly
7454 preceding the function prologue. The generated code is similar to this:
7458 .ascii "arm_poke_function_name", 0
7461 .word 0xff000000 + (t1 - t0)
7462 arm_poke_function_name
7464 stmfd sp!, @{fp, ip, lr, pc@}
7468 When performing a stack backtrace, code can inspect the value of
7469 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7470 location @code{pc - 12} and the top 8 bits are set, then we know that
7471 there is a function name embedded immediately preceding this location
7472 and has length @code{((pc[-3]) & 0xff000000)}.
7476 Generate code for the 16-bit Thumb instruction set. The default is to
7477 use the 32-bit ARM instruction set.
7480 @opindex mtpcs-frame
7481 Generate a stack frame that is compliant with the Thumb Procedure Call
7482 Standard for all non-leaf functions. (A leaf function is one that does
7483 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7485 @item -mtpcs-leaf-frame
7486 @opindex mtpcs-leaf-frame
7487 Generate a stack frame that is compliant with the Thumb Procedure Call
7488 Standard for all leaf functions. (A leaf function is one that does
7489 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7491 @item -mcallee-super-interworking
7492 @opindex mcallee-super-interworking
7493 Gives all externally visible functions in the file being compiled an ARM
7494 instruction set header which switches to Thumb mode before executing the
7495 rest of the function. This allows these functions to be called from
7496 non-interworking code.
7498 @item -mcaller-super-interworking
7499 @opindex mcaller-super-interworking
7500 Allows calls via function pointers (including virtual functions) to
7501 execute correctly regardless of whether the target code has been
7502 compiled for interworking or not. There is a small overhead in the cost
7503 of executing a function pointer if this option is enabled.
7505 @item -mtp=@var{name}
7507 Specify the access model for the thread local storage pointer. The valid
7508 models are @option{soft}, which generates calls to @code{__aeabi_read_tp},
7509 @option{cp15}, which fetches the thread pointer from @code{cp15} directly
7510 (supported in the arm6k architecture), and @option{auto}, which uses the
7511 best available method for the selected processor. The default setting is
7517 @subsection AVR Options
7520 These options are defined for AVR implementations:
7523 @item -mmcu=@var{mcu}
7525 Specify ATMEL AVR instruction set or MCU type.
7527 Instruction set avr1 is for the minimal AVR core, not supported by the C
7528 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7529 attiny11, attiny12, attiny15, attiny28).
7531 Instruction set avr2 (default) is for the classic AVR core with up to
7532 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7533 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7534 at90c8534, at90s8535).
7536 Instruction set avr3 is for the classic AVR core with up to 128K program
7537 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7539 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7540 memory space (MCU types: atmega8, atmega83, atmega85).
7542 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7543 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7544 atmega64, atmega128, at43usb355, at94k).
7548 Output instruction sizes to the asm file.
7550 @item -minit-stack=@var{N}
7551 @opindex minit-stack
7552 Specify the initial stack address, which may be a symbol or numeric value,
7553 @samp{__stack} is the default.
7555 @item -mno-interrupts
7556 @opindex mno-interrupts
7557 Generated code is not compatible with hardware interrupts.
7558 Code size will be smaller.
7560 @item -mcall-prologues
7561 @opindex mcall-prologues
7562 Functions prologues/epilogues expanded as call to appropriate
7563 subroutines. Code size will be smaller.
7565 @item -mno-tablejump
7566 @opindex mno-tablejump
7567 Do not generate tablejump insns which sometimes increase code size.
7570 @opindex mtiny-stack
7571 Change only the low 8 bits of the stack pointer.
7575 Assume int to be 8 bit integer. This affects the sizes of all types: A
7576 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7577 and long long will be 4 bytes. Please note that this option does not
7578 comply to the C standards, but it will provide you with smaller code
7582 @node Blackfin Options
7583 @subsection Blackfin Options
7584 @cindex Blackfin Options
7587 @item -momit-leaf-frame-pointer
7588 @opindex momit-leaf-frame-pointer
7589 Don't keep the frame pointer in a register for leaf functions. This
7590 avoids the instructions to save, set up and restore frame pointers and
7591 makes an extra register available in leaf functions. The option
7592 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7593 which might make debugging harder.
7595 @item -mspecld-anomaly
7596 @opindex mspecld-anomaly
7597 When enabled, the compiler will ensure that the generated code does not
7598 contain speculative loads after jump instructions. This option is enabled
7601 @item -mno-specld-anomaly
7602 @opindex mno-specld-anomaly
7603 Don't generate extra code to prevent speculative loads from occurring.
7605 @item -mcsync-anomaly
7606 @opindex mcsync-anomaly
7607 When enabled, the compiler will ensure that the generated code does not
7608 contain CSYNC or SSYNC instructions too soon after conditional branches.
7609 This option is enabled by default.
7611 @item -mno-csync-anomaly
7612 @opindex mno-csync-anomaly
7613 Don't generate extra code to prevent CSYNC or SSYNC instructions from
7614 occurring too soon after a conditional branch.
7618 When enabled, the compiler is free to take advantage of the knowledge that
7619 the entire program fits into the low 64k of memory.
7622 @opindex mno-low-64k
7623 Assume that the program is arbitrarily large. This is the default.
7625 @item -mid-shared-library
7626 @opindex mid-shared-library
7627 Generate code that supports shared libraries via the library ID method.
7628 This allows for execute in place and shared libraries in an environment
7629 without virtual memory management. This option implies @option{-fPIC}.
7631 @item -mno-id-shared-library
7632 @opindex mno-id-shared-library
7633 Generate code that doesn't assume ID based shared libraries are being used.
7634 This is the default.
7636 @item -mshared-library-id=n
7637 @opindex mshared-library-id
7638 Specified the identification number of the ID based shared library being
7639 compiled. Specifying a value of 0 will generate more compact code, specifying
7640 other values will force the allocation of that number to the current
7641 library but is no more space or time efficient than omitting this option.
7644 @itemx -mno-long-calls
7645 @opindex mlong-calls
7646 @opindex mno-long-calls
7647 Tells the compiler to perform function calls by first loading the
7648 address of the function into a register and then performing a subroutine
7649 call on this register. This switch is needed if the target function
7650 will lie outside of the 24 bit addressing range of the offset based
7651 version of subroutine call instruction.
7653 This feature is not enabled by default. Specifying
7654 @option{-mno-long-calls} will restore the default behavior. Note these
7655 switches have no effect on how the compiler generates code to handle
7656 function calls via function pointers.
7660 @subsection CRIS Options
7661 @cindex CRIS Options
7663 These options are defined specifically for the CRIS ports.
7666 @item -march=@var{architecture-type}
7667 @itemx -mcpu=@var{architecture-type}
7670 Generate code for the specified architecture. The choices for
7671 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7672 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7673 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7676 @item -mtune=@var{architecture-type}
7678 Tune to @var{architecture-type} everything applicable about the generated
7679 code, except for the ABI and the set of available instructions. The
7680 choices for @var{architecture-type} are the same as for
7681 @option{-march=@var{architecture-type}}.
7683 @item -mmax-stack-frame=@var{n}
7684 @opindex mmax-stack-frame
7685 Warn when the stack frame of a function exceeds @var{n} bytes.
7687 @item -melinux-stacksize=@var{n}
7688 @opindex melinux-stacksize
7689 Only available with the @samp{cris-axis-aout} target. Arranges for
7690 indications in the program to the kernel loader that the stack of the
7691 program should be set to @var{n} bytes.
7697 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7698 @option{-march=v3} and @option{-march=v8} respectively.
7700 @item -mmul-bug-workaround
7701 @itemx -mno-mul-bug-workaround
7702 @opindex mmul-bug-workaround
7703 @opindex mno-mul-bug-workaround
7704 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7705 models where it applies. This option is active by default.
7709 Enable CRIS-specific verbose debug-related information in the assembly
7710 code. This option also has the effect to turn off the @samp{#NO_APP}
7711 formatted-code indicator to the assembler at the beginning of the
7716 Do not use condition-code results from previous instruction; always emit
7717 compare and test instructions before use of condition codes.
7719 @item -mno-side-effects
7720 @opindex mno-side-effects
7721 Do not emit instructions with side-effects in addressing modes other than
7725 @itemx -mno-stack-align
7727 @itemx -mno-data-align
7728 @itemx -mconst-align
7729 @itemx -mno-const-align
7730 @opindex mstack-align
7731 @opindex mno-stack-align
7732 @opindex mdata-align
7733 @opindex mno-data-align
7734 @opindex mconst-align
7735 @opindex mno-const-align
7736 These options (no-options) arranges (eliminate arrangements) for the
7737 stack-frame, individual data and constants to be aligned for the maximum
7738 single data access size for the chosen CPU model. The default is to
7739 arrange for 32-bit alignment. ABI details such as structure layout are
7740 not affected by these options.
7748 Similar to the stack- data- and const-align options above, these options
7749 arrange for stack-frame, writable data and constants to all be 32-bit,
7750 16-bit or 8-bit aligned. The default is 32-bit alignment.
7752 @item -mno-prologue-epilogue
7753 @itemx -mprologue-epilogue
7754 @opindex mno-prologue-epilogue
7755 @opindex mprologue-epilogue
7756 With @option{-mno-prologue-epilogue}, the normal function prologue and
7757 epilogue that sets up the stack-frame are omitted and no return
7758 instructions or return sequences are generated in the code. Use this
7759 option only together with visual inspection of the compiled code: no
7760 warnings or errors are generated when call-saved registers must be saved,
7761 or storage for local variable needs to be allocated.
7767 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7768 instruction sequences that load addresses for functions from the PLT part
7769 of the GOT rather than (traditional on other architectures) calls to the
7770 PLT@. The default is @option{-mgotplt}.
7774 Legacy no-op option only recognized with the cris-axis-aout target.
7778 Legacy no-op option only recognized with the cris-axis-elf and
7779 cris-axis-linux-gnu targets.
7783 Only recognized with the cris-axis-aout target, where it selects a
7784 GNU/linux-like multilib, include files and instruction set for
7789 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7793 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7794 to link with input-output functions from a simulator library. Code,
7795 initialized data and zero-initialized data are allocated consecutively.
7799 Like @option{-sim}, but pass linker options to locate initialized data at
7800 0x40000000 and zero-initialized data at 0x80000000.
7804 @subsection CRX Options
7807 These options are defined specifically for the CRX ports.
7813 Enable the use of multiply-accumulate instructions. Disabled by default.
7817 Push instructions will be used to pass outgoing arguments when functions
7818 are called. Enabled by default.
7821 @node Darwin Options
7822 @subsection Darwin Options
7823 @cindex Darwin options
7825 These options are defined for all architectures running the Darwin operating
7828 FSF GCC on Darwin does not create ``fat'' object files; it will create
7829 an object file for the single architecture that it was built to
7830 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7831 @option{-arch} options are used; it does so by running the compiler or
7832 linker multiple times and joining the results together with
7835 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7836 @samp{i686}) is determined by the flags that specify the ISA
7837 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7838 @option{-force_cpusubtype_ALL} option can be used to override this.
7840 The Darwin tools vary in their behavior when presented with an ISA
7841 mismatch. The assembler, @file{as}, will only permit instructions to
7842 be used that are valid for the subtype of the file it is generating,
7843 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7844 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7845 and print an error if asked to create a shared library with a less
7846 restrictive subtype than its input files (for instance, trying to put
7847 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7848 for executables, @file{ld}, will quietly give the executable the most
7849 restrictive subtype of any of its input files.
7854 Add the framework directory @var{dir} to the head of the list of
7855 directories to be searched for header files. These directories are
7856 interleaved with those specified by @option{-I} options and are
7857 scanned in a left-to-right order.
7859 A framework directory is a directory with frameworks in it. A
7860 framework is a directory with a @samp{"Headers"} and/or
7861 @samp{"PrivateHeaders"} directory contained directly in it that ends
7862 in @samp{".framework"}. The name of a framework is the name of this
7863 directory excluding the @samp{".framework"}. Headers associated with
7864 the framework are found in one of those two directories, with
7865 @samp{"Headers"} being searched first. A subframework is a framework
7866 directory that is in a framework's @samp{"Frameworks"} directory.
7867 Includes of subframework headers can only appear in a header of a
7868 framework that contains the subframework, or in a sibling subframework
7869 header. Two subframeworks are siblings if they occur in the same
7870 framework. A subframework should not have the same name as a
7871 framework, a warning will be issued if this is violated. Currently a
7872 subframework cannot have subframeworks, in the future, the mechanism
7873 may be extended to support this. The standard frameworks can be found
7874 in @samp{"/System/Library/Frameworks"} and
7875 @samp{"/Library/Frameworks"}. An example include looks like
7876 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7877 the name of the framework and header.h is found in the
7878 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7882 Emit debugging information for symbols that are used. For STABS
7883 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7884 This is by default ON@.
7888 Emit debugging information for all symbols and types.
7890 @item -mmacosx-version-min=@var{version}
7891 The earliest version of MacOS X that this executable will run on
7892 is @var{version}. Typical values of @var{version} include @code{10.1},
7893 @code{10.2}, and @code{10.3.9}.
7895 The default for this option is to make choices that seem to be most
7898 @item -mone-byte-bool
7899 @opindex -mone-byte-bool
7900 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7901 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7902 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7903 option has no effect on x86.
7905 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7906 to generate code that is not binary compatible with code generated
7907 without that switch. Using this switch may require recompiling all
7908 other modules in a program, including system libraries. Use this
7909 switch to conform to a non-default data model.
7911 @item -mfix-and-continue
7912 @itemx -ffix-and-continue
7913 @itemx -findirect-data
7914 @opindex mfix-and-continue
7915 @opindex ffix-and-continue
7916 @opindex findirect-data
7917 Generate code suitable for fast turn around development. Needed to
7918 enable gdb to dynamically load @code{.o} files into already running
7919 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7920 are provided for backwards compatibility.
7924 Loads all members of static archive libraries.
7925 See man ld(1) for more information.
7927 @item -arch_errors_fatal
7928 @opindex arch_errors_fatal
7929 Cause the errors having to do with files that have the wrong architecture
7933 @opindex bind_at_load
7934 Causes the output file to be marked such that the dynamic linker will
7935 bind all undefined references when the file is loaded or launched.
7939 Produce a Mach-o bundle format file.
7940 See man ld(1) for more information.
7942 @item -bundle_loader @var{executable}
7943 @opindex bundle_loader
7944 This option specifies the @var{executable} that will be loading the build
7945 output file being linked. See man ld(1) for more information.
7948 @opindex -dynamiclib
7949 When passed this option, GCC will produce a dynamic library instead of
7950 an executable when linking, using the Darwin @file{libtool} command.
7952 @item -force_cpusubtype_ALL
7953 @opindex -force_cpusubtype_ALL
7954 This causes GCC's output file to have the @var{ALL} subtype, instead of
7955 one controlled by the @option{-mcpu} or @option{-march} option.
7957 @item -allowable_client @var{client_name}
7959 @itemx -compatibility_version
7960 @itemx -current_version
7962 @itemx -dependency-file
7964 @itemx -dylinker_install_name
7966 @itemx -exported_symbols_list
7968 @itemx -flat_namespace
7969 @itemx -force_flat_namespace
7970 @itemx -headerpad_max_install_names
7973 @itemx -install_name
7974 @itemx -keep_private_externs
7975 @itemx -multi_module
7976 @itemx -multiply_defined
7977 @itemx -multiply_defined_unused
7979 @itemx -no_dead_strip_inits_and_terms
7980 @itemx -nofixprebinding
7983 @itemx -noseglinkedit
7984 @itemx -pagezero_size
7986 @itemx -prebind_all_twolevel_modules
7987 @itemx -private_bundle
7988 @itemx -read_only_relocs
7990 @itemx -sectobjectsymbols
7994 @itemx -sectobjectsymbols
7997 @itemx -segs_read_only_addr
7998 @itemx -segs_read_write_addr
7999 @itemx -seg_addr_table
8000 @itemx -seg_addr_table_filename
8003 @itemx -segs_read_only_addr
8004 @itemx -segs_read_write_addr
8005 @itemx -single_module
8008 @itemx -sub_umbrella
8009 @itemx -twolevel_namespace
8012 @itemx -unexported_symbols_list
8013 @itemx -weak_reference_mismatches
8016 @opindex allowable_client
8017 @opindex client_name
8018 @opindex compatibility_version
8019 @opindex current_version
8021 @opindex dependency-file
8023 @opindex dylinker_install_name
8025 @opindex exported_symbols_list
8027 @opindex flat_namespace
8028 @opindex force_flat_namespace
8029 @opindex headerpad_max_install_names
8032 @opindex install_name
8033 @opindex keep_private_externs
8034 @opindex multi_module
8035 @opindex multiply_defined
8036 @opindex multiply_defined_unused
8038 @opindex no_dead_strip_inits_and_terms
8039 @opindex nofixprebinding
8040 @opindex nomultidefs
8042 @opindex noseglinkedit
8043 @opindex pagezero_size
8045 @opindex prebind_all_twolevel_modules
8046 @opindex private_bundle
8047 @opindex read_only_relocs
8049 @opindex sectobjectsymbols
8053 @opindex sectobjectsymbols
8056 @opindex segs_read_only_addr
8057 @opindex segs_read_write_addr
8058 @opindex seg_addr_table
8059 @opindex seg_addr_table_filename
8060 @opindex seglinkedit
8062 @opindex segs_read_only_addr
8063 @opindex segs_read_write_addr
8064 @opindex single_module
8066 @opindex sub_library
8067 @opindex sub_umbrella
8068 @opindex twolevel_namespace
8071 @opindex unexported_symbols_list
8072 @opindex weak_reference_mismatches
8073 @opindex whatsloaded
8075 These options are passed to the Darwin linker. The Darwin linker man page
8076 describes them in detail.
8079 @node DEC Alpha Options
8080 @subsection DEC Alpha Options
8082 These @samp{-m} options are defined for the DEC Alpha implementations:
8085 @item -mno-soft-float
8087 @opindex mno-soft-float
8088 @opindex msoft-float
8089 Use (do not use) the hardware floating-point instructions for
8090 floating-point operations. When @option{-msoft-float} is specified,
8091 functions in @file{libgcc.a} will be used to perform floating-point
8092 operations. Unless they are replaced by routines that emulate the
8093 floating-point operations, or compiled in such a way as to call such
8094 emulations routines, these routines will issue floating-point
8095 operations. If you are compiling for an Alpha without floating-point
8096 operations, you must ensure that the library is built so as not to call
8099 Note that Alpha implementations without floating-point operations are
8100 required to have floating-point registers.
8105 @opindex mno-fp-regs
8106 Generate code that uses (does not use) the floating-point register set.
8107 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
8108 register set is not used, floating point operands are passed in integer
8109 registers as if they were integers and floating-point results are passed
8110 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
8111 so any function with a floating-point argument or return value called by code
8112 compiled with @option{-mno-fp-regs} must also be compiled with that
8115 A typical use of this option is building a kernel that does not use,
8116 and hence need not save and restore, any floating-point registers.
8120 The Alpha architecture implements floating-point hardware optimized for
8121 maximum performance. It is mostly compliant with the IEEE floating
8122 point standard. However, for full compliance, software assistance is
8123 required. This option generates code fully IEEE compliant code
8124 @emph{except} that the @var{inexact-flag} is not maintained (see below).
8125 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
8126 defined during compilation. The resulting code is less efficient but is
8127 able to correctly support denormalized numbers and exceptional IEEE
8128 values such as not-a-number and plus/minus infinity. Other Alpha
8129 compilers call this option @option{-ieee_with_no_inexact}.
8131 @item -mieee-with-inexact
8132 @opindex mieee-with-inexact
8133 This is like @option{-mieee} except the generated code also maintains
8134 the IEEE @var{inexact-flag}. Turning on this option causes the
8135 generated code to implement fully-compliant IEEE math. In addition to
8136 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
8137 macro. On some Alpha implementations the resulting code may execute
8138 significantly slower than the code generated by default. Since there is
8139 very little code that depends on the @var{inexact-flag}, you should
8140 normally not specify this option. Other Alpha compilers call this
8141 option @option{-ieee_with_inexact}.
8143 @item -mfp-trap-mode=@var{trap-mode}
8144 @opindex mfp-trap-mode
8145 This option controls what floating-point related traps are enabled.
8146 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
8147 The trap mode can be set to one of four values:
8151 This is the default (normal) setting. The only traps that are enabled
8152 are the ones that cannot be disabled in software (e.g., division by zero
8156 In addition to the traps enabled by @samp{n}, underflow traps are enabled
8160 Like @samp{su}, but the instructions are marked to be safe for software
8161 completion (see Alpha architecture manual for details).
8164 Like @samp{su}, but inexact traps are enabled as well.
8167 @item -mfp-rounding-mode=@var{rounding-mode}
8168 @opindex mfp-rounding-mode
8169 Selects the IEEE rounding mode. Other Alpha compilers call this option
8170 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
8175 Normal IEEE rounding mode. Floating point numbers are rounded towards
8176 the nearest machine number or towards the even machine number in case
8180 Round towards minus infinity.
8183 Chopped rounding mode. Floating point numbers are rounded towards zero.
8186 Dynamic rounding mode. A field in the floating point control register
8187 (@var{fpcr}, see Alpha architecture reference manual) controls the
8188 rounding mode in effect. The C library initializes this register for
8189 rounding towards plus infinity. Thus, unless your program modifies the
8190 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
8193 @item -mtrap-precision=@var{trap-precision}
8194 @opindex mtrap-precision
8195 In the Alpha architecture, floating point traps are imprecise. This
8196 means without software assistance it is impossible to recover from a
8197 floating trap and program execution normally needs to be terminated.
8198 GCC can generate code that can assist operating system trap handlers
8199 in determining the exact location that caused a floating point trap.
8200 Depending on the requirements of an application, different levels of
8201 precisions can be selected:
8205 Program precision. This option is the default and means a trap handler
8206 can only identify which program caused a floating point exception.
8209 Function precision. The trap handler can determine the function that
8210 caused a floating point exception.
8213 Instruction precision. The trap handler can determine the exact
8214 instruction that caused a floating point exception.
8217 Other Alpha compilers provide the equivalent options called
8218 @option{-scope_safe} and @option{-resumption_safe}.
8220 @item -mieee-conformant
8221 @opindex mieee-conformant
8222 This option marks the generated code as IEEE conformant. You must not
8223 use this option unless you also specify @option{-mtrap-precision=i} and either
8224 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
8225 is to emit the line @samp{.eflag 48} in the function prologue of the
8226 generated assembly file. Under DEC Unix, this has the effect that
8227 IEEE-conformant math library routines will be linked in.
8229 @item -mbuild-constants
8230 @opindex mbuild-constants
8231 Normally GCC examines a 32- or 64-bit integer constant to
8232 see if it can construct it from smaller constants in two or three
8233 instructions. If it cannot, it will output the constant as a literal and
8234 generate code to load it from the data segment at runtime.
8236 Use this option to require GCC to construct @emph{all} integer constants
8237 using code, even if it takes more instructions (the maximum is six).
8239 You would typically use this option to build a shared library dynamic
8240 loader. Itself a shared library, it must relocate itself in memory
8241 before it can find the variables and constants in its own data segment.
8247 Select whether to generate code to be assembled by the vendor-supplied
8248 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
8266 Indicate whether GCC should generate code to use the optional BWX,
8267 CIX, FIX and MAX instruction sets. The default is to use the instruction
8268 sets supported by the CPU type specified via @option{-mcpu=} option or that
8269 of the CPU on which GCC was built if none was specified.
8274 @opindex mfloat-ieee
8275 Generate code that uses (does not use) VAX F and G floating point
8276 arithmetic instead of IEEE single and double precision.
8278 @item -mexplicit-relocs
8279 @itemx -mno-explicit-relocs
8280 @opindex mexplicit-relocs
8281 @opindex mno-explicit-relocs
8282 Older Alpha assemblers provided no way to generate symbol relocations
8283 except via assembler macros. Use of these macros does not allow
8284 optimal instruction scheduling. GNU binutils as of version 2.12
8285 supports a new syntax that allows the compiler to explicitly mark
8286 which relocations should apply to which instructions. This option
8287 is mostly useful for debugging, as GCC detects the capabilities of
8288 the assembler when it is built and sets the default accordingly.
8292 @opindex msmall-data
8293 @opindex mlarge-data
8294 When @option{-mexplicit-relocs} is in effect, static data is
8295 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
8296 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
8297 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
8298 16-bit relocations off of the @code{$gp} register. This limits the
8299 size of the small data area to 64KB, but allows the variables to be
8300 directly accessed via a single instruction.
8302 The default is @option{-mlarge-data}. With this option the data area
8303 is limited to just below 2GB@. Programs that require more than 2GB of
8304 data must use @code{malloc} or @code{mmap} to allocate the data in the
8305 heap instead of in the program's data segment.
8307 When generating code for shared libraries, @option{-fpic} implies
8308 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
8312 @opindex msmall-text
8313 @opindex mlarge-text
8314 When @option{-msmall-text} is used, the compiler assumes that the
8315 code of the entire program (or shared library) fits in 4MB, and is
8316 thus reachable with a branch instruction. When @option{-msmall-data}
8317 is used, the compiler can assume that all local symbols share the
8318 same @code{$gp} value, and thus reduce the number of instructions
8319 required for a function call from 4 to 1.
8321 The default is @option{-mlarge-text}.
8323 @item -mcpu=@var{cpu_type}
8325 Set the instruction set and instruction scheduling parameters for
8326 machine type @var{cpu_type}. You can specify either the @samp{EV}
8327 style name or the corresponding chip number. GCC supports scheduling
8328 parameters for the EV4, EV5 and EV6 family of processors and will
8329 choose the default values for the instruction set from the processor
8330 you specify. If you do not specify a processor type, GCC will default
8331 to the processor on which the compiler was built.
8333 Supported values for @var{cpu_type} are
8339 Schedules as an EV4 and has no instruction set extensions.
8343 Schedules as an EV5 and has no instruction set extensions.
8347 Schedules as an EV5 and supports the BWX extension.
8352 Schedules as an EV5 and supports the BWX and MAX extensions.
8356 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
8360 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
8363 @item -mtune=@var{cpu_type}
8365 Set only the instruction scheduling parameters for machine type
8366 @var{cpu_type}. The instruction set is not changed.
8368 @item -mmemory-latency=@var{time}
8369 @opindex mmemory-latency
8370 Sets the latency the scheduler should assume for typical memory
8371 references as seen by the application. This number is highly
8372 dependent on the memory access patterns used by the application
8373 and the size of the external cache on the machine.
8375 Valid options for @var{time} are
8379 A decimal number representing clock cycles.
8385 The compiler contains estimates of the number of clock cycles for
8386 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
8387 (also called Dcache, Scache, and Bcache), as well as to main memory.
8388 Note that L3 is only valid for EV5.
8393 @node DEC Alpha/VMS Options
8394 @subsection DEC Alpha/VMS Options
8396 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
8399 @item -mvms-return-codes
8400 @opindex mvms-return-codes
8401 Return VMS condition codes from main. The default is to return POSIX
8402 style condition (e.g.@ error) codes.
8406 @subsection FRV Options
8413 Only use the first 32 general purpose registers.
8418 Use all 64 general purpose registers.
8423 Use only the first 32 floating point registers.
8428 Use all 64 floating point registers
8431 @opindex mhard-float
8433 Use hardware instructions for floating point operations.
8436 @opindex msoft-float
8438 Use library routines for floating point operations.
8443 Dynamically allocate condition code registers.
8448 Do not try to dynamically allocate condition code registers, only
8449 use @code{icc0} and @code{fcc0}.
8454 Change ABI to use double word insns.
8459 Do not use double word instructions.
8464 Use floating point double instructions.
8469 Do not use floating point double instructions.
8474 Use media instructions.
8479 Do not use media instructions.
8484 Use multiply and add/subtract instructions.
8489 Do not use multiply and add/subtract instructions.
8494 Select the FDPIC ABI, that uses function descriptors to represent
8495 pointers to functions. Without any PIC/PIE-related options, it
8496 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8497 assumes GOT entries and small data are within a 12-bit range from the
8498 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8499 are computed with 32 bits.
8502 @opindex minline-plt
8504 Enable inlining of PLT entries in function calls to functions that are
8505 not known to bind locally. It has no effect without @option{-mfdpic}.
8506 It's enabled by default if optimizing for speed and compiling for
8507 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8508 optimization option such as @option{-O3} or above is present in the
8514 Assume a large TLS segment when generating thread-local code.
8519 Do not assume a large TLS segment when generating thread-local code.
8524 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8525 that is known to be in read-only sections. It's enabled by default,
8526 except for @option{-fpic} or @option{-fpie}: even though it may help
8527 make the global offset table smaller, it trades 1 instruction for 4.
8528 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8529 one of which may be shared by multiple symbols, and it avoids the need
8530 for a GOT entry for the referenced symbol, so it's more likely to be a
8531 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8533 @item -multilib-library-pic
8534 @opindex multilib-library-pic
8536 Link with the (library, not FD) pic libraries. It's implied by
8537 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8538 @option{-fpic} without @option{-mfdpic}. You should never have to use
8544 Follow the EABI requirement of always creating a frame pointer whenever
8545 a stack frame is allocated. This option is enabled by default and can
8546 be disabled with @option{-mno-linked-fp}.
8549 @opindex mlong-calls
8551 Use indirect addressing to call functions outside the current
8552 compilation unit. This allows the functions to be placed anywhere
8553 within the 32-bit address space.
8555 @item -malign-labels
8556 @opindex malign-labels
8558 Try to align labels to an 8-byte boundary by inserting nops into the
8559 previous packet. This option only has an effect when VLIW packing
8560 is enabled. It doesn't create new packets; it merely adds nops to
8564 @opindex mlibrary-pic
8566 Generate position-independent EABI code.
8571 Use only the first four media accumulator registers.
8576 Use all eight media accumulator registers.
8581 Pack VLIW instructions.
8586 Do not pack VLIW instructions.
8591 Do not mark ABI switches in e_flags.
8596 Enable the use of conditional-move instructions (default).
8598 This switch is mainly for debugging the compiler and will likely be removed
8599 in a future version.
8601 @item -mno-cond-move
8602 @opindex mno-cond-move
8604 Disable the use of conditional-move instructions.
8606 This switch is mainly for debugging the compiler and will likely be removed
8607 in a future version.
8612 Enable the use of conditional set instructions (default).
8614 This switch is mainly for debugging the compiler and will likely be removed
8615 in a future version.
8620 Disable the use of conditional set instructions.
8622 This switch is mainly for debugging the compiler and will likely be removed
8623 in a future version.
8628 Enable the use of conditional execution (default).
8630 This switch is mainly for debugging the compiler and will likely be removed
8631 in a future version.
8633 @item -mno-cond-exec
8634 @opindex mno-cond-exec
8636 Disable the use of conditional execution.
8638 This switch is mainly for debugging the compiler and will likely be removed
8639 in a future version.
8642 @opindex mvliw-branch
8644 Run a pass to pack branches into VLIW instructions (default).
8646 This switch is mainly for debugging the compiler and will likely be removed
8647 in a future version.
8649 @item -mno-vliw-branch
8650 @opindex mno-vliw-branch
8652 Do not run a pass to pack branches into VLIW instructions.
8654 This switch is mainly for debugging the compiler and will likely be removed
8655 in a future version.
8657 @item -mmulti-cond-exec
8658 @opindex mmulti-cond-exec
8660 Enable optimization of @code{&&} and @code{||} in conditional execution
8663 This switch is mainly for debugging the compiler and will likely be removed
8664 in a future version.
8666 @item -mno-multi-cond-exec
8667 @opindex mno-multi-cond-exec
8669 Disable optimization of @code{&&} and @code{||} in conditional execution.
8671 This switch is mainly for debugging the compiler and will likely be removed
8672 in a future version.
8674 @item -mnested-cond-exec
8675 @opindex mnested-cond-exec
8677 Enable nested conditional execution optimizations (default).
8679 This switch is mainly for debugging the compiler and will likely be removed
8680 in a future version.
8682 @item -mno-nested-cond-exec
8683 @opindex mno-nested-cond-exec
8685 Disable nested conditional execution optimizations.
8687 This switch is mainly for debugging the compiler and will likely be removed
8688 in a future version.
8690 @item -moptimize-membar
8691 @opindex moptimize-membar
8693 This switch removes redundant @code{membar} instructions from the
8694 compiler generated code. It is enabled by default.
8696 @item -mno-optimize-membar
8697 @opindex mno-optimize-membar
8699 This switch disables the automatic removal of redundant @code{membar}
8700 instructions from the generated code.
8702 @item -mtomcat-stats
8703 @opindex mtomcat-stats
8705 Cause gas to print out tomcat statistics.
8707 @item -mcpu=@var{cpu}
8710 Select the processor type for which to generate code. Possible values are
8711 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8712 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8716 @node H8/300 Options
8717 @subsection H8/300 Options
8719 These @samp{-m} options are defined for the H8/300 implementations:
8724 Shorten some address references at link time, when possible; uses the
8725 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8726 ld, Using ld}, for a fuller description.
8730 Generate code for the H8/300H@.
8734 Generate code for the H8S@.
8738 Generate code for the H8S and H8/300H in the normal mode. This switch
8739 must be used either with @option{-mh} or @option{-ms}.
8743 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8747 Make @code{int} data 32 bits by default.
8751 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8752 The default for the H8/300H and H8S is to align longs and floats on 4
8754 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8755 This option has no effect on the H8/300.
8759 @subsection HPPA Options
8760 @cindex HPPA Options
8762 These @samp{-m} options are defined for the HPPA family of computers:
8765 @item -march=@var{architecture-type}
8767 Generate code for the specified architecture. The choices for
8768 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8769 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8770 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8771 architecture option for your machine. Code compiled for lower numbered
8772 architectures will run on higher numbered architectures, but not the
8776 @itemx -mpa-risc-1-1
8777 @itemx -mpa-risc-2-0
8778 @opindex mpa-risc-1-0
8779 @opindex mpa-risc-1-1
8780 @opindex mpa-risc-2-0
8781 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8784 @opindex mbig-switch
8785 Generate code suitable for big switch tables. Use this option only if
8786 the assembler/linker complain about out of range branches within a switch
8789 @item -mjump-in-delay
8790 @opindex mjump-in-delay
8791 Fill delay slots of function calls with unconditional jump instructions
8792 by modifying the return pointer for the function call to be the target
8793 of the conditional jump.
8795 @item -mdisable-fpregs
8796 @opindex mdisable-fpregs
8797 Prevent floating point registers from being used in any manner. This is
8798 necessary for compiling kernels which perform lazy context switching of
8799 floating point registers. If you use this option and attempt to perform
8800 floating point operations, the compiler will abort.
8802 @item -mdisable-indexing
8803 @opindex mdisable-indexing
8804 Prevent the compiler from using indexing address modes. This avoids some
8805 rather obscure problems when compiling MIG generated code under MACH@.
8807 @item -mno-space-regs
8808 @opindex mno-space-regs
8809 Generate code that assumes the target has no space registers. This allows
8810 GCC to generate faster indirect calls and use unscaled index address modes.
8812 Such code is suitable for level 0 PA systems and kernels.
8814 @item -mfast-indirect-calls
8815 @opindex mfast-indirect-calls
8816 Generate code that assumes calls never cross space boundaries. This
8817 allows GCC to emit code which performs faster indirect calls.
8819 This option will not work in the presence of shared libraries or nested
8822 @item -mfixed-range=@var{register-range}
8823 @opindex mfixed-range
8824 Generate code treating the given register range as fixed registers.
8825 A fixed register is one that the register allocator can not use. This is
8826 useful when compiling kernel code. A register range is specified as
8827 two registers separated by a dash. Multiple register ranges can be
8828 specified separated by a comma.
8830 @item -mlong-load-store
8831 @opindex mlong-load-store
8832 Generate 3-instruction load and store sequences as sometimes required by
8833 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8836 @item -mportable-runtime
8837 @opindex mportable-runtime
8838 Use the portable calling conventions proposed by HP for ELF systems.
8842 Enable the use of assembler directives only GAS understands.
8844 @item -mschedule=@var{cpu-type}
8846 Schedule code according to the constraints for the machine type
8847 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8848 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8849 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8850 proper scheduling option for your machine. The default scheduling is
8854 @opindex mlinker-opt
8855 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8856 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8857 linkers in which they give bogus error messages when linking some programs.
8860 @opindex msoft-float
8861 Generate output containing library calls for floating point.
8862 @strong{Warning:} the requisite libraries are not available for all HPPA
8863 targets. Normally the facilities of the machine's usual C compiler are
8864 used, but this cannot be done directly in cross-compilation. You must make
8865 your own arrangements to provide suitable library functions for
8866 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8867 does provide software floating point support.
8869 @option{-msoft-float} changes the calling convention in the output file;
8870 therefore, it is only useful if you compile @emph{all} of a program with
8871 this option. In particular, you need to compile @file{libgcc.a}, the
8872 library that comes with GCC, with @option{-msoft-float} in order for
8877 Generate the predefine, @code{_SIO}, for server IO@. The default is
8878 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8879 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8880 options are available under HP-UX and HI-UX@.
8884 Use GNU ld specific options. This passes @option{-shared} to ld when
8885 building a shared library. It is the default when GCC is configured,
8886 explicitly or implicitly, with the GNU linker. This option does not
8887 have any affect on which ld is called, it only changes what parameters
8888 are passed to that ld. The ld that is called is determined by the
8889 @option{--with-ld} configure option, GCC's program search path, and
8890 finally by the user's @env{PATH}. The linker used by GCC can be printed
8891 using @samp{which `gcc -print-prog-name=ld`}. This option is only available
8892 on the 64 bit HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8896 Use HP ld specific options. This passes @option{-b} to ld when building
8897 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8898 links. It is the default when GCC is configured, explicitly or
8899 implicitly, with the HP linker. This option does not have any affect on
8900 which ld is called, it only changes what parameters are passed to that
8901 ld. The ld that is called is determined by the @option{--with-ld}
8902 configure option, GCC's program search path, and finally by the user's
8903 @env{PATH}. The linker used by GCC can be printed using @samp{which
8904 `gcc -print-prog-name=ld`}. This option is only available on the 64 bit
8905 HP-UX GCC, i.e. configured with @samp{hppa*64*-*-hpux*}.
8908 @opindex mno-long-calls
8909 Generate code that uses long call sequences. This ensures that a call
8910 is always able to reach linker generated stubs. The default is to generate
8911 long calls only when the distance from the call site to the beginning
8912 of the function or translation unit, as the case may be, exceeds a
8913 predefined limit set by the branch type being used. The limits for
8914 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8915 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8918 Distances are measured from the beginning of functions when using the
8919 @option{-ffunction-sections} option, or when using the @option{-mgas}
8920 and @option{-mno-portable-runtime} options together under HP-UX with
8923 It is normally not desirable to use this option as it will degrade
8924 performance. However, it may be useful in large applications,
8925 particularly when partial linking is used to build the application.
8927 The types of long calls used depends on the capabilities of the
8928 assembler and linker, and the type of code being generated. The
8929 impact on systems that support long absolute calls, and long pic
8930 symbol-difference or pc-relative calls should be relatively small.
8931 However, an indirect call is used on 32-bit ELF systems in pic code
8932 and it is quite long.
8934 @item -munix=@var{unix-std}
8936 Generate compiler predefines and select a startfile for the specified
8937 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8938 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8939 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8940 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8941 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8944 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8945 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8946 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8947 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8948 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8949 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8951 It is @emph{important} to note that this option changes the interfaces
8952 for various library routines. It also affects the operational behavior
8953 of the C library. Thus, @emph{extreme} care is needed in using this
8956 Library code that is intended to operate with more than one UNIX
8957 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8958 as appropriate. Most GNU software doesn't provide this capability.
8962 Suppress the generation of link options to search libdld.sl when the
8963 @option{-static} option is specified on HP-UX 10 and later.
8967 The HP-UX implementation of setlocale in libc has a dependency on
8968 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8969 when the @option{-static} option is specified, special link options
8970 are needed to resolve this dependency.
8972 On HP-UX 10 and later, the GCC driver adds the necessary options to
8973 link with libdld.sl when the @option{-static} option is specified.
8974 This causes the resulting binary to be dynamic. On the 64-bit port,
8975 the linkers generate dynamic binaries by default in any case. The
8976 @option{-nolibdld} option can be used to prevent the GCC driver from
8977 adding these link options.
8981 Add support for multithreading with the @dfn{dce thread} library
8982 under HP-UX@. This option sets flags for both the preprocessor and
8986 @node i386 and x86-64 Options
8987 @subsection Intel 386 and AMD x86-64 Options
8988 @cindex i386 Options
8989 @cindex x86-64 Options
8990 @cindex Intel 386 Options
8991 @cindex AMD x86-64 Options
8993 These @samp{-m} options are defined for the i386 and x86-64 family of
8997 @item -mtune=@var{cpu-type}
8999 Tune to @var{cpu-type} everything applicable about the generated code, except
9000 for the ABI and the set of available instructions. The choices for
9004 Original Intel's i386 CPU@.
9006 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
9008 Intel Pentium CPU with no MMX support.
9010 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
9011 @item i686, pentiumpro
9012 Intel PentiumPro CPU@.
9014 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
9015 @item pentium3, pentium3m
9016 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
9019 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
9020 support. Used by Centrino notebooks.
9021 @item pentium4, pentium4m
9022 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
9024 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
9027 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
9028 SSE2 and SSE3 instruction set support.
9030 AMD K6 CPU with MMX instruction set support.
9032 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
9033 @item athlon, athlon-tbird
9034 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
9036 @item athlon-4, athlon-xp, athlon-mp
9037 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
9038 instruction set support.
9039 @item k8, opteron, athlon64, athlon-fx
9040 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
9041 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
9043 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
9046 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
9047 instruction set support.
9049 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
9050 implemented for this chip.)
9052 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
9053 implemented for this chip.)
9056 While picking a specific @var{cpu-type} will schedule things appropriately
9057 for that particular chip, the compiler will not generate any code that
9058 does not run on the i386 without the @option{-march=@var{cpu-type}} option
9061 @item -march=@var{cpu-type}
9063 Generate instructions for the machine type @var{cpu-type}. The choices
9064 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
9065 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
9067 @item -mcpu=@var{cpu-type}
9069 A deprecated synonym for @option{-mtune}.
9078 @opindex mpentiumpro
9079 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
9080 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
9081 These synonyms are deprecated.
9083 @item -mfpmath=@var{unit}
9085 Generate floating point arithmetics for selected unit @var{unit}. The choices
9090 Use the standard 387 floating point coprocessor present majority of chips and
9091 emulated otherwise. Code compiled with this option will run almost everywhere.
9092 The temporary results are computed in 80bit precision instead of precision
9093 specified by the type resulting in slightly different results compared to most
9094 of other chips. See @option{-ffloat-store} for more detailed description.
9096 This is the default choice for i386 compiler.
9099 Use scalar floating point instructions present in the SSE instruction set.
9100 This instruction set is supported by Pentium3 and newer chips, in the AMD line
9101 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
9102 instruction set supports only single precision arithmetics, thus the double and
9103 extended precision arithmetics is still done using 387. Later version, present
9104 only in Pentium4 and the future AMD x86-64 chips supports double precision
9107 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
9108 or @option{-msse2} switches to enable SSE extensions and make this option
9109 effective. For the x86-64 compiler, these extensions are enabled by default.
9111 The resulting code should be considerably faster in the majority of cases and avoid
9112 the numerical instability problems of 387 code, but may break some existing
9113 code that expects temporaries to be 80bit.
9115 This is the default choice for the x86-64 compiler.
9118 Attempt to utilize both instruction sets at once. This effectively double the
9119 amount of available registers and on chips with separate execution units for
9120 387 and SSE the execution resources too. Use this option with care, as it is
9121 still experimental, because the GCC register allocator does not model separate
9122 functional units well resulting in instable performance.
9125 @item -masm=@var{dialect}
9126 @opindex masm=@var{dialect}
9127 Output asm instructions using selected @var{dialect}. Supported
9128 choices are @samp{intel} or @samp{att} (the default one). Darwin does
9129 not support @samp{intel}.
9134 @opindex mno-ieee-fp
9135 Control whether or not the compiler uses IEEE floating point
9136 comparisons. These handle correctly the case where the result of a
9137 comparison is unordered.
9140 @opindex msoft-float
9141 Generate output containing library calls for floating point.
9142 @strong{Warning:} the requisite libraries are not part of GCC@.
9143 Normally the facilities of the machine's usual C compiler are used, but
9144 this can't be done directly in cross-compilation. You must make your
9145 own arrangements to provide suitable library functions for
9148 On machines where a function returns floating point results in the 80387
9149 register stack, some floating point opcodes may be emitted even if
9150 @option{-msoft-float} is used.
9152 @item -mno-fp-ret-in-387
9153 @opindex mno-fp-ret-in-387
9154 Do not use the FPU registers for return values of functions.
9156 The usual calling convention has functions return values of types
9157 @code{float} and @code{double} in an FPU register, even if there
9158 is no FPU@. The idea is that the operating system should emulate
9161 The option @option{-mno-fp-ret-in-387} causes such values to be returned
9162 in ordinary CPU registers instead.
9164 @item -mno-fancy-math-387
9165 @opindex mno-fancy-math-387
9166 Some 387 emulators do not support the @code{sin}, @code{cos} and
9167 @code{sqrt} instructions for the 387. Specify this option to avoid
9168 generating those instructions. This option is the default on FreeBSD,
9169 OpenBSD and NetBSD@. This option is overridden when @option{-march}
9170 indicates that the target cpu will always have an FPU and so the
9171 instruction will not need emulation. As of revision 2.6.1, these
9172 instructions are not generated unless you also use the
9173 @option{-funsafe-math-optimizations} switch.
9175 @item -malign-double
9176 @itemx -mno-align-double
9177 @opindex malign-double
9178 @opindex mno-align-double
9179 Control whether GCC aligns @code{double}, @code{long double}, and
9180 @code{long long} variables on a two word boundary or a one word
9181 boundary. Aligning @code{double} variables on a two word boundary will
9182 produce code that runs somewhat faster on a @samp{Pentium} at the
9183 expense of more memory.
9185 @strong{Warning:} if you use the @option{-malign-double} switch,
9186 structures containing the above types will be aligned differently than
9187 the published application binary interface specifications for the 386
9188 and will not be binary compatible with structures in code compiled
9189 without that switch.
9191 @item -m96bit-long-double
9192 @itemx -m128bit-long-double
9193 @opindex m96bit-long-double
9194 @opindex m128bit-long-double
9195 These switches control the size of @code{long double} type. The i386
9196 application binary interface specifies the size to be 96 bits,
9197 so @option{-m96bit-long-double} is the default in 32 bit mode.
9199 Modern architectures (Pentium and newer) would prefer @code{long double}
9200 to be aligned to an 8 or 16 byte boundary. In arrays or structures
9201 conforming to the ABI, this would not be possible. So specifying a
9202 @option{-m128bit-long-double} will align @code{long double}
9203 to a 16 byte boundary by padding the @code{long double} with an additional
9206 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
9207 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
9209 Notice that neither of these options enable any extra precision over the x87
9210 standard of 80 bits for a @code{long double}.
9212 @strong{Warning:} if you override the default value for your target ABI, the
9213 structures and arrays containing @code{long double} variables will change
9214 their size as well as function calling convention for function taking
9215 @code{long double} will be modified. Hence they will not be binary
9216 compatible with arrays or structures in code compiled without that switch.
9218 @item -mmlarge-data-threshold=@var{number}
9219 @opindex mlarge-data-threshold=@var{number}
9220 When @option{-mcmodel=medium} is specified, the data greater than
9221 @var{threshold} are placed in large data section. This value must be the
9222 same across all object linked into the binary and defaults to 65535.
9225 @itemx -mno-svr3-shlib
9226 @opindex msvr3-shlib
9227 @opindex mno-svr3-shlib
9228 Control whether GCC places uninitialized local variables into the
9229 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
9230 into @code{bss}. These options are meaningful only on System V Release 3.
9234 Use a different function-calling convention, in which functions that
9235 take a fixed number of arguments return with the @code{ret} @var{num}
9236 instruction, which pops their arguments while returning. This saves one
9237 instruction in the caller since there is no need to pop the arguments
9240 You can specify that an individual function is called with this calling
9241 sequence with the function attribute @samp{stdcall}. You can also
9242 override the @option{-mrtd} option by using the function attribute
9243 @samp{cdecl}. @xref{Function Attributes}.
9245 @strong{Warning:} this calling convention is incompatible with the one
9246 normally used on Unix, so you cannot use it if you need to call
9247 libraries compiled with the Unix compiler.
9249 Also, you must provide function prototypes for all functions that
9250 take variable numbers of arguments (including @code{printf});
9251 otherwise incorrect code will be generated for calls to those
9254 In addition, seriously incorrect code will result if you call a
9255 function with too many arguments. (Normally, extra arguments are
9256 harmlessly ignored.)
9258 @item -mregparm=@var{num}
9260 Control how many registers are used to pass integer arguments. By
9261 default, no registers are used to pass arguments, and at most 3
9262 registers can be used. You can control this behavior for a specific
9263 function by using the function attribute @samp{regparm}.
9264 @xref{Function Attributes}.
9266 @strong{Warning:} if you use this switch, and
9267 @var{num} is nonzero, then you must build all modules with the same
9268 value, including any libraries. This includes the system libraries and
9272 @opindex msseregparm
9273 Use SSE register passing conventions for float and double arguments
9274 and return values. You can control this behavior for a specific
9275 function by using the function attribute @samp{sseregparm}.
9276 @xref{Function Attributes}.
9278 @strong{Warning:} if you use this switch then you must build all
9279 modules with the same value, including any libraries. This includes
9280 the system libraries and startup modules.
9282 @item -mpreferred-stack-boundary=@var{num}
9283 @opindex mpreferred-stack-boundary
9284 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
9285 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
9286 the default is 4 (16 bytes or 128 bits), except when optimizing for code
9287 size (@option{-Os}), in which case the default is the minimum correct
9288 alignment (4 bytes for x86, and 8 bytes for x86-64).
9290 On Pentium and PentiumPro, @code{double} and @code{long double} values
9291 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
9292 suffer significant run time performance penalties. On Pentium III, the
9293 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
9294 penalties if it is not 16 byte aligned.
9296 To ensure proper alignment of this values on the stack, the stack boundary
9297 must be as aligned as that required by any value stored on the stack.
9298 Further, every function must be generated such that it keeps the stack
9299 aligned. Thus calling a function compiled with a higher preferred
9300 stack boundary from a function compiled with a lower preferred stack
9301 boundary will most likely misalign the stack. It is recommended that
9302 libraries that use callbacks always use the default setting.
9304 This extra alignment does consume extra stack space, and generally
9305 increases code size. Code that is sensitive to stack space usage, such
9306 as embedded systems and operating system kernels, may want to reduce the
9307 preferred alignment to @option{-mpreferred-stack-boundary=2}.
9325 These switches enable or disable the use of instructions in the MMX,
9326 SSE, SSE2 or 3DNow! extended instruction sets. These extensions are
9327 also available as built-in functions: see @ref{X86 Built-in Functions},
9328 for details of the functions enabled and disabled by these switches.
9330 To have SSE/SSE2 instructions generated automatically from floating-point
9331 code (as opposed to 387 instructions), see @option{-mfpmath=sse}.
9333 These options will enable GCC to use these extended instructions in
9334 generated code, even without @option{-mfpmath=sse}. Applications which
9335 perform runtime CPU detection must compile separate files for each
9336 supported architecture, using the appropriate flags. In particular,
9337 the file containing the CPU detection code should be compiled without
9341 @itemx -mno-push-args
9343 @opindex mno-push-args
9344 Use PUSH operations to store outgoing parameters. This method is shorter
9345 and usually equally fast as method using SUB/MOV operations and is enabled
9346 by default. In some cases disabling it may improve performance because of
9347 improved scheduling and reduced dependencies.
9349 @item -maccumulate-outgoing-args
9350 @opindex maccumulate-outgoing-args
9351 If enabled, the maximum amount of space required for outgoing arguments will be
9352 computed in the function prologue. This is faster on most modern CPUs
9353 because of reduced dependencies, improved scheduling and reduced stack usage
9354 when preferred stack boundary is not equal to 2. The drawback is a notable
9355 increase in code size. This switch implies @option{-mno-push-args}.
9359 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
9360 on thread-safe exception handling must compile and link all code with the
9361 @option{-mthreads} option. When compiling, @option{-mthreads} defines
9362 @option{-D_MT}; when linking, it links in a special thread helper library
9363 @option{-lmingwthrd} which cleans up per thread exception handling data.
9365 @item -mno-align-stringops
9366 @opindex mno-align-stringops
9367 Do not align destination of inlined string operations. This switch reduces
9368 code size and improves performance in case the destination is already aligned,
9369 but GCC doesn't know about it.
9371 @item -minline-all-stringops
9372 @opindex minline-all-stringops
9373 By default GCC inlines string operations only when destination is known to be
9374 aligned at least to 4 byte boundary. This enables more inlining, increase code
9375 size, but may improve performance of code that depends on fast memcpy, strlen
9376 and memset for short lengths.
9378 @item -momit-leaf-frame-pointer
9379 @opindex momit-leaf-frame-pointer
9380 Don't keep the frame pointer in a register for leaf functions. This
9381 avoids the instructions to save, set up and restore frame pointers and
9382 makes an extra register available in leaf functions. The option
9383 @option{-fomit-frame-pointer} removes the frame pointer for all functions
9384 which might make debugging harder.
9386 @item -mtls-direct-seg-refs
9387 @itemx -mno-tls-direct-seg-refs
9388 @opindex mtls-direct-seg-refs
9389 Controls whether TLS variables may be accessed with offsets from the
9390 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
9391 or whether the thread base pointer must be added. Whether or not this
9392 is legal depends on the operating system, and whether it maps the
9393 segment to cover the entire TLS area.
9395 For systems that use GNU libc, the default is on.
9398 These @samp{-m} switches are supported in addition to the above
9399 on AMD x86-64 processors in 64-bit environments.
9406 Generate code for a 32-bit or 64-bit environment.
9407 The 32-bit environment sets int, long and pointer to 32 bits and
9408 generates code that runs on any i386 system.
9409 The 64-bit environment sets int to 32 bits and long and pointer
9410 to 64 bits and generates code for AMD's x86-64 architecture.
9413 @opindex no-red-zone
9414 Do not use a so called red zone for x86-64 code. The red zone is mandated
9415 by the x86-64 ABI, it is a 128-byte area beyond the location of the
9416 stack pointer that will not be modified by signal or interrupt handlers
9417 and therefore can be used for temporary data without adjusting the stack
9418 pointer. The flag @option{-mno-red-zone} disables this red zone.
9420 @item -mcmodel=small
9421 @opindex mcmodel=small
9422 Generate code for the small code model: the program and its symbols must
9423 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
9424 Programs can be statically or dynamically linked. This is the default
9427 @item -mcmodel=kernel
9428 @opindex mcmodel=kernel
9429 Generate code for the kernel code model. The kernel runs in the
9430 negative 2 GB of the address space.
9431 This model has to be used for Linux kernel code.
9433 @item -mcmodel=medium
9434 @opindex mcmodel=medium
9435 Generate code for the medium model: The program is linked in the lower 2
9436 GB of the address space but symbols can be located anywhere in the
9437 address space. Programs can be statically or dynamically linked, but
9438 building of shared libraries are not supported with the medium model.
9440 @item -mcmodel=large
9441 @opindex mcmodel=large
9442 Generate code for the large model: This model makes no assumptions
9443 about addresses and sizes of sections. Currently GCC does not implement
9448 @subsection IA-64 Options
9449 @cindex IA-64 Options
9451 These are the @samp{-m} options defined for the Intel IA-64 architecture.
9455 @opindex mbig-endian
9456 Generate code for a big endian target. This is the default for HP-UX@.
9458 @item -mlittle-endian
9459 @opindex mlittle-endian
9460 Generate code for a little endian target. This is the default for AIX5
9467 Generate (or don't) code for the GNU assembler. This is the default.
9468 @c Also, this is the default if the configure option @option{--with-gnu-as}
9475 Generate (or don't) code for the GNU linker. This is the default.
9476 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9481 Generate code that does not use a global pointer register. The result
9482 is not position independent code, and violates the IA-64 ABI@.
9484 @item -mvolatile-asm-stop
9485 @itemx -mno-volatile-asm-stop
9486 @opindex mvolatile-asm-stop
9487 @opindex mno-volatile-asm-stop
9488 Generate (or don't) a stop bit immediately before and after volatile asm
9491 @item -mregister-names
9492 @itemx -mno-register-names
9493 @opindex mregister-names
9494 @opindex mno-register-names
9495 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9496 the stacked registers. This may make assembler output more readable.
9502 Disable (or enable) optimizations that use the small data section. This may
9503 be useful for working around optimizer bugs.
9506 @opindex mconstant-gp
9507 Generate code that uses a single constant global pointer value. This is
9508 useful when compiling kernel code.
9512 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9513 This is useful when compiling firmware code.
9515 @item -minline-float-divide-min-latency
9516 @opindex minline-float-divide-min-latency
9517 Generate code for inline divides of floating point values
9518 using the minimum latency algorithm.
9520 @item -minline-float-divide-max-throughput
9521 @opindex minline-float-divide-max-throughput
9522 Generate code for inline divides of floating point values
9523 using the maximum throughput algorithm.
9525 @item -minline-int-divide-min-latency
9526 @opindex minline-int-divide-min-latency
9527 Generate code for inline divides of integer values
9528 using the minimum latency algorithm.
9530 @item -minline-int-divide-max-throughput
9531 @opindex minline-int-divide-max-throughput
9532 Generate code for inline divides of integer values
9533 using the maximum throughput algorithm.
9535 @item -minline-sqrt-min-latency
9536 @opindex minline-sqrt-min-latency
9537 Generate code for inline square roots
9538 using the minimum latency algorithm.
9540 @item -minline-sqrt-max-throughput
9541 @opindex minline-sqrt-max-throughput
9542 Generate code for inline square roots
9543 using the maximum throughput algorithm.
9545 @item -mno-dwarf2-asm
9547 @opindex mno-dwarf2-asm
9548 @opindex mdwarf2-asm
9549 Don't (or do) generate assembler code for the DWARF2 line number debugging
9550 info. This may be useful when not using the GNU assembler.
9552 @item -mearly-stop-bits
9553 @itemx -mno-early-stop-bits
9554 @opindex mearly-stop-bits
9555 @opindex mno-early-stop-bits
9556 Allow stop bits to be placed earlier than immediately preceding the
9557 instruction that triggered the stop bit. This can improve instruction
9558 scheduling, but does not always do so.
9560 @item -mfixed-range=@var{register-range}
9561 @opindex mfixed-range
9562 Generate code treating the given register range as fixed registers.
9563 A fixed register is one that the register allocator can not use. This is
9564 useful when compiling kernel code. A register range is specified as
9565 two registers separated by a dash. Multiple register ranges can be
9566 specified separated by a comma.
9568 @item -mtls-size=@var{tls-size}
9570 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9573 @item -mtune=@var{cpu-type}
9575 Tune the instruction scheduling for a particular CPU, Valid values are
9576 itanium, itanium1, merced, itanium2, and mckinley.
9582 Add support for multithreading using the POSIX threads library. This
9583 option sets flags for both the preprocessor and linker. It does
9584 not affect the thread safety of object code produced by the compiler or
9585 that of libraries supplied with it. These are HP-UX specific flags.
9591 Generate code for a 32-bit or 64-bit environment.
9592 The 32-bit environment sets int, long and pointer to 32 bits.
9593 The 64-bit environment sets int to 32 bits and long and pointer
9594 to 64 bits. These are HP-UX specific flags.
9599 @subsection M32C Options
9600 @cindex M32C options
9603 @item -mcpu=@var{name}
9605 Select the CPU for which code is generated. @var{name} may be one of
9606 @samp{r8c} for the R8C/Tiny series, @samp{m16c} for the M16C (up to
9607 /60) series, @samp{m32cm} for the M16C/80 series, or @samp{m32c} for
9612 Specifies that the program will be run on the simulator. This causes
9613 an alternate runtime library to be linked in which supports, for
9614 example, file I/O. You must not use this option when generating
9615 programs that will run on real hardware; you must provide your own
9616 runtime library for whatever I/O functions are needed.
9618 @item -memregs=@var{number}
9620 Specifies the number of memory-based pseudo-registers GCC will use
9621 during code generation. These pseudo-registers will be used like real
9622 registers, so there is a tradeoff between GCC's ability to fit the
9623 code into available registers, and the performance penalty of using
9624 memory instead of registers. Note that all modules in a program must
9625 be compiled with the same value for this option. Because of that, you
9626 must not use this option with the default runtime libraries gcc
9631 @node M32R/D Options
9632 @subsection M32R/D Options
9633 @cindex M32R/D options
9635 These @option{-m} options are defined for Renesas M32R/D architectures:
9640 Generate code for the M32R/2@.
9644 Generate code for the M32R/X@.
9648 Generate code for the M32R@. This is the default.
9651 @opindex mmodel=small
9652 Assume all objects live in the lower 16MB of memory (so that their addresses
9653 can be loaded with the @code{ld24} instruction), and assume all subroutines
9654 are reachable with the @code{bl} instruction.
9655 This is the default.
9657 The addressability of a particular object can be set with the
9658 @code{model} attribute.
9660 @item -mmodel=medium
9661 @opindex mmodel=medium
9662 Assume objects may be anywhere in the 32-bit address space (the compiler
9663 will generate @code{seth/add3} instructions to load their addresses), and
9664 assume all subroutines are reachable with the @code{bl} instruction.
9667 @opindex mmodel=large
9668 Assume objects may be anywhere in the 32-bit address space (the compiler
9669 will generate @code{seth/add3} instructions to load their addresses), and
9670 assume subroutines may not be reachable with the @code{bl} instruction
9671 (the compiler will generate the much slower @code{seth/add3/jl}
9672 instruction sequence).
9675 @opindex msdata=none
9676 Disable use of the small data area. Variables will be put into
9677 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9678 @code{section} attribute has been specified).
9679 This is the default.
9681 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9682 Objects may be explicitly put in the small data area with the
9683 @code{section} attribute using one of these sections.
9686 @opindex msdata=sdata
9687 Put small global and static data in the small data area, but do not
9688 generate special code to reference them.
9692 Put small global and static data in the small data area, and generate
9693 special instructions to reference them.
9697 @cindex smaller data references
9698 Put global and static objects less than or equal to @var{num} bytes
9699 into the small data or bss sections instead of the normal data or bss
9700 sections. The default value of @var{num} is 8.
9701 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9702 for this option to have any effect.
9704 All modules should be compiled with the same @option{-G @var{num}} value.
9705 Compiling with different values of @var{num} may or may not work; if it
9706 doesn't the linker will give an error message---incorrect code will not be
9711 Makes the M32R specific code in the compiler display some statistics
9712 that might help in debugging programs.
9715 @opindex malign-loops
9716 Align all loops to a 32-byte boundary.
9718 @item -mno-align-loops
9719 @opindex mno-align-loops
9720 Do not enforce a 32-byte alignment for loops. This is the default.
9722 @item -missue-rate=@var{number}
9723 @opindex missue-rate=@var{number}
9724 Issue @var{number} instructions per cycle. @var{number} can only be 1
9727 @item -mbranch-cost=@var{number}
9728 @opindex mbranch-cost=@var{number}
9729 @var{number} can only be 1 or 2. If it is 1 then branches will be
9730 preferred over conditional code, if it is 2, then the opposite will
9733 @item -mflush-trap=@var{number}
9734 @opindex mflush-trap=@var{number}
9735 Specifies the trap number to use to flush the cache. The default is
9736 12. Valid numbers are between 0 and 15 inclusive.
9738 @item -mno-flush-trap
9739 @opindex mno-flush-trap
9740 Specifies that the cache cannot be flushed by using a trap.
9742 @item -mflush-func=@var{name}
9743 @opindex mflush-func=@var{name}
9744 Specifies the name of the operating system function to call to flush
9745 the cache. The default is @emph{_flush_cache}, but a function call
9746 will only be used if a trap is not available.
9748 @item -mno-flush-func
9749 @opindex mno-flush-func
9750 Indicates that there is no OS function for flushing the cache.
9754 @node M680x0 Options
9755 @subsection M680x0 Options
9756 @cindex M680x0 options
9758 These are the @samp{-m} options defined for the 68000 series. The default
9759 values for these options depends on which style of 68000 was selected when
9760 the compiler was configured; the defaults for the most common choices are
9768 Generate output for a 68000. This is the default
9769 when the compiler is configured for 68000-based systems.
9771 Use this option for microcontrollers with a 68000 or EC000 core,
9772 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9778 Generate output for a 68020. This is the default
9779 when the compiler is configured for 68020-based systems.
9783 Generate output containing 68881 instructions for floating point.
9784 This is the default for most 68020 systems unless @option{--nfp} was
9785 specified when the compiler was configured.
9789 Generate output for a 68030. This is the default when the compiler is
9790 configured for 68030-based systems.
9794 Generate output for a 68040. This is the default when the compiler is
9795 configured for 68040-based systems.
9797 This option inhibits the use of 68881/68882 instructions that have to be
9798 emulated by software on the 68040. Use this option if your 68040 does not
9799 have code to emulate those instructions.
9803 Generate output for a 68060. This is the default when the compiler is
9804 configured for 68060-based systems.
9806 This option inhibits the use of 68020 and 68881/68882 instructions that
9807 have to be emulated by software on the 68060. Use this option if your 68060
9808 does not have code to emulate those instructions.
9812 Generate output for a CPU32. This is the default
9813 when the compiler is configured for CPU32-based systems.
9815 Use this option for microcontrollers with a
9816 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9817 68336, 68340, 68341, 68349 and 68360.
9821 Generate output for a 520X ``coldfire'' family cpu. This is the default
9822 when the compiler is configured for 520X-based systems.
9824 Use this option for microcontroller with a 5200 core, including
9825 the MCF5202, MCF5203, MCF5204 and MCF5202.
9830 Generate output for a 68040, without using any of the new instructions.
9831 This results in code which can run relatively efficiently on either a
9832 68020/68881 or a 68030 or a 68040. The generated code does use the
9833 68881 instructions that are emulated on the 68040.
9837 Generate output for a 68060, without using any of the new instructions.
9838 This results in code which can run relatively efficiently on either a
9839 68020/68881 or a 68030 or a 68040. The generated code does use the
9840 68881 instructions that are emulated on the 68060.
9843 @opindex msoft-float
9844 Generate output containing library calls for floating point.
9845 @strong{Warning:} the requisite libraries are not available for all m68k
9846 targets. Normally the facilities of the machine's usual C compiler are
9847 used, but this can't be done directly in cross-compilation. You must
9848 make your own arrangements to provide suitable library functions for
9849 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9850 @samp{m68k-*-coff} do provide software floating point support.
9854 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9855 Additionally, parameters passed on the stack are also aligned to a
9856 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9859 @opindex mnobitfield
9860 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9861 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9865 Do use the bit-field instructions. The @option{-m68020} option implies
9866 @option{-mbitfield}. This is the default if you use a configuration
9867 designed for a 68020.
9871 Use a different function-calling convention, in which functions
9872 that take a fixed number of arguments return with the @code{rtd}
9873 instruction, which pops their arguments while returning. This
9874 saves one instruction in the caller since there is no need to pop
9875 the arguments there.
9877 This calling convention is incompatible with the one normally
9878 used on Unix, so you cannot use it if you need to call libraries
9879 compiled with the Unix compiler.
9881 Also, you must provide function prototypes for all functions that
9882 take variable numbers of arguments (including @code{printf});
9883 otherwise incorrect code will be generated for calls to those
9886 In addition, seriously incorrect code will result if you call a
9887 function with too many arguments. (Normally, extra arguments are
9888 harmlessly ignored.)
9890 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9891 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9894 @itemx -mno-align-int
9896 @opindex mno-align-int
9897 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9898 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9899 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9900 Aligning variables on 32-bit boundaries produces code that runs somewhat
9901 faster on processors with 32-bit busses at the expense of more memory.
9903 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9904 align structures containing the above types differently than
9905 most published application binary interface specifications for the m68k.
9909 Use the pc-relative addressing mode of the 68000 directly, instead of
9910 using a global offset table. At present, this option implies @option{-fpic},
9911 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9912 not presently supported with @option{-mpcrel}, though this could be supported for
9913 68020 and higher processors.
9915 @item -mno-strict-align
9916 @itemx -mstrict-align
9917 @opindex mno-strict-align
9918 @opindex mstrict-align
9919 Do not (do) assume that unaligned memory references will be handled by
9923 Generate code that allows the data segment to be located in a different
9924 area of memory from the text segment. This allows for execute in place in
9925 an environment without virtual memory management. This option implies
9929 Generate code that assumes that the data segment follows the text segment.
9930 This is the default.
9932 @item -mid-shared-library
9933 Generate code that supports shared libraries via the library ID method.
9934 This allows for execute in place and shared libraries in an environment
9935 without virtual memory management. This option implies @option{-fPIC}.
9937 @item -mno-id-shared-library
9938 Generate code that doesn't assume ID based shared libraries are being used.
9939 This is the default.
9941 @item -mshared-library-id=n
9942 Specified the identification number of the ID based shared library being
9943 compiled. Specifying a value of 0 will generate more compact code, specifying
9944 other values will force the allocation of that number to the current
9945 library but is no more space or time efficient than omitting this option.
9949 @node M68hc1x Options
9950 @subsection M68hc1x Options
9951 @cindex M68hc1x options
9953 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9954 microcontrollers. The default values for these options depends on
9955 which style of microcontroller was selected when the compiler was configured;
9956 the defaults for the most common choices are given below.
9963 Generate output for a 68HC11. This is the default
9964 when the compiler is configured for 68HC11-based systems.
9970 Generate output for a 68HC12. This is the default
9971 when the compiler is configured for 68HC12-based systems.
9977 Generate output for a 68HCS12.
9980 @opindex mauto-incdec
9981 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9988 Enable the use of 68HC12 min and max instructions.
9991 @itemx -mno-long-calls
9992 @opindex mlong-calls
9993 @opindex mno-long-calls
9994 Treat all calls as being far away (near). If calls are assumed to be
9995 far away, the compiler will use the @code{call} instruction to
9996 call a function and the @code{rtc} instruction for returning.
10000 Consider type @code{int} to be 16 bits wide, like @code{short int}.
10002 @item -msoft-reg-count=@var{count}
10003 @opindex msoft-reg-count
10004 Specify the number of pseudo-soft registers which are used for the
10005 code generation. The maximum number is 32. Using more pseudo-soft
10006 register may or may not result in better code depending on the program.
10007 The default is 4 for 68HC11 and 2 for 68HC12.
10011 @node MCore Options
10012 @subsection MCore Options
10013 @cindex MCore options
10015 These are the @samp{-m} options defined for the Motorola M*Core
10021 @itemx -mno-hardlit
10023 @opindex mno-hardlit
10024 Inline constants into the code stream if it can be done in two
10025 instructions or less.
10031 Use the divide instruction. (Enabled by default).
10033 @item -mrelax-immediate
10034 @itemx -mno-relax-immediate
10035 @opindex mrelax-immediate
10036 @opindex mno-relax-immediate
10037 Allow arbitrary sized immediates in bit operations.
10039 @item -mwide-bitfields
10040 @itemx -mno-wide-bitfields
10041 @opindex mwide-bitfields
10042 @opindex mno-wide-bitfields
10043 Always treat bit-fields as int-sized.
10045 @item -m4byte-functions
10046 @itemx -mno-4byte-functions
10047 @opindex m4byte-functions
10048 @opindex mno-4byte-functions
10049 Force all functions to be aligned to a four byte boundary.
10051 @item -mcallgraph-data
10052 @itemx -mno-callgraph-data
10053 @opindex mcallgraph-data
10054 @opindex mno-callgraph-data
10055 Emit callgraph information.
10058 @itemx -mno-slow-bytes
10059 @opindex mslow-bytes
10060 @opindex mno-slow-bytes
10061 Prefer word access when reading byte quantities.
10063 @item -mlittle-endian
10064 @itemx -mbig-endian
10065 @opindex mlittle-endian
10066 @opindex mbig-endian
10067 Generate code for a little endian target.
10073 Generate code for the 210 processor.
10077 @subsection MIPS Options
10078 @cindex MIPS options
10084 Generate big-endian code.
10088 Generate little-endian code. This is the default for @samp{mips*el-*-*}
10091 @item -march=@var{arch}
10093 Generate code that will run on @var{arch}, which can be the name of a
10094 generic MIPS ISA, or the name of a particular processor.
10096 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
10097 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
10098 The processor names are:
10099 @samp{4kc}, @samp{4km}, @samp{4kp},
10100 @samp{5kc}, @samp{5kf},
10102 @samp{24k}, @samp{24kc}, @samp{24kf}, @samp{24kx},
10105 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
10106 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000},
10107 @samp{rm7000}, @samp{rm9000},
10110 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
10111 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
10112 The special value @samp{from-abi} selects the
10113 most compatible architecture for the selected ABI (that is,
10114 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
10116 In processor names, a final @samp{000} can be abbreviated as @samp{k}
10117 (for example, @samp{-march=r2k}). Prefixes are optional, and
10118 @samp{vr} may be written @samp{r}.
10120 GCC defines two macros based on the value of this option. The first
10121 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
10122 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
10123 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
10124 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
10125 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
10127 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
10128 above. In other words, it will have the full prefix and will not
10129 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
10130 the macro names the resolved architecture (either @samp{"mips1"} or
10131 @samp{"mips3"}). It names the default architecture when no
10132 @option{-march} option is given.
10134 @item -mtune=@var{arch}
10136 Optimize for @var{arch}. Among other things, this option controls
10137 the way instructions are scheduled, and the perceived cost of arithmetic
10138 operations. The list of @var{arch} values is the same as for
10141 When this option is not used, GCC will optimize for the processor
10142 specified by @option{-march}. By using @option{-march} and
10143 @option{-mtune} together, it is possible to generate code that will
10144 run on a family of processors, but optimize the code for one
10145 particular member of that family.
10147 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
10148 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
10149 @samp{-march} ones described above.
10153 Equivalent to @samp{-march=mips1}.
10157 Equivalent to @samp{-march=mips2}.
10161 Equivalent to @samp{-march=mips3}.
10165 Equivalent to @samp{-march=mips4}.
10169 Equivalent to @samp{-march=mips32}.
10173 Equivalent to @samp{-march=mips32r2}.
10177 Equivalent to @samp{-march=mips64}.
10182 @opindex mno-mips16
10183 Generate (do not generate) MIPS16 code. If GCC is targetting a
10184 MIPS32 or MIPS64 architecture, it will make use of the MIPS16e ASE@.
10196 Generate code for the given ABI@.
10198 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
10199 generates 64-bit code when you select a 64-bit architecture, but you
10200 can use @option{-mgp32} to get 32-bit code instead.
10202 For information about the O64 ABI, see
10203 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
10206 @itemx -mno-abicalls
10208 @opindex mno-abicalls
10209 Generate (do not generate) SVR4-style position-independent code.
10210 @option{-mabicalls} is the default for SVR4-based systems.
10216 Lift (do not lift) the usual restrictions on the size of the global
10219 GCC normally uses a single instruction to load values from the GOT@.
10220 While this is relatively efficient, it will only work if the GOT
10221 is smaller than about 64k. Anything larger will cause the linker
10222 to report an error such as:
10224 @cindex relocation truncated to fit (MIPS)
10226 relocation truncated to fit: R_MIPS_GOT16 foobar
10229 If this happens, you should recompile your code with @option{-mxgot}.
10230 It should then work with very large GOTs, although it will also be
10231 less efficient, since it will take three instructions to fetch the
10232 value of a global symbol.
10234 Note that some linkers can create multiple GOTs. If you have such a
10235 linker, you should only need to use @option{-mxgot} when a single object
10236 file accesses more than 64k's worth of GOT entries. Very few do.
10238 These options have no effect unless GCC is generating position
10243 Assume that general-purpose registers are 32 bits wide.
10247 Assume that general-purpose registers are 64 bits wide.
10251 Assume that floating-point registers are 32 bits wide.
10255 Assume that floating-point registers are 64 bits wide.
10258 @opindex mhard-float
10259 Use floating-point coprocessor instructions.
10262 @opindex msoft-float
10263 Do not use floating-point coprocessor instructions. Implement
10264 floating-point calculations using library calls instead.
10266 @item -msingle-float
10267 @opindex msingle-float
10268 Assume that the floating-point coprocessor only supports single-precision
10271 @itemx -mdouble-float
10272 @opindex mdouble-float
10273 Assume that the floating-point coprocessor supports double-precision
10274 operations. This is the default.
10280 Use (do not use) the MIPS DSP ASE. @xref{MIPS DSP Built-in Functions}.
10282 @itemx -mpaired-single
10283 @itemx -mno-paired-single
10284 @opindex mpaired-single
10285 @opindex mno-paired-single
10286 Use (do not use) paired-single floating-point instructions.
10287 @xref{MIPS Paired-Single Support}. This option can only be used
10288 when generating 64-bit code and requires hardware floating-point
10289 support to be enabled.
10294 @opindex mno-mips3d
10295 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
10296 The option @option{-mips3d} implies @option{-mpaired-single}.
10300 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
10301 an explanation of the default and the way that the pointer size is
10306 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
10308 The default size of @code{int}s, @code{long}s and pointers depends on
10309 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
10310 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
10311 32-bit @code{long}s. Pointers are the same size as @code{long}s,
10312 or the same size as integer registers, whichever is smaller.
10318 Assume (do not assume) that all symbols have 32-bit values, regardless
10319 of the selected ABI@. This option is useful in combination with
10320 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
10321 to generate shorter and faster references to symbolic addresses.
10325 @cindex smaller data references (MIPS)
10326 @cindex gp-relative references (MIPS)
10327 Put global and static items less than or equal to @var{num} bytes into
10328 the small data or bss section instead of the normal data or bss section.
10329 This allows the data to be accessed using a single instruction.
10331 All modules should be compiled with the same @option{-G @var{num}}
10334 @item -membedded-data
10335 @itemx -mno-embedded-data
10336 @opindex membedded-data
10337 @opindex mno-embedded-data
10338 Allocate variables to the read-only data section first if possible, then
10339 next in the small data section if possible, otherwise in data. This gives
10340 slightly slower code than the default, but reduces the amount of RAM required
10341 when executing, and thus may be preferred for some embedded systems.
10343 @item -muninit-const-in-rodata
10344 @itemx -mno-uninit-const-in-rodata
10345 @opindex muninit-const-in-rodata
10346 @opindex mno-uninit-const-in-rodata
10347 Put uninitialized @code{const} variables in the read-only data section.
10348 This option is only meaningful in conjunction with @option{-membedded-data}.
10350 @item -msplit-addresses
10351 @itemx -mno-split-addresses
10352 @opindex msplit-addresses
10353 @opindex mno-split-addresses
10354 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
10355 relocation operators. This option has been superseded by
10356 @option{-mexplicit-relocs} but is retained for backwards compatibility.
10358 @item -mexplicit-relocs
10359 @itemx -mno-explicit-relocs
10360 @opindex mexplicit-relocs
10361 @opindex mno-explicit-relocs
10362 Use (do not use) assembler relocation operators when dealing with symbolic
10363 addresses. The alternative, selected by @option{-mno-explicit-relocs},
10364 is to use assembler macros instead.
10366 @option{-mexplicit-relocs} is the default if GCC was configured
10367 to use an assembler that supports relocation operators.
10369 @item -mcheck-zero-division
10370 @itemx -mno-check-zero-division
10371 @opindex mcheck-zero-division
10372 @opindex mno-check-zero-division
10373 Trap (do not trap) on integer division by zero. The default is
10374 @option{-mcheck-zero-division}.
10376 @item -mdivide-traps
10377 @itemx -mdivide-breaks
10378 @opindex mdivide-traps
10379 @opindex mdivide-breaks
10380 MIPS systems check for division by zero by generating either a
10381 conditional trap or a break instruction. Using traps results in
10382 smaller code, but is only supported on MIPS II and later. Also, some
10383 versions of the Linux kernel have a bug that prevents trap from
10384 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
10385 allow conditional traps on architectures that support them and
10386 @option{-mdivide-breaks} to force the use of breaks.
10388 The default is usually @option{-mdivide-traps}, but this can be
10389 overridden at configure time using @option{--with-divide=breaks}.
10390 Divide-by-zero checks can be completely disabled using
10391 @option{-mno-check-zero-division}.
10396 @opindex mno-memcpy
10397 Force (do not force) the use of @code{memcpy()} for non-trivial block
10398 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
10399 most constant-sized copies.
10402 @itemx -mno-long-calls
10403 @opindex mlong-calls
10404 @opindex mno-long-calls
10405 Disable (do not disable) use of the @code{jal} instruction. Calling
10406 functions using @code{jal} is more efficient but requires the caller
10407 and callee to be in the same 256 megabyte segment.
10409 This option has no effect on abicalls code. The default is
10410 @option{-mno-long-calls}.
10416 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
10417 instructions, as provided by the R4650 ISA@.
10420 @itemx -mno-fused-madd
10421 @opindex mfused-madd
10422 @opindex mno-fused-madd
10423 Enable (disable) use of the floating point multiply-accumulate
10424 instructions, when they are available. The default is
10425 @option{-mfused-madd}.
10427 When multiply-accumulate instructions are used, the intermediate
10428 product is calculated to infinite precision and is not subject to
10429 the FCSR Flush to Zero bit. This may be undesirable in some
10434 Tell the MIPS assembler to not run its preprocessor over user
10435 assembler files (with a @samp{.s} suffix) when assembling them.
10438 @itemx -mno-fix-r4000
10439 @opindex mfix-r4000
10440 @opindex mno-fix-r4000
10441 Work around certain R4000 CPU errata:
10444 A double-word or a variable shift may give an incorrect result if executed
10445 immediately after starting an integer division.
10447 A double-word or a variable shift may give an incorrect result if executed
10448 while an integer multiplication is in progress.
10450 An integer division may give an incorrect result if started in a delay slot
10451 of a taken branch or a jump.
10455 @itemx -mno-fix-r4400
10456 @opindex mfix-r4400
10457 @opindex mno-fix-r4400
10458 Work around certain R4400 CPU errata:
10461 A double-word or a variable shift may give an incorrect result if executed
10462 immediately after starting an integer division.
10466 @itemx -mno-fix-vr4120
10467 @opindex mfix-vr4120
10468 Work around certain VR4120 errata:
10471 @code{dmultu} does not always produce the correct result.
10473 @code{div} and @code{ddiv} do not always produce the correct result if one
10474 of the operands is negative.
10476 The workarounds for the division errata rely on special functions in
10477 @file{libgcc.a}. At present, these functions are only provided by
10478 the @code{mips64vr*-elf} configurations.
10480 Other VR4120 errata require a nop to be inserted between certain pairs of
10481 instructions. These errata are handled by the assembler, not by GCC itself.
10484 @opindex mfix-vr4130
10485 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
10486 workarounds are implemented by the assembler rather than by GCC,
10487 although GCC will avoid using @code{mflo} and @code{mfhi} if the
10488 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
10489 instructions are available instead.
10492 @itemx -mno-fix-sb1
10494 Work around certain SB-1 CPU core errata.
10495 (This flag currently works around the SB-1 revision 2
10496 ``F1'' and ``F2'' floating point errata.)
10498 @item -mflush-func=@var{func}
10499 @itemx -mno-flush-func
10500 @opindex mflush-func
10501 Specifies the function to call to flush the I and D caches, or to not
10502 call any such function. If called, the function must take the same
10503 arguments as the common @code{_flush_func()}, that is, the address of the
10504 memory range for which the cache is being flushed, the size of the
10505 memory range, and the number 3 (to flush both caches). The default
10506 depends on the target GCC was configured for, but commonly is either
10507 @samp{_flush_func} or @samp{__cpu_flush}.
10509 @item -mbranch-likely
10510 @itemx -mno-branch-likely
10511 @opindex mbranch-likely
10512 @opindex mno-branch-likely
10513 Enable or disable use of Branch Likely instructions, regardless of the
10514 default for the selected architecture. By default, Branch Likely
10515 instructions may be generated if they are supported by the selected
10516 architecture. An exception is for the MIPS32 and MIPS64 architectures
10517 and processors which implement those architectures; for those, Branch
10518 Likely instructions will not be generated by default because the MIPS32
10519 and MIPS64 architectures specifically deprecate their use.
10521 @item -mfp-exceptions
10522 @itemx -mno-fp-exceptions
10523 @opindex mfp-exceptions
10524 Specifies whether FP exceptions are enabled. This affects how we schedule
10525 FP instructions for some processors. The default is that FP exceptions are
10528 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10529 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10532 @item -mvr4130-align
10533 @itemx -mno-vr4130-align
10534 @opindex mvr4130-align
10535 The VR4130 pipeline is two-way superscalar, but can only issue two
10536 instructions together if the first one is 8-byte aligned. When this
10537 option is enabled, GCC will align pairs of instructions that it
10538 thinks should execute in parallel.
10540 This option only has an effect when optimizing for the VR4130.
10541 It normally makes code faster, but at the expense of making it bigger.
10542 It is enabled by default at optimization level @option{-O3}.
10546 @subsection MMIX Options
10547 @cindex MMIX Options
10549 These options are defined for the MMIX:
10553 @itemx -mno-libfuncs
10555 @opindex mno-libfuncs
10556 Specify that intrinsic library functions are being compiled, passing all
10557 values in registers, no matter the size.
10560 @itemx -mno-epsilon
10562 @opindex mno-epsilon
10563 Generate floating-point comparison instructions that compare with respect
10564 to the @code{rE} epsilon register.
10566 @item -mabi=mmixware
10568 @opindex mabi-mmixware
10570 Generate code that passes function parameters and return values that (in
10571 the called function) are seen as registers @code{$0} and up, as opposed to
10572 the GNU ABI which uses global registers @code{$231} and up.
10574 @item -mzero-extend
10575 @itemx -mno-zero-extend
10576 @opindex mzero-extend
10577 @opindex mno-zero-extend
10578 When reading data from memory in sizes shorter than 64 bits, use (do not
10579 use) zero-extending load instructions by default, rather than
10580 sign-extending ones.
10583 @itemx -mno-knuthdiv
10585 @opindex mno-knuthdiv
10586 Make the result of a division yielding a remainder have the same sign as
10587 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10588 remainder follows the sign of the dividend. Both methods are
10589 arithmetically valid, the latter being almost exclusively used.
10591 @item -mtoplevel-symbols
10592 @itemx -mno-toplevel-symbols
10593 @opindex mtoplevel-symbols
10594 @opindex mno-toplevel-symbols
10595 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10596 code can be used with the @code{PREFIX} assembly directive.
10600 Generate an executable in the ELF format, rather than the default
10601 @samp{mmo} format used by the @command{mmix} simulator.
10603 @item -mbranch-predict
10604 @itemx -mno-branch-predict
10605 @opindex mbranch-predict
10606 @opindex mno-branch-predict
10607 Use (do not use) the probable-branch instructions, when static branch
10608 prediction indicates a probable branch.
10610 @item -mbase-addresses
10611 @itemx -mno-base-addresses
10612 @opindex mbase-addresses
10613 @opindex mno-base-addresses
10614 Generate (do not generate) code that uses @emph{base addresses}. Using a
10615 base address automatically generates a request (handled by the assembler
10616 and the linker) for a constant to be set up in a global register. The
10617 register is used for one or more base address requests within the range 0
10618 to 255 from the value held in the register. The generally leads to short
10619 and fast code, but the number of different data items that can be
10620 addressed is limited. This means that a program that uses lots of static
10621 data may require @option{-mno-base-addresses}.
10623 @item -msingle-exit
10624 @itemx -mno-single-exit
10625 @opindex msingle-exit
10626 @opindex mno-single-exit
10627 Force (do not force) generated code to have a single exit point in each
10631 @node MN10300 Options
10632 @subsection MN10300 Options
10633 @cindex MN10300 options
10635 These @option{-m} options are defined for Matsushita MN10300 architectures:
10640 Generate code to avoid bugs in the multiply instructions for the MN10300
10641 processors. This is the default.
10643 @item -mno-mult-bug
10644 @opindex mno-mult-bug
10645 Do not generate code to avoid bugs in the multiply instructions for the
10646 MN10300 processors.
10650 Generate code which uses features specific to the AM33 processor.
10654 Do not generate code which uses features specific to the AM33 processor. This
10657 @item -mreturn-pointer-on-d0
10658 @opindex mreturn-pointer-on-d0
10659 When generating a function which returns a pointer, return the pointer
10660 in both @code{a0} and @code{d0}. Otherwise, the pointer is returned
10661 only in a0, and attempts to call such functions without a prototype
10662 would result in errors. Note that this option is on by default; use
10663 @option{-mno-return-pointer-on-d0} to disable it.
10667 Do not link in the C run-time initialization object file.
10671 Indicate to the linker that it should perform a relaxation optimization pass
10672 to shorten branches, calls and absolute memory addresses. This option only
10673 has an effect when used on the command line for the final link step.
10675 This option makes symbolic debugging impossible.
10679 @subsection MT Options
10682 These @option{-m} options are defined for Morpho MT architectures:
10686 @item -march=@var{cpu-type}
10688 Generate code that will run on @var{cpu-type}, which is the name of a system
10689 representing a certain processor type. Possible values for
10690 @var{cpu-type} are @samp{ms1-64-001}, @samp{ms1-16-002},
10691 @samp{ms1-16-003} and @samp{ms2}.
10693 When this option is not used, the default is @option{-march=ms1-16-002}.
10697 Use byte loads and stores when generating code.
10701 Do not use byte loads and stores when generating code.
10705 Use simulator runtime
10709 Do not link in the C run-time initialization object file
10710 @file{crti.o}. Other run-time initialization and termination files
10711 such as @file{startup.o} and @file{exit.o} are still included on the
10712 linker command line.
10716 @node PDP-11 Options
10717 @subsection PDP-11 Options
10718 @cindex PDP-11 Options
10720 These options are defined for the PDP-11:
10725 Use hardware FPP floating point. This is the default. (FIS floating
10726 point on the PDP-11/40 is not supported.)
10729 @opindex msoft-float
10730 Do not use hardware floating point.
10734 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10738 Return floating-point results in memory. This is the default.
10742 Generate code for a PDP-11/40.
10746 Generate code for a PDP-11/45. This is the default.
10750 Generate code for a PDP-11/10.
10752 @item -mbcopy-builtin
10753 @opindex bcopy-builtin
10754 Use inline @code{movmemhi} patterns for copying memory. This is the
10759 Do not use inline @code{movmemhi} patterns for copying memory.
10765 Use 16-bit @code{int}. This is the default.
10771 Use 32-bit @code{int}.
10774 @itemx -mno-float32
10776 @opindex mno-float32
10777 Use 64-bit @code{float}. This is the default.
10780 @itemx -mno-float64
10782 @opindex mno-float64
10783 Use 32-bit @code{float}.
10787 Use @code{abshi2} pattern. This is the default.
10791 Do not use @code{abshi2} pattern.
10793 @item -mbranch-expensive
10794 @opindex mbranch-expensive
10795 Pretend that branches are expensive. This is for experimenting with
10796 code generation only.
10798 @item -mbranch-cheap
10799 @opindex mbranch-cheap
10800 Do not pretend that branches are expensive. This is the default.
10804 Generate code for a system with split I&D@.
10808 Generate code for a system without split I&D@. This is the default.
10812 Use Unix assembler syntax. This is the default when configured for
10813 @samp{pdp11-*-bsd}.
10817 Use DEC assembler syntax. This is the default when configured for any
10818 PDP-11 target other than @samp{pdp11-*-bsd}.
10821 @node PowerPC Options
10822 @subsection PowerPC Options
10823 @cindex PowerPC options
10825 These are listed under @xref{RS/6000 and PowerPC Options}.
10827 @node RS/6000 and PowerPC Options
10828 @subsection IBM RS/6000 and PowerPC Options
10829 @cindex RS/6000 and PowerPC Options
10830 @cindex IBM RS/6000 and PowerPC Options
10832 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10839 @itemx -mno-powerpc
10840 @itemx -mpowerpc-gpopt
10841 @itemx -mno-powerpc-gpopt
10842 @itemx -mpowerpc-gfxopt
10843 @itemx -mno-powerpc-gfxopt
10845 @itemx -mno-powerpc64
10849 @itemx -mno-popcntb
10855 @opindex mno-power2
10857 @opindex mno-powerpc
10858 @opindex mpowerpc-gpopt
10859 @opindex mno-powerpc-gpopt
10860 @opindex mpowerpc-gfxopt
10861 @opindex mno-powerpc-gfxopt
10862 @opindex mpowerpc64
10863 @opindex mno-powerpc64
10867 @opindex mno-popcntb
10870 GCC supports two related instruction set architectures for the
10871 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10872 instructions supported by the @samp{rios} chip set used in the original
10873 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10874 architecture of the Freescale MPC5xx, MPC6xx, MPC8xx microprocessors, and
10875 the IBM 4xx, 6xx, and follow-on microprocessors.
10877 Neither architecture is a subset of the other. However there is a
10878 large common subset of instructions supported by both. An MQ
10879 register is included in processors supporting the POWER architecture.
10881 You use these options to specify which instructions are available on the
10882 processor you are using. The default value of these options is
10883 determined when configuring GCC@. Specifying the
10884 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10885 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10886 rather than the options listed above.
10888 The @option{-mpower} option allows GCC to generate instructions that
10889 are found only in the POWER architecture and to use the MQ register.
10890 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10891 to generate instructions that are present in the POWER2 architecture but
10892 not the original POWER architecture.
10894 The @option{-mpowerpc} option allows GCC to generate instructions that
10895 are found only in the 32-bit subset of the PowerPC architecture.
10896 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10897 GCC to use the optional PowerPC architecture instructions in the
10898 General Purpose group, including floating-point square root. Specifying
10899 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10900 use the optional PowerPC architecture instructions in the Graphics
10901 group, including floating-point select.
10903 The @option{-mmfcrf} option allows GCC to generate the move from
10904 condition register field instruction implemented on the POWER4
10905 processor and other processors that support the PowerPC V2.01
10907 The @option{-mpopcntb} option allows GCC to generate the popcount and
10908 double precision FP reciprocal estimate instruction implemented on the
10909 POWER5 processor and other processors that support the PowerPC V2.02
10911 The @option{-mfprnd} option allows GCC to generate the FP round to
10912 integer instructions implemented on the POWER5+ processor and other
10913 processors that support the PowerPC V2.03 architecture.
10915 The @option{-mpowerpc64} option allows GCC to generate the additional
10916 64-bit instructions that are found in the full PowerPC64 architecture
10917 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10918 @option{-mno-powerpc64}.
10920 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10921 will use only the instructions in the common subset of both
10922 architectures plus some special AIX common-mode calls, and will not use
10923 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10924 permits GCC to use any instruction from either architecture and to
10925 allow use of the MQ register; specify this for the Motorola MPC601.
10927 @item -mnew-mnemonics
10928 @itemx -mold-mnemonics
10929 @opindex mnew-mnemonics
10930 @opindex mold-mnemonics
10931 Select which mnemonics to use in the generated assembler code. With
10932 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10933 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10934 assembler mnemonics defined for the POWER architecture. Instructions
10935 defined in only one architecture have only one mnemonic; GCC uses that
10936 mnemonic irrespective of which of these options is specified.
10938 GCC defaults to the mnemonics appropriate for the architecture in
10939 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10940 value of these option. Unless you are building a cross-compiler, you
10941 should normally not specify either @option{-mnew-mnemonics} or
10942 @option{-mold-mnemonics}, but should instead accept the default.
10944 @item -mcpu=@var{cpu_type}
10946 Set architecture type, register usage, choice of mnemonics, and
10947 instruction scheduling parameters for machine type @var{cpu_type}.
10948 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10949 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10950 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10951 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10952 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10953 @samp{860}, @samp{970}, @samp{8540}, @samp{ec603e}, @samp{G3},
10954 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10955 @samp{power4}, @samp{power5}, @samp{power5+},
10956 @samp{common}, @samp{powerpc}, @samp{powerpc64},
10957 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64}.
10959 @option{-mcpu=common} selects a completely generic processor. Code
10960 generated under this option will run on any POWER or PowerPC processor.
10961 GCC will use only the instructions in the common subset of both
10962 architectures, and will not use the MQ register. GCC assumes a generic
10963 processor model for scheduling purposes.
10965 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10966 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10967 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10968 types, with an appropriate, generic processor model assumed for
10969 scheduling purposes.
10971 The other options specify a specific processor. Code generated under
10972 those options will run best on that processor, and may not run at all on
10975 The @option{-mcpu} options automatically enable or disable the
10976 following options: @option{-maltivec}, @option{-mfprnd},
10977 @option{-mhard-float}, @option{-mmfcrf}, @option{-mmultiple},
10978 @option{-mnew-mnemonics}, @option{-mpopcntb}, @option{-mpower},
10979 @option{-mpower2}, @option{-mpowerpc64}, @option{-mpowerpc-gpopt},
10980 @option{-mpowerpc-gfxopt}, @option{-mstring}. The particular options
10981 set for any particular CPU will vary between compiler versions,
10982 depending on what setting seems to produce optimal code for that CPU;
10983 it doesn't necessarily reflect the actual hardware's capabilities. If
10984 you wish to set an individual option to a particular value, you may
10985 specify it after the @option{-mcpu} option, like @samp{-mcpu=970
10988 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10989 not enabled or disabled by the @option{-mcpu} option at present because
10990 AIX does not have full support for these options. You may still
10991 enable or disable them individually if you're sure it'll work in your
10994 @item -mtune=@var{cpu_type}
10996 Set the instruction scheduling parameters for machine type
10997 @var{cpu_type}, but do not set the architecture type, register usage, or
10998 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10999 values for @var{cpu_type} are used for @option{-mtune} as for
11000 @option{-mcpu}. If both are specified, the code generated will use the
11001 architecture, registers, and mnemonics set by @option{-mcpu}, but the
11002 scheduling parameters set by @option{-mtune}.
11008 Generate code to compute division as reciprocal estimate and iterative
11009 refinement, creating opportunities for increased throughput. This
11010 feature requires: optional PowerPC Graphics instruction set for single
11011 precision and FRE instruction for double precision, assuming divides
11012 cannot generate user-visible traps, and the domain values not include
11013 Infinities, denormals or zero denominator.
11016 @itemx -mno-altivec
11018 @opindex mno-altivec
11019 Generate code that uses (does not use) AltiVec instructions, and also
11020 enable the use of built-in functions that allow more direct access to
11021 the AltiVec instruction set. You may also need to set
11022 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
11028 @opindex mno-vrsave
11029 Generate VRSAVE instructions when generating AltiVec code.
11032 @opindex msecure-plt
11033 Generate code that allows ld and ld.so to build executables and shared
11034 libraries with non-exec .plt and .got sections. This is a PowerPC
11035 32-bit SYSV ABI option.
11039 Generate code that uses a BSS .plt section that ld.so fills in, and
11040 requires .plt and .got sections that are both writable and executable.
11041 This is a PowerPC 32-bit SYSV ABI option.
11047 This switch enables or disables the generation of ISEL instructions.
11049 @item -misel=@var{yes/no}
11050 This switch has been deprecated. Use @option{-misel} and
11051 @option{-mno-isel} instead.
11057 This switch enables or disables the generation of SPE simd
11060 @item -mspe=@var{yes/no}
11061 This option has been deprecated. Use @option{-mspe} and
11062 @option{-mno-spe} instead.
11064 @item -mfloat-gprs=@var{yes/single/double/no}
11065 @itemx -mfloat-gprs
11066 @opindex mfloat-gprs
11067 This switch enables or disables the generation of floating point
11068 operations on the general purpose registers for architectures that
11071 The argument @var{yes} or @var{single} enables the use of
11072 single-precision floating point operations.
11074 The argument @var{double} enables the use of single and
11075 double-precision floating point operations.
11077 The argument @var{no} disables floating point operations on the
11078 general purpose registers.
11080 This option is currently only available on the MPC854x.
11086 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
11087 targets (including GNU/Linux). The 32-bit environment sets int, long
11088 and pointer to 32 bits and generates code that runs on any PowerPC
11089 variant. The 64-bit environment sets int to 32 bits and long and
11090 pointer to 64 bits, and generates code for PowerPC64, as for
11091 @option{-mpowerpc64}.
11094 @itemx -mno-fp-in-toc
11095 @itemx -mno-sum-in-toc
11096 @itemx -mminimal-toc
11098 @opindex mno-fp-in-toc
11099 @opindex mno-sum-in-toc
11100 @opindex mminimal-toc
11101 Modify generation of the TOC (Table Of Contents), which is created for
11102 every executable file. The @option{-mfull-toc} option is selected by
11103 default. In that case, GCC will allocate at least one TOC entry for
11104 each unique non-automatic variable reference in your program. GCC
11105 will also place floating-point constants in the TOC@. However, only
11106 16,384 entries are available in the TOC@.
11108 If you receive a linker error message that saying you have overflowed
11109 the available TOC space, you can reduce the amount of TOC space used
11110 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
11111 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
11112 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
11113 generate code to calculate the sum of an address and a constant at
11114 run-time instead of putting that sum into the TOC@. You may specify one
11115 or both of these options. Each causes GCC to produce very slightly
11116 slower and larger code at the expense of conserving TOC space.
11118 If you still run out of space in the TOC even when you specify both of
11119 these options, specify @option{-mminimal-toc} instead. This option causes
11120 GCC to make only one TOC entry for every file. When you specify this
11121 option, GCC will produce code that is slower and larger but which
11122 uses extremely little TOC space. You may wish to use this option
11123 only on files that contain less frequently executed code.
11129 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
11130 @code{long} type, and the infrastructure needed to support them.
11131 Specifying @option{-maix64} implies @option{-mpowerpc64} and
11132 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
11133 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
11136 @itemx -mno-xl-compat
11137 @opindex mxl-compat
11138 @opindex mno-xl-compat
11139 Produce code that conforms more closely to IBM XL compiler semantics
11140 when using AIX-compatible ABI. Pass floating-point arguments to
11141 prototyped functions beyond the register save area (RSA) on the stack
11142 in addition to argument FPRs. Do not assume that most significant
11143 double in 128-bit long double value is properly rounded when comparing
11144 values and converting to double. Use XL symbol names for long double
11147 The AIX calling convention was extended but not initially documented to
11148 handle an obscure K&R C case of calling a function that takes the
11149 address of its arguments with fewer arguments than declared. IBM XL
11150 compilers access floating point arguments which do not fit in the
11151 RSA from the stack when a subroutine is compiled without
11152 optimization. Because always storing floating-point arguments on the
11153 stack is inefficient and rarely needed, this option is not enabled by
11154 default and only is necessary when calling subroutines compiled by IBM
11155 XL compilers without optimization.
11159 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
11160 application written to use message passing with special startup code to
11161 enable the application to run. The system must have PE installed in the
11162 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
11163 must be overridden with the @option{-specs=} option to specify the
11164 appropriate directory location. The Parallel Environment does not
11165 support threads, so the @option{-mpe} option and the @option{-pthread}
11166 option are incompatible.
11168 @item -malign-natural
11169 @itemx -malign-power
11170 @opindex malign-natural
11171 @opindex malign-power
11172 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
11173 @option{-malign-natural} overrides the ABI-defined alignment of larger
11174 types, such as floating-point doubles, on their natural size-based boundary.
11175 The option @option{-malign-power} instructs GCC to follow the ABI-specified
11176 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
11178 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
11182 @itemx -mhard-float
11183 @opindex msoft-float
11184 @opindex mhard-float
11185 Generate code that does not use (uses) the floating-point register set.
11186 Software floating point emulation is provided if you use the
11187 @option{-msoft-float} option, and pass the option to GCC when linking.
11190 @itemx -mno-multiple
11192 @opindex mno-multiple
11193 Generate code that uses (does not use) the load multiple word
11194 instructions and the store multiple word instructions. These
11195 instructions are generated by default on POWER systems, and not
11196 generated on PowerPC systems. Do not use @option{-mmultiple} on little
11197 endian PowerPC systems, since those instructions do not work when the
11198 processor is in little endian mode. The exceptions are PPC740 and
11199 PPC750 which permit the instructions usage in little endian mode.
11204 @opindex mno-string
11205 Generate code that uses (does not use) the load string instructions
11206 and the store string word instructions to save multiple registers and
11207 do small block moves. These instructions are generated by default on
11208 POWER systems, and not generated on PowerPC systems. Do not use
11209 @option{-mstring} on little endian PowerPC systems, since those
11210 instructions do not work when the processor is in little endian mode.
11211 The exceptions are PPC740 and PPC750 which permit the instructions
11212 usage in little endian mode.
11217 @opindex mno-update
11218 Generate code that uses (does not use) the load or store instructions
11219 that update the base register to the address of the calculated memory
11220 location. These instructions are generated by default. If you use
11221 @option{-mno-update}, there is a small window between the time that the
11222 stack pointer is updated and the address of the previous frame is
11223 stored, which means code that walks the stack frame across interrupts or
11224 signals may get corrupted data.
11227 @itemx -mno-fused-madd
11228 @opindex mfused-madd
11229 @opindex mno-fused-madd
11230 Generate code that uses (does not use) the floating point multiply and
11231 accumulate instructions. These instructions are generated by default if
11232 hardware floating is used.
11234 @item -mno-bit-align
11236 @opindex mno-bit-align
11237 @opindex mbit-align
11238 On System V.4 and embedded PowerPC systems do not (do) force structures
11239 and unions that contain bit-fields to be aligned to the base type of the
11242 For example, by default a structure containing nothing but 8
11243 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
11244 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
11245 the structure would be aligned to a 1 byte boundary and be one byte in
11248 @item -mno-strict-align
11249 @itemx -mstrict-align
11250 @opindex mno-strict-align
11251 @opindex mstrict-align
11252 On System V.4 and embedded PowerPC systems do not (do) assume that
11253 unaligned memory references will be handled by the system.
11255 @item -mrelocatable
11256 @itemx -mno-relocatable
11257 @opindex mrelocatable
11258 @opindex mno-relocatable
11259 On embedded PowerPC systems generate code that allows (does not allow)
11260 the program to be relocated to a different address at runtime. If you
11261 use @option{-mrelocatable} on any module, all objects linked together must
11262 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
11264 @item -mrelocatable-lib
11265 @itemx -mno-relocatable-lib
11266 @opindex mrelocatable-lib
11267 @opindex mno-relocatable-lib
11268 On embedded PowerPC systems generate code that allows (does not allow)
11269 the program to be relocated to a different address at runtime. Modules
11270 compiled with @option{-mrelocatable-lib} can be linked with either modules
11271 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
11272 with modules compiled with the @option{-mrelocatable} options.
11278 On System V.4 and embedded PowerPC systems do not (do) assume that
11279 register 2 contains a pointer to a global area pointing to the addresses
11280 used in the program.
11283 @itemx -mlittle-endian
11285 @opindex mlittle-endian
11286 On System V.4 and embedded PowerPC systems compile code for the
11287 processor in little endian mode. The @option{-mlittle-endian} option is
11288 the same as @option{-mlittle}.
11291 @itemx -mbig-endian
11293 @opindex mbig-endian
11294 On System V.4 and embedded PowerPC systems compile code for the
11295 processor in big endian mode. The @option{-mbig-endian} option is
11296 the same as @option{-mbig}.
11298 @item -mdynamic-no-pic
11299 @opindex mdynamic-no-pic
11300 On Darwin and Mac OS X systems, compile code so that it is not
11301 relocatable, but that its external references are relocatable. The
11302 resulting code is suitable for applications, but not shared
11305 @item -mprioritize-restricted-insns=@var{priority}
11306 @opindex mprioritize-restricted-insns
11307 This option controls the priority that is assigned to
11308 dispatch-slot restricted instructions during the second scheduling
11309 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
11310 @var{no/highest/second-highest} priority to dispatch slot restricted
11313 @item -msched-costly-dep=@var{dependence_type}
11314 @opindex msched-costly-dep
11315 This option controls which dependences are considered costly
11316 by the target during instruction scheduling. The argument
11317 @var{dependence_type} takes one of the following values:
11318 @var{no}: no dependence is costly,
11319 @var{all}: all dependences are costly,
11320 @var{true_store_to_load}: a true dependence from store to load is costly,
11321 @var{store_to_load}: any dependence from store to load is costly,
11322 @var{number}: any dependence which latency >= @var{number} is costly.
11324 @item -minsert-sched-nops=@var{scheme}
11325 @opindex minsert-sched-nops
11326 This option controls which nop insertion scheme will be used during
11327 the second scheduling pass. The argument @var{scheme} takes one of the
11329 @var{no}: Don't insert nops.
11330 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
11331 according to the scheduler's grouping.
11332 @var{regroup_exact}: Insert nops to force costly dependent insns into
11333 separate groups. Insert exactly as many nops as needed to force an insn
11334 to a new group, according to the estimated processor grouping.
11335 @var{number}: Insert nops to force costly dependent insns into
11336 separate groups. Insert @var{number} nops to force an insn to a new group.
11339 @opindex mcall-sysv
11340 On System V.4 and embedded PowerPC systems compile code using calling
11341 conventions that adheres to the March 1995 draft of the System V
11342 Application Binary Interface, PowerPC processor supplement. This is the
11343 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
11345 @item -mcall-sysv-eabi
11346 @opindex mcall-sysv-eabi
11347 Specify both @option{-mcall-sysv} and @option{-meabi} options.
11349 @item -mcall-sysv-noeabi
11350 @opindex mcall-sysv-noeabi
11351 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
11353 @item -mcall-solaris
11354 @opindex mcall-solaris
11355 On System V.4 and embedded PowerPC systems compile code for the Solaris
11359 @opindex mcall-linux
11360 On System V.4 and embedded PowerPC systems compile code for the
11361 Linux-based GNU system.
11365 On System V.4 and embedded PowerPC systems compile code for the
11366 Hurd-based GNU system.
11368 @item -mcall-netbsd
11369 @opindex mcall-netbsd
11370 On System V.4 and embedded PowerPC systems compile code for the
11371 NetBSD operating system.
11373 @item -maix-struct-return
11374 @opindex maix-struct-return
11375 Return all structures in memory (as specified by the AIX ABI)@.
11377 @item -msvr4-struct-return
11378 @opindex msvr4-struct-return
11379 Return structures smaller than 8 bytes in registers (as specified by the
11382 @item -mabi=@var{abi-type}
11384 Extend the current ABI with a particular extension, or remove such extension.
11385 Valid values are @var{altivec}, @var{no-altivec}, @var{spe},
11386 @var{no-spe}, @var{ibmlongdouble}, @var{ieeelongdouble}@.
11390 Extend the current ABI with SPE ABI extensions. This does not change
11391 the default ABI, instead it adds the SPE ABI extensions to the current
11395 @opindex mabi=no-spe
11396 Disable Booke SPE ABI extensions for the current ABI@.
11398 @item -mabi=ibmlongdouble
11399 @opindex mabi=ibmlongdouble
11400 Change the current ABI to use IBM extended precision long double.
11401 This is a PowerPC 32-bit SYSV ABI option.
11403 @item -mabi=ieeelongdouble
11404 @opindex mabi=ieeelongdouble
11405 Change the current ABI to use IEEE extended precision long double.
11406 This is a PowerPC 32-bit Linux ABI option.
11409 @itemx -mno-prototype
11410 @opindex mprototype
11411 @opindex mno-prototype
11412 On System V.4 and embedded PowerPC systems assume that all calls to
11413 variable argument functions are properly prototyped. Otherwise, the
11414 compiler must insert an instruction before every non prototyped call to
11415 set or clear bit 6 of the condition code register (@var{CR}) to
11416 indicate whether floating point values were passed in the floating point
11417 registers in case the function takes a variable arguments. With
11418 @option{-mprototype}, only calls to prototyped variable argument functions
11419 will set or clear the bit.
11423 On embedded PowerPC systems, assume that the startup module is called
11424 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
11425 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
11430 On embedded PowerPC systems, assume that the startup module is called
11431 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
11436 On embedded PowerPC systems, assume that the startup module is called
11437 @file{crt0.o} and the standard C libraries are @file{libads.a} and
11440 @item -myellowknife
11441 @opindex myellowknife
11442 On embedded PowerPC systems, assume that the startup module is called
11443 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
11448 On System V.4 and embedded PowerPC systems, specify that you are
11449 compiling for a VxWorks system.
11453 Specify that you are compiling for the WindISS simulation environment.
11457 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
11458 header to indicate that @samp{eabi} extended relocations are used.
11464 On System V.4 and embedded PowerPC systems do (do not) adhere to the
11465 Embedded Applications Binary Interface (eabi) which is a set of
11466 modifications to the System V.4 specifications. Selecting @option{-meabi}
11467 means that the stack is aligned to an 8 byte boundary, a function
11468 @code{__eabi} is called to from @code{main} to set up the eabi
11469 environment, and the @option{-msdata} option can use both @code{r2} and
11470 @code{r13} to point to two separate small data areas. Selecting
11471 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
11472 do not call an initialization function from @code{main}, and the
11473 @option{-msdata} option will only use @code{r13} to point to a single
11474 small data area. The @option{-meabi} option is on by default if you
11475 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
11478 @opindex msdata=eabi
11479 On System V.4 and embedded PowerPC systems, put small initialized
11480 @code{const} global and static data in the @samp{.sdata2} section, which
11481 is pointed to by register @code{r2}. Put small initialized
11482 non-@code{const} global and static data in the @samp{.sdata} section,
11483 which is pointed to by register @code{r13}. Put small uninitialized
11484 global and static data in the @samp{.sbss} section, which is adjacent to
11485 the @samp{.sdata} section. The @option{-msdata=eabi} option is
11486 incompatible with the @option{-mrelocatable} option. The
11487 @option{-msdata=eabi} option also sets the @option{-memb} option.
11490 @opindex msdata=sysv
11491 On System V.4 and embedded PowerPC systems, put small global and static
11492 data in the @samp{.sdata} section, which is pointed to by register
11493 @code{r13}. Put small uninitialized global and static data in the
11494 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11495 The @option{-msdata=sysv} option is incompatible with the
11496 @option{-mrelocatable} option.
11498 @item -msdata=default
11500 @opindex msdata=default
11502 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11503 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11504 same as @option{-msdata=sysv}.
11507 @opindex msdata-data
11508 On System V.4 and embedded PowerPC systems, put small global and static
11509 data in the @samp{.sdata} section. Put small uninitialized global and
11510 static data in the @samp{.sbss} section. Do not use register @code{r13}
11511 to address small data however. This is the default behavior unless
11512 other @option{-msdata} options are used.
11516 @opindex msdata=none
11518 On embedded PowerPC systems, put all initialized global and static data
11519 in the @samp{.data} section, and all uninitialized data in the
11520 @samp{.bss} section.
11524 @cindex smaller data references (PowerPC)
11525 @cindex .sdata/.sdata2 references (PowerPC)
11526 On embedded PowerPC systems, put global and static items less than or
11527 equal to @var{num} bytes into the small data or bss sections instead of
11528 the normal data or bss section. By default, @var{num} is 8. The
11529 @option{-G @var{num}} switch is also passed to the linker.
11530 All modules should be compiled with the same @option{-G @var{num}} value.
11533 @itemx -mno-regnames
11535 @opindex mno-regnames
11536 On System V.4 and embedded PowerPC systems do (do not) emit register
11537 names in the assembly language output using symbolic forms.
11540 @itemx -mno-longcall
11542 @opindex mno-longcall
11543 Default to making all function calls indirectly, using a register, so
11544 that functions which reside further than 32 megabytes (33,554,432
11545 bytes) from the current location can be called. This setting can be
11546 overridden by the @code{shortcall} function attribute, or by
11547 @code{#pragma longcall(0)}.
11549 Some linkers are capable of detecting out-of-range calls and generating
11550 glue code on the fly. On these systems, long calls are unnecessary and
11551 generate slower code. As of this writing, the AIX linker can do this,
11552 as can the GNU linker for PowerPC/64. It is planned to add this feature
11553 to the GNU linker for 32-bit PowerPC systems as well.
11555 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11556 callee, L42'', plus a ``branch island'' (glue code). The two target
11557 addresses represent the callee and the ``branch island''. The
11558 Darwin/PPC linker will prefer the first address and generate a ``bl
11559 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11560 otherwise, the linker will generate ``bl L42'' to call the ``branch
11561 island''. The ``branch island'' is appended to the body of the
11562 calling function; it computes the full 32-bit address of the callee
11565 On Mach-O (Darwin) systems, this option directs the compiler emit to
11566 the glue for every direct call, and the Darwin linker decides whether
11567 to use or discard it.
11569 In the future, we may cause GCC to ignore all longcall specifications
11570 when the linker is known to generate glue.
11574 Adds support for multithreading with the @dfn{pthreads} library.
11575 This option sets flags for both the preprocessor and linker.
11579 @node S/390 and zSeries Options
11580 @subsection S/390 and zSeries Options
11581 @cindex S/390 and zSeries Options
11583 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11587 @itemx -msoft-float
11588 @opindex mhard-float
11589 @opindex msoft-float
11590 Use (do not use) the hardware floating-point instructions and registers
11591 for floating-point operations. When @option{-msoft-float} is specified,
11592 functions in @file{libgcc.a} will be used to perform floating-point
11593 operations. When @option{-mhard-float} is specified, the compiler
11594 generates IEEE floating-point instructions. This is the default.
11596 @item -mlong-double-64
11597 @itemx -mlong-double-128
11598 @opindex mlong-double-64
11599 @opindex mlong-double-128
11600 These switches control the size of @code{long double} type. A size
11601 of 64bit makes the @code{long double} type equivalent to the @code{double}
11602 type. This is the default.
11605 @itemx -mno-backchain
11606 @opindex mbackchain
11607 @opindex mno-backchain
11608 Store (do not store) the address of the caller's frame as backchain pointer
11609 into the callee's stack frame.
11610 A backchain may be needed to allow debugging using tools that do not understand
11611 DWARF-2 call frame information.
11612 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11613 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11614 the backchain is placed into the topmost word of the 96/160 byte register
11617 In general, code compiled with @option{-mbackchain} is call-compatible with
11618 code compiled with @option{-mmo-backchain}; however, use of the backchain
11619 for debugging purposes usually requires that the whole binary is built with
11620 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11621 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11622 to build a linux kernel use @option{-msoft-float}.
11624 The default is to not maintain the backchain.
11626 @item -mpacked-stack
11627 @item -mno-packed-stack
11628 @opindex mpacked-stack
11629 @opindex mno-packed-stack
11630 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11631 specified, the compiler uses the all fields of the 96/160 byte register save
11632 area only for their default purpose; unused fields still take up stack space.
11633 When @option{-mpacked-stack} is specified, register save slots are densely
11634 packed at the top of the register save area; unused space is reused for other
11635 purposes, allowing for more efficient use of the available stack space.
11636 However, when @option{-mbackchain} is also in effect, the topmost word of
11637 the save area is always used to store the backchain, and the return address
11638 register is always saved two words below the backchain.
11640 As long as the stack frame backchain is not used, code generated with
11641 @option{-mpacked-stack} is call-compatible with code generated with
11642 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11643 S/390 or zSeries generated code that uses the stack frame backchain at run
11644 time, not just for debugging purposes. Such code is not call-compatible
11645 with code compiled with @option{-mpacked-stack}. Also, note that the
11646 combination of @option{-mbackchain},
11647 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11648 to build a linux kernel use @option{-msoft-float}.
11650 The default is to not use the packed stack layout.
11653 @itemx -mno-small-exec
11654 @opindex msmall-exec
11655 @opindex mno-small-exec
11656 Generate (or do not generate) code using the @code{bras} instruction
11657 to do subroutine calls.
11658 This only works reliably if the total executable size does not
11659 exceed 64k. The default is to use the @code{basr} instruction instead,
11660 which does not have this limitation.
11666 When @option{-m31} is specified, generate code compliant to the
11667 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11668 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11669 particular to generate 64-bit instructions. For the @samp{s390}
11670 targets, the default is @option{-m31}, while the @samp{s390x}
11671 targets default to @option{-m64}.
11677 When @option{-mzarch} is specified, generate code using the
11678 instructions available on z/Architecture.
11679 When @option{-mesa} is specified, generate code using the
11680 instructions available on ESA/390. Note that @option{-mesa} is
11681 not possible with @option{-m64}.
11682 When generating code compliant to the GNU/Linux for S/390 ABI,
11683 the default is @option{-mesa}. When generating code compliant
11684 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11690 Generate (or do not generate) code using the @code{mvcle} instruction
11691 to perform block moves. When @option{-mno-mvcle} is specified,
11692 use a @code{mvc} loop instead. This is the default unless optimizing for
11699 Print (or do not print) additional debug information when compiling.
11700 The default is to not print debug information.
11702 @item -march=@var{cpu-type}
11704 Generate code that will run on @var{cpu-type}, which is the name of a system
11705 representing a certain processor type. Possible values for
11706 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11707 When generating code using the instructions available on z/Architecture,
11708 the default is @option{-march=z900}. Otherwise, the default is
11709 @option{-march=g5}.
11711 @item -mtune=@var{cpu-type}
11713 Tune to @var{cpu-type} everything applicable about the generated code,
11714 except for the ABI and the set of available instructions.
11715 The list of @var{cpu-type} values is the same as for @option{-march}.
11716 The default is the value used for @option{-march}.
11719 @itemx -mno-tpf-trace
11720 @opindex mtpf-trace
11721 @opindex mno-tpf-trace
11722 Generate code that adds (does not add) in TPF OS specific branches to trace
11723 routines in the operating system. This option is off by default, even
11724 when compiling for the TPF OS@.
11727 @itemx -mno-fused-madd
11728 @opindex mfused-madd
11729 @opindex mno-fused-madd
11730 Generate code that uses (does not use) the floating point multiply and
11731 accumulate instructions. These instructions are generated by default if
11732 hardware floating point is used.
11734 @item -mwarn-framesize=@var{framesize}
11735 @opindex mwarn-framesize
11736 Emit a warning if the current function exceeds the given frame size. Because
11737 this is a compile time check it doesn't need to be a real problem when the program
11738 runs. It is intended to identify functions which most probably cause
11739 a stack overflow. It is useful to be used in an environment with limited stack
11740 size e.g.@: the linux kernel.
11742 @item -mwarn-dynamicstack
11743 @opindex mwarn-dynamicstack
11744 Emit a warning if the function calls alloca or uses dynamically
11745 sized arrays. This is generally a bad idea with a limited stack size.
11747 @item -mstack-guard=@var{stack-guard}
11748 @item -mstack-size=@var{stack-size}
11749 @opindex mstack-guard
11750 @opindex mstack-size
11751 These arguments always have to be used in conjunction. If they are present the s390
11752 back end emits additional instructions in the function prologue which trigger a trap
11753 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11754 (remember that the stack on s390 grows downward). These options are intended to
11755 be used to help debugging stack overflow problems. The additionally emitted code
11756 causes only little overhead and hence can also be used in production like systems
11757 without greater performance degradation. The given values have to be exact
11758 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard} without
11760 In order to be efficient the extra code makes the assumption that the stack starts
11761 at an address aligned to the value given by @var{stack-size}.
11765 @subsection SH Options
11767 These @samp{-m} options are defined for the SH implementations:
11772 Generate code for the SH1.
11776 Generate code for the SH2.
11779 Generate code for the SH2e.
11783 Generate code for the SH3.
11787 Generate code for the SH3e.
11791 Generate code for the SH4 without a floating-point unit.
11793 @item -m4-single-only
11794 @opindex m4-single-only
11795 Generate code for the SH4 with a floating-point unit that only
11796 supports single-precision arithmetic.
11800 Generate code for the SH4 assuming the floating-point unit is in
11801 single-precision mode by default.
11805 Generate code for the SH4.
11809 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11810 floating-point unit is not used.
11812 @item -m4a-single-only
11813 @opindex m4a-single-only
11814 Generate code for the SH4a, in such a way that no double-precision
11815 floating point operations are used.
11818 @opindex m4a-single
11819 Generate code for the SH4a assuming the floating-point unit is in
11820 single-precision mode by default.
11824 Generate code for the SH4a.
11828 Same as @option{-m4a-nofpu}, except that it implicitly passes
11829 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11830 instructions at the moment.
11834 Compile code for the processor in big endian mode.
11838 Compile code for the processor in little endian mode.
11842 Align doubles at 64-bit boundaries. Note that this changes the calling
11843 conventions, and thus some functions from the standard C library will
11844 not work unless you recompile it first with @option{-mdalign}.
11848 Shorten some address references at link time, when possible; uses the
11849 linker option @option{-relax}.
11853 Use 32-bit offsets in @code{switch} tables. The default is to use
11858 Enable the use of the instruction @code{fmovd}.
11862 Comply with the calling conventions defined by Renesas.
11866 Comply with the calling conventions defined by Renesas.
11870 Comply with the calling conventions defined for GCC before the Renesas
11871 conventions were available. This option is the default for all
11872 targets of the SH toolchain except for @samp{sh-symbianelf}.
11875 @opindex mnomacsave
11876 Mark the @code{MAC} register as call-clobbered, even if
11877 @option{-mhitachi} is given.
11881 Increase IEEE-compliance of floating-point code.
11882 At the moment, this is equivalent to @option{-fno-finite-math-only}.
11883 When generating 16 bit SH opcodes, getting IEEE-conforming results for
11884 comparisons of NANs / infinities incurs extra overhead in every
11885 floating point comparison, therefore the default is set to
11886 @option{-ffinite-math-only}.
11890 Dump instruction size and location in the assembly code.
11893 @opindex mpadstruct
11894 This option is deprecated. It pads structures to multiple of 4 bytes,
11895 which is incompatible with the SH ABI@.
11899 Optimize for space instead of speed. Implied by @option{-Os}.
11902 @opindex mprefergot
11903 When generating position-independent code, emit function calls using
11904 the Global Offset Table instead of the Procedure Linkage Table.
11908 Generate a library function call to invalidate instruction cache
11909 entries, after fixing up a trampoline. This library function call
11910 doesn't assume it can write to the whole memory address space. This
11911 is the default when the target is @code{sh-*-linux*}.
11913 @item -multcost=@var{number}
11914 @opindex multcost=@var{number}
11915 Set the cost to assume for a multiply insn.
11917 @item -mdiv=@var{strategy}
11918 @opindex mdiv=@var{strategy}
11919 Set the division strategy to use for SHmedia code. @var{strategy} must be
11920 one of: call, call2, fp, inv, inv:minlat, inv20u, inv20l, inv:call,
11921 inv:call2, inv:fp .
11922 "fp" performs the operation in floating point. This has a very high latency,
11923 but needs only a few instructions, so it might be a good choice if
11924 your code has enough easily exploitable ILP to allow the compiler to
11925 schedule the floating point instructions together with other instructions.
11926 Division by zero causes a floating point exception.
11927 "inv" uses integer operations to calculate the inverse of the divisor,
11928 and then multiplies the dividend with the inverse. This strategy allows
11929 cse and hoisting of the inverse calculation. Division by zero calculates
11930 an unspecified result, but does not trap.
11931 "inv:minlat" is a variant of "inv" where if no cse / hoisting opportunities
11932 have been found, or if the entire operation has been hoisted to the same
11933 place, the last stages of the inverse calculation are intertwined with the
11934 final multiply to reduce the overall latency, at the expense of using a few
11935 more instructions, and thus offering fewer scheduling opportunities with
11937 "call" calls a library function that usually implements the inv:minlat
11939 This gives high code density for m5-*media-nofpu compilations.
11940 "call2" uses a different entry point of the same library function, where it
11941 assumes that a pointer to a lookup table has already been set up, which
11942 exposes the pointer load to cse / code hoisting optimizations.
11943 "inv:call", "inv:call2" and "inv:fp" all use the "inv" algorithm for initial
11944 code generation, but if the code stays unoptimized, revert to the "call",
11945 "call2", or "fp" strategies, respectively. Note that the
11946 potentially-trapping side effect of division by zero is carried by a
11947 separate instruction, so it is possible that all the integer instructions
11948 are hoisted out, but the marker for the side effect stays where it is.
11949 A recombination to fp operations or a call is not possible in that case.
11950 "inv20u" and "inv20l" are variants of the "inv:minlat" strategy. In the case
11951 that the inverse calculation was nor separated from the multiply, they speed
11952 up division where the dividend fits into 20 bits (plus sign where applicable),
11953 by inserting a test to skip a number of operations in this case; this test
11954 slows down the case of larger dividends. inv20u assumes the case of a such
11955 a small dividend to be unlikely, and inv20l assumes it to be likely.
11957 @item -mdivsi3_libfunc=@var{name}
11958 @opindex mdivsi3_libfunc=@var{name}
11959 Set the name of the library function used for 32 bit signed division to
11960 @var{name}. This only affect the name used in the call and inv:call
11961 division strategies, and the compiler will still expect the same
11962 sets of input/output/clobbered registers as if this option was not present.
11964 @item -madjust-unroll
11965 @opindex madjust-unroll
11966 Throttle unrolling to avoid thrashing target registers.
11967 This option only has an effect if the gcc code base supports the
11968 TARGET_ADJUST_UNROLL_MAX target hook.
11970 @item -mindexed-addressing
11971 @opindex mindexed-addressing
11972 Enable the use of the indexed addressing mode for SHmedia32/SHcompact.
11973 This is only safe if the hardware and/or OS implement 32 bit wrap-around
11974 semantics for the indexed addressing mode. The architecture allows the
11975 implementation of processors with 64 bit MMU, which the OS could use to
11976 get 32 bit addressing, but since no current hardware implementation supports
11977 this or any other way to make the indexed addressing mode safe to use in
11978 the 32 bit ABI, the default is -mno-indexed-addressing.
11980 @item -mgettrcost=@var{number}
11981 @opindex mgettrcost=@var{number}
11982 Set the cost assumed for the gettr instruction to @var{number}.
11983 The default is 2 if @option{-mpt-fixed} is in effect, 100 otherwise.
11987 Assume pt* instructions won't trap. This will generally generate better
11988 scheduled code, but is unsafe on current hardware. The current architecture
11989 definition says that ptabs and ptrel trap when the target anded with 3 is 3.
11990 This has the unintentional effect of making it unsafe to schedule ptabs /
11991 ptrel before a branch, or hoist it out of a loop. For example,
11992 __do_global_ctors, a part of libgcc that runs constructors at program
11993 startup, calls functions in a list which is delimited by -1. With the
11994 -mpt-fixed option, the ptabs will be done before testing against -1.
11995 That means that all the constructors will be run a bit quicker, but when
11996 the loop comes to the end of the list, the program crashes because ptabs
11997 loads -1 into a target register. Since this option is unsafe for any
11998 hardware implementing the current architecture specification, the default
11999 is -mno-pt-fixed. Unless the user specifies a specific cost with
12000 @option{-mgettrcost}, -mno-pt-fixed also implies @option{-mgettrcost=100};
12001 this deters register allocation using target registers for storing
12004 @item -minvalid-symbols
12005 @opindex minvalid-symbols
12006 Assume symbols might be invalid. Ordinary function symbols generated by
12007 the compiler will always be valid to load with movi/shori/ptabs or
12008 movi/shori/ptrel, but with assembler and/or linker tricks it is possible
12009 to generate symbols that will cause ptabs / ptrel to trap.
12010 This option is only meaningful when @option{-mno-pt-fixed} is in effect.
12011 It will then prevent cross-basic-block cse, hoisting and most scheduling
12012 of symbol loads. The default is @option{-mno-invalid-symbols}.
12015 @node SPARC Options
12016 @subsection SPARC Options
12017 @cindex SPARC options
12019 These @samp{-m} options are supported on the SPARC:
12022 @item -mno-app-regs
12024 @opindex mno-app-regs
12026 Specify @option{-mapp-regs} to generate output using the global registers
12027 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
12030 To be fully SVR4 ABI compliant at the cost of some performance loss,
12031 specify @option{-mno-app-regs}. You should compile libraries and system
12032 software with this option.
12035 @itemx -mhard-float
12037 @opindex mhard-float
12038 Generate output containing floating point instructions. This is the
12042 @itemx -msoft-float
12044 @opindex msoft-float
12045 Generate output containing library calls for floating point.
12046 @strong{Warning:} the requisite libraries are not available for all SPARC
12047 targets. Normally the facilities of the machine's usual C compiler are
12048 used, but this cannot be done directly in cross-compilation. You must make
12049 your own arrangements to provide suitable library functions for
12050 cross-compilation. The embedded targets @samp{sparc-*-aout} and
12051 @samp{sparclite-*-*} do provide software floating point support.
12053 @option{-msoft-float} changes the calling convention in the output file;
12054 therefore, it is only useful if you compile @emph{all} of a program with
12055 this option. In particular, you need to compile @file{libgcc.a}, the
12056 library that comes with GCC, with @option{-msoft-float} in order for
12059 @item -mhard-quad-float
12060 @opindex mhard-quad-float
12061 Generate output containing quad-word (long double) floating point
12064 @item -msoft-quad-float
12065 @opindex msoft-quad-float
12066 Generate output containing library calls for quad-word (long double)
12067 floating point instructions. The functions called are those specified
12068 in the SPARC ABI@. This is the default.
12070 As of this writing, there are no SPARC implementations that have hardware
12071 support for the quad-word floating point instructions. They all invoke
12072 a trap handler for one of these instructions, and then the trap handler
12073 emulates the effect of the instruction. Because of the trap handler overhead,
12074 this is much slower than calling the ABI library routines. Thus the
12075 @option{-msoft-quad-float} option is the default.
12077 @item -mno-unaligned-doubles
12078 @itemx -munaligned-doubles
12079 @opindex mno-unaligned-doubles
12080 @opindex munaligned-doubles
12081 Assume that doubles have 8 byte alignment. This is the default.
12083 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
12084 alignment only if they are contained in another type, or if they have an
12085 absolute address. Otherwise, it assumes they have 4 byte alignment.
12086 Specifying this option avoids some rare compatibility problems with code
12087 generated by other compilers. It is not the default because it results
12088 in a performance loss, especially for floating point code.
12090 @item -mno-faster-structs
12091 @itemx -mfaster-structs
12092 @opindex mno-faster-structs
12093 @opindex mfaster-structs
12094 With @option{-mfaster-structs}, the compiler assumes that structures
12095 should have 8 byte alignment. This enables the use of pairs of
12096 @code{ldd} and @code{std} instructions for copies in structure
12097 assignment, in place of twice as many @code{ld} and @code{st} pairs.
12098 However, the use of this changed alignment directly violates the SPARC
12099 ABI@. Thus, it's intended only for use on targets where the developer
12100 acknowledges that their resulting code will not be directly in line with
12101 the rules of the ABI@.
12103 @item -mimpure-text
12104 @opindex mimpure-text
12105 @option{-mimpure-text}, used in addition to @option{-shared}, tells
12106 the compiler to not pass @option{-z text} to the linker when linking a
12107 shared object. Using this option, you can link position-dependent
12108 code into a shared object.
12110 @option{-mimpure-text} suppresses the ``relocations remain against
12111 allocatable but non-writable sections'' linker error message.
12112 However, the necessary relocations will trigger copy-on-write, and the
12113 shared object is not actually shared across processes. Instead of
12114 using @option{-mimpure-text}, you should compile all source code with
12115 @option{-fpic} or @option{-fPIC}.
12117 This option is only available on SunOS and Solaris.
12119 @item -mcpu=@var{cpu_type}
12121 Set the instruction set, register set, and instruction scheduling parameters
12122 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
12123 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
12124 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
12125 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
12126 @samp{ultrasparc3}.
12128 Default instruction scheduling parameters are used for values that select
12129 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
12130 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
12132 Here is a list of each supported architecture and their supported
12137 v8: supersparc, hypersparc
12138 sparclite: f930, f934, sparclite86x
12140 v9: ultrasparc, ultrasparc3
12143 By default (unless configured otherwise), GCC generates code for the V7
12144 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
12145 additionally optimizes it for the Cypress CY7C602 chip, as used in the
12146 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
12147 SPARCStation 1, 2, IPX etc.
12149 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
12150 architecture. The only difference from V7 code is that the compiler emits
12151 the integer multiply and integer divide instructions which exist in SPARC-V8
12152 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
12153 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
12156 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
12157 the SPARC architecture. This adds the integer multiply, integer divide step
12158 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
12159 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
12160 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
12161 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
12162 MB86934 chip, which is the more recent SPARClite with FPU@.
12164 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
12165 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
12166 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
12167 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
12168 optimizes it for the TEMIC SPARClet chip.
12170 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
12171 architecture. This adds 64-bit integer and floating-point move instructions,
12172 3 additional floating-point condition code registers and conditional move
12173 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
12174 optimizes it for the Sun UltraSPARC I/II chips. With
12175 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
12176 Sun UltraSPARC III chip.
12178 @item -mtune=@var{cpu_type}
12180 Set the instruction scheduling parameters for machine type
12181 @var{cpu_type}, but do not set the instruction set or register set that the
12182 option @option{-mcpu=@var{cpu_type}} would.
12184 The same values for @option{-mcpu=@var{cpu_type}} can be used for
12185 @option{-mtune=@var{cpu_type}}, but the only useful values are those
12186 that select a particular cpu implementation. Those are @samp{cypress},
12187 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
12188 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
12189 @samp{ultrasparc3}.
12194 @opindex mno-v8plus
12195 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
12196 difference from the V8 ABI is that the global and out registers are
12197 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
12198 mode for all SPARC-V9 processors.
12204 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
12205 Visual Instruction Set extensions. The default is @option{-mno-vis}.
12208 These @samp{-m} options are supported in addition to the above
12209 on SPARC-V9 processors in 64-bit environments:
12212 @item -mlittle-endian
12213 @opindex mlittle-endian
12214 Generate code for a processor running in little-endian mode. It is only
12215 available for a few configurations and most notably not on Solaris and Linux.
12221 Generate code for a 32-bit or 64-bit environment.
12222 The 32-bit environment sets int, long and pointer to 32 bits.
12223 The 64-bit environment sets int to 32 bits and long and pointer
12226 @item -mcmodel=medlow
12227 @opindex mcmodel=medlow
12228 Generate code for the Medium/Low code model: 64-bit addresses, programs
12229 must be linked in the low 32 bits of memory. Programs can be statically
12230 or dynamically linked.
12232 @item -mcmodel=medmid
12233 @opindex mcmodel=medmid
12234 Generate code for the Medium/Middle code model: 64-bit addresses, programs
12235 must be linked in the low 44 bits of memory, the text and data segments must
12236 be less than 2GB in size and the data segment must be located within 2GB of
12239 @item -mcmodel=medany
12240 @opindex mcmodel=medany
12241 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
12242 may be linked anywhere in memory, the text and data segments must be less
12243 than 2GB in size and the data segment must be located within 2GB of the
12246 @item -mcmodel=embmedany
12247 @opindex mcmodel=embmedany
12248 Generate code for the Medium/Anywhere code model for embedded systems:
12249 64-bit addresses, the text and data segments must be less than 2GB in
12250 size, both starting anywhere in memory (determined at link time). The
12251 global register %g4 points to the base of the data segment. Programs
12252 are statically linked and PIC is not supported.
12255 @itemx -mno-stack-bias
12256 @opindex mstack-bias
12257 @opindex mno-stack-bias
12258 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
12259 frame pointer if present, are offset by @minus{}2047 which must be added back
12260 when making stack frame references. This is the default in 64-bit mode.
12261 Otherwise, assume no such offset is present.
12264 These switches are supported in addition to the above on Solaris:
12269 Add support for multithreading using the Solaris threads library. This
12270 option sets flags for both the preprocessor and linker. This option does
12271 not affect the thread safety of object code produced by the compiler or
12272 that of libraries supplied with it.
12276 Add support for multithreading using the POSIX threads library. This
12277 option sets flags for both the preprocessor and linker. This option does
12278 not affect the thread safety of object code produced by the compiler or
12279 that of libraries supplied with it.
12283 This is a synonym for @option{-pthreads}.
12286 @node System V Options
12287 @subsection Options for System V
12289 These additional options are available on System V Release 4 for
12290 compatibility with other compilers on those systems:
12295 Create a shared object.
12296 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
12300 Identify the versions of each tool used by the compiler, in a
12301 @code{.ident} assembler directive in the output.
12305 Refrain from adding @code{.ident} directives to the output file (this is
12308 @item -YP,@var{dirs}
12310 Search the directories @var{dirs}, and no others, for libraries
12311 specified with @option{-l}.
12313 @item -Ym,@var{dir}
12315 Look in the directory @var{dir} to find the M4 preprocessor.
12316 The assembler uses this option.
12317 @c This is supposed to go with a -Yd for predefined M4 macro files, but
12318 @c the generic assembler that comes with Solaris takes just -Ym.
12321 @node TMS320C3x/C4x Options
12322 @subsection TMS320C3x/C4x Options
12323 @cindex TMS320C3x/C4x Options
12325 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
12329 @item -mcpu=@var{cpu_type}
12331 Set the instruction set, register set, and instruction scheduling
12332 parameters for machine type @var{cpu_type}. Supported values for
12333 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
12334 @samp{c44}. The default is @samp{c40} to generate code for the
12339 @itemx -msmall-memory
12341 @opindex mbig-memory
12343 @opindex msmall-memory
12345 Generates code for the big or small memory model. The small memory
12346 model assumed that all data fits into one 64K word page. At run-time
12347 the data page (DP) register must be set to point to the 64K page
12348 containing the .bss and .data program sections. The big memory model is
12349 the default and requires reloading of the DP register for every direct
12356 Allow (disallow) allocation of general integer operands into the block
12357 count register BK@.
12363 Enable (disable) generation of code using decrement and branch,
12364 DBcond(D), instructions. This is enabled by default for the C4x. To be
12365 on the safe side, this is disabled for the C3x, since the maximum
12366 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
12367 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
12368 that it can utilize the decrement and branch instruction, but will give
12369 up if there is more than one memory reference in the loop. Thus a loop
12370 where the loop counter is decremented can generate slightly more
12371 efficient code, in cases where the RPTB instruction cannot be utilized.
12373 @item -mdp-isr-reload
12375 @opindex mdp-isr-reload
12377 Force the DP register to be saved on entry to an interrupt service
12378 routine (ISR), reloaded to point to the data section, and restored on
12379 exit from the ISR@. This should not be required unless someone has
12380 violated the small memory model by modifying the DP register, say within
12387 For the C3x use the 24-bit MPYI instruction for integer multiplies
12388 instead of a library call to guarantee 32-bit results. Note that if one
12389 of the operands is a constant, then the multiplication will be performed
12390 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
12391 then squaring operations are performed inline instead of a library call.
12394 @itemx -mno-fast-fix
12396 @opindex mno-fast-fix
12397 The C3x/C4x FIX instruction to convert a floating point value to an
12398 integer value chooses the nearest integer less than or equal to the
12399 floating point value rather than to the nearest integer. Thus if the
12400 floating point number is negative, the result will be incorrectly
12401 truncated an additional code is necessary to detect and correct this
12402 case. This option can be used to disable generation of the additional
12403 code required to correct the result.
12409 Enable (disable) generation of repeat block sequences using the RPTB
12410 instruction for zero overhead looping. The RPTB construct is only used
12411 for innermost loops that do not call functions or jump across the loop
12412 boundaries. There is no advantage having nested RPTB loops due to the
12413 overhead required to save and restore the RC, RS, and RE registers.
12414 This is enabled by default with @option{-O2}.
12416 @item -mrpts=@var{count}
12420 Enable (disable) the use of the single instruction repeat instruction
12421 RPTS@. If a repeat block contains a single instruction, and the loop
12422 count can be guaranteed to be less than the value @var{count}, GCC will
12423 emit a RPTS instruction instead of a RPTB@. If no value is specified,
12424 then a RPTS will be emitted even if the loop count cannot be determined
12425 at compile time. Note that the repeated instruction following RPTS does
12426 not have to be reloaded from memory each iteration, thus freeing up the
12427 CPU buses for operands. However, since interrupts are blocked by this
12428 instruction, it is disabled by default.
12430 @item -mloop-unsigned
12431 @itemx -mno-loop-unsigned
12432 @opindex mloop-unsigned
12433 @opindex mno-loop-unsigned
12434 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
12435 is @math{2^{31} + 1} since these instructions test if the iteration count is
12436 negative to terminate the loop. If the iteration count is unsigned
12437 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
12438 exceeded. This switch allows an unsigned iteration count.
12442 Try to emit an assembler syntax that the TI assembler (asm30) is happy
12443 with. This also enforces compatibility with the API employed by the TI
12444 C3x C compiler. For example, long doubles are passed as structures
12445 rather than in floating point registers.
12451 Generate code that uses registers (stack) for passing arguments to functions.
12452 By default, arguments are passed in registers where possible rather
12453 than by pushing arguments on to the stack.
12455 @item -mparallel-insns
12456 @itemx -mno-parallel-insns
12457 @opindex mparallel-insns
12458 @opindex mno-parallel-insns
12459 Allow the generation of parallel instructions. This is enabled by
12460 default with @option{-O2}.
12462 @item -mparallel-mpy
12463 @itemx -mno-parallel-mpy
12464 @opindex mparallel-mpy
12465 @opindex mno-parallel-mpy
12466 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
12467 provided @option{-mparallel-insns} is also specified. These instructions have
12468 tight register constraints which can pessimize the code generation
12469 of large functions.
12474 @subsection V850 Options
12475 @cindex V850 Options
12477 These @samp{-m} options are defined for V850 implementations:
12481 @itemx -mno-long-calls
12482 @opindex mlong-calls
12483 @opindex mno-long-calls
12484 Treat all calls as being far away (near). If calls are assumed to be
12485 far away, the compiler will always load the functions address up into a
12486 register, and call indirect through the pointer.
12492 Do not optimize (do optimize) basic blocks that use the same index
12493 pointer 4 or more times to copy pointer into the @code{ep} register, and
12494 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
12495 option is on by default if you optimize.
12497 @item -mno-prolog-function
12498 @itemx -mprolog-function
12499 @opindex mno-prolog-function
12500 @opindex mprolog-function
12501 Do not use (do use) external functions to save and restore registers
12502 at the prologue and epilogue of a function. The external functions
12503 are slower, but use less code space if more than one function saves
12504 the same number of registers. The @option{-mprolog-function} option
12505 is on by default if you optimize.
12509 Try to make the code as small as possible. At present, this just turns
12510 on the @option{-mep} and @option{-mprolog-function} options.
12512 @item -mtda=@var{n}
12514 Put static or global variables whose size is @var{n} bytes or less into
12515 the tiny data area that register @code{ep} points to. The tiny data
12516 area can hold up to 256 bytes in total (128 bytes for byte references).
12518 @item -msda=@var{n}
12520 Put static or global variables whose size is @var{n} bytes or less into
12521 the small data area that register @code{gp} points to. The small data
12522 area can hold up to 64 kilobytes.
12524 @item -mzda=@var{n}
12526 Put static or global variables whose size is @var{n} bytes or less into
12527 the first 32 kilobytes of memory.
12531 Specify that the target processor is the V850.
12534 @opindex mbig-switch
12535 Generate code suitable for big switch tables. Use this option only if
12536 the assembler/linker complain about out of range branches within a switch
12541 This option will cause r2 and r5 to be used in the code generated by
12542 the compiler. This setting is the default.
12544 @item -mno-app-regs
12545 @opindex mno-app-regs
12546 This option will cause r2 and r5 to be treated as fixed registers.
12550 Specify that the target processor is the V850E1. The preprocessor
12551 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
12552 this option is used.
12556 Specify that the target processor is the V850E@. The preprocessor
12557 constant @samp{__v850e__} will be defined if this option is used.
12559 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
12560 are defined then a default target processor will be chosen and the
12561 relevant @samp{__v850*__} preprocessor constant will be defined.
12563 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
12564 defined, regardless of which processor variant is the target.
12566 @item -mdisable-callt
12567 @opindex mdisable-callt
12568 This option will suppress generation of the CALLT instruction for the
12569 v850e and v850e1 flavors of the v850 architecture. The default is
12570 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
12575 @subsection VAX Options
12576 @cindex VAX options
12578 These @samp{-m} options are defined for the VAX:
12583 Do not output certain jump instructions (@code{aobleq} and so on)
12584 that the Unix assembler for the VAX cannot handle across long
12589 Do output those jump instructions, on the assumption that you
12590 will assemble with the GNU assembler.
12594 Output code for g-format floating point numbers instead of d-format.
12597 @node x86-64 Options
12598 @subsection x86-64 Options
12599 @cindex x86-64 options
12601 These are listed under @xref{i386 and x86-64 Options}.
12603 @node Xstormy16 Options
12604 @subsection Xstormy16 Options
12605 @cindex Xstormy16 Options
12607 These options are defined for Xstormy16:
12612 Choose startup files and linker script suitable for the simulator.
12615 @node Xtensa Options
12616 @subsection Xtensa Options
12617 @cindex Xtensa Options
12619 These options are supported for Xtensa targets:
12623 @itemx -mno-const16
12625 @opindex mno-const16
12626 Enable or disable use of @code{CONST16} instructions for loading
12627 constant values. The @code{CONST16} instruction is currently not a
12628 standard option from Tensilica. When enabled, @code{CONST16}
12629 instructions are always used in place of the standard @code{L32R}
12630 instructions. The use of @code{CONST16} is enabled by default only if
12631 the @code{L32R} instruction is not available.
12634 @itemx -mno-fused-madd
12635 @opindex mfused-madd
12636 @opindex mno-fused-madd
12637 Enable or disable use of fused multiply/add and multiply/subtract
12638 instructions in the floating-point option. This has no effect if the
12639 floating-point option is not also enabled. Disabling fused multiply/add
12640 and multiply/subtract instructions forces the compiler to use separate
12641 instructions for the multiply and add/subtract operations. This may be
12642 desirable in some cases where strict IEEE 754-compliant results are
12643 required: the fused multiply add/subtract instructions do not round the
12644 intermediate result, thereby producing results with @emph{more} bits of
12645 precision than specified by the IEEE standard. Disabling fused multiply
12646 add/subtract instructions also ensures that the program output is not
12647 sensitive to the compiler's ability to combine multiply and add/subtract
12650 @item -mtext-section-literals
12651 @itemx -mno-text-section-literals
12652 @opindex mtext-section-literals
12653 @opindex mno-text-section-literals
12654 Control the treatment of literal pools. The default is
12655 @option{-mno-text-section-literals}, which places literals in a separate
12656 section in the output file. This allows the literal pool to be placed
12657 in a data RAM/ROM, and it also allows the linker to combine literal
12658 pools from separate object files to remove redundant literals and
12659 improve code size. With @option{-mtext-section-literals}, the literals
12660 are interspersed in the text section in order to keep them as close as
12661 possible to their references. This may be necessary for large assembly
12664 @item -mtarget-align
12665 @itemx -mno-target-align
12666 @opindex mtarget-align
12667 @opindex mno-target-align
12668 When this option is enabled, GCC instructs the assembler to
12669 automatically align instructions to reduce branch penalties at the
12670 expense of some code density. The assembler attempts to widen density
12671 instructions to align branch targets and the instructions following call
12672 instructions. If there are not enough preceding safe density
12673 instructions to align a target, no widening will be performed. The
12674 default is @option{-mtarget-align}. These options do not affect the
12675 treatment of auto-aligned instructions like @code{LOOP}, which the
12676 assembler will always align, either by widening density instructions or
12677 by inserting no-op instructions.
12680 @itemx -mno-longcalls
12681 @opindex mlongcalls
12682 @opindex mno-longcalls
12683 When this option is enabled, GCC instructs the assembler to translate
12684 direct calls to indirect calls unless it can determine that the target
12685 of a direct call is in the range allowed by the call instruction. This
12686 translation typically occurs for calls to functions in other source
12687 files. Specifically, the assembler translates a direct @code{CALL}
12688 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12689 The default is @option{-mno-longcalls}. This option should be used in
12690 programs where the call target can potentially be out of range. This
12691 option is implemented in the assembler, not the compiler, so the
12692 assembly code generated by GCC will still show direct call
12693 instructions---look at the disassembled object code to see the actual
12694 instructions. Note that the assembler will use an indirect call for
12695 every cross-file call, not just those that really will be out of range.
12698 @node zSeries Options
12699 @subsection zSeries Options
12700 @cindex zSeries options
12702 These are listed under @xref{S/390 and zSeries Options}.
12704 @node Code Gen Options
12705 @section Options for Code Generation Conventions
12706 @cindex code generation conventions
12707 @cindex options, code generation
12708 @cindex run-time options
12710 These machine-independent options control the interface conventions
12711 used in code generation.
12713 Most of them have both positive and negative forms; the negative form
12714 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12715 one of the forms is listed---the one which is not the default. You
12716 can figure out the other form by either removing @samp{no-} or adding
12720 @item -fbounds-check
12721 @opindex fbounds-check
12722 For front-ends that support it, generate additional code to check that
12723 indices used to access arrays are within the declared range. This is
12724 currently only supported by the Java and Fortran 77 front-ends, where
12725 this option defaults to true and false respectively.
12729 This option generates traps for signed overflow on addition, subtraction,
12730 multiplication operations.
12734 This option instructs the compiler to assume that signed arithmetic
12735 overflow of addition, subtraction and multiplication wraps around
12736 using twos-complement representation. This flag enables some optimizations
12737 and disables others. This option is enabled by default for the Java
12738 front-end, as required by the Java language specification.
12741 @opindex fexceptions
12742 Enable exception handling. Generates extra code needed to propagate
12743 exceptions. For some targets, this implies GCC will generate frame
12744 unwind information for all functions, which can produce significant data
12745 size overhead, although it does not affect execution. If you do not
12746 specify this option, GCC will enable it by default for languages like
12747 C++ which normally require exception handling, and disable it for
12748 languages like C that do not normally require it. However, you may need
12749 to enable this option when compiling C code that needs to interoperate
12750 properly with exception handlers written in C++. You may also wish to
12751 disable this option if you are compiling older C++ programs that don't
12752 use exception handling.
12754 @item -fnon-call-exceptions
12755 @opindex fnon-call-exceptions
12756 Generate code that allows trapping instructions to throw exceptions.
12757 Note that this requires platform-specific runtime support that does
12758 not exist everywhere. Moreover, it only allows @emph{trapping}
12759 instructions to throw exceptions, i.e.@: memory references or floating
12760 point instructions. It does not allow exceptions to be thrown from
12761 arbitrary signal handlers such as @code{SIGALRM}.
12763 @item -funwind-tables
12764 @opindex funwind-tables
12765 Similar to @option{-fexceptions}, except that it will just generate any needed
12766 static data, but will not affect the generated code in any other way.
12767 You will normally not enable this option; instead, a language processor
12768 that needs this handling would enable it on your behalf.
12770 @item -fasynchronous-unwind-tables
12771 @opindex fasynchronous-unwind-tables
12772 Generate unwind table in dwarf2 format, if supported by target machine. The
12773 table is exact at each instruction boundary, so it can be used for stack
12774 unwinding from asynchronous events (such as debugger or garbage collector).
12776 @item -fpcc-struct-return
12777 @opindex fpcc-struct-return
12778 Return ``short'' @code{struct} and @code{union} values in memory like
12779 longer ones, rather than in registers. This convention is less
12780 efficient, but it has the advantage of allowing intercallability between
12781 GCC-compiled files and files compiled with other compilers, particularly
12782 the Portable C Compiler (pcc).
12784 The precise convention for returning structures in memory depends
12785 on the target configuration macros.
12787 Short structures and unions are those whose size and alignment match
12788 that of some integer type.
12790 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12791 switch is not binary compatible with code compiled with the
12792 @option{-freg-struct-return} switch.
12793 Use it to conform to a non-default application binary interface.
12795 @item -freg-struct-return
12796 @opindex freg-struct-return
12797 Return @code{struct} and @code{union} values in registers when possible.
12798 This is more efficient for small structures than
12799 @option{-fpcc-struct-return}.
12801 If you specify neither @option{-fpcc-struct-return} nor
12802 @option{-freg-struct-return}, GCC defaults to whichever convention is
12803 standard for the target. If there is no standard convention, GCC
12804 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12805 the principal compiler. In those cases, we can choose the standard, and
12806 we chose the more efficient register return alternative.
12808 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12809 switch is not binary compatible with code compiled with the
12810 @option{-fpcc-struct-return} switch.
12811 Use it to conform to a non-default application binary interface.
12813 @item -fshort-enums
12814 @opindex fshort-enums
12815 Allocate to an @code{enum} type only as many bytes as it needs for the
12816 declared range of possible values. Specifically, the @code{enum} type
12817 will be equivalent to the smallest integer type which has enough room.
12819 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12820 code that is not binary compatible with code generated without that switch.
12821 Use it to conform to a non-default application binary interface.
12823 @item -fshort-double
12824 @opindex fshort-double
12825 Use the same size for @code{double} as for @code{float}.
12827 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12828 code that is not binary compatible with code generated without that switch.
12829 Use it to conform to a non-default application binary interface.
12831 @item -fshort-wchar
12832 @opindex fshort-wchar
12833 Override the underlying type for @samp{wchar_t} to be @samp{short
12834 unsigned int} instead of the default for the target. This option is
12835 useful for building programs to run under WINE@.
12837 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12838 code that is not binary compatible with code generated without that switch.
12839 Use it to conform to a non-default application binary interface.
12841 @item -fshared-data
12842 @opindex fshared-data
12843 Requests that the data and non-@code{const} variables of this
12844 compilation be shared data rather than private data. The distinction
12845 makes sense only on certain operating systems, where shared data is
12846 shared between processes running the same program, while private data
12847 exists in one copy per process.
12850 @opindex fno-common
12851 In C, allocate even uninitialized global variables in the data section of the
12852 object file, rather than generating them as common blocks. This has the
12853 effect that if the same variable is declared (without @code{extern}) in
12854 two different compilations, you will get an error when you link them.
12855 The only reason this might be useful is if you wish to verify that the
12856 program will work on other systems which always work this way.
12860 Ignore the @samp{#ident} directive.
12862 @item -finhibit-size-directive
12863 @opindex finhibit-size-directive
12864 Don't output a @code{.size} assembler directive, or anything else that
12865 would cause trouble if the function is split in the middle, and the
12866 two halves are placed at locations far apart in memory. This option is
12867 used when compiling @file{crtstuff.c}; you should not need to use it
12870 @item -fverbose-asm
12871 @opindex fverbose-asm
12872 Put extra commentary information in the generated assembly code to
12873 make it more readable. This option is generally only of use to those
12874 who actually need to read the generated assembly code (perhaps while
12875 debugging the compiler itself).
12877 @option{-fno-verbose-asm}, the default, causes the
12878 extra information to be omitted and is useful when comparing two assembler
12883 @cindex global offset table
12885 Generate position-independent code (PIC) suitable for use in a shared
12886 library, if supported for the target machine. Such code accesses all
12887 constant addresses through a global offset table (GOT)@. The dynamic
12888 loader resolves the GOT entries when the program starts (the dynamic
12889 loader is not part of GCC; it is part of the operating system). If
12890 the GOT size for the linked executable exceeds a machine-specific
12891 maximum size, you get an error message from the linker indicating that
12892 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12893 instead. (These maximums are 8k on the SPARC and 32k
12894 on the m68k and RS/6000. The 386 has no such limit.)
12896 Position-independent code requires special support, and therefore works
12897 only on certain machines. For the 386, GCC supports PIC for System V
12898 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12899 position-independent.
12903 If supported for the target machine, emit position-independent code,
12904 suitable for dynamic linking and avoiding any limit on the size of the
12905 global offset table. This option makes a difference on the m68k,
12906 PowerPC and SPARC@.
12908 Position-independent code requires special support, and therefore works
12909 only on certain machines.
12915 These options are similar to @option{-fpic} and @option{-fPIC}, but
12916 generated position independent code can be only linked into executables.
12917 Usually these options are used when @option{-pie} GCC option will be
12918 used during linking.
12920 @item -fno-jump-tables
12921 @opindex fno-jump-tables
12922 Do not use jump tables for switch statements even where it would be
12923 more efficient than other code generation strategies. This option is
12924 of use in conjunction with @option{-fpic} or @option{-fPIC} for
12925 building code which forms part of a dynamic linker and cannot
12926 reference the address of a jump table. On some targets, jump tables
12927 do not require a GOT and this option is not needed.
12929 @item -ffixed-@var{reg}
12931 Treat the register named @var{reg} as a fixed register; generated code
12932 should never refer to it (except perhaps as a stack pointer, frame
12933 pointer or in some other fixed role).
12935 @var{reg} must be the name of a register. The register names accepted
12936 are machine-specific and are defined in the @code{REGISTER_NAMES}
12937 macro in the machine description macro file.
12939 This flag does not have a negative form, because it specifies a
12942 @item -fcall-used-@var{reg}
12943 @opindex fcall-used
12944 Treat the register named @var{reg} as an allocable register that is
12945 clobbered by function calls. It may be allocated for temporaries or
12946 variables that do not live across a call. Functions compiled this way
12947 will not save and restore the register @var{reg}.
12949 It is an error to used this flag with the frame pointer or stack pointer.
12950 Use of this flag for other registers that have fixed pervasive roles in
12951 the machine's execution model will produce disastrous results.
12953 This flag does not have a negative form, because it specifies a
12956 @item -fcall-saved-@var{reg}
12957 @opindex fcall-saved
12958 Treat the register named @var{reg} as an allocable register saved by
12959 functions. It may be allocated even for temporaries or variables that
12960 live across a call. Functions compiled this way will save and restore
12961 the register @var{reg} if they use it.
12963 It is an error to used this flag with the frame pointer or stack pointer.
12964 Use of this flag for other registers that have fixed pervasive roles in
12965 the machine's execution model will produce disastrous results.
12967 A different sort of disaster will result from the use of this flag for
12968 a register in which function values may be returned.
12970 This flag does not have a negative form, because it specifies a
12973 @item -fpack-struct[=@var{n}]
12974 @opindex fpack-struct
12975 Without a value specified, pack all structure members together without
12976 holes. When a value is specified (which must be a small power of two), pack
12977 structure members according to this value, representing the maximum
12978 alignment (that is, objects with default alignment requirements larger than
12979 this will be output potentially unaligned at the next fitting location.
12981 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12982 code that is not binary compatible with code generated without that switch.
12983 Additionally, it makes the code suboptimal.
12984 Use it to conform to a non-default application binary interface.
12986 @item -finstrument-functions
12987 @opindex finstrument-functions
12988 Generate instrumentation calls for entry and exit to functions. Just
12989 after function entry and just before function exit, the following
12990 profiling functions will be called with the address of the current
12991 function and its call site. (On some platforms,
12992 @code{__builtin_return_address} does not work beyond the current
12993 function, so the call site information may not be available to the
12994 profiling functions otherwise.)
12997 void __cyg_profile_func_enter (void *this_fn,
12999 void __cyg_profile_func_exit (void *this_fn,
13003 The first argument is the address of the start of the current function,
13004 which may be looked up exactly in the symbol table.
13006 This instrumentation is also done for functions expanded inline in other
13007 functions. The profiling calls will indicate where, conceptually, the
13008 inline function is entered and exited. This means that addressable
13009 versions of such functions must be available. If all your uses of a
13010 function are expanded inline, this may mean an additional expansion of
13011 code size. If you use @samp{extern inline} in your C code, an
13012 addressable version of such functions must be provided. (This is
13013 normally the case anyways, but if you get lucky and the optimizer always
13014 expands the functions inline, you might have gotten away without
13015 providing static copies.)
13017 A function may be given the attribute @code{no_instrument_function}, in
13018 which case this instrumentation will not be done. This can be used, for
13019 example, for the profiling functions listed above, high-priority
13020 interrupt routines, and any functions from which the profiling functions
13021 cannot safely be called (perhaps signal handlers, if the profiling
13022 routines generate output or allocate memory).
13024 @item -fstack-check
13025 @opindex fstack-check
13026 Generate code to verify that you do not go beyond the boundary of the
13027 stack. You should specify this flag if you are running in an
13028 environment with multiple threads, but only rarely need to specify it in
13029 a single-threaded environment since stack overflow is automatically
13030 detected on nearly all systems if there is only one stack.
13032 Note that this switch does not actually cause checking to be done; the
13033 operating system must do that. The switch causes generation of code
13034 to ensure that the operating system sees the stack being extended.
13036 @item -fstack-limit-register=@var{reg}
13037 @itemx -fstack-limit-symbol=@var{sym}
13038 @itemx -fno-stack-limit
13039 @opindex fstack-limit-register
13040 @opindex fstack-limit-symbol
13041 @opindex fno-stack-limit
13042 Generate code to ensure that the stack does not grow beyond a certain value,
13043 either the value of a register or the address of a symbol. If the stack
13044 would grow beyond the value, a signal is raised. For most targets,
13045 the signal is raised before the stack overruns the boundary, so
13046 it is possible to catch the signal without taking special precautions.
13048 For instance, if the stack starts at absolute address @samp{0x80000000}
13049 and grows downwards, you can use the flags
13050 @option{-fstack-limit-symbol=__stack_limit} and
13051 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
13052 of 128KB@. Note that this may only work with the GNU linker.
13054 @cindex aliasing of parameters
13055 @cindex parameters, aliased
13056 @item -fargument-alias
13057 @itemx -fargument-noalias
13058 @itemx -fargument-noalias-global
13059 @opindex fargument-alias
13060 @opindex fargument-noalias
13061 @opindex fargument-noalias-global
13062 Specify the possible relationships among parameters and between
13063 parameters and global data.
13065 @option{-fargument-alias} specifies that arguments (parameters) may
13066 alias each other and may alias global storage.@*
13067 @option{-fargument-noalias} specifies that arguments do not alias
13068 each other, but may alias global storage.@*
13069 @option{-fargument-noalias-global} specifies that arguments do not
13070 alias each other and do not alias global storage.
13072 Each language will automatically use whatever option is required by
13073 the language standard. You should not need to use these options yourself.
13075 @item -fleading-underscore
13076 @opindex fleading-underscore
13077 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
13078 change the way C symbols are represented in the object file. One use
13079 is to help link with legacy assembly code.
13081 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
13082 generate code that is not binary compatible with code generated without that
13083 switch. Use it to conform to a non-default application binary interface.
13084 Not all targets provide complete support for this switch.
13086 @item -ftls-model=@var{model}
13087 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
13088 The @var{model} argument should be one of @code{global-dynamic},
13089 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
13091 The default without @option{-fpic} is @code{initial-exec}; with
13092 @option{-fpic} the default is @code{global-dynamic}.
13094 @item -fvisibility=@var{default|internal|hidden|protected}
13095 @opindex fvisibility
13096 Set the default ELF image symbol visibility to the specified option---all
13097 symbols will be marked with this unless overridden within the code.
13098 Using this feature can very substantially improve linking and
13099 load times of shared object libraries, produce more optimized
13100 code, provide near-perfect API export and prevent symbol clashes.
13101 It is @strong{strongly} recommended that you use this in any shared objects
13104 Despite the nomenclature, @code{default} always means public ie;
13105 available to be linked against from outside the shared object.
13106 @code{protected} and @code{internal} are pretty useless in real-world
13107 usage so the only other commonly used option will be @code{hidden}.
13108 The default if @option{-fvisibility} isn't specified is
13109 @code{default}, i.e., make every
13110 symbol public---this causes the same behavior as previous versions of
13113 A good explanation of the benefits offered by ensuring ELF
13114 symbols have the correct visibility is given by ``How To Write
13115 Shared Libraries'' by Ulrich Drepper (which can be found at
13116 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
13117 solution made possible by this option to marking things hidden when
13118 the default is public is to make the default hidden and mark things
13119 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
13120 and @code{__attribute__ ((visibility("default")))} instead of
13121 @code{__declspec(dllexport)} you get almost identical semantics with
13122 identical syntax. This is a great boon to those working with
13123 cross-platform projects.
13125 For those adding visibility support to existing code, you may find
13126 @samp{#pragma GCC visibility} of use. This works by you enclosing
13127 the declarations you wish to set visibility for with (for example)
13128 @samp{#pragma GCC visibility push(hidden)} and
13129 @samp{#pragma GCC visibility pop}.
13130 Bear in mind that symbol visibility should be viewed @strong{as
13131 part of the API interface contract} and thus all new code should
13132 always specify visibility when it is not the default ie; declarations
13133 only for use within the local DSO should @strong{always} be marked explicitly
13134 as hidden as so to avoid PLT indirection overheads---making this
13135 abundantly clear also aids readability and self-documentation of the code.
13136 Note that due to ISO C++ specification requirements, operator new and
13137 operator delete must always be of default visibility.
13139 An overview of these techniques, their benefits and how to use them
13140 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
13146 @node Environment Variables
13147 @section Environment Variables Affecting GCC
13148 @cindex environment variables
13150 @c man begin ENVIRONMENT
13151 This section describes several environment variables that affect how GCC
13152 operates. Some of them work by specifying directories or prefixes to use
13153 when searching for various kinds of files. Some are used to specify other
13154 aspects of the compilation environment.
13156 Note that you can also specify places to search using options such as
13157 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
13158 take precedence over places specified using environment variables, which
13159 in turn take precedence over those specified by the configuration of GCC@.
13160 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
13161 GNU Compiler Collection (GCC) Internals}.
13166 @c @itemx LC_COLLATE
13168 @c @itemx LC_MONETARY
13169 @c @itemx LC_NUMERIC
13174 @c @findex LC_COLLATE
13175 @findex LC_MESSAGES
13176 @c @findex LC_MONETARY
13177 @c @findex LC_NUMERIC
13181 These environment variables control the way that GCC uses
13182 localization information that allow GCC to work with different
13183 national conventions. GCC inspects the locale categories
13184 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
13185 so. These locale categories can be set to any value supported by your
13186 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
13187 Kingdom encoded in UTF-8.
13189 The @env{LC_CTYPE} environment variable specifies character
13190 classification. GCC uses it to determine the character boundaries in
13191 a string; this is needed for some multibyte encodings that contain quote
13192 and escape characters that would otherwise be interpreted as a string
13195 The @env{LC_MESSAGES} environment variable specifies the language to
13196 use in diagnostic messages.
13198 If the @env{LC_ALL} environment variable is set, it overrides the value
13199 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
13200 and @env{LC_MESSAGES} default to the value of the @env{LANG}
13201 environment variable. If none of these variables are set, GCC
13202 defaults to traditional C English behavior.
13206 If @env{TMPDIR} is set, it specifies the directory to use for temporary
13207 files. GCC uses temporary files to hold the output of one stage of
13208 compilation which is to be used as input to the next stage: for example,
13209 the output of the preprocessor, which is the input to the compiler
13212 @item GCC_EXEC_PREFIX
13213 @findex GCC_EXEC_PREFIX
13214 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
13215 names of the subprograms executed by the compiler. No slash is added
13216 when this prefix is combined with the name of a subprogram, but you can
13217 specify a prefix that ends with a slash if you wish.
13219 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
13220 an appropriate prefix to use based on the pathname it was invoked with.
13222 If GCC cannot find the subprogram using the specified prefix, it
13223 tries looking in the usual places for the subprogram.
13225 The default value of @env{GCC_EXEC_PREFIX} is
13226 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
13227 of @code{prefix} when you ran the @file{configure} script.
13229 Other prefixes specified with @option{-B} take precedence over this prefix.
13231 This prefix is also used for finding files such as @file{crt0.o} that are
13234 In addition, the prefix is used in an unusual way in finding the
13235 directories to search for header files. For each of the standard
13236 directories whose name normally begins with @samp{/usr/local/lib/gcc}
13237 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
13238 replacing that beginning with the specified prefix to produce an
13239 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
13240 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
13241 These alternate directories are searched first; the standard directories
13244 @item COMPILER_PATH
13245 @findex COMPILER_PATH
13246 The value of @env{COMPILER_PATH} is a colon-separated list of
13247 directories, much like @env{PATH}. GCC tries the directories thus
13248 specified when searching for subprograms, if it can't find the
13249 subprograms using @env{GCC_EXEC_PREFIX}.
13252 @findex LIBRARY_PATH
13253 The value of @env{LIBRARY_PATH} is a colon-separated list of
13254 directories, much like @env{PATH}. When configured as a native compiler,
13255 GCC tries the directories thus specified when searching for special
13256 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
13257 using GCC also uses these directories when searching for ordinary
13258 libraries for the @option{-l} option (but directories specified with
13259 @option{-L} come first).
13263 @cindex locale definition
13264 This variable is used to pass locale information to the compiler. One way in
13265 which this information is used is to determine the character set to be used
13266 when character literals, string literals and comments are parsed in C and C++.
13267 When the compiler is configured to allow multibyte characters,
13268 the following values for @env{LANG} are recognized:
13272 Recognize JIS characters.
13274 Recognize SJIS characters.
13276 Recognize EUCJP characters.
13279 If @env{LANG} is not defined, or if it has some other value, then the
13280 compiler will use mblen and mbtowc as defined by the default locale to
13281 recognize and translate multibyte characters.
13285 Some additional environments variables affect the behavior of the
13288 @include cppenv.texi
13292 @node Precompiled Headers
13293 @section Using Precompiled Headers
13294 @cindex precompiled headers
13295 @cindex speed of compilation
13297 Often large projects have many header files that are included in every
13298 source file. The time the compiler takes to process these header files
13299 over and over again can account for nearly all of the time required to
13300 build the project. To make builds faster, GCC allows users to
13301 `precompile' a header file; then, if builds can use the precompiled
13302 header file they will be much faster.
13304 To create a precompiled header file, simply compile it as you would any
13305 other file, if necessary using the @option{-x} option to make the driver
13306 treat it as a C or C++ header file. You will probably want to use a
13307 tool like @command{make} to keep the precompiled header up-to-date when
13308 the headers it contains change.
13310 A precompiled header file will be searched for when @code{#include} is
13311 seen in the compilation. As it searches for the included file
13312 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
13313 compiler looks for a precompiled header in each directory just before it
13314 looks for the include file in that directory. The name searched for is
13315 the name specified in the @code{#include} with @samp{.gch} appended. If
13316 the precompiled header file can't be used, it is ignored.
13318 For instance, if you have @code{#include "all.h"}, and you have
13319 @file{all.h.gch} in the same directory as @file{all.h}, then the
13320 precompiled header file will be used if possible, and the original
13321 header will be used otherwise.
13323 Alternatively, you might decide to put the precompiled header file in a
13324 directory and use @option{-I} to ensure that directory is searched
13325 before (or instead of) the directory containing the original header.
13326 Then, if you want to check that the precompiled header file is always
13327 used, you can put a file of the same name as the original header in this
13328 directory containing an @code{#error} command.
13330 This also works with @option{-include}. So yet another way to use
13331 precompiled headers, good for projects not designed with precompiled
13332 header files in mind, is to simply take most of the header files used by
13333 a project, include them from another header file, precompile that header
13334 file, and @option{-include} the precompiled header. If the header files
13335 have guards against multiple inclusion, they will be skipped because
13336 they've already been included (in the precompiled header).
13338 If you need to precompile the same header file for different
13339 languages, targets, or compiler options, you can instead make a
13340 @emph{directory} named like @file{all.h.gch}, and put each precompiled
13341 header in the directory, perhaps using @option{-o}. It doesn't matter
13342 what you call the files in the directory, every precompiled header in
13343 the directory will be considered. The first precompiled header
13344 encountered in the directory that is valid for this compilation will
13345 be used; they're searched in no particular order.
13347 There are many other possibilities, limited only by your imagination,
13348 good sense, and the constraints of your build system.
13350 A precompiled header file can be used only when these conditions apply:
13354 Only one precompiled header can be used in a particular compilation.
13357 A precompiled header can't be used once the first C token is seen. You
13358 can have preprocessor directives before a precompiled header; you can
13359 even include a precompiled header from inside another header, so long as
13360 there are no C tokens before the @code{#include}.
13363 The precompiled header file must be produced for the same language as
13364 the current compilation. You can't use a C precompiled header for a C++
13368 The precompiled header file must have been produced by the same compiler
13369 binary as the current compilation is using.
13372 Any macros defined before the precompiled header is included must
13373 either be defined in the same way as when the precompiled header was
13374 generated, or must not affect the precompiled header, which usually
13375 means that they don't appear in the precompiled header at all.
13377 The @option{-D} option is one way to define a macro before a
13378 precompiled header is included; using a @code{#define} can also do it.
13379 There are also some options that define macros implicitly, like
13380 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
13383 @item If debugging information is output when using the precompiled
13384 header, using @option{-g} or similar, the same kind of debugging information
13385 must have been output when building the precompiled header. However,
13386 a precompiled header built using @option{-g} can be used in a compilation
13387 when no debugging information is being output.
13389 @item The same @option{-m} options must generally be used when building
13390 and using the precompiled header. @xref{Submodel Options},
13391 for any cases where this rule is relaxed.
13393 @item Each of the following options must be the same when building and using
13394 the precompiled header:
13396 @gccoptlist{-fexceptions -funit-at-a-time}
13399 Some other command-line options starting with @option{-f},
13400 @option{-p}, or @option{-O} must be defined in the same way as when
13401 the precompiled header was generated. At present, it's not clear
13402 which options are safe to change and which are not; the safest choice
13403 is to use exactly the same options when generating and using the
13404 precompiled header. The following are known to be safe:
13406 @gccoptlist{-fmessage-length= -fpreprocessed
13407 -fsched-interblock -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous
13408 -fsched-verbose=<number> -fschedule-insns -fvisibility=
13413 For all of these except the last, the compiler will automatically
13414 ignore the precompiled header if the conditions aren't met. If you
13415 find an option combination that doesn't work and doesn't cause the
13416 precompiled header to be ignored, please consider filing a bug report,
13419 If you do use differing options when generating and using the
13420 precompiled header, the actual behavior will be a mixture of the
13421 behavior for the options. For instance, if you use @option{-g} to
13422 generate the precompiled header but not when using it, you may or may
13423 not get debugging information for routines in the precompiled header.
13425 @node Running Protoize
13426 @section Running Protoize
13428 The program @code{protoize} is an optional part of GCC@. You can use
13429 it to add prototypes to a program, thus converting the program to ISO
13430 C in one respect. The companion program @code{unprotoize} does the
13431 reverse: it removes argument types from any prototypes that are found.
13433 When you run these programs, you must specify a set of source files as
13434 command line arguments. The conversion programs start out by compiling
13435 these files to see what functions they define. The information gathered
13436 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
13438 After scanning comes actual conversion. The specified files are all
13439 eligible to be converted; any files they include (whether sources or
13440 just headers) are eligible as well.
13442 But not all the eligible files are converted. By default,
13443 @code{protoize} and @code{unprotoize} convert only source and header
13444 files in the current directory. You can specify additional directories
13445 whose files should be converted with the @option{-d @var{directory}}
13446 option. You can also specify particular files to exclude with the
13447 @option{-x @var{file}} option. A file is converted if it is eligible, its
13448 directory name matches one of the specified directory names, and its
13449 name within the directory has not been excluded.
13451 Basic conversion with @code{protoize} consists of rewriting most
13452 function definitions and function declarations to specify the types of
13453 the arguments. The only ones not rewritten are those for varargs
13456 @code{protoize} optionally inserts prototype declarations at the
13457 beginning of the source file, to make them available for any calls that
13458 precede the function's definition. Or it can insert prototype
13459 declarations with block scope in the blocks where undeclared functions
13462 Basic conversion with @code{unprotoize} consists of rewriting most
13463 function declarations to remove any argument types, and rewriting
13464 function definitions to the old-style pre-ISO form.
13466 Both conversion programs print a warning for any function declaration or
13467 definition that they can't convert. You can suppress these warnings
13470 The output from @code{protoize} or @code{unprotoize} replaces the
13471 original source file. The original file is renamed to a name ending
13472 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
13473 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
13474 for DOS) file already exists, then the source file is simply discarded.
13476 @code{protoize} and @code{unprotoize} both depend on GCC itself to
13477 scan the program and collect information about the functions it uses.
13478 So neither of these programs will work until GCC is installed.
13480 Here is a table of the options you can use with @code{protoize} and
13481 @code{unprotoize}. Each option works with both programs unless
13485 @item -B @var{directory}
13486 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
13487 usual directory (normally @file{/usr/local/lib}). This file contains
13488 prototype information about standard system functions. This option
13489 applies only to @code{protoize}.
13491 @item -c @var{compilation-options}
13492 Use @var{compilation-options} as the options when running @command{gcc} to
13493 produce the @samp{.X} files. The special option @option{-aux-info} is
13494 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
13496 Note that the compilation options must be given as a single argument to
13497 @code{protoize} or @code{unprotoize}. If you want to specify several
13498 @command{gcc} options, you must quote the entire set of compilation options
13499 to make them a single word in the shell.
13501 There are certain @command{gcc} arguments that you cannot use, because they
13502 would produce the wrong kind of output. These include @option{-g},
13503 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
13504 the @var{compilation-options}, they are ignored.
13507 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
13508 systems) instead of @samp{.c}. This is convenient if you are converting
13509 a C program to C++. This option applies only to @code{protoize}.
13512 Add explicit global declarations. This means inserting explicit
13513 declarations at the beginning of each source file for each function
13514 that is called in the file and was not declared. These declarations
13515 precede the first function definition that contains a call to an
13516 undeclared function. This option applies only to @code{protoize}.
13518 @item -i @var{string}
13519 Indent old-style parameter declarations with the string @var{string}.
13520 This option applies only to @code{protoize}.
13522 @code{unprotoize} converts prototyped function definitions to old-style
13523 function definitions, where the arguments are declared between the
13524 argument list and the initial @samp{@{}. By default, @code{unprotoize}
13525 uses five spaces as the indentation. If you want to indent with just
13526 one space instead, use @option{-i " "}.
13529 Keep the @samp{.X} files. Normally, they are deleted after conversion
13533 Add explicit local declarations. @code{protoize} with @option{-l} inserts
13534 a prototype declaration for each function in each block which calls the
13535 function without any declaration. This option applies only to
13539 Make no real changes. This mode just prints information about the conversions
13540 that would have been done without @option{-n}.
13543 Make no @samp{.save} files. The original files are simply deleted.
13544 Use this option with caution.
13546 @item -p @var{program}
13547 Use the program @var{program} as the compiler. Normally, the name
13548 @file{gcc} is used.
13551 Work quietly. Most warnings are suppressed.
13554 Print the version number, just like @option{-v} for @command{gcc}.
13557 If you need special compiler options to compile one of your program's
13558 source files, then you should generate that file's @samp{.X} file
13559 specially, by running @command{gcc} on that source file with the
13560 appropriate options and the option @option{-aux-info}. Then run
13561 @code{protoize} on the entire set of files. @code{protoize} will use
13562 the existing @samp{.X} file because it is newer than the source file.
13566 gcc -Dfoo=bar file1.c -aux-info file1.X
13571 You need to include the special files along with the rest in the
13572 @code{protoize} command, even though their @samp{.X} files already
13573 exist, because otherwise they won't get converted.
13575 @xref{Protoize Caveats}, for more information on how to use
13576 @code{protoize} successfully.