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, @option{-fforce-mem},
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 -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 @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-exceptions @gol
201 -freplace-objc-classes @gol
204 -Wno-protocol -Wselector -Wundeclared-selector}
206 @item Language Independent Options
207 @xref{Language Independent Options,,Options to Control Diagnostic Messages Formatting}.
208 @gccoptlist{-fmessage-length=@var{n} @gol
209 -fdiagnostics-show-location=@r{[}once@r{|}every-line@r{]}}
211 @item Warning Options
212 @xref{Warning Options,,Options to Request or Suppress Warnings}.
213 @gccoptlist{-fsyntax-only -pedantic -pedantic-errors @gol
214 -w -Wextra -Wall -Waggregate-return @gol
215 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment @gol
216 -Wconversion -Wno-deprecated-declarations @gol
217 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels @gol
218 -Werror -Werror-implicit-function-declaration @gol
219 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2 @gol
220 -Wno-format-extra-args -Wformat-nonliteral @gol
221 -Wformat-security -Wformat-y2k @gol
222 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int @gol
223 -Wimport -Wno-import -Winit-self -Winline @gol
224 -Wno-invalid-offsetof -Winvalid-pch @gol
225 -Wlarger-than-@var{len} -Wlong-long @gol
226 -Wmain -Wmissing-braces -Wmissing-field-initializers @gol
227 -Wmissing-format-attribute -Wmissing-include-dirs @gol
228 -Wmissing-noreturn @gol
229 -Wno-multichar -Wnonnull -Wpacked -Wpadded @gol
230 -Wparentheses -Wpointer-arith -Wredundant-decls @gol
231 -Wreturn-type -Wsequence-point -Wshadow @gol
232 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2 @gol
233 -Wswitch -Wswitch-default -Wswitch-enum @gol
234 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized @gol
235 -Wunknown-pragmas -Wunreachable-code @gol
236 -Wunused -Wunused-function -Wunused-label -Wunused-parameter @gol
237 -Wunused-value -Wunused-variable -Wwrite-strings @gol
240 @item C-only Warning Options
241 @gccoptlist{-Wbad-function-cast -Wmissing-declarations @gol
242 -Wmissing-prototypes -Wnested-externs -Wold-style-definition @gol
243 -Wstrict-prototypes -Wtraditional @gol
244 -Wdeclaration-after-statement -Wno-pointer-sign}
246 @item Debugging Options
247 @xref{Debugging Options,,Options for Debugging Your Program or GCC}.
248 @gccoptlist{-d@var{letters} -dumpspecs -dumpmachine -dumpversion @gol
249 -fdump-unnumbered -fdump-translation-unit@r{[}-@var{n}@r{]} @gol
250 -fdump-class-hierarchy@r{[}-@var{n}@r{]} @gol
251 -fdump-ipa-all -fdump-ipa-cgraph @gol
253 -fdump-tree-original@r{[}-@var{n}@r{]} @gol
254 -fdump-tree-optimized@r{[}-@var{n}@r{]} @gol
255 -fdump-tree-inlined@r{[}-@var{n}@r{]} @gol
256 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias @gol
258 -fdump-tree-ssa@r{[}-@var{n}@r{]} -fdump-tree-pre@r{[}-@var{n}@r{]} @gol
259 -fdump-tree-ccp@r{[}-@var{n}@r{]} -fdump-tree-dce@r{[}-@var{n}@r{]} @gol
260 -fdump-tree-gimple@r{[}-raw@r{]} -fdump-tree-mudflap@r{[}-@var{n}@r{]} @gol
261 -fdump-tree-dom@r{[}-@var{n}@r{]} @gol
262 -fdump-tree-dse@r{[}-@var{n}@r{]} @gol
263 -fdump-tree-phiopt@r{[}-@var{n}@r{]} @gol
264 -fdump-tree-forwprop@r{[}-@var{n}@r{]} @gol
265 -fdump-tree-copyrename@r{[}-@var{n}@r{]} @gol
266 -fdump-tree-nrv -fdump-tree-vect @gol
267 -fdump-tree-sink @gol
268 -fdump-tree-sra@r{[}-@var{n}@r{]} @gol
269 -fdump-tree-salias @gol
270 -fdump-tree-fre@r{[}-@var{n}@r{]} @gol
271 -ftree-vectorizer-verbose=@var{n} @gol
272 -feliminate-dwarf2-dups -feliminate-unused-debug-types @gol
273 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling @gol
274 -frandom-seed=@var{string} -fsched-verbose=@var{n} @gol
275 -ftest-coverage -ftime-report -fvar-tracking @gol
276 -g -g@var{level} -gcoff -gdwarf-2 @gol
277 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+ @gol
278 -p -pg -print-file-name=@var{library} -print-libgcc-file-name @gol
279 -print-multi-directory -print-multi-lib @gol
280 -print-prog-name=@var{program} -print-search-dirs -Q @gol
283 @item Optimization Options
284 @xref{Optimize Options,,Options that Control Optimization}.
285 @gccoptlist{-falign-functions=@var{n} -falign-jumps=@var{n} @gol
286 -falign-labels=@var{n} -falign-loops=@var{n} @gol
287 -fbounds-check -fmudflap -fmudflapth -fmudflapir @gol
288 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize @gol
289 -fbranch-target-load-optimize2 -fbtr-bb-exclusive @gol
290 -fcaller-saves -fcprop-registers -fcse-follow-jumps @gol
291 -fcse-skip-blocks -fcx-limited-range -fdata-sections @gol
292 -fdelayed-branch -fdelete-null-pointer-checks @gol
293 -fexpensive-optimizations -ffast-math -ffloat-store @gol
294 -fforce-addr -fforce-mem -ffunction-sections @gol
295 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload @gol
296 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2 @gol
297 -finline-functions -finline-limit=@var{n} -fkeep-inline-functions @gol
298 -fkeep-static-consts -fmerge-constants -fmerge-all-constants @gol
299 -fmodulo-sched -fno-branch-count-reg @gol
300 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants @gol
301 -fno-function-cse -fno-guess-branch-probability @gol
302 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2 @gol
303 -funsafe-math-optimizations -ffinite-math-only @gol
304 -fno-trapping-math -fno-zero-initialized-in-bss @gol
305 -fomit-frame-pointer -foptimize-register-move @gol
306 -foptimize-sibling-calls -fprefetch-loop-arrays @gol
307 -fprofile-generate -fprofile-use @gol
308 -fregmove -frename-registers @gol
309 -freorder-blocks -freorder-blocks-and-partition -freorder-functions @gol
310 -frerun-cse-after-loop -frerun-loop-opt @gol
311 -frounding-math -fschedule-insns -fschedule-insns2 @gol
312 -fno-sched-interblock -fno-sched-spec -fsched-spec-load @gol
313 -fsched-spec-load-dangerous @gol
314 -fsched-stalled-insns=@var{n} -sched-stalled-insns-dep=@var{n} @gol
315 -fsched2-use-superblocks @gol
316 -fsched2-use-traces -freschedule-modulo-scheduled-loops @gol
317 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching @gol
318 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps @gol
319 -funroll-all-loops -funroll-loops -fpeel-loops @gol
320 -fsplit-ivs-in-unroller -funswitch-loops @gol
321 -fvariable-expansion-in-unroller @gol
322 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize @gol
323 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts @gol
324 -ftree-dominator-opts -ftree-dse -ftree-copyrename -ftree-sink @gol
325 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize @gol
326 -ftree-salias -fweb @gol
327 --param @var{name}=@var{value}
328 -O -O0 -O1 -O2 -O3 -Os}
330 @item Preprocessor Options
331 @xref{Preprocessor Options,,Options Controlling the Preprocessor}.
332 @gccoptlist{-A@var{question}=@var{answer} @gol
333 -A-@var{question}@r{[}=@var{answer}@r{]} @gol
334 -C -dD -dI -dM -dN @gol
335 -D@var{macro}@r{[}=@var{defn}@r{]} -E -H @gol
336 -idirafter @var{dir} @gol
337 -include @var{file} -imacros @var{file} @gol
338 -iprefix @var{file} -iwithprefix @var{dir} @gol
339 -iwithprefixbefore @var{dir} -isystem @var{dir} @gol
340 -M -MM -MF -MG -MP -MQ -MT -nostdinc @gol
341 -P -fworking-directory -remap @gol
342 -trigraphs -undef -U@var{macro} -Wp,@var{option} @gol
343 -Xpreprocessor @var{option}}
345 @item Assembler Option
346 @xref{Assembler Options,,Passing Options to the Assembler}.
347 @gccoptlist{-Wa,@var{option} -Xassembler @var{option}}
350 @xref{Link Options,,Options for Linking}.
351 @gccoptlist{@var{object-file-name} -l@var{library} @gol
352 -nostartfiles -nodefaultlibs -nostdlib -pie @gol
353 -s -static -static-libgcc -shared -shared-libgcc -symbolic @gol
354 -Wl,@var{option} -Xlinker @var{option} @gol
357 @item Directory Options
358 @xref{Directory Options,,Options for Directory Search}.
359 @gccoptlist{-B@var{prefix} -I@var{dir} -iquote@var{dir} -L@var{dir} -specs=@var{file} -I-}
362 @c I wrote this xref this way to avoid overfull hbox. -- rms
363 @xref{Target Options}.
364 @gccoptlist{-V @var{version} -b @var{machine}}
366 @item Machine Dependent Options
367 @xref{Submodel Options,,Hardware Models and Configurations}.
368 @c This list is ordered alphanumerically by subsection name.
369 @c Try and put the significant identifier (CPU or system) first,
370 @c so users have a clue at guessing where the ones they want will be.
373 @gccoptlist{-EB -EL @gol
374 -mmangle-cpu -mcpu=@var{cpu} -mtext=@var{text-section} @gol
375 -mdata=@var{data-section} -mrodata=@var{readonly-data-section}}
378 @gccoptlist{-mapcs-frame -mno-apcs-frame @gol
379 -mabi=@var{name} @gol
380 -mapcs-stack-check -mno-apcs-stack-check @gol
381 -mapcs-float -mno-apcs-float @gol
382 -mapcs-reentrant -mno-apcs-reentrant @gol
383 -msched-prolog -mno-sched-prolog @gol
384 -mlittle-endian -mbig-endian -mwords-little-endian @gol
385 -mfloat-abi=@var{name} -msoft-float -mhard-float -mfpe @gol
386 -mthumb-interwork -mno-thumb-interwork @gol
387 -mcpu=@var{name} -march=@var{name} -mfpu=@var{name} @gol
388 -mstructure-size-boundary=@var{n} @gol
389 -mabort-on-noreturn @gol
390 -mlong-calls -mno-long-calls @gol
391 -msingle-pic-base -mno-single-pic-base @gol
392 -mpic-register=@var{reg} @gol
393 -mnop-fun-dllimport @gol
394 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns @gol
395 -mpoke-function-name @gol
397 -mtpcs-frame -mtpcs-leaf-frame @gol
398 -mcaller-super-interworking -mcallee-super-interworking}
401 @gccoptlist{-mmcu=@var{mcu} -msize -minit-stack=@var{n} -mno-interrupts @gol
402 -mcall-prologues -mno-tablejump -mtiny-stack -mint8}
404 @emph{Blackfin Options}
405 @gccoptlist{-momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync @gol
406 -mno-csync -mlow-64k -mno-low64k -mid-shared-library @gol
407 -mno-id-shared-library -mshared-library-id=@var{n} @gol}
410 @gccoptlist{-mcpu=@var{cpu} -march=@var{cpu} -mtune=@var{cpu} @gol
411 -mmax-stack-frame=@var{n} -melinux-stacksize=@var{n} @gol
412 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects @gol
413 -mstack-align -mdata-align -mconst-align @gol
414 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt @gol
415 -melf -maout -melinux -mlinux -sim -sim2 @gol
416 -mmul-bug-workaround -mno-mul-bug-workaround}
418 @emph{Darwin Options}
419 @gccoptlist{-all_load -allowable_client -arch -arch_errors_fatal @gol
420 -arch_only -bind_at_load -bundle -bundle_loader @gol
421 -client_name -compatibility_version -current_version @gol
423 -dependency-file -dylib_file -dylinker_install_name @gol
424 -dynamic -dynamiclib -exported_symbols_list @gol
425 -filelist -flat_namespace -force_cpusubtype_ALL @gol
426 -force_flat_namespace -headerpad_max_install_names @gol
427 -image_base -init -install_name -keep_private_externs @gol
428 -multi_module -multiply_defined -multiply_defined_unused @gol
429 -noall_load -no_dead_strip_inits_and_terms @gol
430 -nofixprebinding -nomultidefs -noprebind -noseglinkedit @gol
431 -pagezero_size -prebind -prebind_all_twolevel_modules @gol
432 -private_bundle -read_only_relocs -sectalign @gol
433 -sectobjectsymbols -whyload -seg1addr @gol
434 -sectcreate -sectobjectsymbols -sectorder @gol
435 -segaddr -segs_read_only_addr -segs_read_write_addr @gol
436 -seg_addr_table -seg_addr_table_filename -seglinkedit @gol
437 -segprot -segs_read_only_addr -segs_read_write_addr @gol
438 -single_module -static -sub_library -sub_umbrella @gol
439 -twolevel_namespace -umbrella -undefined @gol
440 -unexported_symbols_list -weak_reference_mismatches @gol
441 -whatsloaded -F -gused -gfull -mone-byte-bool}
443 @emph{DEC Alpha Options}
444 @gccoptlist{-mno-fp-regs -msoft-float -malpha-as -mgas @gol
445 -mieee -mieee-with-inexact -mieee-conformant @gol
446 -mfp-trap-mode=@var{mode} -mfp-rounding-mode=@var{mode} @gol
447 -mtrap-precision=@var{mode} -mbuild-constants @gol
448 -mcpu=@var{cpu-type} -mtune=@var{cpu-type} @gol
449 -mbwx -mmax -mfix -mcix @gol
450 -mfloat-vax -mfloat-ieee @gol
451 -mexplicit-relocs -msmall-data -mlarge-data @gol
452 -msmall-text -mlarge-text @gol
453 -mmemory-latency=@var{time}}
455 @emph{DEC Alpha/VMS Options}
456 @gccoptlist{-mvms-return-codes}
459 @gccoptlist{-mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 @gol
460 -mhard-float -msoft-float @gol
461 -malloc-cc -mfixed-cc -mdword -mno-dword @gol
462 -mdouble -mno-double @gol
463 -mmedia -mno-media -mmuladd -mno-muladd @gol
464 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic @gol
465 -mlinked-fp -mlong-calls -malign-labels @gol
466 -mlibrary-pic -macc-4 -macc-8 @gol
467 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move @gol
468 -mscc -mno-scc -mcond-exec -mno-cond-exec @gol
469 -mvliw-branch -mno-vliw-branch @gol
470 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec @gol
471 -mno-nested-cond-exec -mtomcat-stats @gol
475 @emph{H8/300 Options}
476 @gccoptlist{-mrelax -mh -ms -mn -mint32 -malign-300}
479 @gccoptlist{-march=@var{architecture-type} @gol
480 -mbig-switch -mdisable-fpregs -mdisable-indexing @gol
481 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld @gol
482 -mfixed-range=@var{register-range} @gol
483 -mjump-in-delay -mlinker-opt -mlong-calls @gol
484 -mlong-load-store -mno-big-switch -mno-disable-fpregs @gol
485 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas @gol
486 -mno-jump-in-delay -mno-long-load-store @gol
487 -mno-portable-runtime -mno-soft-float @gol
488 -mno-space-regs -msoft-float -mpa-risc-1-0 @gol
489 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime @gol
490 -mschedule=@var{cpu-type} -mspace-regs -msio -mwsio @gol
491 -munix=@var{unix-std} -nolibdld -static -threads}
493 @emph{i386 and x86-64 Options}
494 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
495 -mfpmath=@var{unit} @gol
496 -masm=@var{dialect} -mno-fancy-math-387 @gol
497 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib @gol
498 -mno-wide-multiply -mrtd -malign-double @gol
499 -mpreferred-stack-boundary=@var{num} @gol
500 -mmmx -msse -msse2 -msse3 -m3dnow @gol
501 -mthreads -mno-align-stringops -minline-all-stringops @gol
502 -mpush-args -maccumulate-outgoing-args -m128bit-long-double @gol
503 -m96bit-long-double -mregparm=@var{num} -momit-leaf-frame-pointer @gol
504 -mno-red-zone -mno-tls-direct-seg-refs @gol
505 -mcmodel=@var{code-model} @gol
509 @gccoptlist{-mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic @gol
510 -mvolatile-asm-stop -mregister-names -mno-sdata @gol
511 -mconstant-gp -mauto-pic -minline-float-divide-min-latency @gol
512 -minline-float-divide-max-throughput @gol
513 -minline-int-divide-min-latency @gol
514 -minline-int-divide-max-throughput @gol
515 -minline-sqrt-min-latency -minline-sqrt-max-throughput @gol
516 -mno-dwarf2-asm -mearly-stop-bits @gol
517 -mfixed-range=@var{register-range} -mtls-size=@var{tls-size} @gol
518 -mtune=@var{cpu-type} -mt -pthread -milp32 -mlp64}
520 @emph{M32R/D Options}
521 @gccoptlist{-m32r2 -m32rx -m32r @gol
523 -malign-loops -mno-align-loops @gol
524 -missue-rate=@var{number} @gol
525 -mbranch-cost=@var{number} @gol
526 -mmodel=@var{code-size-model-type} @gol
527 -msdata=@var{sdata-type} @gol
528 -mno-flush-func -mflush-func=@var{name} @gol
529 -mno-flush-trap -mflush-trap=@var{number} @gol
532 @emph{M680x0 Options}
533 @gccoptlist{-m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 @gol
534 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020 @gol
535 -mnobitfield -mrtd -mshort -msoft-float -mpcrel @gol
536 -malign-int -mstrict-align -msep-data -mno-sep-data @gol
537 -mshared-library-id=n -mid-shared-library -mno-id-shared-library}
539 @emph{M68hc1x Options}
540 @gccoptlist{-m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12 @gol
541 -mauto-incdec -minmax -mlong-calls -mshort @gol
542 -msoft-reg-count=@var{count}}
545 @gccoptlist{-mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates @gol
546 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields @gol
547 -m4byte-functions -mno-4byte-functions -mcallgraph-data @gol
548 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim @gol
549 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment}
552 @gccoptlist{-EL -EB -march=@var{arch} -mtune=@var{arch} @gol
553 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64 @gol
554 -mips16 -mno-mips16 -mabi=@var{abi} -mabicalls -mno-abicalls @gol
555 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64 @gol
556 -mhard-float -msoft-float -msingle-float -mdouble-float @gol
557 -mpaired-single -mips3d @gol
558 -mint64 -mlong64 -mlong32 -msym32 -mno-sym32 @gol
559 -G@var{num} -membedded-data -mno-embedded-data @gol
560 -muninit-const-in-rodata -mno-uninit-const-in-rodata @gol
561 -msplit-addresses -mno-split-addresses @gol
562 -mexplicit-relocs -mno-explicit-relocs @gol
563 -mcheck-zero-division -mno-check-zero-division @gol
564 -mdivide-traps -mdivide-breaks @gol
565 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls @gol
566 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp @gol
567 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 @gol
568 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130 @gol
569 -mfix-sb1 -mno-fix-sb1 @gol
570 -mflush-func=@var{func} -mno-flush-func @gol
571 -mbranch-likely -mno-branch-likely @gol
572 -mfp-exceptions -mno-fp-exceptions @gol
573 -mvr4130-align -mno-vr4130-align}
576 @gccoptlist{-mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu @gol
577 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols @gol
578 -melf -mbranch-predict -mno-branch-predict -mbase-addresses @gol
579 -mno-base-addresses -msingle-exit -mno-single-exit}
581 @emph{MN10300 Options}
582 @gccoptlist{-mmult-bug -mno-mult-bug @gol
583 -mam33 -mno-am33 @gol
584 -mam33-2 -mno-am33-2 @gol
588 @gccoptlist{-m32032 -m32332 -m32532 -m32081 -m32381 @gol
589 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd @gol
590 -mregparam -mnoregparam -msb -mnosb @gol
591 -mbitfield -mnobitfield -mhimem -mnohimem}
593 @emph{PDP-11 Options}
594 @gccoptlist{-mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 @gol
595 -mbcopy -mbcopy-builtin -mint32 -mno-int16 @gol
596 -mint16 -mno-int32 -mfloat32 -mno-float64 @gol
597 -mfloat64 -mno-float32 -mabshi -mno-abshi @gol
598 -mbranch-expensive -mbranch-cheap @gol
599 -msplit -mno-split -munix-asm -mdec-asm}
601 @emph{PowerPC Options}
602 See RS/6000 and PowerPC Options.
604 @emph{RS/6000 and PowerPC Options}
605 @gccoptlist{-mcpu=@var{cpu-type} @gol
606 -mtune=@var{cpu-type} @gol
607 -mpower -mno-power -mpower2 -mno-power2 @gol
608 -mpowerpc -mpowerpc64 -mno-powerpc @gol
609 -maltivec -mno-altivec @gol
610 -mpowerpc-gpopt -mno-powerpc-gpopt @gol
611 -mpowerpc-gfxopt -mno-powerpc-gfxopt @gol
612 -mnew-mnemonics -mold-mnemonics @gol
613 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc @gol
614 -m64 -m32 -mxl-compat -mno-xl-compat -mpe @gol
615 -malign-power -malign-natural @gol
616 -msoft-float -mhard-float -mmultiple -mno-multiple @gol
617 -mstring -mno-string -mupdate -mno-update @gol
618 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align @gol
619 -mstrict-align -mno-strict-align -mrelocatable @gol
620 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib @gol
621 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian @gol
622 -mdynamic-no-pic @gol
623 -mprioritize-restricted-insns=@var{priority} @gol
624 -msched-costly-dep=@var{dependence_type} @gol
625 -minsert-sched-nops=@var{scheme} @gol
626 -mcall-sysv -mcall-netbsd @gol
627 -maix-struct-return -msvr4-struct-return @gol
628 -mabi=altivec -mabi=no-altivec @gol
629 -mabi=spe -mabi=no-spe @gol
630 -misel=yes -misel=no @gol
631 -mspe=yes -mspe=no @gol
632 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double @gol
633 -mprototype -mno-prototype @gol
634 -msim -mmvme -mads -myellowknife -memb -msdata @gol
635 -msdata=@var{opt} -mvxworks -mwindiss -G @var{num} -pthread}
637 @emph{S/390 and zSeries Options}
638 @gccoptlist{-mtune=@var{cpu-type} -march=@var{cpu-type} @gol
639 -mhard-float -msoft-float -mbackchain -mno-backchain @gol
640 -mpacked-stack -mno-packed-stack @gol
641 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle @gol
642 -m64 -m31 -mdebug -mno-debug -mesa -mzarch @gol
643 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd @gol
644 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard}
647 @gccoptlist{-m1 -m2 -m2e -m3 -m3e @gol
648 -m4-nofpu -m4-single-only -m4-single -m4 @gol
649 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al @gol
650 -m5-64media -m5-64media-nofpu @gol
651 -m5-32media -m5-32media-nofpu @gol
652 -m5-compact -m5-compact-nofpu @gol
653 -mb -ml -mdalign -mrelax @gol
654 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave @gol
655 -mieee -misize -mpadstruct -mspace @gol
656 -mprefergot -musermode}
659 @gccoptlist{-mcpu=@var{cpu-type} @gol
660 -mtune=@var{cpu-type} @gol
661 -mcmodel=@var{code-model} @gol
662 -m32 -m64 -mapp-regs -mno-app-regs @gol
663 -mfaster-structs -mno-faster-structs @gol
664 -mfpu -mno-fpu -mhard-float -msoft-float @gol
665 -mhard-quad-float -msoft-quad-float @gol
666 -mimpure-text -mno-impure-text -mlittle-endian @gol
667 -mstack-bias -mno-stack-bias @gol
668 -munaligned-doubles -mno-unaligned-doubles @gol
669 -mv8plus -mno-v8plus -mvis -mno-vis
672 @emph{System V Options}
673 @gccoptlist{-Qy -Qn -YP,@var{paths} -Ym,@var{dir}}
675 @emph{TMS320C3x/C4x Options}
676 @gccoptlist{-mcpu=@var{cpu} -mbig -msmall -mregparm -mmemparm @gol
677 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload @gol
678 -mrpts=@var{count} -mrptb -mdb -mloop-unsigned @gol
679 -mparallel-insns -mparallel-mpy -mpreserve-float}
682 @gccoptlist{-mlong-calls -mno-long-calls -mep -mno-ep @gol
683 -mprolog-function -mno-prolog-function -mspace @gol
684 -mtda=@var{n} -msda=@var{n} -mzda=@var{n} @gol
685 -mapp-regs -mno-app-regs @gol
686 -mdisable-callt -mno-disable-callt @gol
692 @gccoptlist{-mg -mgnu -munix}
694 @emph{x86-64 Options}
695 See i386 and x86-64 Options.
697 @emph{Xstormy16 Options}
700 @emph{Xtensa Options}
701 @gccoptlist{-mconst16 -mno-const16 @gol
702 -mfused-madd -mno-fused-madd @gol
703 -mtext-section-literals -mno-text-section-literals @gol
704 -mtarget-align -mno-target-align @gol
705 -mlongcalls -mno-longcalls}
707 @emph{zSeries Options}
708 See S/390 and zSeries Options.
710 @item Code Generation Options
711 @xref{Code Gen Options,,Options for Code Generation Conventions}.
712 @gccoptlist{-fcall-saved-@var{reg} -fcall-used-@var{reg} @gol
713 -ffixed-@var{reg} -fexceptions @gol
714 -fnon-call-exceptions -funwind-tables @gol
715 -fasynchronous-unwind-tables @gol
716 -finhibit-size-directive -finstrument-functions @gol
717 -fno-common -fno-ident @gol
718 -fpcc-struct-return -fpic -fPIC -fpie -fPIE @gol
719 -freg-struct-return -fshared-data -fshort-enums @gol
720 -fshort-double -fshort-wchar @gol
721 -fverbose-asm -fpack-struct[=@var{n}] -fstack-check @gol
722 -fstack-limit-register=@var{reg} -fstack-limit-symbol=@var{sym} @gol
723 -fargument-alias -fargument-noalias @gol
724 -fargument-noalias-global -fleading-underscore @gol
725 -ftls-model=@var{model} @gol
726 -ftrapv -fwrapv -fbounds-check @gol
731 * Overall Options:: Controlling the kind of output:
732 an executable, object files, assembler files,
733 or preprocessed source.
734 * C Dialect Options:: Controlling the variant of C language compiled.
735 * C++ Dialect Options:: Variations on C++.
736 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
738 * Language Independent Options:: Controlling how diagnostics should be
740 * Warning Options:: How picky should the compiler be?
741 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
742 * Optimize Options:: How much optimization?
743 * Preprocessor Options:: Controlling header files and macro definitions.
744 Also, getting dependency information for Make.
745 * Assembler Options:: Passing options to the assembler.
746 * Link Options:: Specifying libraries and so on.
747 * Directory Options:: Where to find header files and libraries.
748 Where to find the compiler executable files.
749 * Spec Files:: How to pass switches to sub-processes.
750 * Target Options:: Running a cross-compiler, or an old version of GCC.
753 @node Overall Options
754 @section Options Controlling the Kind of Output
756 Compilation can involve up to four stages: preprocessing, compilation
757 proper, assembly and linking, always in that order. GCC is capable of
758 preprocessing and compiling several files either into several
759 assembler input files, or into one assembler input file; then each
760 assembler input file produces an object file, and linking combines all
761 the object files (those newly compiled, and those specified as input)
762 into an executable file.
764 @cindex file name suffix
765 For any given input file, the file name suffix determines what kind of
770 C source code which must be preprocessed.
773 C source code which should not be preprocessed.
776 C++ source code which should not be preprocessed.
779 Objective-C source code. Note that you must link with the @file{libobjc}
780 library to make an Objective-C program work.
783 Objective-C source code which should not be preprocessed.
787 Objective-C++ source code. Note that you must link with the @file{libobjc}
788 library to make an Objective-C++ program work. Note that @samp{.M} refers
789 to a literal capital M@.
792 Objective-C++ source code which should not be preprocessed.
795 C, C++, Objective-C or Objective-C++ header file to be turned into a
800 @itemx @var{file}.cxx
801 @itemx @var{file}.cpp
802 @itemx @var{file}.CPP
803 @itemx @var{file}.c++
805 C++ source code which must be preprocessed. Note that in @samp{.cxx},
806 the last two letters must both be literally @samp{x}. Likewise,
807 @samp{.C} refers to a literal capital C@.
811 C++ header file to be turned into a precompiled header.
814 @itemx @var{file}.for
815 @itemx @var{file}.FOR
816 Fortran source code which should not be preprocessed.
819 @itemx @var{file}.fpp
820 @itemx @var{file}.FPP
821 Fortran source code which must be preprocessed (with the traditional
825 Fortran source code which must be preprocessed with a RATFOR
826 preprocessor (not included with GCC)@.
829 @itemx @var{file}.f95
830 Fortran 90/95 source code which should not be preprocessed.
832 @c FIXME: Descriptions of Java file types.
839 Ada source code file which contains a library unit declaration (a
840 declaration of a package, subprogram, or generic, or a generic
841 instantiation), or a library unit renaming declaration (a package,
842 generic, or subprogram renaming declaration). Such files are also
845 @itemx @var{file}.adb
846 Ada source code file containing a library unit body (a subprogram or
847 package body). Such files are also called @dfn{bodies}.
849 @c GCC also knows about some suffixes for languages not yet included:
858 Assembler code which must be preprocessed.
861 An object file to be fed straight into linking.
862 Any file name with no recognized suffix is treated this way.
866 You can specify the input language explicitly with the @option{-x} option:
869 @item -x @var{language}
870 Specify explicitly the @var{language} for the following input files
871 (rather than letting the compiler choose a default based on the file
872 name suffix). This option applies to all following input files until
873 the next @option{-x} option. Possible values for @var{language} are:
875 c c-header c-cpp-output
876 c++ c++-header c++-cpp-output
877 objective-c objective-c-header objective-c-cpp-output
878 objective-c++ objective-c++-header objective-c++-cpp-output
879 assembler assembler-with-cpp
881 f77 f77-cpp-input ratfor
888 Turn off any specification of a language, so that subsequent files are
889 handled according to their file name suffixes (as they are if @option{-x}
890 has not been used at all).
892 @item -pass-exit-codes
893 @opindex pass-exit-codes
894 Normally the @command{gcc} program will exit with the code of 1 if any
895 phase of the compiler returns a non-success return code. If you specify
896 @option{-pass-exit-codes}, the @command{gcc} program will instead return with
897 numerically highest error produced by any phase that returned an error
901 If you only want some of the stages of compilation, you can use
902 @option{-x} (or filename suffixes) to tell @command{gcc} where to start, and
903 one of the options @option{-c}, @option{-S}, or @option{-E} to say where
904 @command{gcc} is to stop. Note that some combinations (for example,
905 @samp{-x cpp-output -E}) instruct @command{gcc} to do nothing at all.
910 Compile or assemble the source files, but do not link. The linking
911 stage simply is not done. The ultimate output is in the form of an
912 object file for each source file.
914 By default, the object file name for a source file is made by replacing
915 the suffix @samp{.c}, @samp{.i}, @samp{.s}, etc., with @samp{.o}.
917 Unrecognized input files, not requiring compilation or assembly, are
922 Stop after the stage of compilation proper; do not assemble. The output
923 is in the form of an assembler code file for each non-assembler input
926 By default, the assembler file name for a source file is made by
927 replacing the suffix @samp{.c}, @samp{.i}, etc., with @samp{.s}.
929 Input files that don't require compilation are ignored.
933 Stop after the preprocessing stage; do not run the compiler proper. The
934 output is in the form of preprocessed source code, which is sent to the
937 Input files which don't require preprocessing are ignored.
939 @cindex output file option
942 Place output in file @var{file}. This applies regardless to whatever
943 sort of output is being produced, whether it be an executable file,
944 an object file, an assembler file or preprocessed C code.
946 If @option{-o} is not specified, the default is to put an executable
947 file in @file{a.out}, the object file for
948 @file{@var{source}.@var{suffix}} in @file{@var{source}.o}, its
949 assembler file in @file{@var{source}.s}, a precompiled header file in
950 @file{@var{source}.@var{suffix}.gch}, and all preprocessed C source on
955 Print (on standard error output) the commands executed to run the stages
956 of compilation. Also print the version number of the compiler driver
957 program and of the preprocessor and the compiler proper.
961 Like @option{-v} except the commands are not executed and all command
962 arguments are quoted. This is useful for shell scripts to capture the
963 driver-generated command lines.
967 Use pipes rather than temporary files for communication between the
968 various stages of compilation. This fails to work on some systems where
969 the assembler is unable to read from a pipe; but the GNU assembler has
974 If you are compiling multiple source files, this option tells the driver
975 to pass all the source files to the compiler at once (for those
976 languages for which the compiler can handle this). This will allow
977 intermodule analysis (IMA) to be performed by the compiler. Currently the only
978 language for which this is supported is C@. If you pass source files for
979 multiple languages to the driver, using this option, the driver will invoke
980 the compiler(s) that support IMA once each, passing each compiler all the
981 source files appropriate for it. For those languages that do not support
982 IMA this option will be ignored, and the compiler will be invoked once for
983 each source file in that language. If you use this option in conjunction
984 with @option{-save-temps}, the compiler will generate multiple
986 (one for each source file), but only one (combined) @file{.o} or
991 Print (on the standard output) a description of the command line options
992 understood by @command{gcc}. If the @option{-v} option is also specified
993 then @option{--help} will also be passed on to the various processes
994 invoked by @command{gcc}, so that they can display the command line options
995 they accept. If the @option{-Wextra} option is also specified then command
996 line options which have no documentation associated with them will also
1000 @opindex target-help
1001 Print (on the standard output) a description of target specific command
1002 line options for each tool.
1006 Display the version number and copyrights of the invoked GCC@.
1010 @section Compiling C++ Programs
1012 @cindex suffixes for C++ source
1013 @cindex C++ source file suffixes
1014 C++ source files conventionally use one of the suffixes @samp{.C},
1015 @samp{.cc}, @samp{.cpp}, @samp{.CPP}, @samp{.c++}, @samp{.cp}, or
1016 @samp{.cxx}; C++ header files often use @samp{.hh} or @samp{.H}; and
1017 preprocessed C++ files use the suffix @samp{.ii}. GCC recognizes
1018 files with these names and compiles them as C++ programs even if you
1019 call the compiler the same way as for compiling C programs (usually
1020 with the name @command{gcc}).
1024 However, C++ programs often require class libraries as well as a
1025 compiler that understands the C++ language---and under some
1026 circumstances, you might want to compile programs or header files from
1027 standard input, or otherwise without a suffix that flags them as C++
1028 programs. You might also like to precompile a C header file with a
1029 @samp{.h} extension to be used in C++ compilations. @command{g++} is a
1030 program that calls GCC with the default language set to C++, and
1031 automatically specifies linking against the C++ library. On many
1032 systems, @command{g++} is also installed with the name @command{c++}.
1034 @cindex invoking @command{g++}
1035 When you compile C++ programs, you may specify many of the same
1036 command-line options that you use for compiling programs in any
1037 language; or command-line options meaningful for C and related
1038 languages; or options that are meaningful only for C++ programs.
1039 @xref{C Dialect Options,,Options Controlling C Dialect}, for
1040 explanations of options for languages related to C@.
1041 @xref{C++ Dialect Options,,Options Controlling C++ Dialect}, for
1042 explanations of options that are meaningful only for C++ programs.
1044 @node C Dialect Options
1045 @section Options Controlling C Dialect
1046 @cindex dialect options
1047 @cindex language dialect options
1048 @cindex options, dialect
1050 The following options control the dialect of C (or languages derived
1051 from C, such as C++, Objective-C and Objective-C++) that the compiler
1055 @cindex ANSI support
1059 In C mode, support all ISO C90 programs. In C++ mode,
1060 remove GNU extensions that conflict with ISO C++.
1062 This turns off certain features of GCC that are incompatible with ISO
1063 C90 (when compiling C code), or of standard C++ (when compiling C++ code),
1064 such as the @code{asm} and @code{typeof} keywords, and
1065 predefined macros such as @code{unix} and @code{vax} that identify the
1066 type of system you are using. It also enables the undesirable and
1067 rarely used ISO trigraph feature. For the C compiler,
1068 it disables recognition of C++ style @samp{//} comments as well as
1069 the @code{inline} keyword.
1071 The alternate keywords @code{__asm__}, @code{__extension__},
1072 @code{__inline__} and @code{__typeof__} continue to work despite
1073 @option{-ansi}. You would not want to use them in an ISO C program, of
1074 course, but it is useful to put them in header files that might be included
1075 in compilations done with @option{-ansi}. Alternate predefined macros
1076 such as @code{__unix__} and @code{__vax__} are also available, with or
1077 without @option{-ansi}.
1079 The @option{-ansi} option does not cause non-ISO programs to be
1080 rejected gratuitously. For that, @option{-pedantic} is required in
1081 addition to @option{-ansi}. @xref{Warning Options}.
1083 The macro @code{__STRICT_ANSI__} is predefined when the @option{-ansi}
1084 option is used. Some header files may notice this macro and refrain
1085 from declaring certain functions or defining certain macros that the
1086 ISO standard doesn't call for; this is to avoid interfering with any
1087 programs that might use these names for other things.
1089 Functions which would normally be built in but do not have semantics
1090 defined by ISO C (such as @code{alloca} and @code{ffs}) are not built-in
1091 functions with @option{-ansi} is used. @xref{Other Builtins,,Other
1092 built-in functions provided by GCC}, for details of the functions
1097 Determine the language standard. This option is currently only
1098 supported when compiling C or C++. A value for this option must be
1099 provided; possible values are
1104 ISO C90 (same as @option{-ansi}).
1106 @item iso9899:199409
1107 ISO C90 as modified in amendment 1.
1113 ISO C99. Note that this standard is not yet fully supported; see
1114 @w{@uref{http://gcc.gnu.org/c99status.html}} for more information. The
1115 names @samp{c9x} and @samp{iso9899:199x} are deprecated.
1118 Default, ISO C90 plus GNU extensions (including some C99 features).
1122 ISO C99 plus GNU extensions. When ISO C99 is fully implemented in GCC,
1123 this will become the default. The name @samp{gnu9x} is deprecated.
1126 The 1998 ISO C++ standard plus amendments.
1129 The same as @option{-std=c++98} plus GNU extensions. This is the
1130 default for C++ code.
1133 Even when this option is not specified, you can still use some of the
1134 features of newer standards in so far as they do not conflict with
1135 previous C standards. For example, you may use @code{__restrict__} even
1136 when @option{-std=c99} is not specified.
1138 The @option{-std} options specifying some version of ISO C have the same
1139 effects as @option{-ansi}, except that features that were not in ISO C90
1140 but are in the specified version (for example, @samp{//} comments and
1141 the @code{inline} keyword in ISO C99) are not disabled.
1143 @xref{Standards,,Language Standards Supported by GCC}, for details of
1144 these standard versions.
1146 @item -aux-info @var{filename}
1148 Output to the given filename prototyped declarations for all functions
1149 declared and/or defined in a translation unit, including those in header
1150 files. This option is silently ignored in any language other than C@.
1152 Besides declarations, the file indicates, in comments, the origin of
1153 each declaration (source file and line), whether the declaration was
1154 implicit, prototyped or unprototyped (@samp{I}, @samp{N} for new or
1155 @samp{O} for old, respectively, in the first character after the line
1156 number and the colon), and whether it came from a declaration or a
1157 definition (@samp{C} or @samp{F}, respectively, in the following
1158 character). In the case of function definitions, a K&R-style list of
1159 arguments followed by their declarations is also provided, inside
1160 comments, after the declaration.
1164 Do not recognize @code{asm}, @code{inline} or @code{typeof} as a
1165 keyword, so that code can use these words as identifiers. You can use
1166 the keywords @code{__asm__}, @code{__inline__} and @code{__typeof__}
1167 instead. @option{-ansi} implies @option{-fno-asm}.
1169 In C++, this switch only affects the @code{typeof} keyword, since
1170 @code{asm} and @code{inline} are standard keywords. You may want to
1171 use the @option{-fno-gnu-keywords} flag instead, which has the same
1172 effect. In C99 mode (@option{-std=c99} or @option{-std=gnu99}), this
1173 switch only affects the @code{asm} and @code{typeof} keywords, since
1174 @code{inline} is a standard keyword in ISO C99.
1177 @itemx -fno-builtin-@var{function}
1178 @opindex fno-builtin
1179 @cindex built-in functions
1180 Don't recognize built-in functions that do not begin with
1181 @samp{__builtin_} as prefix. @xref{Other Builtins,,Other built-in
1182 functions provided by GCC}, for details of the functions affected,
1183 including those which are not built-in functions when @option{-ansi} or
1184 @option{-std} options for strict ISO C conformance are used because they
1185 do not have an ISO standard meaning.
1187 GCC normally generates special code to handle certain built-in functions
1188 more efficiently; for instance, calls to @code{alloca} may become single
1189 instructions that adjust the stack directly, and calls to @code{memcpy}
1190 may become inline copy loops. The resulting code is often both smaller
1191 and faster, but since the function calls no longer appear as such, you
1192 cannot set a breakpoint on those calls, nor can you change the behavior
1193 of the functions by linking with a different library. In addition,
1194 when a function is recognized as a built-in function, GCC may use
1195 information about that function to warn about problems with calls to
1196 that function, or to generate more efficient code, even if the
1197 resulting code still contains calls to that function. For example,
1198 warnings are given with @option{-Wformat} for bad calls to
1199 @code{printf}, when @code{printf} is built in, and @code{strlen} is
1200 known not to modify global memory.
1202 With the @option{-fno-builtin-@var{function}} option
1203 only the built-in function @var{function} is
1204 disabled. @var{function} must not begin with @samp{__builtin_}. If a
1205 function is named this is not built-in in this version of GCC, this
1206 option is ignored. There is no corresponding
1207 @option{-fbuiltin-@var{function}} option; if you wish to enable
1208 built-in functions selectively when using @option{-fno-builtin} or
1209 @option{-ffreestanding}, you may define macros such as:
1212 #define abs(n) __builtin_abs ((n))
1213 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1218 @cindex hosted environment
1220 Assert that compilation takes place in a hosted environment. This implies
1221 @option{-fbuiltin}. A hosted environment is one in which the
1222 entire standard library is available, and in which @code{main} has a return
1223 type of @code{int}. Examples are nearly everything except a kernel.
1224 This is equivalent to @option{-fno-freestanding}.
1226 @item -ffreestanding
1227 @opindex ffreestanding
1228 @cindex hosted environment
1230 Assert that compilation takes place in a freestanding environment. This
1231 implies @option{-fno-builtin}. A freestanding environment
1232 is one in which the standard library may not exist, and program startup may
1233 not necessarily be at @code{main}. The most obvious example is an OS kernel.
1234 This is equivalent to @option{-fno-hosted}.
1236 @xref{Standards,,Language Standards Supported by GCC}, for details of
1237 freestanding and hosted environments.
1239 @item -fms-extensions
1240 @opindex fms-extensions
1241 Accept some non-standard constructs used in Microsoft header files.
1243 Some cases of unnamed fields in structures and unions are only
1244 accepted with this option. @xref{Unnamed Fields,,Unnamed struct/union
1245 fields within structs/unions}, for details.
1249 Support ISO C trigraphs. The @option{-ansi} option (and @option{-std}
1250 options for strict ISO C conformance) implies @option{-trigraphs}.
1252 @item -no-integrated-cpp
1253 @opindex no-integrated-cpp
1254 Performs a compilation in two passes: preprocessing and compiling. This
1255 option allows a user supplied "cc1", "cc1plus", or "cc1obj" via the
1256 @option{-B} option. The user supplied compilation step can then add in
1257 an additional preprocessing step after normal preprocessing but before
1258 compiling. The default is to use the integrated cpp (internal cpp)
1260 The semantics of this option will change if "cc1", "cc1plus", and
1261 "cc1obj" are merged.
1263 @cindex traditional C language
1264 @cindex C language, traditional
1266 @itemx -traditional-cpp
1267 @opindex traditional-cpp
1268 @opindex traditional
1269 Formerly, these options caused GCC to attempt to emulate a pre-standard
1270 C compiler. They are now only supported with the @option{-E} switch.
1271 The preprocessor continues to support a pre-standard mode. See the GNU
1272 CPP manual for details.
1274 @item -fcond-mismatch
1275 @opindex fcond-mismatch
1276 Allow conditional expressions with mismatched types in the second and
1277 third arguments. The value of such an expression is void. This option
1278 is not supported for C++.
1280 @item -funsigned-char
1281 @opindex funsigned-char
1282 Let the type @code{char} be unsigned, like @code{unsigned char}.
1284 Each kind of machine has a default for what @code{char} should
1285 be. It is either like @code{unsigned char} by default or like
1286 @code{signed char} by default.
1288 Ideally, a portable program should always use @code{signed char} or
1289 @code{unsigned char} when it depends on the signedness of an object.
1290 But many programs have been written to use plain @code{char} and
1291 expect it to be signed, or expect it to be unsigned, depending on the
1292 machines they were written for. This option, and its inverse, let you
1293 make such a program work with the opposite default.
1295 The type @code{char} is always a distinct type from each of
1296 @code{signed char} or @code{unsigned char}, even though its behavior
1297 is always just like one of those two.
1300 @opindex fsigned-char
1301 Let the type @code{char} be signed, like @code{signed char}.
1303 Note that this is equivalent to @option{-fno-unsigned-char}, which is
1304 the negative form of @option{-funsigned-char}. Likewise, the option
1305 @option{-fno-signed-char} is equivalent to @option{-funsigned-char}.
1307 @item -fsigned-bitfields
1308 @itemx -funsigned-bitfields
1309 @itemx -fno-signed-bitfields
1310 @itemx -fno-unsigned-bitfields
1311 @opindex fsigned-bitfields
1312 @opindex funsigned-bitfields
1313 @opindex fno-signed-bitfields
1314 @opindex fno-unsigned-bitfields
1315 These options control whether a bit-field is signed or unsigned, when the
1316 declaration does not use either @code{signed} or @code{unsigned}. By
1317 default, such a bit-field is signed, because this is consistent: the
1318 basic integer types such as @code{int} are signed types.
1321 @node C++ Dialect Options
1322 @section Options Controlling C++ Dialect
1324 @cindex compiler options, C++
1325 @cindex C++ options, command line
1326 @cindex options, C++
1327 This section describes the command-line options that are only meaningful
1328 for C++ programs; but you can also use most of the GNU compiler options
1329 regardless of what language your program is in. For example, you
1330 might compile a file @code{firstClass.C} like this:
1333 g++ -g -frepo -O -c firstClass.C
1337 In this example, only @option{-frepo} is an option meant
1338 only for C++ programs; you can use the other options with any
1339 language supported by GCC@.
1341 Here is a list of options that are @emph{only} for compiling C++ programs:
1345 @item -fabi-version=@var{n}
1346 @opindex fabi-version
1347 Use version @var{n} of the C++ ABI@. Version 2 is the version of the
1348 C++ ABI that first appeared in G++ 3.4. Version 1 is the version of
1349 the C++ ABI that first appeared in G++ 3.2. Version 0 will always be
1350 the version that conforms most closely to the C++ ABI specification.
1351 Therefore, the ABI obtained using version 0 will change as ABI bugs
1354 The default is version 2.
1356 @item -fno-access-control
1357 @opindex fno-access-control
1358 Turn off all access checking. This switch is mainly useful for working
1359 around bugs in the access control code.
1363 Check that the pointer returned by @code{operator new} is non-null
1364 before attempting to modify the storage allocated. This check is
1365 normally unnecessary because the C++ standard specifies that
1366 @code{operator new} will only return @code{0} if it is declared
1367 @samp{throw()}, in which case the compiler will always check the
1368 return value even without this option. In all other cases, when
1369 @code{operator new} has a non-empty exception specification, memory
1370 exhaustion is signalled by throwing @code{std::bad_alloc}. See also
1371 @samp{new (nothrow)}.
1373 @item -fconserve-space
1374 @opindex fconserve-space
1375 Put uninitialized or runtime-initialized global variables into the
1376 common segment, as C does. This saves space in the executable at the
1377 cost of not diagnosing duplicate definitions. If you compile with this
1378 flag and your program mysteriously crashes after @code{main()} has
1379 completed, you may have an object that is being destroyed twice because
1380 two definitions were merged.
1382 This option is no longer useful on most targets, now that support has
1383 been added for putting variables into BSS without making them common.
1385 @item -fno-const-strings
1386 @opindex fno-const-strings
1387 Give string constants type @code{char *} instead of type @code{const
1388 char *}. By default, G++ uses type @code{const char *} as required by
1389 the standard. Even if you use @option{-fno-const-strings}, you cannot
1390 actually modify the value of a string constant.
1392 This option might be removed in a future release of G++. For maximum
1393 portability, you should structure your code so that it works with
1394 string constants that have type @code{const char *}.
1396 @item -fno-elide-constructors
1397 @opindex fno-elide-constructors
1398 The C++ standard allows an implementation to omit creating a temporary
1399 which is only used to initialize another object of the same type.
1400 Specifying this option disables that optimization, and forces G++ to
1401 call the copy constructor in all cases.
1403 @item -fno-enforce-eh-specs
1404 @opindex fno-enforce-eh-specs
1405 Don't check for violation of exception specifications at runtime. This
1406 option violates the C++ standard, but may be useful for reducing code
1407 size in production builds, much like defining @samp{NDEBUG}. The compiler
1408 will still optimize based on the exception specifications.
1411 @itemx -fno-for-scope
1413 @opindex fno-for-scope
1414 If @option{-ffor-scope} is specified, the scope of variables declared in
1415 a @i{for-init-statement} is limited to the @samp{for} loop itself,
1416 as specified by the C++ standard.
1417 If @option{-fno-for-scope} is specified, the scope of variables declared in
1418 a @i{for-init-statement} extends to the end of the enclosing scope,
1419 as was the case in old versions of G++, and other (traditional)
1420 implementations of C++.
1422 The default if neither flag is given to follow the standard,
1423 but to allow and give a warning for old-style code that would
1424 otherwise be invalid, or have different behavior.
1426 @item -fno-gnu-keywords
1427 @opindex fno-gnu-keywords
1428 Do not recognize @code{typeof} as a keyword, so that code can use this
1429 word as an identifier. You can use the keyword @code{__typeof__} instead.
1430 @option{-ansi} implies @option{-fno-gnu-keywords}.
1432 @item -fno-implicit-templates
1433 @opindex fno-implicit-templates
1434 Never emit code for non-inline templates which are instantiated
1435 implicitly (i.e.@: by use); only emit code for explicit instantiations.
1436 @xref{Template Instantiation}, for more information.
1438 @item -fno-implicit-inline-templates
1439 @opindex fno-implicit-inline-templates
1440 Don't emit code for implicit instantiations of inline templates, either.
1441 The default is to handle inlines differently so that compiles with and
1442 without optimization will need the same set of explicit instantiations.
1444 @item -fno-implement-inlines
1445 @opindex fno-implement-inlines
1446 To save space, do not emit out-of-line copies of inline functions
1447 controlled by @samp{#pragma implementation}. This will cause linker
1448 errors if these functions are not inlined everywhere they are called.
1450 @item -fms-extensions
1451 @opindex fms-extensions
1452 Disable pedantic warnings about constructs used in MFC, such as implicit
1453 int and getting a pointer to member function via non-standard syntax.
1455 @item -fno-nonansi-builtins
1456 @opindex fno-nonansi-builtins
1457 Disable built-in declarations of functions that are not mandated by
1458 ANSI/ISO C@. These include @code{ffs}, @code{alloca}, @code{_exit},
1459 @code{index}, @code{bzero}, @code{conjf}, and other related functions.
1461 @item -fno-operator-names
1462 @opindex fno-operator-names
1463 Do not treat the operator name keywords @code{and}, @code{bitand},
1464 @code{bitor}, @code{compl}, @code{not}, @code{or} and @code{xor} as
1465 synonyms as keywords.
1467 @item -fno-optional-diags
1468 @opindex fno-optional-diags
1469 Disable diagnostics that the standard says a compiler does not need to
1470 issue. Currently, the only such diagnostic issued by G++ is the one for
1471 a name having multiple meanings within a class.
1474 @opindex fpermissive
1475 Downgrade some diagnostics about nonconformant code from errors to
1476 warnings. Thus, using @option{-fpermissive} will allow some
1477 nonconforming code to compile.
1481 Enable automatic template instantiation at link time. This option also
1482 implies @option{-fno-implicit-templates}. @xref{Template
1483 Instantiation}, for more information.
1487 Disable generation of information about every class with virtual
1488 functions for use by the C++ runtime type identification features
1489 (@samp{dynamic_cast} and @samp{typeid}). If you don't use those parts
1490 of the language, you can save some space by using this flag. Note that
1491 exception handling uses the same information, but it will generate it as
1496 Emit statistics about front-end processing at the end of the compilation.
1497 This information is generally only useful to the G++ development team.
1499 @item -ftemplate-depth-@var{n}
1500 @opindex ftemplate-depth
1501 Set the maximum instantiation depth for template classes to @var{n}.
1502 A limit on the template instantiation depth is needed to detect
1503 endless recursions during template class instantiation. ANSI/ISO C++
1504 conforming programs must not rely on a maximum depth greater than 17.
1506 @item -fno-threadsafe-statics
1507 @opindex fno-threadsafe-statics
1508 Do not emit the extra code to use the routines specified in the C++
1509 ABI for thread-safe initialization of local statics. You can use this
1510 option to reduce code size slightly in code that doesn't need to be
1513 @item -fuse-cxa-atexit
1514 @opindex fuse-cxa-atexit
1515 Register destructors for objects with static storage duration with the
1516 @code{__cxa_atexit} function rather than the @code{atexit} function.
1517 This option is required for fully standards-compliant handling of static
1518 destructors, but will only work if your C library supports
1519 @code{__cxa_atexit}.
1521 @item -fvisibility-inlines-hidden
1522 @opindex fvisibility-inlines-hidden
1523 Causes all inlined methods to be marked with
1524 @code{__attribute__ ((visibility ("hidden")))} so that they do not
1525 appear in the export table of a DSO and do not require a PLT indirection
1526 when used within the DSO@. Enabling this option can have a dramatic effect
1527 on load and link times of a DSO as it massively reduces the size of the
1528 dynamic export table when the library makes heavy use of templates. While
1529 it can cause bloating through duplication of code within each DSO where
1530 it is used, often the wastage is less than the considerable space occupied
1531 by a long symbol name in the export table which is typical when using
1532 templates and namespaces. For even more savings, combine with the
1533 @option{-fvisibility=hidden} switch.
1537 Do not use weak symbol support, even if it is provided by the linker.
1538 By default, G++ will use weak symbols if they are available. This
1539 option exists only for testing, and should not be used by end-users;
1540 it will result in inferior code and has no benefits. This option may
1541 be removed in a future release of G++.
1545 Do not search for header files in the standard directories specific to
1546 C++, but do still search the other standard directories. (This option
1547 is used when building the C++ library.)
1550 In addition, these optimization, warning, and code generation options
1551 have meanings only for C++ programs:
1554 @item -fno-default-inline
1555 @opindex fno-default-inline
1556 Do not assume @samp{inline} for functions defined inside a class scope.
1557 @xref{Optimize Options,,Options That Control Optimization}. Note that these
1558 functions will have linkage like inline functions; they just won't be
1561 @item -Wabi @r{(C++ only)}
1563 Warn when G++ generates code that is probably not compatible with the
1564 vendor-neutral C++ ABI@. Although an effort has been made to warn about
1565 all such cases, there are probably some cases that are not warned about,
1566 even though G++ is generating incompatible code. There may also be
1567 cases where warnings are emitted even though the code that is generated
1570 You should rewrite your code to avoid these warnings if you are
1571 concerned about the fact that code generated by G++ may not be binary
1572 compatible with code generated by other compilers.
1574 The known incompatibilities at this point include:
1579 Incorrect handling of tail-padding for bit-fields. G++ may attempt to
1580 pack data into the same byte as a base class. For example:
1583 struct A @{ virtual void f(); int f1 : 1; @};
1584 struct B : public A @{ int f2 : 1; @};
1588 In this case, G++ will place @code{B::f2} into the same byte
1589 as@code{A::f1}; other compilers will not. You can avoid this problem
1590 by explicitly padding @code{A} so that its size is a multiple of the
1591 byte size on your platform; that will cause G++ and other compilers to
1592 layout @code{B} identically.
1595 Incorrect handling of tail-padding for virtual bases. G++ does not use
1596 tail padding when laying out virtual bases. For example:
1599 struct A @{ virtual void f(); char c1; @};
1600 struct B @{ B(); char c2; @};
1601 struct C : public A, public virtual B @{@};
1605 In this case, G++ will not place @code{B} into the tail-padding for
1606 @code{A}; other compilers will. You can avoid this problem by
1607 explicitly padding @code{A} so that its size is a multiple of its
1608 alignment (ignoring virtual base classes); that will cause G++ and other
1609 compilers to layout @code{C} identically.
1612 Incorrect handling of bit-fields with declared widths greater than that
1613 of their underlying types, when the bit-fields appear in a union. For
1617 union U @{ int i : 4096; @};
1621 Assuming that an @code{int} does not have 4096 bits, G++ will make the
1622 union too small by the number of bits in an @code{int}.
1625 Empty classes can be placed at incorrect offsets. For example:
1635 struct C : public B, public A @{@};
1639 G++ will place the @code{A} base class of @code{C} at a nonzero offset;
1640 it should be placed at offset zero. G++ mistakenly believes that the
1641 @code{A} data member of @code{B} is already at offset zero.
1644 Names of template functions whose types involve @code{typename} or
1645 template template parameters can be mangled incorrectly.
1648 template <typename Q>
1649 void f(typename Q::X) @{@}
1651 template <template <typename> class Q>
1652 void f(typename Q<int>::X) @{@}
1656 Instantiations of these templates may be mangled incorrectly.
1660 @item -Wctor-dtor-privacy @r{(C++ only)}
1661 @opindex Wctor-dtor-privacy
1662 Warn when a class seems unusable because all the constructors or
1663 destructors in that class are private, and it has neither friends nor
1664 public static member functions.
1666 @item -Wnon-virtual-dtor @r{(C++ only)}
1667 @opindex Wnon-virtual-dtor
1668 Warn when a class appears to be polymorphic, thereby requiring a virtual
1669 destructor, yet it declares a non-virtual one.
1670 This warning is enabled by @option{-Wall}.
1672 @item -Wreorder @r{(C++ only)}
1674 @cindex reordering, warning
1675 @cindex warning for reordering of member initializers
1676 Warn when the order of member initializers given in the code does not
1677 match the order in which they must be executed. For instance:
1683 A(): j (0), i (1) @{ @}
1687 The compiler will rearrange the member initializers for @samp{i}
1688 and @samp{j} to match the declaration order of the members, emitting
1689 a warning to that effect. This warning is enabled by @option{-Wall}.
1692 The following @option{-W@dots{}} options are not affected by @option{-Wall}.
1695 @item -Weffc++ @r{(C++ only)}
1697 Warn about violations of the following style guidelines from Scott Meyers'
1698 @cite{Effective C++} book:
1702 Item 11: Define a copy constructor and an assignment operator for classes
1703 with dynamically allocated memory.
1706 Item 12: Prefer initialization to assignment in constructors.
1709 Item 14: Make destructors virtual in base classes.
1712 Item 15: Have @code{operator=} return a reference to @code{*this}.
1715 Item 23: Don't try to return a reference when you must return an object.
1719 Also warn about violations of the following style guidelines from
1720 Scott Meyers' @cite{More Effective C++} book:
1724 Item 6: Distinguish between prefix and postfix forms of increment and
1725 decrement operators.
1728 Item 7: Never overload @code{&&}, @code{||}, or @code{,}.
1732 When selecting this option, be aware that the standard library
1733 headers do not obey all of these guidelines; use @samp{grep -v}
1734 to filter out those warnings.
1736 @item -Wno-deprecated @r{(C++ only)}
1737 @opindex Wno-deprecated
1738 Do not warn about usage of deprecated features. @xref{Deprecated Features}.
1740 @item -Wno-non-template-friend @r{(C++ only)}
1741 @opindex Wno-non-template-friend
1742 Disable warnings when non-templatized friend functions are declared
1743 within a template. Since the advent of explicit template specification
1744 support in G++, if the name of the friend is an unqualified-id (i.e.,
1745 @samp{friend foo(int)}), the C++ language specification demands that the
1746 friend declare or define an ordinary, nontemplate function. (Section
1747 14.5.3). Before G++ implemented explicit specification, unqualified-ids
1748 could be interpreted as a particular specialization of a templatized
1749 function. Because this non-conforming behavior is no longer the default
1750 behavior for G++, @option{-Wnon-template-friend} allows the compiler to
1751 check existing code for potential trouble spots and is on by default.
1752 This new compiler behavior can be turned off with
1753 @option{-Wno-non-template-friend} which keeps the conformant compiler code
1754 but disables the helpful warning.
1756 @item -Wold-style-cast @r{(C++ only)}
1757 @opindex Wold-style-cast
1758 Warn if an old-style (C-style) cast to a non-void type is used within
1759 a C++ program. The new-style casts (@samp{static_cast},
1760 @samp{reinterpret_cast}, and @samp{const_cast}) are less vulnerable to
1761 unintended effects and much easier to search for.
1763 @item -Woverloaded-virtual @r{(C++ only)}
1764 @opindex Woverloaded-virtual
1765 @cindex overloaded virtual fn, warning
1766 @cindex warning for overloaded virtual fn
1767 Warn when a function declaration hides virtual functions from a
1768 base class. For example, in:
1775 struct B: public A @{
1780 the @code{A} class version of @code{f} is hidden in @code{B}, and code
1788 will fail to compile.
1790 @item -Wno-pmf-conversions @r{(C++ only)}
1791 @opindex Wno-pmf-conversions
1792 Disable the diagnostic for converting a bound pointer to member function
1795 @item -Wsign-promo @r{(C++ only)}
1796 @opindex Wsign-promo
1797 Warn when overload resolution chooses a promotion from unsigned or
1798 enumerated type to a signed type, over a conversion to an unsigned type of
1799 the same size. Previous versions of G++ would try to preserve
1800 unsignedness, but the standard mandates the current behavior.
1805 A& operator = (int);
1815 In this example, G++ will synthesize a default @samp{A& operator =
1816 (const A&);}, while cfront will use the user-defined @samp{operator =}.
1819 @node Objective-C and Objective-C++ Dialect Options
1820 @section Options Controlling Objective-C and Objective-C++ Dialects
1822 @cindex compiler options, Objective-C and Objective-C++
1823 @cindex Objective-C and Objective-C++ options, command line
1824 @cindex options, Objective-C and Objective-C++
1825 (NOTE: This manual does not describe the Objective-C and Objective-C++
1826 languages themselves. See @xref{Standards,,Language Standards
1827 Supported by GCC}, for references.)
1829 This section describes the command-line options that are only meaningful
1830 for Objective-C and Objective-C++ programs, but you can also use most of
1831 the language-independent GNU compiler options.
1832 For example, you might compile a file @code{some_class.m} like this:
1835 gcc -g -fgnu-runtime -O -c some_class.m
1839 In this example, @option{-fgnu-runtime} is an option meant only for
1840 Objective-C and Objective-C++ programs; you can use the other options with
1841 any language supported by GCC@.
1843 Note that since Objective-C is an extension of the C language, Objective-C
1844 compilations may also use options specific to the C front-end (e.g.,
1845 @option{-Wtraditional}). Similarly, Objective-C++ compilations may use
1846 C++-specific options (e.g., @option{-Wabi}).
1848 Here is a list of options that are @emph{only} for compiling Objective-C
1849 and Objective-C++ programs:
1852 @item -fconstant-string-class=@var{class-name}
1853 @opindex fconstant-string-class
1854 Use @var{class-name} as the name of the class to instantiate for each
1855 literal string specified with the syntax @code{@@"@dots{}"}. The default
1856 class name is @code{NXConstantString} if the GNU runtime is being used, and
1857 @code{NSConstantString} if the NeXT runtime is being used (see below). The
1858 @option{-fconstant-cfstrings} option, if also present, will override the
1859 @option{-fconstant-string-class} setting and cause @code{@@"@dots{}"} literals
1860 to be laid out as constant CoreFoundation strings.
1863 @opindex fgnu-runtime
1864 Generate object code compatible with the standard GNU Objective-C
1865 runtime. This is the default for most types of systems.
1867 @item -fnext-runtime
1868 @opindex fnext-runtime
1869 Generate output compatible with the NeXT runtime. This is the default
1870 for NeXT-based systems, including Darwin and Mac OS X@. The macro
1871 @code{__NEXT_RUNTIME__} is predefined if (and only if) this option is
1874 @item -fno-nil-receivers
1875 @opindex fno-nil-receivers
1876 Assume that all Objective-C message dispatches (e.g.,
1877 @code{[receiver message:arg]}) in this translation unit ensure that the receiver
1878 is not @code{nil}. This allows for more efficient entry points in the runtime
1879 to be used. Currently, this option is only available in conjunction with
1880 the NeXT runtime on Mac OS X 10.3 and later.
1882 @item -fobjc-exceptions
1883 @opindex fobjc-exceptions
1884 Enable syntactic support for structured exception handling in Objective-C,
1885 similar to what is offered by C++ and Java. Currently, this option is only
1886 available in conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1894 @@catch (AnObjCClass *exc) @{
1901 @@catch (AnotherClass *exc) @{
1904 @@catch (id allOthers) @{
1914 The @code{@@throw} statement may appear anywhere in an Objective-C or
1915 Objective-C++ program; when used inside of a @code{@@catch} block, the
1916 @code{@@throw} may appear without an argument (as shown above), in which case
1917 the object caught by the @code{@@catch} will be rethrown.
1919 Note that only (pointers to) Objective-C objects may be thrown and
1920 caught using this scheme. When an object is thrown, it will be caught
1921 by the nearest @code{@@catch} clause capable of handling objects of that type,
1922 analogously to how @code{catch} blocks work in C++ and Java. A
1923 @code{@@catch(id @dots{})} clause (as shown above) may also be provided to catch
1924 any and all Objective-C exceptions not caught by previous @code{@@catch}
1927 The @code{@@finally} clause, if present, will be executed upon exit from the
1928 immediately preceding @code{@@try @dots{} @@catch} section. This will happen
1929 regardless of whether any exceptions are thrown, caught or rethrown
1930 inside the @code{@@try @dots{} @@catch} section, analogously to the behavior
1931 of the @code{finally} clause in Java.
1933 There are several caveats to using the new exception mechanism:
1937 Although currently designed to be binary compatible with @code{NS_HANDLER}-style
1938 idioms provided by the @code{NSException} class, the new
1939 exceptions can only be used on Mac OS X 10.3 (Panther) and later
1940 systems, due to additional functionality needed in the (NeXT) Objective-C
1944 As mentioned above, the new exceptions do not support handling
1945 types other than Objective-C objects. Furthermore, when used from
1946 Objective-C++, the Objective-C exception model does not interoperate with C++
1947 exceptions at this time. This means you cannot @code{@@throw} an exception
1948 from Objective-C and @code{catch} it in C++, or vice versa
1949 (i.e., @code{throw @dots{} @@catch}).
1952 The @option{-fobjc-exceptions} switch also enables the use of synchronization
1953 blocks for thread-safe execution:
1956 @@synchronized (ObjCClass *guard) @{
1961 Upon entering the @code{@@synchronized} block, a thread of execution shall
1962 first check whether a lock has been placed on the corresponding @code{guard}
1963 object by another thread. If it has, the current thread shall wait until
1964 the other thread relinquishes its lock. Once @code{guard} becomes available,
1965 the current thread will place its own lock on it, execute the code contained in
1966 the @code{@@synchronized} block, and finally relinquish the lock (thereby
1967 making @code{guard} available to other threads).
1969 Unlike Java, Objective-C does not allow for entire methods to be marked
1970 @code{@@synchronized}. Note that throwing exceptions out of
1971 @code{@@synchronized} blocks is allowed, and will cause the guarding object
1972 to be unlocked properly.
1974 @item -freplace-objc-classes
1975 @opindex freplace-objc-classes
1976 Emit a special marker instructing @command{ld(1)} not to statically link in
1977 the resulting object file, and allow @command{dyld(1)} to load it in at
1978 run time instead. This is used in conjunction with the Fix-and-Continue
1979 debugging mode, where the object file in question may be recompiled and
1980 dynamically reloaded in the course of program execution, without the need
1981 to restart the program itself. Currently, Fix-and-Continue functionality
1982 is only available in conjunction with the NeXT runtime on Mac OS X 10.3
1987 When compiling for the NeXT runtime, the compiler ordinarily replaces calls
1988 to @code{objc_getClass("@dots{}")} (when the name of the class is known at
1989 compile time) with static class references that get initialized at load time,
1990 which improves run-time performance. Specifying the @option{-fzero-link} flag
1991 suppresses this behavior and causes calls to @code{objc_getClass("@dots{}")}
1992 to be retained. This is useful in Zero-Link debugging mode, since it allows
1993 for individual class implementations to be modified during program execution.
1997 Dump interface declarations for all classes seen in the source file to a
1998 file named @file{@var{sourcename}.decl}.
2001 @opindex Wno-protocol
2002 If a class is declared to implement a protocol, a warning is issued for
2003 every method in the protocol that is not implemented by the class. The
2004 default behavior is to issue a warning for every method not explicitly
2005 implemented in the class, even if a method implementation is inherited
2006 from the superclass. If you use the @option{-Wno-protocol} option, then
2007 methods inherited from the superclass are considered to be implemented,
2008 and no warning is issued for them.
2012 Warn if multiple methods of different types for the same selector are
2013 found during compilation. The check is performed on the list of methods
2014 in the final stage of compilation. Additionally, a check is performed
2015 for each selector appearing in a @code{@@selector(@dots{})}
2016 expression, and a corresponding method for that selector has been found
2017 during compilation. Because these checks scan the method table only at
2018 the end of compilation, these warnings are not produced if the final
2019 stage of compilation is not reached, for example because an error is
2020 found during compilation, or because the @option{-fsyntax-only} option is
2023 @item -Wundeclared-selector
2024 @opindex Wundeclared-selector
2025 Warn if a @code{@@selector(@dots{})} expression referring to an
2026 undeclared selector is found. A selector is considered undeclared if no
2027 method with that name has been declared before the
2028 @code{@@selector(@dots{})} expression, either explicitly in an
2029 @code{@@interface} or @code{@@protocol} declaration, or implicitly in
2030 an @code{@@implementation} section. This option always performs its
2031 checks as soon as a @code{@@selector(@dots{})} expression is found,
2032 while @option{-Wselector} only performs its checks in the final stage of
2033 compilation. This also enforces the coding style convention
2034 that methods and selectors must be declared before being used.
2036 @item -print-objc-runtime-info
2037 @opindex print-objc-runtime-info
2038 Generate C header describing the largest structure that is passed by
2043 @node Language Independent Options
2044 @section Options to Control Diagnostic Messages Formatting
2045 @cindex options to control diagnostics formatting
2046 @cindex diagnostic messages
2047 @cindex message formatting
2049 Traditionally, diagnostic messages have been formatted irrespective of
2050 the output device's aspect (e.g.@: its width, @dots{}). The options described
2051 below can be used to control the diagnostic messages formatting
2052 algorithm, e.g.@: how many characters per line, how often source location
2053 information should be reported. Right now, only the C++ front end can
2054 honor these options. However it is expected, in the near future, that
2055 the remaining front ends would be able to digest them correctly.
2058 @item -fmessage-length=@var{n}
2059 @opindex fmessage-length
2060 Try to format error messages so that they fit on lines of about @var{n}
2061 characters. The default is 72 characters for @command{g++} and 0 for the rest of
2062 the front ends supported by GCC@. If @var{n} is zero, then no
2063 line-wrapping will be done; each error message will appear on a single
2066 @opindex fdiagnostics-show-location
2067 @item -fdiagnostics-show-location=once
2068 Only meaningful in line-wrapping mode. Instructs the diagnostic messages
2069 reporter to emit @emph{once} source location information; that is, in
2070 case the message is too long to fit on a single physical line and has to
2071 be wrapped, the source location won't be emitted (as prefix) again,
2072 over and over, in subsequent continuation lines. This is the default
2075 @item -fdiagnostics-show-location=every-line
2076 Only meaningful in line-wrapping mode. Instructs the diagnostic
2077 messages reporter to emit the same source location information (as
2078 prefix) for physical lines that result from the process of breaking
2079 a message which is too long to fit on a single line.
2083 @node Warning Options
2084 @section Options to Request or Suppress Warnings
2085 @cindex options to control warnings
2086 @cindex warning messages
2087 @cindex messages, warning
2088 @cindex suppressing warnings
2090 Warnings are diagnostic messages that report constructions which
2091 are not inherently erroneous but which are risky or suggest there
2092 may have been an error.
2094 You can request many specific warnings with options beginning @samp{-W},
2095 for example @option{-Wimplicit} to request warnings on implicit
2096 declarations. Each of these specific warning options also has a
2097 negative form beginning @samp{-Wno-} to turn off warnings;
2098 for example, @option{-Wno-implicit}. This manual lists only one of the
2099 two forms, whichever is not the default.
2101 The following options control the amount and kinds of warnings produced
2102 by GCC; for further, language-specific options also refer to
2103 @ref{C++ Dialect Options} and @ref{Objective-C and Objective-C++ Dialect
2107 @cindex syntax checking
2109 @opindex fsyntax-only
2110 Check the code for syntax errors, but don't do anything beyond that.
2114 Issue all the warnings demanded by strict ISO C and ISO C++;
2115 reject all programs that use forbidden extensions, and some other
2116 programs that do not follow ISO C and ISO C++. For ISO C, follows the
2117 version of the ISO C standard specified by any @option{-std} option used.
2119 Valid ISO C and ISO C++ programs should compile properly with or without
2120 this option (though a rare few will require @option{-ansi} or a
2121 @option{-std} option specifying the required version of ISO C)@. However,
2122 without this option, certain GNU extensions and traditional C and C++
2123 features are supported as well. With this option, they are rejected.
2125 @option{-pedantic} does not cause warning messages for use of the
2126 alternate keywords whose names begin and end with @samp{__}. Pedantic
2127 warnings are also disabled in the expression that follows
2128 @code{__extension__}. However, only system header files should use
2129 these escape routes; application programs should avoid them.
2130 @xref{Alternate Keywords}.
2132 Some users try to use @option{-pedantic} to check programs for strict ISO
2133 C conformance. They soon find that it does not do quite what they want:
2134 it finds some non-ISO practices, but not all---only those for which
2135 ISO C @emph{requires} a diagnostic, and some others for which
2136 diagnostics have been added.
2138 A feature to report any failure to conform to ISO C might be useful in
2139 some instances, but would require considerable additional work and would
2140 be quite different from @option{-pedantic}. We don't have plans to
2141 support such a feature in the near future.
2143 Where the standard specified with @option{-std} represents a GNU
2144 extended dialect of C, such as @samp{gnu89} or @samp{gnu99}, there is a
2145 corresponding @dfn{base standard}, the version of ISO C on which the GNU
2146 extended dialect is based. Warnings from @option{-pedantic} are given
2147 where they are required by the base standard. (It would not make sense
2148 for such warnings to be given only for features not in the specified GNU
2149 C dialect, since by definition the GNU dialects of C include all
2150 features the compiler supports with the given option, and there would be
2151 nothing to warn about.)
2153 @item -pedantic-errors
2154 @opindex pedantic-errors
2155 Like @option{-pedantic}, except that errors are produced rather than
2160 Inhibit all warning messages.
2164 Inhibit warning messages about the use of @samp{#import}.
2166 @item -Wchar-subscripts
2167 @opindex Wchar-subscripts
2168 Warn if an array subscript has type @code{char}. This is a common cause
2169 of error, as programmers often forget that this type is signed on some
2171 This warning is enabled by @option{-Wall}.
2175 Warn whenever a comment-start sequence @samp{/*} appears in a @samp{/*}
2176 comment, or whenever a Backslash-Newline appears in a @samp{//} comment.
2177 This warning is enabled by @option{-Wall}.
2179 @item -Wfatal-errors
2180 @opindex Wfatal-errors
2181 This option causes the compiler to abort compilation on the first error
2182 occurred rather than trying to keep going and printing further error
2187 @opindex ffreestanding
2188 @opindex fno-builtin
2189 Check calls to @code{printf} and @code{scanf}, etc., to make sure that
2190 the arguments supplied have types appropriate to the format string
2191 specified, and that the conversions specified in the format string make
2192 sense. This includes standard functions, and others specified by format
2193 attributes (@pxref{Function Attributes}), in the @code{printf},
2194 @code{scanf}, @code{strftime} and @code{strfmon} (an X/Open extension,
2195 not in the C standard) families (or other target-specific families).
2196 Which functions are checked without format attributes having been
2197 specified depends on the standard version selected, and such checks of
2198 functions without the attribute specified are disabled by
2199 @option{-ffreestanding} or @option{-fno-builtin}.
2201 The formats are checked against the format features supported by GNU
2202 libc version 2.2. These include all ISO C90 and C99 features, as well
2203 as features from the Single Unix Specification and some BSD and GNU
2204 extensions. Other library implementations may not support all these
2205 features; GCC does not support warning about features that go beyond a
2206 particular library's limitations. However, if @option{-pedantic} is used
2207 with @option{-Wformat}, warnings will be given about format features not
2208 in the selected standard version (but not for @code{strfmon} formats,
2209 since those are not in any version of the C standard). @xref{C Dialect
2210 Options,,Options Controlling C Dialect}.
2212 Since @option{-Wformat} also checks for null format arguments for
2213 several functions, @option{-Wformat} also implies @option{-Wnonnull}.
2215 @option{-Wformat} is included in @option{-Wall}. For more control over some
2216 aspects of format checking, the options @option{-Wformat-y2k},
2217 @option{-Wno-format-extra-args}, @option{-Wno-format-zero-length},
2218 @option{-Wformat-nonliteral}, @option{-Wformat-security}, and
2219 @option{-Wformat=2} are available, but are not included in @option{-Wall}.
2222 @opindex Wformat-y2k
2223 If @option{-Wformat} is specified, also warn about @code{strftime}
2224 formats which may yield only a two-digit year.
2226 @item -Wno-format-extra-args
2227 @opindex Wno-format-extra-args
2228 If @option{-Wformat} is specified, do not warn about excess arguments to a
2229 @code{printf} or @code{scanf} format function. The C standard specifies
2230 that such arguments are ignored.
2232 Where the unused arguments lie between used arguments that are
2233 specified with @samp{$} operand number specifications, normally
2234 warnings are still given, since the implementation could not know what
2235 type to pass to @code{va_arg} to skip the unused arguments. However,
2236 in the case of @code{scanf} formats, this option will suppress the
2237 warning if the unused arguments are all pointers, since the Single
2238 Unix Specification says that such unused arguments are allowed.
2240 @item -Wno-format-zero-length
2241 @opindex Wno-format-zero-length
2242 If @option{-Wformat} is specified, do not warn about zero-length formats.
2243 The C standard specifies that zero-length formats are allowed.
2245 @item -Wformat-nonliteral
2246 @opindex Wformat-nonliteral
2247 If @option{-Wformat} is specified, also warn if the format string is not a
2248 string literal and so cannot be checked, unless the format function
2249 takes its format arguments as a @code{va_list}.
2251 @item -Wformat-security
2252 @opindex Wformat-security
2253 If @option{-Wformat} is specified, also warn about uses of format
2254 functions that represent possible security problems. At present, this
2255 warns about calls to @code{printf} and @code{scanf} functions where the
2256 format string is not a string literal and there are no format arguments,
2257 as in @code{printf (foo);}. This may be a security hole if the format
2258 string came from untrusted input and contains @samp{%n}. (This is
2259 currently a subset of what @option{-Wformat-nonliteral} warns about, but
2260 in future warnings may be added to @option{-Wformat-security} that are not
2261 included in @option{-Wformat-nonliteral}.)
2265 Enable @option{-Wformat} plus format checks not included in
2266 @option{-Wformat}. Currently equivalent to @samp{-Wformat
2267 -Wformat-nonliteral -Wformat-security -Wformat-y2k}.
2271 Warn about passing a null pointer for arguments marked as
2272 requiring a non-null value by the @code{nonnull} function attribute.
2274 @option{-Wnonnull} is included in @option{-Wall} and @option{-Wformat}. It
2275 can be disabled with the @option{-Wno-nonnull} option.
2277 @item -Winit-self @r{(C, C++, Objective-C and Objective-C++ only)}
2279 Warn about uninitialized variables which are initialized with themselves.
2280 Note this option can only be used with the @option{-Wuninitialized} option,
2281 which in turn only works with @option{-O1} and above.
2283 For example, GCC will warn about @code{i} being uninitialized in the
2284 following snippet only when @option{-Winit-self} has been specified:
2295 @item -Wimplicit-int
2296 @opindex Wimplicit-int
2297 Warn when a declaration does not specify a type.
2298 This warning is enabled by @option{-Wall}.
2300 @item -Wimplicit-function-declaration
2301 @itemx -Werror-implicit-function-declaration
2302 @opindex Wimplicit-function-declaration
2303 @opindex Werror-implicit-function-declaration
2304 Give a warning (or error) whenever a function is used before being
2305 declared. The form @option{-Wno-error-implicit-function-declaration}
2307 This warning is enabled by @option{-Wall} (as a warning, not an error).
2311 Same as @option{-Wimplicit-int} and @option{-Wimplicit-function-declaration}.
2312 This warning is enabled by @option{-Wall}.
2316 Warn if the type of @samp{main} is suspicious. @samp{main} should be a
2317 function with external linkage, returning int, taking either zero
2318 arguments, two, or three arguments of appropriate types.
2319 This warning is enabled by @option{-Wall}.
2321 @item -Wmissing-braces
2322 @opindex Wmissing-braces
2323 Warn if an aggregate or union initializer is not fully bracketed. In
2324 the following example, the initializer for @samp{a} is not fully
2325 bracketed, but that for @samp{b} is fully bracketed.
2328 int a[2][2] = @{ 0, 1, 2, 3 @};
2329 int b[2][2] = @{ @{ 0, 1 @}, @{ 2, 3 @} @};
2332 This warning is enabled by @option{-Wall}.
2334 @item -Wmissing-include-dirs @r{(C, C++, Objective-C and Objective-C++ only)}
2335 @opindex Wmissing-include-dirs
2336 Warn if a user-supplied include directory does not exist.
2339 @opindex Wparentheses
2340 Warn if parentheses are omitted in certain contexts, such
2341 as when there is an assignment in a context where a truth value
2342 is expected, or when operators are nested whose precedence people
2343 often get confused about. Only the warning for an assignment used as
2344 a truth value is supported when compiling C++; the other warnings are
2345 only supported when compiling C@.
2347 Also warn if a comparison like @samp{x<=y<=z} appears; this is
2348 equivalent to @samp{(x<=y ? 1 : 0) <= z}, which is a different
2349 interpretation from that of ordinary mathematical notation.
2351 Also warn about constructions where there may be confusion to which
2352 @code{if} statement an @code{else} branch belongs. Here is an example of
2367 In C, every @code{else} branch belongs to the innermost possible @code{if}
2368 statement, which in this example is @code{if (b)}. This is often not
2369 what the programmer expected, as illustrated in the above example by
2370 indentation the programmer chose. When there is the potential for this
2371 confusion, GCC will issue a warning when this flag is specified.
2372 To eliminate the warning, add explicit braces around the innermost
2373 @code{if} statement so there is no way the @code{else} could belong to
2374 the enclosing @code{if}. The resulting code would look like this:
2390 This warning is enabled by @option{-Wall}.
2392 @item -Wsequence-point
2393 @opindex Wsequence-point
2394 Warn about code that may have undefined semantics because of violations
2395 of sequence point rules in the C standard.
2397 The C standard defines the order in which expressions in a C program are
2398 evaluated in terms of @dfn{sequence points}, which represent a partial
2399 ordering between the execution of parts of the program: those executed
2400 before the sequence point, and those executed after it. These occur
2401 after the evaluation of a full expression (one which is not part of a
2402 larger expression), after the evaluation of the first operand of a
2403 @code{&&}, @code{||}, @code{? :} or @code{,} (comma) operator, before a
2404 function is called (but after the evaluation of its arguments and the
2405 expression denoting the called function), and in certain other places.
2406 Other than as expressed by the sequence point rules, the order of
2407 evaluation of subexpressions of an expression is not specified. All
2408 these rules describe only a partial order rather than a total order,
2409 since, for example, if two functions are called within one expression
2410 with no sequence point between them, the order in which the functions
2411 are called is not specified. However, the standards committee have
2412 ruled that function calls do not overlap.
2414 It is not specified when between sequence points modifications to the
2415 values of objects take effect. Programs whose behavior depends on this
2416 have undefined behavior; the C standard specifies that ``Between the
2417 previous and next sequence point an object shall have its stored value
2418 modified at most once by the evaluation of an expression. Furthermore,
2419 the prior value shall be read only to determine the value to be
2420 stored.''. If a program breaks these rules, the results on any
2421 particular implementation are entirely unpredictable.
2423 Examples of code with undefined behavior are @code{a = a++;}, @code{a[n]
2424 = b[n++]} and @code{a[i++] = i;}. Some more complicated cases are not
2425 diagnosed by this option, and it may give an occasional false positive
2426 result, but in general it has been found fairly effective at detecting
2427 this sort of problem in programs.
2429 The present implementation of this option only works for C programs. A
2430 future implementation may also work for C++ programs.
2432 The C standard is worded confusingly, therefore there is some debate
2433 over the precise meaning of the sequence point rules in subtle cases.
2434 Links to discussions of the problem, including proposed formal
2435 definitions, may be found on the GCC readings page, at
2436 @w{@uref{http://gcc.gnu.org/readings.html}}.
2438 This warning is enabled by @option{-Wall}.
2441 @opindex Wreturn-type
2442 Warn whenever a function is defined with a return-type that defaults to
2443 @code{int}. Also warn about any @code{return} statement with no
2444 return-value in a function whose return-type is not @code{void}.
2446 For C, also warn if the return type of a function has a type qualifier
2447 such as @code{const}. Such a type qualifier has no effect, since the
2448 value returned by a function is not an lvalue. ISO C prohibits
2449 qualified @code{void} return types on function definitions, so such
2450 return types always receive a warning even without this option.
2452 For C++, a function without return type always produces a diagnostic
2453 message, even when @option{-Wno-return-type} is specified. The only
2454 exceptions are @samp{main} and functions defined in system headers.
2456 This warning is enabled by @option{-Wall}.
2460 Warn whenever a @code{switch} statement has an index of enumerated type
2461 and lacks a @code{case} for one or more of the named codes of that
2462 enumeration. (The presence of a @code{default} label prevents this
2463 warning.) @code{case} labels outside the enumeration range also
2464 provoke warnings when this option is used.
2465 This warning is enabled by @option{-Wall}.
2467 @item -Wswitch-default
2468 @opindex Wswitch-switch
2469 Warn whenever a @code{switch} statement does not have a @code{default}
2473 @opindex Wswitch-enum
2474 Warn whenever a @code{switch} statement has an index of enumerated type
2475 and lacks a @code{case} for one or more of the named codes of that
2476 enumeration. @code{case} labels outside the enumeration range also
2477 provoke warnings when this option is used.
2481 Warn if any trigraphs are encountered that might change the meaning of
2482 the program (trigraphs within comments are not warned about).
2483 This warning is enabled by @option{-Wall}.
2485 @item -Wunused-function
2486 @opindex Wunused-function
2487 Warn whenever a static function is declared but not defined or a
2488 non\-inline static function is unused.
2489 This warning is enabled by @option{-Wall}.
2491 @item -Wunused-label
2492 @opindex Wunused-label
2493 Warn whenever a label is declared but not used.
2494 This warning is enabled by @option{-Wall}.
2496 To suppress this warning use the @samp{unused} attribute
2497 (@pxref{Variable Attributes}).
2499 @item -Wunused-parameter
2500 @opindex Wunused-parameter
2501 Warn whenever a function parameter is unused aside from its declaration.
2503 To suppress this warning use the @samp{unused} attribute
2504 (@pxref{Variable Attributes}).
2506 @item -Wunused-variable
2507 @opindex Wunused-variable
2508 Warn whenever a local variable or non-constant static variable is unused
2509 aside from its declaration
2510 This warning is enabled by @option{-Wall}.
2512 To suppress this warning use the @samp{unused} attribute
2513 (@pxref{Variable Attributes}).
2515 @item -Wunused-value
2516 @opindex Wunused-value
2517 Warn whenever a statement computes a result that is explicitly not used.
2518 This warning is enabled by @option{-Wall}.
2520 To suppress this warning cast the expression to @samp{void}.
2524 All the above @option{-Wunused} options combined.
2526 In order to get a warning about an unused function parameter, you must
2527 either specify @samp{-Wextra -Wunused} (note that @samp{-Wall} implies
2528 @samp{-Wunused}), or separately specify @option{-Wunused-parameter}.
2530 @item -Wuninitialized
2531 @opindex Wuninitialized
2532 Warn if an automatic variable is used without first being initialized or
2533 if a variable may be clobbered by a @code{setjmp} call.
2535 These warnings are possible only in optimizing compilation,
2536 because they require data flow information that is computed only
2537 when optimizing. If you don't specify @option{-O}, you simply won't
2540 If you want to warn about code which uses the uninitialized value of the
2541 variable in its own initializer, use the @option{-Winit-self} option.
2543 These warnings occur only for variables that are candidates for
2544 register allocation. Therefore, they do not occur for a variable that
2545 is declared @code{volatile}, or whose address is taken, or whose size
2546 is other than 1, 2, 4 or 8 bytes. Also, they do not occur for
2547 structures, unions or arrays, even when they are in registers.
2549 Note that there may be no warning about a variable that is used only
2550 to compute a value that itself is never used, because such
2551 computations may be deleted by data flow analysis before the warnings
2554 These warnings are made optional because GCC is not smart
2555 enough to see all the reasons why the code might be correct
2556 despite appearing to have an error. Here is one example of how
2577 If the value of @code{y} is always 1, 2 or 3, then @code{x} is
2578 always initialized, but GCC doesn't know this. Here is
2579 another common case:
2584 if (change_y) save_y = y, y = new_y;
2586 if (change_y) y = save_y;
2591 This has no bug because @code{save_y} is used only if it is set.
2593 @cindex @code{longjmp} warnings
2594 This option also warns when a non-volatile automatic variable might be
2595 changed by a call to @code{longjmp}. These warnings as well are possible
2596 only in optimizing compilation.
2598 The compiler sees only the calls to @code{setjmp}. It cannot know
2599 where @code{longjmp} will be called; in fact, a signal handler could
2600 call it at any point in the code. As a result, you may get a warning
2601 even when there is in fact no problem because @code{longjmp} cannot
2602 in fact be called at the place which would cause a problem.
2604 Some spurious warnings can be avoided if you declare all the functions
2605 you use that never return as @code{noreturn}. @xref{Function
2608 This warning is enabled by @option{-Wall}.
2610 @item -Wunknown-pragmas
2611 @opindex Wunknown-pragmas
2612 @cindex warning for unknown pragmas
2613 @cindex unknown pragmas, warning
2614 @cindex pragmas, warning of unknown
2615 Warn when a #pragma directive is encountered which is not understood by
2616 GCC@. If this command line option is used, warnings will even be issued
2617 for unknown pragmas in system header files. This is not the case if
2618 the warnings were only enabled by the @option{-Wall} command line option.
2620 @item -Wstrict-aliasing
2621 @opindex Wstrict-aliasing
2622 This option is only active when @option{-fstrict-aliasing} is active.
2623 It warns about code which might break the strict aliasing rules that the
2624 compiler is using for optimization. The warning does not catch all
2625 cases, but does attempt to catch the more common pitfalls. It is
2626 included in @option{-Wall}.
2628 @item -Wstrict-aliasing=2
2629 @opindex Wstrict-aliasing=2
2630 This option is only active when @option{-fstrict-aliasing} is active.
2631 It warns about all code which might break the strict aliasing rules that the
2632 compiler is using for optimization. This warning catches all cases, but
2633 it will also give a warning for some ambiguous cases that are safe.
2637 All of the above @samp{-W} options combined. This enables all the
2638 warnings about constructions that some users consider questionable, and
2639 that are easy to avoid (or modify to prevent the warning), even in
2640 conjunction with macros. This also enables some language-specific
2641 warnings described in @ref{C++ Dialect Options} and
2642 @ref{Objective-C and Objective-C++ Dialect Options}.
2645 The following @option{-W@dots{}} options are not implied by @option{-Wall}.
2646 Some of them warn about constructions that users generally do not
2647 consider questionable, but which occasionally you might wish to check
2648 for; others warn about constructions that are necessary or hard to avoid
2649 in some cases, and there is no simple way to modify the code to suppress
2656 (This option used to be called @option{-W}. The older name is still
2657 supported, but the newer name is more descriptive.) Print extra warning
2658 messages for these events:
2662 A function can return either with or without a value. (Falling
2663 off the end of the function body is considered returning without
2664 a value.) For example, this function would evoke such a
2678 An expression-statement or the left-hand side of a comma expression
2679 contains no side effects.
2680 To suppress the warning, cast the unused expression to void.
2681 For example, an expression such as @samp{x[i,j]} will cause a warning,
2682 but @samp{x[(void)i,j]} will not.
2685 An unsigned value is compared against zero with @samp{<} or @samp{>=}.
2688 Storage-class specifiers like @code{static} are not the first things in
2689 a declaration. According to the C Standard, this usage is obsolescent.
2692 If @option{-Wall} or @option{-Wunused} is also specified, warn about unused
2696 A comparison between signed and unsigned values could produce an
2697 incorrect result when the signed value is converted to unsigned.
2698 (But don't warn if @option{-Wno-sign-compare} is also specified.)
2701 An aggregate has an initializer which does not initialize all members.
2702 This warning can be independently controlled by
2703 @option{-Wmissing-field-initializers}.
2706 A function parameter is declared without a type specifier in K&R-style
2714 An empty body occurs in an @samp{if} or @samp{else} statement.
2717 A pointer is compared against integer zero with @samp{<}, @samp{<=},
2718 @samp{>}, or @samp{>=}.
2721 A variable might be changed by @samp{longjmp} or @samp{vfork}.
2724 Any of several floating-point events that often indicate errors, such as
2725 overflow, underflow, loss of precision, etc.
2727 @item @r{(C++ only)}
2728 An enumerator and a non-enumerator both appear in a conditional expression.
2730 @item @r{(C++ only)}
2731 A non-static reference or non-static @samp{const} member appears in a
2732 class without constructors.
2734 @item @r{(C++ only)}
2735 Ambiguous virtual bases.
2737 @item @r{(C++ only)}
2738 Subscripting an array which has been declared @samp{register}.
2740 @item @r{(C++ only)}
2741 Taking the address of a variable which has been declared @samp{register}.
2743 @item @r{(C++ only)}
2744 A base class is not initialized in a derived class' copy constructor.
2747 @item -Wno-div-by-zero
2748 @opindex Wno-div-by-zero
2749 @opindex Wdiv-by-zero
2750 Do not warn about compile-time integer division by zero. Floating point
2751 division by zero is not warned about, as it can be a legitimate way of
2752 obtaining infinities and NaNs.
2754 @item -Wsystem-headers
2755 @opindex Wsystem-headers
2756 @cindex warnings from system headers
2757 @cindex system headers, warnings from
2758 Print warning messages for constructs found in system header files.
2759 Warnings from system headers are normally suppressed, on the assumption
2760 that they usually do not indicate real problems and would only make the
2761 compiler output harder to read. Using this command line option tells
2762 GCC to emit warnings from system headers as if they occurred in user
2763 code. However, note that using @option{-Wall} in conjunction with this
2764 option will @emph{not} warn about unknown pragmas in system
2765 headers---for that, @option{-Wunknown-pragmas} must also be used.
2768 @opindex Wfloat-equal
2769 Warn if floating point values are used in equality comparisons.
2771 The idea behind this is that sometimes it is convenient (for the
2772 programmer) to consider floating-point values as approximations to
2773 infinitely precise real numbers. If you are doing this, then you need
2774 to compute (by analyzing the code, or in some other way) the maximum or
2775 likely maximum error that the computation introduces, and allow for it
2776 when performing comparisons (and when producing output, but that's a
2777 different problem). In particular, instead of testing for equality, you
2778 would check to see whether the two values have ranges that overlap; and
2779 this is done with the relational operators, so equality comparisons are
2782 @item -Wtraditional @r{(C only)}
2783 @opindex Wtraditional
2784 Warn about certain constructs that behave differently in traditional and
2785 ISO C@. Also warn about ISO C constructs that have no traditional C
2786 equivalent, and/or problematic constructs which should be avoided.
2790 Macro parameters that appear within string literals in the macro body.
2791 In traditional C macro replacement takes place within string literals,
2792 but does not in ISO C@.
2795 In traditional C, some preprocessor directives did not exist.
2796 Traditional preprocessors would only consider a line to be a directive
2797 if the @samp{#} appeared in column 1 on the line. Therefore
2798 @option{-Wtraditional} warns about directives that traditional C
2799 understands but would ignore because the @samp{#} does not appear as the
2800 first character on the line. It also suggests you hide directives like
2801 @samp{#pragma} not understood by traditional C by indenting them. Some
2802 traditional implementations would not recognize @samp{#elif}, so it
2803 suggests avoiding it altogether.
2806 A function-like macro that appears without arguments.
2809 The unary plus operator.
2812 The @samp{U} integer constant suffix, or the @samp{F} or @samp{L} floating point
2813 constant suffixes. (Traditional C does support the @samp{L} suffix on integer
2814 constants.) Note, these suffixes appear in macros defined in the system
2815 headers of most modern systems, e.g.@: the @samp{_MIN}/@samp{_MAX} macros in @code{<limits.h>}.
2816 Use of these macros in user code might normally lead to spurious
2817 warnings, however GCC's integrated preprocessor has enough context to
2818 avoid warning in these cases.
2821 A function declared external in one block and then used after the end of
2825 A @code{switch} statement has an operand of type @code{long}.
2828 A non-@code{static} function declaration follows a @code{static} one.
2829 This construct is not accepted by some traditional C compilers.
2832 The ISO type of an integer constant has a different width or
2833 signedness from its traditional type. This warning is only issued if
2834 the base of the constant is ten. I.e.@: hexadecimal or octal values, which
2835 typically represent bit patterns, are not warned about.
2838 Usage of ISO string concatenation is detected.
2841 Initialization of automatic aggregates.
2844 Identifier conflicts with labels. Traditional C lacks a separate
2845 namespace for labels.
2848 Initialization of unions. If the initializer is zero, the warning is
2849 omitted. This is done under the assumption that the zero initializer in
2850 user code appears conditioned on e.g.@: @code{__STDC__} to avoid missing
2851 initializer warnings and relies on default initialization to zero in the
2855 Conversions by prototypes between fixed/floating point values and vice
2856 versa. The absence of these prototypes when compiling with traditional
2857 C would cause serious problems. This is a subset of the possible
2858 conversion warnings, for the full set use @option{-Wconversion}.
2861 Use of ISO C style function definitions. This warning intentionally is
2862 @emph{not} issued for prototype declarations or variadic functions
2863 because these ISO C features will appear in your code when using
2864 libiberty's traditional C compatibility macros, @code{PARAMS} and
2865 @code{VPARAMS}. This warning is also bypassed for nested functions
2866 because that feature is already a GCC extension and thus not relevant to
2867 traditional C compatibility.
2870 @item -Wdeclaration-after-statement @r{(C only)}
2871 @opindex Wdeclaration-after-statement
2872 Warn when a declaration is found after a statement in a block. This
2873 construct, known from C++, was introduced with ISO C99 and is by default
2874 allowed in GCC@. It is not supported by ISO C90 and was not supported by
2875 GCC versions before GCC 3.0. @xref{Mixed Declarations}.
2879 Warn if an undefined identifier is evaluated in an @samp{#if} directive.
2881 @item -Wno-endif-labels
2882 @opindex Wno-endif-labels
2883 @opindex Wendif-labels
2884 Do not warn whenever an @samp{#else} or an @samp{#endif} are followed by text.
2888 Warn whenever a local variable shadows another local variable, parameter or
2889 global variable or whenever a built-in function is shadowed.
2891 @item -Wlarger-than-@var{len}
2892 @opindex Wlarger-than
2893 Warn whenever an object of larger than @var{len} bytes is defined.
2895 @item -Wpointer-arith
2896 @opindex Wpointer-arith
2897 Warn about anything that depends on the ``size of'' a function type or
2898 of @code{void}. GNU C assigns these types a size of 1, for
2899 convenience in calculations with @code{void *} pointers and pointers
2902 @item -Wbad-function-cast @r{(C only)}
2903 @opindex Wbad-function-cast
2904 Warn whenever a function call is cast to a non-matching type.
2905 For example, warn if @code{int malloc()} is cast to @code{anything *}.
2909 Warn whenever a pointer is cast so as to remove a type qualifier from
2910 the target type. For example, warn if a @code{const char *} is cast
2911 to an ordinary @code{char *}.
2914 @opindex Wcast-align
2915 Warn whenever a pointer is cast such that the required alignment of the
2916 target is increased. For example, warn if a @code{char *} is cast to
2917 an @code{int *} on machines where integers can only be accessed at
2918 two- or four-byte boundaries.
2920 @item -Wwrite-strings
2921 @opindex Wwrite-strings
2922 When compiling C, give string constants the type @code{const
2923 char[@var{length}]} so that
2924 copying the address of one into a non-@code{const} @code{char *}
2925 pointer will get a warning; when compiling C++, warn about the
2926 deprecated conversion from string constants to @code{char *}.
2927 These warnings will help you find at
2928 compile time code that can try to write into a string constant, but
2929 only if you have been very careful about using @code{const} in
2930 declarations and prototypes. Otherwise, it will just be a nuisance;
2931 this is why we did not make @option{-Wall} request these warnings.
2934 @opindex Wconversion
2935 Warn if a prototype causes a type conversion that is different from what
2936 would happen to the same argument in the absence of a prototype. This
2937 includes conversions of fixed point to floating and vice versa, and
2938 conversions changing the width or signedness of a fixed point argument
2939 except when the same as the default promotion.
2941 Also, warn if a negative integer constant expression is implicitly
2942 converted to an unsigned type. For example, warn about the assignment
2943 @code{x = -1} if @code{x} is unsigned. But do not warn about explicit
2944 casts like @code{(unsigned) -1}.
2946 @item -Wsign-compare
2947 @opindex Wsign-compare
2948 @cindex warning for comparison of signed and unsigned values
2949 @cindex comparison of signed and unsigned values, warning
2950 @cindex signed and unsigned values, comparison warning
2951 Warn when a comparison between signed and unsigned values could produce
2952 an incorrect result when the signed value is converted to unsigned.
2953 This warning is also enabled by @option{-Wextra}; to get the other warnings
2954 of @option{-Wextra} without this warning, use @samp{-Wextra -Wno-sign-compare}.
2956 @item -Waggregate-return
2957 @opindex Waggregate-return
2958 Warn if any functions that return structures or unions are defined or
2959 called. (In languages where you can return an array, this also elicits
2962 @item -Wstrict-prototypes @r{(C only)}
2963 @opindex Wstrict-prototypes
2964 Warn if a function is declared or defined without specifying the
2965 argument types. (An old-style function definition is permitted without
2966 a warning if preceded by a declaration which specifies the argument
2969 @item -Wold-style-definition @r{(C only)}
2970 @opindex Wold-style-definition
2971 Warn if an old-style function definition is used. A warning is given
2972 even if there is a previous prototype.
2974 @item -Wmissing-prototypes @r{(C only)}
2975 @opindex Wmissing-prototypes
2976 Warn if a global function is defined without a previous prototype
2977 declaration. This warning is issued even if the definition itself
2978 provides a prototype. The aim is to detect global functions that fail
2979 to be declared in header files.
2981 @item -Wmissing-declarations @r{(C only)}
2982 @opindex Wmissing-declarations
2983 Warn if a global function is defined without a previous declaration.
2984 Do so even if the definition itself provides a prototype.
2985 Use this option to detect global functions that are not declared in
2988 @item -Wmissing-field-initializers
2989 @opindex Wmissing-field-initializers
2992 Warn if a structure's initializer has some fields missing. For
2993 example, the following code would cause such a warning, because
2994 @code{x.h} is implicitly zero:
2997 struct s @{ int f, g, h; @};
2998 struct s x = @{ 3, 4 @};
3001 This option does not warn about designated initializers, so the following
3002 modification would not trigger a warning:
3005 struct s @{ int f, g, h; @};
3006 struct s x = @{ .f = 3, .g = 4 @};
3009 This warning is included in @option{-Wextra}. To get other @option{-Wextra}
3010 warnings without this one, use @samp{-Wextra -Wno-missing-field-initializers}.
3012 @item -Wmissing-noreturn
3013 @opindex Wmissing-noreturn
3014 Warn about functions which might be candidates for attribute @code{noreturn}.
3015 Note these are only possible candidates, not absolute ones. Care should
3016 be taken to manually verify functions actually do not ever return before
3017 adding the @code{noreturn} attribute, otherwise subtle code generation
3018 bugs could be introduced. You will not get a warning for @code{main} in
3019 hosted C environments.
3021 @item -Wmissing-format-attribute
3022 @opindex Wmissing-format-attribute
3024 If @option{-Wformat} is enabled, also warn about functions which might be
3025 candidates for @code{format} attributes. Note these are only possible
3026 candidates, not absolute ones. GCC will guess that @code{format}
3027 attributes might be appropriate for any function that calls a function
3028 like @code{vprintf} or @code{vscanf}, but this might not always be the
3029 case, and some functions for which @code{format} attributes are
3030 appropriate may not be detected. This option has no effect unless
3031 @option{-Wformat} is enabled (possibly by @option{-Wall}).
3033 @item -Wno-multichar
3034 @opindex Wno-multichar
3036 Do not warn if a multicharacter constant (@samp{'FOOF'}) is used.
3037 Usually they indicate a typo in the user's code, as they have
3038 implementation-defined values, and should not be used in portable code.
3040 @item -Wnormalized=<none|id|nfc|nfkc>
3041 @opindex Wnormalized
3044 @cindex character set, input normalization
3045 In ISO C and ISO C++, two identifiers are different if they are
3046 different sequences of characters. However, sometimes when characters
3047 outside the basic ASCII character set are used, you can have two
3048 different character sequences that look the same. To avoid confusion,
3049 the ISO 10646 standard sets out some @dfn{normalization rules} which
3050 when applied ensure that two sequences that look the same are turned into
3051 the same sequence. GCC can warn you if you are using identifiers which
3052 have not been normalized; this option controls that warning.
3054 There are four levels of warning that GCC supports. The default is
3055 @option{-Wnormalized=nfc}, which warns about any identifier which is
3056 not in the ISO 10646 ``C'' normalized form, @dfn{NFC}. NFC is the
3057 recommended form for most uses.
3059 Unfortunately, there are some characters which ISO C and ISO C++ allow
3060 in identifiers that when turned into NFC aren't allowable as
3061 identifiers. That is, there's no way to use these symbols in portable
3062 ISO C or C++ and have all your identifiers in NFC.
3063 @option{-Wnormalized=id} suppresses the warning for these characters.
3064 It is hoped that future versions of the standards involved will correct
3065 this, which is why this option is not the default.
3067 You can switch the warning off for all characters by writing
3068 @option{-Wnormalized=none}. You would only want to do this if you
3069 were using some other normalization scheme (like ``D''), because
3070 otherwise you can easily create bugs that are literally impossible to see.
3072 Some characters in ISO 10646 have distinct meanings but look identical
3073 in some fonts or display methodologies, especially once formatting has
3074 been applied. For instance @code{\u207F}, ``SUPERSCRIPT LATIN SMALL
3075 LETTER N'', will display just like a regular @code{n} which has been
3076 placed in a superscript. ISO 10646 defines the @dfn{NFKC}
3077 normalisation scheme to convert all these into a standard form as
3078 well, and GCC will warn if your code is not in NFKC if you use
3079 @option{-Wnormalized=nfkc}. This warning is comparable to warning
3080 about every identifier that contains the letter O because it might be
3081 confused with the digit 0, and so is not the default, but may be
3082 useful as a local coding convention if the programming environment is
3083 unable to be fixed to display these characters distinctly.
3085 @item -Wno-deprecated-declarations
3086 @opindex Wno-deprecated-declarations
3087 Do not warn about uses of functions, variables, and types marked as
3088 deprecated by using the @code{deprecated} attribute.
3089 (@pxref{Function Attributes}, @pxref{Variable Attributes},
3090 @pxref{Type Attributes}.)
3094 Warn if a structure is given the packed attribute, but the packed
3095 attribute has no effect on the layout or size of the structure.
3096 Such structures may be mis-aligned for little benefit. For
3097 instance, in this code, the variable @code{f.x} in @code{struct bar}
3098 will be misaligned even though @code{struct bar} does not itself
3099 have the packed attribute:
3106 @} __attribute__((packed));
3116 Warn if padding is included in a structure, either to align an element
3117 of the structure or to align the whole structure. Sometimes when this
3118 happens it is possible to rearrange the fields of the structure to
3119 reduce the padding and so make the structure smaller.
3121 @item -Wredundant-decls
3122 @opindex Wredundant-decls
3123 Warn if anything is declared more than once in the same scope, even in
3124 cases where multiple declaration is valid and changes nothing.
3126 @item -Wnested-externs @r{(C only)}
3127 @opindex Wnested-externs
3128 Warn if an @code{extern} declaration is encountered within a function.
3130 @item -Wunreachable-code
3131 @opindex Wunreachable-code
3132 Warn if the compiler detects that code will never be executed.
3134 This option is intended to warn when the compiler detects that at
3135 least a whole line of source code will never be executed, because
3136 some condition is never satisfied or because it is after a
3137 procedure that never returns.
3139 It is possible for this option to produce a warning even though there
3140 are circumstances under which part of the affected line can be executed,
3141 so care should be taken when removing apparently-unreachable code.
3143 For instance, when a function is inlined, a warning may mean that the
3144 line is unreachable in only one inlined copy of the function.
3146 This option is not made part of @option{-Wall} because in a debugging
3147 version of a program there is often substantial code which checks
3148 correct functioning of the program and is, hopefully, unreachable
3149 because the program does work. Another common use of unreachable
3150 code is to provide behavior which is selectable at compile-time.
3154 Warn if a function can not be inlined and it was declared as inline.
3155 Even with this option, the compiler will not warn about failures to
3156 inline functions declared in system headers.
3158 The compiler uses a variety of heuristics to determine whether or not
3159 to inline a function. For example, the compiler takes into account
3160 the size of the function being inlined and the the amount of inlining
3161 that has already been done in the current function. Therefore,
3162 seemingly insignificant changes in the source program can cause the
3163 warnings produced by @option{-Winline} to appear or disappear.
3165 @item -Wno-invalid-offsetof @r{(C++ only)}
3166 @opindex Wno-invalid-offsetof
3167 Suppress warnings from applying the @samp{offsetof} macro to a non-POD
3168 type. According to the 1998 ISO C++ standard, applying @samp{offsetof}
3169 to a non-POD type is undefined. In existing C++ implementations,
3170 however, @samp{offsetof} typically gives meaningful results even when
3171 applied to certain kinds of non-POD types. (Such as a simple
3172 @samp{struct} that fails to be a POD type only by virtue of having a
3173 constructor.) This flag is for users who are aware that they are
3174 writing nonportable code and who have deliberately chosen to ignore the
3177 The restrictions on @samp{offsetof} may be relaxed in a future version
3178 of the C++ standard.
3181 @opindex Winvalid-pch
3182 Warn if a precompiled header (@pxref{Precompiled Headers}) is found in
3183 the search path but can't be used.
3187 @opindex Wno-long-long
3188 Warn if @samp{long long} type is used. This is default. To inhibit
3189 the warning messages, use @option{-Wno-long-long}. Flags
3190 @option{-Wlong-long} and @option{-Wno-long-long} are taken into account
3191 only when @option{-pedantic} flag is used.
3193 @item -Wvariadic-macros
3194 @opindex Wvariadic-macros
3195 @opindex Wno-variadic-macros
3196 Warn if variadic macros are used in pedantic ISO C90 mode, or the GNU
3197 alternate syntax when in pedantic ISO C99 mode. This is default.
3198 To inhibit the warning messages, use @option{-Wno-variadic-macros}.
3200 @item -Wdisabled-optimization
3201 @opindex Wdisabled-optimization
3202 Warn if a requested optimization pass is disabled. This warning does
3203 not generally indicate that there is anything wrong with your code; it
3204 merely indicates that GCC's optimizers were unable to handle the code
3205 effectively. Often, the problem is that your code is too big or too
3206 complex; GCC will refuse to optimize programs when the optimization
3207 itself is likely to take inordinate amounts of time.
3209 @item -Wno-pointer-sign
3210 @opindex Wno-pointer-sign
3211 Don't warn for pointer argument passing or assignment with different signedness.
3212 Only useful in the negative form since this warning is enabled by default.
3213 This option is only supported for C and Objective-C@.
3217 Make all warnings into errors.
3220 @node Debugging Options
3221 @section Options for Debugging Your Program or GCC
3222 @cindex options, debugging
3223 @cindex debugging information options
3225 GCC has various special options that are used for debugging
3226 either your program or GCC:
3231 Produce debugging information in the operating system's native format
3232 (stabs, COFF, XCOFF, or DWARF 2)@. GDB can work with this debugging
3235 On most systems that use stabs format, @option{-g} enables use of extra
3236 debugging information that only GDB can use; this extra information
3237 makes debugging work better in GDB but will probably make other debuggers
3239 refuse to read the program. If you want to control for certain whether
3240 to generate the extra information, use @option{-gstabs+}, @option{-gstabs},
3241 @option{-gxcoff+}, @option{-gxcoff}, or @option{-gvms} (see below).
3243 GCC allows you to use @option{-g} with
3244 @option{-O}. The shortcuts taken by optimized code may occasionally
3245 produce surprising results: some variables you declared may not exist
3246 at all; flow of control may briefly move where you did not expect it;
3247 some statements may not be executed because they compute constant
3248 results or their values were already at hand; some statements may
3249 execute in different places because they were moved out of loops.
3251 Nevertheless it proves possible to debug optimized output. This makes
3252 it reasonable to use the optimizer for programs that might have bugs.
3254 The following options are useful when GCC is generated with the
3255 capability for more than one debugging format.
3259 Produce debugging information for use by GDB@. This means to use the
3260 most expressive format available (DWARF 2, stabs, or the native format
3261 if neither of those are supported), including GDB extensions if at all
3266 Produce debugging information in stabs format (if that is supported),
3267 without GDB extensions. This is the format used by DBX on most BSD
3268 systems. On MIPS, Alpha and System V Release 4 systems this option
3269 produces stabs debugging output which is not understood by DBX or SDB@.
3270 On System V Release 4 systems this option requires the GNU assembler.
3272 @item -feliminate-unused-debug-symbols
3273 @opindex feliminate-unused-debug-symbols
3274 Produce debugging information in stabs format (if that is supported),
3275 for only symbols that are actually used.
3279 Produce debugging information in stabs format (if that is supported),
3280 using GNU extensions understood only by the GNU debugger (GDB)@. The
3281 use of these extensions is likely to make other debuggers crash or
3282 refuse to read the program.
3286 Produce debugging information in COFF format (if that is supported).
3287 This is the format used by SDB on most System V systems prior to
3292 Produce debugging information in XCOFF format (if that is supported).
3293 This is the format used by the DBX debugger on IBM RS/6000 systems.
3297 Produce debugging information in XCOFF format (if that is supported),
3298 using GNU extensions understood only by the GNU debugger (GDB)@. The
3299 use of these extensions is likely to make other debuggers crash or
3300 refuse to read the program, and may cause assemblers other than the GNU
3301 assembler (GAS) to fail with an error.
3305 Produce debugging information in DWARF version 2 format (if that is
3306 supported). This is the format used by DBX on IRIX 6. With this
3307 option, GCC uses features of DWARF version 3 when they are useful;
3308 version 3 is upward compatible with version 2, but may still cause
3309 problems for older debuggers.
3313 Produce debugging information in VMS debug format (if that is
3314 supported). This is the format used by DEBUG on VMS systems.
3317 @itemx -ggdb@var{level}
3318 @itemx -gstabs@var{level}
3319 @itemx -gcoff@var{level}
3320 @itemx -gxcoff@var{level}
3321 @itemx -gvms@var{level}
3322 Request debugging information and also use @var{level} to specify how
3323 much information. The default level is 2.
3325 Level 1 produces minimal information, enough for making backtraces in
3326 parts of the program that you don't plan to debug. This includes
3327 descriptions of functions and external variables, but no information
3328 about local variables and no line numbers.
3330 Level 3 includes extra information, such as all the macro definitions
3331 present in the program. Some debuggers support macro expansion when
3332 you use @option{-g3}.
3334 @option{-gdwarf-2} does not accept a concatenated debug level, because
3335 GCC used to support an option @option{-gdwarf} that meant to generate
3336 debug information in version 1 of the DWARF format (which is very
3337 different from version 2), and it would have been too confusing. That
3338 debug format is long obsolete, but the option cannot be changed now.
3339 Instead use an additional @option{-g@var{level}} option to change the
3340 debug level for DWARF2.
3342 @item -feliminate-dwarf2-dups
3343 @opindex feliminate-dwarf2-dups
3344 Compress DWARF2 debugging information by eliminating duplicated
3345 information about each symbol. This option only makes sense when
3346 generating DWARF2 debugging information with @option{-gdwarf-2}.
3348 @cindex @command{prof}
3351 Generate extra code to write profile information suitable for the
3352 analysis program @command{prof}. You must use this option when compiling
3353 the source files you want data about, and you must also use it when
3356 @cindex @command{gprof}
3359 Generate extra code to write profile information suitable for the
3360 analysis program @command{gprof}. You must use this option when compiling
3361 the source files you want data about, and you must also use it when
3366 Makes the compiler print out each function name as it is compiled, and
3367 print some statistics about each pass when it finishes.
3370 @opindex ftime-report
3371 Makes the compiler print some statistics about the time consumed by each
3372 pass when it finishes.
3375 @opindex fmem-report
3376 Makes the compiler print some statistics about permanent memory
3377 allocation when it finishes.
3379 @item -fprofile-arcs
3380 @opindex fprofile-arcs
3381 Add code so that program flow @dfn{arcs} are instrumented. During
3382 execution the program records how many times each branch and call is
3383 executed and how many times it is taken or returns. When the compiled
3384 program exits it saves this data to a file called
3385 @file{@var{auxname}.gcda} for each source file. The data may be used for
3386 profile-directed optimizations (@option{-fbranch-probabilities}), or for
3387 test coverage analysis (@option{-ftest-coverage}). Each object file's
3388 @var{auxname} is generated from the name of the output file, if
3389 explicitly specified and it is not the final executable, otherwise it is
3390 the basename of the source file. In both cases any suffix is removed
3391 (e.g.@: @file{foo.gcda} for input file @file{dir/foo.c}, or
3392 @file{dir/foo.gcda} for output file specified as @option{-o dir/foo.o}).
3394 @cindex @command{gcov}
3398 This option is used to compile and link code instrumented for coverage
3399 analysis. The option is a synonym for @option{-fprofile-arcs}
3400 @option{-ftest-coverage} (when compiling) and @option{-lgcov} (when
3401 linking). See the documentation for those options for more details.
3406 Compile the source files with @option{-fprofile-arcs} plus optimization
3407 and code generation options. For test coverage analysis, use the
3408 additional @option{-ftest-coverage} option. You do not need to profile
3409 every source file in a program.
3412 Link your object files with @option{-lgcov} or @option{-fprofile-arcs}
3413 (the latter implies the former).
3416 Run the program on a representative workload to generate the arc profile
3417 information. This may be repeated any number of times. You can run
3418 concurrent instances of your program, and provided that the file system
3419 supports locking, the data files will be correctly updated. Also
3420 @code{fork} calls are detected and correctly handled (double counting
3424 For profile-directed optimizations, compile the source files again with
3425 the same optimization and code generation options plus
3426 @option{-fbranch-probabilities} (@pxref{Optimize Options,,Options that
3427 Control Optimization}).
3430 For test coverage analysis, use @command{gcov} to produce human readable
3431 information from the @file{.gcno} and @file{.gcda} files. Refer to the
3432 @command{gcov} documentation for further information.
3436 With @option{-fprofile-arcs}, for each function of your program GCC
3437 creates a program flow graph, then finds a spanning tree for the graph.
3438 Only arcs that are not on the spanning tree have to be instrumented: the
3439 compiler adds code to count the number of times that these arcs are
3440 executed. When an arc is the only exit or only entrance to a block, the
3441 instrumentation code can be added to the block; otherwise, a new basic
3442 block must be created to hold the instrumentation code.
3444 @item -ftree-based-profiling
3445 @opindex ftree-based-profiling
3446 This option is used in addition to @option{-fprofile-arcs} or
3447 @option{-fbranch-probabilities} to control whether those optimizations
3448 are performed on a tree-based or rtl-based internal representation.
3449 If you use this option when compiling with @option{-fprofile-arcs},
3450 you must also use it when compiling later with @option{-fbranch-probabilities}.
3451 Currently the tree-based optimization is in an early stage of
3452 development, and this option is recommended only for those people
3453 working on improving it.
3456 @item -ftest-coverage
3457 @opindex ftest-coverage
3458 Produce a notes file that the @command{gcov} code-coverage utility
3459 (@pxref{Gcov,, @command{gcov}---a Test Coverage Program}) can use to
3460 show program coverage. Each source file's note file is called
3461 @file{@var{auxname}.gcno}. Refer to the @option{-fprofile-arcs} option
3462 above for a description of @var{auxname} and instructions on how to
3463 generate test coverage data. Coverage data will match the source files
3464 more closely, if you do not optimize.
3466 @item -d@var{letters}
3467 @item -fdump-rtl-@var{pass}
3469 Says to make debugging dumps during compilation at times specified by
3470 @var{letters}. This is used for debugging the RTL-based passes of the
3471 compiler. The file names for most of the dumps are made by appending a
3472 pass number and a word to the @var{dumpname}. @var{dumpname} is generated
3473 from the name of the output file, if explicitly specified and it is not
3474 an executable, otherwise it is the basename of the source file.
3476 Most debug dumps can be enabled either passing a letter to the @option{-d}
3477 option, or with a long @option{-fdump-rtl} switch; here are the possible
3478 letters for use in @var{letters} and @var{pass}, and their meanings:
3483 Annotate the assembler output with miscellaneous debugging information.
3486 @itemx -fdump-rtl-bp
3488 @opindex fdump-rtl-bp
3489 Dump after computing branch probabilities, to @file{@var{file}.09.bp}.
3492 @itemx -fdump-rtl-bbro
3494 @opindex fdump-rtl-bbro
3495 Dump after block reordering, to @file{@var{file}.30.bbro}.
3498 @itemx -fdump-rtl-combine
3500 @opindex fdump-rtl-combine
3501 Dump after instruction combination, to the file @file{@var{file}.17.combine}.
3504 @itemx -fdump-rtl-ce1
3505 @itemx -fdump-rtl-ce2
3507 @opindex fdump-rtl-ce1
3508 @opindex fdump-rtl-ce2
3509 @option{-dC} and @option{-fdump-rtl-ce1} enable dumping after the
3510 first if conversion, to the file @file{@var{file}.11.ce1}. @option{-dC}
3511 and @option{-fdump-rtl-ce2} enable dumping after the second if
3512 conversion, to the file @file{@var{file}.18.ce2}.
3515 @itemx -fdump-rtl-btl
3516 @itemx -fdump-rtl-dbr
3518 @opindex fdump-rtl-btl
3519 @opindex fdump-rtl-dbr
3520 @option{-dd} and @option{-fdump-rtl-btl} enable dumping after branch
3521 target load optimization, to to @file{@var{file}.31.btl}. @option{-dd}
3522 and @option{-fdump-rtl-dbr} enable dumping after delayed branch
3523 scheduling, to @file{@var{file}.36.dbr}.
3527 Dump all macro definitions, at the end of preprocessing, in addition to
3531 @itemx -fdump-rtl-ce3
3533 @opindex fdump-rtl-ce3
3534 Dump after the third if conversion, to @file{@var{file}.28.ce3}.
3537 @itemx -fdump-rtl-cfg
3538 @itemx -fdump-rtl-life
3540 @opindex fdump-rtl-cfg
3541 @opindex fdump-rtl-life
3542 @option{-df} and @option{-fdump-rtl-cfg} enable dumping after control
3543 and data flow analysis, to @file{@var{file}.08.cfg}. @option{-df}
3544 and @option{-fdump-rtl-cfg} enable dumping dump after life analysis,
3545 to @file{@var{file}.16.life}.
3548 @itemx -fdump-rtl-greg
3550 @opindex fdump-rtl-greg
3551 Dump after global register allocation, to @file{@var{file}.23.greg}.
3554 @itemx -fdump-rtl-gcse
3555 @itemx -fdump-rtl-bypass
3557 @opindex fdump-rtl-gcse
3558 @opindex fdump-rtl-bypass
3559 @option{-dG} and @option{-fdump-rtl-gcse} enable dumping after GCSE, to
3560 @file{@var{file}.05.gcse}. @option{-dG} and @option{-fdump-rtl-bypass}
3561 enable dumping after jump bypassing and control flow optimizations, to
3562 @file{@var{file}.07.bypass}.
3565 @itemx -fdump-rtl-eh
3567 @opindex fdump-rtl-eh
3568 Dump after finalization of EH handling code, to @file{@var{file}.02.eh}.
3571 @itemx -fdump-rtl-sibling
3573 @opindex fdump-rtl-sibling
3574 Dump after sibling call optimizations, to @file{@var{file}.01.sibling}.
3577 @itemx -fdump-rtl-jump
3579 @opindex fdump-rtl-jump
3580 Dump after the first jump optimization, to @file{@var{file}.03.jump}.
3583 @itemx -fdump-rtl-stack
3585 @opindex fdump-rtl-stack
3586 Dump after conversion from registers to stack, to @file{@var{file}.33.stack}.
3589 @itemx -fdump-rtl-lreg
3591 @opindex fdump-rtl-lreg
3592 Dump after local register allocation, to @file{@var{file}.22.lreg}.
3595 @itemx -fdump-rtl-loop
3596 @itemx -fdump-rtl-loop2
3598 @opindex fdump-rtl-loop
3599 @opindex fdump-rtl-loop2
3600 @option{-dL} and @option{-fdump-rtl-loop} enable dumping after the first
3601 loop optimization pass, to @file{@var{file}.06.loop}. @option{-dL} and
3602 @option{-fdump-rtl-loop2} enable dumping after the second pass, to
3603 @file{@var{file}.13.loop2}.
3606 @itemx -fdump-rtl-sms
3608 @opindex fdump-rtl-sms
3609 Dump after modulo scheduling, to @file{@var{file}.20.sms}.
3612 @itemx -fdump-rtl-mach
3614 @opindex fdump-rtl-mach
3615 Dump after performing the machine dependent reorganization pass, to
3616 @file{@var{file}.35.mach}.
3619 @itemx -fdump-rtl-rnreg
3621 @opindex fdump-rtl-rnreg
3622 Dump after register renumbering, to @file{@var{file}.29.rnreg}.
3625 @itemx -fdump-rtl-regmove
3627 @opindex fdump-rtl-regmove
3628 Dump after the register move pass, to @file{@var{file}.19.regmove}.
3631 @itemx -fdump-rtl-postreload
3633 @opindex fdump-rtl-postreload
3634 Dump after post-reload optimizations, to @file{@var{file}.24.postreload}.
3637 @itemx -fdump-rtl-expand
3639 @opindex fdump-rtl-expand
3640 Dump after RTL generation, to @file{@var{file}.00.expand}.
3643 @itemx -fdump-rtl-sched2
3645 @opindex fdump-rtl-sched2
3646 Dump after the second scheduling pass, to @file{@var{file}.32.sched2}.
3649 @itemx -fdump-rtl-cse
3651 @opindex fdump-rtl-cse
3652 Dump after CSE (including the jump optimization that sometimes follows
3653 CSE), to @file{@var{file}.04.cse}.
3656 @itemx -fdump-rtl-sched
3658 @opindex fdump-rtl-sched
3659 Dump after the first scheduling pass, to @file{@var{file}.21.sched}.
3662 @itemx -fdump-rtl-cse2
3664 @opindex fdump-rtl-cse2
3665 Dump after the second CSE pass (including the jump optimization that
3666 sometimes follows CSE), to @file{@var{file}.15.cse2}.
3669 @itemx -fdump-rtl-tracer
3671 @opindex fdump-rtl-tracer
3672 Dump after running tracer, to @file{@var{file}.12.tracer}.
3675 @itemx -fdump-rtl-vpt
3676 @itemx -fdump-rtl-vartrack
3678 @opindex fdump-rtl-vpt
3679 @opindex fdump-rtl-vartrack
3680 @option{-dV} and @option{-fdump-rtl-vpt} enable dumping after the value
3681 profile transformations, to @file{@var{file}.10.vpt}. @option{-dV}
3682 and @option{-fdump-rtl-vartrack} enable dumping after variable tracking,
3683 to @file{@var{file}.34.vartrack}.
3686 @itemx -fdump-rtl-flow2
3688 @opindex fdump-rtl-flow2
3689 Dump after the second flow pass, to @file{@var{file}.26.flow2}.
3692 @itemx -fdump-rtl-peephole2
3694 @opindex fdump-rtl-peephole2
3695 Dump after the peephole pass, to @file{@var{file}.27.peephole2}.
3698 @itemx -fdump-rtl-web
3700 @opindex fdump-rtl-web
3701 Dump after live range splitting, to @file{@var{file}.14.web}.
3704 @itemx -fdump-rtl-all
3706 @opindex fdump-rtl-all
3707 Produce all the dumps listed above.
3711 Produce a core dump whenever an error occurs.
3715 Print statistics on memory usage, at the end of the run, to
3720 Annotate the assembler output with a comment indicating which
3721 pattern and alternative was used. The length of each instruction is
3726 Dump the RTL in the assembler output as a comment before each instruction.
3727 Also turns on @option{-dp} annotation.
3731 For each of the other indicated dump files (either with @option{-d} or
3732 @option{-fdump-rtl-@var{pass}}), dump a representation of the control flow
3733 graph suitable for viewing with VCG to @file{@var{file}.@var{pass}.vcg}.
3737 Just generate RTL for a function instead of compiling it. Usually used
3738 with @samp{r} (@option{-fdump-rtl-expand}).
3742 Dump debugging information during parsing, to standard error.
3745 @item -fdump-unnumbered
3746 @opindex fdump-unnumbered
3747 When doing debugging dumps (see @option{-d} option above), suppress instruction
3748 numbers and line number note output. This makes it more feasible to
3749 use diff on debugging dumps for compiler invocations with different
3750 options, in particular with and without @option{-g}.
3752 @item -fdump-translation-unit @r{(C and C++ only)}
3753 @itemx -fdump-translation-unit-@var{options} @r{(C and C++ only)}
3754 @opindex fdump-translation-unit
3755 Dump a representation of the tree structure for the entire translation
3756 unit to a file. The file name is made by appending @file{.tu} to the
3757 source file name. If the @samp{-@var{options}} form is used, @var{options}
3758 controls the details of the dump as described for the
3759 @option{-fdump-tree} options.
3761 @item -fdump-class-hierarchy @r{(C++ only)}
3762 @itemx -fdump-class-hierarchy-@var{options} @r{(C++ only)}
3763 @opindex fdump-class-hierarchy
3764 Dump a representation of each class's hierarchy and virtual function
3765 table layout to a file. The file name is made by appending @file{.class}
3766 to the source file name. If the @samp{-@var{options}} form is used,
3767 @var{options} controls the details of the dump as described for the
3768 @option{-fdump-tree} options.
3770 @item -fdump-ipa-@var{switch}
3772 Control the dumping at various stages of inter-procedural analysis
3773 language tree to a file. The file name is generated by appending a switch
3774 specific suffix to the source file name. The following dumps are possible:
3778 Enables all inter-procedural analysis dumps; currently the only produced
3779 dump is the @samp{cgraph} dump.
3782 Dumps information about call-graph optimization, unused function removal,
3783 and inlining decisions.
3786 @item -fdump-tree-@var{switch} @r{(C and C++ only)}
3787 @itemx -fdump-tree-@var{switch}-@var{options} @r{(C and C++ only)}
3789 Control the dumping at various stages of processing the intermediate
3790 language tree to a file. The file name is generated by appending a switch
3791 specific suffix to the source file name. If the @samp{-@var{options}}
3792 form is used, @var{options} is a list of @samp{-} separated options that
3793 control the details of the dump. Not all options are applicable to all
3794 dumps, those which are not meaningful will be ignored. The following
3795 options are available
3799 Print the address of each node. Usually this is not meaningful as it
3800 changes according to the environment and source file. Its primary use
3801 is for tying up a dump file with a debug environment.
3803 Inhibit dumping of members of a scope or body of a function merely
3804 because that scope has been reached. Only dump such items when they
3805 are directly reachable by some other path. When dumping pretty-printed
3806 trees, this option inhibits dumping the bodies of control structures.
3808 Print a raw representation of the tree. By default, trees are
3809 pretty-printed into a C-like representation.
3811 Enable more detailed dumps (not honored by every dump option).
3813 Enable dumping various statistics about the pass (not honored by every dump
3816 Enable showing basic block boundaries (disabled in raw dumps).
3818 Enable showing virtual operands for every statement.
3820 Enable showing line numbers for statements.
3822 Enable showing the unique ID (@code{DECL_UID}) for each variable.
3824 Turn on all options, except @option{raw}, @option{slim} and @option{lineno}.
3827 The following tree dumps are possible:
3831 Dump before any tree based optimization, to @file{@var{file}.original}.
3834 Dump after all tree based optimization, to @file{@var{file}.optimized}.
3837 Dump after function inlining, to @file{@var{file}.inlined}.
3840 @opindex fdump-tree-gimple
3841 Dump each function before and after the gimplification pass to a file. The
3842 file name is made by appending @file{.gimple} to the source file name.
3845 @opindex fdump-tree-cfg
3846 Dump the control flow graph of each function to a file. The file name is
3847 made by appending @file{.cfg} to the source file name.
3850 @opindex fdump-tree-vcg
3851 Dump the control flow graph of each function to a file in VCG format. The
3852 file name is made by appending @file{.vcg} to the source file name. Note
3853 that if the file contains more than one function, the generated file cannot
3854 be used directly by VCG@. You will need to cut and paste each function's
3855 graph into its own separate file first.
3858 @opindex fdump-tree-ch
3859 Dump each function after copying loop headers. The file name is made by
3860 appending @file{.ch} to the source file name.
3863 @opindex fdump-tree-ssa
3864 Dump SSA related information to a file. The file name is made by appending
3865 @file{.ssa} to the source file name.
3868 @opindex fdump-tree-salias
3869 Dump structure aliasing variable information to a file. This file name
3870 is made by appending @file{.salias} to the source file name.
3873 @opindex fdump-tree-alias
3874 Dump aliasing information for each function. The file name is made by
3875 appending @file{.alias} to the source file name.
3878 @opindex fdump-tree-ccp
3879 Dump each function after CCP@. The file name is made by appending
3880 @file{.ccp} to the source file name.
3883 @opindex fdump-tree-pre
3884 Dump trees after partial redundancy elimination. The file name is made
3885 by appending @file{.pre} to the source file name.
3888 @opindex fdump-tree-fre
3889 Dump trees after full redundancy elimination. The file name is made
3890 by appending @file{.fre} to the source file name.
3893 @opindex fdump-tree-dce
3894 Dump each function after dead code elimination. The file name is made by
3895 appending @file{.dce} to the source file name.
3898 @opindex fdump-tree-mudflap
3899 Dump each function after adding mudflap instrumentation. The file name is
3900 made by appending @file{.mudflap} to the source file name.
3903 @opindex fdump-tree-sra
3904 Dump each function after performing scalar replacement of aggregates. The
3905 file name is made by appending @file{.sra} to the source file name.
3908 @opindex fdump-tree-sink
3909 Dump each function after performing code sinking. The file name is made
3910 by appending @file{.sink} to the source file name.
3913 @opindex fdump-tree-dom
3914 Dump each function after applying dominator tree optimizations. The file
3915 name is made by appending @file{.dom} to the source file name.
3918 @opindex fdump-tree-dse
3919 Dump each function after applying dead store elimination. The file
3920 name is made by appending @file{.dse} to the source file name.
3923 @opindex fdump-tree-phiopt
3924 Dump each function after optimizing PHI nodes into straightline code. The file
3925 name is made by appending @file{.phiopt} to the source file name.
3928 @opindex fdump-tree-forwprop
3929 Dump each function after forward propagating single use variables. The file
3930 name is made by appending @file{.forwprop} to the source file name.
3933 @opindex fdump-tree-copyrename
3934 Dump each function after applying the copy rename optimization. The file
3935 name is made by appending @file{.copyrename} to the source file name.
3938 @opindex fdump-tree-nrv
3939 Dump each function after applying the named return value optimization on
3940 generic trees. The file name is made by appending @file{.nrv} to the source
3944 @opindex fdump-tree-vect
3945 Dump each function after applying vectorization of loops. The file name is
3946 made by appending @file{.vect} to the source file name.
3949 @opindex fdump-tree-all
3950 Enable all the available tree dumps with the flags provided in this option.
3953 @item -ftree-vectorizer-verbose=@var{n}
3954 @opindex ftree-vectorizer-verbose
3955 This option controls the amount of debugging output the vectorizer prints.
3956 This information is written to standard error, unless @option{-fdump-tree-all}
3957 or @option{-fdump-tree-vect} is specified, in which case it is output to the
3958 usual dump listing file, @file{.vect}.
3960 @item -frandom-seed=@var{string}
3961 @opindex frandom-string
3962 This option provides a seed that GCC uses when it would otherwise use
3963 random numbers. It is used to generate certain symbol names
3964 that have to be different in every compiled file. It is also used to
3965 place unique stamps in coverage data files and the object files that
3966 produce them. You can use the @option{-frandom-seed} option to produce
3967 reproducibly identical object files.
3969 The @var{string} should be different for every file you compile.
3971 @item -fsched-verbose=@var{n}
3972 @opindex fsched-verbose
3973 On targets that use instruction scheduling, this option controls the
3974 amount of debugging output the scheduler prints. This information is
3975 written to standard error, unless @option{-dS} or @option{-dR} is
3976 specified, in which case it is output to the usual dump
3977 listing file, @file{.sched} or @file{.sched2} respectively. However
3978 for @var{n} greater than nine, the output is always printed to standard
3981 For @var{n} greater than zero, @option{-fsched-verbose} outputs the
3982 same information as @option{-dRS}. For @var{n} greater than one, it
3983 also output basic block probabilities, detailed ready list information
3984 and unit/insn info. For @var{n} greater than two, it includes RTL
3985 at abort point, control-flow and regions info. And for @var{n} over
3986 four, @option{-fsched-verbose} also includes dependence info.
3990 Store the usual ``temporary'' intermediate files permanently; place them
3991 in the current directory and name them based on the source file. Thus,
3992 compiling @file{foo.c} with @samp{-c -save-temps} would produce files
3993 @file{foo.i} and @file{foo.s}, as well as @file{foo.o}. This creates a
3994 preprocessed @file{foo.i} output file even though the compiler now
3995 normally uses an integrated preprocessor.
3997 When used in combination with the @option{-x} command line option,
3998 @option{-save-temps} is sensible enough to avoid over writing an
3999 input source file with the same extension as an intermediate file.
4000 The corresponding intermediate file may be obtained by renaming the
4001 source file before using @option{-save-temps}.
4005 Report the CPU time taken by each subprocess in the compilation
4006 sequence. For C source files, this is the compiler proper and assembler
4007 (plus the linker if linking is done). The output looks like this:
4014 The first number on each line is the ``user time'', that is time spent
4015 executing the program itself. The second number is ``system time'',
4016 time spent executing operating system routines on behalf of the program.
4017 Both numbers are in seconds.
4019 @item -fvar-tracking
4020 @opindex fvar-tracking
4021 Run variable tracking pass. It computes where variables are stored at each
4022 position in code. Better debugging information is then generated
4023 (if the debugging information format supports this information).
4025 It is enabled by default when compiling with optimization (@option{-Os},
4026 @option{-O}, @option{-O2}, ...), debugging information (@option{-g}) and
4027 the debug info format supports it.
4029 @item -print-file-name=@var{library}
4030 @opindex print-file-name
4031 Print the full absolute name of the library file @var{library} that
4032 would be used when linking---and don't do anything else. With this
4033 option, GCC does not compile or link anything; it just prints the
4036 @item -print-multi-directory
4037 @opindex print-multi-directory
4038 Print the directory name corresponding to the multilib selected by any
4039 other switches present in the command line. This directory is supposed
4040 to exist in @env{GCC_EXEC_PREFIX}.
4042 @item -print-multi-lib
4043 @opindex print-multi-lib
4044 Print the mapping from multilib directory names to compiler switches
4045 that enable them. The directory name is separated from the switches by
4046 @samp{;}, and each switch starts with an @samp{@@} instead of the
4047 @samp{-}, without spaces between multiple switches. This is supposed to
4048 ease shell-processing.
4050 @item -print-prog-name=@var{program}
4051 @opindex print-prog-name
4052 Like @option{-print-file-name}, but searches for a program such as @samp{cpp}.
4054 @item -print-libgcc-file-name
4055 @opindex print-libgcc-file-name
4056 Same as @option{-print-file-name=libgcc.a}.
4058 This is useful when you use @option{-nostdlib} or @option{-nodefaultlibs}
4059 but you do want to link with @file{libgcc.a}. You can do
4062 gcc -nostdlib @var{files}@dots{} `gcc -print-libgcc-file-name`
4065 @item -print-search-dirs
4066 @opindex print-search-dirs
4067 Print the name of the configured installation directory and a list of
4068 program and library directories @command{gcc} will search---and don't do anything else.
4070 This is useful when @command{gcc} prints the error message
4071 @samp{installation problem, cannot exec cpp0: No such file or directory}.
4072 To resolve this you either need to put @file{cpp0} and the other compiler
4073 components where @command{gcc} expects to find them, or you can set the environment
4074 variable @env{GCC_EXEC_PREFIX} to the directory where you installed them.
4075 Don't forget the trailing @samp{/}.
4076 @xref{Environment Variables}.
4079 @opindex dumpmachine
4080 Print the compiler's target machine (for example,
4081 @samp{i686-pc-linux-gnu})---and don't do anything else.
4084 @opindex dumpversion
4085 Print the compiler version (for example, @samp{3.0})---and don't do
4090 Print the compiler's built-in specs---and don't do anything else. (This
4091 is used when GCC itself is being built.) @xref{Spec Files}.
4093 @item -feliminate-unused-debug-types
4094 @opindex feliminate-unused-debug-types
4095 Normally, when producing DWARF2 output, GCC will emit debugging
4096 information for all types declared in a compilation
4097 unit, regardless of whether or not they are actually used
4098 in that compilation unit. Sometimes this is useful, such as
4099 if, in the debugger, you want to cast a value to a type that is
4100 not actually used in your program (but is declared). More often,
4101 however, this results in a significant amount of wasted space.
4102 With this option, GCC will avoid producing debug symbol output
4103 for types that are nowhere used in the source file being compiled.
4106 @node Optimize Options
4107 @section Options That Control Optimization
4108 @cindex optimize options
4109 @cindex options, optimization
4111 These options control various sorts of optimizations.
4113 Without any optimization option, the compiler's goal is to reduce the
4114 cost of compilation and to make debugging produce the expected
4115 results. Statements are independent: if you stop the program with a
4116 breakpoint between statements, you can then assign a new value to any
4117 variable or change the program counter to any other statement in the
4118 function and get exactly the results you would expect from the source
4121 Turning on optimization flags makes the compiler attempt to improve
4122 the performance and/or code size at the expense of compilation time
4123 and possibly the ability to debug the program.
4125 The compiler performs optimization based on the knowledge it has of
4126 the program. Optimization levels @option{-O2} and above, in
4127 particular, enable @emph{unit-at-a-time} mode, which allows the
4128 compiler to consider information gained from later functions in
4129 the file when compiling a function. Compiling multiple files at
4130 once to a single output file in @emph{unit-at-a-time} mode allows
4131 the compiler to use information gained from all of the files when
4132 compiling each of them.
4134 Not all optimizations are controlled directly by a flag. Only
4135 optimizations that have a flag are listed.
4142 Optimize. Optimizing compilation takes somewhat more time, and a lot
4143 more memory for a large function.
4145 With @option{-O}, the compiler tries to reduce code size and execution
4146 time, without performing any optimizations that take a great deal of
4149 @option{-O} turns on the following optimization flags:
4150 @gccoptlist{-fdefer-pop @gol
4151 -fdelayed-branch @gol
4152 -fguess-branch-probability @gol
4153 -fcprop-registers @gol
4154 -floop-optimize @gol
4155 -fif-conversion @gol
4156 -fif-conversion2 @gol
4164 -ftree-copyrename @gol
4169 @option{-O} also turns on @option{-fomit-frame-pointer} on machines
4170 where doing so does not interfere with debugging.
4174 Optimize even more. GCC performs nearly all supported optimizations
4175 that do not involve a space-speed tradeoff. The compiler does not
4176 perform loop unrolling or function inlining when you specify @option{-O2}.
4177 As compared to @option{-O}, this option increases both compilation time
4178 and the performance of the generated code.
4180 @option{-O2} turns on all optimization flags specified by @option{-O}. It
4181 also turns on the following optimization flags:
4182 @gccoptlist{-fthread-jumps @gol
4184 -foptimize-sibling-calls @gol
4185 -fcse-follow-jumps -fcse-skip-blocks @gol
4186 -fgcse -fgcse-lm @gol
4187 -fexpensive-optimizations @gol
4188 -fstrength-reduce @gol
4189 -frerun-cse-after-loop -frerun-loop-opt @gol
4193 -fschedule-insns -fschedule-insns2 @gol
4194 -fsched-interblock -fsched-spec @gol
4196 -fstrict-aliasing @gol
4197 -fdelete-null-pointer-checks @gol
4198 -freorder-blocks -freorder-functions @gol
4199 -funit-at-a-time @gol
4200 -falign-functions -falign-jumps @gol
4201 -falign-loops -falign-labels @gol
4204 Please note the warning under @option{-fgcse} about
4205 invoking @option{-O2} on programs that use computed gotos.
4209 Optimize yet more. @option{-O3} turns on all optimizations specified by
4210 @option{-O2} and also turns on the @option{-finline-functions},
4211 @option{-funswitch-loops} and @option{-fgcse-after-reload} options.
4215 Do not optimize. This is the default.
4219 Optimize for size. @option{-Os} enables all @option{-O2} optimizations that
4220 do not typically increase code size. It also performs further
4221 optimizations designed to reduce code size.
4223 @option{-Os} disables the following optimization flags:
4224 @gccoptlist{-falign-functions -falign-jumps -falign-loops @gol
4225 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays}
4227 If you use multiple @option{-O} options, with or without level numbers,
4228 the last such option is the one that is effective.
4231 Options of the form @option{-f@var{flag}} specify machine-independent
4232 flags. Most flags have both positive and negative forms; the negative
4233 form of @option{-ffoo} would be @option{-fno-foo}. In the table
4234 below, only one of the forms is listed---the one you typically will
4235 use. You can figure out the other form by either removing @samp{no-}
4238 The following options control specific optimizations. They are either
4239 activated by @option{-O} options or are related to ones that are. You
4240 can use the following flags in the rare cases when ``fine-tuning'' of
4241 optimizations to be performed is desired.
4244 @item -fno-default-inline
4245 @opindex fno-default-inline
4246 Do not make member functions inline by default merely because they are
4247 defined inside the class scope (C++ only). Otherwise, when you specify
4248 @w{@option{-O}}, member functions defined inside class scope are compiled
4249 inline by default; i.e., you don't need to add @samp{inline} in front of
4250 the member function name.
4252 @item -fno-defer-pop
4253 @opindex fno-defer-pop
4254 Always pop the arguments to each function call as soon as that function
4255 returns. For machines which must pop arguments after a function call,
4256 the compiler normally lets arguments accumulate on the stack for several
4257 function calls and pops them all at once.
4259 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4263 Force memory operands to be copied into registers before doing
4264 arithmetic on them. This produces better code by making all memory
4265 references potential common subexpressions. When they are not common
4266 subexpressions, instruction combination should eliminate the separate
4269 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4272 @opindex fforce-addr
4273 Force memory address constants to be copied into registers before
4274 doing arithmetic on them. This may produce better code just as
4275 @option{-fforce-mem} may.
4277 @item -fomit-frame-pointer
4278 @opindex fomit-frame-pointer
4279 Don't keep the frame pointer in a register for functions that
4280 don't need one. This avoids the instructions to save, set up and
4281 restore frame pointers; it also makes an extra register available
4282 in many functions. @strong{It also makes debugging impossible on
4285 On some machines, such as the VAX, this flag has no effect, because
4286 the standard calling sequence automatically handles the frame pointer
4287 and nothing is saved by pretending it doesn't exist. The
4288 machine-description macro @code{FRAME_POINTER_REQUIRED} controls
4289 whether a target machine supports this flag. @xref{Registers,,Register
4290 Usage, gccint, GNU Compiler Collection (GCC) Internals}.
4292 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4294 @item -foptimize-sibling-calls
4295 @opindex foptimize-sibling-calls
4296 Optimize sibling and tail recursive calls.
4298 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4302 Don't pay attention to the @code{inline} keyword. Normally this option
4303 is used to keep the compiler from expanding any functions inline.
4304 Note that if you are not optimizing, no functions can be expanded inline.
4306 @item -finline-functions
4307 @opindex finline-functions
4308 Integrate all simple functions into their callers. The compiler
4309 heuristically decides which functions are simple enough to be worth
4310 integrating in this way.
4312 If all calls to a given function are integrated, and the function is
4313 declared @code{static}, then the function is normally not output as
4314 assembler code in its own right.
4316 Enabled at level @option{-O3}.
4318 @item -finline-limit=@var{n}
4319 @opindex finline-limit
4320 By default, GCC limits the size of functions that can be inlined. This flag
4321 allows the control of this limit for functions that are explicitly marked as
4322 inline (i.e., marked with the inline keyword or defined within the class
4323 definition in c++). @var{n} is the size of functions that can be inlined in
4324 number of pseudo instructions (not counting parameter handling). The default
4325 value of @var{n} is 600.
4326 Increasing this value can result in more inlined code at
4327 the cost of compilation time and memory consumption. Decreasing usually makes
4328 the compilation faster and less code will be inlined (which presumably
4329 means slower programs). This option is particularly useful for programs that
4330 use inlining heavily such as those based on recursive templates with C++.
4332 Inlining is actually controlled by a number of parameters, which may be
4333 specified individually by using @option{--param @var{name}=@var{value}}.
4334 The @option{-finline-limit=@var{n}} option sets some of these parameters
4338 @item max-inline-insns-single
4339 is set to @var{n}/2.
4340 @item max-inline-insns-auto
4341 is set to @var{n}/2.
4342 @item min-inline-insns
4343 is set to 130 or @var{n}/4, whichever is smaller.
4344 @item max-inline-insns-rtl
4348 See below for a documentation of the individual
4349 parameters controlling inlining.
4351 @emph{Note:} pseudo instruction represents, in this particular context, an
4352 abstract measurement of function's size. In no way, it represents a count
4353 of assembly instructions and as such its exact meaning might change from one
4354 release to an another.
4356 @item -fkeep-inline-functions
4357 @opindex fkeep-inline-functions
4358 In C, emit @code{static} functions that are declared @code{inline}
4359 into the object file, even if the function has been inlined into all
4360 of its callers. This switch does not affect functions using the
4361 @code{extern inline} extension in GNU C@. In C++, emit any and all
4362 inline functions into the object file.
4364 @item -fkeep-static-consts
4365 @opindex fkeep-static-consts
4366 Emit variables declared @code{static const} when optimization isn't turned
4367 on, even if the variables aren't referenced.
4369 GCC enables this option by default. If you want to force the compiler to
4370 check if the variable was referenced, regardless of whether or not
4371 optimization is turned on, use the @option{-fno-keep-static-consts} option.
4373 @item -fmerge-constants
4374 Attempt to merge identical constants (string constants and floating point
4375 constants) across compilation units.
4377 This option is the default for optimized compilation if the assembler and
4378 linker support it. Use @option{-fno-merge-constants} to inhibit this
4381 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4383 @item -fmerge-all-constants
4384 Attempt to merge identical constants and identical variables.
4386 This option implies @option{-fmerge-constants}. In addition to
4387 @option{-fmerge-constants} this considers e.g.@: even constant initialized
4388 arrays or initialized constant variables with integral or floating point
4389 types. Languages like C or C++ require each non-automatic variable to
4390 have distinct location, so using this option will result in non-conforming
4393 @item -fmodulo-sched
4394 @opindex fmodulo-sched
4395 Perform swing modulo scheduling immediately before the first scheduling
4396 pass. This pass looks at innermost loops and reorders their
4397 instructions by overlapping different iterations.
4399 @item -fno-branch-count-reg
4400 @opindex fno-branch-count-reg
4401 Do not use ``decrement and branch'' instructions on a count register,
4402 but instead generate a sequence of instructions that decrement a
4403 register, compare it against zero, then branch based upon the result.
4404 This option is only meaningful on architectures that support such
4405 instructions, which include x86, PowerPC, IA-64 and S/390.
4407 The default is @option{-fbranch-count-reg}, enabled when
4408 @option{-fstrength-reduce} is enabled.
4410 @item -fno-function-cse
4411 @opindex fno-function-cse
4412 Do not put function addresses in registers; make each instruction that
4413 calls a constant function contain the function's address explicitly.
4415 This option results in less efficient code, but some strange hacks
4416 that alter the assembler output may be confused by the optimizations
4417 performed when this option is not used.
4419 The default is @option{-ffunction-cse}
4421 @item -fno-zero-initialized-in-bss
4422 @opindex fno-zero-initialized-in-bss
4423 If the target supports a BSS section, GCC by default puts variables that
4424 are initialized to zero into BSS@. This can save space in the resulting
4427 This option turns off this behavior because some programs explicitly
4428 rely on variables going to the data section. E.g., so that the
4429 resulting executable can find the beginning of that section and/or make
4430 assumptions based on that.
4432 The default is @option{-fzero-initialized-in-bss}.
4434 @item -fbounds-check
4435 @opindex fbounds-check
4436 For front-ends that support it, generate additional code to check that
4437 indices used to access arrays are within the declared range. This is
4438 currently only supported by the Java and Fortran front-ends, where
4439 this option defaults to true and false respectively.
4441 @item -fmudflap -fmudflapth -fmudflapir
4445 @cindex bounds checking
4447 For front-ends that support it (C and C++), instrument all risky
4448 pointer/array dereferencing operations, some standard library
4449 string/heap functions, and some other associated constructs with
4450 range/validity tests. Modules so instrumented should be immune to
4451 buffer overflows, invalid heap use, and some other classes of C/C++
4452 programming errors. The instrumentation relies on a separate runtime
4453 library (@file{libmudflap}), which will be linked into a program if
4454 @option{-fmudflap} is given at link time. Run-time behavior of the
4455 instrumented program is controlled by the @env{MUDFLAP_OPTIONS}
4456 environment variable. See @code{env MUDFLAP_OPTIONS=-help a.out}
4459 Use @option{-fmudflapth} instead of @option{-fmudflap} to compile and to
4460 link if your program is multi-threaded. Use @option{-fmudflapir}, in
4461 addition to @option{-fmudflap} or @option{-fmudflapth}, if
4462 instrumentation should ignore pointer reads. This produces less
4463 instrumentation (and therefore faster execution) and still provides
4464 some protection against outright memory corrupting writes, but allows
4465 erroneously read data to propagate within a program.
4467 @item -fstrength-reduce
4468 @opindex fstrength-reduce
4469 Perform the optimizations of loop strength reduction and
4470 elimination of iteration variables.
4472 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4474 @item -fthread-jumps
4475 @opindex fthread-jumps
4476 Perform optimizations where we check to see if a jump branches to a
4477 location where another comparison subsumed by the first is found. If
4478 so, the first branch is redirected to either the destination of the
4479 second branch or a point immediately following it, depending on whether
4480 the condition is known to be true or false.
4482 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4484 @item -fcse-follow-jumps
4485 @opindex fcse-follow-jumps
4486 In common subexpression elimination, scan through jump instructions
4487 when the target of the jump is not reached by any other path. For
4488 example, when CSE encounters an @code{if} statement with an
4489 @code{else} clause, CSE will follow the jump when the condition
4492 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4494 @item -fcse-skip-blocks
4495 @opindex fcse-skip-blocks
4496 This is similar to @option{-fcse-follow-jumps}, but causes CSE to
4497 follow jumps which conditionally skip over blocks. When CSE
4498 encounters a simple @code{if} statement with no else clause,
4499 @option{-fcse-skip-blocks} causes CSE to follow the jump around the
4500 body of the @code{if}.
4502 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4504 @item -frerun-cse-after-loop
4505 @opindex frerun-cse-after-loop
4506 Re-run common subexpression elimination after loop optimizations has been
4509 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4511 @item -frerun-loop-opt
4512 @opindex frerun-loop-opt
4513 Run the loop optimizer twice.
4515 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4519 Perform a global common subexpression elimination pass.
4520 This pass also performs global constant and copy propagation.
4522 @emph{Note:} When compiling a program using computed gotos, a GCC
4523 extension, you may get better runtime performance if you disable
4524 the global common subexpression elimination pass by adding
4525 @option{-fno-gcse} to the command line.
4527 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4531 When @option{-fgcse-lm} is enabled, global common subexpression elimination will
4532 attempt to move loads which are only killed by stores into themselves. This
4533 allows a loop containing a load/store sequence to be changed to a load outside
4534 the loop, and a copy/store within the loop.
4536 Enabled by default when gcse is enabled.
4540 When @option{-fgcse-sm} is enabled, a store motion pass is run after
4541 global common subexpression elimination. This pass will attempt to move
4542 stores out of loops. When used in conjunction with @option{-fgcse-lm},
4543 loops containing a load/store sequence can be changed to a load before
4544 the loop and a store after the loop.
4546 Not enabled at any optimization level.
4550 When @option{-fgcse-las} is enabled, the global common subexpression
4551 elimination pass eliminates redundant loads that come after stores to the
4552 same memory location (both partial and full redundancies).
4554 Not enabled at any optimization level.
4556 @item -fgcse-after-reload
4557 @opindex fgcse-after-reload
4558 When @option{-fgcse-after-reload} is enabled, a redundant load elimination
4559 pass is performed after reload. The purpose of this pass is to cleanup
4562 @item -floop-optimize
4563 @opindex floop-optimize
4564 Perform loop optimizations: move constant expressions out of loops, simplify
4565 exit test conditions and optionally do strength-reduction as well.
4567 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4569 @item -floop-optimize2
4570 @opindex floop-optimize2
4571 Perform loop optimizations using the new loop optimizer. The optimizations
4572 (loop unrolling, peeling and unswitching, loop invariant motion) are enabled
4575 @item -fcrossjumping
4576 @opindex crossjumping
4577 Perform cross-jumping transformation. This transformation unifies equivalent code and save code size. The
4578 resulting code may or may not perform better than without cross-jumping.
4580 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4582 @item -fif-conversion
4583 @opindex if-conversion
4584 Attempt to transform conditional jumps into branch-less equivalents. This
4585 include use of conditional moves, min, max, set flags and abs instructions, and
4586 some tricks doable by standard arithmetics. The use of conditional execution
4587 on chips where it is available is controlled by @code{if-conversion2}.
4589 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4591 @item -fif-conversion2
4592 @opindex if-conversion2
4593 Use conditional execution (where available) to transform conditional jumps into
4594 branch-less equivalents.
4596 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4598 @item -fdelete-null-pointer-checks
4599 @opindex fdelete-null-pointer-checks
4600 Use global dataflow analysis to identify and eliminate useless checks
4601 for null pointers. The compiler assumes that dereferencing a null
4602 pointer would have halted the program. If a pointer is checked after
4603 it has already been dereferenced, it cannot be null.
4605 In some environments, this assumption is not true, and programs can
4606 safely dereference null pointers. Use
4607 @option{-fno-delete-null-pointer-checks} to disable this optimization
4608 for programs which depend on that behavior.
4610 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4612 @item -fexpensive-optimizations
4613 @opindex fexpensive-optimizations
4614 Perform a number of minor optimizations that are relatively expensive.
4616 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4618 @item -foptimize-register-move
4620 @opindex foptimize-register-move
4622 Attempt to reassign register numbers in move instructions and as
4623 operands of other simple instructions in order to maximize the amount of
4624 register tying. This is especially helpful on machines with two-operand
4627 Note @option{-fregmove} and @option{-foptimize-register-move} are the same
4630 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4632 @item -fdelayed-branch
4633 @opindex fdelayed-branch
4634 If supported for the target machine, attempt to reorder instructions
4635 to exploit instruction slots available after delayed branch
4638 Enabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4640 @item -fschedule-insns
4641 @opindex fschedule-insns
4642 If supported for the target machine, attempt to reorder instructions to
4643 eliminate execution stalls due to required data being unavailable. This
4644 helps machines that have slow floating point or memory load instructions
4645 by allowing other instructions to be issued until the result of the load
4646 or floating point instruction is required.
4648 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4650 @item -fschedule-insns2
4651 @opindex fschedule-insns2
4652 Similar to @option{-fschedule-insns}, but requests an additional pass of
4653 instruction scheduling after register allocation has been done. This is
4654 especially useful on machines with a relatively small number of
4655 registers and where memory load instructions take more than one cycle.
4657 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4659 @item -fno-sched-interblock
4660 @opindex fno-sched-interblock
4661 Don't schedule instructions across basic blocks. This is normally
4662 enabled by default when scheduling before register allocation, i.e.@:
4663 with @option{-fschedule-insns} or at @option{-O2} or higher.
4665 @item -fno-sched-spec
4666 @opindex fno-sched-spec
4667 Don't allow speculative motion of non-load instructions. This is normally
4668 enabled by default when scheduling before register allocation, i.e.@:
4669 with @option{-fschedule-insns} or at @option{-O2} or higher.
4671 @item -fsched-spec-load
4672 @opindex fsched-spec-load
4673 Allow speculative motion of some load instructions. This only makes
4674 sense when scheduling before register allocation, i.e.@: with
4675 @option{-fschedule-insns} or at @option{-O2} or higher.
4677 @item -fsched-spec-load-dangerous
4678 @opindex fsched-spec-load-dangerous
4679 Allow speculative motion of more load instructions. This only makes
4680 sense when scheduling before register allocation, i.e.@: with
4681 @option{-fschedule-insns} or at @option{-O2} or higher.
4683 @item -fsched-stalled-insns=@var{n}
4684 @opindex fsched-stalled-insns
4685 Define how many insns (if any) can be moved prematurely from the queue
4686 of stalled insns into the ready list, during the second scheduling pass.
4688 @item -fsched-stalled-insns-dep=@var{n}
4689 @opindex fsched-stalled-insns-dep
4690 Define how many insn groups (cycles) will be examined for a dependency
4691 on a stalled insn that is candidate for premature removal from the queue
4692 of stalled insns. Has an effect only during the second scheduling pass,
4693 and only if @option{-fsched-stalled-insns} is used and its value is not zero.
4695 @item -fsched2-use-superblocks
4696 @opindex fsched2-use-superblocks
4697 When scheduling after register allocation, do use superblock scheduling
4698 algorithm. Superblock scheduling allows motion across basic block boundaries
4699 resulting on faster schedules. This option is experimental, as not all machine
4700 descriptions used by GCC model the CPU closely enough to avoid unreliable
4701 results from the algorithm.
4703 This only makes sense when scheduling after register allocation, i.e.@: with
4704 @option{-fschedule-insns2} or at @option{-O2} or higher.
4706 @item -fsched2-use-traces
4707 @opindex fsched2-use-traces
4708 Use @option{-fsched2-use-superblocks} algorithm when scheduling after register
4709 allocation and additionally perform code duplication in order to increase the
4710 size of superblocks using tracer pass. See @option{-ftracer} for details on
4713 This mode should produce faster but significantly longer programs. Also
4714 without @option{-fbranch-probabilities} the traces constructed may not
4715 match the reality and hurt the performance. This only makes
4716 sense when scheduling after register allocation, i.e.@: with
4717 @option{-fschedule-insns2} or at @option{-O2} or higher.
4719 @item -freschedule-modulo-scheduled-loops
4720 @opindex fscheduling-in-modulo-scheduled-loops
4721 The modulo scheduling comes before the traditional scheduling, if a loop was modulo scheduled
4722 we may want to prevent the later scheduling passes from changing its schedule, we use this
4723 option to control that.
4725 @item -fcaller-saves
4726 @opindex fcaller-saves
4727 Enable values to be allocated in registers that will be clobbered by
4728 function calls, by emitting extra instructions to save and restore the
4729 registers around such calls. Such allocation is done only when it
4730 seems to result in better code than would otherwise be produced.
4732 This option is always enabled by default on certain machines, usually
4733 those which have no call-preserved registers to use instead.
4735 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4738 Perform Partial Redundancy Elimination (PRE) on trees. This flag is
4739 enabled by default at @option{-O2} and @option{-O3}.
4742 Perform Full Redundancy Elimination (FRE) on trees. The difference
4743 between FRE and PRE is that FRE only considers expressions
4744 that are computed on all paths leading to the redundant computation.
4745 This analysis faster than PRE, though it exposes fewer redundancies.
4746 This flag is enabled by default at @option{-O} and higher.
4749 Perform structural alias analysis on trees. This flag
4750 is enabled by default at @option{-O} and higher.
4753 Perform forward store motion on trees. This flag is
4754 enabled by default at @option{-O} and higher.
4757 Perform sparse conditional constant propagation (CCP) on trees. This flag
4758 is enabled by default at @option{-O} and higher.
4761 Perform dead code elimination (DCE) on trees. This flag is enabled by
4762 default at @option{-O} and higher.
4764 @item -ftree-dominator-opts
4765 Perform dead code elimination (DCE) on trees. This flag is enabled by
4766 default at @option{-O} and higher.
4769 Perform loop header copying on trees. This is beneficial since it increases
4770 effectiveness of code motion optimizations. It also saves one jump. This flag
4771 is enabled by default at @option{-O} and higher. It is not enabled
4772 for @option{-Os}, since it usually increases code size.
4774 @item -ftree-loop-optimize
4775 Perform loop optimizations on trees. This flag is enabled by default
4776 at @option{-O} and higher.
4778 @item -ftree-loop-linear
4779 Perform linear loop transformations on tree. This flag can improve cache
4780 performance and allow further loop optimizations to take place.
4782 @item -ftree-loop-im
4783 Perform loop invariant motion on trees. This pass moves only invariants that
4784 would be hard to handle on rtl level (function calls, operations that expand to
4785 nontrivial sequences of insns). With @option{-funswitch-loops} it also moves
4786 operands of conditions that are invariant out of the loop, so that we can use
4787 just trivial invariantness analysis in loop unswitching. The pass also includes
4790 @item -ftree-loop-ivcanon
4791 Create a canonical counter for number of iterations in the loop for that
4792 determining number of iterations requires complicated analysis. Later
4793 optimizations then may determine the number easily. Useful especially
4794 in connection with unrolling.
4797 Perform induction variable optimizations (strength reduction, induction
4798 variable merging and induction variable elimination) on trees.
4801 Perform scalar replacement of aggregates. This pass replaces structure
4802 references with scalars to prevent committing structures to memory too
4803 early. This flag is enabled by default at @option{-O} and higher.
4805 @item -ftree-copyrename
4806 Perform copy renaming on trees. This pass attempts to rename compiler
4807 temporaries to other variables at copy locations, usually resulting in
4808 variable names which more closely resemble the original variables. This flag
4809 is enabled by default at @option{-O} and higher.
4812 Perform temporary expression replacement during the SSA->normal phase. Single
4813 use/single def temporaries are replaced at their use location with their
4814 defining expression. This results in non-GIMPLE code, but gives the expanders
4815 much more complex trees to work on resulting in better RTL generation. This is
4816 enabled by default at @option{-O} and higher.
4819 Perform live range splitting during the SSA->normal phase. Distinct live
4820 ranges of a variable are split into unique variables, allowing for better
4821 optimization later. This is enabled by default at @option{-O} and higher.
4823 @item -ftree-vectorize
4824 Perform loop vectorization on trees.
4828 Perform tail duplication to enlarge superblock size. This transformation
4829 simplifies the control flow of the function allowing other optimizations to do
4832 @item -funroll-loops
4833 @opindex funroll-loops
4834 Unroll loops whose number of iterations can be determined at compile
4835 time or upon entry to the loop. @option{-funroll-loops} implies both
4836 @option{-fstrength-reduce} and @option{-frerun-cse-after-loop}. This
4837 option makes code larger, and may or may not make it run faster.
4839 @item -funroll-all-loops
4840 @opindex funroll-all-loops
4841 Unroll all loops, even if their number of iterations is uncertain when
4842 the loop is entered. This usually makes programs run more slowly.
4843 @option{-funroll-all-loops} implies the same options as
4844 @option{-funroll-loops},
4846 @item -fsplit-ivs-in-unroller
4847 @opindex -fsplit-ivs-in-unroller
4848 Enables expressing of values of induction variables in later iterations
4849 of the unrolled loop using the value in the first iteration. This breaks
4850 long dependency chains, thus improving efficiency of the scheduling passes
4851 (for best results, @option{-fweb} should be used as well).
4853 Combination of @option{-fweb} and CSE is often sufficient to obtain the
4854 same effect. However in cases the loop body is more complicated than
4855 a single basic block, this is not reliable. It also does not work at all
4856 on some of the architectures due to restrictions in the CSE pass.
4858 This optimization is enabled by default.
4860 @item -fvariable-expansion-in-unroller
4861 @opindex -fvariable-expansion-in-unroller
4862 With this option, the compiler will create multiple copies of some
4863 local variables when unrolling a loop which can result in superior code.
4865 @item -fprefetch-loop-arrays
4866 @opindex fprefetch-loop-arrays
4867 If supported by the target machine, generate instructions to prefetch
4868 memory to improve the performance of loops that access large arrays.
4870 These options may generate better or worse code; results are highly
4871 dependent on the structure of loops within the source code.
4874 @itemx -fno-peephole2
4875 @opindex fno-peephole
4876 @opindex fno-peephole2
4877 Disable any machine-specific peephole optimizations. The difference
4878 between @option{-fno-peephole} and @option{-fno-peephole2} is in how they
4879 are implemented in the compiler; some targets use one, some use the
4880 other, a few use both.
4882 @option{-fpeephole} is enabled by default.
4883 @option{-fpeephole2} enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4885 @item -fno-guess-branch-probability
4886 @opindex fno-guess-branch-probability
4887 Do not guess branch probabilities using heuristics.
4889 GCC will use heuristics to guess branch probabilities if they are
4890 not provided by profiling feedback (@option{-fprofile-arcs}). These
4891 heuristics are based on the control flow graph. If some branch probabilities
4892 are specified by @samp{__builtin_expect}, then the heuristics will be
4893 used to guess branch probabilities for the rest of the control flow graph,
4894 taking the @samp{__builtin_expect} info into account. The interactions
4895 between the heuristics and @samp{__builtin_expect} can be complex, and in
4896 some cases, it may be useful to disable the heuristics so that the effects
4897 of @samp{__builtin_expect} are easier to understand.
4899 The default is @option{-fguess-branch-probability} at levels
4900 @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
4902 @item -freorder-blocks
4903 @opindex freorder-blocks
4904 Reorder basic blocks in the compiled function in order to reduce number of
4905 taken branches and improve code locality.
4907 Enabled at levels @option{-O2}, @option{-O3}.
4909 @item -freorder-blocks-and-partition
4910 @opindex freorder-blocks-and-partition
4911 In addition to reordering basic blocks in the compiled function, in order
4912 to reduce number of taken branches, partitions hot and cold basic blocks
4913 into separate sections of the assembly and .o files, to improve
4914 paging and cache locality performance.
4916 This optimization is automatically turned off in the presence of
4917 exception handling, for linkonce sections, for functions with a user-defined
4918 section attribute and on any architecture that does not support named
4921 @item -freorder-functions
4922 @opindex freorder-functions
4923 Reorder functions in the object file in order to
4924 improve code locality. This is implemented by using special
4925 subsections @code{.text.hot} for most frequently executed functions and
4926 @code{.text.unlikely} for unlikely executed functions. Reordering is done by
4927 the linker so object file format must support named sections and linker must
4928 place them in a reasonable way.
4930 Also profile feedback must be available in to make this option effective. See
4931 @option{-fprofile-arcs} for details.
4933 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4935 @item -fstrict-aliasing
4936 @opindex fstrict-aliasing
4937 Allows the compiler to assume the strictest aliasing rules applicable to
4938 the language being compiled. For C (and C++), this activates
4939 optimizations based on the type of expressions. In particular, an
4940 object of one type is assumed never to reside at the same address as an
4941 object of a different type, unless the types are almost the same. For
4942 example, an @code{unsigned int} can alias an @code{int}, but not a
4943 @code{void*} or a @code{double}. A character type may alias any other
4946 Pay special attention to code like this:
4959 The practice of reading from a different union member than the one most
4960 recently written to (called ``type-punning'') is common. Even with
4961 @option{-fstrict-aliasing}, type-punning is allowed, provided the memory
4962 is accessed through the union type. So, the code above will work as
4963 expected. However, this code might not:
4974 Every language that wishes to perform language-specific alias analysis
4975 should define a function that computes, given an @code{tree}
4976 node, an alias set for the node. Nodes in different alias sets are not
4977 allowed to alias. For an example, see the C front-end function
4978 @code{c_get_alias_set}.
4980 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os}.
4982 @item -falign-functions
4983 @itemx -falign-functions=@var{n}
4984 @opindex falign-functions
4985 Align the start of functions to the next power-of-two greater than
4986 @var{n}, skipping up to @var{n} bytes. For instance,
4987 @option{-falign-functions=32} aligns functions to the next 32-byte
4988 boundary, but @option{-falign-functions=24} would align to the next
4989 32-byte boundary only if this can be done by skipping 23 bytes or less.
4991 @option{-fno-align-functions} and @option{-falign-functions=1} are
4992 equivalent and mean that functions will not be aligned.
4994 Some assemblers only support this flag when @var{n} is a power of two;
4995 in that case, it is rounded up.
4997 If @var{n} is not specified or is zero, use a machine-dependent default.
4999 Enabled at levels @option{-O2}, @option{-O3}.
5001 @item -falign-labels
5002 @itemx -falign-labels=@var{n}
5003 @opindex falign-labels
5004 Align all branch targets to a power-of-two boundary, skipping up to
5005 @var{n} bytes like @option{-falign-functions}. This option can easily
5006 make code slower, because it must insert dummy operations for when the
5007 branch target is reached in the usual flow of the code.
5009 @option{-fno-align-labels} and @option{-falign-labels=1} are
5010 equivalent and mean that labels will not be aligned.
5012 If @option{-falign-loops} or @option{-falign-jumps} are applicable and
5013 are greater than this value, then their values are used instead.
5015 If @var{n} is not specified or is zero, use a machine-dependent default
5016 which is very likely to be @samp{1}, meaning no alignment.
5018 Enabled at levels @option{-O2}, @option{-O3}.
5021 @itemx -falign-loops=@var{n}
5022 @opindex falign-loops
5023 Align loops to a power-of-two boundary, skipping up to @var{n} bytes
5024 like @option{-falign-functions}. The hope is that the loop will be
5025 executed many times, which will make up for any execution of the dummy
5028 @option{-fno-align-loops} and @option{-falign-loops=1} are
5029 equivalent and mean that loops will not be aligned.
5031 If @var{n} is not specified or is zero, use a machine-dependent default.
5033 Enabled at levels @option{-O2}, @option{-O3}.
5036 @itemx -falign-jumps=@var{n}
5037 @opindex falign-jumps
5038 Align branch targets to a power-of-two boundary, for branch targets
5039 where the targets can only be reached by jumping, skipping up to @var{n}
5040 bytes like @option{-falign-functions}. In this case, no dummy operations
5043 @option{-fno-align-jumps} and @option{-falign-jumps=1} are
5044 equivalent and mean that loops will not be aligned.
5046 If @var{n} is not specified or is zero, use a machine-dependent default.
5048 Enabled at levels @option{-O2}, @option{-O3}.
5050 @item -funit-at-a-time
5051 @opindex funit-at-a-time
5052 Parse the whole compilation unit before starting to produce code.
5053 This allows some extra optimizations to take place but consumes
5054 more memory (in general). There are some compatibility issues
5055 with @emph{unit-at-at-time} mode:
5058 enabling @emph{unit-at-a-time} mode may change the order
5059 in which functions, variables, and top-level @code{asm} statements
5060 are emitted, and will likely break code relying on some particular
5061 ordering. The majority of such top-level @code{asm} statements,
5062 though, can be replaced by @code{section} attributes.
5065 @emph{unit-at-a-time} mode removes unreferenced static variables
5066 and functions are removed. This may result in undefined references
5067 when an @code{asm} statement refers directly to variables or functions
5068 that are otherwise unused. In that case either the variable/function
5069 shall be listed as an operand of the @code{asm} statement operand or,
5070 in the case of top-level @code{asm} statements the attribute @code{used}
5071 shall be used on the declaration.
5074 Static functions now can use non-standard passing conventions that
5075 may break @code{asm} statements calling functions directly. Again,
5076 attribute @code{used} will prevent this behavior.
5079 As a temporary workaround, @option{-fno-unit-at-a-time} can be used,
5080 but this scheme may not be supported by future releases of GCC@.
5082 Enabled at levels @option{-O2}, @option{-O3}.
5086 Constructs webs as commonly used for register allocation purposes and assign
5087 each web individual pseudo register. This allows the register allocation pass
5088 to operate on pseudos directly, but also strengthens several other optimization
5089 passes, such as CSE, loop optimizer and trivial dead code remover. It can,
5090 however, make debugging impossible, since variables will no longer stay in a
5093 Enabled at levels @option{-O2}, @option{-O3}, @option{-Os},
5094 on targets where the default format for debugging information supports
5097 @item -fno-cprop-registers
5098 @opindex fno-cprop-registers
5099 After register allocation and post-register allocation instruction splitting,
5100 we perform a copy-propagation pass to try to reduce scheduling dependencies
5101 and occasionally eliminate the copy.
5103 Disabled at levels @option{-O}, @option{-O2}, @option{-O3}, @option{-Os}.
5105 @item -fprofile-generate
5106 @opindex fprofile-generate
5108 Enable options usually used for instrumenting application to produce
5109 profile useful for later recompilation with profile feedback based
5110 optimization. You must use @option{-fprofile-generate} both when
5111 compiling and when linking your program.
5113 The following options are enabled: @code{-fprofile-arcs}, @code{-fprofile-values}, @code{-fvpt}.
5116 @opindex fprofile-use
5117 Enable profile feedback directed optimizations, and optimizations
5118 generally profitable only with profile feedback available.
5120 The following options are enabled: @code{-fbranch-probabilities},
5121 @code{-fvpt}, @code{-funroll-loops}, @code{-fpeel-loops}, @code{-ftracer}.
5125 The following options control compiler behavior regarding floating
5126 point arithmetic. These options trade off between speed and
5127 correctness. All must be specifically enabled.
5131 @opindex ffloat-store
5132 Do not store floating point variables in registers, and inhibit other
5133 options that might change whether a floating point value is taken from a
5136 @cindex floating point precision
5137 This option prevents undesirable excess precision on machines such as
5138 the 68000 where the floating registers (of the 68881) keep more
5139 precision than a @code{double} is supposed to have. Similarly for the
5140 x86 architecture. For most programs, the excess precision does only
5141 good, but a few programs rely on the precise definition of IEEE floating
5142 point. Use @option{-ffloat-store} for such programs, after modifying
5143 them to store all pertinent intermediate computations into variables.
5147 Sets @option{-fno-math-errno}, @option{-funsafe-math-optimizations}, @*
5148 @option{-fno-trapping-math}, @option{-ffinite-math-only},
5149 @option{-fno-rounding-math}, @option{-fno-signaling-nans}
5150 and @option{fcx-limited-range}.
5152 This option causes the preprocessor macro @code{__FAST_MATH__} to be defined.
5154 This option should never be turned on by any @option{-O} option since
5155 it can result in incorrect output for programs which depend on
5156 an exact implementation of IEEE or ISO rules/specifications for
5159 @item -fno-math-errno
5160 @opindex fno-math-errno
5161 Do not set ERRNO after calling math functions that are executed
5162 with a single instruction, e.g., sqrt. A program that relies on
5163 IEEE exceptions for math error handling may want to use this flag
5164 for speed while maintaining IEEE arithmetic compatibility.
5166 This option should never be turned on by any @option{-O} option since
5167 it can result in incorrect output for programs which depend on
5168 an exact implementation of IEEE or ISO rules/specifications for
5171 The default is @option{-fmath-errno}.
5173 @item -funsafe-math-optimizations
5174 @opindex funsafe-math-optimizations
5175 Allow optimizations for floating-point arithmetic that (a) assume
5176 that arguments and results are valid and (b) may violate IEEE or
5177 ANSI standards. When used at link-time, it may include libraries
5178 or startup files that change the default FPU control word or other
5179 similar optimizations.
5181 This option should never be turned on by any @option{-O} option since
5182 it can result in incorrect output for programs which depend on
5183 an exact implementation of IEEE or ISO rules/specifications for
5186 The default is @option{-fno-unsafe-math-optimizations}.
5188 @item -ffinite-math-only
5189 @opindex ffinite-math-only
5190 Allow optimizations for floating-point arithmetic that assume
5191 that arguments and results are not NaNs or +-Infs.
5193 This option should never be turned on by any @option{-O} option since
5194 it can result in incorrect output for programs which depend on
5195 an exact implementation of IEEE or ISO rules/specifications.
5197 The default is @option{-fno-finite-math-only}.
5199 @item -fno-trapping-math
5200 @opindex fno-trapping-math
5201 Compile code assuming that floating-point operations cannot generate
5202 user-visible traps. These traps include division by zero, overflow,
5203 underflow, inexact result and invalid operation. This option implies
5204 @option{-fno-signaling-nans}. Setting this option may allow faster
5205 code if one relies on ``non-stop'' IEEE arithmetic, for example.
5207 This option should never be turned on by any @option{-O} option since
5208 it can result in incorrect output for programs which depend on
5209 an exact implementation of IEEE or ISO rules/specifications for
5212 The default is @option{-ftrapping-math}.
5214 @item -frounding-math
5215 @opindex frounding-math
5216 Disable transformations and optimizations that assume default floating
5217 point rounding behavior. This is round-to-zero for all floating point
5218 to integer conversions, and round-to-nearest for all other arithmetic
5219 truncations. This option should be specified for programs that change
5220 the FP rounding mode dynamically, or that may be executed with a
5221 non-default rounding mode. This option disables constant folding of
5222 floating point expressions at compile-time (which may be affected by
5223 rounding mode) and arithmetic transformations that are unsafe in the
5224 presence of sign-dependent rounding modes.
5226 The default is @option{-fno-rounding-math}.
5228 This option is experimental and does not currently guarantee to
5229 disable all GCC optimizations that are affected by rounding mode.
5230 Future versions of GCC may provide finer control of this setting
5231 using C99's @code{FENV_ACCESS} pragma. This command line option
5232 will be used to specify the default state for @code{FENV_ACCESS}.
5234 @item -fsignaling-nans
5235 @opindex fsignaling-nans
5236 Compile code assuming that IEEE signaling NaNs may generate user-visible
5237 traps during floating-point operations. Setting this option disables
5238 optimizations that may change the number of exceptions visible with
5239 signaling NaNs. This option implies @option{-ftrapping-math}.
5241 This option causes the preprocessor macro @code{__SUPPORT_SNAN__} to
5244 The default is @option{-fno-signaling-nans}.
5246 This option is experimental and does not currently guarantee to
5247 disable all GCC optimizations that affect signaling NaN behavior.
5249 @item -fsingle-precision-constant
5250 @opindex fsingle-precision-constant
5251 Treat floating point constant as single precision constant instead of
5252 implicitly converting it to double precision constant.
5254 @item -fcx-limited-range
5255 @itemx -fno-cx-limited-range
5256 @opindex fcx-limited-range
5257 @opindex fno-cx-limited-range
5258 When enabled, this option states that a range reduction step is not
5259 needed when performing complex division. The default is
5260 @option{-fno-cx-limited-range}, but is enabled by @option{-ffast-math}.
5262 This option controls the default setting of the ISO C99
5263 @code{CX_LIMITED_RANGE} pragma. Nevertheless, the option applies to
5268 The following options control optimizations that may improve
5269 performance, but are not enabled by any @option{-O} options. This
5270 section includes experimental options that may produce broken code.
5273 @item -fbranch-probabilities
5274 @opindex fbranch-probabilities
5275 After running a program compiled with @option{-fprofile-arcs}
5276 (@pxref{Debugging Options,, Options for Debugging Your Program or
5277 @command{gcc}}), you can compile it a second time using
5278 @option{-fbranch-probabilities}, to improve optimizations based on
5279 the number of times each branch was taken. When the program
5280 compiled with @option{-fprofile-arcs} exits it saves arc execution
5281 counts to a file called @file{@var{sourcename}.gcda} for each source
5282 file The information in this data file is very dependent on the
5283 structure of the generated code, so you must use the same source code
5284 and the same optimization options for both compilations.
5286 With @option{-fbranch-probabilities}, GCC puts a
5287 @samp{REG_BR_PROB} note on each @samp{JUMP_INSN} and @samp{CALL_INSN}.
5288 These can be used to improve optimization. Currently, they are only
5289 used in one place: in @file{reorg.c}, instead of guessing which path a
5290 branch is mostly to take, the @samp{REG_BR_PROB} values are used to
5291 exactly determine which path is taken more often.
5293 @item -fprofile-values
5294 @opindex fprofile-values
5295 If combined with @option{-fprofile-arcs}, it adds code so that some
5296 data about values of expressions in the program is gathered.
5298 With @option{-fbranch-probabilities}, it reads back the data gathered
5299 from profiling values of expressions and adds @samp{REG_VALUE_PROFILE}
5300 notes to instructions for their later usage in optimizations.
5302 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5306 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5307 a code to gather information about values of expressions.
5309 With @option{-fbranch-probabilities}, it reads back the data gathered
5310 and actually performs the optimizations based on them.
5311 Currently the optimizations include specialization of division operation
5312 using the knowledge about the value of the denominator.
5314 @item -fspeculative-prefetching
5315 @opindex fspeculative-prefetching
5316 If combined with @option{-fprofile-arcs}, it instructs the compiler to add
5317 a code to gather information about addresses of memory references in the
5320 With @option{-fbranch-probabilities}, it reads back the data gathered
5321 and issues prefetch instructions according to them. In addition to the opportunities
5322 noticed by @option{-fprefetch-loop-arrays}, it also notices more complicated
5323 memory access patterns---for example accesses to the data stored in linked
5324 list whose elements are usually allocated sequentially.
5326 In order to prevent issuing double prefetches, usage of
5327 @option{-fspeculative-prefetching} implies @option{-fno-prefetch-loop-arrays}.
5329 Enabled with @option{-fprofile-generate} and @option{-fprofile-use}.
5331 @item -frename-registers
5332 @opindex frename-registers
5333 Attempt to avoid false dependencies in scheduled code by making use
5334 of registers left over after register allocation. This optimization
5335 will most benefit processors with lots of registers. Depending on the
5336 debug information format adopted by the target, however, it can
5337 make debugging impossible, since variables will no longer stay in
5338 a ``home register''.
5340 Not enabled by default at any level because it has known bugs.
5344 Perform tail duplication to enlarge superblock size. This transformation
5345 simplifies the control flow of the function allowing other optimizations to do
5348 Enabled with @option{-fprofile-use}.
5350 @item -funroll-loops
5351 @opindex funroll-loops
5352 Unroll loops whose number of iterations can be determined at compile time or
5353 upon entry to the loop. @option{-funroll-loops} implies
5354 @option{-frerun-cse-after-loop}. It also turns on complete loop peeling
5355 (i.e.@: complete removal of loops with small constant number of iterations).
5356 This option makes code larger, and may or may not make it run faster.
5358 Enabled with @option{-fprofile-use}.
5360 @item -funroll-all-loops
5361 @opindex funroll-all-loops
5362 Unroll all loops, even if their number of iterations is uncertain when
5363 the loop is entered. This usually makes programs run more slowly.
5364 @option{-funroll-all-loops} implies the same options as
5365 @option{-funroll-loops}.
5368 @opindex fpeel-loops
5369 Peels the loops for that there is enough information that they do not
5370 roll much (from profile feedback). It also turns on complete loop peeling
5371 (i.e.@: complete removal of loops with small constant number of iterations).
5373 Enabled with @option{-fprofile-use}.
5375 @item -fmove-loop-invariants
5376 @opindex fmove-loop-invariants
5377 Enables the loop invariant motion pass in the new loop optimizer. Enabled
5378 at level @option{-O1}
5380 @item -funswitch-loops
5381 @opindex funswitch-loops
5382 Move branches with loop invariant conditions out of the loop, with duplicates
5383 of the loop on both branches (modified according to result of the condition).
5385 @item -fprefetch-loop-arrays
5386 @opindex fprefetch-loop-arrays
5387 If supported by the target machine, generate instructions to prefetch
5388 memory to improve the performance of loops that access large arrays.
5390 Disabled at level @option{-Os}.
5392 @item -ffunction-sections
5393 @itemx -fdata-sections
5394 @opindex ffunction-sections
5395 @opindex fdata-sections
5396 Place each function or data item into its own section in the output
5397 file if the target supports arbitrary sections. The name of the
5398 function or the name of the data item determines the section's name
5401 Use these options on systems where the linker can perform optimizations
5402 to improve locality of reference in the instruction space. Most systems
5403 using the ELF object format and SPARC processors running Solaris 2 have
5404 linkers with such optimizations. AIX may have these optimizations in
5407 Only use these options when there are significant benefits from doing
5408 so. When you specify these options, the assembler and linker will
5409 create larger object and executable files and will also be slower.
5410 You will not be able to use @code{gprof} on all systems if you
5411 specify this option and you may have problems with debugging if
5412 you specify both this option and @option{-g}.
5414 @item -fbranch-target-load-optimize
5415 @opindex fbranch-target-load-optimize
5416 Perform branch target register load optimization before prologue / epilogue
5418 The use of target registers can typically be exposed only during reload,
5419 thus hoisting loads out of loops and doing inter-block scheduling needs
5420 a separate optimization pass.
5422 @item -fbranch-target-load-optimize2
5423 @opindex fbranch-target-load-optimize2
5424 Perform branch target register load optimization after prologue / epilogue
5427 @item -fbtr-bb-exclusive
5428 @opindex fbtr-bb-exclusive
5429 When performing branch target register load optimization, don't reuse
5430 branch target registers in within any basic block.
5432 @item --param @var{name}=@var{value}
5434 In some places, GCC uses various constants to control the amount of
5435 optimization that is done. For example, GCC will not inline functions
5436 that contain more that a certain number of instructions. You can
5437 control some of these constants on the command-line using the
5438 @option{--param} option.
5440 The names of specific parameters, and the meaning of the values, are
5441 tied to the internals of the compiler, and are subject to change
5442 without notice in future releases.
5444 In each case, the @var{value} is an integer. The allowable choices for
5445 @var{name} are given in the following table:
5448 @item salias-max-implicit-fields
5449 The maximum number of fields in a variable without direct
5450 structure accesses for which structure aliasing will consider trying
5451 to track each field. The default is 5
5453 @item sra-max-structure-size
5454 The maximum structure size, in bytes, at which the scalar replacement
5455 of aggregates (SRA) optimization will perform block copies. The
5456 default value, 0, implies that GCC will select the most appropriate
5459 @item sra-field-structure-ratio
5460 The threshold ratio (as a percentage) between instantiated fields and
5461 the complete structure size. We say that if the ratio of the number
5462 of bytes in instantiated fields to the number of bytes in the complete
5463 structure exceeds this parameter, then block copies are not used. The
5466 @item max-crossjump-edges
5467 The maximum number of incoming edges to consider for crossjumping.
5468 The algorithm used by @option{-fcrossjumping} is @math{O(N^2)} in
5469 the number of edges incoming to each block. Increasing values mean
5470 more aggressive optimization, making the compile time increase with
5471 probably small improvement in executable size.
5473 @item min-crossjump-insns
5474 The minimum number of instructions which must be matched at the end
5475 of two blocks before crossjumping will be performed on them. This
5476 value is ignored in the case where all instructions in the block being
5477 crossjumped from are matched. The default value is 5.
5479 @item max-goto-duplication-insns
5480 The maximum number of instructions to duplicate to a block that jumps
5481 to a computed goto. To avoid @math{O(N^2)} behavior in a number of
5482 passes, GCC factors computed gotos early in the compilation process,
5483 and unfactors them as late as possible. Only computed jumps at the
5484 end of a basic blocks with no more than max-goto-duplication-insns are
5485 unfactored. The default value is 8.
5487 @item max-delay-slot-insn-search
5488 The maximum number of instructions to consider when looking for an
5489 instruction to fill a delay slot. If more than this arbitrary number of
5490 instructions is searched, the time savings from filling the delay slot
5491 will be minimal so stop searching. Increasing values mean more
5492 aggressive optimization, making the compile time increase with probably
5493 small improvement in executable run time.
5495 @item max-delay-slot-live-search
5496 When trying to fill delay slots, the maximum number of instructions to
5497 consider when searching for a block with valid live register
5498 information. Increasing this arbitrarily chosen value means more
5499 aggressive optimization, increasing the compile time. This parameter
5500 should be removed when the delay slot code is rewritten to maintain the
5503 @item max-gcse-memory
5504 The approximate maximum amount of memory that will be allocated in
5505 order to perform the global common subexpression elimination
5506 optimization. If more memory than specified is required, the
5507 optimization will not be done.
5509 @item max-gcse-passes
5510 The maximum number of passes of GCSE to run. The default is 1.
5512 @item max-pending-list-length
5513 The maximum number of pending dependencies scheduling will allow
5514 before flushing the current state and starting over. Large functions
5515 with few branches or calls can create excessively large lists which
5516 needlessly consume memory and resources.
5518 @item max-inline-insns-single
5519 Several parameters control the tree inliner used in gcc.
5520 This number sets the maximum number of instructions (counted in GCC's
5521 internal representation) in a single function that the tree inliner
5522 will consider for inlining. This only affects functions declared
5523 inline and methods implemented in a class declaration (C++).
5524 The default value is 450.
5526 @item max-inline-insns-auto
5527 When you use @option{-finline-functions} (included in @option{-O3}),
5528 a lot of functions that would otherwise not be considered for inlining
5529 by the compiler will be investigated. To those functions, a different
5530 (more restrictive) limit compared to functions declared inline can
5532 The default value is 90.
5534 @item large-function-insns
5535 The limit specifying really large functions. For functions larger than this
5536 limit after inlining inlining is constrained by
5537 @option{--param large-function-growth}. This parameter is useful primarily
5538 to avoid extreme compilation time caused by non-linear algorithms used by the
5540 This parameter is ignored when @option{-funit-at-a-time} is not used.
5541 The default value is 2700.
5543 @item large-function-growth
5544 Specifies maximal growth of large function caused by inlining in percents.
5545 This parameter is ignored when @option{-funit-at-a-time} is not used.
5546 The default value is 100 which limits large function growth to 2.0 times
5549 @item inline-unit-growth
5550 Specifies maximal overall growth of the compilation unit caused by inlining.
5551 This parameter is ignored when @option{-funit-at-a-time} is not used.
5552 The default value is 50 which limits unit growth to 1.5 times the original
5555 @item max-inline-insns-recursive
5556 @itemx max-inline-insns-recursive-auto
5557 Specifies maximum number of instructions out-of-line copy of self recursive inline
5558 function can grow into by performing recursive inlining.
5560 For functions declared inline @option{--param max-inline-insns-recursive} is
5561 taken into acount. For function not declared inline, recursive inlining
5562 happens only when @option{-finline-functions} (included in @option{-O3}) is
5563 enabled and @option{--param max-inline-insns-recursive-auto} is used. The
5564 default value is 450.
5566 @item max-inline-recursive-depth
5567 @itemx max-inline-recursive-depth-auto
5568 Specifies maximum recursion depth used by the recursive inlining.
5570 For functions declared inline @option{--param max-inline-recursive-depth} is
5571 taken into acount. For function not declared inline, recursive inlining
5572 happens only when @option{-finline-functions} (included in @option{-O3}) is
5573 enabled and @option{--param max-inline-recursive-depth-auto} is used. The
5574 default value is 450.
5576 @item inline-call-cost
5577 Specify cost of call instruction relative to simple arithmetics operations
5578 (having cost of 1). Increasing this cost disqualifies inlining of non-leaf
5579 functions and at the same time increases size of leaf function that is believed to
5580 reduce function size by being inlined. In effect it increases amount of
5581 inlining for code having large abstraction penalty (many functions that just
5582 pass the arguments to other functions) and decrease inlining for code with low
5583 abstraction penalty. The default value is 16.
5585 @item max-unrolled-insns
5586 The maximum number of instructions that a loop should have if that loop
5587 is unrolled, and if the loop is unrolled, it determines how many times
5588 the loop code is unrolled.
5590 @item max-average-unrolled-insns
5591 The maximum number of instructions biased by probabilities of their execution
5592 that a loop should have if that loop is unrolled, and if the loop is unrolled,
5593 it determines how many times the loop code is unrolled.
5595 @item max-unroll-times
5596 The maximum number of unrollings of a single loop.
5598 @item max-peeled-insns
5599 The maximum number of instructions that a loop should have if that loop
5600 is peeled, and if the loop is peeled, it determines how many times
5601 the loop code is peeled.
5603 @item max-peel-times
5604 The maximum number of peelings of a single loop.
5606 @item max-completely-peeled-insns
5607 The maximum number of insns of a completely peeled loop.
5609 @item max-completely-peel-times
5610 The maximum number of iterations of a loop to be suitable for complete peeling.
5612 @item max-unswitch-insns
5613 The maximum number of insns of an unswitched loop.
5615 @item max-unswitch-level
5616 The maximum number of branches unswitched in a single loop.
5619 The minimum cost of an expensive expression in the loop invariant motion.
5621 @item iv-consider-all-candidates-bound
5622 Bound on number of candidates for induction variables below that
5623 all candidates are considered for each use in induction variable
5624 optimizations. Only the most relevant candidates are considered
5625 if there are more candidates, to avoid quadratic time complexity.
5627 @item iv-max-considered-uses
5628 The induction variable optimizations give up on loops that contain more
5629 induction variable uses.
5631 @item iv-always-prune-cand-set-bound
5632 If number of candidates in the set is smaller than this value,
5633 we always try to remove unnecessary ivs from the set during its
5634 optimization when a new iv is added to the set.
5636 @item max-iterations-to-track
5638 The maximum number of iterations of a loop the brute force algorithm
5639 for analysis of # of iterations of the loop tries to evaluate.
5641 @item hot-bb-count-fraction
5642 Select fraction of the maximal count of repetitions of basic block in program
5643 given basic block needs to have to be considered hot.
5645 @item hot-bb-frequency-fraction
5646 Select fraction of the maximal frequency of executions of basic block in
5647 function given basic block needs to have to be considered hot
5649 @item tracer-dynamic-coverage
5650 @itemx tracer-dynamic-coverage-feedback
5652 This value is used to limit superblock formation once the given percentage of
5653 executed instructions is covered. This limits unnecessary code size
5656 The @option{tracer-dynamic-coverage-feedback} is used only when profile
5657 feedback is available. The real profiles (as opposed to statically estimated
5658 ones) are much less balanced allowing the threshold to be larger value.
5660 @item tracer-max-code-growth
5661 Stop tail duplication once code growth has reached given percentage. This is
5662 rather hokey argument, as most of the duplicates will be eliminated later in
5663 cross jumping, so it may be set to much higher values than is the desired code
5666 @item tracer-min-branch-ratio
5668 Stop reverse growth when the reverse probability of best edge is less than this
5669 threshold (in percent).
5671 @item tracer-min-branch-ratio
5672 @itemx tracer-min-branch-ratio-feedback
5674 Stop forward growth if the best edge do have probability lower than this
5677 Similarly to @option{tracer-dynamic-coverage} two values are present, one for
5678 compilation for profile feedback and one for compilation without. The value
5679 for compilation with profile feedback needs to be more conservative (higher) in
5680 order to make tracer effective.
5682 @item max-cse-path-length
5684 Maximum number of basic blocks on path that cse considers. The default is 10.
5686 @item global-var-threshold
5688 Counts the number of function calls (@var{n}) and the number of
5689 call-clobbered variables (@var{v}). If @var{n}x@var{v} is larger than this limit, a
5690 single artificial variable will be created to represent all the
5691 call-clobbered variables at function call sites. This artificial
5692 variable will then be made to alias every call-clobbered variable.
5693 (done as @code{int * size_t} on the host machine; beware overflow).
5695 @item max-aliased-vops
5697 Maximum number of virtual operands allowed to represent aliases
5698 before triggering the alias grouping heuristic. Alias grouping
5699 reduces compile times and memory consumption needed for aliasing at
5700 the expense of precision loss in alias information.
5702 @item ggc-min-expand
5704 GCC uses a garbage collector to manage its own memory allocation. This
5705 parameter specifies the minimum percentage by which the garbage
5706 collector's heap should be allowed to expand between collections.
5707 Tuning this may improve compilation speed; it has no effect on code
5710 The default is 30% + 70% * (RAM/1GB) with an upper bound of 100% when
5711 RAM >= 1GB@. If @code{getrlimit} is available, the notion of "RAM" is
5712 the smallest of actual RAM and @code{RLIMIT_DATA} or @code{RLIMIT_AS}. If
5713 GCC is not able to calculate RAM on a particular platform, the lower
5714 bound of 30% is used. Setting this parameter and
5715 @option{ggc-min-heapsize} to zero causes a full collection to occur at
5716 every opportunity. This is extremely slow, but can be useful for
5719 @item ggc-min-heapsize
5721 Minimum size of the garbage collector's heap before it begins bothering
5722 to collect garbage. The first collection occurs after the heap expands
5723 by @option{ggc-min-expand}% beyond @option{ggc-min-heapsize}. Again,
5724 tuning this may improve compilation speed, and has no effect on code
5727 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit which
5728 tries to ensure that RLIMIT_DATA or RLIMIT_AS are not exceeded, but
5729 with a lower bound of 4096 (four megabytes) and an upper bound of
5730 131072 (128 megabytes). If GCC is not able to calculate RAM on a
5731 particular platform, the lower bound is used. Setting this parameter
5732 very large effectively disables garbage collection. Setting this
5733 parameter and @option{ggc-min-expand} to zero causes a full collection
5734 to occur at every opportunity.
5736 @item max-reload-search-insns
5737 The maximum number of instruction reload should look backward for equivalent
5738 register. Increasing values mean more aggressive optimization, making the
5739 compile time increase with probably slightly better performance. The default
5742 @item max-cselib-memory-location
5743 The maximum number of memory locations cselib should take into acount.
5744 Increasing values mean more aggressive optimization, making the compile time
5745 increase with probably slightly better performance. The default value is 500.
5747 @item reorder-blocks-duplicate
5748 @itemx reorder-blocks-duplicate-feedback
5750 Used by basic block reordering pass to decide whether to use unconditional
5751 branch or duplicate the code on its destination. Code is duplicated when its
5752 estimated size is smaller than this value multiplied by the estimated size of
5753 unconditional jump in the hot spots of the program.
5755 The @option{reorder-block-duplicate-feedback} is used only when profile
5756 feedback is available and may be set to higher values than
5757 @option{reorder-block-duplicate} since information about the hot spots is more
5760 @item max-sched-region-blocks
5761 The maximum number of blocks in a region to be considered for
5762 interblock scheduling. The default value is 10.
5764 @item max-sched-region-insns
5765 The maximum number of insns in a region to be considered for
5766 interblock scheduling. The default value is 100.
5768 @item max-last-value-rtl
5770 The maximum size measured as number of RTLs that can be recorded in an expression
5771 in combiner for a pseudo register as last known value of that register. The default
5774 @item integer-share-limit
5775 Small integer constants can use a shared data structure, reducing the
5776 compiler's memory usage and increasing its speed. This sets the maximum
5777 value of a shared integer constant's. The default value is 256.
5782 @node Preprocessor Options
5783 @section Options Controlling the Preprocessor
5784 @cindex preprocessor options
5785 @cindex options, preprocessor
5787 These options control the C preprocessor, which is run on each C source
5788 file before actual compilation.
5790 If you use the @option{-E} option, nothing is done except preprocessing.
5791 Some of these options make sense only together with @option{-E} because
5792 they cause the preprocessor output to be unsuitable for actual
5797 You can use @option{-Wp,@var{option}} to bypass the compiler driver
5798 and pass @var{option} directly through to the preprocessor. If
5799 @var{option} contains commas, it is split into multiple options at the
5800 commas. However, many options are modified, translated or interpreted
5801 by the compiler driver before being passed to the preprocessor, and
5802 @option{-Wp} forcibly bypasses this phase. The preprocessor's direct
5803 interface is undocumented and subject to change, so whenever possible
5804 you should avoid using @option{-Wp} and let the driver handle the
5807 @item -Xpreprocessor @var{option}
5808 @opindex preprocessor
5809 Pass @var{option} as an option to the preprocessor. You can use this to
5810 supply system-specific preprocessor options which GCC does not know how to
5813 If you want to pass an option that takes an argument, you must use
5814 @option{-Xpreprocessor} twice, once for the option and once for the argument.
5817 @include cppopts.texi
5819 @node Assembler Options
5820 @section Passing Options to the Assembler
5822 @c prevent bad page break with this line
5823 You can pass options to the assembler.
5826 @item -Wa,@var{option}
5828 Pass @var{option} as an option to the assembler. If @var{option}
5829 contains commas, it is split into multiple options at the commas.
5831 @item -Xassembler @var{option}
5833 Pass @var{option} as an option to the assembler. You can use this to
5834 supply system-specific assembler options which GCC does not know how to
5837 If you want to pass an option that takes an argument, you must use
5838 @option{-Xassembler} twice, once for the option and once for the argument.
5843 @section Options for Linking
5844 @cindex link options
5845 @cindex options, linking
5847 These options come into play when the compiler links object files into
5848 an executable output file. They are meaningless if the compiler is
5849 not doing a link step.
5853 @item @var{object-file-name}
5854 A file name that does not end in a special recognized suffix is
5855 considered to name an object file or library. (Object files are
5856 distinguished from libraries by the linker according to the file
5857 contents.) If linking is done, these object files are used as input
5866 If any of these options is used, then the linker is not run, and
5867 object file names should not be used as arguments. @xref{Overall
5871 @item -l@var{library}
5872 @itemx -l @var{library}
5874 Search the library named @var{library} when linking. (The second
5875 alternative with the library as a separate argument is only for
5876 POSIX compliance and is not recommended.)
5878 It makes a difference where in the command you write this option; the
5879 linker searches and processes libraries and object files in the order they
5880 are specified. Thus, @samp{foo.o -lz bar.o} searches library @samp{z}
5881 after file @file{foo.o} but before @file{bar.o}. If @file{bar.o} refers
5882 to functions in @samp{z}, those functions may not be loaded.
5884 The linker searches a standard list of directories for the library,
5885 which is actually a file named @file{lib@var{library}.a}. The linker
5886 then uses this file as if it had been specified precisely by name.
5888 The directories searched include several standard system directories
5889 plus any that you specify with @option{-L}.
5891 Normally the files found this way are library files---archive files
5892 whose members are object files. The linker handles an archive file by
5893 scanning through it for members which define symbols that have so far
5894 been referenced but not defined. But if the file that is found is an
5895 ordinary object file, it is linked in the usual fashion. The only
5896 difference between using an @option{-l} option and specifying a file name
5897 is that @option{-l} surrounds @var{library} with @samp{lib} and @samp{.a}
5898 and searches several directories.
5902 You need this special case of the @option{-l} option in order to
5903 link an Objective-C or Objective-C++ program.
5906 @opindex nostartfiles
5907 Do not use the standard system startup files when linking.
5908 The standard system libraries are used normally, unless @option{-nostdlib}
5909 or @option{-nodefaultlibs} is used.
5911 @item -nodefaultlibs
5912 @opindex nodefaultlibs
5913 Do not use the standard system libraries when linking.
5914 Only the libraries you specify will be passed to the linker.
5915 The standard startup files are used normally, unless @option{-nostartfiles}
5916 is used. The compiler may generate calls to @code{memcmp},
5917 @code{memset}, @code{memcpy} and @code{memmove}.
5918 These entries are usually resolved by entries in
5919 libc. These entry points should be supplied through some other
5920 mechanism when this option is specified.
5924 Do not use the standard system startup files or libraries when linking.
5925 No startup files and only the libraries you specify will be passed to
5926 the linker. The compiler may generate calls to @code{memcmp}, @code{memset},
5927 @code{memcpy} and @code{memmove}.
5928 These entries are usually resolved by entries in
5929 libc. These entry points should be supplied through some other
5930 mechanism when this option is specified.
5932 @cindex @option{-lgcc}, use with @option{-nostdlib}
5933 @cindex @option{-nostdlib} and unresolved references
5934 @cindex unresolved references and @option{-nostdlib}
5935 @cindex @option{-lgcc}, use with @option{-nodefaultlibs}
5936 @cindex @option{-nodefaultlibs} and unresolved references
5937 @cindex unresolved references and @option{-nodefaultlibs}
5938 One of the standard libraries bypassed by @option{-nostdlib} and
5939 @option{-nodefaultlibs} is @file{libgcc.a}, a library of internal subroutines
5940 that GCC uses to overcome shortcomings of particular machines, or special
5941 needs for some languages.
5942 (@xref{Interface,,Interfacing to GCC Output,gccint,GNU Compiler
5943 Collection (GCC) Internals},
5944 for more discussion of @file{libgcc.a}.)
5945 In most cases, you need @file{libgcc.a} even when you want to avoid
5946 other standard libraries. In other words, when you specify @option{-nostdlib}
5947 or @option{-nodefaultlibs} you should usually specify @option{-lgcc} as well.
5948 This ensures that you have no unresolved references to internal GCC
5949 library subroutines. (For example, @samp{__main}, used to ensure C++
5950 constructors will be called; @pxref{Collect2,,@code{collect2}, gccint,
5951 GNU Compiler Collection (GCC) Internals}.)
5955 Produce a position independent executable on targets which support it.
5956 For predictable results, you must also specify the same set of options
5957 that were used to generate code (@option{-fpie}, @option{-fPIE},
5958 or model suboptions) when you specify this option.
5962 Remove all symbol table and relocation information from the executable.
5966 On systems that support dynamic linking, this prevents linking with the shared
5967 libraries. On other systems, this option has no effect.
5971 Produce a shared object which can then be linked with other objects to
5972 form an executable. Not all systems support this option. For predictable
5973 results, you must also specify the same set of options that were used to
5974 generate code (@option{-fpic}, @option{-fPIC}, or model suboptions)
5975 when you specify this option.@footnote{On some systems, @samp{gcc -shared}
5976 needs to build supplementary stub code for constructors to work. On
5977 multi-libbed systems, @samp{gcc -shared} must select the correct support
5978 libraries to link against. Failing to supply the correct flags may lead
5979 to subtle defects. Supplying them in cases where they are not necessary
5982 @item -shared-libgcc
5983 @itemx -static-libgcc
5984 @opindex shared-libgcc
5985 @opindex static-libgcc
5986 On systems that provide @file{libgcc} as a shared library, these options
5987 force the use of either the shared or static version respectively.
5988 If no shared version of @file{libgcc} was built when the compiler was
5989 configured, these options have no effect.
5991 There are several situations in which an application should use the
5992 shared @file{libgcc} instead of the static version. The most common
5993 of these is when the application wishes to throw and catch exceptions
5994 across different shared libraries. In that case, each of the libraries
5995 as well as the application itself should use the shared @file{libgcc}.
5997 Therefore, the G++ and GCJ drivers automatically add
5998 @option{-shared-libgcc} whenever you build a shared library or a main
5999 executable, because C++ and Java programs typically use exceptions, so
6000 this is the right thing to do.
6002 If, instead, you use the GCC driver to create shared libraries, you may
6003 find that they will not always be linked with the shared @file{libgcc}.
6004 If GCC finds, at its configuration time, that you have a non-GNU linker
6005 or a GNU linker that does not support option @option{--eh-frame-hdr},
6006 it will link the shared version of @file{libgcc} into shared libraries
6007 by default. Otherwise, it will take advantage of the linker and optimize
6008 away the linking with the shared version of @file{libgcc}, linking with
6009 the static version of libgcc by default. This allows exceptions to
6010 propagate through such shared libraries, without incurring relocation
6011 costs at library load time.
6013 However, if a library or main executable is supposed to throw or catch
6014 exceptions, you must link it using the G++ or GCJ driver, as appropriate
6015 for the languages used in the program, or using the option
6016 @option{-shared-libgcc}, such that it is linked with the shared
6021 Bind references to global symbols when building a shared object. Warn
6022 about any unresolved references (unless overridden by the link editor
6023 option @samp{-Xlinker -z -Xlinker defs}). Only a few systems support
6026 @item -Xlinker @var{option}
6028 Pass @var{option} as an option to the linker. You can use this to
6029 supply system-specific linker options which GCC does not know how to
6032 If you want to pass an option that takes an argument, you must use
6033 @option{-Xlinker} twice, once for the option and once for the argument.
6034 For example, to pass @option{-assert definitions}, you must write
6035 @samp{-Xlinker -assert -Xlinker definitions}. It does not work to write
6036 @option{-Xlinker "-assert definitions"}, because this passes the entire
6037 string as a single argument, which is not what the linker expects.
6039 @item -Wl,@var{option}
6041 Pass @var{option} as an option to the linker. If @var{option} contains
6042 commas, it is split into multiple options at the commas.
6044 @item -u @var{symbol}
6046 Pretend the symbol @var{symbol} is undefined, to force linking of
6047 library modules to define it. You can use @option{-u} multiple times with
6048 different symbols to force loading of additional library modules.
6051 @node Directory Options
6052 @section Options for Directory Search
6053 @cindex directory options
6054 @cindex options, directory search
6057 These options specify directories to search for header files, for
6058 libraries and for parts of the compiler:
6063 Add the directory @var{dir} to the head of the list of directories to be
6064 searched for header files. This can be used to override a system header
6065 file, substituting your own version, since these directories are
6066 searched before the system header file directories. However, you should
6067 not use this option to add directories that contain vendor-supplied
6068 system header files (use @option{-isystem} for that). If you use more than
6069 one @option{-I} option, the directories are scanned in left-to-right
6070 order; the standard system directories come after.
6072 If a standard system include directory, or a directory specified with
6073 @option{-isystem}, is also specified with @option{-I}, the @option{-I}
6074 option will be ignored. The directory will still be searched but as a
6075 system directory at its normal position in the system include chain.
6076 This is to ensure that GCC's procedure to fix buggy system headers and
6077 the ordering for the include_next directive are not inadvertently changed.
6078 If you really need to change the search order for system directories,
6079 use the @option{-nostdinc} and/or @option{-isystem} options.
6081 @item -iquote@var{dir}
6083 Add the directory @var{dir} to the head of the list of directories to
6084 be searched for header files only for the case of @samp{#include
6085 "@var{file}"}; they are not searched for @samp{#include <@var{file}>},
6086 otherwise just like @option{-I}.
6090 Add directory @var{dir} to the list of directories to be searched
6093 @item -B@var{prefix}
6095 This option specifies where to find the executables, libraries,
6096 include files, and data files of the compiler itself.
6098 The compiler driver program runs one or more of the subprograms
6099 @file{cpp}, @file{cc1}, @file{as} and @file{ld}. It tries
6100 @var{prefix} as a prefix for each program it tries to run, both with and
6101 without @samp{@var{machine}/@var{version}/} (@pxref{Target Options}).
6103 For each subprogram to be run, the compiler driver first tries the
6104 @option{-B} prefix, if any. If that name is not found, or if @option{-B}
6105 was not specified, the driver tries two standard prefixes, which are
6106 @file{/usr/lib/gcc/} and @file{/usr/local/lib/gcc/}. If neither of
6107 those results in a file name that is found, the unmodified program
6108 name is searched for using the directories specified in your
6109 @env{PATH} environment variable.
6111 The compiler will check to see if the path provided by the @option{-B}
6112 refers to a directory, and if necessary it will add a directory
6113 separator character at the end of the path.
6115 @option{-B} prefixes that effectively specify directory names also apply
6116 to libraries in the linker, because the compiler translates these
6117 options into @option{-L} options for the linker. They also apply to
6118 includes files in the preprocessor, because the compiler translates these
6119 options into @option{-isystem} options for the preprocessor. In this case,
6120 the compiler appends @samp{include} to the prefix.
6122 The run-time support file @file{libgcc.a} can also be searched for using
6123 the @option{-B} prefix, if needed. If it is not found there, the two
6124 standard prefixes above are tried, and that is all. The file is left
6125 out of the link if it is not found by those means.
6127 Another way to specify a prefix much like the @option{-B} prefix is to use
6128 the environment variable @env{GCC_EXEC_PREFIX}. @xref{Environment
6131 As a special kludge, if the path provided by @option{-B} is
6132 @file{[dir/]stage@var{N}/}, where @var{N} is a number in the range 0 to
6133 9, then it will be replaced by @file{[dir/]include}. This is to help
6134 with boot-strapping the compiler.
6136 @item -specs=@var{file}
6138 Process @var{file} after the compiler reads in the standard @file{specs}
6139 file, in order to override the defaults that the @file{gcc} driver
6140 program uses when determining what switches to pass to @file{cc1},
6141 @file{cc1plus}, @file{as}, @file{ld}, etc. More than one
6142 @option{-specs=@var{file}} can be specified on the command line, and they
6143 are processed in order, from left to right.
6147 This option has been deprecated. Please use @option{-iquote} instead for
6148 @option{-I} directories before the @option{-I-} and remove the @option{-I-}.
6149 Any directories you specify with @option{-I} options before the @option{-I-}
6150 option are searched only for the case of @samp{#include "@var{file}"};
6151 they are not searched for @samp{#include <@var{file}>}.
6153 If additional directories are specified with @option{-I} options after
6154 the @option{-I-}, these directories are searched for all @samp{#include}
6155 directives. (Ordinarily @emph{all} @option{-I} directories are used
6158 In addition, the @option{-I-} option inhibits the use of the current
6159 directory (where the current input file came from) as the first search
6160 directory for @samp{#include "@var{file}"}. There is no way to
6161 override this effect of @option{-I-}. With @option{-I.} you can specify
6162 searching the directory which was current when the compiler was
6163 invoked. That is not exactly the same as what the preprocessor does
6164 by default, but it is often satisfactory.
6166 @option{-I-} does not inhibit the use of the standard system directories
6167 for header files. Thus, @option{-I-} and @option{-nostdinc} are
6174 @section Specifying subprocesses and the switches to pass to them
6177 @command{gcc} is a driver program. It performs its job by invoking a
6178 sequence of other programs to do the work of compiling, assembling and
6179 linking. GCC interprets its command-line parameters and uses these to
6180 deduce which programs it should invoke, and which command-line options
6181 it ought to place on their command lines. This behavior is controlled
6182 by @dfn{spec strings}. In most cases there is one spec string for each
6183 program that GCC can invoke, but a few programs have multiple spec
6184 strings to control their behavior. The spec strings built into GCC can
6185 be overridden by using the @option{-specs=} command-line switch to specify
6188 @dfn{Spec files} are plaintext files that are used to construct spec
6189 strings. They consist of a sequence of directives separated by blank
6190 lines. The type of directive is determined by the first non-whitespace
6191 character on the line and it can be one of the following:
6194 @item %@var{command}
6195 Issues a @var{command} to the spec file processor. The commands that can
6199 @item %include <@var{file}>
6201 Search for @var{file} and insert its text at the current point in the
6204 @item %include_noerr <@var{file}>
6205 @cindex %include_noerr
6206 Just like @samp{%include}, but do not generate an error message if the include
6207 file cannot be found.
6209 @item %rename @var{old_name} @var{new_name}
6211 Rename the spec string @var{old_name} to @var{new_name}.
6215 @item *[@var{spec_name}]:
6216 This tells the compiler to create, override or delete the named spec
6217 string. All lines after this directive up to the next directive or
6218 blank line are considered to be the text for the spec string. If this
6219 results in an empty string then the spec will be deleted. (Or, if the
6220 spec did not exist, then nothing will happened.) Otherwise, if the spec
6221 does not currently exist a new spec will be created. If the spec does
6222 exist then its contents will be overridden by the text of this
6223 directive, unless the first character of that text is the @samp{+}
6224 character, in which case the text will be appended to the spec.
6226 @item [@var{suffix}]:
6227 Creates a new @samp{[@var{suffix}] spec} pair. All lines after this directive
6228 and up to the next directive or blank line are considered to make up the
6229 spec string for the indicated suffix. When the compiler encounters an
6230 input file with the named suffix, it will processes the spec string in
6231 order to work out how to compile that file. For example:
6238 This says that any input file whose name ends in @samp{.ZZ} should be
6239 passed to the program @samp{z-compile}, which should be invoked with the
6240 command-line switch @option{-input} and with the result of performing the
6241 @samp{%i} substitution. (See below.)
6243 As an alternative to providing a spec string, the text that follows a
6244 suffix directive can be one of the following:
6247 @item @@@var{language}
6248 This says that the suffix is an alias for a known @var{language}. This is
6249 similar to using the @option{-x} command-line switch to GCC to specify a
6250 language explicitly. For example:
6257 Says that .ZZ files are, in fact, C++ source files.
6260 This causes an error messages saying:
6263 @var{name} compiler not installed on this system.
6267 GCC already has an extensive list of suffixes built into it.
6268 This directive will add an entry to the end of the list of suffixes, but
6269 since the list is searched from the end backwards, it is effectively
6270 possible to override earlier entries using this technique.
6274 GCC has the following spec strings built into it. Spec files can
6275 override these strings or create their own. Note that individual
6276 targets can also add their own spec strings to this list.
6279 asm Options to pass to the assembler
6280 asm_final Options to pass to the assembler post-processor
6281 cpp Options to pass to the C preprocessor
6282 cc1 Options to pass to the C compiler
6283 cc1plus Options to pass to the C++ compiler
6284 endfile Object files to include at the end of the link
6285 link Options to pass to the linker
6286 lib Libraries to include on the command line to the linker
6287 libgcc Decides which GCC support library to pass to the linker
6288 linker Sets the name of the linker
6289 predefines Defines to be passed to the C preprocessor
6290 signed_char Defines to pass to CPP to say whether @code{char} is signed
6292 startfile Object files to include at the start of the link
6295 Here is a small example of a spec file:
6301 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6304 This example renames the spec called @samp{lib} to @samp{old_lib} and
6305 then overrides the previous definition of @samp{lib} with a new one.
6306 The new definition adds in some extra command-line options before
6307 including the text of the old definition.
6309 @dfn{Spec strings} are a list of command-line options to be passed to their
6310 corresponding program. In addition, the spec strings can contain
6311 @samp{%}-prefixed sequences to substitute variable text or to
6312 conditionally insert text into the command line. Using these constructs
6313 it is possible to generate quite complex command lines.
6315 Here is a table of all defined @samp{%}-sequences for spec
6316 strings. Note that spaces are not generated automatically around the
6317 results of expanding these sequences. Therefore you can concatenate them
6318 together or combine them with constant text in a single argument.
6322 Substitute one @samp{%} into the program name or argument.
6325 Substitute the name of the input file being processed.
6328 Substitute the basename of the input file being processed.
6329 This is the substring up to (and not including) the last period
6330 and not including the directory.
6333 This is the same as @samp{%b}, but include the file suffix (text after
6337 Marks the argument containing or following the @samp{%d} as a
6338 temporary file name, so that that file will be deleted if GCC exits
6339 successfully. Unlike @samp{%g}, this contributes no text to the
6342 @item %g@var{suffix}
6343 Substitute a file name that has suffix @var{suffix} and is chosen
6344 once per compilation, and mark the argument in the same way as
6345 @samp{%d}. To reduce exposure to denial-of-service attacks, the file
6346 name is now chosen in a way that is hard to predict even when previously
6347 chosen file names are known. For example, @samp{%g.s @dots{} %g.o @dots{} %g.s}
6348 might turn into @samp{ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s}. @var{suffix} matches
6349 the regexp @samp{[.A-Za-z]*} or the special string @samp{%O}, which is
6350 treated exactly as if @samp{%O} had been preprocessed. Previously, @samp{%g}
6351 was simply substituted with a file name chosen once per compilation,
6352 without regard to any appended suffix (which was therefore treated
6353 just like ordinary text), making such attacks more likely to succeed.
6355 @item %u@var{suffix}
6356 Like @samp{%g}, but generates a new temporary file name even if
6357 @samp{%u@var{suffix}} was already seen.
6359 @item %U@var{suffix}
6360 Substitutes the last file name generated with @samp{%u@var{suffix}}, generating a
6361 new one if there is no such last file name. In the absence of any
6362 @samp{%u@var{suffix}}, this is just like @samp{%g@var{suffix}}, except they don't share
6363 the same suffix @emph{space}, so @samp{%g.s @dots{} %U.s @dots{} %g.s @dots{} %U.s}
6364 would involve the generation of two distinct file names, one
6365 for each @samp{%g.s} and another for each @samp{%U.s}. Previously, @samp{%U} was
6366 simply substituted with a file name chosen for the previous @samp{%u},
6367 without regard to any appended suffix.
6369 @item %j@var{suffix}
6370 Substitutes the name of the @code{HOST_BIT_BUCKET}, if any, and if it is
6371 writable, and if save-temps is off; otherwise, substitute the name
6372 of a temporary file, just like @samp{%u}. This temporary file is not
6373 meant for communication between processes, but rather as a junk
6376 @item %|@var{suffix}
6377 @itemx %m@var{suffix}
6378 Like @samp{%g}, except if @option{-pipe} is in effect. In that case
6379 @samp{%|} substitutes a single dash and @samp{%m} substitutes nothing at
6380 all. These are the two most common ways to instruct a program that it
6381 should read from standard input or write to standard output. If you
6382 need something more elaborate you can use an @samp{%@{pipe:@code{X}@}}
6383 construct: see for example @file{f/lang-specs.h}.
6385 @item %.@var{SUFFIX}
6386 Substitutes @var{.SUFFIX} for the suffixes of a matched switch's args
6387 when it is subsequently output with @samp{%*}. @var{SUFFIX} is
6388 terminated by the next space or %.
6391 Marks the argument containing or following the @samp{%w} as the
6392 designated output file of this compilation. This puts the argument
6393 into the sequence of arguments that @samp{%o} will substitute later.
6396 Substitutes the names of all the output files, with spaces
6397 automatically placed around them. You should write spaces
6398 around the @samp{%o} as well or the results are undefined.
6399 @samp{%o} is for use in the specs for running the linker.
6400 Input files whose names have no recognized suffix are not compiled
6401 at all, but they are included among the output files, so they will
6405 Substitutes the suffix for object files. Note that this is
6406 handled specially when it immediately follows @samp{%g, %u, or %U},
6407 because of the need for those to form complete file names. The
6408 handling is such that @samp{%O} is treated exactly as if it had already
6409 been substituted, except that @samp{%g, %u, and %U} do not currently
6410 support additional @var{suffix} characters following @samp{%O} as they would
6411 following, for example, @samp{.o}.
6414 Substitutes the standard macro predefinitions for the
6415 current target machine. Use this when running @code{cpp}.
6418 Like @samp{%p}, but puts @samp{__} before and after the name of each
6419 predefined macro, except for macros that start with @samp{__} or with
6420 @samp{_@var{L}}, where @var{L} is an uppercase letter. This is for ISO
6424 Substitute any of @option{-iprefix} (made from @env{GCC_EXEC_PREFIX}),
6425 @option{-isysroot} (made from @env{TARGET_SYSTEM_ROOT}), and
6426 @option{-isystem} (made from @env{COMPILER_PATH} and @option{-B} options)
6430 Current argument is the name of a library or startup file of some sort.
6431 Search for that file in a standard list of directories and substitute
6432 the full name found.
6435 Print @var{str} as an error message. @var{str} is terminated by a newline.
6436 Use this when inconsistent options are detected.
6439 Substitute the contents of spec string @var{name} at this point.
6442 Like @samp{%(@dots{})} but put @samp{__} around @option{-D} arguments.
6444 @item %x@{@var{option}@}
6445 Accumulate an option for @samp{%X}.
6448 Output the accumulated linker options specified by @option{-Wl} or a @samp{%x}
6452 Output the accumulated assembler options specified by @option{-Wa}.
6455 Output the accumulated preprocessor options specified by @option{-Wp}.
6458 Process the @code{asm} spec. This is used to compute the
6459 switches to be passed to the assembler.
6462 Process the @code{asm_final} spec. This is a spec string for
6463 passing switches to an assembler post-processor, if such a program is
6467 Process the @code{link} spec. This is the spec for computing the
6468 command line passed to the linker. Typically it will make use of the
6469 @samp{%L %G %S %D and %E} sequences.
6472 Dump out a @option{-L} option for each directory that GCC believes might
6473 contain startup files. If the target supports multilibs then the
6474 current multilib directory will be prepended to each of these paths.
6477 Process the @code{lib} spec. This is a spec string for deciding which
6478 libraries should be included on the command line to the linker.
6481 Process the @code{libgcc} spec. This is a spec string for deciding
6482 which GCC support library should be included on the command line to the linker.
6485 Process the @code{startfile} spec. This is a spec for deciding which
6486 object files should be the first ones passed to the linker. Typically
6487 this might be a file named @file{crt0.o}.
6490 Process the @code{endfile} spec. This is a spec string that specifies
6491 the last object files that will be passed to the linker.
6494 Process the @code{cpp} spec. This is used to construct the arguments
6495 to be passed to the C preprocessor.
6498 Process the @code{cc1} spec. This is used to construct the options to be
6499 passed to the actual C compiler (@samp{cc1}).
6502 Process the @code{cc1plus} spec. This is used to construct the options to be
6503 passed to the actual C++ compiler (@samp{cc1plus}).
6506 Substitute the variable part of a matched option. See below.
6507 Note that each comma in the substituted string is replaced by
6511 Remove all occurrences of @code{-S} from the command line. Note---this
6512 command is position dependent. @samp{%} commands in the spec string
6513 before this one will see @code{-S}, @samp{%} commands in the spec string
6514 after this one will not.
6516 @item %:@var{function}(@var{args})
6517 Call the named function @var{function}, passing it @var{args}.
6518 @var{args} is first processed as a nested spec string, then split
6519 into an argument vector in the usual fashion. The function returns
6520 a string which is processed as if it had appeared literally as part
6521 of the current spec.
6523 The following built-in spec functions are provided:
6526 @item @code{if-exists}
6527 The @code{if-exists} spec function takes one argument, an absolute
6528 pathname to a file. If the file exists, @code{if-exists} returns the
6529 pathname. Here is a small example of its usage:
6533 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6536 @item @code{if-exists-else}
6537 The @code{if-exists-else} spec function is similar to the @code{if-exists}
6538 spec function, except that it takes two arguments. The first argument is
6539 an absolute pathname to a file. If the file exists, @code{if-exists-else}
6540 returns the pathname. If it does not exist, it returns the second argument.
6541 This way, @code{if-exists-else} can be used to select one file or another,
6542 based on the existence of the first. Here is a small example of its usage:
6546 crt0%O%s %:if-exists(crti%O%s) \
6547 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6550 @item @code{replace-outfile}
6551 The @code{replace-outfile} spec function takes two arguments. It looks for the
6552 first argument in the outfiles array and replaces it with the second argument. Here
6553 is a small example of its usage:
6556 %@{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)@}
6562 Substitutes the @code{-S} switch, if that switch was given to GCC@.
6563 If that switch was not specified, this substitutes nothing. Note that
6564 the leading dash is omitted when specifying this option, and it is
6565 automatically inserted if the substitution is performed. Thus the spec
6566 string @samp{%@{foo@}} would match the command-line option @option{-foo}
6567 and would output the command line option @option{-foo}.
6569 @item %W@{@code{S}@}
6570 Like %@{@code{S}@} but mark last argument supplied within as a file to be
6573 @item %@{@code{S}*@}
6574 Substitutes all the switches specified to GCC whose names start
6575 with @code{-S}, but which also take an argument. This is used for
6576 switches like @option{-o}, @option{-D}, @option{-I}, etc.
6577 GCC considers @option{-o foo} as being
6578 one switch whose names starts with @samp{o}. %@{o*@} would substitute this
6579 text, including the space. Thus two arguments would be generated.
6581 @item %@{@code{S}*&@code{T}*@}
6582 Like %@{@code{S}*@}, but preserve order of @code{S} and @code{T} options
6583 (the order of @code{S} and @code{T} in the spec is not significant).
6584 There can be any number of ampersand-separated variables; for each the
6585 wild card is optional. Useful for CPP as @samp{%@{D*&U*&A*@}}.
6587 @item %@{@code{S}:@code{X}@}
6588 Substitutes @code{X}, if the @samp{-S} switch was given to GCC@.
6590 @item %@{!@code{S}:@code{X}@}
6591 Substitutes @code{X}, if the @samp{-S} switch was @emph{not} given to GCC@.
6593 @item %@{@code{S}*:@code{X}@}
6594 Substitutes @code{X} if one or more switches whose names start with
6595 @code{-S} are specified to GCC@. Normally @code{X} is substituted only
6596 once, no matter how many such switches appeared. However, if @code{%*}
6597 appears somewhere in @code{X}, then @code{X} will be substituted once
6598 for each matching switch, with the @code{%*} replaced by the part of
6599 that switch that matched the @code{*}.
6601 @item %@{.@code{S}:@code{X}@}
6602 Substitutes @code{X}, if processing a file with suffix @code{S}.
6604 @item %@{!.@code{S}:@code{X}@}
6605 Substitutes @code{X}, if @emph{not} processing a file with suffix @code{S}.
6607 @item %@{@code{S}|@code{P}:@code{X}@}
6608 Substitutes @code{X} if either @code{-S} or @code{-P} was given to GCC@.
6609 This may be combined with @samp{!}, @samp{.}, and @code{*} sequences as well,
6610 although they have a stronger binding than the @samp{|}. If @code{%*}
6611 appears in @code{X}, all of the alternatives must be starred, and only
6612 the first matching alternative is substituted.
6614 For example, a spec string like this:
6617 %@{.c:-foo@} %@{!.c:-bar@} %@{.c|d:-baz@} %@{!.c|d:-boggle@}
6620 will output the following command-line options from the following input
6621 command-line options:
6626 -d fred.c -foo -baz -boggle
6627 -d jim.d -bar -baz -boggle
6630 @item %@{S:X; T:Y; :D@}
6632 If @code{S} was given to GCC, substitutes @code{X}; else if @code{T} was
6633 given to GCC, substitutes @code{Y}; else substitutes @code{D}. There can
6634 be as many clauses as you need. This may be combined with @code{.},
6635 @code{!}, @code{|}, and @code{*} as needed.
6640 The conditional text @code{X} in a %@{@code{S}:@code{X}@} or similar
6641 construct may contain other nested @samp{%} constructs or spaces, or
6642 even newlines. They are processed as usual, as described above.
6643 Trailing white space in @code{X} is ignored. White space may also
6644 appear anywhere on the left side of the colon in these constructs,
6645 except between @code{.} or @code{*} and the corresponding word.
6647 The @option{-O}, @option{-f}, @option{-m}, and @option{-W} switches are
6648 handled specifically in these constructs. If another value of
6649 @option{-O} or the negated form of a @option{-f}, @option{-m}, or
6650 @option{-W} switch is found later in the command line, the earlier
6651 switch value is ignored, except with @{@code{S}*@} where @code{S} is
6652 just one letter, which passes all matching options.
6654 The character @samp{|} at the beginning of the predicate text is used to
6655 indicate that a command should be piped to the following command, but
6656 only if @option{-pipe} is specified.
6658 It is built into GCC which switches take arguments and which do not.
6659 (You might think it would be useful to generalize this to allow each
6660 compiler's spec to say which switches take arguments. But this cannot
6661 be done in a consistent fashion. GCC cannot even decide which input
6662 files have been specified without knowing which switches take arguments,
6663 and it must know which input files to compile in order to tell which
6666 GCC also knows implicitly that arguments starting in @option{-l} are to be
6667 treated as compiler output files, and passed to the linker in their
6668 proper position among the other output files.
6670 @c man begin OPTIONS
6672 @node Target Options
6673 @section Specifying Target Machine and Compiler Version
6674 @cindex target options
6675 @cindex cross compiling
6676 @cindex specifying machine version
6677 @cindex specifying compiler version and target machine
6678 @cindex compiler version, specifying
6679 @cindex target machine, specifying
6681 The usual way to run GCC is to run the executable called @file{gcc}, or
6682 @file{<machine>-gcc} when cross-compiling, or
6683 @file{<machine>-gcc-<version>} to run a version other than the one that
6684 was installed last. Sometimes this is inconvenient, so GCC provides
6685 options that will switch to another cross-compiler or version.
6688 @item -b @var{machine}
6690 The argument @var{machine} specifies the target machine for compilation.
6692 The value to use for @var{machine} is the same as was specified as the
6693 machine type when configuring GCC as a cross-compiler. For
6694 example, if a cross-compiler was configured with @samp{configure
6695 i386v}, meaning to compile for an 80386 running System V, then you
6696 would specify @option{-b i386v} to run that cross compiler.
6698 @item -V @var{version}
6700 The argument @var{version} specifies which version of GCC to run.
6701 This is useful when multiple versions are installed. For example,
6702 @var{version} might be @samp{2.0}, meaning to run GCC version 2.0.
6705 The @option{-V} and @option{-b} options work by running the
6706 @file{<machine>-gcc-<version>} executable, so there's no real reason to
6707 use them if you can just run that directly.
6709 @node Submodel Options
6710 @section Hardware Models and Configurations
6711 @cindex submodel options
6712 @cindex specifying hardware config
6713 @cindex hardware models and configurations, specifying
6714 @cindex machine dependent options
6716 Earlier we discussed the standard option @option{-b} which chooses among
6717 different installed compilers for completely different target
6718 machines, such as VAX vs.@: 68000 vs.@: 80386.
6720 In addition, each of these target machine types can have its own
6721 special options, starting with @samp{-m}, to choose among various
6722 hardware models or configurations---for example, 68010 vs 68020,
6723 floating coprocessor or none. A single installed version of the
6724 compiler can compile for any model or configuration, according to the
6727 Some configurations of the compiler also support additional special
6728 options, usually for compatibility with other compilers on the same
6731 These options are defined by the macro @code{TARGET_SWITCHES} in the
6732 machine description. The default for the options is also defined by
6733 that macro, which enables you to change the defaults.
6735 @c This list is ordered alphanumerically by subsection name.
6736 @c It should be the same order and spelling as these options are listed
6737 @c in Machine Dependent Options
6743 * Blackfin Options::
6746 * DEC Alpha Options::
6747 * DEC Alpha/VMS Options::
6751 * i386 and x86-64 Options::
6763 * RS/6000 and PowerPC Options::
6764 * S/390 and zSeries Options::
6767 * System V Options::
6768 * TMS320C3x/C4x Options::
6772 * Xstormy16 Options::
6778 @subsection ARC Options
6781 These options are defined for ARC implementations:
6786 Compile code for little endian mode. This is the default.
6790 Compile code for big endian mode.
6793 @opindex mmangle-cpu
6794 Prepend the name of the cpu to all public symbol names.
6795 In multiple-processor systems, there are many ARC variants with different
6796 instruction and register set characteristics. This flag prevents code
6797 compiled for one cpu to be linked with code compiled for another.
6798 No facility exists for handling variants that are ``almost identical''.
6799 This is an all or nothing option.
6801 @item -mcpu=@var{cpu}
6803 Compile code for ARC variant @var{cpu}.
6804 Which variants are supported depend on the configuration.
6805 All variants support @option{-mcpu=base}, this is the default.
6807 @item -mtext=@var{text-section}
6808 @itemx -mdata=@var{data-section}
6809 @itemx -mrodata=@var{readonly-data-section}
6813 Put functions, data, and readonly data in @var{text-section},
6814 @var{data-section}, and @var{readonly-data-section} respectively
6815 by default. This can be overridden with the @code{section} attribute.
6816 @xref{Variable Attributes}.
6821 @subsection ARM Options
6824 These @samp{-m} options are defined for Advanced RISC Machines (ARM)
6828 @item -mabi=@var{name}
6830 Generate code for the specified ABI@. Permissible values are: @samp{apcs-gnu},
6831 @samp{atpcs}, @samp{aapcs} and @samp{iwmmxt}.
6834 @opindex mapcs-frame
6835 Generate a stack frame that is compliant with the ARM Procedure Call
6836 Standard for all functions, even if this is not strictly necessary for
6837 correct execution of the code. Specifying @option{-fomit-frame-pointer}
6838 with this option will cause the stack frames not to be generated for
6839 leaf functions. The default is @option{-mno-apcs-frame}.
6843 This is a synonym for @option{-mapcs-frame}.
6846 @c not currently implemented
6847 @item -mapcs-stack-check
6848 @opindex mapcs-stack-check
6849 Generate code to check the amount of stack space available upon entry to
6850 every function (that actually uses some stack space). If there is
6851 insufficient space available then either the function
6852 @samp{__rt_stkovf_split_small} or @samp{__rt_stkovf_split_big} will be
6853 called, depending upon the amount of stack space required. The run time
6854 system is required to provide these functions. The default is
6855 @option{-mno-apcs-stack-check}, since this produces smaller code.
6857 @c not currently implemented
6859 @opindex mapcs-float
6860 Pass floating point arguments using the float point registers. This is
6861 one of the variants of the APCS@. This option is recommended if the
6862 target hardware has a floating point unit or if a lot of floating point
6863 arithmetic is going to be performed by the code. The default is
6864 @option{-mno-apcs-float}, since integer only code is slightly increased in
6865 size if @option{-mapcs-float} is used.
6867 @c not currently implemented
6868 @item -mapcs-reentrant
6869 @opindex mapcs-reentrant
6870 Generate reentrant, position independent code. The default is
6871 @option{-mno-apcs-reentrant}.
6874 @item -mthumb-interwork
6875 @opindex mthumb-interwork
6876 Generate code which supports calling between the ARM and Thumb
6877 instruction sets. Without this option the two instruction sets cannot
6878 be reliably used inside one program. The default is
6879 @option{-mno-thumb-interwork}, since slightly larger code is generated
6880 when @option{-mthumb-interwork} is specified.
6882 @item -mno-sched-prolog
6883 @opindex mno-sched-prolog
6884 Prevent the reordering of instructions in the function prolog, or the
6885 merging of those instruction with the instructions in the function's
6886 body. This means that all functions will start with a recognizable set
6887 of instructions (or in fact one of a choice from a small set of
6888 different function prologues), and this information can be used to
6889 locate the start if functions inside an executable piece of code. The
6890 default is @option{-msched-prolog}.
6893 @opindex mhard-float
6894 Generate output containing floating point instructions. This is the
6898 @opindex msoft-float
6899 Generate output containing library calls for floating point.
6900 @strong{Warning:} the requisite libraries are not available for all ARM
6901 targets. Normally the facilities of the machine's usual C compiler are
6902 used, but this cannot be done directly in cross-compilation. You must make
6903 your own arrangements to provide suitable library functions for
6906 @option{-msoft-float} changes the calling convention in the output file;
6907 therefore, it is only useful if you compile @emph{all} of a program with
6908 this option. In particular, you need to compile @file{libgcc.a}, the
6909 library that comes with GCC, with @option{-msoft-float} in order for
6912 @item -mfloat-abi=@var{name}
6914 Specifies which ABI to use for floating point values. Permissible values
6915 are: @samp{soft}, @samp{softfp} and @samp{hard}.
6917 @samp{soft} and @samp{hard} are equivalent to @option{-msoft-float}
6918 and @option{-mhard-float} respectively. @samp{softfp} allows the generation
6919 of floating point instructions, but still uses the soft-float calling
6922 @item -mlittle-endian
6923 @opindex mlittle-endian
6924 Generate code for a processor running in little-endian mode. This is
6925 the default for all standard configurations.
6928 @opindex mbig-endian
6929 Generate code for a processor running in big-endian mode; the default is
6930 to compile code for a little-endian processor.
6932 @item -mwords-little-endian
6933 @opindex mwords-little-endian
6934 This option only applies when generating code for big-endian processors.
6935 Generate code for a little-endian word order but a big-endian byte
6936 order. That is, a byte order of the form @samp{32107654}. Note: this
6937 option should only be used if you require compatibility with code for
6938 big-endian ARM processors generated by versions of the compiler prior to
6941 @item -mcpu=@var{name}
6943 This specifies the name of the target ARM processor. GCC uses this name
6944 to determine what kind of instructions it can emit when generating
6945 assembly code. Permissible names are: @samp{arm2}, @samp{arm250},
6946 @samp{arm3}, @samp{arm6}, @samp{arm60}, @samp{arm600}, @samp{arm610},
6947 @samp{arm620}, @samp{arm7}, @samp{arm7m}, @samp{arm7d}, @samp{arm7dm},
6948 @samp{arm7di}, @samp{arm7dmi}, @samp{arm70}, @samp{arm700},
6949 @samp{arm700i}, @samp{arm710}, @samp{arm710c}, @samp{arm7100},
6950 @samp{arm7500}, @samp{arm7500fe}, @samp{arm7tdmi}, @samp{arm7tdmi-s},
6951 @samp{arm8}, @samp{strongarm}, @samp{strongarm110}, @samp{strongarm1100},
6952 @samp{arm8}, @samp{arm810}, @samp{arm9}, @samp{arm9e}, @samp{arm920},
6953 @samp{arm920t}, @samp{arm922t}, @samp{arm946e-s}, @samp{arm966e-s},
6954 @samp{arm968e-s}, @samp{arm926ej-s}, @samp{arm940t}, @samp{arm9tdmi},
6955 @samp{arm10tdmi}, @samp{arm1020t}, @samp{arm1026ej-s},
6956 @samp{arm10e}, @samp{arm1020e}, @samp{arm1022e},
6957 @samp{arm1136j-s}, @samp{arm1136jf-s}, @samp{mpcore}, @samp{mpcorenovfp},
6958 @samp{arm1176jz-s}, @samp{arm1176jzf-s}, @samp{xscale}, @samp{iwmmxt},
6961 @itemx -mtune=@var{name}
6963 This option is very similar to the @option{-mcpu=} option, except that
6964 instead of specifying the actual target processor type, and hence
6965 restricting which instructions can be used, it specifies that GCC should
6966 tune the performance of the code as if the target were of the type
6967 specified in this option, but still choosing the instructions that it
6968 will generate based on the cpu specified by a @option{-mcpu=} option.
6969 For some ARM implementations better performance can be obtained by using
6972 @item -march=@var{name}
6974 This specifies the name of the target ARM architecture. GCC uses this
6975 name to determine what kind of instructions it can emit when generating
6976 assembly code. This option can be used in conjunction with or instead
6977 of the @option{-mcpu=} option. Permissible names are: @samp{armv2},
6978 @samp{armv2a}, @samp{armv3}, @samp{armv3m}, @samp{armv4}, @samp{armv4t},
6979 @samp{armv5}, @samp{armv5t}, @samp{armv5te}, @samp{armv6}, @samp{armv6j},
6980 @samp{iwmmxt}, @samp{ep9312}.
6982 @item -mfpu=@var{name}
6983 @itemx -mfpe=@var{number}
6984 @itemx -mfp=@var{number}
6988 This specifies what floating point hardware (or hardware emulation) is
6989 available on the target. Permissible names are: @samp{fpa}, @samp{fpe2},
6990 @samp{fpe3}, @samp{maverick}, @samp{vfp}. @option{-mfp} and @option{-mfpe}
6991 are synonyms for @option{-mfpu}=@samp{fpe}@var{number}, for compatibility
6992 with older versions of GCC@.
6994 If @option{-msoft-float} is specified this specifies the format of
6995 floating point values.
6997 @item -mstructure-size-boundary=@var{n}
6998 @opindex mstructure-size-boundary
6999 The size of all structures and unions will be rounded up to a multiple
7000 of the number of bits set by this option. Permissible values are 8, 32
7001 and 64. The default value varies for different toolchains. For the COFF
7002 targeted toolchain the default value is 8. A value of 64 is only allowed
7003 if the underlying ABI supports it.
7005 Specifying the larger number can produce faster, more efficient code, but
7006 can also increase the size of the program. Different values are potentially
7007 incompatible. Code compiled with one value cannot necessarily expect to
7008 work with code or libraries compiled with another value, if they exchange
7009 information using structures or unions.
7011 @item -mabort-on-noreturn
7012 @opindex mabort-on-noreturn
7013 Generate a call to the function @code{abort} at the end of a
7014 @code{noreturn} function. It will be executed if the function tries to
7018 @itemx -mno-long-calls
7019 @opindex mlong-calls
7020 @opindex mno-long-calls
7021 Tells the compiler to perform function calls by first loading the
7022 address of the function into a register and then performing a subroutine
7023 call on this register. This switch is needed if the target function
7024 will lie outside of the 64 megabyte addressing range of the offset based
7025 version of subroutine call instruction.
7027 Even if this switch is enabled, not all function calls will be turned
7028 into long calls. The heuristic is that static functions, functions
7029 which have the @samp{short-call} attribute, functions that are inside
7030 the scope of a @samp{#pragma no_long_calls} directive and functions whose
7031 definitions have already been compiled within the current compilation
7032 unit, will not be turned into long calls. The exception to this rule is
7033 that weak function definitions, functions with the @samp{long-call}
7034 attribute or the @samp{section} attribute, and functions that are within
7035 the scope of a @samp{#pragma long_calls} directive, will always be
7036 turned into long calls.
7038 This feature is not enabled by default. Specifying
7039 @option{-mno-long-calls} will restore the default behavior, as will
7040 placing the function calls within the scope of a @samp{#pragma
7041 long_calls_off} directive. Note these switches have no effect on how
7042 the compiler generates code to handle function calls via function
7045 @item -mnop-fun-dllimport
7046 @opindex mnop-fun-dllimport
7047 Disable support for the @code{dllimport} attribute.
7049 @item -msingle-pic-base
7050 @opindex msingle-pic-base
7051 Treat the register used for PIC addressing as read-only, rather than
7052 loading it in the prologue for each function. The run-time system is
7053 responsible for initializing this register with an appropriate value
7054 before execution begins.
7056 @item -mpic-register=@var{reg}
7057 @opindex mpic-register
7058 Specify the register to be used for PIC addressing. The default is R10
7059 unless stack-checking is enabled, when R9 is used.
7061 @item -mcirrus-fix-invalid-insns
7062 @opindex mcirrus-fix-invalid-insns
7063 @opindex mno-cirrus-fix-invalid-insns
7064 Insert NOPs into the instruction stream to in order to work around
7065 problems with invalid Maverick instruction combinations. This option
7066 is only valid if the @option{-mcpu=ep9312} option has been used to
7067 enable generation of instructions for the Cirrus Maverick floating
7068 point co-processor. This option is not enabled by default, since the
7069 problem is only present in older Maverick implementations. The default
7070 can be re-enabled by use of the @option{-mno-cirrus-fix-invalid-insns}
7073 @item -mpoke-function-name
7074 @opindex mpoke-function-name
7075 Write the name of each function into the text section, directly
7076 preceding the function prologue. The generated code is similar to this:
7080 .ascii "arm_poke_function_name", 0
7083 .word 0xff000000 + (t1 - t0)
7084 arm_poke_function_name
7086 stmfd sp!, @{fp, ip, lr, pc@}
7090 When performing a stack backtrace, code can inspect the value of
7091 @code{pc} stored at @code{fp + 0}. If the trace function then looks at
7092 location @code{pc - 12} and the top 8 bits are set, then we know that
7093 there is a function name embedded immediately preceding this location
7094 and has length @code{((pc[-3]) & 0xff000000)}.
7098 Generate code for the 16-bit Thumb instruction set. The default is to
7099 use the 32-bit ARM instruction set.
7102 @opindex mtpcs-frame
7103 Generate a stack frame that is compliant with the Thumb Procedure Call
7104 Standard for all non-leaf functions. (A leaf function is one that does
7105 not call any other functions.) The default is @option{-mno-tpcs-frame}.
7107 @item -mtpcs-leaf-frame
7108 @opindex mtpcs-leaf-frame
7109 Generate a stack frame that is compliant with the Thumb Procedure Call
7110 Standard for all leaf functions. (A leaf function is one that does
7111 not call any other functions.) The default is @option{-mno-apcs-leaf-frame}.
7113 @item -mcallee-super-interworking
7114 @opindex mcallee-super-interworking
7115 Gives all externally visible functions in the file being compiled an ARM
7116 instruction set header which switches to Thumb mode before executing the
7117 rest of the function. This allows these functions to be called from
7118 non-interworking code.
7120 @item -mcaller-super-interworking
7121 @opindex mcaller-super-interworking
7122 Allows calls via function pointers (including virtual functions) to
7123 execute correctly regardless of whether the target code has been
7124 compiled for interworking or not. There is a small overhead in the cost
7125 of executing a function pointer if this option is enabled.
7130 @subsection AVR Options
7133 These options are defined for AVR implementations:
7136 @item -mmcu=@var{mcu}
7138 Specify ATMEL AVR instruction set or MCU type.
7140 Instruction set avr1 is for the minimal AVR core, not supported by the C
7141 compiler, only for assembler programs (MCU types: at90s1200, attiny10,
7142 attiny11, attiny12, attiny15, attiny28).
7144 Instruction set avr2 (default) is for the classic AVR core with up to
7145 8K program memory space (MCU types: at90s2313, at90s2323, attiny22,
7146 at90s2333, at90s2343, at90s4414, at90s4433, at90s4434, at90s8515,
7147 at90c8534, at90s8535).
7149 Instruction set avr3 is for the classic AVR core with up to 128K program
7150 memory space (MCU types: atmega103, atmega603, at43usb320, at76c711).
7152 Instruction set avr4 is for the enhanced AVR core with up to 8K program
7153 memory space (MCU types: atmega8, atmega83, atmega85).
7155 Instruction set avr5 is for the enhanced AVR core with up to 128K program
7156 memory space (MCU types: atmega16, atmega161, atmega163, atmega32, atmega323,
7157 atmega64, atmega128, at43usb355, at94k).
7161 Output instruction sizes to the asm file.
7163 @item -minit-stack=@var{N}
7164 @opindex minit-stack
7165 Specify the initial stack address, which may be a symbol or numeric value,
7166 @samp{__stack} is the default.
7168 @item -mno-interrupts
7169 @opindex mno-interrupts
7170 Generated code is not compatible with hardware interrupts.
7171 Code size will be smaller.
7173 @item -mcall-prologues
7174 @opindex mcall-prologues
7175 Functions prologues/epilogues expanded as call to appropriate
7176 subroutines. Code size will be smaller.
7178 @item -mno-tablejump
7179 @opindex mno-tablejump
7180 Do not generate tablejump insns which sometimes increase code size.
7183 @opindex mtiny-stack
7184 Change only the low 8 bits of the stack pointer.
7188 Assume int to be 8 bit integer. This affects the sizes of all types: A
7189 char will be 1 byte, an int will be 1 byte, an long will be 2 bytes
7190 and long long will be 4 bytes. Please note that this option does not
7191 comply to the C standards, but it will provide you with smaller code
7195 @node Blackfin Options
7196 @subsection Blackfin Options
7197 @cindex Blackfin Options
7200 @item -momit-leaf-frame-pointer
7201 @opindex momit-leaf-frame-pointer
7202 Don't keep the frame pointer in a register for leaf functions. This
7203 avoids the instructions to save, set up and restore frame pointers and
7204 makes an extra register available in leaf functions. The option
7205 @option{-fomit-frame-pointer} removes the frame pointer for all functions
7206 which might make debugging harder.
7210 When enabled, the compiler will ensure that the generated code does not
7211 contain speculative loads after jump instructions. This option is enabled
7216 Don't generate extra code to prevent speculative loads from occurring.
7220 When enabled, the compiler is free to take advantage of the knowledge that
7221 the entire program fits into the low 64k of memory.
7224 @opindex mno-low-64k
7225 Assume that the program is arbitrarily large. This is the default.
7227 @item -mid-shared-library
7228 @opindex mid-shared-library
7229 Generate code that supports shared libraries via the library ID method.
7230 This allows for execute in place and shared libraries in an environment
7231 without virtual memory management. This option implies @option{-fPIC}.
7233 @item -mno-id-shared-library
7234 @opindex mno-id-shared-library
7235 Generate code that doesn't assume ID based shared libraries are being used.
7236 This is the default.
7238 @item -mshared-library-id=n
7239 @opindex mshared-library-id
7240 Specified the identification number of the ID based shared library being
7241 compiled. Specifying a value of 0 will generate more compact code, specifying
7242 other values will force the allocation of that number to the current
7243 library but is no more space or time efficient than omitting this option.
7247 @subsection CRIS Options
7248 @cindex CRIS Options
7250 These options are defined specifically for the CRIS ports.
7253 @item -march=@var{architecture-type}
7254 @itemx -mcpu=@var{architecture-type}
7257 Generate code for the specified architecture. The choices for
7258 @var{architecture-type} are @samp{v3}, @samp{v8} and @samp{v10} for
7259 respectively ETRAX@w{ }4, ETRAX@w{ }100, and ETRAX@w{ }100@w{ }LX@.
7260 Default is @samp{v0} except for cris-axis-linux-gnu, where the default is
7263 @item -mtune=@var{architecture-type}
7265 Tune to @var{architecture-type} everything applicable about the generated
7266 code, except for the ABI and the set of available instructions. The
7267 choices for @var{architecture-type} are the same as for
7268 @option{-march=@var{architecture-type}}.
7270 @item -mmax-stack-frame=@var{n}
7271 @opindex mmax-stack-frame
7272 Warn when the stack frame of a function exceeds @var{n} bytes.
7274 @item -melinux-stacksize=@var{n}
7275 @opindex melinux-stacksize
7276 Only available with the @samp{cris-axis-aout} target. Arranges for
7277 indications in the program to the kernel loader that the stack of the
7278 program should be set to @var{n} bytes.
7284 The options @option{-metrax4} and @option{-metrax100} are synonyms for
7285 @option{-march=v3} and @option{-march=v8} respectively.
7287 @item -mmul-bug-workaround
7288 @itemx -mno-mul-bug-workaround
7289 @opindex mmul-bug-workaround
7290 @opindex mno-mul-bug-workaround
7291 Work around a bug in the @code{muls} and @code{mulu} instructions for CPU
7292 models where it applies. This option is active by default.
7296 Enable CRIS-specific verbose debug-related information in the assembly
7297 code. This option also has the effect to turn off the @samp{#NO_APP}
7298 formatted-code indicator to the assembler at the beginning of the
7303 Do not use condition-code results from previous instruction; always emit
7304 compare and test instructions before use of condition codes.
7306 @item -mno-side-effects
7307 @opindex mno-side-effects
7308 Do not emit instructions with side-effects in addressing modes other than
7312 @itemx -mno-stack-align
7314 @itemx -mno-data-align
7315 @itemx -mconst-align
7316 @itemx -mno-const-align
7317 @opindex mstack-align
7318 @opindex mno-stack-align
7319 @opindex mdata-align
7320 @opindex mno-data-align
7321 @opindex mconst-align
7322 @opindex mno-const-align
7323 These options (no-options) arranges (eliminate arrangements) for the
7324 stack-frame, individual data and constants to be aligned for the maximum
7325 single data access size for the chosen CPU model. The default is to
7326 arrange for 32-bit alignment. ABI details such as structure layout are
7327 not affected by these options.
7335 Similar to the stack- data- and const-align options above, these options
7336 arrange for stack-frame, writable data and constants to all be 32-bit,
7337 16-bit or 8-bit aligned. The default is 32-bit alignment.
7339 @item -mno-prologue-epilogue
7340 @itemx -mprologue-epilogue
7341 @opindex mno-prologue-epilogue
7342 @opindex mprologue-epilogue
7343 With @option{-mno-prologue-epilogue}, the normal function prologue and
7344 epilogue that sets up the stack-frame are omitted and no return
7345 instructions or return sequences are generated in the code. Use this
7346 option only together with visual inspection of the compiled code: no
7347 warnings or errors are generated when call-saved registers must be saved,
7348 or storage for local variable needs to be allocated.
7354 With @option{-fpic} and @option{-fPIC}, don't generate (do generate)
7355 instruction sequences that load addresses for functions from the PLT part
7356 of the GOT rather than (traditional on other architectures) calls to the
7357 PLT@. The default is @option{-mgotplt}.
7361 Legacy no-op option only recognized with the cris-axis-aout target.
7365 Legacy no-op option only recognized with the cris-axis-elf and
7366 cris-axis-linux-gnu targets.
7370 Only recognized with the cris-axis-aout target, where it selects a
7371 GNU/linux-like multilib, include files and instruction set for
7376 Legacy no-op option only recognized with the cris-axis-linux-gnu target.
7380 This option, recognized for the cris-axis-aout and cris-axis-elf arranges
7381 to link with input-output functions from a simulator library. Code,
7382 initialized data and zero-initialized data are allocated consecutively.
7386 Like @option{-sim}, but pass linker options to locate initialized data at
7387 0x40000000 and zero-initialized data at 0x80000000.
7390 @node Darwin Options
7391 @subsection Darwin Options
7392 @cindex Darwin options
7394 These options are defined for all architectures running the Darwin operating
7397 FSF GCC on Darwin does not create ``fat'' object files; it will create
7398 an object file for the single architecture that it was built to
7399 target. Apple's GCC on Darwin does create ``fat'' files if multiple
7400 @option{-arch} options are used; it does so by running the compiler or
7401 linker multiple times and joining the results together with
7404 The subtype of the file created (like @samp{ppc7400} or @samp{ppc970} or
7405 @samp{i686}) is determined by the flags that specify the ISA
7406 that GCC is targetting, like @option{-mcpu} or @option{-march}. The
7407 @option{-force_cpusubtype_ALL} option can be used to override this.
7409 The Darwin tools vary in their behavior when presented with an ISA
7410 mismatch. The assembler, @file{as}, will only permit instructions to
7411 be used that are valid for the subtype of the file it is generating,
7412 so you cannot put 64-bit instructions in an @samp{ppc750} object file.
7413 The linker for shared libraries, @file{/usr/bin/libtool}, will fail
7414 and print an error if asked to create a shared library with a less
7415 restrictive subtype than its input files (for instance, trying to put
7416 a @samp{ppc970} object file in a @samp{ppc7400} library). The linker
7417 for executables, @file{ld}, will quietly give the executable the most
7418 restrictive subtype of any of its input files.
7423 Add the framework directory @var{dir} to the head of the list of
7424 directories to be searched for header files. These directories are
7425 interleaved with those specified by @option{-I} options and are
7426 scanned in a left-to-right order.
7428 A framework directory is a directory with frameworks in it. A
7429 framework is a directory with a @samp{"Headers"} and/or
7430 @samp{"PrivateHeaders"} directory contained directly in it that ends
7431 in @samp{".framework"}. The name of a framework is the name of this
7432 directory excluding the @samp{".framework"}. Headers associated with
7433 the framework are found in one of those two directories, with
7434 @samp{"Headers"} being searched first. A subframework is a framework
7435 directory that is in a framework's @samp{"Frameworks"} directory.
7436 Includes of subframework headers can only appear in a header of a
7437 framework that contains the subframework, or in a sibling subframework
7438 header. Two subframeworks are siblings if they occur in the same
7439 framework. A subframework should not have the same name as a
7440 framework, a warning will be issued if this is violated. Currently a
7441 subframework cannot have subframeworks, in the future, the mechanism
7442 may be extended to support this. The standard frameworks can be found
7443 in @samp{"/System/Library/Frameworks"} and
7444 @samp{"/Library/Frameworks"}. An example include looks like
7445 @code{#include <Framework/header.h>}, where @samp{Framework} denotes
7446 the name of the framework and header.h is found in the
7447 @samp{"PrivateHeaders"} or @samp{"Headers"} directory.
7451 Emit debugging information for symbols that are used. For STABS
7452 debugging format, this enables @option{-feliminate-unused-debug-symbols}.
7453 This is by default ON@.
7457 Emit debugging information for all symbols and types.
7459 @item -mone-byte-bool
7460 @opindex -mone-byte-bool
7461 Override the defaults for @samp{bool} so that @samp{sizeof(bool)==1}.
7462 By default @samp{sizeof(bool)} is @samp{4} when compiling for
7463 Darwin/PowerPC and @samp{1} when compiling for Darwin/x86, so this
7464 option has no effect on x86.
7466 @strong{Warning:} The @option{-mone-byte-bool} switch causes GCC
7467 to generate code that is not binary compatible with code generated
7468 without that switch. Using this switch may require recompiling all
7469 other modules in a program, including system libraries. Use this
7470 switch to conform to a non-default data model.
7472 @item -mfix-and-continue
7473 @itemx -ffix-and-continue
7474 @itemx -findirect-data
7475 @opindex mfix-and-continue
7476 @opindex ffix-and-continue
7477 @opindex findirect-data
7478 Generate code suitable for fast turn around development. Needed to
7479 enable gdb to dynamically load @code{.o} files into already running
7480 programs. @option{-findirect-data} and @option{-ffix-and-continue}
7481 are provided for backwards compatibility.
7485 Loads all members of static archive libraries.
7486 See man ld(1) for more information.
7488 @item -arch_errors_fatal
7489 @opindex arch_errors_fatal
7490 Cause the errors having to do with files that have the wrong architecture
7494 @opindex bind_at_load
7495 Causes the output file to be marked such that the dynamic linker will
7496 bind all undefined references when the file is loaded or launched.
7500 Produce a Mach-o bundle format file.
7501 See man ld(1) for more information.
7503 @item -bundle_loader @var{executable}
7504 @opindex bundle_loader
7505 This option specifies the @var{executable} that will be loading the build
7506 output file being linked. See man ld(1) for more information.
7509 @opindex -dynamiclib
7510 When passed this option, GCC will produce a dynamic library instead of
7511 an executable when linking, using the Darwin @file{libtool} command.
7513 @item -force_cpusubtype_ALL
7514 @opindex -force_cpusubtype_ALL
7515 This causes GCC's output file to have the @var{ALL} subtype, instead of
7516 one controlled by the @option{-mcpu} or @option{-march} option.
7518 @item -allowable_client @var{client_name}
7520 @itemx -compatibility_version
7521 @itemx -current_version
7523 @itemx -dependency-file
7525 @itemx -dylinker_install_name
7527 @itemx -exported_symbols_list
7529 @itemx -flat_namespace
7530 @itemx -force_flat_namespace
7531 @itemx -headerpad_max_install_names
7534 @itemx -install_name
7535 @itemx -keep_private_externs
7536 @itemx -multi_module
7537 @itemx -multiply_defined
7538 @itemx -multiply_defined_unused
7540 @itemx -no_dead_strip_inits_and_terms
7541 @itemx -nofixprebinding
7544 @itemx -noseglinkedit
7545 @itemx -pagezero_size
7547 @itemx -prebind_all_twolevel_modules
7548 @itemx -private_bundle
7549 @itemx -read_only_relocs
7551 @itemx -sectobjectsymbols
7555 @itemx -sectobjectsymbols
7558 @itemx -segs_read_only_addr
7559 @itemx -segs_read_write_addr
7560 @itemx -seg_addr_table
7561 @itemx -seg_addr_table_filename
7564 @itemx -segs_read_only_addr
7565 @itemx -segs_read_write_addr
7566 @itemx -single_module
7569 @itemx -sub_umbrella
7570 @itemx -twolevel_namespace
7573 @itemx -unexported_symbols_list
7574 @itemx -weak_reference_mismatches
7577 @opindex allowable_client
7578 @opindex client_name
7579 @opindex compatibility_version
7580 @opindex current_version
7582 @opindex dependency-file
7584 @opindex dylinker_install_name
7586 @opindex exported_symbols_list
7588 @opindex flat_namespace
7589 @opindex force_flat_namespace
7590 @opindex headerpad_max_install_names
7593 @opindex install_name
7594 @opindex keep_private_externs
7595 @opindex multi_module
7596 @opindex multiply_defined
7597 @opindex multiply_defined_unused
7599 @opindex no_dead_strip_inits_and_terms
7600 @opindex nofixprebinding
7601 @opindex nomultidefs
7603 @opindex noseglinkedit
7604 @opindex pagezero_size
7606 @opindex prebind_all_twolevel_modules
7607 @opindex private_bundle
7608 @opindex read_only_relocs
7610 @opindex sectobjectsymbols
7614 @opindex sectobjectsymbols
7617 @opindex segs_read_only_addr
7618 @opindex segs_read_write_addr
7619 @opindex seg_addr_table
7620 @opindex seg_addr_table_filename
7621 @opindex seglinkedit
7623 @opindex segs_read_only_addr
7624 @opindex segs_read_write_addr
7625 @opindex single_module
7627 @opindex sub_library
7628 @opindex sub_umbrella
7629 @opindex twolevel_namespace
7632 @opindex unexported_symbols_list
7633 @opindex weak_reference_mismatches
7634 @opindex whatsloaded
7636 These options are passed to the Darwin linker. The Darwin linker man page
7637 describes them in detail.
7640 @node DEC Alpha Options
7641 @subsection DEC Alpha Options
7643 These @samp{-m} options are defined for the DEC Alpha implementations:
7646 @item -mno-soft-float
7648 @opindex mno-soft-float
7649 @opindex msoft-float
7650 Use (do not use) the hardware floating-point instructions for
7651 floating-point operations. When @option{-msoft-float} is specified,
7652 functions in @file{libgcc.a} will be used to perform floating-point
7653 operations. Unless they are replaced by routines that emulate the
7654 floating-point operations, or compiled in such a way as to call such
7655 emulations routines, these routines will issue floating-point
7656 operations. If you are compiling for an Alpha without floating-point
7657 operations, you must ensure that the library is built so as not to call
7660 Note that Alpha implementations without floating-point operations are
7661 required to have floating-point registers.
7666 @opindex mno-fp-regs
7667 Generate code that uses (does not use) the floating-point register set.
7668 @option{-mno-fp-regs} implies @option{-msoft-float}. If the floating-point
7669 register set is not used, floating point operands are passed in integer
7670 registers as if they were integers and floating-point results are passed
7671 in @code{$0} instead of @code{$f0}. This is a non-standard calling sequence,
7672 so any function with a floating-point argument or return value called by code
7673 compiled with @option{-mno-fp-regs} must also be compiled with that
7676 A typical use of this option is building a kernel that does not use,
7677 and hence need not save and restore, any floating-point registers.
7681 The Alpha architecture implements floating-point hardware optimized for
7682 maximum performance. It is mostly compliant with the IEEE floating
7683 point standard. However, for full compliance, software assistance is
7684 required. This option generates code fully IEEE compliant code
7685 @emph{except} that the @var{inexact-flag} is not maintained (see below).
7686 If this option is turned on, the preprocessor macro @code{_IEEE_FP} is
7687 defined during compilation. The resulting code is less efficient but is
7688 able to correctly support denormalized numbers and exceptional IEEE
7689 values such as not-a-number and plus/minus infinity. Other Alpha
7690 compilers call this option @option{-ieee_with_no_inexact}.
7692 @item -mieee-with-inexact
7693 @opindex mieee-with-inexact
7694 This is like @option{-mieee} except the generated code also maintains
7695 the IEEE @var{inexact-flag}. Turning on this option causes the
7696 generated code to implement fully-compliant IEEE math. In addition to
7697 @code{_IEEE_FP}, @code{_IEEE_FP_EXACT} is defined as a preprocessor
7698 macro. On some Alpha implementations the resulting code may execute
7699 significantly slower than the code generated by default. Since there is
7700 very little code that depends on the @var{inexact-flag}, you should
7701 normally not specify this option. Other Alpha compilers call this
7702 option @option{-ieee_with_inexact}.
7704 @item -mfp-trap-mode=@var{trap-mode}
7705 @opindex mfp-trap-mode
7706 This option controls what floating-point related traps are enabled.
7707 Other Alpha compilers call this option @option{-fptm @var{trap-mode}}.
7708 The trap mode can be set to one of four values:
7712 This is the default (normal) setting. The only traps that are enabled
7713 are the ones that cannot be disabled in software (e.g., division by zero
7717 In addition to the traps enabled by @samp{n}, underflow traps are enabled
7721 Like @samp{su}, but the instructions are marked to be safe for software
7722 completion (see Alpha architecture manual for details).
7725 Like @samp{su}, but inexact traps are enabled as well.
7728 @item -mfp-rounding-mode=@var{rounding-mode}
7729 @opindex mfp-rounding-mode
7730 Selects the IEEE rounding mode. Other Alpha compilers call this option
7731 @option{-fprm @var{rounding-mode}}. The @var{rounding-mode} can be one
7736 Normal IEEE rounding mode. Floating point numbers are rounded towards
7737 the nearest machine number or towards the even machine number in case
7741 Round towards minus infinity.
7744 Chopped rounding mode. Floating point numbers are rounded towards zero.
7747 Dynamic rounding mode. A field in the floating point control register
7748 (@var{fpcr}, see Alpha architecture reference manual) controls the
7749 rounding mode in effect. The C library initializes this register for
7750 rounding towards plus infinity. Thus, unless your program modifies the
7751 @var{fpcr}, @samp{d} corresponds to round towards plus infinity.
7754 @item -mtrap-precision=@var{trap-precision}
7755 @opindex mtrap-precision
7756 In the Alpha architecture, floating point traps are imprecise. This
7757 means without software assistance it is impossible to recover from a
7758 floating trap and program execution normally needs to be terminated.
7759 GCC can generate code that can assist operating system trap handlers
7760 in determining the exact location that caused a floating point trap.
7761 Depending on the requirements of an application, different levels of
7762 precisions can be selected:
7766 Program precision. This option is the default and means a trap handler
7767 can only identify which program caused a floating point exception.
7770 Function precision. The trap handler can determine the function that
7771 caused a floating point exception.
7774 Instruction precision. The trap handler can determine the exact
7775 instruction that caused a floating point exception.
7778 Other Alpha compilers provide the equivalent options called
7779 @option{-scope_safe} and @option{-resumption_safe}.
7781 @item -mieee-conformant
7782 @opindex mieee-conformant
7783 This option marks the generated code as IEEE conformant. You must not
7784 use this option unless you also specify @option{-mtrap-precision=i} and either
7785 @option{-mfp-trap-mode=su} or @option{-mfp-trap-mode=sui}. Its only effect
7786 is to emit the line @samp{.eflag 48} in the function prologue of the
7787 generated assembly file. Under DEC Unix, this has the effect that
7788 IEEE-conformant math library routines will be linked in.
7790 @item -mbuild-constants
7791 @opindex mbuild-constants
7792 Normally GCC examines a 32- or 64-bit integer constant to
7793 see if it can construct it from smaller constants in two or three
7794 instructions. If it cannot, it will output the constant as a literal and
7795 generate code to load it from the data segment at runtime.
7797 Use this option to require GCC to construct @emph{all} integer constants
7798 using code, even if it takes more instructions (the maximum is six).
7800 You would typically use this option to build a shared library dynamic
7801 loader. Itself a shared library, it must relocate itself in memory
7802 before it can find the variables and constants in its own data segment.
7808 Select whether to generate code to be assembled by the vendor-supplied
7809 assembler (@option{-malpha-as}) or by the GNU assembler @option{-mgas}.
7827 Indicate whether GCC should generate code to use the optional BWX,
7828 CIX, FIX and MAX instruction sets. The default is to use the instruction
7829 sets supported by the CPU type specified via @option{-mcpu=} option or that
7830 of the CPU on which GCC was built if none was specified.
7835 @opindex mfloat-ieee
7836 Generate code that uses (does not use) VAX F and G floating point
7837 arithmetic instead of IEEE single and double precision.
7839 @item -mexplicit-relocs
7840 @itemx -mno-explicit-relocs
7841 @opindex mexplicit-relocs
7842 @opindex mno-explicit-relocs
7843 Older Alpha assemblers provided no way to generate symbol relocations
7844 except via assembler macros. Use of these macros does not allow
7845 optimal instruction scheduling. GNU binutils as of version 2.12
7846 supports a new syntax that allows the compiler to explicitly mark
7847 which relocations should apply to which instructions. This option
7848 is mostly useful for debugging, as GCC detects the capabilities of
7849 the assembler when it is built and sets the default accordingly.
7853 @opindex msmall-data
7854 @opindex mlarge-data
7855 When @option{-mexplicit-relocs} is in effect, static data is
7856 accessed via @dfn{gp-relative} relocations. When @option{-msmall-data}
7857 is used, objects 8 bytes long or smaller are placed in a @dfn{small data area}
7858 (the @code{.sdata} and @code{.sbss} sections) and are accessed via
7859 16-bit relocations off of the @code{$gp} register. This limits the
7860 size of the small data area to 64KB, but allows the variables to be
7861 directly accessed via a single instruction.
7863 The default is @option{-mlarge-data}. With this option the data area
7864 is limited to just below 2GB@. Programs that require more than 2GB of
7865 data must use @code{malloc} or @code{mmap} to allocate the data in the
7866 heap instead of in the program's data segment.
7868 When generating code for shared libraries, @option{-fpic} implies
7869 @option{-msmall-data} and @option{-fPIC} implies @option{-mlarge-data}.
7873 @opindex msmall-text
7874 @opindex mlarge-text
7875 When @option{-msmall-text} is used, the compiler assumes that the
7876 code of the entire program (or shared library) fits in 4MB, and is
7877 thus reachable with a branch instruction. When @option{-msmall-data}
7878 is used, the compiler can assume that all local symbols share the
7879 same @code{$gp} value, and thus reduce the number of instructions
7880 required for a function call from 4 to 1.
7882 The default is @option{-mlarge-text}.
7884 @item -mcpu=@var{cpu_type}
7886 Set the instruction set and instruction scheduling parameters for
7887 machine type @var{cpu_type}. You can specify either the @samp{EV}
7888 style name or the corresponding chip number. GCC supports scheduling
7889 parameters for the EV4, EV5 and EV6 family of processors and will
7890 choose the default values for the instruction set from the processor
7891 you specify. If you do not specify a processor type, GCC will default
7892 to the processor on which the compiler was built.
7894 Supported values for @var{cpu_type} are
7900 Schedules as an EV4 and has no instruction set extensions.
7904 Schedules as an EV5 and has no instruction set extensions.
7908 Schedules as an EV5 and supports the BWX extension.
7913 Schedules as an EV5 and supports the BWX and MAX extensions.
7917 Schedules as an EV6 and supports the BWX, FIX, and MAX extensions.
7921 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX extensions.
7924 @item -mtune=@var{cpu_type}
7926 Set only the instruction scheduling parameters for machine type
7927 @var{cpu_type}. The instruction set is not changed.
7929 @item -mmemory-latency=@var{time}
7930 @opindex mmemory-latency
7931 Sets the latency the scheduler should assume for typical memory
7932 references as seen by the application. This number is highly
7933 dependent on the memory access patterns used by the application
7934 and the size of the external cache on the machine.
7936 Valid options for @var{time} are
7940 A decimal number representing clock cycles.
7946 The compiler contains estimates of the number of clock cycles for
7947 ``typical'' EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7948 (also called Dcache, Scache, and Bcache), as well as to main memory.
7949 Note that L3 is only valid for EV5.
7954 @node DEC Alpha/VMS Options
7955 @subsection DEC Alpha/VMS Options
7957 These @samp{-m} options are defined for the DEC Alpha/VMS implementations:
7960 @item -mvms-return-codes
7961 @opindex mvms-return-codes
7962 Return VMS condition codes from main. The default is to return POSIX
7963 style condition (e.g.@ error) codes.
7967 @subsection FRV Options
7974 Only use the first 32 general purpose registers.
7979 Use all 64 general purpose registers.
7984 Use only the first 32 floating point registers.
7989 Use all 64 floating point registers
7992 @opindex mhard-float
7994 Use hardware instructions for floating point operations.
7997 @opindex msoft-float
7999 Use library routines for floating point operations.
8004 Dynamically allocate condition code registers.
8009 Do not try to dynamically allocate condition code registers, only
8010 use @code{icc0} and @code{fcc0}.
8015 Change ABI to use double word insns.
8020 Do not use double word instructions.
8025 Use floating point double instructions.
8030 Do not use floating point double instructions.
8035 Use media instructions.
8040 Do not use media instructions.
8045 Use multiply and add/subtract instructions.
8050 Do not use multiply and add/subtract instructions.
8055 Select the FDPIC ABI, that uses function descriptors to represent
8056 pointers to functions. Without any PIC/PIE-related options, it
8057 implies @option{-fPIE}. With @option{-fpic} or @option{-fpie}, it
8058 assumes GOT entries and small data are within a 12-bit range from the
8059 GOT base address; with @option{-fPIC} or @option{-fPIE}, GOT offsets
8060 are computed with 32 bits.
8063 @opindex minline-plt
8065 Enable inlining of PLT entries in function calls to functions that are
8066 not known to bind locally. It has no effect without @option{-mfdpic}.
8067 It's enabled by default if optimizing for speed and compiling for
8068 shared libraries (i.e., @option{-fPIC} or @option{-fpic}), or when an
8069 optimization option such as @option{-O3} or above is present in the
8075 Assume a large TLS segment when generating thread-local code.
8080 Do not assume a large TLS segment when generating thread-local code.
8085 Enable the use of @code{GPREL} relocations in the FDPIC ABI for data
8086 that is known to be in read-only sections. It's enabled by default,
8087 except for @option{-fpic} or @option{-fpie}: even though it may help
8088 make the global offset table smaller, it trades 1 instruction for 4.
8089 With @option{-fPIC} or @option{-fPIE}, it trades 3 instructions for 4,
8090 one of which may be shared by multiple symbols, and it avoids the need
8091 for a GOT entry for the referenced symbol, so it's more likely to be a
8092 win. If it is not, @option{-mno-gprel-ro} can be used to disable it.
8094 @item -multilib-library-pic
8095 @opindex multilib-library-pic
8097 Link with the (library, not FD) pic libraries. It's implied by
8098 @option{-mlibrary-pic}, as well as by @option{-fPIC} and
8099 @option{-fpic} without @option{-mfdpic}. You should never have to use
8105 Follow the EABI requirement of always creating a frame pointer whenever
8106 a stack frame is allocated. This option is enabled by default and can
8107 be disabled with @option{-mno-linked-fp}.
8110 @opindex mlong-calls
8112 Use indirect addressing to call functions outside the current
8113 compilation unit. This allows the functions to be placed anywhere
8114 within the 32-bit address space.
8116 @item -malign-labels
8117 @opindex malign-labels
8119 Try to align labels to an 8-byte boundary by inserting nops into the
8120 previous packet. This option only has an effect when VLIW packing
8121 is enabled. It doesn't create new packets; it merely adds nops to
8125 @opindex mlibrary-pic
8127 Generate position-independent EABI code.
8132 Use only the first four media accumulator registers.
8137 Use all eight media accumulator registers.
8142 Pack VLIW instructions.
8147 Do not pack VLIW instructions.
8152 Do not mark ABI switches in e_flags.
8157 Enable the use of conditional-move instructions (default).
8159 This switch is mainly for debugging the compiler and will likely be removed
8160 in a future version.
8162 @item -mno-cond-move
8163 @opindex mno-cond-move
8165 Disable the use of conditional-move instructions.
8167 This switch is mainly for debugging the compiler and will likely be removed
8168 in a future version.
8173 Enable the use of conditional set instructions (default).
8175 This switch is mainly for debugging the compiler and will likely be removed
8176 in a future version.
8181 Disable the use of conditional set instructions.
8183 This switch is mainly for debugging the compiler and will likely be removed
8184 in a future version.
8189 Enable the use of conditional execution (default).
8191 This switch is mainly for debugging the compiler and will likely be removed
8192 in a future version.
8194 @item -mno-cond-exec
8195 @opindex mno-cond-exec
8197 Disable the use of conditional execution.
8199 This switch is mainly for debugging the compiler and will likely be removed
8200 in a future version.
8203 @opindex mvliw-branch
8205 Run a pass to pack branches into VLIW instructions (default).
8207 This switch is mainly for debugging the compiler and will likely be removed
8208 in a future version.
8210 @item -mno-vliw-branch
8211 @opindex mno-vliw-branch
8213 Do not run a pass to pack branches into VLIW instructions.
8215 This switch is mainly for debugging the compiler and will likely be removed
8216 in a future version.
8218 @item -mmulti-cond-exec
8219 @opindex mmulti-cond-exec
8221 Enable optimization of @code{&&} and @code{||} in conditional execution
8224 This switch is mainly for debugging the compiler and will likely be removed
8225 in a future version.
8227 @item -mno-multi-cond-exec
8228 @opindex mno-multi-cond-exec
8230 Disable optimization of @code{&&} and @code{||} in conditional execution.
8232 This switch is mainly for debugging the compiler and will likely be removed
8233 in a future version.
8235 @item -mnested-cond-exec
8236 @opindex mnested-cond-exec
8238 Enable nested conditional execution optimizations (default).
8240 This switch is mainly for debugging the compiler and will likely be removed
8241 in a future version.
8243 @item -mno-nested-cond-exec
8244 @opindex mno-nested-cond-exec
8246 Disable nested conditional execution optimizations.
8248 This switch is mainly for debugging the compiler and will likely be removed
8249 in a future version.
8251 @item -mtomcat-stats
8252 @opindex mtomcat-stats
8254 Cause gas to print out tomcat statistics.
8256 @item -mcpu=@var{cpu}
8259 Select the processor type for which to generate code. Possible values are
8260 @samp{frv}, @samp{fr550}, @samp{tomcat}, @samp{fr500}, @samp{fr450},
8261 @samp{fr405}, @samp{fr400}, @samp{fr300} and @samp{simple}.
8265 @node H8/300 Options
8266 @subsection H8/300 Options
8268 These @samp{-m} options are defined for the H8/300 implementations:
8273 Shorten some address references at link time, when possible; uses the
8274 linker option @option{-relax}. @xref{H8/300,, @code{ld} and the H8/300,
8275 ld, Using ld}, for a fuller description.
8279 Generate code for the H8/300H@.
8283 Generate code for the H8S@.
8287 Generate code for the H8S and H8/300H in the normal mode. This switch
8288 must be used either with @option{-mh} or @option{-ms}.
8292 Generate code for the H8S/2600. This switch must be used with @option{-ms}.
8296 Make @code{int} data 32 bits by default.
8300 On the H8/300H and H8S, use the same alignment rules as for the H8/300.
8301 The default for the H8/300H and H8S is to align longs and floats on 4
8303 @option{-malign-300} causes them to be aligned on 2 byte boundaries.
8304 This option has no effect on the H8/300.
8308 @subsection HPPA Options
8309 @cindex HPPA Options
8311 These @samp{-m} options are defined for the HPPA family of computers:
8314 @item -march=@var{architecture-type}
8316 Generate code for the specified architecture. The choices for
8317 @var{architecture-type} are @samp{1.0} for PA 1.0, @samp{1.1} for PA
8318 1.1, and @samp{2.0} for PA 2.0 processors. Refer to
8319 @file{/usr/lib/sched.models} on an HP-UX system to determine the proper
8320 architecture option for your machine. Code compiled for lower numbered
8321 architectures will run on higher numbered architectures, but not the
8325 @itemx -mpa-risc-1-1
8326 @itemx -mpa-risc-2-0
8327 @opindex mpa-risc-1-0
8328 @opindex mpa-risc-1-1
8329 @opindex mpa-risc-2-0
8330 Synonyms for @option{-march=1.0}, @option{-march=1.1}, and @option{-march=2.0} respectively.
8333 @opindex mbig-switch
8334 Generate code suitable for big switch tables. Use this option only if
8335 the assembler/linker complain about out of range branches within a switch
8338 @item -mjump-in-delay
8339 @opindex mjump-in-delay
8340 Fill delay slots of function calls with unconditional jump instructions
8341 by modifying the return pointer for the function call to be the target
8342 of the conditional jump.
8344 @item -mdisable-fpregs
8345 @opindex mdisable-fpregs
8346 Prevent floating point registers from being used in any manner. This is
8347 necessary for compiling kernels which perform lazy context switching of
8348 floating point registers. If you use this option and attempt to perform
8349 floating point operations, the compiler will abort.
8351 @item -mdisable-indexing
8352 @opindex mdisable-indexing
8353 Prevent the compiler from using indexing address modes. This avoids some
8354 rather obscure problems when compiling MIG generated code under MACH@.
8356 @item -mno-space-regs
8357 @opindex mno-space-regs
8358 Generate code that assumes the target has no space registers. This allows
8359 GCC to generate faster indirect calls and use unscaled index address modes.
8361 Such code is suitable for level 0 PA systems and kernels.
8363 @item -mfast-indirect-calls
8364 @opindex mfast-indirect-calls
8365 Generate code that assumes calls never cross space boundaries. This
8366 allows GCC to emit code which performs faster indirect calls.
8368 This option will not work in the presence of shared libraries or nested
8371 @item -mfixed-range=@var{register-range}
8372 @opindex mfixed-range
8373 Generate code treating the given register range as fixed registers.
8374 A fixed register is one that the register allocator can not use. This is
8375 useful when compiling kernel code. A register range is specified as
8376 two registers separated by a dash. Multiple register ranges can be
8377 specified separated by a comma.
8379 @item -mlong-load-store
8380 @opindex mlong-load-store
8381 Generate 3-instruction load and store sequences as sometimes required by
8382 the HP-UX 10 linker. This is equivalent to the @samp{+k} option to
8385 @item -mportable-runtime
8386 @opindex mportable-runtime
8387 Use the portable calling conventions proposed by HP for ELF systems.
8391 Enable the use of assembler directives only GAS understands.
8393 @item -mschedule=@var{cpu-type}
8395 Schedule code according to the constraints for the machine type
8396 @var{cpu-type}. The choices for @var{cpu-type} are @samp{700}
8397 @samp{7100}, @samp{7100LC}, @samp{7200}, @samp{7300} and @samp{8000}. Refer
8398 to @file{/usr/lib/sched.models} on an HP-UX system to determine the
8399 proper scheduling option for your machine. The default scheduling is
8403 @opindex mlinker-opt
8404 Enable the optimization pass in the HP-UX linker. Note this makes symbolic
8405 debugging impossible. It also triggers a bug in the HP-UX 8 and HP-UX 9
8406 linkers in which they give bogus error messages when linking some programs.
8409 @opindex msoft-float
8410 Generate output containing library calls for floating point.
8411 @strong{Warning:} the requisite libraries are not available for all HPPA
8412 targets. Normally the facilities of the machine's usual C compiler are
8413 used, but this cannot be done directly in cross-compilation. You must make
8414 your own arrangements to provide suitable library functions for
8415 cross-compilation. The embedded target @samp{hppa1.1-*-pro}
8416 does provide software floating point support.
8418 @option{-msoft-float} changes the calling convention in the output file;
8419 therefore, it is only useful if you compile @emph{all} of a program with
8420 this option. In particular, you need to compile @file{libgcc.a}, the
8421 library that comes with GCC, with @option{-msoft-float} in order for
8426 Generate the predefine, @code{_SIO}, for server IO@. The default is
8427 @option{-mwsio}. This generates the predefines, @code{__hp9000s700},
8428 @code{__hp9000s700__} and @code{_WSIO}, for workstation IO@. These
8429 options are available under HP-UX and HI-UX@.
8433 Use GNU ld specific options. This passes @option{-shared} to ld when
8434 building a shared library. It is the default when GCC is configured,
8435 explicitly or implicitly, with the GNU linker. This option does not
8436 have any affect on which ld is called, it only changes what parameters
8437 are passed to that ld. The ld that is called is determined by the
8438 @option{--with-ld} configure option, GCC's program search path, and
8439 finally by the user's @env{PATH}. The linker used by GCC can be printed
8440 using @samp{which `gcc -print-prog-name=ld`}.
8444 Use HP ld specific options. This passes @option{-b} to ld when building
8445 a shared library and passes @option{+Accept TypeMismatch} to ld on all
8446 links. It is the default when GCC is configured, explicitly or
8447 implicitly, with the HP linker. This option does not have any affect on
8448 which ld is called, it only changes what parameters are passed to that
8449 ld. The ld that is called is determined by the @option{--with-ld}
8450 configure option, GCC's program search path, and finally by the user's
8451 @env{PATH}. The linker used by GCC can be printed using @samp{which
8452 `gcc -print-prog-name=ld`}.
8455 @opindex mno-long-calls
8456 Generate code that uses long call sequences. This ensures that a call
8457 is always able to reach linker generated stubs. The default is to generate
8458 long calls only when the distance from the call site to the beginning
8459 of the function or translation unit, as the case may be, exceeds a
8460 predefined limit set by the branch type being used. The limits for
8461 normal calls are 7,600,000 and 240,000 bytes, respectively for the
8462 PA 2.0 and PA 1.X architectures. Sibcalls are always limited at
8465 Distances are measured from the beginning of functions when using the
8466 @option{-ffunction-sections} option, or when using the @option{-mgas}
8467 and @option{-mno-portable-runtime} options together under HP-UX with
8470 It is normally not desirable to use this option as it will degrade
8471 performance. However, it may be useful in large applications,
8472 particularly when partial linking is used to build the application.
8474 The types of long calls used depends on the capabilities of the
8475 assembler and linker, and the type of code being generated. The
8476 impact on systems that support long absolute calls, and long pic
8477 symbol-difference or pc-relative calls should be relatively small.
8478 However, an indirect call is used on 32-bit ELF systems in pic code
8479 and it is quite long.
8481 @item -munix=@var{unix-std}
8483 Generate compiler predefines and select a startfile for the specified
8484 UNIX standard. The choices for @var{unix-std} are @samp{93}, @samp{95}
8485 and @samp{98}. @samp{93} is supported on all HP-UX versions. @samp{95}
8486 is available on HP-UX 10.10 and later. @samp{98} is available on HP-UX
8487 11.11 and later. The default values are @samp{93} for HP-UX 10.00,
8488 @samp{95} for HP-UX 10.10 though to 11.00, and @samp{98} for HP-UX 11.11
8491 @option{-munix=93} provides the same predefines as GCC 3.3 and 3.4.
8492 @option{-munix=95} provides additional predefines for @code{XOPEN_UNIX}
8493 and @code{_XOPEN_SOURCE_EXTENDED}, and the startfile @file{unix95.o}.
8494 @option{-munix=98} provides additional predefines for @code{_XOPEN_UNIX},
8495 @code{_XOPEN_SOURCE_EXTENDED}, @code{_INCLUDE__STDC_A1_SOURCE} and
8496 @code{_INCLUDE_XOPEN_SOURCE_500}, and the startfile @file{unix98.o}.
8498 It is @emph{important} to note that this option changes the interfaces
8499 for various library routines. It also affects the operational behavior
8500 of the C library. Thus, @emph{extreme} care is needed in using this
8503 Library code that is intended to operate with more than one UNIX
8504 standard must test, set and restore the variable @var{__xpg4_extended_mask}
8505 as appropriate. Most GNU software doesn't provide this capability.
8509 Suppress the generation of link options to search libdld.sl when the
8510 @option{-static} option is specified on HP-UX 10 and later.
8514 The HP-UX implementation of setlocale in libc has a dependency on
8515 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8516 when the @option{-static} option is specified, special link options
8517 are needed to resolve this dependency.
8519 On HP-UX 10 and later, the GCC driver adds the necessary options to
8520 link with libdld.sl when the @option{-static} option is specified.
8521 This causes the resulting binary to be dynamic. On the 64-bit port,
8522 the linkers generate dynamic binaries by default in any case. The
8523 @option{-nolibdld} option can be used to prevent the GCC driver from
8524 adding these link options.
8528 Add support for multithreading with the @dfn{dce thread} library
8529 under HP-UX@. This option sets flags for both the preprocessor and
8533 @node i386 and x86-64 Options
8534 @subsection Intel 386 and AMD x86-64 Options
8535 @cindex i386 Options
8536 @cindex x86-64 Options
8537 @cindex Intel 386 Options
8538 @cindex AMD x86-64 Options
8540 These @samp{-m} options are defined for the i386 and x86-64 family of
8544 @item -mtune=@var{cpu-type}
8546 Tune to @var{cpu-type} everything applicable about the generated code, except
8547 for the ABI and the set of available instructions. The choices for
8551 Original Intel's i386 CPU@.
8553 Intel's i486 CPU@. (No scheduling is implemented for this chip.)
8555 Intel Pentium CPU with no MMX support.
8557 Intel PentiumMMX CPU based on Pentium core with MMX instruction set support.
8558 @item i686, pentiumpro
8559 Intel PentiumPro CPU@.
8561 Intel Pentium2 CPU based on PentiumPro core with MMX instruction set support.
8562 @item pentium3, pentium3m
8563 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE instruction set
8566 Low power version of Intel Pentium3 CPU with MMX, SSE and SSE2 instruction set
8567 support. Used by Centrino notebooks.
8568 @item pentium4, pentium4m
8569 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set support.
8571 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2 and SSE3 instruction
8574 Improved version of Intel Pentium4 CPU with 64-bit extensions, MMX, SSE,
8575 SSE2 and SSE3 instruction set support.
8577 AMD K6 CPU with MMX instruction set support.
8579 Improved versions of AMD K6 CPU with MMX and 3dNOW! instruction set support.
8580 @item athlon, athlon-tbird
8581 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE prefetch instructions
8583 @item athlon-4, athlon-xp, athlon-mp
8584 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and full SSE
8585 instruction set support.
8586 @item k8, opteron, athlon64, athlon-fx
8587 AMD K8 core based CPUs with x86-64 instruction set support. (This supersets
8588 MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and 64-bit instruction set extensions.)
8590 IDT Winchip C6 CPU, dealt in same way as i486 with additional MMX instruction
8593 IDT Winchip2 CPU, dealt in same way as i486 with additional MMX and 3dNOW!
8594 instruction set support.
8596 Via C3 CPU with MMX and 3dNOW! instruction set support. (No scheduling is
8597 implemented for this chip.)
8599 Via C3-2 CPU with MMX and SSE instruction set support. (No scheduling is
8600 implemented for this chip.)
8603 While picking a specific @var{cpu-type} will schedule things appropriately
8604 for that particular chip, the compiler will not generate any code that
8605 does not run on the i386 without the @option{-march=@var{cpu-type}} option
8608 @item -march=@var{cpu-type}
8610 Generate instructions for the machine type @var{cpu-type}. The choices
8611 for @var{cpu-type} are the same as for @option{-mtune}. Moreover,
8612 specifying @option{-march=@var{cpu-type}} implies @option{-mtune=@var{cpu-type}}.
8614 @item -mcpu=@var{cpu-type}
8616 A deprecated synonym for @option{-mtune}.
8625 @opindex mpentiumpro
8626 These options are synonyms for @option{-mtune=i386}, @option{-mtune=i486},
8627 @option{-mtune=pentium}, and @option{-mtune=pentiumpro} respectively.
8628 These synonyms are deprecated.
8630 @item -mfpmath=@var{unit}
8632 Generate floating point arithmetics for selected unit @var{unit}. The choices
8637 Use the standard 387 floating point coprocessor present majority of chips and
8638 emulated otherwise. Code compiled with this option will run almost everywhere.
8639 The temporary results are computed in 80bit precision instead of precision
8640 specified by the type resulting in slightly different results compared to most
8641 of other chips. See @option{-ffloat-store} for more detailed description.
8643 This is the default choice for i386 compiler.
8646 Use scalar floating point instructions present in the SSE instruction set.
8647 This instruction set is supported by Pentium3 and newer chips, in the AMD line
8648 by Athlon-4, Athlon-xp and Athlon-mp chips. The earlier version of SSE
8649 instruction set supports only single precision arithmetics, thus the double and
8650 extended precision arithmetics is still done using 387. Later version, present
8651 only in Pentium4 and the future AMD x86-64 chips supports double precision
8654 For the i386 compiler, you need to use @option{-march=@var{cpu-type}}, @option{-msse}
8655 or @option{-msse2} switches to enable SSE extensions and make this option
8656 effective. For the x86-64 compiler, these extensions are enabled by default.
8658 The resulting code should be considerably faster in the majority of cases and avoid
8659 the numerical instability problems of 387 code, but may break some existing
8660 code that expects temporaries to be 80bit.
8662 This is the default choice for the x86-64 compiler.
8665 Attempt to utilize both instruction sets at once. This effectively double the
8666 amount of available registers and on chips with separate execution units for
8667 387 and SSE the execution resources too. Use this option with care, as it is
8668 still experimental, because the GCC register allocator does not model separate
8669 functional units well resulting in instable performance.
8672 @item -masm=@var{dialect}
8673 @opindex masm=@var{dialect}
8674 Output asm instructions using selected @var{dialect}. Supported choices are
8675 @samp{intel} or @samp{att} (the default one).
8680 @opindex mno-ieee-fp
8681 Control whether or not the compiler uses IEEE floating point
8682 comparisons. These handle correctly the case where the result of a
8683 comparison is unordered.
8686 @opindex msoft-float
8687 Generate output containing library calls for floating point.
8688 @strong{Warning:} the requisite libraries are not part of GCC@.
8689 Normally the facilities of the machine's usual C compiler are used, but
8690 this can't be done directly in cross-compilation. You must make your
8691 own arrangements to provide suitable library functions for
8694 On machines where a function returns floating point results in the 80387
8695 register stack, some floating point opcodes may be emitted even if
8696 @option{-msoft-float} is used.
8698 @item -mno-fp-ret-in-387
8699 @opindex mno-fp-ret-in-387
8700 Do not use the FPU registers for return values of functions.
8702 The usual calling convention has functions return values of types
8703 @code{float} and @code{double} in an FPU register, even if there
8704 is no FPU@. The idea is that the operating system should emulate
8707 The option @option{-mno-fp-ret-in-387} causes such values to be returned
8708 in ordinary CPU registers instead.
8710 @item -mno-fancy-math-387
8711 @opindex mno-fancy-math-387
8712 Some 387 emulators do not support the @code{sin}, @code{cos} and
8713 @code{sqrt} instructions for the 387. Specify this option to avoid
8714 generating those instructions. This option is the default on FreeBSD,
8715 OpenBSD and NetBSD@. This option is overridden when @option{-march}
8716 indicates that the target cpu will always have an FPU and so the
8717 instruction will not need emulation. As of revision 2.6.1, these
8718 instructions are not generated unless you also use the
8719 @option{-funsafe-math-optimizations} switch.
8721 @item -malign-double
8722 @itemx -mno-align-double
8723 @opindex malign-double
8724 @opindex mno-align-double
8725 Control whether GCC aligns @code{double}, @code{long double}, and
8726 @code{long long} variables on a two word boundary or a one word
8727 boundary. Aligning @code{double} variables on a two word boundary will
8728 produce code that runs somewhat faster on a @samp{Pentium} at the
8729 expense of more memory.
8731 @strong{Warning:} if you use the @option{-malign-double} switch,
8732 structures containing the above types will be aligned differently than
8733 the published application binary interface specifications for the 386
8734 and will not be binary compatible with structures in code compiled
8735 without that switch.
8737 @item -m96bit-long-double
8738 @itemx -m128bit-long-double
8739 @opindex m96bit-long-double
8740 @opindex m128bit-long-double
8741 These switches control the size of @code{long double} type. The i386
8742 application binary interface specifies the size to be 96 bits,
8743 so @option{-m96bit-long-double} is the default in 32 bit mode.
8745 Modern architectures (Pentium and newer) would prefer @code{long double}
8746 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8747 conforming to the ABI, this would not be possible. So specifying a
8748 @option{-m128bit-long-double} will align @code{long double}
8749 to a 16 byte boundary by padding the @code{long double} with an additional
8752 In the x86-64 compiler, @option{-m128bit-long-double} is the default choice as
8753 its ABI specifies that @code{long double} is to be aligned on 16 byte boundary.
8755 Notice that neither of these options enable any extra precision over the x87
8756 standard of 80 bits for a @code{long double}.
8758 @strong{Warning:} if you override the default value for your target ABI, the
8759 structures and arrays containing @code{long double} variables will change
8760 their size as well as function calling convention for function taking
8761 @code{long double} will be modified. Hence they will not be binary
8762 compatible with arrays or structures in code compiled without that switch.
8766 @itemx -mno-svr3-shlib
8767 @opindex msvr3-shlib
8768 @opindex mno-svr3-shlib
8769 Control whether GCC places uninitialized local variables into the
8770 @code{bss} or @code{data} segments. @option{-msvr3-shlib} places them
8771 into @code{bss}. These options are meaningful only on System V Release 3.
8775 Use a different function-calling convention, in which functions that
8776 take a fixed number of arguments return with the @code{ret} @var{num}
8777 instruction, which pops their arguments while returning. This saves one
8778 instruction in the caller since there is no need to pop the arguments
8781 You can specify that an individual function is called with this calling
8782 sequence with the function attribute @samp{stdcall}. You can also
8783 override the @option{-mrtd} option by using the function attribute
8784 @samp{cdecl}. @xref{Function Attributes}.
8786 @strong{Warning:} this calling convention is incompatible with the one
8787 normally used on Unix, so you cannot use it if you need to call
8788 libraries compiled with the Unix compiler.
8790 Also, you must provide function prototypes for all functions that
8791 take variable numbers of arguments (including @code{printf});
8792 otherwise incorrect code will be generated for calls to those
8795 In addition, seriously incorrect code will result if you call a
8796 function with too many arguments. (Normally, extra arguments are
8797 harmlessly ignored.)
8799 @item -mregparm=@var{num}
8801 Control how many registers are used to pass integer arguments. By
8802 default, no registers are used to pass arguments, and at most 3
8803 registers can be used. You can control this behavior for a specific
8804 function by using the function attribute @samp{regparm}.
8805 @xref{Function Attributes}.
8807 @strong{Warning:} if you use this switch, and
8808 @var{num} is nonzero, then you must build all modules with the same
8809 value, including any libraries. This includes the system libraries and
8812 @item -mpreferred-stack-boundary=@var{num}
8813 @opindex mpreferred-stack-boundary
8814 Attempt to keep the stack boundary aligned to a 2 raised to @var{num}
8815 byte boundary. If @option{-mpreferred-stack-boundary} is not specified,
8816 the default is 4 (16 bytes or 128 bits), except when optimizing for code
8817 size (@option{-Os}), in which case the default is the minimum correct
8818 alignment (4 bytes for x86, and 8 bytes for x86-64).
8820 On Pentium and PentiumPro, @code{double} and @code{long double} values
8821 should be aligned to an 8 byte boundary (see @option{-malign-double}) or
8822 suffer significant run time performance penalties. On Pentium III, the
8823 Streaming SIMD Extension (SSE) data type @code{__m128} suffers similar
8824 penalties if it is not 16 byte aligned.
8826 To ensure proper alignment of this values on the stack, the stack boundary
8827 must be as aligned as that required by any value stored on the stack.
8828 Further, every function must be generated such that it keeps the stack
8829 aligned. Thus calling a function compiled with a higher preferred
8830 stack boundary from a function compiled with a lower preferred stack
8831 boundary will most likely misalign the stack. It is recommended that
8832 libraries that use callbacks always use the default setting.
8834 This extra alignment does consume extra stack space, and generally
8835 increases code size. Code that is sensitive to stack space usage, such
8836 as embedded systems and operating system kernels, may want to reduce the
8837 preferred alignment to @option{-mpreferred-stack-boundary=2}.
8855 These switches enable or disable the use of built-in functions that allow
8856 direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow extensions of the
8859 @xref{X86 Built-in Functions}, for details of the functions enabled
8860 and disabled by these switches.
8862 To have SSE/SSE2 instructions generated automatically from floating-point
8863 code, see @option{-mfpmath=sse}.
8866 @itemx -mno-push-args
8868 @opindex mno-push-args
8869 Use PUSH operations to store outgoing parameters. This method is shorter
8870 and usually equally fast as method using SUB/MOV operations and is enabled
8871 by default. In some cases disabling it may improve performance because of
8872 improved scheduling and reduced dependencies.
8874 @item -maccumulate-outgoing-args
8875 @opindex maccumulate-outgoing-args
8876 If enabled, the maximum amount of space required for outgoing arguments will be
8877 computed in the function prologue. This is faster on most modern CPUs
8878 because of reduced dependencies, improved scheduling and reduced stack usage
8879 when preferred stack boundary is not equal to 2. The drawback is a notable
8880 increase in code size. This switch implies @option{-mno-push-args}.
8884 Support thread-safe exception handling on @samp{Mingw32}. Code that relies
8885 on thread-safe exception handling must compile and link all code with the
8886 @option{-mthreads} option. When compiling, @option{-mthreads} defines
8887 @option{-D_MT}; when linking, it links in a special thread helper library
8888 @option{-lmingwthrd} which cleans up per thread exception handling data.
8890 @item -mno-align-stringops
8891 @opindex mno-align-stringops
8892 Do not align destination of inlined string operations. This switch reduces
8893 code size and improves performance in case the destination is already aligned,
8894 but GCC doesn't know about it.
8896 @item -minline-all-stringops
8897 @opindex minline-all-stringops
8898 By default GCC inlines string operations only when destination is known to be
8899 aligned at least to 4 byte boundary. This enables more inlining, increase code
8900 size, but may improve performance of code that depends on fast memcpy, strlen
8901 and memset for short lengths.
8903 @item -momit-leaf-frame-pointer
8904 @opindex momit-leaf-frame-pointer
8905 Don't keep the frame pointer in a register for leaf functions. This
8906 avoids the instructions to save, set up and restore frame pointers and
8907 makes an extra register available in leaf functions. The option
8908 @option{-fomit-frame-pointer} removes the frame pointer for all functions
8909 which might make debugging harder.
8911 @item -mtls-direct-seg-refs
8912 @itemx -mno-tls-direct-seg-refs
8913 @opindex mtls-direct-seg-refs
8914 Controls whether TLS variables may be accessed with offsets from the
8915 TLS segment register (@code{%gs} for 32-bit, @code{%fs} for 64-bit),
8916 or whether the thread base pointer must be added. Whether or not this
8917 is legal depends on the operating system, and whether it maps the
8918 segment to cover the entire TLS area.
8920 For systems that use GNU libc, the default is on.
8923 These @samp{-m} switches are supported in addition to the above
8924 on AMD x86-64 processors in 64-bit environments.
8931 Generate code for a 32-bit or 64-bit environment.
8932 The 32-bit environment sets int, long and pointer to 32 bits and
8933 generates code that runs on any i386 system.
8934 The 64-bit environment sets int to 32 bits and long and pointer
8935 to 64 bits and generates code for AMD's x86-64 architecture.
8938 @opindex no-red-zone
8939 Do not use a so called red zone for x86-64 code. The red zone is mandated
8940 by the x86-64 ABI, it is a 128-byte area beyond the location of the
8941 stack pointer that will not be modified by signal or interrupt handlers
8942 and therefore can be used for temporary data without adjusting the stack
8943 pointer. The flag @option{-mno-red-zone} disables this red zone.
8945 @item -mcmodel=small
8946 @opindex mcmodel=small
8947 Generate code for the small code model: the program and its symbols must
8948 be linked in the lower 2 GB of the address space. Pointers are 64 bits.
8949 Programs can be statically or dynamically linked. This is the default
8952 @item -mcmodel=kernel
8953 @opindex mcmodel=kernel
8954 Generate code for the kernel code model. The kernel runs in the
8955 negative 2 GB of the address space.
8956 This model has to be used for Linux kernel code.
8958 @item -mcmodel=medium
8959 @opindex mcmodel=medium
8960 Generate code for the medium model: The program is linked in the lower 2
8961 GB of the address space but symbols can be located anywhere in the
8962 address space. Programs can be statically or dynamically linked, but
8963 building of shared libraries are not supported with the medium model.
8965 @item -mcmodel=large
8966 @opindex mcmodel=large
8967 Generate code for the large model: This model makes no assumptions
8968 about addresses and sizes of sections. Currently GCC does not implement
8973 @subsection IA-64 Options
8974 @cindex IA-64 Options
8976 These are the @samp{-m} options defined for the Intel IA-64 architecture.
8980 @opindex mbig-endian
8981 Generate code for a big endian target. This is the default for HP-UX@.
8983 @item -mlittle-endian
8984 @opindex mlittle-endian
8985 Generate code for a little endian target. This is the default for AIX5
8992 Generate (or don't) code for the GNU assembler. This is the default.
8993 @c Also, this is the default if the configure option @option{--with-gnu-as}
9000 Generate (or don't) code for the GNU linker. This is the default.
9001 @c Also, this is the default if the configure option @option{--with-gnu-ld}
9006 Generate code that does not use a global pointer register. The result
9007 is not position independent code, and violates the IA-64 ABI@.
9009 @item -mvolatile-asm-stop
9010 @itemx -mno-volatile-asm-stop
9011 @opindex mvolatile-asm-stop
9012 @opindex mno-volatile-asm-stop
9013 Generate (or don't) a stop bit immediately before and after volatile asm
9016 @item -mregister-names
9017 @itemx -mno-register-names
9018 @opindex mregister-names
9019 @opindex mno-register-names
9020 Generate (or don't) @samp{in}, @samp{loc}, and @samp{out} register names for
9021 the stacked registers. This may make assembler output more readable.
9027 Disable (or enable) optimizations that use the small data section. This may
9028 be useful for working around optimizer bugs.
9031 @opindex mconstant-gp
9032 Generate code that uses a single constant global pointer value. This is
9033 useful when compiling kernel code.
9037 Generate code that is self-relocatable. This implies @option{-mconstant-gp}.
9038 This is useful when compiling firmware code.
9040 @item -minline-float-divide-min-latency
9041 @opindex minline-float-divide-min-latency
9042 Generate code for inline divides of floating point values
9043 using the minimum latency algorithm.
9045 @item -minline-float-divide-max-throughput
9046 @opindex minline-float-divide-max-throughput
9047 Generate code for inline divides of floating point values
9048 using the maximum throughput algorithm.
9050 @item -minline-int-divide-min-latency
9051 @opindex minline-int-divide-min-latency
9052 Generate code for inline divides of integer values
9053 using the minimum latency algorithm.
9055 @item -minline-int-divide-max-throughput
9056 @opindex minline-int-divide-max-throughput
9057 Generate code for inline divides of integer values
9058 using the maximum throughput algorithm.
9060 @item -minline-sqrt-min-latency
9061 @opindex minline-sqrt-min-latency
9062 Generate code for inline square roots
9063 using the minimum latency algorithm.
9065 @item -minline-sqrt-max-throughput
9066 @opindex minline-sqrt-max-throughput
9067 Generate code for inline square roots
9068 using the maximum throughput algorithm.
9070 @item -mno-dwarf2-asm
9072 @opindex mno-dwarf2-asm
9073 @opindex mdwarf2-asm
9074 Don't (or do) generate assembler code for the DWARF2 line number debugging
9075 info. This may be useful when not using the GNU assembler.
9077 @item -mearly-stop-bits
9078 @itemx -mno-early-stop-bits
9079 @opindex mearly-stop-bits
9080 @opindex mno-early-stop-bits
9081 Allow stop bits to be placed earlier than immediately preceding the
9082 instruction that triggered the stop bit. This can improve instruction
9083 scheduling, but does not always do so.
9085 @item -mfixed-range=@var{register-range}
9086 @opindex mfixed-range
9087 Generate code treating the given register range as fixed registers.
9088 A fixed register is one that the register allocator can not use. This is
9089 useful when compiling kernel code. A register range is specified as
9090 two registers separated by a dash. Multiple register ranges can be
9091 specified separated by a comma.
9093 @item -mtls-size=@var{tls-size}
9095 Specify bit size of immediate TLS offsets. Valid values are 14, 22, and
9098 @item -mtune-arch=@var{cpu-type}
9100 Tune the instruction scheduling for a particular CPU, Valid values are
9101 itanium, itanium1, merced, itanium2, and mckinley.
9107 Add support for multithreading using the POSIX threads library. This
9108 option sets flags for both the preprocessor and linker. It does
9109 not affect the thread safety of object code produced by the compiler or
9110 that of libraries supplied with it. These are HP-UX specific flags.
9116 Generate code for a 32-bit or 64-bit environment.
9117 The 32-bit environment sets int, long and pointer to 32 bits.
9118 The 64-bit environment sets int to 32 bits and long and pointer
9119 to 64 bits. These are HP-UX specific flags.
9123 @node M32R/D Options
9124 @subsection M32R/D Options
9125 @cindex M32R/D options
9127 These @option{-m} options are defined for Renesas M32R/D architectures:
9132 Generate code for the M32R/2@.
9136 Generate code for the M32R/X@.
9140 Generate code for the M32R@. This is the default.
9143 @opindex mmodel=small
9144 Assume all objects live in the lower 16MB of memory (so that their addresses
9145 can be loaded with the @code{ld24} instruction), and assume all subroutines
9146 are reachable with the @code{bl} instruction.
9147 This is the default.
9149 The addressability of a particular object can be set with the
9150 @code{model} attribute.
9152 @item -mmodel=medium
9153 @opindex mmodel=medium
9154 Assume objects may be anywhere in the 32-bit address space (the compiler
9155 will generate @code{seth/add3} instructions to load their addresses), and
9156 assume all subroutines are reachable with the @code{bl} instruction.
9159 @opindex mmodel=large
9160 Assume objects may be anywhere in the 32-bit address space (the compiler
9161 will generate @code{seth/add3} instructions to load their addresses), and
9162 assume subroutines may not be reachable with the @code{bl} instruction
9163 (the compiler will generate the much slower @code{seth/add3/jl}
9164 instruction sequence).
9167 @opindex msdata=none
9168 Disable use of the small data area. Variables will be put into
9169 one of @samp{.data}, @samp{bss}, or @samp{.rodata} (unless the
9170 @code{section} attribute has been specified).
9171 This is the default.
9173 The small data area consists of sections @samp{.sdata} and @samp{.sbss}.
9174 Objects may be explicitly put in the small data area with the
9175 @code{section} attribute using one of these sections.
9178 @opindex msdata=sdata
9179 Put small global and static data in the small data area, but do not
9180 generate special code to reference them.
9184 Put small global and static data in the small data area, and generate
9185 special instructions to reference them.
9189 @cindex smaller data references
9190 Put global and static objects less than or equal to @var{num} bytes
9191 into the small data or bss sections instead of the normal data or bss
9192 sections. The default value of @var{num} is 8.
9193 The @option{-msdata} option must be set to one of @samp{sdata} or @samp{use}
9194 for this option to have any effect.
9196 All modules should be compiled with the same @option{-G @var{num}} value.
9197 Compiling with different values of @var{num} may or may not work; if it
9198 doesn't the linker will give an error message---incorrect code will not be
9203 Makes the M32R specific code in the compiler display some statistics
9204 that might help in debugging programs.
9207 @opindex malign-loops
9208 Align all loops to a 32-byte boundary.
9210 @item -mno-align-loops
9211 @opindex mno-align-loops
9212 Do not enforce a 32-byte alignment for loops. This is the default.
9214 @item -missue-rate=@var{number}
9215 @opindex missue-rate=@var{number}
9216 Issue @var{number} instructions per cycle. @var{number} can only be 1
9219 @item -mbranch-cost=@var{number}
9220 @opindex mbranch-cost=@var{number}
9221 @var{number} can only be 1 or 2. If it is 1 then branches will be
9222 preferred over conditional code, if it is 2, then the opposite will
9225 @item -mflush-trap=@var{number}
9226 @opindex mflush-trap=@var{number}
9227 Specifies the trap number to use to flush the cache. The default is
9228 12. Valid numbers are between 0 and 15 inclusive.
9230 @item -mno-flush-trap
9231 @opindex mno-flush-trap
9232 Specifies that the cache cannot be flushed by using a trap.
9234 @item -mflush-func=@var{name}
9235 @opindex mflush-func=@var{name}
9236 Specifies the name of the operating system function to call to flush
9237 the cache. The default is @emph{_flush_cache}, but a function call
9238 will only be used if a trap is not available.
9240 @item -mno-flush-func
9241 @opindex mno-flush-func
9242 Indicates that there is no OS function for flushing the cache.
9246 @node M680x0 Options
9247 @subsection M680x0 Options
9248 @cindex M680x0 options
9250 These are the @samp{-m} options defined for the 68000 series. The default
9251 values for these options depends on which style of 68000 was selected when
9252 the compiler was configured; the defaults for the most common choices are
9260 Generate output for a 68000. This is the default
9261 when the compiler is configured for 68000-based systems.
9263 Use this option for microcontrollers with a 68000 or EC000 core,
9264 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9270 Generate output for a 68020. This is the default
9271 when the compiler is configured for 68020-based systems.
9275 Generate output containing 68881 instructions for floating point.
9276 This is the default for most 68020 systems unless @option{--nfp} was
9277 specified when the compiler was configured.
9281 Generate output for a 68030. This is the default when the compiler is
9282 configured for 68030-based systems.
9286 Generate output for a 68040. This is the default when the compiler is
9287 configured for 68040-based systems.
9289 This option inhibits the use of 68881/68882 instructions that have to be
9290 emulated by software on the 68040. Use this option if your 68040 does not
9291 have code to emulate those instructions.
9295 Generate output for a 68060. This is the default when the compiler is
9296 configured for 68060-based systems.
9298 This option inhibits the use of 68020 and 68881/68882 instructions that
9299 have to be emulated by software on the 68060. Use this option if your 68060
9300 does not have code to emulate those instructions.
9304 Generate output for a CPU32. This is the default
9305 when the compiler is configured for CPU32-based systems.
9307 Use this option for microcontrollers with a
9308 CPU32 or CPU32+ core, including the 68330, 68331, 68332, 68333, 68334,
9309 68336, 68340, 68341, 68349 and 68360.
9313 Generate output for a 520X ``coldfire'' family cpu. This is the default
9314 when the compiler is configured for 520X-based systems.
9316 Use this option for microcontroller with a 5200 core, including
9317 the MCF5202, MCF5203, MCF5204 and MCF5202.
9322 Generate output for a 68040, without using any of the new instructions.
9323 This results in code which can run relatively efficiently on either a
9324 68020/68881 or a 68030 or a 68040. The generated code does use the
9325 68881 instructions that are emulated on the 68040.
9329 Generate output for a 68060, without using any of the new instructions.
9330 This results in code which can run relatively efficiently on either a
9331 68020/68881 or a 68030 or a 68040. The generated code does use the
9332 68881 instructions that are emulated on the 68060.
9335 @opindex msoft-float
9336 Generate output containing library calls for floating point.
9337 @strong{Warning:} the requisite libraries are not available for all m68k
9338 targets. Normally the facilities of the machine's usual C compiler are
9339 used, but this can't be done directly in cross-compilation. You must
9340 make your own arrangements to provide suitable library functions for
9341 cross-compilation. The embedded targets @samp{m68k-*-aout} and
9342 @samp{m68k-*-coff} do provide software floating point support.
9346 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9347 Additionally, parameters passed on the stack are also aligned to a
9348 16-bit boundary even on targets whose API mandates promotion to 32-bit.
9351 @opindex mnobitfield
9352 Do not use the bit-field instructions. The @option{-m68000}, @option{-mcpu32}
9353 and @option{-m5200} options imply @w{@option{-mnobitfield}}.
9357 Do use the bit-field instructions. The @option{-m68020} option implies
9358 @option{-mbitfield}. This is the default if you use a configuration
9359 designed for a 68020.
9363 Use a different function-calling convention, in which functions
9364 that take a fixed number of arguments return with the @code{rtd}
9365 instruction, which pops their arguments while returning. This
9366 saves one instruction in the caller since there is no need to pop
9367 the arguments there.
9369 This calling convention is incompatible with the one normally
9370 used on Unix, so you cannot use it if you need to call libraries
9371 compiled with the Unix compiler.
9373 Also, you must provide function prototypes for all functions that
9374 take variable numbers of arguments (including @code{printf});
9375 otherwise incorrect code will be generated for calls to those
9378 In addition, seriously incorrect code will result if you call a
9379 function with too many arguments. (Normally, extra arguments are
9380 harmlessly ignored.)
9382 The @code{rtd} instruction is supported by the 68010, 68020, 68030,
9383 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9386 @itemx -mno-align-int
9388 @opindex mno-align-int
9389 Control whether GCC aligns @code{int}, @code{long}, @code{long long},
9390 @code{float}, @code{double}, and @code{long double} variables on a 32-bit
9391 boundary (@option{-malign-int}) or a 16-bit boundary (@option{-mno-align-int}).
9392 Aligning variables on 32-bit boundaries produces code that runs somewhat
9393 faster on processors with 32-bit busses at the expense of more memory.
9395 @strong{Warning:} if you use the @option{-malign-int} switch, GCC will
9396 align structures containing the above types differently than
9397 most published application binary interface specifications for the m68k.
9401 Use the pc-relative addressing mode of the 68000 directly, instead of
9402 using a global offset table. At present, this option implies @option{-fpic},
9403 allowing at most a 16-bit offset for pc-relative addressing. @option{-fPIC} is
9404 not presently supported with @option{-mpcrel}, though this could be supported for
9405 68020 and higher processors.
9407 @item -mno-strict-align
9408 @itemx -mstrict-align
9409 @opindex mno-strict-align
9410 @opindex mstrict-align
9411 Do not (do) assume that unaligned memory references will be handled by
9415 Generate code that allows the data segment to be located in a different
9416 area of memory from the text segment. This allows for execute in place in
9417 an environment without virtual memory management. This option implies
9421 Generate code that assumes that the data segment follows the text segment.
9422 This is the default.
9424 @item -mid-shared-library
9425 Generate code that supports shared libraries via the library ID method.
9426 This allows for execute in place and shared libraries in an environment
9427 without virtual memory management. This option implies @option{-fPIC}.
9429 @item -mno-id-shared-library
9430 Generate code that doesn't assume ID based shared libraries are being used.
9431 This is the default.
9433 @item -mshared-library-id=n
9434 Specified the identification number of the ID based shared library being
9435 compiled. Specifying a value of 0 will generate more compact code, specifying
9436 other values will force the allocation of that number to the current
9437 library but is no more space or time efficient than omitting this option.
9441 @node M68hc1x Options
9442 @subsection M68hc1x Options
9443 @cindex M68hc1x options
9445 These are the @samp{-m} options defined for the 68hc11 and 68hc12
9446 microcontrollers. The default values for these options depends on
9447 which style of microcontroller was selected when the compiler was configured;
9448 the defaults for the most common choices are given below.
9455 Generate output for a 68HC11. This is the default
9456 when the compiler is configured for 68HC11-based systems.
9462 Generate output for a 68HC12. This is the default
9463 when the compiler is configured for 68HC12-based systems.
9469 Generate output for a 68HCS12.
9472 @opindex mauto-incdec
9473 Enable the use of 68HC12 pre and post auto-increment and auto-decrement
9480 Enable the use of 68HC12 min and max instructions.
9483 @itemx -mno-long-calls
9484 @opindex mlong-calls
9485 @opindex mno-long-calls
9486 Treat all calls as being far away (near). If calls are assumed to be
9487 far away, the compiler will use the @code{call} instruction to
9488 call a function and the @code{rtc} instruction for returning.
9492 Consider type @code{int} to be 16 bits wide, like @code{short int}.
9494 @item -msoft-reg-count=@var{count}
9495 @opindex msoft-reg-count
9496 Specify the number of pseudo-soft registers which are used for the
9497 code generation. The maximum number is 32. Using more pseudo-soft
9498 register may or may not result in better code depending on the program.
9499 The default is 4 for 68HC11 and 2 for 68HC12.
9504 @subsection MCore Options
9505 @cindex MCore options
9507 These are the @samp{-m} options defined for the Motorola M*Core
9515 @opindex mno-hardlit
9516 Inline constants into the code stream if it can be done in two
9517 instructions or less.
9523 Use the divide instruction. (Enabled by default).
9525 @item -mrelax-immediate
9526 @itemx -mno-relax-immediate
9527 @opindex mrelax-immediate
9528 @opindex mno-relax-immediate
9529 Allow arbitrary sized immediates in bit operations.
9531 @item -mwide-bitfields
9532 @itemx -mno-wide-bitfields
9533 @opindex mwide-bitfields
9534 @opindex mno-wide-bitfields
9535 Always treat bit-fields as int-sized.
9537 @item -m4byte-functions
9538 @itemx -mno-4byte-functions
9539 @opindex m4byte-functions
9540 @opindex mno-4byte-functions
9541 Force all functions to be aligned to a four byte boundary.
9543 @item -mcallgraph-data
9544 @itemx -mno-callgraph-data
9545 @opindex mcallgraph-data
9546 @opindex mno-callgraph-data
9547 Emit callgraph information.
9550 @itemx -mno-slow-bytes
9551 @opindex mslow-bytes
9552 @opindex mno-slow-bytes
9553 Prefer word access when reading byte quantities.
9555 @item -mlittle-endian
9557 @opindex mlittle-endian
9558 @opindex mbig-endian
9559 Generate code for a little endian target.
9565 Generate code for the 210 processor.
9569 @subsection MIPS Options
9570 @cindex MIPS options
9576 Generate big-endian code.
9580 Generate little-endian code. This is the default for @samp{mips*el-*-*}
9583 @item -march=@var{arch}
9585 Generate code that will run on @var{arch}, which can be the name of a
9586 generic MIPS ISA, or the name of a particular processor.
9588 @samp{mips1}, @samp{mips2}, @samp{mips3}, @samp{mips4},
9589 @samp{mips32}, @samp{mips32r2}, and @samp{mips64}.
9590 The processor names are:
9591 @samp{4kc}, @samp{4kp}, @samp{5kc}, @samp{20kc},
9593 @samp{r2000}, @samp{r3000}, @samp{r3900}, @samp{r4000}, @samp{r4400},
9594 @samp{r4600}, @samp{r4650}, @samp{r6000}, @samp{r8000}, @samp{rm7000},
9598 @samp{vr4100}, @samp{vr4111}, @samp{vr4120}, @samp{vr4130}, @samp{vr4300},
9599 @samp{vr5000}, @samp{vr5400} and @samp{vr5500}.
9600 The special value @samp{from-abi} selects the
9601 most compatible architecture for the selected ABI (that is,
9602 @samp{mips1} for 32-bit ABIs and @samp{mips3} for 64-bit ABIs)@.
9604 In processor names, a final @samp{000} can be abbreviated as @samp{k}
9605 (for example, @samp{-march=r2k}). Prefixes are optional, and
9606 @samp{vr} may be written @samp{r}.
9608 GCC defines two macros based on the value of this option. The first
9609 is @samp{_MIPS_ARCH}, which gives the name of target architecture, as
9610 a string. The second has the form @samp{_MIPS_ARCH_@var{foo}},
9611 where @var{foo} is the capitalized value of @samp{_MIPS_ARCH}@.
9612 For example, @samp{-march=r2000} will set @samp{_MIPS_ARCH}
9613 to @samp{"r2000"} and define the macro @samp{_MIPS_ARCH_R2000}.
9615 Note that the @samp{_MIPS_ARCH} macro uses the processor names given
9616 above. In other words, it will have the full prefix and will not
9617 abbreviate @samp{000} as @samp{k}. In the case of @samp{from-abi},
9618 the macro names the resolved architecture (either @samp{"mips1"} or
9619 @samp{"mips3"}). It names the default architecture when no
9620 @option{-march} option is given.
9622 @item -mtune=@var{arch}
9624 Optimize for @var{arch}. Among other things, this option controls
9625 the way instructions are scheduled, and the perceived cost of arithmetic
9626 operations. The list of @var{arch} values is the same as for
9629 When this option is not used, GCC will optimize for the processor
9630 specified by @option{-march}. By using @option{-march} and
9631 @option{-mtune} together, it is possible to generate code that will
9632 run on a family of processors, but optimize the code for one
9633 particular member of that family.
9635 @samp{-mtune} defines the macros @samp{_MIPS_TUNE} and
9636 @samp{_MIPS_TUNE_@var{foo}}, which work in the same way as the
9637 @samp{-march} ones described above.
9641 Equivalent to @samp{-march=mips1}.
9645 Equivalent to @samp{-march=mips2}.
9649 Equivalent to @samp{-march=mips3}.
9653 Equivalent to @samp{-march=mips4}.
9657 Equivalent to @samp{-march=mips32}.
9661 Equivalent to @samp{-march=mips32r2}.
9665 Equivalent to @samp{-march=mips64}.
9671 Use (do not use) the MIPS16 ISA@.
9683 Generate code for the given ABI@.
9685 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9686 generates 64-bit code when you select a 64-bit architecture, but you
9687 can use @option{-mgp32} to get 32-bit code instead.
9689 For information about the O64 ABI, see
9690 @w{@uref{http://gcc.gnu.org/projects/mipso64-abi.html}}.
9693 @itemx -mno-abicalls
9695 @opindex mno-abicalls
9696 Generate (do not generate) SVR4-style position-independent code.
9697 @option{-mabicalls} is the default for SVR4-based systems.
9703 Lift (do not lift) the usual restrictions on the size of the global
9706 GCC normally uses a single instruction to load values from the GOT@.
9707 While this is relatively efficient, it will only work if the GOT
9708 is smaller than about 64k. Anything larger will cause the linker
9709 to report an error such as:
9711 @cindex relocation truncated to fit (MIPS)
9713 relocation truncated to fit: R_MIPS_GOT16 foobar
9716 If this happens, you should recompile your code with @option{-mxgot}.
9717 It should then work with very large GOTs, although it will also be
9718 less efficient, since it will take three instructions to fetch the
9719 value of a global symbol.
9721 Note that some linkers can create multiple GOTs. If you have such a
9722 linker, you should only need to use @option{-mxgot} when a single object
9723 file accesses more than 64k's worth of GOT entries. Very few do.
9725 These options have no effect unless GCC is generating position
9730 Assume that general-purpose registers are 32 bits wide.
9734 Assume that general-purpose registers are 64 bits wide.
9738 Assume that floating-point registers are 32 bits wide.
9742 Assume that floating-point registers are 64 bits wide.
9745 @opindex mhard-float
9746 Use floating-point coprocessor instructions.
9749 @opindex msoft-float
9750 Do not use floating-point coprocessor instructions. Implement
9751 floating-point calculations using library calls instead.
9753 @item -msingle-float
9754 @opindex msingle-float
9755 Assume that the floating-point coprocessor only supports single-precision
9758 @itemx -mdouble-float
9759 @opindex mdouble-float
9760 Assume that the floating-point coprocessor supports double-precision
9761 operations. This is the default.
9763 @itemx -mpaired-single
9764 @itemx -mno-paired-single
9765 @opindex mpaired-single
9766 @opindex mno-paired-single
9767 Use (do not use) paired-single floating-point instructions.
9768 @xref{MIPS Paired-Single Support}. This option can only be used
9769 when generating 64-bit code and requires hardware floating-point
9770 support to be enabled.
9776 Use (do not use) the MIPS-3D ASE@. @xref{MIPS-3D Built-in Functions}.
9777 The option @option{-mips3d} implies @option{-mpaired-single}.
9781 Force @code{int} and @code{long} types to be 64 bits wide. See
9782 @option{-mlong32} for an explanation of the default and the way
9783 that the pointer size is determined.
9785 This option has been deprecated and will be removed in a future release.
9789 Force @code{long} types to be 64 bits wide. See @option{-mlong32} for
9790 an explanation of the default and the way that the pointer size is
9795 Force @code{long}, @code{int}, and pointer types to be 32 bits wide.
9797 The default size of @code{int}s, @code{long}s and pointers depends on
9798 the ABI@. All the supported ABIs use 32-bit @code{int}s. The n64 ABI
9799 uses 64-bit @code{long}s, as does the 64-bit EABI; the others use
9800 32-bit @code{long}s. Pointers are the same size as @code{long}s,
9801 or the same size as integer registers, whichever is smaller.
9807 Assume (do not assume) that all symbols have 32-bit values, regardless
9808 of the selected ABI@. This option is useful in combination with
9809 @option{-mabi=64} and @option{-mno-abicalls} because it allows GCC
9810 to generate shorter and faster references to symbolic addresses.
9814 @cindex smaller data references (MIPS)
9815 @cindex gp-relative references (MIPS)
9816 Put global and static items less than or equal to @var{num} bytes into
9817 the small data or bss section instead of the normal data or bss section.
9818 This allows the data to be accessed using a single instruction.
9820 All modules should be compiled with the same @option{-G @var{num}}
9823 @item -membedded-data
9824 @itemx -mno-embedded-data
9825 @opindex membedded-data
9826 @opindex mno-embedded-data
9827 Allocate variables to the read-only data section first if possible, then
9828 next in the small data section if possible, otherwise in data. This gives
9829 slightly slower code than the default, but reduces the amount of RAM required
9830 when executing, and thus may be preferred for some embedded systems.
9832 @item -muninit-const-in-rodata
9833 @itemx -mno-uninit-const-in-rodata
9834 @opindex muninit-const-in-rodata
9835 @opindex mno-uninit-const-in-rodata
9836 Put uninitialized @code{const} variables in the read-only data section.
9837 This option is only meaningful in conjunction with @option{-membedded-data}.
9839 @item -msplit-addresses
9840 @itemx -mno-split-addresses
9841 @opindex msplit-addresses
9842 @opindex mno-split-addresses
9843 Enable (disable) use of the @code{%hi()} and @code{%lo()} assembler
9844 relocation operators. This option has been superseded by
9845 @option{-mexplicit-relocs} but is retained for backwards compatibility.
9847 @item -mexplicit-relocs
9848 @itemx -mno-explicit-relocs
9849 @opindex mexplicit-relocs
9850 @opindex mno-explicit-relocs
9851 Use (do not use) assembler relocation operators when dealing with symbolic
9852 addresses. The alternative, selected by @option{-mno-explicit-relocs},
9853 is to use assembler macros instead.
9855 @option{-mexplicit-relocs} is the default if GCC was configured
9856 to use an assembler that supports relocation operators.
9858 @item -mcheck-zero-division
9859 @itemx -mno-check-zero-division
9860 @opindex mcheck-zero-division
9861 @opindex mno-check-zero-division
9862 Trap (do not trap) on integer division by zero. The default is
9863 @option{-mcheck-zero-division}.
9865 @item -mdivide-traps
9866 @itemx -mdivide-breaks
9867 @opindex mdivide-traps
9868 @opindex mdivide-breaks
9869 MIPS systems check for division by zero by generating either a
9870 conditional trap or a break instruction. Using traps results in
9871 smaller code, but is only supported on MIPS II and later. Also, some
9872 versions of the Linux kernel have a bug that prevents trap from
9873 generating the proper signal (@code{SIGFPE}). Use @option{-mdivide-traps} to
9874 allow conditional traps on architectures that support them and
9875 @option{-mdivide-breaks} to force the use of breaks.
9877 The default is usually @option{-mdivide-traps}, but this can be
9878 overridden at configure time using @option{--with-divide=breaks}.
9879 Divide-by-zero checks can be completely disabled using
9880 @option{-mno-check-zero-division}.
9886 Force (do not force) the use of @code{memcpy()} for non-trivial block
9887 moves. The default is @option{-mno-memcpy}, which allows GCC to inline
9888 most constant-sized copies.
9891 @itemx -mno-long-calls
9892 @opindex mlong-calls
9893 @opindex mno-long-calls
9894 Disable (do not disable) use of the @code{jal} instruction. Calling
9895 functions using @code{jal} is more efficient but requires the caller
9896 and callee to be in the same 256 megabyte segment.
9898 This option has no effect on abicalls code. The default is
9899 @option{-mno-long-calls}.
9905 Enable (disable) use of the @code{mad}, @code{madu} and @code{mul}
9906 instructions, as provided by the R4650 ISA@.
9909 @itemx -mno-fused-madd
9910 @opindex mfused-madd
9911 @opindex mno-fused-madd
9912 Enable (disable) use of the floating point multiply-accumulate
9913 instructions, when they are available. The default is
9914 @option{-mfused-madd}.
9916 When multiply-accumulate instructions are used, the intermediate
9917 product is calculated to infinite precision and is not subject to
9918 the FCSR Flush to Zero bit. This may be undesirable in some
9923 Tell the MIPS assembler to not run its preprocessor over user
9924 assembler files (with a @samp{.s} suffix) when assembling them.
9927 @itemx -mno-fix-r4000
9929 @opindex mno-fix-r4000
9930 Work around certain R4000 CPU errata:
9933 A double-word or a variable shift may give an incorrect result if executed
9934 immediately after starting an integer division.
9936 A double-word or a variable shift may give an incorrect result if executed
9937 while an integer multiplication is in progress.
9939 An integer division may give an incorrect result if started in a delay slot
9940 of a taken branch or a jump.
9944 @itemx -mno-fix-r4400
9946 @opindex mno-fix-r4400
9947 Work around certain R4400 CPU errata:
9950 A double-word or a variable shift may give an incorrect result if executed
9951 immediately after starting an integer division.
9955 @itemx -mno-fix-vr4120
9956 @opindex mfix-vr4120
9957 Work around certain VR4120 errata:
9960 @code{dmultu} does not always produce the correct result.
9962 @code{div} and @code{ddiv} do not always produce the correct result if one
9963 of the operands is negative.
9965 The workarounds for the division errata rely on special functions in
9966 @file{libgcc.a}. At present, these functions are only provided by
9967 the @code{mips64vr*-elf} configurations.
9969 Other VR4120 errata require a nop to be inserted between certain pairs of
9970 instructions. These errata are handled by the assembler, not by GCC itself.
9973 @opindex mfix-vr4130
9974 Work around the VR4130 @code{mflo}/@code{mfhi} errata. The
9975 workarounds are implemented by the assembler rather than by GCC,
9976 although GCC will avoid using @code{mflo} and @code{mfhi} if the
9977 VR4130 @code{macc}, @code{macchi}, @code{dmacc} and @code{dmacchi}
9978 instructions are available instead.
9983 Work around certain SB-1 CPU core errata.
9984 (This flag currently works around the SB-1 revision 2
9985 ``F1'' and ``F2'' floating point errata.)
9987 @item -mflush-func=@var{func}
9988 @itemx -mno-flush-func
9989 @opindex mflush-func
9990 Specifies the function to call to flush the I and D caches, or to not
9991 call any such function. If called, the function must take the same
9992 arguments as the common @code{_flush_func()}, that is, the address of the
9993 memory range for which the cache is being flushed, the size of the
9994 memory range, and the number 3 (to flush both caches). The default
9995 depends on the target GCC was configured for, but commonly is either
9996 @samp{_flush_func} or @samp{__cpu_flush}.
9998 @item -mbranch-likely
9999 @itemx -mno-branch-likely
10000 @opindex mbranch-likely
10001 @opindex mno-branch-likely
10002 Enable or disable use of Branch Likely instructions, regardless of the
10003 default for the selected architecture. By default, Branch Likely
10004 instructions may be generated if they are supported by the selected
10005 architecture. An exception is for the MIPS32 and MIPS64 architectures
10006 and processors which implement those architectures; for those, Branch
10007 Likely instructions will not be generated by default because the MIPS32
10008 and MIPS64 architectures specifically deprecate their use.
10010 @item -mfp-exceptions
10011 @itemx -mno-fp-exceptions
10012 @opindex mfp-exceptions
10013 Specifies whether FP exceptions are enabled. This affects how we schedule
10014 FP instructions for some processors. The default is that FP exceptions are
10017 For instance, on the SB-1, if FP exceptions are disabled, and we are emitting
10018 64-bit code, then we can use both FP pipes. Otherwise, we can only use one
10021 @item -mvr4130-align
10022 @itemx -mno-vr4130-align
10023 @opindex mvr4130-align
10024 The VR4130 pipeline is two-way superscalar, but can only issue two
10025 instructions together if the first one is 8-byte aligned. When this
10026 option is enabled, GCC will align pairs of instructions that it
10027 thinks should execute in parallel.
10029 This option only has an effect when optimizing for the VR4130.
10030 It normally makes code faster, but at the expense of making it bigger.
10031 It is enabled by default at optimization level @option{-O3}.
10035 @subsection MMIX Options
10036 @cindex MMIX Options
10038 These options are defined for the MMIX:
10042 @itemx -mno-libfuncs
10044 @opindex mno-libfuncs
10045 Specify that intrinsic library functions are being compiled, passing all
10046 values in registers, no matter the size.
10049 @itemx -mno-epsilon
10051 @opindex mno-epsilon
10052 Generate floating-point comparison instructions that compare with respect
10053 to the @code{rE} epsilon register.
10055 @item -mabi=mmixware
10057 @opindex mabi-mmixware
10059 Generate code that passes function parameters and return values that (in
10060 the called function) are seen as registers @code{$0} and up, as opposed to
10061 the GNU ABI which uses global registers @code{$231} and up.
10063 @item -mzero-extend
10064 @itemx -mno-zero-extend
10065 @opindex mzero-extend
10066 @opindex mno-zero-extend
10067 When reading data from memory in sizes shorter than 64 bits, use (do not
10068 use) zero-extending load instructions by default, rather than
10069 sign-extending ones.
10072 @itemx -mno-knuthdiv
10074 @opindex mno-knuthdiv
10075 Make the result of a division yielding a remainder have the same sign as
10076 the divisor. With the default, @option{-mno-knuthdiv}, the sign of the
10077 remainder follows the sign of the dividend. Both methods are
10078 arithmetically valid, the latter being almost exclusively used.
10080 @item -mtoplevel-symbols
10081 @itemx -mno-toplevel-symbols
10082 @opindex mtoplevel-symbols
10083 @opindex mno-toplevel-symbols
10084 Prepend (do not prepend) a @samp{:} to all global symbols, so the assembly
10085 code can be used with the @code{PREFIX} assembly directive.
10089 Generate an executable in the ELF format, rather than the default
10090 @samp{mmo} format used by the @command{mmix} simulator.
10092 @item -mbranch-predict
10093 @itemx -mno-branch-predict
10094 @opindex mbranch-predict
10095 @opindex mno-branch-predict
10096 Use (do not use) the probable-branch instructions, when static branch
10097 prediction indicates a probable branch.
10099 @item -mbase-addresses
10100 @itemx -mno-base-addresses
10101 @opindex mbase-addresses
10102 @opindex mno-base-addresses
10103 Generate (do not generate) code that uses @emph{base addresses}. Using a
10104 base address automatically generates a request (handled by the assembler
10105 and the linker) for a constant to be set up in a global register. The
10106 register is used for one or more base address requests within the range 0
10107 to 255 from the value held in the register. The generally leads to short
10108 and fast code, but the number of different data items that can be
10109 addressed is limited. This means that a program that uses lots of static
10110 data may require @option{-mno-base-addresses}.
10112 @item -msingle-exit
10113 @itemx -mno-single-exit
10114 @opindex msingle-exit
10115 @opindex mno-single-exit
10116 Force (do not force) generated code to have a single exit point in each
10120 @node MN10300 Options
10121 @subsection MN10300 Options
10122 @cindex MN10300 options
10124 These @option{-m} options are defined for Matsushita MN10300 architectures:
10129 Generate code to avoid bugs in the multiply instructions for the MN10300
10130 processors. This is the default.
10132 @item -mno-mult-bug
10133 @opindex mno-mult-bug
10134 Do not generate code to avoid bugs in the multiply instructions for the
10135 MN10300 processors.
10139 Generate code which uses features specific to the AM33 processor.
10143 Do not generate code which uses features specific to the AM33 processor. This
10148 Do not link in the C run-time initialization object file.
10152 Indicate to the linker that it should perform a relaxation optimization pass
10153 to shorten branches, calls and absolute memory addresses. This option only
10154 has an effect when used on the command line for the final link step.
10156 This option makes symbolic debugging impossible.
10159 @node NS32K Options
10160 @subsection NS32K Options
10161 @cindex NS32K options
10163 These are the @samp{-m} options defined for the 32000 series. The default
10164 values for these options depends on which style of 32000 was selected when
10165 the compiler was configured; the defaults for the most common choices are
10173 Generate output for a 32032. This is the default
10174 when the compiler is configured for 32032 and 32016 based systems.
10180 Generate output for a 32332. This is the default
10181 when the compiler is configured for 32332-based systems.
10187 Generate output for a 32532. This is the default
10188 when the compiler is configured for 32532-based systems.
10192 Generate output containing 32081 instructions for floating point.
10193 This is the default for all systems.
10197 Generate output containing 32381 instructions for floating point. This
10198 also implies @option{-m32081}. The 32381 is only compatible with the 32332
10199 and 32532 cpus. This is the default for the pc532-netbsd configuration.
10202 @opindex mmulti-add
10203 Try and generate multiply-add floating point instructions @code{polyF}
10204 and @code{dotF}. This option is only available if the @option{-m32381}
10205 option is in effect. Using these instructions requires changes to
10206 register allocation which generally has a negative impact on
10207 performance. This option should only be enabled when compiling code
10208 particularly likely to make heavy use of multiply-add instructions.
10210 @item -mnomulti-add
10211 @opindex mnomulti-add
10212 Do not try and generate multiply-add floating point instructions
10213 @code{polyF} and @code{dotF}. This is the default on all platforms.
10216 @opindex msoft-float
10217 Generate output containing library calls for floating point.
10218 @strong{Warning:} the requisite libraries may not be available.
10220 @item -mieee-compare
10221 @itemx -mno-ieee-compare
10222 @opindex mieee-compare
10223 @opindex mno-ieee-compare
10224 Control whether or not the compiler uses IEEE floating point
10225 comparisons. These handle correctly the case where the result of a
10226 comparison is unordered.
10227 @strong{Warning:} the requisite kernel support may not be available.
10230 @opindex mnobitfield
10231 Do not use the bit-field instructions. On some machines it is faster to
10232 use shifting and masking operations. This is the default for the pc532.
10236 Do use the bit-field instructions. This is the default for all platforms
10241 Use a different function-calling convention, in which functions
10242 that take a fixed number of arguments return pop their
10243 arguments on return with the @code{ret} instruction.
10245 This calling convention is incompatible with the one normally
10246 used on Unix, so you cannot use it if you need to call libraries
10247 compiled with the Unix compiler.
10249 Also, you must provide function prototypes for all functions that
10250 take variable numbers of arguments (including @code{printf});
10251 otherwise incorrect code will be generated for calls to those
10254 In addition, seriously incorrect code will result if you call a
10255 function with too many arguments. (Normally, extra arguments are
10256 harmlessly ignored.)
10258 This option takes its name from the 680x0 @code{rtd} instruction.
10263 Use a different function-calling convention where the first two arguments
10264 are passed in registers.
10266 This calling convention is incompatible with the one normally
10267 used on Unix, so you cannot use it if you need to call libraries
10268 compiled with the Unix compiler.
10271 @opindex mnoregparam
10272 Do not pass any arguments in registers. This is the default for all
10277 It is OK to use the sb as an index register which is always loaded with
10278 zero. This is the default for the pc532-netbsd target.
10282 The sb register is not available for use or has not been initialized to
10283 zero by the run time system. This is the default for all targets except
10284 the pc532-netbsd. It is also implied whenever @option{-mhimem} or
10285 @option{-fpic} is set.
10289 Many ns32000 series addressing modes use displacements of up to 512MB@.
10290 If an address is above 512MB then displacements from zero can not be used.
10291 This option causes code to be generated which can be loaded above 512MB@.
10292 This may be useful for operating systems or ROM code.
10296 Assume code will be loaded in the first 512MB of virtual address space.
10297 This is the default for all platforms.
10301 @node PDP-11 Options
10302 @subsection PDP-11 Options
10303 @cindex PDP-11 Options
10305 These options are defined for the PDP-11:
10310 Use hardware FPP floating point. This is the default. (FIS floating
10311 point on the PDP-11/40 is not supported.)
10314 @opindex msoft-float
10315 Do not use hardware floating point.
10319 Return floating-point results in ac0 (fr0 in Unix assembler syntax).
10323 Return floating-point results in memory. This is the default.
10327 Generate code for a PDP-11/40.
10331 Generate code for a PDP-11/45. This is the default.
10335 Generate code for a PDP-11/10.
10337 @item -mbcopy-builtin
10338 @opindex bcopy-builtin
10339 Use inline @code{movmemhi} patterns for copying memory. This is the
10344 Do not use inline @code{movmemhi} patterns for copying memory.
10350 Use 16-bit @code{int}. This is the default.
10356 Use 32-bit @code{int}.
10359 @itemx -mno-float32
10361 @opindex mno-float32
10362 Use 64-bit @code{float}. This is the default.
10365 @itemx -mno-float64
10367 @opindex mno-float64
10368 Use 32-bit @code{float}.
10372 Use @code{abshi2} pattern. This is the default.
10376 Do not use @code{abshi2} pattern.
10378 @item -mbranch-expensive
10379 @opindex mbranch-expensive
10380 Pretend that branches are expensive. This is for experimenting with
10381 code generation only.
10383 @item -mbranch-cheap
10384 @opindex mbranch-cheap
10385 Do not pretend that branches are expensive. This is the default.
10389 Generate code for a system with split I&D@.
10393 Generate code for a system without split I&D@. This is the default.
10397 Use Unix assembler syntax. This is the default when configured for
10398 @samp{pdp11-*-bsd}.
10402 Use DEC assembler syntax. This is the default when configured for any
10403 PDP-11 target other than @samp{pdp11-*-bsd}.
10406 @node PowerPC Options
10407 @subsection PowerPC Options
10408 @cindex PowerPC options
10410 These are listed under @xref{RS/6000 and PowerPC Options}.
10412 @node RS/6000 and PowerPC Options
10413 @subsection IBM RS/6000 and PowerPC Options
10414 @cindex RS/6000 and PowerPC Options
10415 @cindex IBM RS/6000 and PowerPC Options
10417 These @samp{-m} options are defined for the IBM RS/6000 and PowerPC:
10424 @itemx -mno-powerpc
10425 @itemx -mpowerpc-gpopt
10426 @itemx -mno-powerpc-gpopt
10427 @itemx -mpowerpc-gfxopt
10428 @itemx -mno-powerpc-gfxopt
10430 @itemx -mno-powerpc64
10434 @opindex mno-power2
10436 @opindex mno-powerpc
10437 @opindex mpowerpc-gpopt
10438 @opindex mno-powerpc-gpopt
10439 @opindex mpowerpc-gfxopt
10440 @opindex mno-powerpc-gfxopt
10441 @opindex mpowerpc64
10442 @opindex mno-powerpc64
10443 GCC supports two related instruction set architectures for the
10444 RS/6000 and PowerPC@. The @dfn{POWER} instruction set are those
10445 instructions supported by the @samp{rios} chip set used in the original
10446 RS/6000 systems and the @dfn{PowerPC} instruction set is the
10447 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx microprocessors, and
10448 the IBM 4xx microprocessors.
10450 Neither architecture is a subset of the other. However there is a
10451 large common subset of instructions supported by both. An MQ
10452 register is included in processors supporting the POWER architecture.
10454 You use these options to specify which instructions are available on the
10455 processor you are using. The default value of these options is
10456 determined when configuring GCC@. Specifying the
10457 @option{-mcpu=@var{cpu_type}} overrides the specification of these
10458 options. We recommend you use the @option{-mcpu=@var{cpu_type}} option
10459 rather than the options listed above.
10461 The @option{-mpower} option allows GCC to generate instructions that
10462 are found only in the POWER architecture and to use the MQ register.
10463 Specifying @option{-mpower2} implies @option{-power} and also allows GCC
10464 to generate instructions that are present in the POWER2 architecture but
10465 not the original POWER architecture.
10467 The @option{-mpowerpc} option allows GCC to generate instructions that
10468 are found only in the 32-bit subset of the PowerPC architecture.
10469 Specifying @option{-mpowerpc-gpopt} implies @option{-mpowerpc} and also allows
10470 GCC to use the optional PowerPC architecture instructions in the
10471 General Purpose group, including floating-point square root. Specifying
10472 @option{-mpowerpc-gfxopt} implies @option{-mpowerpc} and also allows GCC to
10473 use the optional PowerPC architecture instructions in the Graphics
10474 group, including floating-point select.
10476 The @option{-mpowerpc64} option allows GCC to generate the additional
10477 64-bit instructions that are found in the full PowerPC64 architecture
10478 and to treat GPRs as 64-bit, doubleword quantities. GCC defaults to
10479 @option{-mno-powerpc64}.
10481 If you specify both @option{-mno-power} and @option{-mno-powerpc}, GCC
10482 will use only the instructions in the common subset of both
10483 architectures plus some special AIX common-mode calls, and will not use
10484 the MQ register. Specifying both @option{-mpower} and @option{-mpowerpc}
10485 permits GCC to use any instruction from either architecture and to
10486 allow use of the MQ register; specify this for the Motorola MPC601.
10488 @item -mnew-mnemonics
10489 @itemx -mold-mnemonics
10490 @opindex mnew-mnemonics
10491 @opindex mold-mnemonics
10492 Select which mnemonics to use in the generated assembler code. With
10493 @option{-mnew-mnemonics}, GCC uses the assembler mnemonics defined for
10494 the PowerPC architecture. With @option{-mold-mnemonics} it uses the
10495 assembler mnemonics defined for the POWER architecture. Instructions
10496 defined in only one architecture have only one mnemonic; GCC uses that
10497 mnemonic irrespective of which of these options is specified.
10499 GCC defaults to the mnemonics appropriate for the architecture in
10500 use. Specifying @option{-mcpu=@var{cpu_type}} sometimes overrides the
10501 value of these option. Unless you are building a cross-compiler, you
10502 should normally not specify either @option{-mnew-mnemonics} or
10503 @option{-mold-mnemonics}, but should instead accept the default.
10505 @item -mcpu=@var{cpu_type}
10507 Set architecture type, register usage, choice of mnemonics, and
10508 instruction scheduling parameters for machine type @var{cpu_type}.
10509 Supported values for @var{cpu_type} are @samp{401}, @samp{403},
10510 @samp{405}, @samp{405fp}, @samp{440}, @samp{440fp}, @samp{505},
10511 @samp{601}, @samp{602}, @samp{603}, @samp{603e}, @samp{604},
10512 @samp{604e}, @samp{620}, @samp{630}, @samp{740}, @samp{7400},
10513 @samp{7450}, @samp{750}, @samp{801}, @samp{821}, @samp{823},
10514 @samp{860}, @samp{970}, @samp{8540}, @samp{common}, @samp{ec603e}, @samp{G3},
10515 @samp{G4}, @samp{G5}, @samp{power}, @samp{power2}, @samp{power3},
10516 @samp{power4}, @samp{power5}, @samp{powerpc}, @samp{powerpc64},
10517 @samp{rios}, @samp{rios1}, @samp{rios2}, @samp{rsc}, and @samp{rs64a}.
10519 @option{-mcpu=common} selects a completely generic processor. Code
10520 generated under this option will run on any POWER or PowerPC processor.
10521 GCC will use only the instructions in the common subset of both
10522 architectures, and will not use the MQ register. GCC assumes a generic
10523 processor model for scheduling purposes.
10525 @option{-mcpu=power}, @option{-mcpu=power2}, @option{-mcpu=powerpc}, and
10526 @option{-mcpu=powerpc64} specify generic POWER, POWER2, pure 32-bit
10527 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10528 types, with an appropriate, generic processor model assumed for
10529 scheduling purposes.
10531 The other options specify a specific processor. Code generated under
10532 those options will run best on that processor, and may not run at all on
10535 The @option{-mcpu} options automatically enable or disable the
10536 following options: @option{-maltivec}, @option{-mhard-float},
10537 @option{-mmfcrf}, @option{-mmultiple}, @option{-mnew-mnemonics},
10538 @option{-mpower}, @option{-mpower2}, @option{-mpowerpc64},
10539 @option{-mpowerpc-gpopt}, @option{-mpowerpc-gfxopt},
10540 @option{-mstring}. The particular options set for any particular CPU
10541 will vary between compiler versions, depending on what setting seems
10542 to produce optimal code for that CPU; it doesn't necessarily reflect
10543 the actual hardware's capabilities. If you wish to set an individual
10544 option to a particular value, you may specify it after the
10545 @option{-mcpu} option, like @samp{-mcpu=970 -mno-altivec}.
10547 On AIX, the @option{-maltivec} and @option{-mpowerpc64} options are
10548 not enabled or disabled by the @option{-mcpu} option at present, since
10549 AIX does not have full support for these options. You may still
10550 enable or disable them individually if you're sure it'll work in your
10553 @item -mtune=@var{cpu_type}
10555 Set the instruction scheduling parameters for machine type
10556 @var{cpu_type}, but do not set the architecture type, register usage, or
10557 choice of mnemonics, as @option{-mcpu=@var{cpu_type}} would. The same
10558 values for @var{cpu_type} are used for @option{-mtune} as for
10559 @option{-mcpu}. If both are specified, the code generated will use the
10560 architecture, registers, and mnemonics set by @option{-mcpu}, but the
10561 scheduling parameters set by @option{-mtune}.
10564 @itemx -mno-altivec
10566 @opindex mno-altivec
10567 Generate code that uses (does not use) AltiVec instructions, and also
10568 enable the use of built-in functions that allow more direct access to
10569 the AltiVec instruction set. You may also need to set
10570 @option{-mabi=altivec} to adjust the current ABI with AltiVec ABI
10575 Extend the current ABI with SPE ABI extensions. This does not change
10576 the default ABI, instead it adds the SPE ABI extensions to the current
10580 @opindex mabi=no-spe
10581 Disable Booke SPE ABI extensions for the current ABI@.
10583 @item -misel=@var{yes/no}
10586 This switch enables or disables the generation of ISEL instructions.
10588 @item -mspe=@var{yes/no}
10591 This switch enables or disables the generation of SPE simd
10594 @item -mfloat-gprs=@var{yes/single/double/no}
10595 @itemx -mfloat-gprs
10596 @opindex mfloat-gprs
10597 This switch enables or disables the generation of floating point
10598 operations on the general purpose registers for architectures that
10601 The argument @var{yes} or @var{single} enables the use of
10602 single-precision floating point operations.
10604 The argument @var{double} enables the use of single and
10605 double-precision floating point operations.
10607 The argument @var{no} disables floating point operations on the
10608 general purpose registers.
10610 This option is currently only available on the MPC854x.
10616 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10617 targets (including GNU/Linux). The 32-bit environment sets int, long
10618 and pointer to 32 bits and generates code that runs on any PowerPC
10619 variant. The 64-bit environment sets int to 32 bits and long and
10620 pointer to 64 bits, and generates code for PowerPC64, as for
10621 @option{-mpowerpc64}.
10624 @itemx -mno-fp-in-toc
10625 @itemx -mno-sum-in-toc
10626 @itemx -mminimal-toc
10628 @opindex mno-fp-in-toc
10629 @opindex mno-sum-in-toc
10630 @opindex mminimal-toc
10631 Modify generation of the TOC (Table Of Contents), which is created for
10632 every executable file. The @option{-mfull-toc} option is selected by
10633 default. In that case, GCC will allocate at least one TOC entry for
10634 each unique non-automatic variable reference in your program. GCC
10635 will also place floating-point constants in the TOC@. However, only
10636 16,384 entries are available in the TOC@.
10638 If you receive a linker error message that saying you have overflowed
10639 the available TOC space, you can reduce the amount of TOC space used
10640 with the @option{-mno-fp-in-toc} and @option{-mno-sum-in-toc} options.
10641 @option{-mno-fp-in-toc} prevents GCC from putting floating-point
10642 constants in the TOC and @option{-mno-sum-in-toc} forces GCC to
10643 generate code to calculate the sum of an address and a constant at
10644 run-time instead of putting that sum into the TOC@. You may specify one
10645 or both of these options. Each causes GCC to produce very slightly
10646 slower and larger code at the expense of conserving TOC space.
10648 If you still run out of space in the TOC even when you specify both of
10649 these options, specify @option{-mminimal-toc} instead. This option causes
10650 GCC to make only one TOC entry for every file. When you specify this
10651 option, GCC will produce code that is slower and larger but which
10652 uses extremely little TOC space. You may wish to use this option
10653 only on files that contain less frequently executed code.
10659 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 64-bit
10660 @code{long} type, and the infrastructure needed to support them.
10661 Specifying @option{-maix64} implies @option{-mpowerpc64} and
10662 @option{-mpowerpc}, while @option{-maix32} disables the 64-bit ABI and
10663 implies @option{-mno-powerpc64}. GCC defaults to @option{-maix32}.
10666 @itemx -mno-xl-compat
10667 @opindex mxl-compat
10668 @opindex mno-xl-compat
10669 Produce code that conforms more closely to IBM XLC semantics when using
10670 AIX-compatible ABI. Pass floating-point arguments to prototyped
10671 functions beyond the register save area (RSA) on the stack in addition
10672 to argument FPRs. Do not assume that most significant double in 128
10673 bit long double value is properly rounded when comparing values.
10675 The AIX calling convention was extended but not initially documented to
10676 handle an obscure K&R C case of calling a function that takes the
10677 address of its arguments with fewer arguments than declared. AIX XL
10678 compilers access floating point arguments which do not fit in the
10679 RSA from the stack when a subroutine is compiled without
10680 optimization. Because always storing floating-point arguments on the
10681 stack is inefficient and rarely needed, this option is not enabled by
10682 default and only is necessary when calling subroutines compiled by AIX
10683 XL compilers without optimization.
10687 Support @dfn{IBM RS/6000 SP} @dfn{Parallel Environment} (PE)@. Link an
10688 application written to use message passing with special startup code to
10689 enable the application to run. The system must have PE installed in the
10690 standard location (@file{/usr/lpp/ppe.poe/}), or the @file{specs} file
10691 must be overridden with the @option{-specs=} option to specify the
10692 appropriate directory location. The Parallel Environment does not
10693 support threads, so the @option{-mpe} option and the @option{-pthread}
10694 option are incompatible.
10696 @item -malign-natural
10697 @itemx -malign-power
10698 @opindex malign-natural
10699 @opindex malign-power
10700 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10701 @option{-malign-natural} overrides the ABI-defined alignment of larger
10702 types, such as floating-point doubles, on their natural size-based boundary.
10703 The option @option{-malign-power} instructs GCC to follow the ABI-specified
10704 alignment rules. GCC defaults to the standard alignment defined in the ABI@.
10706 On 64-bit Darwin, natural alignment is the default, and @option{-malign-power}
10710 @itemx -mhard-float
10711 @opindex msoft-float
10712 @opindex mhard-float
10713 Generate code that does not use (uses) the floating-point register set.
10714 Software floating point emulation is provided if you use the
10715 @option{-msoft-float} option, and pass the option to GCC when linking.
10718 @itemx -mno-multiple
10720 @opindex mno-multiple
10721 Generate code that uses (does not use) the load multiple word
10722 instructions and the store multiple word instructions. These
10723 instructions are generated by default on POWER systems, and not
10724 generated on PowerPC systems. Do not use @option{-mmultiple} on little
10725 endian PowerPC systems, since those instructions do not work when the
10726 processor is in little endian mode. The exceptions are PPC740 and
10727 PPC750 which permit the instructions usage in little endian mode.
10732 @opindex mno-string
10733 Generate code that uses (does not use) the load string instructions
10734 and the store string word instructions to save multiple registers and
10735 do small block moves. These instructions are generated by default on
10736 POWER systems, and not generated on PowerPC systems. Do not use
10737 @option{-mstring} on little endian PowerPC systems, since those
10738 instructions do not work when the processor is in little endian mode.
10739 The exceptions are PPC740 and PPC750 which permit the instructions
10740 usage in little endian mode.
10745 @opindex mno-update
10746 Generate code that uses (does not use) the load or store instructions
10747 that update the base register to the address of the calculated memory
10748 location. These instructions are generated by default. If you use
10749 @option{-mno-update}, there is a small window between the time that the
10750 stack pointer is updated and the address of the previous frame is
10751 stored, which means code that walks the stack frame across interrupts or
10752 signals may get corrupted data.
10755 @itemx -mno-fused-madd
10756 @opindex mfused-madd
10757 @opindex mno-fused-madd
10758 Generate code that uses (does not use) the floating point multiply and
10759 accumulate instructions. These instructions are generated by default if
10760 hardware floating is used.
10762 @item -mno-bit-align
10764 @opindex mno-bit-align
10765 @opindex mbit-align
10766 On System V.4 and embedded PowerPC systems do not (do) force structures
10767 and unions that contain bit-fields to be aligned to the base type of the
10770 For example, by default a structure containing nothing but 8
10771 @code{unsigned} bit-fields of length 1 would be aligned to a 4 byte
10772 boundary and have a size of 4 bytes. By using @option{-mno-bit-align},
10773 the structure would be aligned to a 1 byte boundary and be one byte in
10776 @item -mno-strict-align
10777 @itemx -mstrict-align
10778 @opindex mno-strict-align
10779 @opindex mstrict-align
10780 On System V.4 and embedded PowerPC systems do not (do) assume that
10781 unaligned memory references will be handled by the system.
10783 @item -mrelocatable
10784 @itemx -mno-relocatable
10785 @opindex mrelocatable
10786 @opindex mno-relocatable
10787 On embedded PowerPC systems generate code that allows (does not allow)
10788 the program to be relocated to a different address at runtime. If you
10789 use @option{-mrelocatable} on any module, all objects linked together must
10790 be compiled with @option{-mrelocatable} or @option{-mrelocatable-lib}.
10792 @item -mrelocatable-lib
10793 @itemx -mno-relocatable-lib
10794 @opindex mrelocatable-lib
10795 @opindex mno-relocatable-lib
10796 On embedded PowerPC systems generate code that allows (does not allow)
10797 the program to be relocated to a different address at runtime. Modules
10798 compiled with @option{-mrelocatable-lib} can be linked with either modules
10799 compiled without @option{-mrelocatable} and @option{-mrelocatable-lib} or
10800 with modules compiled with the @option{-mrelocatable} options.
10806 On System V.4 and embedded PowerPC systems do not (do) assume that
10807 register 2 contains a pointer to a global area pointing to the addresses
10808 used in the program.
10811 @itemx -mlittle-endian
10813 @opindex mlittle-endian
10814 On System V.4 and embedded PowerPC systems compile code for the
10815 processor in little endian mode. The @option{-mlittle-endian} option is
10816 the same as @option{-mlittle}.
10819 @itemx -mbig-endian
10821 @opindex mbig-endian
10822 On System V.4 and embedded PowerPC systems compile code for the
10823 processor in big endian mode. The @option{-mbig-endian} option is
10824 the same as @option{-mbig}.
10826 @item -mdynamic-no-pic
10827 @opindex mdynamic-no-pic
10828 On Darwin and Mac OS X systems, compile code so that it is not
10829 relocatable, but that its external references are relocatable. The
10830 resulting code is suitable for applications, but not shared
10833 @item -mprioritize-restricted-insns=@var{priority}
10834 @opindex mprioritize-restricted-insns
10835 This option controls the priority that is assigned to
10836 dispatch-slot restricted instructions during the second scheduling
10837 pass. The argument @var{priority} takes the value @var{0/1/2} to assign
10838 @var{no/highest/second-highest} priority to dispatch slot restricted
10841 @item -msched-costly-dep=@var{dependence_type}
10842 @opindex msched-costly-dep
10843 This option controls which dependences are considered costly
10844 by the target during instruction scheduling. The argument
10845 @var{dependence_type} takes one of the following values:
10846 @var{no}: no dependence is costly,
10847 @var{all}: all dependences are costly,
10848 @var{true_store_to_load}: a true dependence from store to load is costly,
10849 @var{store_to_load}: any dependence from store to load is costly,
10850 @var{number}: any dependence which latency >= @var{number} is costly.
10852 @item -minsert-sched-nops=@var{scheme}
10853 @opindex minsert-sched-nops
10854 This option controls which nop insertion scheme will be used during
10855 the second scheduling pass. The argument @var{scheme} takes one of the
10857 @var{no}: Don't insert nops.
10858 @var{pad}: Pad with nops any dispatch group which has vacant issue slots,
10859 according to the scheduler's grouping.
10860 @var{regroup_exact}: Insert nops to force costly dependent insns into
10861 separate groups. Insert exactly as many nops as needed to force an insn
10862 to a new group, according to the estimated processor grouping.
10863 @var{number}: Insert nops to force costly dependent insns into
10864 separate groups. Insert @var{number} nops to force an insn to a new group.
10867 @opindex mcall-sysv
10868 On System V.4 and embedded PowerPC systems compile code using calling
10869 conventions that adheres to the March 1995 draft of the System V
10870 Application Binary Interface, PowerPC processor supplement. This is the
10871 default unless you configured GCC using @samp{powerpc-*-eabiaix}.
10873 @item -mcall-sysv-eabi
10874 @opindex mcall-sysv-eabi
10875 Specify both @option{-mcall-sysv} and @option{-meabi} options.
10877 @item -mcall-sysv-noeabi
10878 @opindex mcall-sysv-noeabi
10879 Specify both @option{-mcall-sysv} and @option{-mno-eabi} options.
10881 @item -mcall-solaris
10882 @opindex mcall-solaris
10883 On System V.4 and embedded PowerPC systems compile code for the Solaris
10887 @opindex mcall-linux
10888 On System V.4 and embedded PowerPC systems compile code for the
10889 Linux-based GNU system.
10893 On System V.4 and embedded PowerPC systems compile code for the
10894 Hurd-based GNU system.
10896 @item -mcall-netbsd
10897 @opindex mcall-netbsd
10898 On System V.4 and embedded PowerPC systems compile code for the
10899 NetBSD operating system.
10901 @item -maix-struct-return
10902 @opindex maix-struct-return
10903 Return all structures in memory (as specified by the AIX ABI)@.
10905 @item -msvr4-struct-return
10906 @opindex msvr4-struct-return
10907 Return structures smaller than 8 bytes in registers (as specified by the
10910 @item -mabi=altivec
10911 @opindex mabi=altivec
10912 Extend the current ABI with AltiVec ABI extensions. This does not
10913 change the default ABI, instead it adds the AltiVec ABI extensions to
10916 @item -mabi=no-altivec
10917 @opindex mabi=no-altivec
10918 Disable AltiVec ABI extensions for the current ABI@.
10921 @itemx -mno-prototype
10922 @opindex mprototype
10923 @opindex mno-prototype
10924 On System V.4 and embedded PowerPC systems assume that all calls to
10925 variable argument functions are properly prototyped. Otherwise, the
10926 compiler must insert an instruction before every non prototyped call to
10927 set or clear bit 6 of the condition code register (@var{CR}) to
10928 indicate whether floating point values were passed in the floating point
10929 registers in case the function takes a variable arguments. With
10930 @option{-mprototype}, only calls to prototyped variable argument functions
10931 will set or clear the bit.
10935 On embedded PowerPC systems, assume that the startup module is called
10936 @file{sim-crt0.o} and that the standard C libraries are @file{libsim.a} and
10937 @file{libc.a}. This is the default for @samp{powerpc-*-eabisim}.
10942 On embedded PowerPC systems, assume that the startup module is called
10943 @file{crt0.o} and the standard C libraries are @file{libmvme.a} and
10948 On embedded PowerPC systems, assume that the startup module is called
10949 @file{crt0.o} and the standard C libraries are @file{libads.a} and
10952 @item -myellowknife
10953 @opindex myellowknife
10954 On embedded PowerPC systems, assume that the startup module is called
10955 @file{crt0.o} and the standard C libraries are @file{libyk.a} and
10960 On System V.4 and embedded PowerPC systems, specify that you are
10961 compiling for a VxWorks system.
10965 Specify that you are compiling for the WindISS simulation environment.
10969 On embedded PowerPC systems, set the @var{PPC_EMB} bit in the ELF flags
10970 header to indicate that @samp{eabi} extended relocations are used.
10976 On System V.4 and embedded PowerPC systems do (do not) adhere to the
10977 Embedded Applications Binary Interface (eabi) which is a set of
10978 modifications to the System V.4 specifications. Selecting @option{-meabi}
10979 means that the stack is aligned to an 8 byte boundary, a function
10980 @code{__eabi} is called to from @code{main} to set up the eabi
10981 environment, and the @option{-msdata} option can use both @code{r2} and
10982 @code{r13} to point to two separate small data areas. Selecting
10983 @option{-mno-eabi} means that the stack is aligned to a 16 byte boundary,
10984 do not call an initialization function from @code{main}, and the
10985 @option{-msdata} option will only use @code{r13} to point to a single
10986 small data area. The @option{-meabi} option is on by default if you
10987 configured GCC using one of the @samp{powerpc*-*-eabi*} options.
10990 @opindex msdata=eabi
10991 On System V.4 and embedded PowerPC systems, put small initialized
10992 @code{const} global and static data in the @samp{.sdata2} section, which
10993 is pointed to by register @code{r2}. Put small initialized
10994 non-@code{const} global and static data in the @samp{.sdata} section,
10995 which is pointed to by register @code{r13}. Put small uninitialized
10996 global and static data in the @samp{.sbss} section, which is adjacent to
10997 the @samp{.sdata} section. The @option{-msdata=eabi} option is
10998 incompatible with the @option{-mrelocatable} option. The
10999 @option{-msdata=eabi} option also sets the @option{-memb} option.
11002 @opindex msdata=sysv
11003 On System V.4 and embedded PowerPC systems, put small global and static
11004 data in the @samp{.sdata} section, which is pointed to by register
11005 @code{r13}. Put small uninitialized global and static data in the
11006 @samp{.sbss} section, which is adjacent to the @samp{.sdata} section.
11007 The @option{-msdata=sysv} option is incompatible with the
11008 @option{-mrelocatable} option.
11010 @item -msdata=default
11012 @opindex msdata=default
11014 On System V.4 and embedded PowerPC systems, if @option{-meabi} is used,
11015 compile code the same as @option{-msdata=eabi}, otherwise compile code the
11016 same as @option{-msdata=sysv}.
11019 @opindex msdata-data
11020 On System V.4 and embedded PowerPC systems, put small global and static
11021 data in the @samp{.sdata} section. Put small uninitialized global and
11022 static data in the @samp{.sbss} section. Do not use register @code{r13}
11023 to address small data however. This is the default behavior unless
11024 other @option{-msdata} options are used.
11028 @opindex msdata=none
11030 On embedded PowerPC systems, put all initialized global and static data
11031 in the @samp{.data} section, and all uninitialized data in the
11032 @samp{.bss} section.
11036 @cindex smaller data references (PowerPC)
11037 @cindex .sdata/.sdata2 references (PowerPC)
11038 On embedded PowerPC systems, put global and static items less than or
11039 equal to @var{num} bytes into the small data or bss sections instead of
11040 the normal data or bss section. By default, @var{num} is 8. The
11041 @option{-G @var{num}} switch is also passed to the linker.
11042 All modules should be compiled with the same @option{-G @var{num}} value.
11045 @itemx -mno-regnames
11047 @opindex mno-regnames
11048 On System V.4 and embedded PowerPC systems do (do not) emit register
11049 names in the assembly language output using symbolic forms.
11052 @itemx -mno-longcall
11054 @opindex mno-longcall
11055 Default to making all function calls indirectly, using a register, so
11056 that functions which reside further than 32 megabytes (33,554,432
11057 bytes) from the current location can be called. This setting can be
11058 overridden by the @code{shortcall} function attribute, or by
11059 @code{#pragma longcall(0)}.
11061 Some linkers are capable of detecting out-of-range calls and generating
11062 glue code on the fly. On these systems, long calls are unnecessary and
11063 generate slower code. As of this writing, the AIX linker can do this,
11064 as can the GNU linker for PowerPC/64. It is planned to add this feature
11065 to the GNU linker for 32-bit PowerPC systems as well.
11067 On Darwin/PPC systems, @code{#pragma longcall} will generate ``jbsr
11068 callee, L42'', plus a ``branch island'' (glue code). The two target
11069 addresses represent the callee and the ``branch island''. The
11070 Darwin/PPC linker will prefer the first address and generate a ``bl
11071 callee'' if the PPC ``bl'' instruction will reach the callee directly;
11072 otherwise, the linker will generate ``bl L42'' to call the ``branch
11073 island''. The ``branch island'' is appended to the body of the
11074 calling function; it computes the full 32-bit address of the callee
11077 On Mach-O (Darwin) systems, this option directs the compiler emit to
11078 the glue for every direct call, and the Darwin linker decides whether
11079 to use or discard it.
11081 In the future, we may cause GCC to ignore all longcall specifications
11082 when the linker is known to generate glue.
11086 Adds support for multithreading with the @dfn{pthreads} library.
11087 This option sets flags for both the preprocessor and linker.
11091 @node S/390 and zSeries Options
11092 @subsection S/390 and zSeries Options
11093 @cindex S/390 and zSeries Options
11095 These are the @samp{-m} options defined for the S/390 and zSeries architecture.
11099 @itemx -msoft-float
11100 @opindex mhard-float
11101 @opindex msoft-float
11102 Use (do not use) the hardware floating-point instructions and registers
11103 for floating-point operations. When @option{-msoft-float} is specified,
11104 functions in @file{libgcc.a} will be used to perform floating-point
11105 operations. When @option{-mhard-float} is specified, the compiler
11106 generates IEEE floating-point instructions. This is the default.
11109 @itemx -mno-backchain
11110 @opindex mbackchain
11111 @opindex mno-backchain
11112 Store (do not store) the address of the caller's frame as backchain pointer
11113 into the callee's stack frame.
11114 A backchain may be needed to allow debugging using tools that do not understand
11115 DWARF-2 call frame information.
11116 When @option{-mno-packed-stack} is in effect, the backchain pointer is stored
11117 at the bottom of the stack frame; when @option{-mpacked-stack} is in effect,
11118 the backchain is placed into the topmost word of the 96/160 byte register
11121 In general, code compiled with @option{-mbackchain} is call-compatible with
11122 code compiled with @option{-mmo-backchain}; however, use of the backchain
11123 for debugging purposes usually requires that the whole binary is built with
11124 @option{-mbackchain}. Note that the combination of @option{-mbackchain},
11125 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11126 to build a linux kernel use @option{-msoft-float}.
11128 The default is to not maintain the backchain.
11130 @item -mpacked-stack
11131 @item -mno-packed-stack
11132 @opindex mpacked-stack
11133 @opindex mno-packed-stack
11134 Use (do not use) the packed stack layout. When @option{-mno-packed-stack} is
11135 specified, the compiler uses the all fields of the 96/160 byte register save
11136 area only for their default purpose; unused fields still take up stack space.
11137 When @option{-mpacked-stack} is specified, register save slots are densely
11138 packed at the top of the register save area; unused space is reused for other
11139 purposes, allowing for more efficient use of the available stack space.
11140 However, when @option{-mbackchain} is also in effect, the topmost word of
11141 the save area is always used to store the backchain, and the return address
11142 register is always saved two words below the backchain.
11144 As long as the stack frame backchain is not used, code generated with
11145 @option{-mpacked-stack} is call-compatible with code generated with
11146 @option{-mno-packed-stack}. Note that some non-FSF releases of GCC 2.95 for
11147 S/390 or zSeries generated code that uses the stack frame backchain at run
11148 time, not just for debugging purposes. Such code is not call-compatible
11149 with code compiled with @option{-mpacked-stack}. Also, note that the
11150 combination of @option{-mbackchain},
11151 @option{-mpacked-stack} and @option{-mhard-float} is not supported. In order
11152 to build a linux kernel use @option{-msoft-float}.
11154 The default is to not use the packed stack layout.
11157 @itemx -mno-small-exec
11158 @opindex msmall-exec
11159 @opindex mno-small-exec
11160 Generate (or do not generate) code using the @code{bras} instruction
11161 to do subroutine calls.
11162 This only works reliably if the total executable size does not
11163 exceed 64k. The default is to use the @code{basr} instruction instead,
11164 which does not have this limitation.
11170 When @option{-m31} is specified, generate code compliant to the
11171 GNU/Linux for S/390 ABI@. When @option{-m64} is specified, generate
11172 code compliant to the GNU/Linux for zSeries ABI@. This allows GCC in
11173 particular to generate 64-bit instructions. For the @samp{s390}
11174 targets, the default is @option{-m31}, while the @samp{s390x}
11175 targets default to @option{-m64}.
11181 When @option{-mzarch} is specified, generate code using the
11182 instructions available on z/Architecture.
11183 When @option{-mesa} is specified, generate code using the
11184 instructions available on ESA/390. Note that @option{-mesa} is
11185 not possible with @option{-m64}.
11186 When generating code compliant to the GNU/Linux for S/390 ABI,
11187 the default is @option{-mesa}. When generating code compliant
11188 to the GNU/Linux for zSeries ABI, the default is @option{-mzarch}.
11194 Generate (or do not generate) code using the @code{mvcle} instruction
11195 to perform block moves. When @option{-mno-mvcle} is specified,
11196 use a @code{mvc} loop instead. This is the default.
11202 Print (or do not print) additional debug information when compiling.
11203 The default is to not print debug information.
11205 @item -march=@var{cpu-type}
11207 Generate code that will run on @var{cpu-type}, which is the name of a system
11208 representing a certain processor type. Possible values for
11209 @var{cpu-type} are @samp{g5}, @samp{g6}, @samp{z900}, and @samp{z990}.
11210 When generating code using the instructions available on z/Architecture,
11211 the default is @option{-march=z900}. Otherwise, the default is
11212 @option{-march=g5}.
11214 @item -mtune=@var{cpu-type}
11216 Tune to @var{cpu-type} everything applicable about the generated code,
11217 except for the ABI and the set of available instructions.
11218 The list of @var{cpu-type} values is the same as for @option{-march}.
11219 The default is the value used for @option{-march}.
11222 @itemx -mno-tpf-trace
11223 @opindex mtpf-trace
11224 @opindex mno-tpf-trace
11225 Generate code that adds (does not add) in TPF OS specific branches to trace
11226 routines in the operating system. This option is off by default, even
11227 when compiling for the TPF OS@.
11230 @itemx -mno-fused-madd
11231 @opindex mfused-madd
11232 @opindex mno-fused-madd
11233 Generate code that uses (does not use) the floating point multiply and
11234 accumulate instructions. These instructions are generated by default if
11235 hardware floating point is used.
11237 @item -mwarn-framesize=@var{framesize}
11238 @opindex mwarn-framesize
11239 Emit a warning if the current function exceeds the given frame size. Because
11240 this is a compile time check it doesn't need to be a real problem when the program
11241 runs. It is intended to identify functions which most probably cause
11242 a stack overflow. It is useful to be used in an environment with limited stack
11243 size e.g.@: the linux kernel.
11245 @item -mwarn-dynamicstack
11246 @opindex mwarn-dynamicstack
11247 Emit a warning if the function calls alloca or uses dynamically
11248 sized arrays. This is generally a bad idea with a limited stack size.
11250 @item -mstack-guard=@var{stack-guard}
11251 @item -mstack-size=@var{stack-size}
11252 @opindex mstack-guard
11253 @opindex mstack-size
11254 These arguments always have to be used in conjunction. If they are present the s390
11255 back end emits additional instructions in the function prologue which trigger a trap
11256 if the stack size is @var{stack-guard} bytes above the @var{stack-size}
11257 (remember that the stack on s390 grows downward). These options are intended to
11258 be used to help debugging stack overflow problems. The additionally emitted code
11259 cause only little overhead and hence can also be used in production like systems
11260 without greater performance degradation. The given values have to be exact
11261 powers of 2 and @var{stack-size} has to be greater than @var{stack-guard}.
11262 In order to be efficient the extra code makes the assumption that the stack starts
11263 at an address aligned to the value given by @var{stack-size}.
11267 @subsection SH Options
11269 These @samp{-m} options are defined for the SH implementations:
11274 Generate code for the SH1.
11278 Generate code for the SH2.
11281 Generate code for the SH2e.
11285 Generate code for the SH3.
11289 Generate code for the SH3e.
11293 Generate code for the SH4 without a floating-point unit.
11295 @item -m4-single-only
11296 @opindex m4-single-only
11297 Generate code for the SH4 with a floating-point unit that only
11298 supports single-precision arithmetic.
11302 Generate code for the SH4 assuming the floating-point unit is in
11303 single-precision mode by default.
11307 Generate code for the SH4.
11311 Generate code for the SH4al-dsp, or for a SH4a in such a way that the
11312 floating-point unit is not used.
11314 @item -m4a-single-only
11315 @opindex m4a-single-only
11316 Generate code for the SH4a, in such a way that no double-precision
11317 floating point operations are used.
11320 @opindex m4a-single
11321 Generate code for the SH4a assuming the floating-point unit is in
11322 single-precision mode by default.
11326 Generate code for the SH4a.
11330 Same as @option{-m4a-nofpu}, except that it implicitly passes
11331 @option{-dsp} to the assembler. GCC doesn't generate any DSP
11332 instructions at the moment.
11336 Compile code for the processor in big endian mode.
11340 Compile code for the processor in little endian mode.
11344 Align doubles at 64-bit boundaries. Note that this changes the calling
11345 conventions, and thus some functions from the standard C library will
11346 not work unless you recompile it first with @option{-mdalign}.
11350 Shorten some address references at link time, when possible; uses the
11351 linker option @option{-relax}.
11355 Use 32-bit offsets in @code{switch} tables. The default is to use
11360 Enable the use of the instruction @code{fmovd}.
11364 Comply with the calling conventions defined by Renesas.
11368 Comply with the calling conventions defined by Renesas.
11372 Comply with the calling conventions defined for GCC before the Renesas
11373 conventions were available. This option is the default for all
11374 targets of the SH toolchain except for @samp{sh-symbianelf}.
11377 @opindex mnomacsave
11378 Mark the @code{MAC} register as call-clobbered, even if
11379 @option{-mhitachi} is given.
11383 Increase IEEE-compliance of floating-point code.
11387 Dump instruction size and location in the assembly code.
11390 @opindex mpadstruct
11391 This option is deprecated. It pads structures to multiple of 4 bytes,
11392 which is incompatible with the SH ABI@.
11396 Optimize for space instead of speed. Implied by @option{-Os}.
11399 @opindex mprefergot
11400 When generating position-independent code, emit function calls using
11401 the Global Offset Table instead of the Procedure Linkage Table.
11405 Generate a library function call to invalidate instruction cache
11406 entries, after fixing up a trampoline. This library function call
11407 doesn't assume it can write to the whole memory address space. This
11408 is the default when the target is @code{sh-*-linux*}.
11411 @node SPARC Options
11412 @subsection SPARC Options
11413 @cindex SPARC options
11415 These @samp{-m} options are supported on the SPARC:
11418 @item -mno-app-regs
11420 @opindex mno-app-regs
11422 Specify @option{-mapp-regs} to generate output using the global registers
11423 2 through 4, which the SPARC SVR4 ABI reserves for applications. This
11426 To be fully SVR4 ABI compliant at the cost of some performance loss,
11427 specify @option{-mno-app-regs}. You should compile libraries and system
11428 software with this option.
11431 @itemx -mhard-float
11433 @opindex mhard-float
11434 Generate output containing floating point instructions. This is the
11438 @itemx -msoft-float
11440 @opindex msoft-float
11441 Generate output containing library calls for floating point.
11442 @strong{Warning:} the requisite libraries are not available for all SPARC
11443 targets. Normally the facilities of the machine's usual C compiler are
11444 used, but this cannot be done directly in cross-compilation. You must make
11445 your own arrangements to provide suitable library functions for
11446 cross-compilation. The embedded targets @samp{sparc-*-aout} and
11447 @samp{sparclite-*-*} do provide software floating point support.
11449 @option{-msoft-float} changes the calling convention in the output file;
11450 therefore, it is only useful if you compile @emph{all} of a program with
11451 this option. In particular, you need to compile @file{libgcc.a}, the
11452 library that comes with GCC, with @option{-msoft-float} in order for
11455 @item -mhard-quad-float
11456 @opindex mhard-quad-float
11457 Generate output containing quad-word (long double) floating point
11460 @item -msoft-quad-float
11461 @opindex msoft-quad-float
11462 Generate output containing library calls for quad-word (long double)
11463 floating point instructions. The functions called are those specified
11464 in the SPARC ABI@. This is the default.
11466 As of this writing, there are no SPARC implementations that have hardware
11467 support for the quad-word floating point instructions. They all invoke
11468 a trap handler for one of these instructions, and then the trap handler
11469 emulates the effect of the instruction. Because of the trap handler overhead,
11470 this is much slower than calling the ABI library routines. Thus the
11471 @option{-msoft-quad-float} option is the default.
11473 @item -mno-unaligned-doubles
11474 @itemx -munaligned-doubles
11475 @opindex mno-unaligned-doubles
11476 @opindex munaligned-doubles
11477 Assume that doubles have 8 byte alignment. This is the default.
11479 With @option{-munaligned-doubles}, GCC assumes that doubles have 8 byte
11480 alignment only if they are contained in another type, or if they have an
11481 absolute address. Otherwise, it assumes they have 4 byte alignment.
11482 Specifying this option avoids some rare compatibility problems with code
11483 generated by other compilers. It is not the default because it results
11484 in a performance loss, especially for floating point code.
11486 @item -mno-faster-structs
11487 @itemx -mfaster-structs
11488 @opindex mno-faster-structs
11489 @opindex mfaster-structs
11490 With @option{-mfaster-structs}, the compiler assumes that structures
11491 should have 8 byte alignment. This enables the use of pairs of
11492 @code{ldd} and @code{std} instructions for copies in structure
11493 assignment, in place of twice as many @code{ld} and @code{st} pairs.
11494 However, the use of this changed alignment directly violates the SPARC
11495 ABI@. Thus, it's intended only for use on targets where the developer
11496 acknowledges that their resulting code will not be directly in line with
11497 the rules of the ABI@.
11499 @item -mimpure-text
11500 @opindex mimpure-text
11501 @option{-mimpure-text}, used in addition to @option{-shared}, tells
11502 the compiler to not pass @option{-z text} to the linker when linking a
11503 shared object. Using this option, you can link position-dependent
11504 code into a shared object.
11506 @option{-mimpure-text} suppresses the ``relocations remain against
11507 allocatable but non-writable sections'' linker error message.
11508 However, the necessary relocations will trigger copy-on-write, and the
11509 shared object is not actually shared across processes. Instead of
11510 using @option{-mimpure-text}, you should compile all source code with
11511 @option{-fpic} or @option{-fPIC}.
11513 This option is only available on SunOS and Solaris.
11515 @item -mcpu=@var{cpu_type}
11517 Set the instruction set, register set, and instruction scheduling parameters
11518 for machine type @var{cpu_type}. Supported values for @var{cpu_type} are
11519 @samp{v7}, @samp{cypress}, @samp{v8}, @samp{supersparc}, @samp{sparclite},
11520 @samp{f930}, @samp{f934}, @samp{hypersparc}, @samp{sparclite86x},
11521 @samp{sparclet}, @samp{tsc701}, @samp{v9}, @samp{ultrasparc}, and
11522 @samp{ultrasparc3}.
11524 Default instruction scheduling parameters are used for values that select
11525 an architecture and not an implementation. These are @samp{v7}, @samp{v8},
11526 @samp{sparclite}, @samp{sparclet}, @samp{v9}.
11528 Here is a list of each supported architecture and their supported
11533 v8: supersparc, hypersparc
11534 sparclite: f930, f934, sparclite86x
11536 v9: ultrasparc, ultrasparc3
11539 By default (unless configured otherwise), GCC generates code for the V7
11540 variant of the SPARC architecture. With @option{-mcpu=cypress}, the compiler
11541 additionally optimizes it for the Cypress CY7C602 chip, as used in the
11542 SPARCStation/SPARCServer 3xx series. This is also appropriate for the older
11543 SPARCStation 1, 2, IPX etc.
11545 With @option{-mcpu=v8}, GCC generates code for the V8 variant of the SPARC
11546 architecture. The only difference from V7 code is that the compiler emits
11547 the integer multiply and integer divide instructions which exist in SPARC-V8
11548 but not in SPARC-V7. With @option{-mcpu=supersparc}, the compiler additionally
11549 optimizes it for the SuperSPARC chip, as used in the SPARCStation 10, 1000 and
11552 With @option{-mcpu=sparclite}, GCC generates code for the SPARClite variant of
11553 the SPARC architecture. This adds the integer multiply, integer divide step
11554 and scan (@code{ffs}) instructions which exist in SPARClite but not in SPARC-V7.
11555 With @option{-mcpu=f930}, the compiler additionally optimizes it for the
11556 Fujitsu MB86930 chip, which is the original SPARClite, with no FPU@. With
11557 @option{-mcpu=f934}, the compiler additionally optimizes it for the Fujitsu
11558 MB86934 chip, which is the more recent SPARClite with FPU@.
11560 With @option{-mcpu=sparclet}, GCC generates code for the SPARClet variant of
11561 the SPARC architecture. This adds the integer multiply, multiply/accumulate,
11562 integer divide step and scan (@code{ffs}) instructions which exist in SPARClet
11563 but not in SPARC-V7. With @option{-mcpu=tsc701}, the compiler additionally
11564 optimizes it for the TEMIC SPARClet chip.
11566 With @option{-mcpu=v9}, GCC generates code for the V9 variant of the SPARC
11567 architecture. This adds 64-bit integer and floating-point move instructions,
11568 3 additional floating-point condition code registers and conditional move
11569 instructions. With @option{-mcpu=ultrasparc}, the compiler additionally
11570 optimizes it for the Sun UltraSPARC I/II chips. With
11571 @option{-mcpu=ultrasparc3}, the compiler additionally optimizes it for the
11572 Sun UltraSPARC III chip.
11574 @item -mtune=@var{cpu_type}
11576 Set the instruction scheduling parameters for machine type
11577 @var{cpu_type}, but do not set the instruction set or register set that the
11578 option @option{-mcpu=@var{cpu_type}} would.
11580 The same values for @option{-mcpu=@var{cpu_type}} can be used for
11581 @option{-mtune=@var{cpu_type}}, but the only useful values are those
11582 that select a particular cpu implementation. Those are @samp{cypress},
11583 @samp{supersparc}, @samp{hypersparc}, @samp{f930}, @samp{f934},
11584 @samp{sparclite86x}, @samp{tsc701}, @samp{ultrasparc}, and
11585 @samp{ultrasparc3}.
11590 @opindex mno-v8plus
11591 With @option{-mv8plus}, GCC generates code for the SPARC-V8+ ABI@. The
11592 difference from the V8 ABI is that the global and out registers are
11593 considered 64-bit wide. This is enabled by default on Solaris in 32-bit
11594 mode for all SPARC-V9 processors.
11600 With @option{-mvis}, GCC generates code that takes advantage of the UltraSPARC
11601 Visual Instruction Set extensions. The default is @option{-mno-vis}.
11604 These @samp{-m} options are supported in addition to the above
11605 on SPARC-V9 processors in 64-bit environments:
11608 @item -mlittle-endian
11609 @opindex mlittle-endian
11610 Generate code for a processor running in little-endian mode. It is only
11611 available for a few configurations and most notably not on Solaris and Linux.
11617 Generate code for a 32-bit or 64-bit environment.
11618 The 32-bit environment sets int, long and pointer to 32 bits.
11619 The 64-bit environment sets int to 32 bits and long and pointer
11622 @item -mcmodel=medlow
11623 @opindex mcmodel=medlow
11624 Generate code for the Medium/Low code model: 64-bit addresses, programs
11625 must be linked in the low 32 bits of memory. Programs can be statically
11626 or dynamically linked.
11628 @item -mcmodel=medmid
11629 @opindex mcmodel=medmid
11630 Generate code for the Medium/Middle code model: 64-bit addresses, programs
11631 must be linked in the low 44 bits of memory, the text and data segments must
11632 be less than 2GB in size and the data segment must be located within 2GB of
11635 @item -mcmodel=medany
11636 @opindex mcmodel=medany
11637 Generate code for the Medium/Anywhere code model: 64-bit addresses, programs
11638 may be linked anywhere in memory, the text and data segments must be less
11639 than 2GB in size and the data segment must be located within 2GB of the
11642 @item -mcmodel=embmedany
11643 @opindex mcmodel=embmedany
11644 Generate code for the Medium/Anywhere code model for embedded systems:
11645 64-bit addresses, the text and data segments must be less than 2GB in
11646 size, both starting anywhere in memory (determined at link time). The
11647 global register %g4 points to the base of the data segment. Programs
11648 are statically linked and PIC is not supported.
11651 @itemx -mno-stack-bias
11652 @opindex mstack-bias
11653 @opindex mno-stack-bias
11654 With @option{-mstack-bias}, GCC assumes that the stack pointer, and
11655 frame pointer if present, are offset by @minus{}2047 which must be added back
11656 when making stack frame references. This is the default in 64-bit mode.
11657 Otherwise, assume no such offset is present.
11660 These switches are supported in addition to the above on Solaris:
11665 Add support for multithreading using the Solaris threads library. This
11666 option sets flags for both the preprocessor and linker. This option does
11667 not affect the thread safety of object code produced by the compiler or
11668 that of libraries supplied with it.
11672 Add support for multithreading using the POSIX threads library. This
11673 option sets flags for both the preprocessor and linker. This option does
11674 not affect the thread safety of object code produced by the compiler or
11675 that of libraries supplied with it.
11678 @node System V Options
11679 @subsection Options for System V
11681 These additional options are available on System V Release 4 for
11682 compatibility with other compilers on those systems:
11687 Create a shared object.
11688 It is recommended that @option{-symbolic} or @option{-shared} be used instead.
11692 Identify the versions of each tool used by the compiler, in a
11693 @code{.ident} assembler directive in the output.
11697 Refrain from adding @code{.ident} directives to the output file (this is
11700 @item -YP,@var{dirs}
11702 Search the directories @var{dirs}, and no others, for libraries
11703 specified with @option{-l}.
11705 @item -Ym,@var{dir}
11707 Look in the directory @var{dir} to find the M4 preprocessor.
11708 The assembler uses this option.
11709 @c This is supposed to go with a -Yd for predefined M4 macro files, but
11710 @c the generic assembler that comes with Solaris takes just -Ym.
11713 @node TMS320C3x/C4x Options
11714 @subsection TMS320C3x/C4x Options
11715 @cindex TMS320C3x/C4x Options
11717 These @samp{-m} options are defined for TMS320C3x/C4x implementations:
11721 @item -mcpu=@var{cpu_type}
11723 Set the instruction set, register set, and instruction scheduling
11724 parameters for machine type @var{cpu_type}. Supported values for
11725 @var{cpu_type} are @samp{c30}, @samp{c31}, @samp{c32}, @samp{c40}, and
11726 @samp{c44}. The default is @samp{c40} to generate code for the
11731 @itemx -msmall-memory
11733 @opindex mbig-memory
11735 @opindex msmall-memory
11737 Generates code for the big or small memory model. The small memory
11738 model assumed that all data fits into one 64K word page. At run-time
11739 the data page (DP) register must be set to point to the 64K page
11740 containing the .bss and .data program sections. The big memory model is
11741 the default and requires reloading of the DP register for every direct
11748 Allow (disallow) allocation of general integer operands into the block
11749 count register BK@.
11755 Enable (disable) generation of code using decrement and branch,
11756 DBcond(D), instructions. This is enabled by default for the C4x. To be
11757 on the safe side, this is disabled for the C3x, since the maximum
11758 iteration count on the C3x is @math{2^{23} + 1} (but who iterates loops more than
11759 @math{2^{23}} times on the C3x?). Note that GCC will try to reverse a loop so
11760 that it can utilize the decrement and branch instruction, but will give
11761 up if there is more than one memory reference in the loop. Thus a loop
11762 where the loop counter is decremented can generate slightly more
11763 efficient code, in cases where the RPTB instruction cannot be utilized.
11765 @item -mdp-isr-reload
11767 @opindex mdp-isr-reload
11769 Force the DP register to be saved on entry to an interrupt service
11770 routine (ISR), reloaded to point to the data section, and restored on
11771 exit from the ISR@. This should not be required unless someone has
11772 violated the small memory model by modifying the DP register, say within
11779 For the C3x use the 24-bit MPYI instruction for integer multiplies
11780 instead of a library call to guarantee 32-bit results. Note that if one
11781 of the operands is a constant, then the multiplication will be performed
11782 using shifts and adds. If the @option{-mmpyi} option is not specified for the C3x,
11783 then squaring operations are performed inline instead of a library call.
11786 @itemx -mno-fast-fix
11788 @opindex mno-fast-fix
11789 The C3x/C4x FIX instruction to convert a floating point value to an
11790 integer value chooses the nearest integer less than or equal to the
11791 floating point value rather than to the nearest integer. Thus if the
11792 floating point number is negative, the result will be incorrectly
11793 truncated an additional code is necessary to detect and correct this
11794 case. This option can be used to disable generation of the additional
11795 code required to correct the result.
11801 Enable (disable) generation of repeat block sequences using the RPTB
11802 instruction for zero overhead looping. The RPTB construct is only used
11803 for innermost loops that do not call functions or jump across the loop
11804 boundaries. There is no advantage having nested RPTB loops due to the
11805 overhead required to save and restore the RC, RS, and RE registers.
11806 This is enabled by default with @option{-O2}.
11808 @item -mrpts=@var{count}
11812 Enable (disable) the use of the single instruction repeat instruction
11813 RPTS@. If a repeat block contains a single instruction, and the loop
11814 count can be guaranteed to be less than the value @var{count}, GCC will
11815 emit a RPTS instruction instead of a RPTB@. If no value is specified,
11816 then a RPTS will be emitted even if the loop count cannot be determined
11817 at compile time. Note that the repeated instruction following RPTS does
11818 not have to be reloaded from memory each iteration, thus freeing up the
11819 CPU buses for operands. However, since interrupts are blocked by this
11820 instruction, it is disabled by default.
11822 @item -mloop-unsigned
11823 @itemx -mno-loop-unsigned
11824 @opindex mloop-unsigned
11825 @opindex mno-loop-unsigned
11826 The maximum iteration count when using RPTS and RPTB (and DB on the C40)
11827 is @math{2^{31} + 1} since these instructions test if the iteration count is
11828 negative to terminate the loop. If the iteration count is unsigned
11829 there is a possibility than the @math{2^{31} + 1} maximum iteration count may be
11830 exceeded. This switch allows an unsigned iteration count.
11834 Try to emit an assembler syntax that the TI assembler (asm30) is happy
11835 with. This also enforces compatibility with the API employed by the TI
11836 C3x C compiler. For example, long doubles are passed as structures
11837 rather than in floating point registers.
11843 Generate code that uses registers (stack) for passing arguments to functions.
11844 By default, arguments are passed in registers where possible rather
11845 than by pushing arguments on to the stack.
11847 @item -mparallel-insns
11848 @itemx -mno-parallel-insns
11849 @opindex mparallel-insns
11850 @opindex mno-parallel-insns
11851 Allow the generation of parallel instructions. This is enabled by
11852 default with @option{-O2}.
11854 @item -mparallel-mpy
11855 @itemx -mno-parallel-mpy
11856 @opindex mparallel-mpy
11857 @opindex mno-parallel-mpy
11858 Allow the generation of MPY||ADD and MPY||SUB parallel instructions,
11859 provided @option{-mparallel-insns} is also specified. These instructions have
11860 tight register constraints which can pessimize the code generation
11861 of large functions.
11866 @subsection V850 Options
11867 @cindex V850 Options
11869 These @samp{-m} options are defined for V850 implementations:
11873 @itemx -mno-long-calls
11874 @opindex mlong-calls
11875 @opindex mno-long-calls
11876 Treat all calls as being far away (near). If calls are assumed to be
11877 far away, the compiler will always load the functions address up into a
11878 register, and call indirect through the pointer.
11884 Do not optimize (do optimize) basic blocks that use the same index
11885 pointer 4 or more times to copy pointer into the @code{ep} register, and
11886 use the shorter @code{sld} and @code{sst} instructions. The @option{-mep}
11887 option is on by default if you optimize.
11889 @item -mno-prolog-function
11890 @itemx -mprolog-function
11891 @opindex mno-prolog-function
11892 @opindex mprolog-function
11893 Do not use (do use) external functions to save and restore registers
11894 at the prologue and epilogue of a function. The external functions
11895 are slower, but use less code space if more than one function saves
11896 the same number of registers. The @option{-mprolog-function} option
11897 is on by default if you optimize.
11901 Try to make the code as small as possible. At present, this just turns
11902 on the @option{-mep} and @option{-mprolog-function} options.
11904 @item -mtda=@var{n}
11906 Put static or global variables whose size is @var{n} bytes or less into
11907 the tiny data area that register @code{ep} points to. The tiny data
11908 area can hold up to 256 bytes in total (128 bytes for byte references).
11910 @item -msda=@var{n}
11912 Put static or global variables whose size is @var{n} bytes or less into
11913 the small data area that register @code{gp} points to. The small data
11914 area can hold up to 64 kilobytes.
11916 @item -mzda=@var{n}
11918 Put static or global variables whose size is @var{n} bytes or less into
11919 the first 32 kilobytes of memory.
11923 Specify that the target processor is the V850.
11926 @opindex mbig-switch
11927 Generate code suitable for big switch tables. Use this option only if
11928 the assembler/linker complain about out of range branches within a switch
11933 This option will cause r2 and r5 to be used in the code generated by
11934 the compiler. This setting is the default.
11936 @item -mno-app-regs
11937 @opindex mno-app-regs
11938 This option will cause r2 and r5 to be treated as fixed registers.
11942 Specify that the target processor is the V850E1. The preprocessor
11943 constants @samp{__v850e1__} and @samp{__v850e__} will be defined if
11944 this option is used.
11948 Specify that the target processor is the V850E@. The preprocessor
11949 constant @samp{__v850e__} will be defined if this option is used.
11951 If neither @option{-mv850} nor @option{-mv850e} nor @option{-mv850e1}
11952 are defined then a default target processor will be chosen and the
11953 relevant @samp{__v850*__} preprocessor constant will be defined.
11955 The preprocessor constants @samp{__v850} and @samp{__v851__} are always
11956 defined, regardless of which processor variant is the target.
11958 @item -mdisable-callt
11959 @opindex mdisable-callt
11960 This option will suppress generation of the CALLT instruction for the
11961 v850e and v850e1 flavors of the v850 architecture. The default is
11962 @option{-mno-disable-callt} which allows the CALLT instruction to be used.
11967 @subsection VAX Options
11968 @cindex VAX options
11970 These @samp{-m} options are defined for the VAX:
11975 Do not output certain jump instructions (@code{aobleq} and so on)
11976 that the Unix assembler for the VAX cannot handle across long
11981 Do output those jump instructions, on the assumption that you
11982 will assemble with the GNU assembler.
11986 Output code for g-format floating point numbers instead of d-format.
11989 @node x86-64 Options
11990 @subsection x86-64 Options
11991 @cindex x86-64 options
11993 These are listed under @xref{i386 and x86-64 Options}.
11995 @node Xstormy16 Options
11996 @subsection Xstormy16 Options
11997 @cindex Xstormy16 Options
11999 These options are defined for Xstormy16:
12004 Choose startup files and linker script suitable for the simulator.
12007 @node Xtensa Options
12008 @subsection Xtensa Options
12009 @cindex Xtensa Options
12011 These options are supported for Xtensa targets:
12015 @itemx -mno-const16
12017 @opindex mno-const16
12018 Enable or disable use of @code{CONST16} instructions for loading
12019 constant values. The @code{CONST16} instruction is currently not a
12020 standard option from Tensilica. When enabled, @code{CONST16}
12021 instructions are always used in place of the standard @code{L32R}
12022 instructions. The use of @code{CONST16} is enabled by default only if
12023 the @code{L32R} instruction is not available.
12026 @itemx -mno-fused-madd
12027 @opindex mfused-madd
12028 @opindex mno-fused-madd
12029 Enable or disable use of fused multiply/add and multiply/subtract
12030 instructions in the floating-point option. This has no effect if the
12031 floating-point option is not also enabled. Disabling fused multiply/add
12032 and multiply/subtract instructions forces the compiler to use separate
12033 instructions for the multiply and add/subtract operations. This may be
12034 desirable in some cases where strict IEEE 754-compliant results are
12035 required: the fused multiply add/subtract instructions do not round the
12036 intermediate result, thereby producing results with @emph{more} bits of
12037 precision than specified by the IEEE standard. Disabling fused multiply
12038 add/subtract instructions also ensures that the program output is not
12039 sensitive to the compiler's ability to combine multiply and add/subtract
12042 @item -mtext-section-literals
12043 @itemx -mno-text-section-literals
12044 @opindex mtext-section-literals
12045 @opindex mno-text-section-literals
12046 Control the treatment of literal pools. The default is
12047 @option{-mno-text-section-literals}, which places literals in a separate
12048 section in the output file. This allows the literal pool to be placed
12049 in a data RAM/ROM, and it also allows the linker to combine literal
12050 pools from separate object files to remove redundant literals and
12051 improve code size. With @option{-mtext-section-literals}, the literals
12052 are interspersed in the text section in order to keep them as close as
12053 possible to their references. This may be necessary for large assembly
12056 @item -mtarget-align
12057 @itemx -mno-target-align
12058 @opindex mtarget-align
12059 @opindex mno-target-align
12060 When this option is enabled, GCC instructs the assembler to
12061 automatically align instructions to reduce branch penalties at the
12062 expense of some code density. The assembler attempts to widen density
12063 instructions to align branch targets and the instructions following call
12064 instructions. If there are not enough preceding safe density
12065 instructions to align a target, no widening will be performed. The
12066 default is @option{-mtarget-align}. These options do not affect the
12067 treatment of auto-aligned instructions like @code{LOOP}, which the
12068 assembler will always align, either by widening density instructions or
12069 by inserting no-op instructions.
12072 @itemx -mno-longcalls
12073 @opindex mlongcalls
12074 @opindex mno-longcalls
12075 When this option is enabled, GCC instructs the assembler to translate
12076 direct calls to indirect calls unless it can determine that the target
12077 of a direct call is in the range allowed by the call instruction. This
12078 translation typically occurs for calls to functions in other source
12079 files. Specifically, the assembler translates a direct @code{CALL}
12080 instruction into an @code{L32R} followed by a @code{CALLX} instruction.
12081 The default is @option{-mno-longcalls}. This option should be used in
12082 programs where the call target can potentially be out of range. This
12083 option is implemented in the assembler, not the compiler, so the
12084 assembly code generated by GCC will still show direct call
12085 instructions---look at the disassembled object code to see the actual
12086 instructions. Note that the assembler will use an indirect call for
12087 every cross-file call, not just those that really will be out of range.
12090 @node zSeries Options
12091 @subsection zSeries Options
12092 @cindex zSeries options
12094 These are listed under @xref{S/390 and zSeries Options}.
12096 @node Code Gen Options
12097 @section Options for Code Generation Conventions
12098 @cindex code generation conventions
12099 @cindex options, code generation
12100 @cindex run-time options
12102 These machine-independent options control the interface conventions
12103 used in code generation.
12105 Most of them have both positive and negative forms; the negative form
12106 of @option{-ffoo} would be @option{-fno-foo}. In the table below, only
12107 one of the forms is listed---the one which is not the default. You
12108 can figure out the other form by either removing @samp{no-} or adding
12112 @item -fbounds-check
12113 @opindex fbounds-check
12114 For front-ends that support it, generate additional code to check that
12115 indices used to access arrays are within the declared range. This is
12116 currently only supported by the Java and Fortran 77 front-ends, where
12117 this option defaults to true and false respectively.
12121 This option generates traps for signed overflow on addition, subtraction,
12122 multiplication operations.
12126 This option instructs the compiler to assume that signed arithmetic
12127 overflow of addition, subtraction and multiplication wraps around
12128 using twos-complement representation. This flag enables some optimizations
12129 and disables other. This option is enabled by default for the Java
12130 front-end, as required by the Java language specification.
12133 @opindex fexceptions
12134 Enable exception handling. Generates extra code needed to propagate
12135 exceptions. For some targets, this implies GCC will generate frame
12136 unwind information for all functions, which can produce significant data
12137 size overhead, although it does not affect execution. If you do not
12138 specify this option, GCC will enable it by default for languages like
12139 C++ which normally require exception handling, and disable it for
12140 languages like C that do not normally require it. However, you may need
12141 to enable this option when compiling C code that needs to interoperate
12142 properly with exception handlers written in C++. You may also wish to
12143 disable this option if you are compiling older C++ programs that don't
12144 use exception handling.
12146 @item -fnon-call-exceptions
12147 @opindex fnon-call-exceptions
12148 Generate code that allows trapping instructions to throw exceptions.
12149 Note that this requires platform-specific runtime support that does
12150 not exist everywhere. Moreover, it only allows @emph{trapping}
12151 instructions to throw exceptions, i.e.@: memory references or floating
12152 point instructions. It does not allow exceptions to be thrown from
12153 arbitrary signal handlers such as @code{SIGALRM}.
12155 @item -funwind-tables
12156 @opindex funwind-tables
12157 Similar to @option{-fexceptions}, except that it will just generate any needed
12158 static data, but will not affect the generated code in any other way.
12159 You will normally not enable this option; instead, a language processor
12160 that needs this handling would enable it on your behalf.
12162 @item -fasynchronous-unwind-tables
12163 @opindex fasynchronous-unwind-tables
12164 Generate unwind table in dwarf2 format, if supported by target machine. The
12165 table is exact at each instruction boundary, so it can be used for stack
12166 unwinding from asynchronous events (such as debugger or garbage collector).
12168 @item -fpcc-struct-return
12169 @opindex fpcc-struct-return
12170 Return ``short'' @code{struct} and @code{union} values in memory like
12171 longer ones, rather than in registers. This convention is less
12172 efficient, but it has the advantage of allowing intercallability between
12173 GCC-compiled files and files compiled with other compilers, particularly
12174 the Portable C Compiler (pcc).
12176 The precise convention for returning structures in memory depends
12177 on the target configuration macros.
12179 Short structures and unions are those whose size and alignment match
12180 that of some integer type.
12182 @strong{Warning:} code compiled with the @option{-fpcc-struct-return}
12183 switch is not binary compatible with code compiled with the
12184 @option{-freg-struct-return} switch.
12185 Use it to conform to a non-default application binary interface.
12187 @item -freg-struct-return
12188 @opindex freg-struct-return
12189 Return @code{struct} and @code{union} values in registers when possible.
12190 This is more efficient for small structures than
12191 @option{-fpcc-struct-return}.
12193 If you specify neither @option{-fpcc-struct-return} nor
12194 @option{-freg-struct-return}, GCC defaults to whichever convention is
12195 standard for the target. If there is no standard convention, GCC
12196 defaults to @option{-fpcc-struct-return}, except on targets where GCC is
12197 the principal compiler. In those cases, we can choose the standard, and
12198 we chose the more efficient register return alternative.
12200 @strong{Warning:} code compiled with the @option{-freg-struct-return}
12201 switch is not binary compatible with code compiled with the
12202 @option{-fpcc-struct-return} switch.
12203 Use it to conform to a non-default application binary interface.
12205 @item -fshort-enums
12206 @opindex fshort-enums
12207 Allocate to an @code{enum} type only as many bytes as it needs for the
12208 declared range of possible values. Specifically, the @code{enum} type
12209 will be equivalent to the smallest integer type which has enough room.
12211 @strong{Warning:} the @option{-fshort-enums} switch causes GCC to generate
12212 code that is not binary compatible with code generated without that switch.
12213 Use it to conform to a non-default application binary interface.
12215 @item -fshort-double
12216 @opindex fshort-double
12217 Use the same size for @code{double} as for @code{float}.
12219 @strong{Warning:} the @option{-fshort-double} switch causes GCC to generate
12220 code that is not binary compatible with code generated without that switch.
12221 Use it to conform to a non-default application binary interface.
12223 @item -fshort-wchar
12224 @opindex fshort-wchar
12225 Override the underlying type for @samp{wchar_t} to be @samp{short
12226 unsigned int} instead of the default for the target. This option is
12227 useful for building programs to run under WINE@.
12229 @strong{Warning:} the @option{-fshort-wchar} switch causes GCC to generate
12230 code that is not binary compatible with code generated without that switch.
12231 Use it to conform to a non-default application binary interface.
12233 @item -fshared-data
12234 @opindex fshared-data
12235 Requests that the data and non-@code{const} variables of this
12236 compilation be shared data rather than private data. The distinction
12237 makes sense only on certain operating systems, where shared data is
12238 shared between processes running the same program, while private data
12239 exists in one copy per process.
12242 @opindex fno-common
12243 In C, allocate even uninitialized global variables in the data section of the
12244 object file, rather than generating them as common blocks. This has the
12245 effect that if the same variable is declared (without @code{extern}) in
12246 two different compilations, you will get an error when you link them.
12247 The only reason this might be useful is if you wish to verify that the
12248 program will work on other systems which always work this way.
12252 Ignore the @samp{#ident} directive.
12254 @item -finhibit-size-directive
12255 @opindex finhibit-size-directive
12256 Don't output a @code{.size} assembler directive, or anything else that
12257 would cause trouble if the function is split in the middle, and the
12258 two halves are placed at locations far apart in memory. This option is
12259 used when compiling @file{crtstuff.c}; you should not need to use it
12262 @item -fverbose-asm
12263 @opindex fverbose-asm
12264 Put extra commentary information in the generated assembly code to
12265 make it more readable. This option is generally only of use to those
12266 who actually need to read the generated assembly code (perhaps while
12267 debugging the compiler itself).
12269 @option{-fno-verbose-asm}, the default, causes the
12270 extra information to be omitted and is useful when comparing two assembler
12275 @cindex global offset table
12277 Generate position-independent code (PIC) suitable for use in a shared
12278 library, if supported for the target machine. Such code accesses all
12279 constant addresses through a global offset table (GOT)@. The dynamic
12280 loader resolves the GOT entries when the program starts (the dynamic
12281 loader is not part of GCC; it is part of the operating system). If
12282 the GOT size for the linked executable exceeds a machine-specific
12283 maximum size, you get an error message from the linker indicating that
12284 @option{-fpic} does not work; in that case, recompile with @option{-fPIC}
12285 instead. (These maximums are 8k on the SPARC and 32k
12286 on the m68k and RS/6000. The 386 has no such limit.)
12288 Position-independent code requires special support, and therefore works
12289 only on certain machines. For the 386, GCC supports PIC for System V
12290 but not for the Sun 386i. Code generated for the IBM RS/6000 is always
12291 position-independent.
12295 If supported for the target machine, emit position-independent code,
12296 suitable for dynamic linking and avoiding any limit on the size of the
12297 global offset table. This option makes a difference on the m68k,
12298 PowerPC and SPARC@.
12300 Position-independent code requires special support, and therefore works
12301 only on certain machines.
12307 These options are similar to @option{-fpic} and @option{-fPIC}, but
12308 generated position independent code can be only linked into executables.
12309 Usually these options are used when @option{-pie} GCC option will be
12310 used during linking.
12312 @item -ffixed-@var{reg}
12314 Treat the register named @var{reg} as a fixed register; generated code
12315 should never refer to it (except perhaps as a stack pointer, frame
12316 pointer or in some other fixed role).
12318 @var{reg} must be the name of a register. The register names accepted
12319 are machine-specific and are defined in the @code{REGISTER_NAMES}
12320 macro in the machine description macro file.
12322 This flag does not have a negative form, because it specifies a
12325 @item -fcall-used-@var{reg}
12326 @opindex fcall-used
12327 Treat the register named @var{reg} as an allocable register that is
12328 clobbered by function calls. It may be allocated for temporaries or
12329 variables that do not live across a call. Functions compiled this way
12330 will not save and restore the register @var{reg}.
12332 It is an error to used this flag with the frame pointer or stack pointer.
12333 Use of this flag for other registers that have fixed pervasive roles in
12334 the machine's execution model will produce disastrous results.
12336 This flag does not have a negative form, because it specifies a
12339 @item -fcall-saved-@var{reg}
12340 @opindex fcall-saved
12341 Treat the register named @var{reg} as an allocable register saved by
12342 functions. It may be allocated even for temporaries or variables that
12343 live across a call. Functions compiled this way will save and restore
12344 the register @var{reg} if they use it.
12346 It is an error to used this flag with the frame pointer or stack pointer.
12347 Use of this flag for other registers that have fixed pervasive roles in
12348 the machine's execution model will produce disastrous results.
12350 A different sort of disaster will result from the use of this flag for
12351 a register in which function values may be returned.
12353 This flag does not have a negative form, because it specifies a
12356 @item -fpack-struct[=@var{n}]
12357 @opindex fpack-struct
12358 Without a value specified, pack all structure members together without
12359 holes. When a value is specified (which must be a small power of two), pack
12360 structure members according to this value, representing the maximum
12361 alignment (that is, objects with default alignment requirements larger than
12362 this will be output potentially unaligned at the next fitting location.
12364 @strong{Warning:} the @option{-fpack-struct} switch causes GCC to generate
12365 code that is not binary compatible with code generated without that switch.
12366 Additionally, it makes the code suboptimal.
12367 Use it to conform to a non-default application binary interface.
12369 @item -finstrument-functions
12370 @opindex finstrument-functions
12371 Generate instrumentation calls for entry and exit to functions. Just
12372 after function entry and just before function exit, the following
12373 profiling functions will be called with the address of the current
12374 function and its call site. (On some platforms,
12375 @code{__builtin_return_address} does not work beyond the current
12376 function, so the call site information may not be available to the
12377 profiling functions otherwise.)
12380 void __cyg_profile_func_enter (void *this_fn,
12382 void __cyg_profile_func_exit (void *this_fn,
12386 The first argument is the address of the start of the current function,
12387 which may be looked up exactly in the symbol table.
12389 This instrumentation is also done for functions expanded inline in other
12390 functions. The profiling calls will indicate where, conceptually, the
12391 inline function is entered and exited. This means that addressable
12392 versions of such functions must be available. If all your uses of a
12393 function are expanded inline, this may mean an additional expansion of
12394 code size. If you use @samp{extern inline} in your C code, an
12395 addressable version of such functions must be provided. (This is
12396 normally the case anyways, but if you get lucky and the optimizer always
12397 expands the functions inline, you might have gotten away without
12398 providing static copies.)
12400 A function may be given the attribute @code{no_instrument_function}, in
12401 which case this instrumentation will not be done. This can be used, for
12402 example, for the profiling functions listed above, high-priority
12403 interrupt routines, and any functions from which the profiling functions
12404 cannot safely be called (perhaps signal handlers, if the profiling
12405 routines generate output or allocate memory).
12407 @item -fstack-check
12408 @opindex fstack-check
12409 Generate code to verify that you do not go beyond the boundary of the
12410 stack. You should specify this flag if you are running in an
12411 environment with multiple threads, but only rarely need to specify it in
12412 a single-threaded environment since stack overflow is automatically
12413 detected on nearly all systems if there is only one stack.
12415 Note that this switch does not actually cause checking to be done; the
12416 operating system must do that. The switch causes generation of code
12417 to ensure that the operating system sees the stack being extended.
12419 @item -fstack-limit-register=@var{reg}
12420 @itemx -fstack-limit-symbol=@var{sym}
12421 @itemx -fno-stack-limit
12422 @opindex fstack-limit-register
12423 @opindex fstack-limit-symbol
12424 @opindex fno-stack-limit
12425 Generate code to ensure that the stack does not grow beyond a certain value,
12426 either the value of a register or the address of a symbol. If the stack
12427 would grow beyond the value, a signal is raised. For most targets,
12428 the signal is raised before the stack overruns the boundary, so
12429 it is possible to catch the signal without taking special precautions.
12431 For instance, if the stack starts at absolute address @samp{0x80000000}
12432 and grows downwards, you can use the flags
12433 @option{-fstack-limit-symbol=__stack_limit} and
12434 @option{-Wl,--defsym,__stack_limit=0x7ffe0000} to enforce a stack limit
12435 of 128KB@. Note that this may only work with the GNU linker.
12437 @cindex aliasing of parameters
12438 @cindex parameters, aliased
12439 @item -fargument-alias
12440 @itemx -fargument-noalias
12441 @itemx -fargument-noalias-global
12442 @opindex fargument-alias
12443 @opindex fargument-noalias
12444 @opindex fargument-noalias-global
12445 Specify the possible relationships among parameters and between
12446 parameters and global data.
12448 @option{-fargument-alias} specifies that arguments (parameters) may
12449 alias each other and may alias global storage.@*
12450 @option{-fargument-noalias} specifies that arguments do not alias
12451 each other, but may alias global storage.@*
12452 @option{-fargument-noalias-global} specifies that arguments do not
12453 alias each other and do not alias global storage.
12455 Each language will automatically use whatever option is required by
12456 the language standard. You should not need to use these options yourself.
12458 @item -fleading-underscore
12459 @opindex fleading-underscore
12460 This option and its counterpart, @option{-fno-leading-underscore}, forcibly
12461 change the way C symbols are represented in the object file. One use
12462 is to help link with legacy assembly code.
12464 @strong{Warning:} the @option{-fleading-underscore} switch causes GCC to
12465 generate code that is not binary compatible with code generated without that
12466 switch. Use it to conform to a non-default application binary interface.
12467 Not all targets provide complete support for this switch.
12469 @item -ftls-model=@var{model}
12470 Alter the thread-local storage model to be used (@pxref{Thread-Local}).
12471 The @var{model} argument should be one of @code{global-dynamic},
12472 @code{local-dynamic}, @code{initial-exec} or @code{local-exec}.
12474 The default without @option{-fpic} is @code{initial-exec}; with
12475 @option{-fpic} the default is @code{global-dynamic}.
12477 @item -fvisibility=@var{default|internal|hidden|protected}
12478 @opindex fvisibility
12479 Set the default ELF image symbol visibility to the specified option---all
12480 symbols will be marked with this unless overridden within the code.
12481 Using this feature can very substantially improve linking and
12482 load times of shared object libraries, produce more optimized
12483 code, provide near-perfect API export and prevent symbol clashes.
12484 It is @strong{strongly} recommended that you use this in any shared objects
12487 Despite the nomenclature, @code{default} always means public ie;
12488 available to be linked against from outside the shared object.
12489 @code{protected} and @code{internal} are pretty useless in real-world
12490 usage so the only other commonly used option will be @code{hidden}.
12491 The default if @option{-fvisibility} isn't specified is
12492 @code{default}, i.e., make every
12493 symbol public---this causes the same behavior as previous versions of
12496 A good explanation of the benefits offered by ensuring ELF
12497 symbols have the correct visibility is given by ``How To Write
12498 Shared Libraries'' by Ulrich Drepper (which can be found at
12499 @w{@uref{http://people.redhat.com/~drepper/}})---however a superior
12500 solution made possible by this option to marking things hidden when
12501 the default is public is to make the default hidden and mark things
12502 public. This is the norm with DLL's on Windows and with @option{-fvisibility=hidden}
12503 and @code{__attribute__ ((visibility("default")))} instead of
12504 @code{__declspec(dllexport)} you get almost identical semantics with
12505 identical syntax. This is a great boon to those working with
12506 cross-platform projects.
12508 For those adding visibility support to existing code, you may find
12509 @samp{#pragma GCC visibility} of use. This works by you enclosing
12510 the declarations you wish to set visibility for with (for example)
12511 @samp{#pragma GCC visibility push(hidden)} and
12512 @samp{#pragma GCC visibility pop}. These can be nested up to sixteen
12513 times. Bear in mind that symbol visibility should be viewed @strong{as
12514 part of the API interface contract} and thus all new code should
12515 always specify visibility when it is not the default ie; declarations
12516 only for use within the local DSO should @strong{always} be marked explicitly
12517 as hidden as so to avoid PLT indirection overheads---making this
12518 abundantly clear also aids readability and self-documentation of the code.
12519 Note that due to ISO C++ specification requirements, operator new and
12520 operator delete must always be of default visibility.
12522 An overview of these techniques, their benefits and how to use them
12523 is at @w{@uref{http://gcc.gnu.org/wiki/Visibility}}.
12529 @node Environment Variables
12530 @section Environment Variables Affecting GCC
12531 @cindex environment variables
12533 @c man begin ENVIRONMENT
12534 This section describes several environment variables that affect how GCC
12535 operates. Some of them work by specifying directories or prefixes to use
12536 when searching for various kinds of files. Some are used to specify other
12537 aspects of the compilation environment.
12539 Note that you can also specify places to search using options such as
12540 @option{-B}, @option{-I} and @option{-L} (@pxref{Directory Options}). These
12541 take precedence over places specified using environment variables, which
12542 in turn take precedence over those specified by the configuration of GCC@.
12543 @xref{Driver,, Controlling the Compilation Driver @file{gcc}, gccint,
12544 GNU Compiler Collection (GCC) Internals}.
12549 @c @itemx LC_COLLATE
12551 @c @itemx LC_MONETARY
12552 @c @itemx LC_NUMERIC
12557 @c @findex LC_COLLATE
12558 @findex LC_MESSAGES
12559 @c @findex LC_MONETARY
12560 @c @findex LC_NUMERIC
12564 These environment variables control the way that GCC uses
12565 localization information that allow GCC to work with different
12566 national conventions. GCC inspects the locale categories
12567 @env{LC_CTYPE} and @env{LC_MESSAGES} if it has been configured to do
12568 so. These locale categories can be set to any value supported by your
12569 installation. A typical value is @samp{en_GB.UTF-8} for English in the United
12570 Kingdom encoded in UTF-8.
12572 The @env{LC_CTYPE} environment variable specifies character
12573 classification. GCC uses it to determine the character boundaries in
12574 a string; this is needed for some multibyte encodings that contain quote
12575 and escape characters that would otherwise be interpreted as a string
12578 The @env{LC_MESSAGES} environment variable specifies the language to
12579 use in diagnostic messages.
12581 If the @env{LC_ALL} environment variable is set, it overrides the value
12582 of @env{LC_CTYPE} and @env{LC_MESSAGES}; otherwise, @env{LC_CTYPE}
12583 and @env{LC_MESSAGES} default to the value of the @env{LANG}
12584 environment variable. If none of these variables are set, GCC
12585 defaults to traditional C English behavior.
12589 If @env{TMPDIR} is set, it specifies the directory to use for temporary
12590 files. GCC uses temporary files to hold the output of one stage of
12591 compilation which is to be used as input to the next stage: for example,
12592 the output of the preprocessor, which is the input to the compiler
12595 @item GCC_EXEC_PREFIX
12596 @findex GCC_EXEC_PREFIX
12597 If @env{GCC_EXEC_PREFIX} is set, it specifies a prefix to use in the
12598 names of the subprograms executed by the compiler. No slash is added
12599 when this prefix is combined with the name of a subprogram, but you can
12600 specify a prefix that ends with a slash if you wish.
12602 If @env{GCC_EXEC_PREFIX} is not set, GCC will attempt to figure out
12603 an appropriate prefix to use based on the pathname it was invoked with.
12605 If GCC cannot find the subprogram using the specified prefix, it
12606 tries looking in the usual places for the subprogram.
12608 The default value of @env{GCC_EXEC_PREFIX} is
12609 @file{@var{prefix}/lib/gcc/} where @var{prefix} is the value
12610 of @code{prefix} when you ran the @file{configure} script.
12612 Other prefixes specified with @option{-B} take precedence over this prefix.
12614 This prefix is also used for finding files such as @file{crt0.o} that are
12617 In addition, the prefix is used in an unusual way in finding the
12618 directories to search for header files. For each of the standard
12619 directories whose name normally begins with @samp{/usr/local/lib/gcc}
12620 (more precisely, with the value of @env{GCC_INCLUDE_DIR}), GCC tries
12621 replacing that beginning with the specified prefix to produce an
12622 alternate directory name. Thus, with @option{-Bfoo/}, GCC will search
12623 @file{foo/bar} where it would normally search @file{/usr/local/lib/bar}.
12624 These alternate directories are searched first; the standard directories
12627 @item COMPILER_PATH
12628 @findex COMPILER_PATH
12629 The value of @env{COMPILER_PATH} is a colon-separated list of
12630 directories, much like @env{PATH}. GCC tries the directories thus
12631 specified when searching for subprograms, if it can't find the
12632 subprograms using @env{GCC_EXEC_PREFIX}.
12635 @findex LIBRARY_PATH
12636 The value of @env{LIBRARY_PATH} is a colon-separated list of
12637 directories, much like @env{PATH}. When configured as a native compiler,
12638 GCC tries the directories thus specified when searching for special
12639 linker files, if it can't find them using @env{GCC_EXEC_PREFIX}. Linking
12640 using GCC also uses these directories when searching for ordinary
12641 libraries for the @option{-l} option (but directories specified with
12642 @option{-L} come first).
12646 @cindex locale definition
12647 This variable is used to pass locale information to the compiler. One way in
12648 which this information is used is to determine the character set to be used
12649 when character literals, string literals and comments are parsed in C and C++.
12650 When the compiler is configured to allow multibyte characters,
12651 the following values for @env{LANG} are recognized:
12655 Recognize JIS characters.
12657 Recognize SJIS characters.
12659 Recognize EUCJP characters.
12662 If @env{LANG} is not defined, or if it has some other value, then the
12663 compiler will use mblen and mbtowc as defined by the default locale to
12664 recognize and translate multibyte characters.
12668 Some additional environments variables affect the behavior of the
12671 @include cppenv.texi
12675 @node Precompiled Headers
12676 @section Using Precompiled Headers
12677 @cindex precompiled headers
12678 @cindex speed of compilation
12680 Often large projects have many header files that are included in every
12681 source file. The time the compiler takes to process these header files
12682 over and over again can account for nearly all of the time required to
12683 build the project. To make builds faster, GCC allows users to
12684 `precompile' a header file; then, if builds can use the precompiled
12685 header file they will be much faster.
12687 @strong{Caution:} There are a few known situations where GCC will
12688 crash when trying to use a precompiled header. If you have trouble
12689 with a precompiled header, you should remove the precompiled header
12690 and compile without it. In addition, please use GCC's on-line
12691 defect-tracking system to report any problems you encounter with
12692 precompiled headers. @xref{Bugs}.
12694 To create a precompiled header file, simply compile it as you would any
12695 other file, if necessary using the @option{-x} option to make the driver
12696 treat it as a C or C++ header file. You will probably want to use a
12697 tool like @command{make} to keep the precompiled header up-to-date when
12698 the headers it contains change.
12700 A precompiled header file will be searched for when @code{#include} is
12701 seen in the compilation. As it searches for the included file
12702 (@pxref{Search Path,,Search Path,cpp,The C Preprocessor}) the
12703 compiler looks for a precompiled header in each directory just before it
12704 looks for the include file in that directory. The name searched for is
12705 the name specified in the @code{#include} with @samp{.gch} appended. If
12706 the precompiled header file can't be used, it is ignored.
12708 For instance, if you have @code{#include "all.h"}, and you have
12709 @file{all.h.gch} in the same directory as @file{all.h}, then the
12710 precompiled header file will be used if possible, and the original
12711 header will be used otherwise.
12713 Alternatively, you might decide to put the precompiled header file in a
12714 directory and use @option{-I} to ensure that directory is searched
12715 before (or instead of) the directory containing the original header.
12716 Then, if you want to check that the precompiled header file is always
12717 used, you can put a file of the same name as the original header in this
12718 directory containing an @code{#error} command.
12720 This also works with @option{-include}. So yet another way to use
12721 precompiled headers, good for projects not designed with precompiled
12722 header files in mind, is to simply take most of the header files used by
12723 a project, include them from another header file, precompile that header
12724 file, and @option{-include} the precompiled header. If the header files
12725 have guards against multiple inclusion, they will be skipped because
12726 they've already been included (in the precompiled header).
12728 If you need to precompile the same header file for different
12729 languages, targets, or compiler options, you can instead make a
12730 @emph{directory} named like @file{all.h.gch}, and put each precompiled
12731 header in the directory, perhaps using @option{-o}. It doesn't matter
12732 what you call the files in the directory, every precompiled header in
12733 the directory will be considered. The first precompiled header
12734 encountered in the directory that is valid for this compilation will
12735 be used; they're searched in no particular order.
12737 There are many other possibilities, limited only by your imagination,
12738 good sense, and the constraints of your build system.
12740 A precompiled header file can be used only when these conditions apply:
12744 Only one precompiled header can be used in a particular compilation.
12747 A precompiled header can't be used once the first C token is seen. You
12748 can have preprocessor directives before a precompiled header; you can
12749 even include a precompiled header from inside another header, so long as
12750 there are no C tokens before the @code{#include}.
12753 The precompiled header file must be produced for the same language as
12754 the current compilation. You can't use a C precompiled header for a C++
12758 The precompiled header file must be produced by the same compiler
12759 version and configuration as the current compilation is using.
12760 The easiest way to guarantee this is to use the same compiler binary
12761 for creating and using precompiled headers.
12764 Any macros defined before the precompiled header is included must
12765 either be defined in the same way as when the precompiled header was
12766 generated, or must not affect the precompiled header, which usually
12767 means that the they don't appear in the precompiled header at all.
12769 The @option{-D} option is one way to define a macro before a
12770 precompiled header is included; using a @code{#define} can also do it.
12771 There are also some options that define macros implicitly, like
12772 @option{-O} and @option{-Wdeprecated}; the same rule applies to macros
12775 @item If debugging information is output when using the precompiled
12776 header, using @option{-g} or similar, the same kind of debugging information
12777 must have been output when building the precompiled header. However,
12778 a precompiled header built using @option{-g} can be used in a compilation
12779 when no debugging information is being output.
12781 @item The same @option{-m} options must generally be used when building
12782 and using the precompiled header. @xref{Submodel Options},
12783 for any cases where this rule is relaxed.
12785 @item Each of the following options must be the same when building and using
12786 the precompiled header:
12788 @gccoptlist{-fexceptions -funit-at-a-time}
12791 Some other command-line options starting with @option{-f},
12792 @option{-p}, or @option{-O} must be defined in the same way as when
12793 the precompiled header was generated. At present, it's not clear
12794 which options are safe to change and which are not; the safest choice
12795 is to use exactly the same options when generating and using the
12796 precompiled header. The following are known to be safe:
12798 @gccoptlist{-fpreprocessed -pedantic-errors}
12802 For all of these except the last, the compiler will automatically
12803 ignore the precompiled header if the conditions aren't met. If you
12804 find an option combination that doesn't work and doesn't cause the
12805 precompiled header to be ignored, please consider filing a bug report,
12808 If you do use differing options when generating and using the
12809 precompiled header, the actual behavior will be a mixture of the
12810 behavior for the options. For instance, if you use @option{-g} to
12811 generate the precompiled header but not when using it, you may or may
12812 not get debugging information for routines in the precompiled header.
12814 @node Running Protoize
12815 @section Running Protoize
12817 The program @code{protoize} is an optional part of GCC@. You can use
12818 it to add prototypes to a program, thus converting the program to ISO
12819 C in one respect. The companion program @code{unprotoize} does the
12820 reverse: it removes argument types from any prototypes that are found.
12822 When you run these programs, you must specify a set of source files as
12823 command line arguments. The conversion programs start out by compiling
12824 these files to see what functions they define. The information gathered
12825 about a file @var{foo} is saved in a file named @file{@var{foo}.X}.
12827 After scanning comes actual conversion. The specified files are all
12828 eligible to be converted; any files they include (whether sources or
12829 just headers) are eligible as well.
12831 But not all the eligible files are converted. By default,
12832 @code{protoize} and @code{unprotoize} convert only source and header
12833 files in the current directory. You can specify additional directories
12834 whose files should be converted with the @option{-d @var{directory}}
12835 option. You can also specify particular files to exclude with the
12836 @option{-x @var{file}} option. A file is converted if it is eligible, its
12837 directory name matches one of the specified directory names, and its
12838 name within the directory has not been excluded.
12840 Basic conversion with @code{protoize} consists of rewriting most
12841 function definitions and function declarations to specify the types of
12842 the arguments. The only ones not rewritten are those for varargs
12845 @code{protoize} optionally inserts prototype declarations at the
12846 beginning of the source file, to make them available for any calls that
12847 precede the function's definition. Or it can insert prototype
12848 declarations with block scope in the blocks where undeclared functions
12851 Basic conversion with @code{unprotoize} consists of rewriting most
12852 function declarations to remove any argument types, and rewriting
12853 function definitions to the old-style pre-ISO form.
12855 Both conversion programs print a warning for any function declaration or
12856 definition that they can't convert. You can suppress these warnings
12859 The output from @code{protoize} or @code{unprotoize} replaces the
12860 original source file. The original file is renamed to a name ending
12861 with @samp{.save} (for DOS, the saved filename ends in @samp{.sav}
12862 without the original @samp{.c} suffix). If the @samp{.save} (@samp{.sav}
12863 for DOS) file already exists, then the source file is simply discarded.
12865 @code{protoize} and @code{unprotoize} both depend on GCC itself to
12866 scan the program and collect information about the functions it uses.
12867 So neither of these programs will work until GCC is installed.
12869 Here is a table of the options you can use with @code{protoize} and
12870 @code{unprotoize}. Each option works with both programs unless
12874 @item -B @var{directory}
12875 Look for the file @file{SYSCALLS.c.X} in @var{directory}, instead of the
12876 usual directory (normally @file{/usr/local/lib}). This file contains
12877 prototype information about standard system functions. This option
12878 applies only to @code{protoize}.
12880 @item -c @var{compilation-options}
12881 Use @var{compilation-options} as the options when running @command{gcc} to
12882 produce the @samp{.X} files. The special option @option{-aux-info} is
12883 always passed in addition, to tell @command{gcc} to write a @samp{.X} file.
12885 Note that the compilation options must be given as a single argument to
12886 @code{protoize} or @code{unprotoize}. If you want to specify several
12887 @command{gcc} options, you must quote the entire set of compilation options
12888 to make them a single word in the shell.
12890 There are certain @command{gcc} arguments that you cannot use, because they
12891 would produce the wrong kind of output. These include @option{-g},
12892 @option{-O}, @option{-c}, @option{-S}, and @option{-o} If you include these in
12893 the @var{compilation-options}, they are ignored.
12896 Rename files to end in @samp{.C} (@samp{.cc} for DOS-based file
12897 systems) instead of @samp{.c}. This is convenient if you are converting
12898 a C program to C++. This option applies only to @code{protoize}.
12901 Add explicit global declarations. This means inserting explicit
12902 declarations at the beginning of each source file for each function
12903 that is called in the file and was not declared. These declarations
12904 precede the first function definition that contains a call to an
12905 undeclared function. This option applies only to @code{protoize}.
12907 @item -i @var{string}
12908 Indent old-style parameter declarations with the string @var{string}.
12909 This option applies only to @code{protoize}.
12911 @code{unprotoize} converts prototyped function definitions to old-style
12912 function definitions, where the arguments are declared between the
12913 argument list and the initial @samp{@{}. By default, @code{unprotoize}
12914 uses five spaces as the indentation. If you want to indent with just
12915 one space instead, use @option{-i " "}.
12918 Keep the @samp{.X} files. Normally, they are deleted after conversion
12922 Add explicit local declarations. @code{protoize} with @option{-l} inserts
12923 a prototype declaration for each function in each block which calls the
12924 function without any declaration. This option applies only to
12928 Make no real changes. This mode just prints information about the conversions
12929 that would have been done without @option{-n}.
12932 Make no @samp{.save} files. The original files are simply deleted.
12933 Use this option with caution.
12935 @item -p @var{program}
12936 Use the program @var{program} as the compiler. Normally, the name
12937 @file{gcc} is used.
12940 Work quietly. Most warnings are suppressed.
12943 Print the version number, just like @option{-v} for @command{gcc}.
12946 If you need special compiler options to compile one of your program's
12947 source files, then you should generate that file's @samp{.X} file
12948 specially, by running @command{gcc} on that source file with the
12949 appropriate options and the option @option{-aux-info}. Then run
12950 @code{protoize} on the entire set of files. @code{protoize} will use
12951 the existing @samp{.X} file because it is newer than the source file.
12955 gcc -Dfoo=bar file1.c -aux-info file1.X
12960 You need to include the special files along with the rest in the
12961 @code{protoize} command, even though their @samp{.X} files already
12962 exist, because otherwise they won't get converted.
12964 @xref{Protoize Caveats}, for more information on how to use
12965 @code{protoize} successfully.